Splitters Versus Lumpers in Historical Linguistics

Warning:  Long, runs to 57 pages. This article is intended at the moment more for the general audience than for specialists,  but specialists may also find it of interest. At the moment, it is not properly formatted or edited to be of use for publication in an academic journal, but perhaps it could be published in such a format some day.

For background into what Historical Linguistics is, see this Wikipedia article. Basically it involves determining which languages are related to each other via various means and once that is determined, reconstructing a proto-language that the related languages descended from, along with, hopefully, regular sound correspondences which supposedly proves the relationship once and for all. The argument in Historical Linguistics now is between conservatives or splitters or progressives or lumpers.

Splitters say that the comparative method – described above as reconstructing a proto-language with regular sound correspondences – is necessary in order to prove that two or more languages are related. However, they also say, probably correctly, that this method is not useful beyond ~6,000 years. Any relationships beyond that time frame would not be provable by the comparative method and hence could never be proven. This effectively shuts down all research into long-range older language families.

Some lumpers say that this method is not necessary and instead relationships can be determined by simply looking at the two or more languages, a process called comparison or mass comparison. I point out below that comparison need not be cursory but could mean deep study of languages over 10, 15, or 20 years.

They tend to focus on core vocabulary, numerals, family terms, pronouns, and deictics, in addition to small morphological particles – all things that are rarely borrowed. Once they find a number of these items that resemble one another greater than chance, they say that the two languages are related because chance and borrowing are ruled out.

They say that this is the way to prove language relatedness, not the comparative method. The comparative method instead is used to learn interesting things about language families that have already been discovered via comparison, such as reconstructing proto-languages and finding regular sound correspondences.

Splitters say that comparison or mass comparison is not a valid way of proving that languages are related and that only the comparative method can be used to prove this. However, as noted, they set a 6,000- year time limit on the method needed to prove this, and this walls off a lot of potential knowledge and about ancient and long-range language relationships as unprovable and hence undiscoverable. In a way, they are shutting the door to new scientific discovery beyond a certain time frame by claiming that the method needed to make these discoveries doesn’t work beyond X thousand years.

Other lumpers disagree that the comparative method has a time limit on it and are attempting to use the comparative method to reconstruct ancient long-range language families and find regular sound correspondences between them. Unfortunately, most of their efforts are in vain as splitters are using increasingly strict criteria for proof of language relationship and hence are shooting down most if not all of these efforts being done “in the proper way.”

So they are saying that proof must be done in a certain way, but when people try to play by the rules and use that way to find proof, they keep moving the goalposts and using increasingly strict, petty, and quibbling methods to in general say that the relationship is not proven.

So the say, “You must use this tool for your proof!” And then people play fair and use the tool, and almost always say, “Sorry, you didn’t prove it!” It all feels like a game that is rigged to fail is most if not all cases.

Hence, the current trend of extreme conservatism in Historical Linguistics has set up rules seem to be designed to prevent the discovery of most if not all new language families, in particular long-range families older than 6-8,000 years.

I am quite certain that long-range language families such as Altaic (with either three families or five), Indo-Uralic, Uralic-Yukaghir, Hokan, Penutian, Mosan, Almosan, Japanese-Korean, Gulf, Yuki-Gulf, Elamite-Dravidian, Quechumaran, Austroasiatic-Hmong Mien, Coahuiltecan, North Caucasian, or Na-Dene will never be proven in my lifetime, and that’s not to mention the more extreme proposals such as Eurasiatic, Nostratic, Dene-Caucasian, Austric, and Amerind, although the evidence for the first and last of these is quite powerful.

There are simply too many emotions tied up in any of these proposals. Further, many linguists have spent a good part of their careers arguing against these proposals. It is doubtful that any amount of evidence will cause them to change their minds. Scientists, like any other humans, don’t like to be shown that they’re wrong.

Lyle Campbell, Maryanne Mithun, Mauricio Mixco, Sarah Grey Thomason, Joanna Nichols, William Poser, Peter Daniels, Dell Hymes, Larry Trask, Gerrit Dimmendaal, Donald Ringe, Juha Janhunen, William Bright, and Paul Sidwell are among the leaders of this new conservatism.

At first I was very angry at what these people were doing, especially the most egregious cases such as Campbell. Then I realized that people lie and misrepresent things all day long every single day in my life and that this behavior is fairly normal behavior in humans, especially in a mushy area like this one where hard truths are hard to come by and most stated facts are more properly matters of opinion or could be construed that way.

I realized that they are simply defending a scientific paradigm and that unfortunately, this is the rather underhanded and emotion-ridden environment that defending paradigms tends to produce.

Though to be completely honest, I should not be singling these people out because the current conservatism is simply consensus and acts as the current paradigm on the language relatedness question in Historical Linguistics. The people listed above are at the top of the profession and are often considered the best historical linguists. They write books on historical linguistics. A number are considered to be ultimate authorities on questions of language relatedness. They are simply the leading edge of the current conservative consensus and paradigm in the field.

Although granted, of all of them, Campbell seems to be the most extreme conservative. He is also one of the top historical linguists in the world. Mixco, Mithun, and Poser are about on the same level as Campbell.

Campbell, Mithun, Thomason, and Mixco are Americanists whose conservatism was set off by the publication of Joseph Greenberg’s Language in the Americas (LIA) in 1987.

All of the linguists above are noted for the excellent scholarship.

The conservatives who are denying most if not all new families are are called splitters.They tend to be very angry if not out and out abusive, engaging in bullying, mockery, ridicule, ostracization, and all of the usual techniques used in science against the proposers of a new paradigm.

The people who propose long-range families are called lumpers. Lumpers are heavily disparaged in the field nowadays such that almost no one wants to be known as a lumper or associated with such. However, many other historical linguists seem to be taking a more moderate fence-sitter stance where they are open to questions of new language families, including long-range families.

Among the long-range families that the moderates are open to considering nowadays are Indo-Uralic, Dene-Yenisien, and Austro-Tai. Some of the smaller long-range families in the Americas even have supporters among the most hardline of splitters. I’m even dubious about well-argued proposals such as Dene-Yenisien.

Thomason takes extreme umbrage to the notion that splitters have a bias that will not allow few if any new families to be discovered after Greenberg compared them with Malcolm Guthrie’s objections to Greenberg’s new classification of Bantu. However, after thinking this over for some time now, I now believe that Greenberg is correct. The splitters have their minds made up. They are going to allow few if any new families to be discovered. A few of them have caved a bit.

I also work in mental health, and it’s pretty obvious to me when something is not right about a scientific debate. I’ve been getting that vibe about the splitters versus lumpers debate from the very start. When a debate in science has degenerated into bias, ideology and ideologues, propaganda, politics, and in particular extreme emotion, it gives off a certain intuitive feel about it. This debate has felt this way from Day One. To put it simply, the debate simply doesn’t smell right. I have a feeling that science left the room along time ago here.

One thing I noticed was that people who have worked on one particular language or family for much of their careers are especially angry and aggressive about the notion that their family could possibly be related to anything else. Indeed famous linguists were remarking on this tendency as early as 1901. Among the reasons given was that they had their hands full already without new work to take on and a disinclination to see their language family related to anything else as this would deny its specialness.

Trask is forceful that Basque could not possibly have any outside relatives.

I saw a debate on the Net some years ago with Trask and a Spanish assistant holding court over a debate over the external relations of Basque. Those who argued for external relations were pushing a relationship with the Caucasian languages, which is possible though not proven in my opinion. Trask and his assistant were very angry and aggressive in holding down the fort. Apparently everything was a Spanish borrowing. The debate didn’t smell right at all.

With a background in psychology, I wonder what is going on here. One possibility is as Greenberg suggests and as was suggested back in 1901 – simple narcissism. When one specializes in a language family for a long time, it probably become blurred with the self such that the self and the family become married to each other, and it’s hard to tell where one ends and the other begins. Yourself and the family you’ve spent your career working on become one and same thing. If your family is not related to anything else, it’s special.

We all think we are special. This is the essence of human narcissism. To say that their favorite language has relatives is to deny its specialness almost as if to say that our egos were not real but were instead extensions of other people’s egos. Actually if you read Sartre or study modern particle physics, that’s not a bad theory, but most people bristle at the notion.

I met Korean and Japanese people when I was doing my Masters. Both beamed when they told me that their language had no known relatives. Of course that made it special in their eyes and played right into their ethnocentrism.

Another problem may be the trajectory of one’s career. If one has been arguing forcefully for 30 years that there are no known relations to your family, your reputation is going to take a huge hit if you have to agree that you were wrong all those years.

There is also a politics question.

Another reason is Politics. We are dealing here with a Paradigm. For a good description of a Scientific Paradigm, see Thomas Kuhn’s The Structure of Scientific Revolutions. Kuhn holds that science is by its nature very conservative, some sciences being more conservative than others. A Paradigm is set up when the field reaches a satisfactory consensus that a particular theory is correct. After a while, serious barriers go up to any challenges to overthrow the proven theory.

The challenges are first ignored, then ridiculed (often severely), then attacked (often ferociously) and then, if the challenge is successful, it is accepted (often slowly and grudgingly). Kuhn pointed out that defenders of the old theory are usually so reluctant to see the paradigm overthrown that we often must wait literally until their deaths to finally overthrow the paradigm. They defend it to their deathbeds. I suggest we are dealing with something more than pure empiricism here.

It is quite risky to challenge a paradigm in science. People’s careers have suffered from it. A supporter of Keynesian economics, then challenging the current paradigm in economics, could not get hired at any university in the US during the 1930’s.

In the splitters versus lumpers debate, we have been in the Anger phase for some time now. We seem to be settling out of it, as many are taking a fence-sitting position and arguing for attempts to resolve the debate to make it less heated.

The Paradigm here involves extreme skepticism about any new language families to the point that any new families are simply going to be rejected on all sorts of grounds. Paradigms involve politics at the academic level. When a Paradigm is set up in science, almost all scientists write and do research within the paradigm. Anything outside of the paradigm is derided as pseudoscience or worse.

The problem is that when a Paradigm in in effect, all scholars are supposed to publish within the Paradigm. Publishing outside the paradigm is regarded as evidence that one is a kook, a crank, is practicing pseudoscience, or that one is crazy or a fool. It is instructive in this debate to note that most of the prominent lumpers are independent scholars operating outside of the politics of academia.

I have had them tell me that the only reason they can take the lumper position that they do is because they are independent and don’t have a university job, so there are no repercussions if they are wrong. They told me that if they had a professorship, they would not be able to do this work. They have also told me that they know for a fact that certain splitters might jeopardize their jobs, careers, and especially their funding if they took a lumper position. This was given as one of the reasons for their dogmatic splitterism.

In addition, science works according to fads, or more properly, standard beliefs. The trends for these beliefs are set by the biggest names in the field. The biggest names in Linguistics are all splitters now. They are the trendsetters, especially in whatever specialty of Historical Linguistics you are working in. Everyone else in the field is dutifully following in their footsteps. As an up and coming young scholar, you are supposed to follow the proper trends and hypotheses of your field to uphold the consensus of scholars in your area of specialty. As you can see there is a lot more than simple empiricism going on here.

With my background, I look for psychological motivations anywhere I can find them. And science is no stranger to bias and emotional psychological motivations driving, or usually distorting it. We are human and humans have emotions. Emotion is the enemy of logic. Logic is the basis of empiricism. Hence, emotions are the enemy of science.

Scientists are supposed to remain objective, but alas, they are humans themselves and subject to all of the emotional psychological motivations that the rest of them are. Scientists are supposed to police themselves for bias, but that’s probably hard to do, especially if the bias is rooted in psychological processes or in particular if it is unconscious, as many such processes are.

Campbell’s case is an extreme one, but I believe it is simply motivated by internal psychological process inside of the man himself.

Campbell is driven by psychological complexes. His entire turn towards extreme conservatism in this debate was set off by the huge feud he had with Greenberg, and everything since has flowed from that. He took a very angry position that LIA was completely false and did his best to trash its reputation far and wide. This disparagement is still the order of the day, and Greenberg’s name is as good as mud in the field.

Then Campbell generalized his extreme splitterist reaction to LIA out to all of the language families in the world because if he allowed any new families elsewhere in the world, he might have to allow them in the Americas, and he could not countenance that. Note also that Campbell has gone out of his way to specifically attack Greenberg’s four-family split in his proposal for language families in Africa.

This proposal, done with Greenberg’s derided method of mass comparison, has had a successful result in Africa and has been proven with the test of time. Campbell cannot allow this because if he admits that Greenberg was right in Africa, he might have to accept that he might be right in the Americas too, and that’s beyond the pale. So in his recent works he has specifically set out to state that Afroasiatic, Nilo-Saharan, Niger-Kordofanian, and Khoisan – the four families of Greenberg’s classification – have not been proven to exist yet. The truth is exactly the opposite, but the psychological process here is bald and naked for all to see.

Here he specifically trashes these language families because they were discovered by Joseph Greenberg, Campbell’s bete noir. Campbell’s agenda is to show the Greenberg is a preposterous kook and crank, although he was one of the greatest linguists of the 20th century. Greenberg’s African work is regarded as true, and this poses a problem if Campbell is to characterize Greenberg as a charlatan.

If Greenberg was right about one thing, could he not be right about another? In order to lay the foundation for the theory that Greenberg’s method doesn’t work and that it cannot discover any language relationships, Campbell will have to deny the method ever had any successes. So he sets about to deny that Greenberg’s four African families are proven.

Splitters have come up with a repertoire of reasons to shoot down proposed language relations and most are pretty poor.

They rely on overuse of the borrowing, chance, sound symbolism, nursery word, and onomatopoeia explanations for non-relatedness. There is also an overuse of the comparative method with excessively strict standards being set up for etymologies and sound correspondences. In a number of cases, linguists are going back to the etymologies of their proto-languages and reducing them by up to half.

In the last 20 years, Uralicists have gone back over the original Proto-Uralic etymologies and gotten rid of fully half of them (from 2,000 down to 1,000) on a variety of very poor reasons, mostly irregular sound correspondences. It appears to me that while there were some obvious bad etymologies in there, most of the ones that were thrown out were perfectly good.

Irregular sound correspondences is a bad reason to throw out an etymology. Keep in mind that 5

This is not just conservatism. It is out and out Reaction. Worse, it is nearly a Conservative Revolution, which I won’t define further. It is akin to a city council declaring that all of the old, beautiful buildings in the city are going to be torn down because they were not constructed properly. Will they be rebuilt? Well, of course not. Most of the top Uralicists are involved in this silly and destructive project.

In a recent paper, George Starostin warned that the splitters were not just conservatives determined to stop all progress. He pointed out that there was actually a trend towards rejection and going backwards in time to dismantle families that have already set up on the grounds that they were not done perfectly enough. As we can see, his warning was prescient.

There are statements being made by moderates that both sides, the splitters and the lumpers, are being equally unreasonable. As one linguist said, the debate is between lazy lumpers (Just believe us, don’t demand that we prove it!) and angry splitters (Not only is this new family false, but all new families proposed from now on will also be shot down!). He suggested that they are both wrong and that the solution lies in a point in the middle. I don’t have a problem with this moderate centrist belief

The splitter notion itself rests on an obvious falsehood, that there are hundreds of language families in the world that have no possible relationship with each other.

According to Campbell, there are 160 language families and isolates in the Americas. The question is where did all of these entities come from. Keep in mind, in Linguistics, the standard view is that these 160 entities are not related to each other in any way, shape, or form. Thinking back, this means that language would have had to have developed in humans 160 times among the Amerindians alone.

The truth is that there was no polygenesis of language.

Sit back and think for a moment. How could language possibly have been independently developed more than one time? Obviously it arose in one group. How could it have arose in other groups too? It couldn’t and it didn’t. Did some of the original speakers go deaf, become mutes, forget all their language, and  then have children, raising them without language, in which case the children devised language for themselves?

Children need comprehensible input to develop language. No language to hear in the environment, no language for the children to acquire on their own. With coclear implants, formerly deaf people are now able to hear for the first time. A woman got hers at age 32. Since she missed the Critical Period for language development, the window of which closes at age 8, she  has not, even at this late  date, been able to acquire language satisfactorily. She missed the boat. No input, no language.

Obviously language arose only once among humans. It had to. And hence, all human languages are related to each other de facto whether we can “prove” it by out fancy methods or not. In other words, all human languages are related. Those 160 language families and  isolates in the Americas? All related. Now we may not be able to prove which languages they are related to specifically and most closely, but we know they are all related to each other.

In the physical sciences, including Evolutionary Psychology, many things are simply assumed because the alternate theories could not have happened. But we have no evidence of much of anything in Evolutionary Psychology or Evolutionary Anthropology. We know our ancestors lived in X place at Y times, but we have no idea what they were doing there. We can’t go back in time to prove that this or that happened.

Using the logic of linguists, since we cannot make time machines to go back in time and make theories about Evolutionary Anthropology and Evolutionary Psychology of these peoples, we can make no statements about this matter, as the only way to prove it would be to see it. In physics, there are particles that we have never seen. We have simply posited their existence because according to our theories, they have to exist. According to linguists, we could not posit the discovery of these particles unless we see it.

Contrary to popular rumor, everything in science does not have to be “proven” by this or that rigorous method. Many things are simply posited, as no real evidence for their existence exists, either because we were not there or because we can’t see them, or in the case of pure physics, we can’t even test out our theories. They exist simply because they have to according to our existing theories, and all competing theories fall down flat.

Well, the Americanists beg to disagree. Greenberg’s theory was so extreme and radical that the entire field erupted in outrage. None of their alternate theories, not even one of them, make the slightest bit of sense.

Despite the fact that these languages are obviously related to each other, in order to “officially prove it” we have to use a method called the comparative method whereby proto-languages and families are reconstructed and regular sound correspondences are shown between the languages being studied.

This is the only way that we can prove one language is related to another. That’s simply absurd for a few reasons.

First of all, I concur with Joanna Nichols that the comparative method does not really work on language families older than 6-8,000 years. Beyond that time, so many sound changes have taken place, semantics have been distorted, and terms fallen out of use that there’s not much of anything left to reconstruct. Furthermore, time has washed away any evidence of sound correspondences.

Although Nichols is a splitter, I have to commend her. First, she’s right above.

Second, realizing this, she says that the comparative method will always fail beyond this time frame. I believe she thinks then that we need to use new methods if we are to prove that long-range families exist. The method she suggests is “individual-identifying evidence,” which seems to be another way of saying odd morpheme paradigms that were probably not borrowed and are hardly existent outside of that family.

This harkens back to Edward Sapir’s “submerged features,” where he says we can prove the existence of language families by these small morphemic resemblances alone.

The rest of the field remain sticks in the mud. They say that we must use the comparative method to discover that languages are related because no other method exists. The problem is that as noted, as splitters themselves note, if the comparative method fails beyond 6,000 years back, all attempts to prove language families that old or older are bound to fail.

The splitters seem positively gleeful that according to their paradigm, few if any new language families will be discovered. This delight in nihilism seems odd and disturbing. What sort of science is gleeful that no new knowledge will be found? Even in the even that this is true, it’s depressing. Why get excited about something so negative?

Many language families in the world were discovered by Greenberg’s “mass comparison” or simply comparing one language to another, which should be called “comparison.” And in fact, many of the smaller language families in the world are still being posited by the means of comparison or mass comparison. Comparison need not be the broad, sweeping, forest for the trees, holistic method Greenberg employs. I argue that it means lining up languages and looking for common features. We could be lining up one language against another and that would also be “comparison.”

It need not be a shallow examination. One could examine a possible language for five, ten, fifteen, or twenty years.

After studying a pair or group of languages for some time, if one finds a group of core vocabulary items that resemble one another and are above the rate found by chance (

I fail to understand why examining a language or group of languages for a long period of time to find resemblances and try to rule out chance or borrowings is a ridiculous method. What’s so ridiculous about that? Sure, it’s nice to reconstruct and get nice sound correspondences going, but it’s not always necessary, especially in long-range comparisons when such methods are doomed to failure.

One more thing: if splitters say that the comparative method fails beyond 6,000 years, why do they keep putting long-range families to the test using the comparative method? After all, the result will always come up negative, right? What’s the point of doing a study you know will come up negative? Just to get your punches in?

There are a number of folks who have bought into the splitters’ arguments and are trying to discover long-range families by the comparative method of reconstructing the proto-language and finding regular sound correspondences between them. A number of them claim to have been successful. There have been attempts to reconstruct proto-languages and find regular sound correspondences with Altaic, Nostratic, Dene-Caucasian, Dene-Yenisien, Austro-Tai, Totonozoquean, and Uralo-Yukaghir.

Altaic, Nostratic, and Dene-Caucasian all have proto-languages reconstructed with good sound correspondences running through them. Altaic and Nostratic have etymological dictionaries containing many words, 2,300 proto-forms in the case of Altaic in a 1,000 page volume. Further, a considerable Nostratic proto-language was reconstructed by Dogopolsky and Illich-Svitych.

All of these efforts claim that they have proven their hypotheses. However, the splitters such as Campbell have rejected all of them. So you see, even when people follow the mandated method and play it by the book the way they are supposed to, the splitters will nearly always say that the efforts come up short. It’s a rigged game.

How about another question? If the comparative method fails is doomed beyond 6,000 years, why don’t we use another method to discover these relationships? The splitter rejoinder is that there is no other method. It’s the comparative method or nothing. But how do they know this? Can they prove that other methods can never be used to successfully discover a language relationship?

The following quotes are from a textbook or general text on Historical Linguistics by Lyle Campbell and Mario Mixco, A Glossary of Historical Linguistics. The purpose of this paper will be misrepresented as critics who will say that I am a lumper who is saying criticizing splitters for their opposition to known language families.

There is some of that here, but more than lumper propaganda, what I am trying to do here more than anything else is to show how Campbell and Mixco have been untruthful about linguistic specialist consensus regarding these families. In most cases, they are openly misrepresenting the state of consensus in the field.

As will be shown, Campbell and Mixco repeatedly seriously distort the state of consensus regarding many language families, particularly long-range ones. They usually favor a more negative and conservative view, saying that a family has little support when it has significant support and saying it is controversial when the consensus in the field is that the family is real. Campbell and Mixco engage in serious distortions of fact all through this text:

Campbell and Mixco:

Afroasiatic: Enjoys wide support among linguists, but it is not uncontroversial, especially with regard to which of the groups assumed to be genetically related to one another are to be considered true members of the phylum.

There is disagreement concerning Cushitic, and Omotic (formerly called Sidama or West Cushitic) is disputed; the great linguistic diversity within Omotic makes it a questionable entity for some. Chadic is held to be uncertain by others. Typological and areal problems contribute to these doubts. For example, some treat Cushitic and Omotic together as a linguistic area (Sprachbund) of seven families within Afroasiatic.

Campbell and Mixco are wrong. Afroasiatic is not controversial at all. There is widespread consensus that the family exists and that all of the subfamilies are correct.

The “we can’t reconstruct the numerals” argument is much in evidence here too. See the Altaic debate below for more on this. One argument against Altaic is “We can’t reconstruct the numerals.” However, Afroasiatic is a recognized family and not only has reconstruction itself proved difficult, but the numerals in particular are a gigantic mess. It seems that one does not need to have a fully reconstructed numeral set after all to have a proven language family.

There is consensus that Cushitic is a valid entity. Granted, there has been some question about Omotic, but in the last 10-15 years, consensus has settled on an agreement that Omotic is part of Afroasiatic.

The great diversity of Omotic is no surprise. Omotic is probably 13,000 years old! It’s amazing that there’s anything left at all after all that time.

Where do we get the idea that a language family cannot possibly be highly diverse? Chadic is also uncontroversial by consensus. I am not aware of any serious proposals to see Cushitic and Omotic as an Altaic-like Sprachbund of mass borrowings. Campbell and Mixco’s comments above are simply not correct. The only people questioning the validity of Afroasiatic or any of its components are Campbell and Mixco, and they are not an experts on the family.

Campbell and Mixco:

Berber is usually believed to be one of the branches of Afroasiatic.

This is far too pessimistic. Berber is recognized by consensus as being one of the branches of Afroasiatic.

Campbell and Mixco:

Niger-Kordofanian (now often just called Niger-Congo): A hypothesis of distant genetic relationship proposed by Joseph H. Greenberg in his classification of African languages. Estimated counts of Niger-Kordofanian languages vary from around 900 to 1,500 languages. Greenberg grouped ‘West Sudanic’ and Bantu into a single large family, which he called Niger-Congo, after the two major rivers, the Niger and the Congo ‘in whose basins these languages predominate’ (Greenberg 1963: 7).

This included the subfamilies already recognized earlier: (1) West Atlantic (to which Greenberg joined Fulani, in a Serer-Wolof-Fulani [Fulfulde] group), (2) Mande (Mandingo) (thirty-five to forty languages), (3) Gur (or Voltaic), (4) Kwa (with Togo Remnant) and (5) Benue-Congo (Benue-Cross), with the addition of (6) Adamawa-Eastern, which had not previously been classified with these languages and whose classification remains controversial.

For Greenberg, Bantu was but a subgroup of Benue-Congo, not a separate subfamily on its own. In 1963 he joined Niger-Congo and the ‘Kordofanian’ languages into a larger postulated phylum, which he called Niger-Kordofanian.

Niger-Kordofanian has numerous supporters but is not well established; the classification of several of the language groups Greenberg assigned to Niger-Kordofanian is rejected or revised, though most scholars accept some form of Niger-Congo as a valid grouping.

As Nurse (1997: 368) points out, it is on the basis of general similarities and the noun-class system that most scholars have accepted Niger-Congo, but ‘the fact remains that no one has yet attempted a rigorous demonstration of the genetic unity of Niger-Congo by means of the Comparative Method.’

There is consensus among scholars that Niger-Kordofanian is a real thing.

Campbell and Mixco:

Nilo-Saharan: One of Greenberg’s four large phyla in his classification of African languages. In dismantling the inaccurate and racially biased ‘Hamitic,’ of which Nilo-Hamitic was held to be part, Greenberg demonstrated the inadequacy of those former classifications and argued for the connection between Nilotic and Eastern Sudanic.

He noted that ‘the Nilotic languages seem to be predominantly isolating, tend to monosyllabism, and employ tonal distinctions’ (Greenberg 1963: 92). To the extent that this classification is based on commonplace shared typology and perhaps areally diffused traits, it does not have a firm foundation. Nilo-Saharan is disputed, and many are not convinced of the proposed genetic relationships. It is generally seen as Greenberg’s wastebasket phylum, into which he placed all the otherwise unaffiliated languages of Africa.

First of all, Nilo-Saharan is not classified based on its language typology which were perhaps areally diffused. There is also a great deal of the more typical evidence in favor of this language family. Second,  it is not true that it lacks a firm foundation and that many are not convinced of its reality. The consensus among experts is that this family exists and the overwhelming majority of the subfamilies and isolates Greenberg put it in are correct.

Saying that it is a wastebasket phylum does not make sense because the Nilo-Saharan languages are only found in  a certain part of Africa. If it was truly such a phylum, there would be languages from all over Africa placed in this family.

According to Roger Bench, a moderate, there is now consensus in the last 10-15 years that Nilo-Saharan is a real thing.

Consensus has formed that 7

Yes, Campbell and Mixco say that Nilo-Saharan is not real, but they are not specialists.

Campbell and Mixco:

Khoisan: A proposed distant genetic relationship associated with Greenberg’s (1963) classification of African languages, which holds some thirty non-Bantu click languages of southern and eastern Africa to be genetically related to one another. Greenberg originally called his Khoisan grouping ‘the Click Languages’ but later changed this to a name based on a created compound of the Hottentots’ name for themselves, Khoi, and their name for the Bushmen, San.

Khoisan is the least accepted of Greenberg’s four African phyla. Several scholars agree in using the term ‘Khoisan’ not to reflect a genetic relationship among the languages but, rather, as a cover term for all the non-Bantu and non-Cushitic click languages.

Although it is probably true that Khoisan is the least accepted of Greenberg’s families, that’s not saying much, as it only means that 8

According to George Starostin, in the last 5-10 years, there is now consensus that Khoisan exists. There are five major Khoisan scholars, and four of them agree that Khoisan is real, with all of them including Sandawe and most including Hadza. There is one, Traill, who says it’s not real, but he is also a notorious Africanist splitter.

Campbell and Mixco:

Eurasiatic: Greenberg’s hypothesis of a distant genetic relationship that would group Indo-European, Uralic–Yukaghir, Altaic, Korean–Japanese–Ainu, Nivkh, Chukotian and Eskimo–Aleut as members of a very large ‘linguistic stock’. While there is considerable overlap in the putative members of Eurasiatic and Nostratic there are also significant differences. Eurasiatic has been sharply criticized and is largely rejected by specialists.

I have no doubt that Eurasiatic has been sharply criticized, but apart from a negative review in Language by Peter Daniels, the controversy seems quite muted compared to the furor over Amerind. I am also not sure that it is largely rejected by specialists. It probably is, but most of them have not even bothered to comment on it. I believe that this family is one of the best long-range proposals out there.

Based on the data from the pronouns alone, it’s obviously a real entity, though I would include Indo-European, Uralic-Yukaghir, Altaic including Japanese and Korean, Chukotian, and Eskimo-Aleut, leaving out Nivki for the time being and certainly leaving out Ainu. Nivki does seem to be a Eurasiatic language but it’s not a separate node. Instead it may be a part of the Chukotian family. Or even better yet, it seems to be part of a family connected to the New World via the Almosan family in the Americas.

I feel that Eurasiatic is a much more solid entity than Nostratic. Not that I am against Nostratic, but it’s more that Eurasiatic is a simple hypothesis to prove and with Nostratic, I’m much less sure of that. On the other hand, to the extent that Nostratic overlaps with Eurasiatic, it is surely correct.

Campbell and Mixco:

Indo-Anatolian: The hypothesis, associated with Edgar Sturtevant, that Hittite (or better said, the Anatolian languages, of which Hittite is the best known member) was the earliest Indo-European language to split off from the others. That is, this hypothesis would have Anatolian and Indo-European as sisters, two branches of a Proto-Indo-Hittite.

The more accepted view is that Anatolian is just one subgroup of Indo-European, albeit perhaps the first to have branched off, hence not ‘Indo-Hittite’ but just ‘Indo-European’ with Anatolian as one of its branches. In fact the two views differ very little in substance, since, in either case, Anatolian ends up being a subfamily distinct from the other branches and in the view of many the first to branch off the family.

The view that Anatolian is just another subgroup of IE is not the more accepted view. In fact, it has been rejected by specialists. Indo-Europeanists have told me that Indo-Anatolian is now the consensus among Indo-Europeanists, so Campbell and Mixco’s statement that Indo-Anatolian is a minority view is false.

Campbell and Mixco:

Nostratic (< Latin nostra ‘our’): A proposed distant genetic relationship that, as formulated in the 1960s by Illich-Svitych, would group Indo-European, Uralic, Altaic, Kartvelian, Dravidian and Hamito-Semitic (later Afroasiatic), though other versions of the hypothesis would include various other languages. Nostratic has a number of supporters, mostly associated with the Moscow school of Nostratic, though a majority of historical linguists do not accept the claims.

There are many problems with the evidence presented on behalf of the Nostratic hypothesis. In several instances the proposed reconstructions do not comply with typological expectations; numerous proposed cognates are lax in semantic associations, involve onomatopoeia, are forms too short to deny chance, include nursery forms and do not follow the sound correspondences formulated by supporters of Nostratic.

A large number of the putative cognate sets are considered problematic or doubtful even by its adherents. More than one-third of the sets are represented in only two of the putative Nostratic branches, though by its founder’s criteria, acceptable cases need to appear in at least three of the Nostratic language families. Numerous sets appear to involve borrowing. (See Campbell 1998, 1999.) It is for reasons of this sort that most historical linguists reject Nostratic.

It is probably correct that consensus among specialists is to reject Nostratic, but serious papers taking apart of the proposal seem to be lacking. Nevertheless, most dismiss it and it is beginning to enter into the emotionally charged terrain of Altaic and Amerind, particularly the former, and belief in it is becoming a thing of ridicule as it is for Altaic. Nevertheless, there have been a few excellent linguists doing work on this very long-range family for decades now.

Campbell and Mixco:

Indo-Uralic: The hypothesis that the Indo-European and Uralic language families are genetically related to one another. While there is some suggestive evidence for the hypothesis, it has not yet been possible to confirm the proposed relationship.

This summary seems too negative. Indo-Uralic is probably one of the most promising long-range proposals out there. I regard the relationship between the two as obvious, but to me it is only a smaller part of the larger Eurasiatic family. Frederick Kortland has done a lot of good work on this idea. Even some hardline splitters are open to this hypothesis.

Campbell and Mixco:

Altaic: While ‘Altaic’ is repeated in encyclopedias and handbooks most specialists in these languages no longer believe that the three traditional supposed Altaic groups, Turkic, Mongolian and Tungusic, are related. In spite of this, Altaic does have a few dedicated followers.

The most serious problems for the Altaic proposal are the extensive lexical borrowing across inner Asia and among the ‘Altaic’ languages, lack of significant numbers of convincing cognates, extensive areal diffusion and typologically commonplace traits presented as evidence of relationship.

The shared ‘Altaic’ traits typically cited include vowel harmony, relatively simple phoneme inventories, agglutination, their exclusively suffixing nature, (S)OV ([Subject]-Object-Verb) word order and the fact that their non-main clauses are mostly non-finite (participial) constructions.

These shared features are not only commonplace typological traits that occur with frequency in unrelated languages of the world and therefore could easily have developed independently, but they are also areal traits shared by a number of languages in surrounding regions the structural properties of which were not well-known when the hypothesis was first framed.

This one is still up in the air, but Campbell and Mixco are lying when they say that idea has been abandoned. Most US linguists regard it as a laughingstock, and if you say you believe in it you will experience intense bullying and taunting from them. Oddly enough, outside the US, in Europe in particular, Altaic is regarded as obviously true. However, notorious anti-Altaicist Alexander Vovin has camped out in Paris and is now spreading his nihilistic doctrine to Europeans there.

The problem is that almost all of the US linguists who will laugh in your face and call you an idiot if you believe in Altaic are not specialists in the language. However, I did a study of Altaic specialists, and 7

So the anti-Altaicists are pushing a massive lie – that critical consensus has completely abandoned Altaic and regards as a laughingstock, but their project is more Politics and Propaganda than Science. In particular, it’s a fad. So Altaic is in the preposterous position where almost all of the people who know nothing about it will laugh in your face and call you an idiot if you believe in it and the overwhelming majority of specialists will say it’s real.

Altaic must be the only nonexistent family that has an incredibly elaborate 1,000 page etymological dictionary, full reconstructions of the proto-languages, etymologies of over 2,000 Altaic terms, and elaborate sound correspondences running through it. The anti-Altaicists use the silly “we can’t reconstruct the numerals so it’s not real” line here.

Altaic is obviously true based on 1-2 person pronoun paradigms at an absolute minimum. The anti-Altaic argument of course, is preposterous. As noted, they dismiss a vast 1,000 page Etymological Dictionary with 2,300 reconstructed etymologies as a hallucinated work.

There are vast parallels in all three families at all levels, in particular in the Mongolic-Tungusic family, which gets a 10

The argument that entire 1-2 pronoun paradigms have been borrowed is particularly preposterous because 1-2 pronouns are almost never borrowed anyway, and there has never been a single case of on Earth of the borrowing of a 1-2 person pronoun paradigm, much less the borrowing of one at the proto-language level. So the anti-Altaicists are arguing that something that has never happened anywhere on Earth not only happened, but happened more than once among different proto-languages. So the anti-Altaic argument is that something that could not possibly have happened actually occurred.

This is the conclusion of every paper the splitters write. Something that has never occurred on Earth and probably could not possibly happen not only occurred, but occurred many times around the globe for thousands of years.

Many regard including Japonic and Koreanic in Altaic as dubious, although having looked over the data, I am certain that they are part of Altaic. But they seem to be further away from the traditional tripartite system than the traditional three families are to each other. If we follow the theory that Japanese and Korean have been split from Proto-Altaic for 8,000 years, this starts to make a lot more sense.

The ridiculous massive borrowings argument specifically fails for geographical reasons. Proto-Turkic was never next door to Proto-Mongolic and Proto-Tungusic. The Proto-Altaic homeland is in the Khingan Mountains in Western Manchuria and Eastern Mongolia. Tungusic split off from Altaic 5,300 years ago, leaving Proto-Turkic-Mongolic in Khingans. 3,400 years ago, Proto-Turkic broke from Proto-Turkic-Mongolic and headed west to Northern Kazakhstan and the southern part of the Western Siberian Plain, leaving Mongolic alone in the Khingans.

Proto-Transeurasian – Khingans 9,000 YBP

Proto-Korean – Liaojiang on the north shore of the Bohai Sea 8,000 YBP.

Proto-Japanese – Northern coast of the Shandong Peninsula on the southern shore of the Bohai Sea 8,000 YBP

Proto-Tungusic – Amur Peninsula 5,300 BP. Breaks apart 2,000 YBP.

Proto-Turkic – Northern Kazakhstan 3,400 BP.

Proto-Mongolic – Khingans 3,400 BP.

Can someone explain to me how Mongolic and Tungusic borrow from Turkic 3,000 miles away in a different place at a different time in this scenario? Can someone explain to me how any of these proto-languages borrowed from each other at all, especially as they were in different places at different times?

Not only that but supposedly both Proto-Mongolic and Proto-Tungusic each borrowed from Proto-Turkic separately. These borrowings included massive amounts of core vocabulary in addition to an entire 1st and 2nd person pronoun paradigm.

Keep in mind that the borrowing of this paradigm, something that has never happened anywhere, supposedly occurred not just once but twice, between Proto-Tungusic 5,300 YBP on the Amur from Proto-Turkic in North Kazakhstan 3,000 miles away 2,000 later, and at the same time, between  Proto-Mongolic in the Khingans and Proto-Turkic in Northern Kazakhstan 3,000 miles away. How exactly did this occur?

And can someone explain to me how Proto-Korean and Proto-Japanese borrow from either of the others under this scenario?

Campbell and Mixco:

Turkic: A family of about thirty languages, spoken across central Asia from China to Lithuania. The family has two branches: Chuvash (of the Volga region) and the non-Chuvash Turkic branch of relatively closely related languages. Some of the Turkic languages are Azeri, Kyrgyz, Tatar, Crimean Tatar, Uighur, Uzbek, Yakut, Tuvan, and Tofa. Turkic is often assigned to the ‘Altaic’ hypothesis, though specialists have largely abandoned Altaic.

As noted above, it is simply incorrect that specialists have largely abandoned Altaic. This is simply carefully crafted propaganda on the part of Campbell and Mixco. In fact, my own study showed that 7

Campbell and Mixco:

Some scholars classify Korean in a single family with Japanese; however, this is a controversial hypothesis. Korean is often said to belong with the Altaic hypothesis, often also with Japanese, though this is not widely supported.

Japonic-Koreanic has considerable support among specialists in these languages, although it is not universally accepted. Campbell and Mixco are excessively negative about the level of support for an expanded Altaic. In fact, an expanded Altaic which includes Japanese and Korean in some part of it has significant though probably not majority support. Perhaps 30-4

Shandong Peninsula with Tianjin and Liaojiang across the Bohai Sea, location of the Proto-Japonic and Proto-Korean homelands.

Proto-Japanic and Proto-Koreanic were both spoken in Northeastern China 8,000 YBP. Proto-Japonic was spoke on the north of the Shandong Peninsula and Proto-Koreanic was spoken across the Bohai Sea in Tianjin and especially across the Bohai Straights on the Liaodong Peninsula. They may have stayed here next to each other for 3,000 years until the Proto-Koreanics moved to the Korean Peninsula 5,000 YBP, displacing the Ainuid types there. Proto-Japonics probably stayed in Shandong until 2,3000 YBP when they left to populate Japan and the Ryukus, displacing the Ainu who were already there.

Campbell and Mixco:

Yeniseian, Yenisseian: Small language family of southern Siberia of which Ket (Khet) is the only surviving member. Yeniseian has no known broader relatives, though some have been hypothesized (see the Dené-Caucasian hypothesis).

Campbell and Mixco state and serious untruth here, including some weasel words. By discussing Dene-Caucasian in the same breath as relatives of Yenisien, they are able to deflect away from the more widely accepted proposal of a link between Yenisien in the Old World and Na-Dene in the New World. This is Edward Vajda’s Dene-Yenisien proposal.

The problem is that this long-range proposal has the support of many people, including splitter Johanna Nichols. Of the 17 experts who weighed in on Dene-Yenisien, 15 of them had a positive view of the hypothesis. Campbell and Mixco are the only two who are negative, but neither are experts on either family. All specialists in either or both families support the proposal. When 15 out of 17 is not enough, one wonders at what point the field reaches a consensus. Must we hold out for Campbell and Mixco’s approval for everything?

Campbell and Mixco:

Nivkh (also called Gilyak): A language isolate spoken in the northern part of Sakhalin Island and along the Amur River of Manchuria, in China. There have been various unsuccessful attempts to link Nivkh genetically with various other language groupings, including Eurasiatic and Nostratic.

Granted, there is no consensus on the affiliation of Nivkhi. However, a recent paper by Sergei Nikolaev proved to me that Nivkhi is related to Algonquian-Wakashan, a family of languages in the Americas. One of these languages is Wakashan, and there has been talk of links between Wakashan and the Old World for some time.

Michael Fortescue places Nivkhi in Chukotko-Kamchatkan. Greenberg places it is Eurasiatic as a separate node. But as Chukotko-Kamchatkan is part of Eurasiatic, they are both saying the same thing in a way. My theory is that Nivkhi is Eurasiatic, possibly related to Chukoto-Kamchatkan, and like Yeniseian, is also connected to languages in North America as some of the Nivkhi probably migrated to North America and became the American Indians. In this way, we can reconcile both hypotheses.

There are three specialist views on Nivkhi. One says it is Eurasiatic, the other that it is Chukotian, and the third that it is part of the Algonquian-Wakashan or Almosan family in the New World. Consensus is that Nivkhi is related to one of two other entities – other languages in Northeastern Asia or a New World Amerindian family. So expert consensus seems to have moved away from the view of Nivkhi as an isolate.

Campbell and Mixco:

Paleosiberian languages (also sometimes called Paleoasiatic, Hyperborean languages): A geographical (not genetic) designation for several otherwise unaffiliated languages (isolates) and small language families of Siberia.

Perhaps the main thing that unites these languages is that they are not Turkic, Russian or Tungusic, the better known languages of Siberia. Languages often listed as Paleosiberian are: Chukchi, Koryak, Kamchadal (Itelmen), Yukaghir, Yeniseian (Ket) and Nivkh (Gilyak). These have no known genetic relationship to one other.

Taken as a broad statement, of course this is true. However, Chukchi, Koryak, and Kamchadal or Itelmen are part of a family called Chukutko-Kamchatkan. This family has even been reconstructed. Campbell and Mixco’s statement that these languages have no known genetic relationship with each other is false.

Campbell and Mixco:

Austroasiatic: A proposed genetic relationship between Mon-Khmer and Munda, accepted as valid by many scholars but not by all.

The fact is that Austroasiatic is not a “proposed genetic relationship.” Instead it is now accepted by consensus. That there may be a few outliers who don’t believe in it is not important. I’m not aware of any linguists who doubt Austroasiatic other than Campbell and Mixco, and neither is a specialist. Austroasiatic-Hmong-Mien is the best long-range proposal for Austroasiatic, but it has probably not yet been proven. Austroasiatic is also part of the expanded version of the Austric hypothesis.

Campbell and Mixco:

Miao-Yao (also called Hmong-Mien): A language family spoken by the Miao and Yao peoples of southern China and Southeast Asia. Some proposals would classify Miao-Yao with Sino-Tibetan, others with Tai or Austronesian; none of these has much support.

This seems to be more weasel wording on the part of the authors. By listing Tai or Austronesian and Sino-Tibetan as possible relatives of Miao-Yao and then correctly dismissing it, they leave out a much better proposal linking Hmong-Mien to Austroasiatic.

This shows some promise, but the relationship is hard to see amidst all of the Chinese borrowing. As noted, the relationship between Hmong-Mien and Sino-Tibetan is one of borrowing. The relationship with Tai or Austronesian is part of Paul Benedict’s original Austric proposal. He later turned against this proposal and supported a more watered down Austric with Austronesian and Tai-Kadai, which seems to be nearing consensus support now.

Campbell and Mixco:

Austric: A mostly discounted hypothesis of distant genetic relationship proposed by Paul Benedict that would group together the Austronesian, Tai-Kadai and Miao-Yao.

More weasel wording. It is correct that Benedict’s original Austric (which also included Austroasiatic) was abandoned even by Benedict himself, a more watered down Austric that he later supported consisting of Austronesian and Tai-Kadai called Austro-Tai has much more support. They get around discussing the watered down Austro-Tai with good support by limiting Austric to Benedict’s own theory which even he rejected later in life. In this sense, they misrepresent the debate, probably deliberately.

In fact, evidence is building towards acceptance of Austro-Tai after papers by Weera Ostapirat and Laurence Sagart seem to have proved the case using the comparative method. Roger Blench also supports the concept. In addition, to Benedict, it is also supported by  Lawrence Reid, Hui Li, and Lawrence Reid. It is opposed by Graham Thurgood, who is a specialist (he was my main academic advisor on my Master’s Degree in Linguistics). It is also opposed by Campbell and Mixco, but they are not specialists. Looking at expert opinion, we have seven arguing for the theory and one arguing against it. Specialist consensus then is that Austro-Tai is a real language family.

Even the larger version of Austric, including all of Benedict’s families plus Ainu and the South Indian isolate Nihali, has some supporters and some suggestive evidence that it may be correct.

Campbell and Mixco:

Tai-Kadai: A large language family, generally but not universally accepted, of languages located in Southeast Asia and southern China. The family includes Tai, Kam-Sui, Kadai and various other languages. The genetic relatedness of several proposed Tai-Kadai languages is not yet settled.

Tai-Kadai is not “mostly but not universally accepted.” It is accepted by consensus as an existent language family. Perhaps whether some languages belong there is in doubt but the proposal itself is not controversial. Campbell and Mixco’s statement that Tai-Kadai remains controversial is a serious distortion of fact.

Campbell and Mixco:

Na-Dene: A disputed proposal of distant genetic relationship, put forward by Sapir, that would group Haida, Tlingit and Eyak-Athabaskan. There is considerable disagreement about whether Haida is related to the others. The relationship between Tlingit and Eyak-Athabaskan seems more likely, and some scholars misleadingly use the name ‘Na-Dené’ to mean a grouping of these two without Haida.

Levine and Michael Krauss, two top Na-Dene experts, are on record as opposing the addition of Haida to Na-Dene for 40 years. A recent conference about Edward Vajda’s Dene-Yenisien concluded that there was no evidence to include Haida in Na-Dene. However, a recent paper by Alexander Manaster-Ramer made the case that Haida is part of Na-Dene. This paper was enough to convince me. Further, the scholar with the most expertise on Haida has said that Haida is part of Na-Dene. So Campbell and Mixco are correct here that the subject is up in the air with both supporters and opponents.

The statement that a relationship between Tlingit and Eyak-Athabaskan seems “more than likely” is an understatement. I believe it is now linguistic consensus that Tlingit is part of Na-Dene, so Campbell and Mixco’s statement is not quite true.

Campbell and Mixco:

Tonkawa: An extinct language isolate of Texas. Proposals to link Tonkawa with the languages of the Coahuiltecan or Hokan-Coahuiltecan hypotheses have not generally been accepted.

I’m sure it is the case that Coahuiltecan and Hokan-Coahuiltecan affiliations of Tonkawa have been rejected. A Coahuiltecan connection was even denied by Manaster-Ramer, who recently proved that the family existed. That said, there are interesting  parallels between Tonkawa and Coahuiltecan that I cannot explain. However, a recent paper by Manaster-Ramer made the much better case that Tonkawa was in fact Na-Dene.

Campbell and Mixco:

Amerind: The Amerind hypothesis is rejected by nearly all practicing American Indianists and by most historical linguists. Specialists maintain that valid methods do not at present permit classification of Native American languages into fewer than about 180 independent language families and isolates. Amerind has been highly criticized on various grounds.There is an excessive number of errors in Greenberg’s data.

Where Greenberg stops – after assembling superficial similarities and declaring them due to common ancestry – is where other linguists begin. Since such similarities can be due to chance similarity, borrowing, onomatopoeia, sound symbolism, nursery words (the mama, papa, nana, dada, caca sort), misanalysis, and much more, for a plausible proposal of remote linguistic relationship one must attempt to eliminate all other possible explanations, leaving a shared common ancestor as the most likely.

Greenberg made no attempt to eliminate these other explanations, and the similarities he amassed appear to be due mostly to accident and a combination of these other factors.

In various instances, Greenberg compared arbitrary segments of words, equated words with very different meanings (for example, ‘excrement/night/grass’), misidentified many languages, failed to analyze the morphology of some words and falsely analyzed that of others, neglected regular sound correspondences, failed to eliminate loanwords and misinterpreted well-established findings.

The Amerind ‘etymologies’ proposed are often limited to a very few languages of the many involved. Finnish, Japanese, Basque and other randomly chosen languages fit Greenberg’s Amerind data as well as or better than do any of the American Indian languages in his ‘etymologies’; Greenberg’s method has proven incapable of distinguishing implausible relationships from Amerind generally. In short, it is with good reason Amerind has been rejected.

The movement into the Americas came in three waves.

The first wave brought the Amerinds. It is here where the 160 language families reside. According to the reigning theory in Linguistics, this group of Amerindians came in one wave that spoke not only 160 different languages but spoke languages that came from 160 different language families, none of which were related to each other. These being language families which, by the way, we can find scarcely a trace of in the Old World.

The second wave was the Na-Dene people who came along the west coast and then went inland.

The last wave were the Inuits.

Greenberg simply lumped all of the 600 languages of the  Americas into a single family. The argument was good, though I’m not sure he proved that every single one of those languages were all part of Amerind. But a lot of them were. The n- m- 1st and 2nd person pronouns are found in 450 of those languages. The ablauted t’ana, t’una, t’ina word, meaning respectively human child  of either sex, all females including family terms, and all males including family terms are extremely common in Amerind.

So t’ana just means child. T’una means girl, woman, and includes various names for all sorts of female relatives – grandmother, cousin, aunt, niece, etc. T’ina means boy, man, and includes the family terms grandfather, brother-in-law, uncle, cousin, and  nephew. This ablauted paradigm is found across a vast number of these Amerind languages, and it is nonexistent in the rest of the world.

Quite probably most to all of those languages having that term are part of a single family. What are the other arguments? That 300 languages independently innovated these terms, in this precise ablauted paradigm, on their own? What is the likelihood of that?

That these items occurring across such vast swathes of languages is due to chance? But this paradigm does not exist anywhere else, so how could it be due to chance? That these core vocabulary items were borrowed massively all across the Americas, when family terms like that are rarely borrowed? That’s not possible. None of the alternate theories make the slightest bit of sense.

Hence, the Amerind languages that have the n- m- pronoun paradigm and the t’ana, t’una, t’ina ablauted names for the sexes and the terms of family relations by sex are quite probably part of a huge language family. I’m well aware that a few of the languages having those terms could be due to chance. I’m pretty sure that about zero of those pronouns and few, if any, of those family terms were borrowed.

However, not all Amerind languages have either the pronoun paradigm or the ablauted sex term. In those cases, I’m unsure if those languages are all part of the same language. But if you can put those languages in families and reconstruct to the proto-languages and end up with the pronoun paradigm or the ablauted family term reconstructed in the proto-language of that family, I’m sure that family would be part of Amerind. That’s about all you have to do to prove relationship in Amerind.

Campbell and Mixco:

Penutian: A very large proposed distant genetic relationship in western North America, suggested originally by Dixon and Kroeber for the Californian language families Wintuan, Maiduan, Yokutsan, and Miwok-Costanoan. The name is based on words for ‘two’, something like pen in Wintuan, Maiduan, and Yokutsan, and uti in Miwok-Costanoan, joined to form Penutian.

Sapir, impressed with the hypothesis, attempted to add an Oregon Penutian (Takelma, Coos, Siuslaw, and ‘Yakonan’), Chinook, Tsimshian, a Plateau Penutian (Sahaptian, ‘Molala-Cayuse,’ and Klamath-Modoc) and a Mexican Penutian (Mixe-Zoquean and Huave).

The Penutian grouping has been influential, and later proposals have attempted to unite various languages from Alaska to Bolivia with it. Nevertheless, it had a shaky foundation based on extremely limited evidence, and, in spite of extensive later research, it did not prove possible to demonstrate any version of the Penutian hypothesis and several prominent Penutian specialists abandoned it. Today it remains controversial and unconfirmed, with some supporters but with many who doubt it.

The statement that today it “remains controversial and unconfirmed, with some supporters but with many who doubt it,”  has no basis in fact. It is surely controversial and it is probably unconfirmed by linguistic consensus. Yes, it has a number of supporters, and there are quite a few who doubt it. However, among those who doubt it, none of them are specialists in these languages. Hence, we are dealing with an Altaic situation here, where the specialists believe in it but the non-specialists insist it’s nonsense.

In fact, the consensus among the specialists on these languages is that Penutian exists. A Penutian family comprising Maiduan, Utian (Miwok-Costanoan), Wintuan, Yokutsan, Coosan, Siuslaw, Takelma, and Kalapuyan and Alsean (Yakonan), Chinookan, Tsimshianic, Klamath-Modoc (Lutuami), Cayuse and Molala (Waiilatpuan), Sahaptian has been proven to my satisfaction. I am uncertain of the Penutian status of Mixe-Zoque and Huave (Mexican Penutian), although I believe that Huave and Mixe-Zoque are related to each other, albeit at a very deep time depth of 9,000 years.

Anti-Penutianists have not published a paper in a long time. The last one I remembered was published by William Shipley, and he’s been gone for a while. I am not aware of one expert on these languages who says Penutian does not exist.

Campbell and Mixco:

Cayuse-Molala: A genetic classification no longer believed that linked Cayuse (of Oregon and Washington) and Molala (of Oregon) in a single assumed family. The evidence for this was later shown to be wrong and the hypothesis was abandoned.

According to Campbell and Mixco, Cayuse is an isolate. I assume they see Molala as an isolate too. There probably is no Cayuse-Molala family, but Molala is part of Plateau Penutian, and Cayuse may be part of the same group. Plateau Penutian is part of the Penutian hypothesis, which appears to be true. By not mentioning these facts, Campbell and Mixco’s statement is quite misleading.

Campbell and Mixco:

Mosan: A now abandoned proposal of distant genetic relationship that would group Salishan, Wakashan and Chimakuan together.

Another part of this proposal was that Mosan was part of a larger family with Algonquian called Almosan. An excellent series of papers was published recently by Sergei Nikolaev that validated Almosan and proved to me that it was related to Nivkhi in the Old World.

Michael Fortescue argued a few years before that Mosan was a valid entity and that was related to the Old World language Nivkhi. Recently, Murray Gell-Mann, Ilia Peiros, and Georgiy Starostin also supported Almosan and grouped it with Chukotko-Kamchatkan and Nivkhi. David Beck recently argued that Mosan is a language area or Sprachbund instead of a genetic family.

So far we have four specialists arguing that Mosan exists, and one saying it does not. The consensus among specialists seems to be that Mosan is a valid language family. At any rate, Campbell and Mixco’s statement that this proposal is “now abandoned” is false.

For Almosan, we have four specialists saying it exists and two apparently saying it does not. Expert consensus on Almosan is optimistic.

Hokan: A controversial hypothesis of distant genetic relationship proposed by Dixon and Kroeber among certain languages of California; the original list included Shastan, Chimariko, Pomoan, Karok, and Yana, to which they soon added Esselen, Yuman, and later Chumashan, Salinan, Seri, and Tequistlatecan. Later scholars, especially Edward Sapir, proposed various additions to Hokan. Many ‘Hokan’ specialists doubt the validity of the hypothesis.

It is not true that many Hokan specialists “doubt the validity of the hypothesis.” I can’t remember the last time I saw an anti-Hokan paper. Yes, Campbell, Mixco, and Mithun say Hokan does not exist, but they are not specialists. The consensus among specialists such as Mikhail Zhikov, Terence Kaufman, and Marcelo Jokelsy is that Hokan exists. I have only found one specialist who disagrees with the Hokan hypothesis, and she merely doubts the existence of Ch’imáriko.

I believe that a Hokan family consisting of Karuk, Shasta-Palaihnihan, Ch’imáriko, Yana, Salinan, Pomoan, Yuman, Seri, and Tequistlatecan exists, although I would leave out Chumashan, Washo, and Jicaquean or Tolan. Chumashan is an isolate, and while Washo and Tolan may be Hokan at a very deep time depth, the few possible cognates are not enough to provide evidence of this. I am agnostic on Esselen, which is only known from a 350 word list collected by friars at a California mission.

I have not seen any evidence that Coahuiltecan is Hokan. There is some evidence, though not probative enough for me, that Lencan and Misumalpan may be Hokan. Nevertheless, Lencan and Misumalpan form a language family that has even been accepted by Campbell himself. This is the only long-range family proposal he has supported since the publication of LIA.

Although Campbell’s opinion on many hypotheses may be waved away as he is not an expert on that family or language, Lencan and Misumalpan are right up his alley as he is an expert in languages in Central America. He has focused mostly on Mayan, but he also knows the other languages of the region well.

Campbell and Mixco:

Cochimí–Yuman: A family of languages from Arizona, California and Baja California, with two branches, extinct Cochimí (of Baja California) and the Yuman subfamily (members of which are Kiliwa, Diegueño, Cocopa, Mojave, Maricopa, Paipai, and Walapai–Havasupai–Yavapai, among others). Cochimí–Yuman is often associated with the controversial Hokan hypothesis, though evidence is insufficient to embrace the proposed relationship.

The consensus among experts in the Cochimí–Yuman family, including Mikhail Zhikov and Terence Kaufman, is that it is part of the Hokan family. Campbell disbelieves in the association but he is not an expert. However, Mixco opposes the Hokan affinity of Cochimi-Yuman, and granted, he is actually a specialist on these languages. So among specialists, we have two who support the Hokan association and one who opposes it. The specialist consensus then would be that they are this association is a promising hypothesis, but it is not yet proven. This is different from Campbell and Mixco’s wording, which is more negative.

Campbell and Mixco:

Coahuiltecan: A hypothesis of distant genetic relationship that proposed to group some languages of south Texas and northern Mexico: Coahuilteco, Comecrudo and Cotoname, and sometimes also Tonkawa, Karankawa, Atakapa and Maratino (with Aranama and Solano assumed to be varieties of Coahuilteco).

Sapir proposed a broader classification of Hokan–Coahuiltecan, joining the Coahuiltecan proposal with the broader Hokan hypothesis, and placed this in his even larger Hokan–Siouan super-stock. None of these proposals has proven sufficiently robust to be accepted generally.

I am not aware of any specialists who have recently argued against the existence of Coahuiltecan. Yes, Campbell and Mixco do not accept it, but they are not specialists. A recent paper by Alexander Manaster-Ramer proved the existence of Coahuiltecan to my satisfaction. I believe that a Coahuiltecan family consisting of Comecrudo, Cotoname, Aranama, Solano, Mamulique, Garza, and Coahuilteco absolutely exists. Karankawa is probably a part of this family. I am not aware that any specialist is arguing against the existence of this family at the moment.

I do not think there is good evidence for other postulated languages such as Atakapa and Tonkowa. First of all, Tonkawa is probably Na-Dene as per another paper by Manaster-Ramer. Atakapa is part of the Gulf family. However, I am not yet convinced that Coahuiltecan is as member of the Hokan language family.

Campbell and Mixco:

Gulf: Hypothesis of a distant genetic relationship proposed by Mary R. Haas that would group Muskogean, Natchez, Tunica, Atakapa and Chitimacha, no longer supported by most linguists.

The notion that Gulf is no longer supported by most linguists is simply incorrect. There have only been four linguists who studied this family.

The first was Mary Haas, who also proposed a relationship with Yuki as Yuki-Gulf. Haas was always dubious about Chitimacha’s addition to Gulf.

Greenberg resurrected Yuki-Gulf in LIA.

Pam Munro is an expert on these languages. A while back she published a paper on Yuki-Gulf. I read that paper. The resemblances are so stunning between Muskogean, Natchez, Tunica, Atakapa and Chitimacha that I was shocked that anyone doubted the relationship. Furthermore, the relationship with Yuki and Wappo, a full 2,500 miles away in Northern California, was shocking.

The fourth was Geoffrey Kimball, who concluded that Gulf was probably a family but that this could not be proven.

There evidence for Gulf in Munro’s paper was good, and there even appeared to be sound correspondences running through the relationship. What was shocking about it was that Yuki and Wappo could not possibly have borrowed from Gulf because Gulf is in Louisiana 2,500 miles away. So how did all these resemblances come in? Chance is ruled out. Borrowing could not have happened. Therefore a relationship at least between Yuki and the Gulf languages is obvious.

Munro’s paper took the position that Greenberg’s Yuki-Gulf hypothesis was correct. However, there are some problems. First, Atakapa as part of Gulf has been controversial, in part because it has also been tied in with Coahuiltecan. Indeed there are resemblances between the two, and they were not spoken next to each other so borrowing can be ruled out.

Perhaps a way of solving the matter is to posit not only Yuki-Gulf but a larger family that includes Coahuiltecan as Greenberg does in LIA. I have no idea how justified this is, but there are certainly surprising resemblances between Atakapa and the Coahuiltecan languages.

Furthermore, whether or not Chitimacha is part of Gulf has been up in the air from the beginning when Haas published her paper. Recent papers have made the case that Chitimacha is related to Mesoamerican language families of Mexico such as Mixe-Zoque and Totonacan. These papers used the comparative method. Campbell has rejected this hypothesis.

That Tunica at the very least shows a close relationship with Muskogean is not even controversial. The idea has a long pedigree and is presently supported by all experts in this family.

Geoffrey Kimball examined the data recently and concluded that from the evidence, it appears that Gulf exists, but we will never be able to prove it, as he puts it. However, he stated that Tunica is almost certainly related to Muskogean. At this point, I would think that Tunica-Muskogean at the very least should be considered consensus among specialists.

Kimball’s paper had a number of problems, mostly that he was operating with a negative stance towards the existence of the family. Further, there were issues with his notions of sound symbolism and borrowing in the paper where his explanations made no sense at all.

Let’s evaluate Campbell and Mixco’s statement that Gulf is no longer supported by most linguists.

We have four specialists on record about whether or not a Gulf family exists.

Mary Haas: Positive, minus Chitimacha

Joseph Greenberg: Positive

Pamela Munro: Positive

Geoffrey Kimball: Probably exists but it’s not possible to prove it.

Brown et al: Chitimacha is a part of the Totonozoquean family, not the Gulf family. The other members of Gulf are not members of this family.

Three out of the four specialists on the Gulf family say that the Gulf family is a reality. The other feels it exists but cannot be proven. And there is uncertainty about whether Chitimacha is probably not part of Gulf. The consensus among experts is that Gulf is a real language family.

Campbell and Mixco’s statement that Gulf is no longer supported by most linguists is simply false.

Furthermore, I would like to point out that a good case can be made for the existence of a Totonozoquean family consisting of the Mixe-Zoque and Totonacan languages. Whether this is consensus among experts is somewhat up in the air.

Campbell and Mixco:

Macro-Gê: A proposed distant genetic relationship composed of several language families and isolates, many now extinct, along the Atlantic coast (primarily of Brazil). These include Chiquitano, Bororoan, Botocudoan, Rikbaktsa, the Gê family proper, Jeikó, Kamakanan, Maxakalían, Purian, Fulnío, Ofayé and Guató. Many are sympathetic to the hypothesis and several of these languages will very probably be demonstrated to be related to one another eventually, though others will probably need to be separated out.

This is much too pessimistic. Macro-Gê is not a proposed long range family -it is a large language family in South America accepted by consensus. It is not true that many are sympathetic to it; instead, the consensus is that it is correct. Nor is it correct to say that it will probably be demonstrated eventually. In fact, it is already an accepted reality.

Campbell and Mixco:

Quechumaran: Proposed distant genetic relationship that would join Quechuan and Aymaran. While considerable evidence has been gathered in support of the hypothesis, it is extremely difficult in this case to distinguish what may be inherited (and therefore evidence of a genetic relationship) from what may be diffused (and therefore not reliable evidence of a genetic connection).

It is true that there is no consensus on the existence of Quechumaran. The consensus seems to be as above that it is not yet proven. Those opposed to the idea throw out the usual borrowing scenario, but they have had to push the large number of borrowings in core vocabulary all the way back to Proto-Aymara and Proto-Quechua. In my opinion, “massive borrowing of core vocabulary at the proto-language level” is simply another word for genetics.

Gerald Clauson, the famous Turkologist opponent of Altaic, had to keep pushing his massive borrowings of core vocabulary further and further back until he eventually had the scenario taking place at the Proto-Turkic, Proto-Tungusic, and Proto-Mongolic levels. See above for my analysis on why these three proto-languages could not possibly have borrowed from each other as they were in different places in different times.

A similar problem exists with opponents of the Uralo-Yukaghir theory, in which they are also forced to deal with a large amount of core vocabulary dating back a long time. Hakkinen tried to solve this problem by pushing the borrowing all the way back to not just Proto-Uralic but Pre-Proto-Uralic. Pre-Proto-Uralic at 8,000 years to me means nothing less than Uralo-Yukaghir. What else could it mean? He has heavy borrowing of core vocabulary between Pre-Proto-Uralic and Proto-Yukaghir. That’s another way of saying genetics.

Campbell and Mixco:

Macro-Guaicuruan (also spelled Macro-Waykuruan, Macro-Waikuruan): A proposed distant genetic relationship that would join the Guaicuruan and Matacoan families of the Gran Chaco in South America in a larger-scale genetic classification. Grammatical similarities, for example in the pronominal systems, have suggested the relationship to some scholars, but the extremely limited lexical evidence raises doubts for others. Some would also add Charruan and Mascoyan to these in an even larger ‘Macro-Waikuruan cluster.’

It is not true that this is a proposed long-range family suggested by some by doubted by others. In fact, Macro-Guaicuruan is accepted by consensus and is as uncontroversial as Macro-Gê, Pama-Nyungan, and other such families. There is however debate about which families are members outside of the Guaicuruan and Mataguayo language families that make up the essence of the family. There have been suggestions to add Lule-Vilela and the Zamucoan, Charruan, and Mascoyan families to this family. I do not feel that these additions are yet warranted.

Campbell and Mixco:

Pama-Nyungan: A very large, widely spread language family of Australia, some 175 languages. The name comes from Kenneth Hale, based on the words pama ‘man’ in the far northeast and nyunga ‘man’ in the southwest. Languages assigned to Pama-Nyungan extend over four-fifths of Australia, most of the continent except northern areas.

Pama-Nyungan is accepted by most Australianists as a legitimate language family, but not uncritically and not universally. It is rejected by Dixon; it is held by others to be plausible but inconclusive based on current evidence. Some Pama-Nyungan languages are Lardil, Kayardilt, Yukulta, Yidiny, Dyirbal, Pitta-Pitta, Arrente, Warlpiri, Western Desert language(s), and there are many more.

Actually, consensus now is that this family of Australian languages does indeed exist. True, Dixon challenged the existence of Pama-Nyungan recently, but his opposition was so outrageous and it prompted a quick surge of papers from Australianists defending the existence of Pama-Nyungan. The notion that other Australianists feel that Pama-Nyungan is possible but presently inconclusive is not correct. I am not aware of a single Australianist other than Dixon who feels this way. Instead, Pama-Nyungan is about as uncontroversial as Macro-Gê, Afroasiatic, or Austroasiatic.

Campbell and Mixco:

‘Papuan’ languages: A term of convenience used to refer to the languages of the western Pacific, most in New Guinea (Papua New Guinea and the Indonesian provinces of Papua and West Irian Jaya), that are neither Austronesian nor Australian. Papuan definitely does not refer to a genetic relationship among these languages for no such relationship can at present be shown.

That is, the term is defined negatively and does not imply a linguistic relationship. While most are spoken on the island of New Guinea, some are found in the Bismark Archipelago, Bougainville Island and the Solomon Islands to the east, and in Halmahera, Timor and the Alor Archipelago to the west.

There are some 800 Papuan languages divided in the a large number of mostly small language families and isolates not demonstrably related to one another.

For what it’s worth, this statement by Campbell and Mixco is correct.

Campbell and Mixco:

One large genetic grouping that has been posited for a number of Papuan languages is the Trans-New Guinea phylum, which is promising but not yet confirmed.

Trans-New Guinea is not “promising but not yet confirmed.” Instead it is an uncontroversial language family accepted by the consensus of all specialists.

References

Beck, David (1997). Mosan III: A Problem of Remote Common Proximity. International Conference on Salish (and Neighbo(u)ring) Languages.
Benedict, Paul K. (1942). “Thai, Kadai, and Indonesian: A New Alignment in Southeastern Asia.” American Anthropologist 44, 4: 576–601.
Benedict, Paul K. (1975). Austro-Thai Language and Culture, with a Glossary of Roots. New Haven: HRAF Press.
Blench, Roger (2008). The Prehistory of the Daic (Tai-Kadai) Speaking Peoples. Presented at the 12th EURASEAA Meeting in Leiden, the Netherlands, 1-5 September 2008.
Blench, Roger (2018). Tai-Kadai and Austronesian Are Related at Multiple Levels and Their Archaeological Interpretation (draft).
Blust, Robert (2014). “The Higher Phylogeny of Austronesian and the Position of Tai-Kadai: Another Look,” in The 14th International Symposium on Chinese Languages and Linguistics (IsCLL-14).
Campbell, Lyle and Marianne Mithun (Eds.) (1979). The Languages of Native America: An Historical and Comparative Assessment.
Campbell, Lyle and Mauricio J. Mixco (2007). A Glossary of Historical Linguistics. Edinburgh University Press.
Campbell, Lyle and William J. Poser (2008). Language Classification: History and Method. Cambridge: Cambridge University Press
Fortescue, M. (1998). Language Relations across Bering Strait: Reappraising the Archaeological and Linguistic Evidence. (Nivkhi is Mosan.)
Fortescue, Michael (2011). “The Relationship of Nivkh to Chukotko-Kamchatkan Revisited.” Lingua 121, 8: 1359-1376. (Nivkhi is Chukoto-Kamchatkan.)
Gell-Mann, Murray; Ilia Peiros, and George Starostin (2009). “Distant Language Relationship: The Current Perspective.” Journal of Language Relationship.
Greenberg, Joseph H. (2000). Indo-European and Its Closest Relatives: The Eurasiatic Language Family. Volume 1, Grammar. Stanford: Stanford University Press.
Greenberg, Joseph H. (2002). Indo-European and Its Closest Relatives: The Eurasiatic Language Family. Volume 2, Lexicon. Stanford: Stanford University Press.
Heine, Bernd (1992). African Languages. International Encyclopedia of Linguistics, ed. by William Bright, Vol. 1, pp. 31-36. Oxford: Oxford University Press. (No such thing as Nilo-Saharan.)
Krauss, Michael E. (1979). Na-Dene and Eskimo-Aleut. The Languages of Native America: Historical and comparative assessment, ed. by Lyle Campbell and Marianne Mithun, pp. 803-901. Austin: University of Texas Press. (Haida not part of Na-Dene.)
Levine, Robert D. (1979). Haida and Na-Dene: A New Look at the evidence. IJAL 45: 157-70. (Haida not part of Na-Dene.)
Li, Hui (李辉) (2005). Genetic Structure of Austro-Tai Populations (Doctoral Dissertation). Fudan University.
Mixco, Mauricio J. (1976). “Kiliwa Texts.” International Journal of American Linguistics Native American Text Series 1: 92-101
Mixco, Mauricio J. (1977). “The Linguistic Affiliation of the Ñakipa and Yakakwal of Lower California”. International Journal of American Linguistics 43: 189-200.
Nicola¨i, Robert (1990). Parent´es Linguistiques (`A Propos du Songhay). Paris: CNRS. (Dimmendaal says Songhay is Nilo-Saharan.)
Nikolaev, S. (2015). Toward the Reconstruction of Proto-Algonquian-Wakashan. Part 1: Proof of the Algonquian-Wakashan Relationship.
Nikolaev, S. (2016). Toward the Reconstruction of Proto-Algonquian-Wakashan. Part 2: Algonquian-Wakashan Sound Correspondences.
Ostapirat, Weera (2005). “Kra-Dai and Austronesian: Notes on Phonological Correspondences and Vocabulary Distribution,”  in Laurent Sagart, Roger Blench and Alicia Sanchez-Mazas, eds. The Peopling of East Asia: Putting Together Archaeology, Linguistics, and Genetics, pp. 107-131. London: Routledge Curzon.
Ostapirat, Weera (2013). Austro-Tai Revisited. Paper Presented at the 23rd Annual Meeting of the Southeast Asian Linguistics Society, 29-31 May 2013, Chulalongkorn University.
Reid, Lawrence A. (2006). “Austro-Tai Hypotheses.” In Keith Brown (Ed.), The Encyclopedia of Language and Linguistics, 2nd Edition, pp. 609–610.
Sagart, Laurent (2005b). “Tai-Kadai as a Subgroup of Austronesian,” in L. Sagart, R. Blench, and A. Sanchez-Mazas (Eds.), The Peopling of East Asia: Putting Together Archaeology, Linguistics, and Genetics, pp. 177-181.
Sagart, Laurent (2019). “A Model of the Origin of Kra-Dai Tones.” Cahiers de Linguistique Asie Orientale. 48, 1: 1–29.
Thurgood, Graham (1994). “Tai-Kadai and Austronesian: The Nature of the Relationship.” Oceanic Linguistics 33: 345-368.

A Look at the Quechua and Aymara Languages

From here. A look at two major South American Indian languages, Quechua and Aymara with a view towards how difficult they are to learn for an English speaker. Aymara, which is actually at least two separate languages, is declining in Bolivia, Chile and Argentina and is only doing well around Lake Titicaca in Peru. Quechua is actually made up of up to 46 different languages spoken in Peru, Ecuador, Colombia, Argentina, Chile and Bolivia. Quechua is probably one of the easiest Amerindian languages to learn, but that’s not saying much. Aymara is probably a lot more difficult. Quechua (actually a large group of languages and not a single language at all) is one of the easiest Amerindian languages to learn. Quechua is a classic example of a highly regular grammar with few exceptions. Its agglutinative system is more straightforward than even that of Turkish. The phonology is dead simple. On the down side, there is a lot of dialectal divergence (these are actually separate languages and not dialects) and a lack of learning materials. Some say that Quechua speakers spend their whole lives learning the language. Quechua has inconsistent orthographies. There is a fight between those who prefer a Spanish-based orthography and those who prefer a more phonemic one. Also there is an argument over whether to use the Ayacucho language or the Cuzco language as a base. Quechua has a difficult feature known as evidential marking. This marker indicates the source of the speaker’s knowledge and how sure they are about the statement. -mi expresses personal knowledge: Tayta Wayllaqawaqa chufirmi. Mr. Huayllacahua is a driver. (I know it for a fact.) -si expresses hearsay knowledge: Tayta Wayllaqawaqa chufirsi. Mr. Huayllacahua is a driver (or so I’ve heard). chá expresses strong possibility: Tayta Wayllaqawaqa chufirchá. Mr. Huayllacahua is a driver (most likely). Quechua is rated 4, very difficult. Aymara has some of the wildest morphophonology out there. Morpheme-final vowel deletion is present in the language as a morphophonological process, and it is dependent on a set of highly complex phonological, morphological and syntactic rules (Kim 2013). For instance, there are three types of suffixes: dominant, recessive and a 3rd class is neither dominant nor recessive. If a stem ends in a vowel, dominant suffixes delete the vowel but recessive suffixes allow the vowel to remain. The third class either deletes or retains the vowel on the stem depending on how many vowels are in the stem. If the root has two vowels, the vowel is retained. If it has three vowels, the vowel is deleted. Aymara gets a 5 rating, hardest of all.

More On The Hardest Languages To Learn – Non-Indo-European Languages

Caution: This post is very long. It runs to 200 pages on the Net. Updated January 17, 2016.

This is a continuation of the earlier post. I split it up into two parts because it had gotten too long.

The post refers to which languages are the hardest for English speakers to learn, though to some extent, the ratings are applicable across languages. Most Chinese speakers would recognize Spanish as being an easy language, despite its alien nature. And even most Chinese, Navajo, Poles or Czechs acknowledge that their languages are hard to learn. To a certain extent, difficulty is independent of linguistic starting point. Some languages are just harder than others, and that’s all there is to it.

Method, Results and Conclusion. See here.

In this case, 73 non-IE languages were examined.

Ratings: Languages are rated 1-6, easiest to hardest. 1 = easiest, 2 = moderately easy to average, 3 = average to moderately difficult, 4 = very  difficult, 5 = extremely difficult, 6 = most difficult of all.

Time needed: Time needed to learn the language “reasonably well”: Level 1 languages = 3 months-1 year. Level 2 languages = 6 months-1 year. Level 3 languages = 1-2 years. Level 4 languages = 2 years. Level 5 languages = 3-4 years, but some may take longer.

Here is a list of the ratings for the languages below as a handy reference.

 

Malagasy 1.0 Bahasa Indonesian 1.5 Aymara 2.0 Malay 2.0 Hawaiian 2.0 Swahili 2.0 Maori 3.0 Turkish 3.5 Quechua 4.0 Maltese 4.0 Tamil 4.0 Tagalog 4.0 Anyi 4.0 Egyptian Arabic 4.5 Moroccan Arabic 4.5 Amharic 4.5 Estonian 4.5 Khmer 4.5 Lao 4.5 Georgian 5.0 Gros Ventre 5.0 Karok 5.0 MSA Arabic 5.0 Hebrew 5.0 Somali 5.0 Malayalam 5.0 Korean 5.0 Japanese 5.0 Finnish 5.0 Skolt Sami 5.0 Hungarian 5.0 Quiang 5.0 Tibetan 5.0 Dzongka 5.0 Vietnamese 5.0 Sedang 5.0 Hmong 5.0 Tsou 5.0 Sakai 5.0 Kwaio 5.0 Thai 5.0 Kam 5.0 Buyang 5.0 Ga 5.0 Ndali 5.0 Xhosa 5.0 Ndebele 5.0 Zulu 5.0 Taa 5.0 Ju|’hoan 5.0 Cherokee 5.5 Lakota 5.5 Classical Japanese 5.5 Mandarin 5.5 Cantonese 5.5 Min Nan 5.5 Dondan Wu 5.5 Basque 5.5 Chechen 6.0 Circassian 6.0 Tsez 6.0 Archi 6.0 Tabasaran 6.0 Ingush 6.0 Ubykh 6.0 Abkhaz 6.0 Burushaski 6.0 Kootenai 6.0 Yuchi 6.0 Tlingit 6.0 Navajo 6.0 Slavey 6.0 Haida 6.0 Salish 6.0 Nuxalk 6.0 Montana Salish 6.0 Straits Salish 6.0 Halkomelem 6.0 Lushootseed 6.0 Cree 6.0 Ojibwa 6.0 Cheyenne 6.0 Arapaho 6.0 Wichita 6.0 Huamelutec 6.0 Hopi 6.0 Nahuatl 6.0 Comanche 6.0 Chinantec 6.0 Jalapa Mazatec 6.0 Tarina 6.0 Bora 6.0 Tuyuca 6.0 Cubeo 6.0 Hixkaryána 6.0 Nambikwara 6.0 Pirahã 6.0 Australian Languages – 6.0 Berik 6.0 Amele 6.0 Valpan 6.0 Tamazight 6.0 Tachelhit 6.0 Dahalo 6.0 Classical Chinese 6.0 Inuktitut 6.0 Kalaallisut 6.0 Chukchi 6.0

Northeast Caucasian, Northwest Caucasian and Kartvelian

Of course the Caucasian languages like Tsez, Tabasaran, Georgian, Chechen, Ingush, Abkhaz and Circassian are some of the hardest languages on Earth to learn.

Chechen and Circassian are rated 6, hardest of all.

Northeast Caucasian

NE Caucasian languages have the uvulars and ejectives of Georgian in addition to pharyngeals, lateral fricatives, and other strangeness. They have noun classes like the Bantu languages (but usually fewer). Nevertheless, they have noun class agreement markers on verbs on adjectives. One thing NE Caucasian has is lots of case. Some languages have 40+ cases. They are built from the ground up via two forms – one a spatial form such as in, on or around and the other a directional motion form such as to, from, through or at.

Tsezic

Tsez has 64-126 different cases, making it by far the most complex case system on Earth! It is one of the few languages on Earth that has two genitive cases – Genitive 1 (-s) and Genitive 2 (-z). Genitive 1 is used when the genitive’s head noun is in absolutive case and Genitive 2 is used when the genitive’s head noun is in any other case. It also has four noun classes. It is said that even native speakers have a hard time picking up the correct inflection to use sometimes.

In Tsez, you need to know a lot Tsez grammar to communicate at a basic level. The sentence:

English: I like your mother.

Tsez: Дāьр деби энийу йетих. (Dǟr debi eniyu yetix.)

In order to speak that sentence in Tsez, you need to know:

• the words themselves (word order is not as important) • that the verb -eti- requires the subject to be in the dative/lative case and the object to be in the absolutive • the noun class for eniyu (class II) • the dative/lative form of di (I), which is dǟr • the genitive 1 form of mi (you), which is debi • the congruence prefix y- that corresponds to the noun class of the absolutive argument of the phrase, in this case mother • the present tense ending for vowel-final verbs -x

Tsez is rated 6, hardest of all.

Lezgic Archi

Archi has an extremely complex phonology and one of the most complicated grammars on Earth. The extreme fusional aspects and the verbal morphology are what make the grammar so difficult. Every verb root has 1,502,839 possible forms! It is also an ergative language, but there is irregularity in its ergative system.

Some verbs take the typical ergative/absolutive case (absolutive for the subject of an intransitive very and ergative for the subject of a transitive verb – where the direct object would be in absolutive). In others the subject is in dative rather than the expected ergative/absolutive case. These are usually verbs of perception like love/want, hear, see, feel, and be bored. For instance, the verb:

-эти- = to love/want must have its subject in dative case instead of the expected absolutive or ergative case.

Among non-click languages, Archi has one of the largest consonant inventories, with only the extinct Ubykh having more. There are 26 vowels and between 76 and 82 consonants, depending on the analysis. Five of the six vowels can occur in five varieties: short, pharyngealized, high tone, long (with high tone), and pharyngealized with high tone.

It has many unusual phonemes, including contrasts between several voiceless velar lateral fricatives, voiceless and ejective velar lateral affricates and a voiced velar lateral fricative. The voiceless velar lateral fricative ʟ̝̊, the voiced velar lateral fricative ʟ̝, and the corresponding voiceless and ejective affricates k͡ʟ̝̊ and k͡ʟ̝̊ʼ are extremely unusual sounds, as velar fricatives are not typically laterals.

There are 15 cases, 10 regular cases, five spatial cases and five directional cases. The Spatial cases are Inessive (in), Intrative (between), superessive (above), Subessive (below) and Pertingent (against). The directional cases are Essive (as), Elative (out of), Lative (to/into), Allative (onto), Terminative (specifies a limit) and Translative (indicates change).

There are four noun classes:

I Male human II Female human III All insects, some animates, and some inanimates IV Abstracts, some animates, and some inanimates that can only be seen via verbal agreement

Archi is rated 6, hardest of all.

Samur Eastern Samur Lezgi–Aghul–Tabasaran

Tabasaran is rated the 3rd most complex grammar in the world, with 48 different noun cases.

Tabasaran is rated 6, hardest of all.

Nakh Vainakh

Ingush has a very difficult phonology, an extremely complex grammar, and furthermore, is extremely irregular. Ingush also has a proximate/obviate distinction and is the only language in the region that has this feature. Ingush along with Chechen both have a closed class of verbs, an unusual feature in the world’s languages. New verbs are formed by adding a noun to the verb do:

shootdo gun

Ingush is rated 6, hardest of all.

Kartvelian Karto-Zan

One problem with Georgian is the strange alphabet: ქართულია ერთ ერთი რთული ენა. It also has lots of glottal stops that are hard for many foreigners to speak; consonant clusters can be huge – up to eight consonants stuck together (CCCCCCCCVC)- and many consonant sounds are strange. In addition, there are uvulars and ejectives. Georgian is one of the hardest languages on Earth to pronounce. It regularly makes it onto craziest phonologies lists.

Its grammar is exceedingly complex. Georgian is both highly agglutinative and highly irregular, which is the worst of two worlds. Other agglutinative languages such as Turkish and Finnish at least have the benefit of being highly regular. The verbs in particular seem nearly random with no pattern to them at all. The system of argument and tense marking on the verb is exceedingly complex, with tense, aspect, mood on the verb, person and number marking for the subject, and direct and indirect objects.

Although it is an ergative language, the ergative (or active-stative case marking as it is called) oddly enough is only used in the aorist and perfect tenses where the agent in the sentence receives a different case, while the aorist also masquerades as imperative. In the present, there is standard nominative-accusative marking. A single verb can have up to 12 different parts, similar to Polish, and there are six cases and six tenses.

Georgian also features something called polypersonal agreement, a highly complex type of morphological feature that is often associated with polysynthetic languages and to a lesser extent with ergativity.

In a polypersonal language, the verb has agreement morphemes attached to it dealing with one or more of the verbs arguments (usually up to four arguments). In a non polypersonal language like English, the verb either shows no agreement or agrees with only one of its arguments, usually the subject. Whereas in a polypersonal language, the verb agrees with one or more of the subject, the direct object, the indirect object, the beneficiary of the verb, etc. The polypersonal marking may be obligatory or optional.

In Georgian, the polypersonal morphemes appear as either suffixes or prefixes, depending on the verb class and the person, number, aspect and tense of the verb. The affixes also modify each other phonologically when they are next to each other. In the Georgian system, the polypersonal affixes convey subject, direct object, indirect object, genitive, locative and causative meanings.

g-mal-av-en = they hide you g-i-mal-av-en = they hide it from you

mal (to hide) is the verb, and the other four forms are polypersonal affixes.

In the case below,

xelebi ga-m-i-tsiv-d-a = My hands got cold.

xelebi means hands. The m marker indicates genitive or my. With intransitive verbs, Georgian often omits my before the subject and instead puts the genitive onto the verb to indicate possession.

Georgian verbs of motion focus on deixis, whether the goal of the motion is towards the speaker or the hearer. You use a particle to signify who the motion is heading towards. If it heading towards neither of you, you use no deixis marker. You specify the path taken to reach the goal through the use or prefixes called preverbs, similar to “verbal case.” These come after the deixis marker:

up             a-
out            ga-
in             sha-
down into      cha-
across/through garda-
thither        mi-
away           c’a-
or down        da-

Hence:

up towards me = amo-. The deixis marker is mo- and up is a-

On the plus side, Georgian has borrowed a great deal of Latinate foreign vocabulary, so that will help anyone coming from a Latinate or Latinate-heavy language background.

Georgian is rated 5, extremely difficult.

Northwest Caucasian

All NW Caucasian languages are characterized by a very small number of vowels (usually only two or three) combined with a vast consonant inventory, the largest consonant inventories on Earth. Almost any consonant can be plain, labialized or palatalized. This is apparently the result of an historical process whereby many vowels were lost and their various features became assigned to consonants. For instance, palatalized consonants may have come from Ci sequences and labialized consonants may have come from Cu sequences.

The grammars of these languages are complex. Unlike the NE Caucasian languages, they have simple noun systems, usually with only a handful of cases.

However, they have some of the complex verbal systems on Earth. These are some of the most synthetic languages in the Old World. Often the entire syntax of the sentence is contained within the verb. All verbs are marked with ergative, absolutive and direct object morphemes in addition to various applicative affixes.

These are akin to what some might call “verbal case.” For instance, in applicative voice systems, applicatives may take forms such as comitative, locative, instrumental, benefactive and malefactive. These roles are similar to the case system in nouns – even the names are the same. So you can see why some call this “verbal case.”

NW Caucasian verbs can be marked for aspect (whether something is momentous, continuous or habitual), mood (if something is certain, likely, desired, potential, or unreal). Other affixes can shape the verb in an adverbial sense, to express pity, excess or emphasis.

Like NE Caucasian, they are also ergative.

NW Caucasian makes it onto a lot of craziest language lists.

These are some of the strangest sounding languages on Earth. Of all of these languages, Abaza has the most consonants. Here is a video in the Abaza language.

Ubykh

Ubykh, a Caucasian language of Turkey, is now extinct, but there is one second language speaker, a linguist who is said to have taught himself the language. It has more consonants than any non-click language on Earth – 84 consonant sounds in all. Furthermore, the phonemic inventory allows some very strange consonant clusters.

Ubykh has many rare consonant sounds. is only also found in two of Ubykh’s relatives, Abkhaz and Abaza and in two other languages, both in the Brazilian Amazon. The pharyngealized labiodental voiced fricative  does not exist in any other language. It often makes it onto weirdest phonologies lists. Ubykh also got a very high score on a study of the weirdest languages on Earth.

Combine that with only two vowel sounds and a highly complex grammar, and you have one tough language.

In addition, Ubykh is both agglutinative and polysynthetic, ergative, and has polypersonal agreement:

Aχʲazbatʂʾaʁawdətʷaajlafaqʾajtʾmadaχ! If only you had not been able to make him take it all out from under me again for them…

There are an incredible 16 morphemes in that nine syllable word.

Ubykh has only four case systems on its nouns, but much case function has shifted over to the verb via preverbs and determinants. It is these preverbs and determinants that make Ubykh monstrously complex. The following are some of the directional preverbs:

  • above and touching
  • above and not touching
  • below and touching
  • below and not touching
  • at the side of
  • through a space
  • through solid matter
  • on a flat horizontal surface
  • on a non-horizontal or vertical surface
  • in a homogeneous mass
  • towards
  • in an upward direction
  • in a downward direction
  • into a tubular space
  • into an enclosed space

There are also some preverbal forms that indicate deixis:

j-  = towards the speaker

Others can indicate ideas that would take up whole phrases in English:

jtɕʷʼaa- = on the Earth, in the Earth

ʁadja ajtɕʷʼaanaaɬqʼa They buried his body. (Lit. They put his body in the earth.)

faa– = out of, into or with regard to a fire.

Amdʒan zatʃətʃaqʲa faastχʷən. I take a brand out of the fire.

Morphemes may be as small as a single phoneme:

wantʷaan They give you to him.

w – 2nd singular absolutive a – 3rd singular dative n – 3rd ergative – to give aa – ergative plural n – present tense

Adverbial suffixes are attached to words to form meanings that are often formed by aspects or tenses in other languages:

asfəpχaI need to drink it. asfəfanI can drink it. asfəɡʲanI drink it all the time. asfəlanI am drinking it all up. asfətɕʷan I drink it too much. asfaajənI drink it again.

Nouns and verbs can transform into each other. Any noun can turn into a stative verb:

məzəchild

səməzəjtʼ I was a child. (Lit. I child-waschild-was is a verb – to be a child.)

By the same token, many verbs can become nouns via the use of a nominal affix:

qʼato say

səqʼa what I say – (Lit. That which I saymy speech, my words, my language, my orders, etc.

Number is marked on the verb via a verbal suffix and is only marked on the noun in the ergative case.

However, it does lack the convoluted case systems of the Caucasian languages next door and there is no grammatical gender.

Ubykh is rated 6, hardest of all.

Abkhaz-Abazin

Abkhaz is an extremely difficult language to learn. Each basic consonant has eight different positions of articulation in the mouth. Imagine how difficult that would be for an Abkhaz child with a speech impediment. Abkhaz seems to put agreement markers on just about everything in the language. Abkhaz makes it onto many craziest language lists, and it recently got a very high score on a weirdest language study.

Abkhaz is rated 6, hardest of all.

Burushaski

Burushaski is often thought to be a language isolate, related to no other languages, however, I think it is Dene-Caucasian. It is spoken in the Himalaya Mountains of far northern Pakistan in an area called the Hunza. It’s verb conjugation is complex, it has a lot of inflections, there are complicated ways of making sentences depending on many factors, and it is an ergative language, which is hard to learn for speakers of non-ergative languages. In addition, there are very few to no cognates for the vocabulary.

Burushaski is rated 6, hardest of all.

American Indian Languages

American Indian languages are also notoriously difficult, though few try to learn them in the US anyway. In the rest of the continent, they are still learned by millions in many different nations. You almost really need to learn these as a kid. It’s going to be quite hard for an adult to get full competence in them.

One problem with these languages is the multiplicity of verb forms. For instance, the standard paradigm for the overwhelming number of regular English verbs is a maximum of five forms:

steal steals stealing stole stolen

Many Amerindian languages have over 1,000 forms of each verb in the language.

Kootenai

Yet the Salishans (see below) always considered the neighboring language Kootenai to be too hard to learn. Kootenai also has a distinction between proximate/obviate along with direct/inverse alignment, probably from contact with Algonquian.

However, the Kootenai direct/inverse system is less complex than Algonquian’s, as it is present only in the 3rd person. Kootenai also has a very strange feature in that they have particles that look like subject pronouns, but these go outside of the full noun phrase. This is a very rare feature in the world’s languages. Kootenai scored very high on a weirdest language survey.

Kootenai is an isolate spoken in Idaho by 100 people.

Kootenai is rated 6, hardest of all.

Yuchi

Yuchi is a language isolate spoken in the Southern US. They were originally located in Eastern Tennessee and were part of the Creek Confederacy at one time. Yuchi is nearly extinct, with only five remaining speakers.

Yuchi has noun genders or classes based on three distinctions of position: standing, sitting or lying. All nouns are either standing, sitting or lying. Trees are standing, and rivers are lying, for instance. It it is taller than it is wide, it is standing. It if is  wider than it is tall, it is lying.

If it is about as about as wide as it is tall, it is sitting. All nouns are one of these three genders, but you can change the gender for humorous or poetic effect. A linguist once asked a group of female speakers whether a penis was standing, sitting or lying. After lots of giggles, they said the default was sitting, but you could say it was standing or lying for poetic effect.

Also all Yuchi pronouns must make a distinction between age (older or younger than the speaker) and ethnicity (Yuchi or non-Yuchi).

Yuchi gets a 6 rating, hardest of all.

Dene-Yeniseian Na-Dene Athabascan-Eyak Tlingit

Tlingit is probably one of the hardest, if not the hardest, language in the world. Tlingit is analyzed as partly synthetic, partly agglutinative, and sometimes polysynthetic. It has not only suffixes and prefixes, but it also has infixes, or affixes in the middle of words.

‘eechto pick

All prefixes must be in proper order for the word to work.

tuyakaoonagadagaxayaeecheen. I am usually picking, on purpose, a long object through the hole while standing on a table.

tuyakaoonagootxayaeecheen. I am usually being forced to pick a long object through the hole while standing on a table.

tuyaoonagootxawa’eecheen. I am usually being picking the edible long object through the hole while standing on a table.

Tlingit has a pretty unusual phonology. For one thing, it is the only language on Earth with no l. This despite the fact that it has five other laterals: dl (), tl (tɬʰ), tl’ (tɬʼ), l (ɬ) and l’ (ɬʼ). The tɬʼ and ɬʼ sounds are rare in the world’s languages. ɬʼ  is only found in the wild NW Caucasian languages. It also has two labialized glottal consonants, ʔʷ and hw ().

Tlingit gets a 6 rating, hardest of all.

Athabascan Southern

Navajo has long, short and nasal vowels, a tone system and a grammar totally unlike anything in Indo-European. A stem of only four letters or so can take enough affixes to fill a whole line of text.

Navajo is a polysynthetic language. In polysynthetic languages, very long words can denote an entire sentence, and it’s quite hard to take the word apart into its parts and figure out exactly what they mean and how they go together. The long words are created because polysynthetic languages have an amazing amount of morphological richness. They put many morpheme together to create a word out of what might be a sentence in a non-polysynthetic language.

Some Navajo dictionaries have thousands of entries of verbs only, with no nouns. Many adjectives have no direct translation into Navajo. Instead, verbs are used as adjectives. A verb has no particular form like in English – to walk. Instead, it assumes various forms depending on whether or not the action is completed, incomplete, in progress, repeated, habitual, one time only, instantaneous, or simply desired. These are called aspects. Navajo must have one of the most complex aspect systems of any language:

The Primary aspects:

Momentaneous – punctually (takes place at one point in time) Continuative – an indefinite span of time & movement with a specified direction Durative – over an indefinite span of time, non-locomotive uninterrupted continuum Repetitive – a continuum of repeated acts or connected series of acts Conclusive – like durative but in perfective terminates with static sequel Semelfactive – a single act in a repetitive series of acts Distributive – a distributive manipulation of objects or performance of actions Diversative – a movement distributed among things (similar to distributive) Reversative – results in directional change Conative – an attempted action Transitional – a shift from one state to another Cursive – progression in a line through time/space (only progressive mode)

The subaspects:

Completive – an event/action simply takes place (similar to the aorist tense) Terminative – a stopping of an action Stative – sequentially durative and static Inceptive – beginning of an action Terminal – an inherently terminal action Prolongative – an arrested beginning or ending of an action Seriative – an interconnected series of successive separate & distinct acts Inchoative – a focus on the beginning of a non-locomotion action Reversionary – a return to a previous state/location Semeliterative – a single repetition of an event/action

The tense system is almost as wild as the aspectual system.

For instance, the verb ndideesh means to pick up or to lift up. But it varies depending on what you are picking up:

ndideeshtiilto pick up a slender stiff object (key, pole) ndideeshleel to pick up a slender flexible object (branch, rope) ndideesh’aalto pick up a roundish or bulky object (bottle, rock) ndideeshgheelto pick up a compact and heavy object (bundle, pack) ndideeshjolto pick up a non-compact or diffuse object (wool, hay) ndideeshteelto pick up something animate (child, dog) ndideeshnil to pick up a few small objects (a couple of berries, nuts) ndideeshjihto pick up a large number of small objects (a pile of berries, nuts) ndideeshtsosto pick up something flexible and flat (blanket, piece of paper) ndideeshjilto pick up something I carry on my back ndideeshkaalto pick up anything in a vessel ndideeshtlohto pick up mushy matter (mud).

But picking up is only one way of handling the 12 different consistencies. One can also bring, take, hang up, keep, carry around, turn over, etc. objects. There are about 28 different verbs one can use for handling objects. If we multiply these verbs by the consistencies, there are over 300 different verbs used just for handling objects.

In Navajo textbooks, there are conjugation tables for inflecting words, but it’s pretty hard to find a pattern there. One of the most frustrating things about Navajo is that every little morpheme you add to a word seems to change everything else around it, even in both directions.

Navajo is said to have a very difficult system for counting numerals.

There is also a noun classifier system with more than a dozen classifiers that affect inflection. This is quite a few classifiers even for a noun classifier language and is similar to African languages like Zulu. In addition, it has the strange direct/inverse system.

To add insult to injury, Navajo is an ergative language.

Navajo also has an honorifics or politeness system similar to Japanese or Korean.

Navajo also has the odd feature where the word niinaabecause can be analyzed as a verb.

X áhóót’įįd biniinaa… Because X happened…

Shiniinaa sits’il. It broke into pieces because of me.

In the latter sentence, the only way we know that 1st singular was involved in because of the person marking on niinaa.

There are 25 different kinds of pronominal prefixes that can be piled onto one another before a verb base.

Navajo has a very strange feature called animacy, where nouns take certain verbs according to their rank in the hierarchy of animation which is a sort of a ranking based on how alive something is. Humans and lightning are at the top, children and large animals are next and abstractions are at the bottom.

All in all, Navajo, even compared to other polysynthetic languages, has some of the most incredibly complicated polysynthetic morphology of any language. On craziest grammar and craziest language lists, Navajo is typically listed.

It is even said that Navajo children have a hard time learning Navajo as compared to children learning other languages, but Navajo kids definitely learn the language. Similarly with Hopi below, even linguists find even the best Navajo grammars difficult or even impossible to understand.

However, Navajo is quite regular, a common feature in Amerindian languages.

Navajo is rated 6, hardest of all.

Northern

Slavey, a Na-Dene language of Canada, is hard to learn. It is similar to Navajo and Apache. Verbs take up to 15 different prefixes. All Athabascan languages have wild verbal systems. It also uses a completely different alphabet, a syllabic one designed for Canadian Indians.

Slavey is rated 6, hardest of all.

Haida

Haida is often thought to be a Na-Dene language, but proof of its status is lacking. If it is Na-Dene, it is the most distant member of the family. Haida is in the competition for the most complicated language on Earth, with 70 different suffixes.

Haida is rated 6, hardest of all.

Salishan

The Salishan languages spoken in the Northwest have a long reputation for being hard to learn, in part because of long strings of consonants, in one case 11 consonants long. Salish languages are the only languages on Earth that allow words without sonorants.

Many of the vowels and consonants are not present in most of the world’s widely spoken languages. The Salish languages are, like Chukchi, polysynthetic. Some translations treat all Salish words are either verbs or phrases. Some say that Salish languages do not contain nouns, though this is controversial. The verbal system of Salish languages is absurdly complex.

All Salishan languages are rated rated 6, hardest of all.

Nuxálk (Bella Coola)

Nuxálk is a notoriously difficult Salishan Amerindian language spoken in British Colombia. It is famous for having some really wild words and even sentences that don’t seem to have any vowels in them at all. For instance:

xłp̓x̣ʷłtłpłłskʷc̓  (xɬpʼχʷɬtʰɬpʰɬːskʷʰt͡sʼ in IPA) He had a bunchberry plant.

sxs seal fat

Here are some more odd words and sentences:

smnmnmuuc mute

Nuyamłamkis timantx tisyuttx ʔułtimnastx. The father sang the song to his son.

Musis tiʔimmllkītx taq̓lsxʷt̓aχ. The boy felt that rope.

However, this word is not typically used by speakers and by no means do most words consist of all consonants. The language sounds odd when spoken. It has been described as “whispering while chewing on a granola bar” (see the video sample under Montana Salish below).

These wild consonant clusters are even crazier than the ones in Ubykh and NW Caucasian. In fact, the nutty consonant clusters in Salish and causing a debate in linguistics about whether or not the syllable is even a universal phenomenon in language as some Salish words and phrases appear to lack syllables. Some Berber dialects have raised similar questions about the syllable.

Nuxálk makes it onto lists of the craziest phonologies on Earth.

Nuxálk is rated 6, hardest of all.

Interior Salish Southern

Montana Salish is said to be just as hard to learn as Nuxálk . Spokane (Montana Salish) has combining and independent forms with the same meaning:

spim’cnmouth -cinmouth

Montana Salish makes it onto a lot of craziest grammars lists.

This link shows an elder on the Flathead Indian Reservation in Montana, Steven Smallsalmon, speaking Montana Salish. He also leads classes in the language. This is probably one of the strangest sounding languages on Earth.

Montana Salish is rated 6, hardest of all.

Central

Straits Salish has an aspectual distinction between persistent and nonpersistent. Persistent means the activity continues after its inception as a state. The persistent morpheme is . The result is similar to English:

figure out – nonpersistent know – persistent

look at – nonpersistent watch – persistent

take – nonpersistent hold – persistent

is referred to as a “parasitic morpheme” and only occurs in stem that has an underlying ə which serves as a “host” for the morpheme.

How strange.

The Saanich dialect of Straits Salish is often listed in the rogue’s gallery of craziest grammars on Earth. The writing system is often listed as one of the worst out there. In addition, Saanich makes it onto craziest grammars lists for the parasitic morphemes and for having no distinction between nouns and verbs!

Straits Salish gets a 6 rating, hardest of all.

Halkomelem, spoken by 570 people around Vancouver, British Colombia, is widely considered to be one of the hardest languages on Earth to learn. In Halkomelem, many verbs have an orientation towards water. You can’t just say, She went home. You have say how she was going home in relation to nearby bodies of water. So depending on where she was walking home in relation to the nearest river, you would say:

She was farther away from the water and going home. She was coming home in the direction away from the water. She was walking parallel to the flow of the water downstream. She was walking parallel to the flow of the water upstream.

Halkomelem gets a 6 rating, hardest of all.

Lushootseed

Lushootseed is said to be just as hard to learn as Nuxálk. Lushootseed is one of the few languages on Earth that has no nasals at all, except in special registers like baby talk and the archaic speech of mythological figures. It also has laryngealized glides and nasals: w ̰ , m̥ ̰ , and n̥ ̰ .

Lushootseed is rated 6, hardest of all.

Iroquoian

All Iroquoian languages are extremely difficult, but Athabaskan is probably even harder. Siouan languages may be equal to Iroquoian in difficulty.

Compare the same phrases in Tlingit (Athabaskan) and and  Cherokee (Iroquoian).

Tlingit:

kutíkusa‘áatIt’s cold outside. kutíkuta‘áatIt’s cold right now.

In Tlingit, you can add or modify affixes at the beginning as prefixes, in the middle as infixes and at the end as suffixes. In the above example, you changed a part of the word within the clause itself.

Cherokee:

doyáditlv uyvtlvIt is cold outside. (Lit. Outside it is cold) ka uyvtlv It is cold now. (Lit. Now it is cold.)

As you can see, Cherokee is easier.

Cherokee

Cherokee is very hard to learn. In addition to everything else, it has a completely different alphabet. It’s polysynthetic, to make matters worse. It is possible to write a Cherokee sentence that somehow lacks a verb. There are five categories of verb classifiers. Verbs needing classifiers must use one. Each regular verb can have an incredible 21,262 inflected forms! All verbs contain a verb root, a pronominal prefix, a modal suffix and an aspect suffix. In addition, verbs inflect for singular, plural and also dual. For instance:

ᎠᎸᎢᎭ   a'lv'íha 

You have 126 different forms:
ᎬᏯᎸᎢᎭ  gvyalv'iha     I tie you up
ᏕᎬᏯᎸᎢᎭ degvyalviha  I'm tying you up
ᏥᏯᎸᎢᎭ  jiyalv'ha        I tie him up
ᎦᎸᎢᎭ                          I tie it
ᏍᏓᏯᎸᎢᎭ sdayalv'iha  I tie you (dual)
ᎢᏨᏯᎢᎭ  ijvyalv'iha    I tie you (pl)
ᎦᏥᏯᎸᎢᎭ gajiyalv'iha  I tie them (animate)
ᏕᎦᎸᎢᎭ                        I tie them up (inanimate)
ᏍᏆᎸᎢᎭ  squahlv'iha    You tie me
ᎯᏯᎸᎢᎭ  hiyalv'iha     You're tying him
ᎭᏢᎢᎭ   hatlv'iha         You tie it
ᏍᎩᎾᎸᎢᎭ skinalv'iha    You're tying me and him
ᎪᎩᎾᏢᎢᎭ goginatlv'iha  They tie me and him etc.

Let us look at another form:

to see

I see myself           gadagotia
I see you                gvgohtia
I see him/               tsigotia
I see it                    tsigotia
I see you two          advgotia
I see you (plural)    istvgotia
I see them (live)    gatsigotia
I see them (things) detsigotia

You see me                     sgigotia
You see yourself              hadagotia
You see him/her              higo(h)tia
You see it                        higotia
You see another and me  sginigotia
You see others and me    isgigotia
You see them (living)      dehigotia
You see them (living)      gahigotia
You see them (things)     detsigotia

He/she sees me                    agigotia
He/she sees you                   tsagotia
He/she sees you                   atsigotia
He/she sees him/her            agotia
He/she sees himself/herself  adagotia
He/she sees you + me          ginigotia
He/she sees you two             sdigotia
He/she sees another + me    oginigotia
He she sees us (them + me) otsigotia
He/she sees you (plural)       itsigotia
He/she sees them                 dagotia

You and I see him/her/it                igigotia
You and I see ourselves                 edadotia
You and I see one another             denadagotia/dosdadagotia
You and I see them (living)           genigotia
You and I see them (living or not) denigotia

You two see me                           sgninigotia
You two see him/her/it                 esdigotia
You two see yourselves                sdadagotia
You two see us (another and me) sginigotia
You two see them                        desdigotia

Another and I see you             sdvgotia
Another and I see him/her       osdigotia
Another and I see it                 osdigotia
Another and I see you-two      sdvgotia
Another and I see ourselves    dosdadagotia
Another and I see you (plural) itsvgotia
Another and I see them           dosdigotia

You (plural) see me        isgigoti
You (plural) see him/her etsigoti

They see me                    gvgigotia
They see you                   getsagotia
They see him/her             anigoti
They see you and me       geginigoti
They see you two             gesdigoti
They see another and me gegigotia/gogenigoti
They see you (plural)       getsigoti
They see them                 danagotia
They see themselves       anadagoti

I will see datsigoi
I saw      agigohvi

He/she will see dvgohi
He/she             sawugohvi

Number is marked for inclusive vs. exclusive and there is a dual. 3rd person plural is marked for animate/inanimate. Verbs take different object forms depending on if the object is solid/alive/indefinite shape/flexible. This is similar to the Navajo system.

Cherokee also has lexical tone, with complex rules about how tones may combine with each other. Tone is not marked in the orthography. The phonology is noted for somehow not having any labial consonants.

However, Cherokee is very regular. It has only three irregular verbs. It is just that there are many complex rules.

Cherokee is rated 5.5, close to most difficult of all.

Iroquoian Northern Iroquoian Five Nations-Huronian-Susquehannock Huronian Huron-Petun

Wyandot, a dormant language that has been extinct for about 50 years, has some unbelievably complex structures. Let us look at one of them. Wyandot is the only language on Earth that allows negative sentences that somehow do not contain a negative morpheme. Wyandot makes it onto craziest grammars lists. (To be continued).

Siouan-Catawban Siouan Mississippi Valley-Ohio Valley Siouan Mississippi Valley Siouan Dakota

Lakota and other Siouan languages may well be as convoluted as Iroquoian. In Lakota, all adjectives are expressed as verbs. Something similar is seen in Nahuatl.

Ógle sápe kiŋ mak’ú. The shirt it is black he gave it to me. He gave me the black shirt.

In the above, it is black is a stative verb and serves as an adjective.

Ógle kiŋ sabyá mak’ú. Shirt the blackly he gave it to me. He gave me the black shirt. (Lit. He gave me the shirt blackly.)

Bkackly is an adverb serving as an adjective above.

Lakota gets a 5.5 rating, hardest of all.

Algic Algonquian

All Algonquian languages have distinctions between animate/inanimate nouns, in addition to having proximate/obviate and direct/inverse distinctions. However, most languages that have proximate/obviate and direct/inverse distinctions are not as difficult as Algonquian.

Proximate/obviative is a way of marking the 3rd person in discourse. It distinguishes between an important 3rd person (proximate) and a more peripheral 3rd person (obviative). Animate nouns and possessor nouns tend to be marked proximate while inanimate nouns and possessed nouns tend to be marked obviative.

Direct/inverse is a way of marking discourse in terms of saliency, topicality or animacy. Whether one noun ranks higher than another in terms of saliency, topicality or animacy means that that nouns ranks higher in terms of person hierarchy. It is used only in transitive clauses. When the subject has a higher ranking than the object, the direct form is used. When the object has a higher ranking than the object, the inverse form is used.

Central Algonquian Cree-Montagnais

Cree is very hard to learn. It are written in a variety of different ways with different alphabets and syllabic systems, complicating matters even further. The syllabic alphabet has many problems and is often listed as one of the worst scripts out there. They are both polysynthetic and have long, short and nasal vowels and aspirated and unaspirated voiceless consonants. Words are divided into metrical feet, the rules for determining stress placement in words are quite complex and there is lots of irregularity. Vowels fall out a lot, or syncopate, within words.

Cree adds noun classifiers to the mix, and both nouns and verbs are marked as animate or inanimate. In addition, verbs are marked for transitive and intransitive. In addition, verbs get different affixes depending on whether they occur in main or subordinate clauses.

Cree is rated 6, hardest of all.

Ojibwa-Patowatomi

Ojibwa is said to be about as hard to learn, as Cree as it is very similar.

Ojibwa is rated 6, hardest of all.

Plains Algonquian Cheyenne

Cheyenne is well-known for being a hard Amerindian language to learn. Like many polysynthetic languages, it can have very long words.

Náohkêsáa’oné’seómepêhévetsêhésto’anéhe. I truly don’t know Cheyenne very well.

However, Cheyenne is quite regular, but has so many complex rules that it is hard to figure them all out.

Cheyenne is rated 6, hardest of all.

Arapahoan

Arapaho has a strange phonology. It lacks phonemic low vowels. The vowel system consists of i, ɨ~,u, ɛ, and ɔ, with no low phonemic vowels. Each vowel also has a corresponding long version. In addition, there are four diphthongs, ei, ou, oe and ie, several triphthongs, eii, oee, and ouu, as well as extended sequences of vowels such as eee with stress on either the first or the last vowel in the combination. Long vowels of various types are common:

Héétbih’ínkúútiinoo. I will turn out the lights.

Honoosóó’. It is raining.

There is a pitch accent system with normal, high and allophonic falling tones. Arapaho words also undergo some very wild sound changes.

Arapaho is rated 6, hardest of all.

Gros Ventre has a similar phonological system and similar elaborate sound changes as Arapaho.

Gros Ventre is rated 5, hardest of all.

Caddoan Northern Wichita

Wichita has many strange phonological traits. It has only one nasal. Labials are rare and appear in only two roots. It also may have only three vowels, i, e, and a, with only height as a distinction. Such a restricted vertical vowel distribution is only found in NW Caucasian and the Papuan Ndu languages. There is apparently a three-way contrast in vowel length – regular, long and extra-long.

This is only found in Mixe and Estonian. There are some interesting tenses. Perfect tense means that an act has been carried out. The strange intentive tense means that one hopes or hoped to to carry out an act. The habitual tense means one regularly engages in the activity, not that one is doing so at the moment.

Long consonant clusters are permitted.

kskhaːɾʔa

nahiʔinckskih while sleeping

There are many cases where a CVɁ sequence has been reduced to due to loss of the vowel, resulting in odd words such as:

ki·sɁ bone

Word order is ordered in accordance with novelty or importance.

hira:wisɁiha:s kiyari:ce:hire: Our ancestors God put us on this Earth.

weɁe hira:rɁ tiɁi na:kirih God put our ancestors on this Earth.

In the sentence above, “our ancestors” is actually the subject, so it makes sense that it comes first.

Wichita has inclusive and exclusive 3rd person plural and has singular, dual and plural. There is an evidential system where if you say you know something, you must say how you know it – whether it is personal knowledge or hearsay.

Wichita gets a 6 rating, hardest of all.

Hokan Tequislatecan Coastal Chantal

Huamelutec or Lowland Oaxaca Chantal has the odd glottalized fricatives , , ɬʼ and as its only glottalized consonants. They alternate with plain f, s, l and x. , ɬʼ and are extremely rare in the world’s languages, usually only found in 2-3 other languages, often in NW Caucasian. occurs only in one other language – Tlingit. is slightly more common, occurring five other languages including Tlingit. In other languages, these odd sounds derived from sequences of consonant + q: Cq -> Cʔ -> glottalized fricative.

Sentence structure is odd:

Hit the ball the man. Hit the man the ball. The man hit the ball.

All mean the same thing.

Huamelutec gets a 6 rating, hardest of all.

Karok

Karok is a language isolate spoken by a few dozen people in northern California. The last native speaker recently died, however, there are ~80 who have varying levels of L2 fluency.

In Karok, you can use a suffix for different types of containment – fire, water or a solid.

pa:θ-kirih throw into a fire

pa:θ-kurih throw into water

pa:θ-ruprih throw through a solid

The suffixes are unrelated to the words for fire, water and solid.

Karok gets a 5 rating, hardest of all.

Uto-Aztecan Northern

Hopi is so difficult that even grammars describing the language are almost impossible to understand. For instance, Hopi has two different words for and depending on whether the noun phrase containing the word and is nominative or accusative.

Hopi is rated 6, hardest of all.

Southern Uto-Aztecan Corachol-Aztecan Core Nahua Nahuatl

In Nahuatl, most adjectives are simply stative verbs. Hence:

Umntu omde waya eTenochtitlan. The man he is tall went to Tenochtitlan. The tall man went to Tenochtitlan.

He is tall is a stative verb in the above.

Nahuatl gets a 6 rating, hardest of all.

Numic Central Numic

Comanche is legendary for being one of the hardest Indian languages of all to learn. Reasons are unknown, but all Amerindian languages are quite difficult. I doubt if Comanche is harder than other Numic languages.

Bizarrely enough, Comanche has very strange sounds called voiceless vowels, which seems to be an oxymoron, as vowels would seem to be inherently voiced. English has something akin to voiceless vowels in the words particular and peculiar, where the bolded vowels act something akin to a voiceless vowel.

Comanche was used for a while by the codespeakers in World War 2 – not all codespeakers were Navajos. Comanche was specifically chosen because it was hard to figure out. The Japanese were never able to break the Comanche code.

Comanche is rated 6, hardest of all.

Oto-Manguean Western Oto-Mangue Oto-Pame-Chinantecan Chinantecan

Chinantec, an Indian language of southwest Mexico, is very hard for non-Chinantecs to learn. The tone system is maddeningly complex, and the syntax and morphology are very intricate.

Chinantec is rated 6, hardest of all.

Popolocan Mazatecan Lowland Valley Southern

Jalapa Mazatec has distinctions between modal, creaky, breathy-voiced vowels along with nasal versions of those three. It also has creaky consonants and voiceless nasals. It has three tones, low, mid and high. Combining the tones results in various contour tones. In addition, it has a 3-way distinction in vowel length. Whistled speech is also possible. It has a phonemic distinction between “ballistic” and “controlled” syllables which is only present on Oto-Manguean.

Ballistic (short) warm nīˑntūslippery tsǣguava hų̄you plural

Controlled (half-long) sūˑblue nīˑntūˑneedle tsǣˑfull hų̄ˑ – six

Jalapa Mazatec is rated 6, hardest of all.

Maipurean Northern Upper Amazon Eastern Nawiki

Tariana is a very difficult language mostly because of the unbelievable amount of information it crams into its morphology and syntax. This is mostly because it is an Arawakan language that has been heavily influenced by neighboring Tucanoan languages, with the result that it has many of the grammatical categories and particles present in both families.

This stems from the widespread bilingualism in the Vaupes Basin of Colombia, where many people grow up bilingual from childhood and often become multilingual by adulthood. Learning up to five different languages is common. Code-switching was frowned upon and anyone using a word from Language Y while speaking Language X would get laughed at. Hence the various languages tended to borrow features from each other quite easily.

For instance, Tariana has both a noun classifier system and a gender system. Noun classifiers and gender are sometimes subsumed under the single category of “noun classifiers.” Yet Tariana has both, presumably from its relationship to two completely different language families. So in Tariana is not unusual to get both demonstratives and verbs marked for both gender and noun classifier. Tariana borrowed such things as serialized perception verbs and the dubitative marker from Tucano.

In addition, Tariana has some very odd sounds, including aspirated nasals mh (), nh (n̺ʰ) and ñh (ɲʰ) and an aspirated w () of all things. They seem to be actually aspirated, not just partially devoiced as many voiceless nasals and liquids are.

Tariana gets 6, hardest of all.

Huitotoan Proto-Bora-Muinane

Bora, a Wintotoan language spoken in Peru and Colombia near the border between the two countries, has a mind-boggling 350 different noun classes. The noun classifier system is actually highly productive and is often used to create new nouns. New nouns can be created very easily, and their meanings are often semantically transparent. In some noun classifier systems, classifiers can be stacked one upon the other. In these cases, typically the last one is used for agreement purposes.

Bora also is a tonal language, but it has only two tones. In addition, nearly all consonantal phonemes have phonemic aspirated and palatalized counterparts. The agreement structure in the language is also quite convoluted. The classifier system effectively replaces much derivational morphology on the noun and noun compounding processes that other languages use to expand the meanings of nominals.

Bora gets a 6 rating, hardest of all.

Tucanoan Eastern Tucanoan Bará-Tuyuka

Tuyuca is a Tucanoan language spoken in by 450 people in the department of Vaupés in Colombia. An article in The Economist magazine concluded that it was the hardest language on Earth to learn.

It has a simple sound system, but it’s agglutinative, and agglutinative languages are pretty hard. For instance, hóabãsiriga means I don’t know how to write. It has two forms of 1st person plural, I and you (inclusive) and I and the others (exclusive). It has between 50-140 noun classes, including strange ones like bark that does not cling closely to a tree, which can be extended to mean baggy trousers or wet plywood that has begun to fall apart.

Like Yamana, a nearly extinct Amerindian language of Chile, Tuyuca marks for evidentiality, that is, how it is that you know something. For instance:

Diga ape-wi. = The boy played soccer. (I saw him playing). Diga ape-hiyi. = The boy played soccer. (I assume he was playing soccer, though I did not see it firsthand).

Evidential marking is obligatory on all Tuyuca verbs and it forces you to think about how you know whatever it is you know.

Tuyuca definitely gets a 6 rating!

Central Tucanoan

Cubeo, a language spoken in the Vaupes of Colombia, has a either SOV or OVS. That would mean that the following:

The man the ball hit. The ball hit the man.

Mean the same things. OVS languages are quite rare.

Morphemes belong to one of four classes:

  1. Nasal (many roots, as well as suffixes like -xã  = associative)
  2. Oral (many roots, as well as suffixes like -pe  = similarity, -du = frustrative)
  3. Unmarked (only suffixes, e.g. -re  = in/direct object)
  4. Oral/Nasal (some roots and some suffixes) /bãˈkaxa-/(mãˈkaxa-) – to defecate and -kebã = suppose

Just by looking at any given consonant-initial suffix, it is impossible to determine which of the first three categories it belongs to. They must be learned one by one.

Cubeo has nasal assimilation, common to many Amazonian languages. In some of these, nasalization is best analyzed at the syllable level – some syllables are nasal and others are not.

dĩ-bI-ko /dĩ-bĩ-ko/ nĩmĩko She recently went.

The underlying form dĩ-bI-ko is realized on the surface as nĩmĩko. The ĩ in dĩ-bI-ko nasalizes the d, the b, and the I on either side of it, so nasal spreading works in both directions. However, it is blocked from the third syllable because k is part of a class of non-nasalizable consonants.

Pretty difficult language.

Cuneo gets a 6 rating, hardest of all.

Carib Waiwai

Hixkaryána is famous for being the only language on Earth to have basic OVS (Object-Verb-Subject) word order.

The sentence Toto yonoye kamara, or The man ate the jaguar, actually means The jaguar ate the man.

Toto yonoye kamara Lit. The man ate the jaguar. Gloss: The jaguar ate the man.

Grammatical suffixes attached to the end of the verb mark not only number but also aspect, mood and tense.

Hixkaryána gets a 6 rating, hardest of all.

Nambikwaran Mamaindê

This is actually a series of closely related languages as opposed to one language, but the Southern Nambikwara language is the most well-known of the family, with 1,200 speakers in the Brazilian Amazon.

Phonology is complex. Consonants distinguish between aspirated, plain and glottalized, common in the Americas. There are strange sounds like prestopped nasals glottalized fricatives. There are nasal vowels and three different tones. All vowels except one have both nasal, creaky-voiced and nasal-creaky counterparts, for a total of 19 vowels.

The grammar is polysynthetic with a complex evidential system.

Reportedly, Nambikwara children do not pick up the language fully until age 10 or so, one of the latest recorded ages for full competence. Nambikwara is sometimes said to be the hardest language on Earth to learn, but it has some competition.

Nambikwara definitely gets a 6 rating, hardest of all!

Muran

Pirahã is a language isolate spoken in the Brazilian Amazon. Recent writings by Daniel Everett indicate that not only is this one of the hardest languages on Earth to learn, but it is also one of the weirdest languages on Earth. It is monumentally complex in nearly every way imaginable. It is commonly listed on the rogue’s gallery of craziest languages and phonologies on Earth.

It has the smallest phonemic inventory on Earth with only seven consonants, three vowels and either two or three tones. Everett recently wrote a paper about it after spending many years with them. Previous missionaries who had spent time with the Pirahã generally failed to learn the language because it was too hard to learn. It took Everett a very long time, but he finally learned it well.

Many of Everett’s claims about Pirahã are astounding: whistled speech, no system for counting, very few Portuguese loans (they deliberately refuse to use Portuguese loans) evidence for the Sapir-Whorf linguistic relativity hypothesis, and evidence that it violates some of Noam Chomsky’s purported language universals such as embedding. It also has the t͡ʙ̥ sound – a bilabially trilled postdental affricate which is only found in two other languages, both in the Brazilian Amazon – Oro Win and Wari’.

Initially, Everett never heard the sound, but they got to know him better, they started to make it more often. Everett believes that they were ridiculed by other groups when they made the odd sound.

Pirahã has the simplest kinship system in any language – there is only word for both mother and father, and the Pirahã do not have any words for anyone other than direct biological relatives.

Pirahã may have only two numerals, or it may lack a numeral system altogether.

Pirahã does not distinguish between singular and plural person. This is highly unusual. The language may have borrowed its entire pronoun set from the Tupian languages Nheengatu and Tenarim, groups the Pirahã had formerly been in contact with. This may be one of the only attested case of the borrowing of a complete pronoun set.

There are mandatory evidentiality markers that must be used in Pirahã discourse. Speakers must say how they know something, whether they saw it themselves, whether it was hearsay or whether they inferred it circumstantially.

There are various strange moods – the desiderative (desire to perform an action) and two types of frustrative – frustration in starting an action (inchoative/incompletive) and frustration in completing an action (causative/incompletive). There are others: immediate/intentive (you are going to do something now/you intend to do it in the future)

There are many verbal aspects: perfect/imperfect (completed/incomplete) telic/atelic (reaching a goal/not reaching a goal), continuative (continuing), repetitive (iterative), and beginning an action (inchoative).

Each Pirahã verb has 262,144 possible forms, or possibly in the many millions, depending on which analysis you use.

The future tense is divided into future/somewhere and future/elsewhere. The past tense is divided into plain past and immediate past.

Pirahã has a closed class of only 90 verb roots, an incredibly small number. But these roots can be combined together to form compound verbs, a much larger category. Here is one example of three verbs strung together to form a compound verb:

xig ab op take turn gobring back, You take something away, you turn around, and you go back to where you got it to return it.

There are no abstract color terms in Pirahã. There are only two words for colors, one for light and one for dark. The only other languages with this restricted of a color sense are in Papua New Guinea. The other color terms are not really color terms, but are more descriptive – red is translated as like blood.

Pirahã can be whistled, hummed or encoded into music. Consonants and vowels can be omitted altogether and meaning conveyed instead via variations in stress, pitch and rhythm. Mothers teach the language to children by repeating musical patterns.

Pirahã may well be one of the hardest languages on Earth to learn.

Pirahã gets a 6 rating, hardest of all.

Quechuan

Quechua (actually a large group of languages and not a single language at all) is one of the easiest Amerindian languages to learn. Quechua is a classic example of a highly regular grammar with few exceptions. Its agglutinative system is more straightforward than even that of Turkish. The phonology is dead simple.

On the down side, there is a lot of dialectal divergence (these are actually separate languages and not dialects) and a lack of learning materials. Some say that Quechua speakers spend their whole lives learning the language.

Quechua has inconsistent orthographies. There is a fight between those who prefer a Spanish-based orthography and those who prefer a more phonemic one. Also there is an argument over whether to use the Ayacucho language or the Cuzco language as a base.

Quechua has a difficult feature known as evidential marking. This marker indicates the source of the speaker’s knowledge and how sure they are about the statement.

-mi expresses personal knowledge:

Tayta Wayllaqawaqa chufirmi. Mr. Huayllacahua is a driver. (I know it for a fact.)

-si expresses hearsay knowledge:

Tayta Wayllaqawaqa chufirsi. Mr. Huayllacahua is a driver (or so I’ve heard).

chá expresses strong possibility:

Tayta Wayllaqawaqa chufirchá. Mr. Huayllacahua is a driver (most likely).

Quechua is rated 4, very difficult.

Aymaran Aymara

Aymara has some of the wildest morphophonology out there. Morpheme-final vowel deletion is present in the language as a morphophonological process, and it is dependent on a set of highly complex phonological, morphological and syntactic rules (Kim 2013).

For instance, there are three types of suffixes: dominant, recessive and a 3rd class is neither dominant nor recessive. If a stem ends in a vowel, dominant suffixes delete the vowel but recessive suffixes allow the vowel to remain. The third class either deletes or retains the vowel on the stem depending on how many vowels are in the stem. If the root has two vowels, the vowel is retained. If it has three vowels, the vowel is deleted.

Although all of this seems quite odd, Finnish has something similar going on, if not a lot worse.

Nevertheless, Aymara is still said to be a very easy language to learn. The Guinness Book of World Records claims it is almost as easy to learn as Esperanto.

Aymara gets a 2 rating, very easy to learn.

Australian

Australian Aborigine languages are some of the hardest languages on Earth to learn, like Amerindian or Caucasian languages. Some Australian languages have phonemic contrasts that few other languages have, such as apico-dental, lamino-dental, apico-post-alveolar, and lamino-postalveolar cononals.

Australian languages tend to be mixed ergative. Ordinary nouns are ergative-absolutive, but 1st and 2nd person pronouns are nominative-accusative. One language has a three way agent-patient-experiencer distinction in the 1st person pronoun. Australian pronouns typically have singular, plural and dual forms along with inclusive and exclusive 1st plural. In some sentences, they have what is known as double case agreement which is rare in the world’s languages:

I gave a spear to my father. I gave a spear mine-to father’s-to.

Both elements of the phrase my father are in both dative and genitive.

However, Aboriginal languages do have the plus of being very regular.

All Australian languages are rated 6, most difficult of all.

Tor-Kwerba Orya-Tor Tor

Berik is a Tor-Orya language spoken in Indonesian colony of Irian Jaya in New Guinea.

Verbs take many strange endings, in many cases mandatory ones, that indicate what time of day something happened, among other things.

TelbenerHe drinks in the evening.

Where a verb takes an object, it will not only be marked for time of day but for the size of the object.

KitobanaHe gives three large objects to a man in the sunlight.

Verbs may also be marked for where the action takes place in reference to the speaker.

GwerantenaTo place a large object in a low place nearby.

Berik is rated 6, hardest of all.

Trans New Guinea Madang Croisilles Gum

Amele is the world’s most complex language as far as verb forms go, with 69,000 finitive and 860 infinitive forms.

Amele is rated 6, hardest of all.

Torricelli Wapei Valman

Valman is a bizarre case where the word and that connects two nouns is actually a verb of all things and is marked with the first noun as subject and the second noun as object.

John (subject) and Mary (object)

John is marked as subject for some reason, and Mary is marked as object, and the and word shows subject agreement with John and object agreement with Mary.

Valman gets a 6 rating, hardest of all.

Afroasiatic Semitic

Semitic languages such as Arabic and Hebrew are notoriously difficult to learn, and Arabic (especially MSA) tops many language learners’ lists as the hardest language they have ever attempted to learn. Although Semitic verbs are notoriously complex, the verbal system does have some advantages especially as compared to IE languages like Slavic. Unlike Slavic, Semitic verbs are not inflected for mood and there is no perfect or imperfect.

Central South Arabic

Arabic has some very irregular manners of noun declension, even in the plural. For instance, the word girls changes in an unpredictable way when you say one girl, two girls and three girls, and there are two different ways to say two girls depending on context. Two girls is marked with the dual, but different dual forms can be used. All languages with duals are relatively difficult for most speakers that lack a dual in their native language. However, the dual is predictable from the singular, so one might argue that you only need to learn how to say one girl and three girls.

Further, it is full of irregular plurals similar to octopus and octopi in English, whereas these forms are rare in English. With any given word, there might be 20 different possible ways to pluralize it, and there is no way to know which of the 20 paradigms to use with that word, and further, there is no way to generalize a plural pattern from a singular pattern. In addition, many words have 2-3 ways of pluralizing them. Some messy Arab plurals:

kalb -> kilaab qalb -> quluub maktab -> makaatib taalib -> tullaab balad -> buldaan

When you say I love you to a man, you say it one way, and when you say it to a woman, you say it another way. On and on.

The Arabic writing system is exceeding difficult and is more of the hardest to use of any on Earth. Soft vowels are omitted. You have to learn where to insert missing vowels, where to double consonants and which vowels to skip in the script. There are 28 different symbols in the alphabet and four different ways to write each symbol depending on its place in the word.

Consonants are written in different ways depending on where they appear in a word. An h is written differently at the beginning of a word than at the end of a word. However, one simple aspect of it is that the medial form is always the same as the initial form. You need to learn not only Arabic words but also the grammar to read Arabic.

Pronouns attach themselves to roots, and there are many different verb conjugation paradigms which simply have to be memorized. For instance, if a verb has a و, a ي, or a ء  in its root, you need to memorize the patters of the derivations, and that is a good chunk of the conjugations right there. The system for measuring quantities is extremely confusing.

The grammar has many odd rules that seem senseless. Unfortunately, most rules have exceptions, and it seems that the exceptions are more common than the rules themselves. Many people, including native speakers, complain about Arabic grammar.

Arabic does have case, but the system is rather simple.

The laryngeals, uvulars and glottalized sounds are hard for many foreigners to make and nearly impossible for them to get right. The ha’(ح ), qa (ق ) and غ sounds and the glottal stop in initial position give a lot of learners headaches.

Arabic is at least as idiomatic as French or English, so it order to speak it right, you have to learn all of the expressionistic nuances.

One of the worst problems with Arabic is the dialects, which in many cases are separate languages altogether. If you learn Arabic, you often have to learn one of the dialects along with classical Arabic. All Arabic speakers speak both an Arabic dialect and Classical Arabic.

In some Arabic as a foreign language classes, even after 1 1/2 years, not one student could yet make a complete and proper sentence that was not memorized.

Adding weight to the commonly held belief that Arabic is hard to learn is research done in Germany in 2005 which showed that Turkish children learn their language at age 2-3, German children at age 4-5, but Arabic kids did not get Arabic until age 12.

Arabic has complex verbal agreement with the subject, masculine and feminine gender in nouns and adjectives, head-initial syntax and a serious restriction to forming compounds. If you come from a language that has similar nature, Arabic may be easier for you than it is for so many others. Its 3 vowel system makes for easy vowels.

MSA Arabic is rated 5, extremely difficult.

Arabic dialects are often somewhat easier to learn than MSA Arabic. At least in Lebanese and Egyptian Arabic, the very difficult q’ sound has been turned into a hamza or glottal stop which is an easier sound to make. Compared to MSA Arabic, the dialectal words tend to be shorter and easier to pronounce.

To attain anywhere near native speaker competency in Egyptian Arabic, you probably need to live in Egypt for 10 years, but Arabic speakers say that few if any second language learners ever come close to native competency. There is a huge vocabulary, and most words have a wealth of possible meanings.

Egyptian Arabic is rated 4.5, very to extremely difficult.

Moroccan Arabic is said to be particularly difficult, with much vowel elision in triconsonantal stems. In addition, all dialectal Arabic is plagued by irrational writing systems.

Moroccan Arabic is rated 4.5, very to extremely difficult.

Maltese is a strange language, basically a Maghrebi Arabic language (similar to Moroccan or Tunisian Arabic) that has very heavy influence from non-Arabic tongues. It shares the problem of Gaelic that often words look one way and are pronounced another.

It has the common Semitic problem of difficult plurals. Although many plurals use common plural endings (-i, -iet, -ijiet, -at), others simply form the plural by having their last vowel dropped or adding an s (English borrowing). There’s no pattern, and you simply have to memorize which ones act which way. Maltese permits the consonant cluster spt, which is surely hard to pronounce.

On the other hand, Maltese has quite a few IE loans from Italian, Sicilian, Spanish, French and increasingly English. If you have knowledge of Romance languages, Maltese is going to be easier than most Arabic dialects.

Maltese is rated 4, very difficult.

South Canaanite

Hebrew is hard to learn according to a number of Israelis. Part of the problem may be the abjad writing system, which often leaves out vowels which must simply be remembered. Also, other than borrowings, the vocabulary is Afroasiatic, hence mostly unknown to speakers of IE languages. There are also difficult consonants as in Arabic such as pharyngeals and uvulars.

The het or glottal h is particularly hard to make. However, most modern Israelis no longer make the het sound or a’ain sounds. Instead, they pronounce the het like the chaf sound and the a’ain like an alef. Almost all Ashkenazi Israeli Jews no longer use the het or a’ain sounds. But most Jews who came from Arab countries (often older people) still use the sound, and some of their children do (Dorani 2013).

Hebrew has complex morphophonological rules. The letters p, b, t, d, k and g change to v, f, dh, th, kh and gh in certain situations. In some environments, pharyngeals change the nature of the vowels around them. The prefix ve-, which means and, is pronounced differently when it precedes certain letters. Hebrew is also quite irregular.

Hebrew has quite a few voices, including active, passive, intensive, intensive passive, etc. It also has a number of tenses such as present, past, and the odd juissive.

Hebrew also has two different noun classes. There are also many suffixes and quite a few prefixes that can be attached to verbs and nouns.

Even most native Hebrew speakers do not speak Hebrew correctly by a long shot.

Quite a few say Hebrew is as hard to learn as MSA or perhaps even harder, but this is controversial.

Hebrew gets a 5 rating for extremely difficult.

Berber Northern Atlas

Berber languages are considered to be very hard to learn. Worse, there are very few language learning resources available.

Tamazight allows doubled consonants at the beginning of a word! How can you possibly make that sound?

Tamazight gets a 6 rating, hardest of all.

In Tachelhit , words like this are possible:

tkkststt You took it off.

tfktstt You gave it.

In addition, there are words which contain only one or two consonants:

ɡ be

ks feed on

Tachelhit gets a 6 rating, hardest of all.

South Ethiopian South Transversal Amharic–Argobba Amharic

Amharic is said to be a very hard language to learn. It is quite complex, and its sentence structures seem strange even to speakers of other Semitic languages. Hebrew speakers say they have a hard time with this language.

There are a multitude of rules which almost seem ridiculous in their complexity, there are numerous conjugation patterns, objects are suffixed to the verb, the alphabet has 274 letters, and the pronunciation seems strange. However, if you already know Hebrew or Arabic, it will be a lot easier. The hardest part of all is the verbal system, as with any Semitic language. It is easier than Arabic.

Amharic gets a 4.5 rating, very hard to extremely hard.

Cushitic East Cushitic

Dahalo is legendary for having some of the wildest consonant phonology on Earth. It has all four airstream mechanisms found in languages: ejectives, implosives, clicks and normal pulmonic sounds. There are both glottal and epiglottal stops and fricatives and laminal and apical stops.

There is also a strange series of nasal clicks and are both glottalized and plain. Some of these clicks are also labialized. It has both voiced and unvoiced prenasalized stops and affricates, and some of the stops are also labialized. There is a weird palatal lateral ejective. There are three different lateral fricatives, including a labialized and palatalized one, and one lateral approximant. It contrasts alveolar and palatal lateral affricates and fricatives, the only language on Earth to do this.

The Dahalo are former elephant hunting hunter gatherers who live in southern Kenya. It is believed that at one time they spoke a language like Sandawe or Hadza, but they switched over to Cushitic at some point. The clicks are thought to be substratum from a time when Dahalo was a Sandawe-Hadza type language.

Dahalo gets a 6 rating, hardest of all.

Somali

Somali has one of the strangest proposition systems on Earth. It actually has no real prepositions at all. Instead it has preverbal particles and possessives that serve as prepositions.

Here is how possessives serve as prepositions:

habeennimada horteeda the night her front before nightfall

kulaylka dartiisa the heat his reason because of the heat

Here we have the use of a preverbal particle serving as a preposition:

kú ríd shandádda Into put the suitcase. Put it into the suitcase.

Somali combines four “prepositions” with four deictic particles to form its prepositions.

There are four basic “prepositions”:

to in from with

These combine with a four different deictic particles:

toward the speaker away from the speaker toward each other away from each other

Hence you put the “prepositions” and the deictic particles together in various ways. Both tend to go in front of and close to the verb:

Nínkíi bàan cèelka xádhig kagá sóo saaray. …well-the rope with-from towards-me I-raised. I pulled the man out of the well with a rope.

Way inoogá warrámi jireen. They us-to-about news gave. They used to give us news about it.

Prepositions are the hardest part of the Somali language for the learner.

Somali deals with verbs of motion via deixis in a similar way that Georgian does. One reference point is the speaker and the other is any other entities discussed. Verbs of motion are formed using adverbs. Entities may move:

towards each other    wada
away from each other  kala
towards the speaker   so
away from the speaker si

Hence:

kala durka separate
si gal     go in (away from the speaker)
so gal     come in (toward the speaker)

Somali lacks orthographic consistency. There are four different orthographic systems in use – the lists.

Somali pluralization makes no sense and must be memorized. There are seven different plurals, and there is no clue in the singular that tells you what form to use in the plural. See here:

Republication:

áf  (language) -> afaf

Suffixation:

hoóyo (mother) -> hoyoóyin

áabbe -> aabayaal

Note the tone shifts in all three of the plurals above.

There are four cases, absolutive, nominative, genitive and vocative. Despite the presences of absolutive and nominative cases, Somali is not an ergative language. Absolutive case is the basic case of the noun, and nominative is the case given to the noun when a verb follows in the sentence. There are different articles depending on whether the noun was mentioned previously or not (similar to the articles a and the in English). The absolutive and nominative are marked not only on the noun but also on the article that precedes it.

In terms of difficulty, Somali is much harder than Persian and probably about as difficult as Arabic.

Somali gets a 5 rating, extremely hard to learn.

Dravidian Southern Tamil-Kannada Tamil-Kodagu Tamil-Malayalam Malayalam

Malayalam, a Dravidian language of India, was has been cited as the hardest language to learn by an language foundation, but the citation is obscure and hard to verify.

Malayalam words are often even hard to look up in a Malayalam dictionary.

For instance, adiyAnkaLAkkikkoNDirikkukayumANello is a word in Malayalam. It means something like I, your servant, am sitting and mixing s.t. (which is why I cannot do what you are asking of me). The part in parentheses is an example of the type of sentence where it might be used.

The above word is composed of many different morphemes, including conjunctions and other affixes, with sandhi going on with some of them so they are eroded away from their basic forms. There doesn’t seem to be any way to look that word up or to write a Malayalam dictionary that lists all the possible forms, including forms like the word above. It would probably be way too huge of a book. However, all agglutinative languages are made up of affixes, and if you know the affixes, it is not particularly hard to parse the word apart.

Malayalam is said to be very hard to pronounce correctly.

Further, few foreigners even try to learn Malayalam, so Malayalam speakers, like the French, might not listen to you and might make fun of you if your Malayalam is not native sounding.

However, Malayalam has the advantage of having many pedagogic materials available for language learning such as audio-visual material and subtitled videos.

Malayalam is rated 5, extremely difficult.

Tamil

Tamil, a Dravidian language is hard, but probably not as difficult as Malayalam is. Tamil has an incredible 247 characters in its alphabet. Nevertheless, most of those are consonant-vowel combinations, so it is almost more of a syllabary than an alphabet. Going by what would traditionally be considered alphabetic symbols, there are probably only 72 real symbols in the alphabet. Nevertheless, Tamil probably has one of the easier Indic scripts as Tamil has fewer characters than other scripts due to its lack of aspiration. Compare to Devanagari’s over 1,000 characters.

But no Indic script is easy. A problem with Tamil is that all of the characters seem to look alike. It is even worse than Devanagari in that regard. However, the more rounded scripts such as Kannada, Sinhala, Telegu and Malayalam have that problem to a worse degree. Tamil has a few sharp corners in the characters that helps to disambiguate them.

In addition, as with other languages, words are written one way and pronounced another. However, there are claims that the difficulty of Tamil’s diglossia is overrated.

Tamil has two different registers for written and spoken speech, but the differences are not large, so this problem is exaggerated. Both Tamil and Malayalam are spoken very fast and have extremely complicated, nearly impenetrable scripts. If Westerners try to speak a Dravidian language in south India, more often than not the Dravidian speaker will simply address them in English rather than try to accommodate them.

Tamil has the odd evidential mood, similar to Bulgarian.

However, on the plus side, the language does seem to be very logical and regular, almost like German in that regard. In addition, there are a lot of language learning materials for Tamil.

Tamil is rated 4, very difficult.

Altaic Korean

Most agree that Korean is a hard language to learn.

The alphabet, Hangul at least is reasonable; in fact, it is quite elegant. But there are four different Romanizations- Lukoff, Yale, Horne, and McCune-Reischauer – which is preposterous. It’s best to just blow off the Romanizations and dive straight into Hangul. This way you can learn a Romanization later, and you won’t mess up your Hangul with spelling errors, as can occur if you go from Romanization to Hangul.

Hangul can be learned very quickly, but learning to read Korean books and newspapers fast is another matter altogether because you really need to know the hanja or Chinese character that are used in addition to the Hangul. After World War 2, the Koreas decided to officially get rid of their Chinese characters, but in practice this was not successful. With the use of Chinese characters in Korean, you can be a lot more precise in terms what you are trying to communicate.

Bizarrely, there are two different numeral sets used, but one is derived from Chinese so it should be familiar to Chinese, Japanese or Thai speakers who use similar or identical systems.

Korean has a wealth of homonyms, and this is one of the tricky aspects of the language. Any given combination of a couple of characters can have multiple meanings. Japanese has a similar problem with homonyms, but at least with Japanese you have the benefit of kanji to help you tell the homonyms apart. With Korean Hangul, you get no such advantage.

Similarly, there seem to be many ways to say the same thing in Korean. The learner will feel when people are using all of these different ways of saying the same thing that they are actually saying something different each time, but that is not the case.

One problem is that the bp, j, ch, t and d are pronounced differently than their English counterparts. The consonants, the pachim system and the morphing consonants at the end of the word that slide into the next word make Korean harder to pronounce than any major European language. Korean has a similar problem with Japanese, that is, if you mess up one vowel in sentence, you render it incomprehensible.

The vocabulary is very difficult for an English speaker who does not have knowledge of either Japanese or Chinese. On the other hand, Japanese or Chinese will help you a lot with Korean.

Korean is agglutinative and has a subject-topic discourse structure, and the logic of these systems is difficult for English speakers to understand. In addition, there are hundreds of ways of conjugating any given verb based on tense, mood, age or seniority. Adjectives also decline and take hundreds of different suffixes.

Meanwhile, Korean has an honorific system that is even wackier than that of Japanese. A single sentence can be said in three different ways depending on the relationship between the speaker and the listener. However, the younger generation is not using the honorifics so much, and a foreigner isn’t expected to know the honorific system anyway.

Maybe 6

Speakers of Korean can learn Japanese fairly easily. Korean seems to be a more difficult language to learn than Japanese. There are maybe twice as many particles as in Japanese, the grammar is dramatically more difficult and the verbs are quite a bit harder. The phonemic inventory in Korean is also larger and includes such oddities as double consonants.

Korean is rated by language professors as being one of the hardest languages to learn.

Korean is rated 5, extremely hard.

Japonic

Japanese also uses a symbolic alphabet, but the symbols themselves are sometime undecipherable in that even Japanese speakers will sometimes encounter written Japanese and will say that they don’t know how to pronounce it. I don’t mean that they mispronounce it; that would make sense. I mean they don’t have the slightest clue how to say the word! This problem is essentially nonexistent in a language like English.

The Japanese orthography is one of the most difficult to use of any orthography.

There are over 2,000 frequently used characters in three different symbolic alphabets that are frequently mixed together in confusing ways. Due to the large number of frequently used symbols, it’s said that even Japanese adults learn a new symbol a day a ways into adulthood.

The Japanese writing system is probably crazier than the Chinese writing system and it often makes it onto lists of worst orthographies. The very idea of writing an agglutinative language in a combination of two syllabaries and an ideography seems wacky right off the bat. Japanese borrowed Chinese characters.

But then they gave each character several pronunciations, and in some cases as many as 24. Next they made two syllabaries using another set of characters, then over the next millennia came up with all sorts of contradictory and often senseless rules about when to use the syllabaries and when to use the character set. Later on they added a Romanization to make things even worse.

Chinese uses 5-6,000 characters regularly, while Japanese only uses around 2,000. But in Chinese, each character has only one or maybe two pronunciations. In Japanese, there are complicated rules about when and how to combine the hiragana with the characters. These rules are so hard that many native speakers still have problems with them. There are also personal and place names (proper nouns) which are given completely arbitrary pronunciations often totally at odds with the usual pronunciation of the character.

There are some writers, typically of literature, who deliberately choose to use kanji that even Japanese people cannot read. For instance, Ryuu Murakami  uses the odd symbols 擽る、, 轢く、and 憑ける.

The Japanese system is made up of three different systems: the katakana and hiragana (the kana) and the kanji, similar to the hanzi used in Chinese. Chinese has at least 85,000 hanzi. The number of kanji is much less than that, but kanji often have more than one meaning in contrast to hanzi.

After WW2, Japan decided to simplify its language. They both simplified and reduced the number of Chinese characters used, and they unified the written and spoken language, which previously had been different.

Speaking Japanese is not as difficult as everyone says, and many say it’s fairly easy. However, there is a problem similar to English in that one word can be pronounced in multiple ways, like read and read in English.

A common problem is that a perfectly grammatically correct sentence uttered by a Japanese language learner, while perfectly correct, is still not acceptable by Japanese speakers because “we just don’t say it that way.” The Japanese speaker often cannot tell why the unacceptable sentence you uttered is not ok. On the other hand, this problem may be common to more languages than Japanese.

There is also a class of Japanese called “honorifics” or “keigo” that is quite hard to master. Honorifics are meant to show respect and to indicate one’s place or status in the social hierarchy. These typically effect verbs but can also affect particles and prefixes. They are usually formed by archaic or highly irregular verbs. However, there are both regular and irregular honorific forms. Furthermore, there are five different levels of honorifics. Honorifics vary depending on who you are and who you are talking to. In addition, gender comes into play.

Although it is true the Japanese young people are said to not understand the intricacies of keigo, it is still expected that they know how to speak this well. Consequently, many young Japanese will opt out of certain conversations because they feel that their keigo is not very good. Books explaining how to use keigo properly have been big sellers among young people in Japan in recent years as young people try to appear classy, refined or cultured.

In addition, Japanese born overseas (especially in the US), while often learning Japanese pretty well, typically have a very poor understanding of keigo. Instead of embarrassing themselves by not using keigo or using it wrong, these Japanese speakers often prefer to speak in English to Japanese people rather than bother with keigo-less Japanese. Overcorrection in keigo is also a problem when hypercorrection leads to someone making errors in keigo due to “trying to hard.” This looks like phony or insincere politeness and is often worse than not using keigo at all.

One wild thing about Japanese is counting forms. You actually use different numeral sets depending on what it is you are counting! There are dozens of different ways of counting things which involve the use of a complex numerical noun classifier system.

Japanese grammar is often said to be simple, but that does not appear to be the case on closer examination. Particles are especially vexing. Verbs engage in all sorts of wild behavior, and adverbs often act like verbs. Nouns can act like adjectives and adverbs. Meanwhile, honorifics change the behavior of all words. There are particles like ha and ga that have many different meanings. One problem is that all noun modifiers, even phrases, must precede the nouns they are modifying.

It’s often said that Japanese has no case, but this is not true. Actually, there are seven cases in Japanese. The aforementioned ga is a clitic meaning nominative, made is terminative case, -no is genitive and -o is accusative.

In this sentence:

The plane that was supposed to arrive at midnight, but which had been delayed by bad weather, finally arrived at 1 AM.

Everything underlined must precede the noun plane:

Was supposed to arrive at midnight, but had been delayed by bad weather, the plane finally arrived at 1 AM.

One of the main problems with Japanese grammar is that it is going to seem to so different from the sort of grammar and English speaker is likely to be used to.

Speaking Japanese is one thing, but reading and writing it is a whole new ballgame. It’s perfectly possible to know the meaning of every kanji and the meaning of every word in a sentence, but you still can’t figure out the meaning of the sentence because you can’t figure out how the sentence is stuck together in such a way as to create meaning.

The real problem is that the Japanese you learn in class is one thing, and the Japanese of the street is another. One problem is that in street Japanese, the subject is typically not stated in a sentence. Instead it is inferred through such things as honorific terms or the choice of words you used in the sentence. Probably no one goes crazier on negatives than the Japanese. Particularly in academic writing, triple and quadruple negatives are common, and can be quite confusing.

Yet there are problems with the agglutinative nature of Japanese. It’s a completely different syntactic structure than English. Often if you translate a sentence from Japanese to English it will just look like a meaningless jumble of words.

However, Japanese grammar has the advantage of being quite regular. For instance, there are only four frequently used irregular verbs.

Like Chinese, the nouns are not marked for number or gender. However, while Chinese is forgiving of errors, if you mess up one vowel in a Japanese sentence, you may end up with incomprehension.

Although many Japanese learners feel it’s fairly easy to learn, surveys of language professors continue to rate Japanese as one of the hardest languages to learn. A study by the US Navy concluded that the hardest language the corpsmen had to learn in the course of service was Japanese. However, it’s generally agreed that Japanese is easier to learn than Korean. Japanese speakers are able to learn Korean pretty easily.

Japanese is rated 5, extremely hard.

Classical Japanese is much harder to read than Modern Japanese. Though you can get by with much less kanji when reading the modern language, you will need a minimum knowledge of 3,000 kanji for reading Classical Japanese, and that’s using a dictionary. There are only about 500-1,000 frequently used characters, but there are countless other words that will come up in your reading especially say special words used in the Imperial Court. Many words have more than one meaning, and unless you know this, you will be lost. 東宮(とうぐう) for instance means Eastern Palace. However, it also means Crown Prince because his residence was to the east of the Emperor’s.

The movie The Seven Samurai (set in the late 1500’s) seems to use some sort of Classical Japanese, or at least Classical vocabulary and syntax with modern pronunciation. Japanese language learners say they can’t understand a word of the archaic Japanese used in this movie.

Classical Japanese gets 5.5, nearly hardest of all.

Turkic Oghuz Western Oghuz

Turkish is often considered to be hard to learn, and it’s rated one of the hardest in surveys of language teachers, however, it’s probably easier than its reputation made it out to be. It is agglutinative, so you can have one long word where in English you might have a sentence of shorter words. One word is

Çekoslovakyalilastiramadiklarimizdanmissiniz? Were you one of those people whom we could not turn into a Czechoslovakian?

Many words have more than one meaning. However, the agglutination is very regular in that each particle of meaning has its own morpheme and falls into an exact place in the word. See here:

göz            eye
göz-lük        glasses
göz-lük-çü     optician
göz-lük-çü-lük the business of an optician

Nevertheless, agglutination means that you can always create new words or add new parts to words, and for this reason even a lot of Turkish adults have problems with their language.

There is no verb to be, which is hard for many foreigners. Instead, the concept is wrapped onto the subject of the sentence as a -dim or -im suffix. Turkish is an imagery-heavy language, and if you try to translate straight from a dictionary, it often won’t make sense.

However, the suffixation in Turkish, along with the vowel harmony, are both precise. Nevertheless, many words have irregular vowel harmony. The rules for making plurals are very regular, with no exceptions (the only exceptions are in foreign loans). In Turkish, incredible as it sounds, you can make a plural out of anything, even a word like what, who or blood. However, there is some irregularity in the strengthening of adjectives, and the forms are not predictable and must be memorized.

Turkish is a language of precision in other ways. For instance, there are eight different forms of subjunctive mood that describe various degrees of uncertainty that one has about what one is talking about. This relates to the evidentiality discussed under Tuyuca above, and Turkish has an evidential form similar to Tamil and Bulgarian. On Turkish news, verbs are generally marked with miş, which means that the announcer believes it to be true though he has not seen it firsthand. The particle miş is interesting because this evidential form is coded into the tense system, which is an unusual use of evidentiality.

The Roman alphabet and almost mathematically precise grammar really help out. Turkish lacks gender and has but a single irregular verb – olmak. Nevertheless, there are many verbal forms. However, this is controversial and it depends on how you define grammatical irregularity. There is some strangeness in some of the verb paradigms, but it is argued that these oddities are rule-based. The aorist tense is said to have irregularity.

There is some irregular morphophonology, but not much. The oblique relative clauses have complex morphosyntax. Turkish has two completely different ways of making relative clauses, one of which may have been borrowed from Persian. There are many gerunds for verbs, and these have many different uses. At the end of the day, Turkish grammar is not as regular or as simple as it is made out to be.

Words are pronounced nearly the same as they are written. A suggestion that Turkish may be easier to learn that many think is the research that shows that Turkish children learn attain basic grammatical mastery of Turkish at age 2-3, as compared to 4-5 for German and 12 for Arabic. The research was conducted in Germany in 2005.

In addition, Turkish has a phonetic orthography.

However, Turkish is hard for an English speaker to learn for a variety of reasons. It is agglutinative like Japanese, and all agglutinative languages are difficult for English speakers to learn. As in Japanese, you start your Turkish sentence the way you would end your English sentence. As in the Japanese example above, the subordinate clause must precede the subject, whereas in English, the subordinate clause must follow the subject. The italicized phrase below is a subordinate clause.

In English, we say, “I hope that he will be on time.”

In Turkish, the sentence would read, “That he will be on time I hope.”

Turkish vowels are unusual to speakers of IE languages, and Turkish learners say the vowels are hard to make or even tell apart from one another.

Turkish is rated 3.5, harder than average to learn.

Uralic

Finno-Ugric

One test of the difficulty of any language is how much of the grammar you must know in order to express yourself on a basic level. On this basis, Finno-Ugric languages are complicated because you need to know quite a bit more grammar to communicate on a basic level in them than in say, German.

Finnic Northern

Finnish is very hard to learn, and even long-time learners often still have problems with it. Famous polyglot Barry Farber said it was one of the hardest languages he learned. You have to know exactly which grammatical forms to use where in a sentence. In addition, Finnish has 15 cases in the singular and 16 in the plural. This is hard to learn for speakers coming from a language with little or no case.

For instance, talothe house

Cases:

talon        house's
taloasome    of the house
taloksiinto  as the house
talossain    the house
talostafrom  inside the house
talooninto   the house
talollaon    to the house
taloltafrom  beside the house
talolleto    the house
taloistafrom the houses
taloissa     in the houses

It gets much worse than that. This web page shows that the noun kauppashop can have 2,253 forms.

A simple adjective + noun type of noun phrase of two words can be conjugated in up to 100 different ways.

Adjectives and nouns belong to 20 different classes. The rules governing their case declension depend on what class the substantive is in.

As with Hungarian, words can be very long. For instance:

lentokonesuihkuturbiinimoottoriapumekaanikkoaliupseerioppilas non-commissioned officer cadet learning to be an assistant mechanic for airplane jet engines

Like Turkish, Finnish agglutination is very regular. Each bit of information has its own morpheme and has an exact place in the word.

Like Turkish, Finnish has vowel harmony, but the vowel harmony is very regular like that of Turkish. Unlike Turkish or Hungarian, consonant gradation forms a major part of Finnish morphology. In order to form a sentence in Finnish, you will need to learn about verb types, cases and consonant gradation, and it can take a while to get your mind around those things.

Finnish, oddly enough, always puts the stress on the first syllable. Finnish vowels will be hard to pronounce for most foreigners.

However, Finnish has the advantage of being pronounced precisely as it is written. This is also part of the problem though, because if you don’t say it just right, the meaning changes. So, similarly with Polish, when you mangle their language, you will only achieve incomprehension. Whereas with say English, if a foreigner mangles the language, you can often winnow some sense out of it.

However, despite that fact that written Finnish can be easily pronounced, when learning Finnish, as in Korean, it is as if you must learn two different languages – the written language and the spoken language. A better way to put it is that there is “one language for writing and another for speaking.” You use different forms whether conversing or putting something on paper.

Some pronunciation is difficult. The the contrast between short and long vowels and consonants is particularly troublesome. Check out these minimal pairs:

sydämellä sydämmellä

jollekin jollekkin

A problem for the English speaker coming to Finnish would be the vocabulary, which is alien to the speaker of an IE language. Finnish language learners often find themselves looking up over half the words they encounter. Obviously, this slows down reading quite a bit!

In the grammar, the partitive case and potential tense can be difficult. Here is an example of how Finnish verb tenses combine with various cases to form words:

I A-Infinitive
Base form mennä

II E-Infinitive
Active inessive    mennessä
Active instructive mennen
Passive inessive   mentäessä

III MA-Infinitive
Inessive            menemässä
Elative             menemästä
Illative            menemään
Adessive            menemällä
Abessive            menemättä
Active instructive  menemän
Passive instructive mentämän

Verbs in Finnish

Finnish verbs are very regular. The irregular verbs can almost be counted on one hand:

juosta käydä olla nähdä tehdä

and a few others. In fact, on the plus side, Finnish in general is very regular.

One easy aspect of Finnish is the way you can build many forms from a base root:

kirj-

kirjabook kirjeletter kirjoittaato write kirjailijawriter

As in many Asian languages, there are no masculine or feminine pronouns, and there is no grammatical gender. The numeral system is quite simple compared to other languages. Finnish has a complete lack of consonant clusters. In addition, the phonology is fairly simple.

Finnish is rated 5, extremely hard to learn.

Southern

Estonian has similar difficulties as Finnish, since they are closely related. However, Estonian is more irregular than Finnish. In particular, the very regular agglutination system described in Finnish seems to have gone awry in Estonian. Estonian has 14 cases, including strange cases such as the abessive, adessive, elative and inessive. On the other hand, all of these cases can simply be analyzed as the genitive case plus a single unvarying suffix for each case. In addition, there is no gender, so the only things you have to worry about when forming cases are singular and plural.

Estonian has a strange mood form called the quotative, often translated as “reported speech.”

tema onhe/she/it is

tema olevatit’s rumored that he/she/it is or he/she/it is said to be

This mood is often used in newspaper reporting and is also used for gossip.

Estonian has an astounding 25 diphthongs. It also has three different varieties of vowel length, which is strange in the world’s languages. There are short, vowels and extra-long vowels and consonants.

linalinen – short n linnathe town’s – long n, written as nn `linnainto the town – extra-long n, not written out!

There are differences in the pronunciation of the three forms above, but in rapid speech, they are hard to hear, though native speakers can make them out. Difficulties are further compounded in that extra-long sonorants (m, n, ng, l, and r) and vowels and are not written out. All in all, phonemic length can be a problem in Estonian, and foreigners never seem to get it completely down.

Estonian pronunciation is not very difficult, though the õ sound can cause problems. However, Estonian has completely lost the vowel harmony system it inherited from Finnish, resulting in words that seem very hard to pronounce.

At least in written form, Estonian is not as complex as Finnish. Estonian can be seen as an abbreviated and modernized form of Finnish. The grammar is also like a simplified version of Finnish grammar and may be much easier to learn.

Estonian is rated 4.5, very to extremely difficult.

Sami Eastern

Skolt Sami‘s Latinization is often listed as one of the worst Latinizations around. The rest of the language is quite similar to, and as difficult as, Finnish.

Skolt Sami gets a 5 rating, extremely hard to learn.

Ugric Hungarian

It’s widely agreed that Hungarian is one of the hardest languages on Earth to learn. Even language professors agree. The British Diplomatic Corps did a study of the languages that its diplomats commonly had to learn and concluded that Hungarian was the hardest. Hungarian grammar is maddeningly complex, and Hungarian is often listed on craziest grammar lists. For one thing, there are many different forms for a single word via word modification. This enables the speaker to make his intended meaning very precise. Looking at nouns, there are about 257 different forms per noun.

Hungarian is said to have from 24-35 different cases (there are charts available showing 31 cases), but the actual number may only be 18. Nearly everything in Hungarian is inflected, similar to Lithuanian or Czech. Similar to Georgian and Basque, Hungarian has the polypersonal agreement, albeit to a lesser degree than those two languages. There are many irregularities in inflections, and even Hungarians have to learn how to spell all of these in school and have a hard time learning this.

The case distinctions alone can create many different words out of one base form. For the word house, we end up with 31 different words using case forms:

házbainto the house házbanin the house házból from [within] the house házraonto the house házonon the house házróloff [from] the house házhozto the house házíguntil/up to the house háználat the house háztól [away] from the house házzá – Translative case, where the house is the end product of a transformation, such as They turned the cave into a house. házkéntas the house, which could be used if you acted in your capacity as a house or disguised yourself as one. He dressed up as a house for Halloween. házértfor the house, specifically things done on its behalf or done to get the house. They spent a lot of time fixing things up (for the house). házul – Essive-modal case. Something like “house-ly” or in the way/manner of a house. The tent served as a house (in a house-ly fashion).

And we do have some basic cases:

ház – Nominative. The house is down the street. házat – Accusative. The ball hit the house. háznak – Dative. The man gave the house to Mary. házzal – Similar to instrumental, but more similar to English with. Refers to both instruments and companions.

The genitive takes 12 different declensions, depending on person and number:

házammy house házaimmy houses házadyour house házaidyour houses házahis/her/its house házai his/her/its houses házunk our house házainkour houses házatok your house házaitok your house házuk their house házaik their houses egyházchurch, as in the Catholic Church. (Literally one-house)

In addition, the genitive suffixes to the possession, which is not how the genitive works in IE.

emberman/person házhouse a(z)the

az ember házathe man’s house (Lit. the man house-his) a házammy house (Lit. the house-my) a házadyour house (Lit. the house-your)

There are also very long words such as this:

megszentségteleníthetetlenségeskedéseitekért… for your (you all possessive) repeated pretensions at being impossible to desecrate

Being an agglutinative language, that word is made up of many small parts of words, or morphemes. That word means something like

The preposition is stuck onto the word in this language, and this will seem strange to speakers of languages with free prepositions.

Hungarian is full of synonyms, similar to English.

For instance, there are 78 different words that mean to move: halad, jár, megy, dülöngél, lépdel, botorkál, kódorog, sétál , andalog, rohan, csörtet, üget, lohol, fut, átvág, vágtat, tipeg, libeg, biceg, poroszkál, vágtázik, somfordál , bóklászik, szedi a lábát, kitér, elszökken, betér , botladozik, őgyeleg, slattyog, bandukol, lófrál, szalad, vánszorog, kószál, kullog, baktat, koslat, kaptat, császkál, totyog, suhan, robog, rohan, kocog, cselleng, csatangol, beslisszol, elinal, elillan, bitangol, lopakodik, sompolyog, lapul, elkotródik, settenkedik, sündörög, eltérül, elódalog, kóborol, lézeng, ődöng, csavarog, lődörög, elvándorol , tekereg, kóvályog, ténfereg, özönlik, tódul, vonul, hömpölyög, ömlik, surran, oson, lépeget, mozog and mozgolódik .

Only about five of those terms are archaic and seldom used, the rest are in current use. However, to be a fair, a Hungarian native speaker might only recognize half of those words.

In addition, while most languages have names for countries that are pretty easy to figure out, in Hungarian even languages of nations are hard because they have changed the names so much. Italy becomes Olazorszag, Germany becomes Nemetzorsag, etc.

As in Russian and Serbo-Croatian, word order is relatively free in Hungarian. It is not completely free as some say but rather is it governed by a set of rules. The problem is that as you reorder the word order in a sentence, you say the same thing but the meaning changes slightly in terms of nuance. Further, there are quite a few dialects in Hungarian. Native speakers can pretty much understand them, but foreigners often have a lot of problems. Accent is very difficult in Hungarian due to the bewildering number of rules used to determine accent. In addition, there are exceptions to all of these rules. Nevertheless, Hungarian is probably more regular than Polish.

Hungarian spelling is also very strange for non-Hungarians, but at least the orthography is phonetic. Nevertheless, the orthography often makes it onto worst orthographies lists.

Hungarian phonetics is also strange. One of the problems with Hungarian phonetics is vowel harmony. Since you stick morphemes together to make a word, the vowels that you have used in the first part of the word will influence the vowels that you will use to make up the morphemes that occur later in the word. The vowel harmony gives Hungarian a “singing effect” when it is spoken. The ty, ny, sz, zs, dzs, dz, ly, cs and gy sounds are hard for many foreigners to make. The á, é, ó, ö, ő, ú, ü, ű, and í vowel sounds are not found in English.

Verbs are marked for object (indefinite, definite and person/number), subject (person and number) tense (past, present and future), mood (indicative, conditional and imperative), and aspect (frequency, potentiality, factitiveness, and reflexiveness.

Elmentegettethetnélek. I could make others save you occasionally (on a disk).

Verbs change depending on whether the object is definite or indefinite.

Olvasok könyvet. I read a book. (indefinite object)

Olvasom a könvyet. I read the book. (definite object)

As noted in the introduction to the Finno-Ugric section, you need to know quite a bit of Hungarian grammar to be able to express yourself on a basic level. For instance, in order to say:

I like your sister.

you will need to understand the following Hungarian forms:

  1. verb conjugation and definite or indefinite forms
  2. possessive suffixes
  3. case
  4. how to combine possessive suffixes with case
  5. word order
  6. explicit pronouns
  7. articles

It’s hard to say, but Hungarian is probably harder to learn than even the hardest Slavic languages like Czech, Serbo-Croatian and Polish. At any rate, it is generally agreed that Hungarian grammar is more complicated than Slavic grammar, which is pretty impressive as Slavic grammar is quite a beast.

Hungarian is rated 5, extremely hard to learn.

Sino-Tibetan Sinitic Chinese Mandarin

It’s fairly easy to learn to speak Mandarin at a basic level, though the tones can be tough. This is because the grammar is very simple – short words, no case, gender, verb inflections or tense. But with Japanese, you can keep learning, and with Chinese, you often tend to hit a wall, often because the syntactic structure is so strangely different from English (isolating).

Actually, the grammar is harder than it seems. At first it seems simple, like a simplified English. No word is capable of declension, and there is no tense, case, and number, nor are there articles. But the simplicity makes it difficult. No tense means there is no easy way to mark time in a sentence. Furthermore, tense is not as easy as it seems. Sure, there are no verb conjugations, but instead you must learn some particles and special word orders that are used to mark tense. Mandarin has 12 different adverbs for which there is no good English translation.

Once you start digging into Chinese, there is a complex layer under all the surface simplicity. There is such things as aspect, serial verbs, a complex classifier system, syntax marked by something called topic-prominence, a strange form called the detrimental passive, preposed relative clauses, use of verbs rather than adverbs to mark direction, and all sorts of strange stuff. Verb complements can be baffling, especially potential and directional complements. The 把, 是 and 的 constructions can be very hard to understand.

The topic-prominence is interesting in that only a few major languages have topic-comment syntax, and most of those are Oriental languages with a lot of Chinese borrowing. Topicalization is not marked morphologically.

There are sentences where the entire meaning changes with the addition of a single character. Chinese sentences are SVO (Subject -Verb – Object) at their base, but that is a bit of an illusion. A sentence that causes you to discuss time duration makes you repeat the verb after the direct object – SVOVT (T= time phrase). In the case of topicalization, sentences can have the structure of OSV (Object – Subject – Verb). Relative clauses and all subordinate clauses come before the noun they modify. In other words:

English: The man who always wore red walked into the room. Chinese: Who always wore red the man walked into the room.

The relative clause in the sentences above is marked in bold.

In Chinese, the prepositional phrase comes between the subject and the verb:

English: The man hit the ball into the yard. Chinese: The man into the yard hit the ball.

The prepositional phrase is bolded in the sentences above.

In Chinese, adjectives are actually stative verbs as in Nahuatl and Lakota.

那个热菜很好吃。 Nàgè rède cài hěnhǎochī. The it is hot food is good to eat. The hot food is delicious.

The symbol turns food hot into food it is hot, an attributive verb. means something like to be.

There are dozens of words called particles which shade the meaning of a sentence ever so slightly.

Chinese phonology is not as easy as some say. There are way too many instances of the zh, ch, sh, j, q, and x sounds in the language such that many of the words seem to sound the same. There is a distinction between aspirated and nonaspirated consonants. There is also the presence of odd retroflex consonants.

Chinese orthography is probably the most hardest orthography of any language. The alphabet uses symbols, so it’s not even a real alphabet. There are at least 85,000 symbols and actually many more, but you only need to know about 3-5,000 of them, and many Chinese don’t even know 1,000. To be highly proficient in Chinese, you need to know 10,000 characters, and probably less than

In addition, the characters have not been changed in 3,000 years, and the alphabet is at least somewhat phonetic, so we run into a serious problem of lack of a spelling reform.

The Communists tried to simplify the system (simplified Mandarin) but instead of making the connections between the phonetic aspects of character more sensible by decreasing their number and increasing their regularity (they did do this somewhat but not enough), they simply decreased the number of strokes needed for each symbol typically without dealing with the phonetic aspect of all. The simplification did not work well, so now you have a mixture of two different types of written Chinese – simplified and traditional.

In addition to all of this, Chinese borrowed a lot from the Japanese symbolic alphabet a full 1,000 years after it had already been developed and had not undergone a spelling reform, adding insult to injury.

Even leaving the characters aside, the stylistic and literary constraints required to write Chinese in an eloquent or formal (literary) manner would make your head swim. And just because you can read Chinese does not mean that you can read Classical Chinese prose. It’s as if it’s written in a different language – actually, it is technically a different language similar to Middle English or Old English. However, few Middle English or Old English texts are read anymore, and Classical Chinese is still widely read.

However, the orthography is at least consistent. 9

Writing the characters is even harder than reading them. One wrong dot or wrong line either completely changes the meaning or turns the symbol into nonsense.

It’s a real problem when you encounter a symbol you don’t know because there is no way to sound out the word. You are really and truly lost and screwed. There is a clue at the right side of the symbol, but it is not always accurate.You need to learn quite a bit of vocabulary just to speak simple sentences.

Similarly, a dictionary is not necessarily helpful when trying to read Chinese. You can have a Chinese sentence in front of you along with a dictionary, and the sentence still might not make sense even after looking it up in the dictionary.

Some Chinese Muslims write Chinese using an Arabic script. This is often considered to be one of the worst orthographies of all.

The tones are often quite difficult for a Westerner to pick up. If you mess up the tones, you have said a completely different word. Often foreigners who know their tones well nevertheless do not say them correctly, and hence, they say one word when they mean another. However, compared to other tone systems around the world, the tonal system in Chinese is comparatively easy.

A major problem with Chinese is homonyms. To some extent, this is true in many tonal languages. Since Chinese uses short words and is disyllabic, there is a limited repertoire of sounds that can be used. At a certain point, all of the sounds are used up, and you are into the realm of homophones.

Tonal distinctions are one way that monosyllabic and disyllabic languages attempt to deal with the homophone problem, but it’s not good enough, since Chinese still has many homophones, and meaning is often discerned by context, stress, rhythm and intonation. Chinese, like French and English, is heavily idiomatic.

It’s little known, but Chinese also uses different forms (classifiers) to count different things, like Japanese.

There is zero common vocabulary between English and Chinese, so you need to learn a whole new set of lexical forms.

In addition, nouns often show relatedness or hierarchy. For instance, in English, you can simply say my brother or my sister, but in Chinese, you cannot do this. You have to indicate whether you are speaking of an older or younger sibling.

mei meiyounger sister jie jieolder sister ge geolder brother di diyounger brother

Mandarin scored very high on a weirdest languages study.

On the positive side, Chinese grammar is fairly regular and word derivation, compound words are sensible and the meaning can be determined by looking at the word. In other languages, compound words are not necessarily so obvious.

Many agree that Chinese is the hardest to learn of all of the major languages. A recent survey of language professors rated Chinese as the hardest language on Earth to learn.

Mandarin gets a 5.5 rating for nearly hardest of all.

However, Cantonese is even harder to learn than Mandarin. Cantonese has eight tones to Mandarin’s four, and in addition, they continue to use a lot of the older traditional Chinese characters that were superseded when China moved to a simplified script in 1949. Furthermore, since non-Mandarin characters are not standardized, Cantonese cannot be written down as it is spoken.

In addition, Cantonese has verbal aspect, possibly up to 20 different varieties. Modal particles are difficult in Cantonese. Clusters of up to the 3 sentence final particles are very common. 我食咗飯 and 我食咗飯架啦喎 are both grammatical for I have had a meal, but the particles add the meaning of I have already had a meal, answering a question or even to imply I have had a meal, so I don’t need to eat anymore.

Cantonese gets a 5.5 rating, nearly hardest of all.

Min Nan is also said to be harder to learn than Mandarin, as it has a more complex tone system, with five tones on three different levels. Even many Taiwanese natives don’t seem to get it right these days, as it is falling out of favor, and many fewer children are being raised speaking it than before.

Min Nan gets a 5.5 rating, nearly hardest of all.

A recent 15 year survey out of Fudan University utilizing both the departments of Linguistics and Anthropology looked at 579 different languages in 91 linguistic families in order to try to find the most complicated language in the world. The result was that a Wu language dialect (or perhaps a separate language) in the Fengxian district of southern Shanghai (Dônđän Wu) was the most phonologically complex language of all, with 20 separate vowels (Wang 2012). The nearest competitor was Norwegian with 16 vowels.

Dônđän Wu gets a 5.5 rating, nearly hardest of all.

Classical Chinese is still read by many Chinese people and Chinese language learners. Unless you have a very good grasp on modern Chinese, classical Chinese will be completely wasted on you. Classical Chinese is much harder to read than reading modern Chinese.

Classical Chinese covers an era extending over 3,000 years, and to attain a reading fluency in this language, you need to be familiar with all of the characters used during this period along with all of the literature of the period so you can understand all the allusions. Even with a knowledge of Classical Chinese, you need to read it in context. If you are good at Classical Chinese and someone throws you a random section of it, it will take you a good amount of time to figure it out unless you know context.

The language is much more to the point than Modern Chinese, but this is not as good as it sounds. This simplicity leaves a room for ambiguity, and context plays an important role. A joke about some obscure historical or literary anecdote will be lost you unless you know what it refers to. For reading modern Chinese, you will need at least 5,000 characters, but even then, you will still need a dictionary. With Classical Chinese, there are no lower limits on the number of characters you need to know. The sky is the limit.

Classical Chinese gets a 6 rating, hardest of all.

Tibeto-Burman Qiangic Northern Qiang

In Quiang, a language of Sichuan Province in China, not only are there rhotic vowels, which are present in only

ʀuɑ +e˞ > ʀuɑ˞kʰ me + w ˞> mw

Rhotic vowels are found in US English – Unstressed ɚ: standard, dinner, Lincolnshire, editor, measure, martyr.

Qiang also has a very bad romanization, so bad that the Qiang will not even use it. Voiced consonants are written by adding a vowel to the symbol for the voiceless consonant. It has long and short vowels, but these are not represented in the system.

Qiang gets a 5 rating, extremely hard to learn.

Western Tibeto-Burman Bodish Central Bodish Central

Tibetan probably has one of the least rational orthographies of any language. The orthography has not changed in ~1,000 years while the language has gone through all sorts of changes. A langauge learner in Tibet can get by using phonetic spelling. The problem comes when you try to spell using the Classical Alphabet. For instance:

Srong rtsan Sgam po (written) soŋtsɛn ɡampo (spoken)

bsgrubs (written)

d`up (spoken)

While the orthography is etymological and completely outdated, it is quite predictable.

Tibetan gets a 5 rating, extremely hard to learn.

Southern

Dzongka, the official language of Bhutan, has some pretty wild phonology, in addition to having the Tibetan writing system, this time using Bhutanese forms of the Tibetan script.

It contrasts all of the following: s, , ʰs, ʰsʰ, ts, ʰts, tsʰ, z, ʱz, dz, ʱdz, ⁿsʰ, ᵐtsʰ, ⁿtsʰ, ⁿdz, ᵖts, ᵖtsʰ, ᵖtsʷʰ, and ᶲs, and in addition it has four tones, but there is no single word that is distinguished by tone only. On top of that, there are 22 different vowels.

Dzongka gets a 5 rating, extremely hard to learn.

Austroasiatic Mon-Khmer Vietic

Vietnamese is also hard to learn because to an outsider, the tones seem hard to tell apart. Therefore, foreigners often make themselves difficult to understand by not getting the tone precisely correct. It also has “creaky-voiced” tones, which are very hard for foreigners to get a grasp on.

Vietnamese grammar is fairly simple, and reading Vietnamese is pretty easy once you figure out the tone marks. Words are short as in Chinese. However, the simple grammar is relative, as you can have 25 or more forms just for I, the 1st person singular pronoun. In addition, the Latin orthography is said to be quite bad. It was invented by missionaries a few centuries ago, and it has never made much sense.

Vietnamese gets 5 rating, extremely hard to learn.

Mon-Khmer Khmer

Khmer has a reputation for being hard to learn. I understand that it has one of the most complex honorifics systems of any language on Earth. Over a dozen different words mean to carry depending on what one is carrying. There are several different words for slave depending on who owned the slave and what the slave did. There are 28-30 different vowels, including sets of long and short vowels and long and short diphthongs. The vowel system is so complicated that there isn’t even agreement on exactly what it looks like. Khmer learners, especially speakers of IE languages, often have a hard time producing or even distinguishing these vowels.

Speaking it is not so bad, but reading and writing it is pretty difficult. For instance, you can put up to five different symbols together in one complex symbol. The orthographic script is even worse than the Thai one. There are actually rules to this mess, but no one seems to know who they are.

Khmer gets a  4.5 rating, very to extremely hard.

Bahnaric North Bahnaric West Sedang-Todrah Sedang

Sedang, a language of Vietnam, has the highest number of vowel sounds of any language on Earth, at 55 distinct vowel sounds.

Sedang gets a 5 rating, extremely hard to learn.

Hmong-Mien Hmongic Chuanqiandian

Hmong is widely spoken in this part of California, but it’s not easy to learn. There are eight tones, and they are not easy to figure out. It’s not obviously related to any other major language but the obscure Mien.

It has some very strange consonants called voiceless nasals. We have them in English as allophones – the m in small is voiceless, but in Hmong, they put them at the front of words – the m in the word Hmong is voiceless. These can be very hard to pronounce.

The romanization is widely criticized for being a lousy one, but the Hmong use it anyway.

Hmong gets a 5 rating, extremely hard to learn.

Austro-Tai Austronesian Tsouic

Tsou is a Taiwanese aborigine language spoken by about 2,000 people in Taiwan. It has the odd feature whereby the underlying glides y and w turn into or surface as non-syllabic mid vowels e̯ and o̯ in certain contexts:

jo~joskɨ -> e̯oˈe̯oskɨ  -= fishes

Tsou is also ergative like most Formosan languages. Tsou is the only language in the world that has no prepositions or anything that looks like a preposition. Instead it uses nouns and verbs in the place of prepositions. Tsou allows more potential consonant clusters than most other languages. About 1/2 of all possible CC clusters are allowed.

Tsou has an inclusive/exclusive distinction in the 1st person plural and a very strange visible and non-visible distinction in the 3rd person singular and plural. Both adjectives and adverbs can turn into verbs and are marked for voice in the same way that verbs are. Verbs are extensively marked for voice. Nouns are marked for a variety of odd cases, often referring to perception, (visible/invisible) person, and place deixis.

‘e –               visible and near speaker si/ta –           visible and near hearer ta –               visible but away from speaker ‘o/to –           invisible and far away, or newly introduced to discourse na/no ~ ne – non-identifiable and non-referential (often when scanning a class of elements)

Tsou gets a 5 rating, extremely hard to learn.

Malayo-Polynesian Malayo-Chamic Malayic Malay

Bahasa Indonesia is an easy language to learn. For one thing, the grammar is dead simple. There are only a handful of prefixes, only two of which might be seen as inflectional. There are also several suffixes. Verbs are not marked for tense at all. And the sound system of these languages, in common with Austronesian in general, is one of the simplest on Earth, with only two dozen phonemes. Bahasa Indonesia has few homonyms, homophones, homographs, or heteronyms. Words in general have only one meaning.

Though the orthography is not completely phonetic, it only has a small number of nonphonetic exceptions. The orthography is one of the easiest on Earth to use.

The system for converting words into either nouns or verbs is regular. To make a plural, you simply repeat a word, so instead of saying pencils, you say pencil pencil.

Bahasa Indonesia gets a 1.5 rating, extremely easy to learn.

Malay is only easy if you learn the standard spoken form or one of the creoles. Learning the literary language is quite a bit more difficult. However, the Jawi script, which is Malay written in Arabic script, is often considered to be perfectly awful.

Malay get a 2 rating for moderately easy.

Philippine Greater Central Philippine Central Philippine Tagalog

However, Tagalog is much harder than Malay or Indonesian. Compared to many European languages, Tagalog syntax, morphology and semantics are often quite different. Also, Tagalog is typically spoken very fast. Unlike Malay, verbs conjugate quite a bit in Tagalog. The main idea of Tagalog grammar is something called focus. Once you figure that out, the language gets pretty easy, but until you understand that concept, you are going to have a hard time.

Everything is affixed in Tagalog.

However, articles and creation of adjectives from nouns is very easy.

Compare:

gandabeauty (noun) magandabeautiful (adjective)

Tagalog gets a 4 rating, very difficult.

Central-Eastern Malayo-Polynesian Eastern Malayo-Polynesian Oceanic Central-Eastern Oceanic Remote Oceanic Central Pacific East Fijian-Polynesian Polynesian Nuclear East Central Tahitic

Maori and other Polynesian languages have a reputation for being quite easy to learn. The main problem for English speakers is that the sentence structure is backwards compared to English. In addition, macrons can cause problems.

One problem with Maori is dialects. The dialects are so diverse that this means that there are multiple words for the same thing. Swiss German has a similar issue, with up to 50 words for each common household item (nearly every major dialect has its own word for common objects):

ngongi, noni, koki, waiwater whiri, rarangi, hiri –  to plait, to twist, to weave pai, maitaigood tu, , tutehu, mātikato stand mau, mouto hold pau, pouto be exhausted ika, tohorāwhale ika, ngohifish kāwei, kāwailine ori, kori, keukeu, koukou, neke, nukuto move haere, hara, here, horo, whanoto go, to come hara, hapa, to be wrong kōrerorero, wānanga, rūnangato discuss tohunga, tahungapriest matikuku, maikukufinger nail kanohi, konohi, mata, whatu, kamo, karueye, face

Entire Maori sentences can be written with vowels only.

E uu aau? Are yours firm?

I uaa ai. It rained as usual.

I ui au ‘i auau aau?’ E uaua! It will be difficult/hard/heavy!

On the plus side, the pronunciation is simple, and there is no gender. The language is as regular as Japanese. No Polynesian language has more than 16 sounds, and they all lack tones. They all have five vowels, which can be either long or short. A consonant must be followed by a vowel, so there are no consonant clusters. All consonants are easy to pronounce.

Maori gets a 3 rating, average difficulty.

Marquesic

Hawaiian is a pretty easy language to learn. It is easy to pronounce, has a simple alphabet, lacks complex morphology and has a fairly simple syntax.

Hawaiian gets a 2 rating, very easy to learn.

North and Central Vanuatu East Santo North

Sakao is a very strange langauge spoken by 4,000 people in Vanuatu.  It is very strange. It is a polysynthetic Austronesian language, which is very weird. It allows extreme consonant clusters. Sakao has an incredible seven degrees of deixis. The language has an amazing four persons: singular, dual, paucal and plural. The neighboring language Tomoko has singular, dual, trial and plural. The trial form is very odd. Sakao’s paucal derived from Tomato’s trial:

jørðœl they, from three to ten

jørðœl løn the five of them (Literally, they three, five)

All nouns are always in the singular except for kinship forms and demonstratives, which only display the plural:

ðjœɣmy mother/aunt -> rðjœɣmy aunts

walðyɣmy child -> raalðyɣmy children

It has a number of nouns that are said to be “inalienably possessed”, that is, whenever they occur, they must be possessed by some possessor. These often take highly irregular inflections:

Sakao 	  English
œsɨŋœ-ɣ   my mouth
œsɨŋœ-m   thy mouth
ɔsɨŋɔ-n   his/her/its mouth
œsœŋ-...  ...'s mouth	

uly-ɣ 	  my hair
uly-m 	  thy hair
ulœ-n 	  his/her/its hair
nøl-...   ...'s hair

Here, mouth is either œsɨŋœ-, ɔsɨŋɔ- or œsœŋ-, and hair is either uly-, ulœ- or nøl-

Sakao, strangely enough, may not even have syllables in the way that we normally think of them. If it does have syllables at all, they would appear to be at least a vowel optionally  surrounded by any number of consonants.

i (V) thou Mhɛrtpr. (CCVCCCC) Having sung and stopped singing thou kept silent.

Sakao has a suffix -in that makes an intransitive verb transitive and makes a transitive verb ditransitive. Ditransitive verbs can take two arguments – a direct object and an instrumental.

Mɨjilɨn amas ara./Mɨjilɨn ara amas. He kills the pig with the club/He kills with the club the pig.

Sakao polysynthesis allows compound verbs, each one having its own instrument or object:

Mɔssɔnɛshɔβrɨn aða ɛðɛ. He-shooting-fish-kept-on-walking with-a-bow the-sea. He walked along the sea shooting the fish with a bow.

Sakao gets a 5 rating, extremely hard to learn.

Central-Eastern Oceanic Southeast Solomonic Malaita–San Cristobal Malaita Northern Malaita

Kwaio is an Austronesian language spoken in the Solomon Islands. It has four different forms of number to mark pronouns – not only the usual singular and plural, but also the rarer dual and the very rare paucal. In addition, there is an inclusive/exclusive contrast in the non-singular forms.

For instance:

1 dual inclusive (you and I) 1 dual exclusive (I and someone else, not you)

1 paucal inclusive (you, I and a few others) 1 paucal exclusive (I and a few others)

1 plural inclusive (I, you and many others) 1 plural exclusive (I and many others)

Pretty wild!

Kwaio gets a 5 rating, extremely hard to learn.

Greater Barito East Barito Malagasy

Malagasy, the official language of Madagascar, has a reputation for being even easier to learn than Indonesian or Malay.

Malagasy gets a 1 rating, easiest of all to learn.

Tai-Kadai Kam-Tai Tai Southwestern

Thai is a pretty hard language to learn. There are 75 symbols in the strange script, there are no spaces between words in the script, and vowels can come before, after, above or below consonants in any given syllable. There seem to be many different glyphs for every consonant, but the different glyphs for the same consonant will sometimes change the sound of the neighboring vowel. The orthography is as insensible as that of English since centuries have gone by with no spelling reforms, in fact, Thai has not changed its system in 1000 years. The wild card of having tone thrown in adds to the insanity.

Consonant pronunciations vary depending on the location of the syllable in the word – for instance, s can change to t. There are many vowels which are spoken but not written. There are many consonants that are pronounced the same – for instance, there are six different t‘s, not counting the s‘s that turn into t‘s. The Thai script is definitely one of the most difficult phonetic scripts. Nevertheless, the Thai script is easier to learn than the Japanese or Chinese character sets. In spite of all of that, the syntax is simple, like Chinese.

There are five tones, including a neutral tone. Tones are determined by a variety of complex things, including a combination of tone marks, the class of consonants, if the syllable ends in a sonorant or a stop and what the tone of the preceding syllable was. Tone marking in the orthography is quite complex.

The vowels are different than in many languages, and there are some unusual diphthongs: eua, euai, aui and uu. There is a contrast between aspirated and unaspirated consonants.

There is a system of noun classifiers for counting various things, similar to Japanese. In addition, common to many Asian languages, there is a complicated honorifics system.

On the plus side, Thai is a regular language, with few exceptions to the rules. However, the rules are quite complex. The syntax is about as complex as that of Chinese, and the grammar is dead simple.

Thai gets a 5 rating, hardest of all to learn.

Lao is very similar to Thai, in fact it is identical to a Thai language spoken by 16 million people in northeast Thailand called Northeastern Thai. The Lao script is similar to Thai, but it has fewer letters so there is somewhat less confusion.

Lao gets a 4.5 rating, very to extremely hard to learn.

Kam-Sui

The Kam languages of the Dong people in southwest China were rated by the Fudan University study referenced above under Wu as the 2nd most phonologically complex on Earth (Wang 2012). There are 32 stem initial consonants, including oddities like , tɕʰ, , pʲʰ, ɕ, , kʷʰ, ŋʷ, tʃʰ, tsʰ. Note the many contrasts between aspirated and unaspirated voiceless consonants, including bilabial palatalized stops, labialized velar stops, and alveolar affricates. There are an incredible 64 different syllable finals, and 14 others that occur only in Chinese loans.

There are an astounding 15 different tones, nine in open syllables and six in checked syllables (entering tones). Main tones are high, high rising, high falling, low, low rising, low falling, mid, dipping and peaking. When they speak, it sounds as if they are singing.

Kam gets a 5 rating, extremely hard to learn.

Kra Paha

According to the Fudan University study quoted above, Buyang in the 3rd most phonologically complex language in the world. Buyang is a cluster of 4 related languages spoken by 1,900 people in Yunnan Province, China. Buyang has a completely wild consonant inventory.

It has a full set of both voiced and voiceless plain and aspirated stops, including voiceless uvulars. The contrast between aspirated and plain voiced stops is peculiar. The stop series also has distinctions between palatalized and rounded stops throughout the series. It has a labialized voiceless palatal fricative and a voiceless dental aspirated lateral, unusual sounds. It has four different voiceless aspirated nasals. It has voiceless y and w, more odd sounds. It also has plain and labialized palatal glides.

That is one heck of a wild phonology.

Buyang gets a 5 rating, extremely hard to learn.

Niger-Kordofanian Niger-Congo Atlantic–Congo Kwa Nyo Ga-Dangme

The African Bantu language Ga has a bad reputation for being a tough nut to crack. It is spoken in Ghana by about 600,000 people. It has two tones and engages in a strange behavior called tone terracing that is common to many West African languages. There is a phonemic distinction between three different types of vowel length. All vowels have 3 different lengths – short, long and extra long. It also has many sounds that are not in any Western languages.

Ga gets a 5 rating, extremely hard to learn.

Potou-Tano Tano Central Bia Northern

Anyi is a language spoken by 610,000 people in Côte d’Ivoire.  It is relatively straightforward as far as African languages go. Probably the hardest part about the language is that it is tonal, and it does have two tones. The phonology does have the unusual +-ATR contrast which will seem very odd. ATR stands for advanced tongue root, so the language has a contrast between vowels with an advanced tongue root and without one. However, the grammar is pretty regular. There are few confusing phonological processes.

Anyi has a simple tense system, with only present, past and future. There is no aspect, mood or voice marking, and it lacks the noun class systems so common in many African languages. It has a plural marker, but it is often optional.

The syntax does have serial verbs, which will seem odd to Westerners. It distinguishes between relative clauses marked with and subordinate clauses marked with .

Anyi gets a 4 rating, very hard to learn.

Volta-Congo Benue-Congo Bantoid Southern Narrow Bantu Central M Nyika-Safwa

Ndali is a Bantu language with 150,000 speakers spoken in Malawi and Tanzania. It has many strange tense forms. For instance, in the past tense:

Past tense A: He went just now. Past tense B: He went sometime earlier today. Past tense C: He went yesterday. Past tense D: He went sometime before yesterday.

Future tense is marked similarly:

Future tense A: He’s going to go right away. Future tense B: He’s going to go sometime later today. Future tense C: He’s going to go tomorrow. Future tense D: He’s going to go sometime after tomorrow.

Ndali gets a 5 rating, extremely hard to learn.

S Nguni

Xhosa, a language of South Africa, is quite difficult, with up to nine click sounds. Clicks only exist in one language outside of Africa – the Australian language Damin – and are extremely difficult to learn. Even native speakers mess up the clicks sometimes. Nelson Mandela said he had problems making some of the click sounds in Xhosa. The phonemics in general of Xhosa are pretty wild.

Xhosa gets a 5 rating, extremely hard to learn.

Zulu and Ndebele also have these impossible click sounds. However, outside of click sounds, the phonology of Nguni languages is straightforward. All Nguni languages are agglutinative. These languages also make plurals by changing the prefix of the noun, and the manner varies according the noun class. If you want to look up a word in the dictionary, first of all you need to discard the prefix. For instance, in Ndebele,

riverumfula riversimifula, but

stoneilitshe stones –  amatsheyet

treeisihlahla trees izihlahla

Ndebele gets a 5 rating, hardest of all.

Zulu has pitch accent, tones and clicks. There are nine different pitch accents, four tones and three clicks, but each click can be pronounced in five different ways. However, tones are not marked in writing, so it’s hard to figure out when to use them. Zulu also has depressor consonants, which lower the tone in the vowel in the following syllable. In addition, Zulu has multiple gender – 15 different genders. And some nouns behave like verbs. It also has 12 different noun classes, but 9

Zulu gets a 5 rating, extremely hard to learn.

G Swahili

For unknown reasons, Swahili is generally considered to be an easy language to learn. The US military ranks it 1, with the easiest of all languages to learn. This seems to be the typical perception. Why Swahili is so easy to learn, I am not sure. It’s a trade language, and trade languages are often fairly easy to learn. There’s also a lot of controversy about whether or not Swahili can be considered a creole, but that has not been proven. For the moment, the reasons why Swahili is so easy to learn will have to remain mysterious.

On the down side, Swahili has many noun classes, but they have the benefit of being more or less logical.

Swahili gets a 2 rating, moderately easy.

Khoisan Southern Africa Southern Hua

!Xóõ (Taa), spoken by only 4,200 Bushmen in Botswana and Namibia, is a notoriously difficult Khoisan language replete with the notoriously impossible to comprehend click sounds. Taa has anywhere from 130 to 164 consonants, the largest phonemic inventory of any language. Of this vast wealth of sounds, there are anywhere from 30-64 different click sounds. There are five basic clicks and 17 accompanying ones. Speakers develop a lump on their larynx from making the click sounds.

In addition, there are four types of vowels: plain, pharyngealized, breathy-voiced and strident. On top of that, there are four tones. Taa appears on many lists of the wildest phonologies and craziest languages period on Earth.

Taa gets a 5 rating, extremely hard to learn.

Northern

Ju|’hoan, a Khoisan language spoken by 5,000 people in Botswana, has one of the study of the weirdest languages on Earth.

Ju|’hoan gets a 5 rating, extremely hard to learn.

Eskimo-Aleut Eskimo Inuit-Inupiaq

Inuktitut is extremely hard to learn. Inuktitut is polysynthetic-agglutinative, and roots can take many suffixes, in some cases up to 700. Verbs have 63 forms of the present indicative, and conjugation involves 252 different inflections. Inuktitut has the complicated polypersonal agreement system discussed under Georgian above and Basque below. In a typical long Inuktitut text, 9

Inuktituusuungutsialaarungnanngittuaraaluuvunga. I truly don’t know how to speak Inuktitut very well.

You may need to analyze up to 10 different bits of information in order to figure out a single word. However, the affixation is all via suffixes (there are no prefixes or infixes) and the suffixation is extremely regular.

Inuktitut is also rated one by linguists one of the hardest languages on Earth to pronounce. Inuktitut may be as hard to learn as Navajo.

Inuktitut is rated 6, hardest of all.

Kalaallisut (Western Greenlandic) is very closely related to Inuktitut. Look at this sentence:

Aliikusersuillammassuaanerartassagaluarpaalli… However, they will say that he is a great entertainer, but …

That word is composed of 12 separate morphemes. A single word can conceptualize what could be an entire sentence in a non-polysynthetic language.

Kalaallisut is rated 6, hardest of all.

Chukotko-Kamchatkan Northern Chukot

Chukchi is a polysynthetic, agglutinating and incorporating language and is often listed as one of the hardest languages on Earth to learn.

Təmeyŋəlevtpəγtərkən. I have a fierce headache.

There are five morphemes in that word, and there are three lexical morphemes (nouns or adjectives) incorporated in that word: meyŋgreat, levthead, and pəγtache.

Chukchi gets a 6 rating, hardest of all.

Basque

Basque, of course, is just a wild language altogether. There is an old saying that the Devil tried to learn Basque, but after seven years, he only learned how to say Hello and Goodbye. Many Basques, including some of the most ardent Basque nationalists, tried to learn Basque as adults. Some of them succeeded, but a very large number of them failed. Based on the number that failed, it does seem that Basque is harder for an adult to learn as an L2 than many other languages are. Basque grammar is maddeningly complex and it often makes it onto craziest grammars and craziest language lists.

There are 11 cases, and each one takes four different forms. The verbs are quite complex. This is because it is an ergative language, so verbs vary according to the number of subjects and the number of objects and if any third person is involved.

This is the same polypersonal agreement system that Georgian has above. Basque’s polypersonal system is a polysynthetic system consisting of two verb types – synthetic and analytical. Only a few verbs use the synthetic form.

Three of Basque’s cases – the absolutive (intransitive verb case), the ergative (intransitive verb case) and the dative – can be marked via affixes to the verb. In Basque, only present simple and past simple synthetic tenses take polypersonal affixes.

The analytical forms are composed of more than one word, while the synthetic forms are all one word. The analytic verbs are built via the synthetic verbs izanbe, ukanhave and egindo.

Synthetic:

d-akar-ki-o-gu = We bring it to him/her. The verb is ekarribring. z-erama-zki-gu-te-n = They took them to us. The verb is eramantake

Analytic:

Ekarriko d-i-o-gu = We’ll bring it to him/her. Literally: We will have-bring it to him/her. The analytic verb is built from ukanhave.

Eraman d-ieza-zki-gu-ke-te = They can take them to us. Literally: They can be taking them to us. The analytic verb is built from izanbe.

Most of the analytic verbs require an auxiliary which carries all sorts of information that is often carried on verbs in other languages – tense, mood, sometimes gender and person for subject, object and indirect object.

Jaten naiz. Eat I-am-doing. I am eating.

Jaten nintekeen. Eat I-was-able-to. I could eat.

Eman geniezazkiake. Give we-might-have-them-to-you-male. We might have given them to you.

In the above, naiz, nintekeen and geniezazkiake are auxiliaries. There are actually 2,640 different forms of these auxiliaries!

A language with ergative morphosyntax in Europe is quite a strange thing, and Basque is the only one of its kind. The ergative itself is quite unusual:

Gizona etorri da.The man has arrived. Gizonak mutila ikusi du.The man saw the boy.

gizonman mutilboy -a = the

The noun gizon takes a different form whether it is the subject of a transitive or intransitive verb. The first sentence is in absolutive case (unmarked) while the second sentence is in the ergative case (marked by the morpheme -k). If you come from a non-ergative IE language, the concept of ergativity itself is difficult enough to conceptualize, much less trying to actually learn an ergative language. Consequently, any ergative language will automatically be more difficult than a non-ergative one for all speakers of IE languages.

Ergativity also works with pronouns.  There are four basic systems:

Nor:           verb has subject only
Nor-Nork:          "    subj. + direct complement
Nor-Nori:          "    subj. + indirect comp.
Nor-Nori-Nork:     "    subj. + indir. + dir. comps.

Some call Basque the most consistently ergative language on Earth.

If you don’t grow up speaking Basque, it’s hard to attain native speaker competence. It’s quite a bit easier to write in Basque than to speak it.

Nevertheless, Basque verbs are quite regular. There are only a few irregularities in conjugations and they have phonetic explanations. In fact, the entire language is quite regular. In addition, most words above the intermediate level are borrowings from large languages, so once you reach intermediate Basque, the rest is not that hard. In addition, pronunciation is straightforward.

Basque is rated 5.5, nearly hardest of all.

References

Dorani, Yakir. Hebrew speaker, Israel. August 2013. Personal communication.

Hewitt, B. G.. 2005. Georgian: A Learner’s Grammar, p. 29.

Kim, Yuni. December 16, 2003. Vowel Elision and the Morphophonology of Dominance in Aymara. UC Berkeley.

Kirk, John William Carnegie. 1905. A Grammar of the Somali Language: With Examples in Prose and Verse and an Account of the Yibir and Midgan Dialects, pp. 73-74.

Rogers, Jean H. 1978. Differential Focusing in Ojibwa Conjunct Verbs: On Circumstances, Participants, and Events. International Journal of American Linguistics 44: 167-179.

Wang, Chuan-Chao et al. 2012. Comment on ”Phonemic Diversity Supports a Serial Founder Effect Model of Language Expansion from Africa.” Science 335:657.

This research takes a lot of time, and I do not get paid anything for it. If you think this website is valuable to you, please consider a a contribution to support more of this valuable research.

error

Enjoy this blog? Please spread the word :)