Question regarding chimpanzees...

[Srelex]

I was thinking about the use of sign language by chimpanzees and gorillas, and got thinking--if it was somehow made possible to grant chimpanzees the same vocal cord capability as humans, would it be possible for them to be taught to speak languages, to the point of even having conversations, of whatever level of complexity?

[madd0ct0r]

yes. exactly the way they use sign language.

[LaCroix]

And at the same, low level. Speech is quite a complex task, and it isn't as if they use sign language to chat about modern philosophy. They could form short sentences, but they'll probably speak like you'd expect from a cave man.

[madd0ct0r]

or a small child.

[LaCroix]

I would almost expect it to be worse. A small child's brain is, after all, at least human. A chimp brain hasn't evolved that far.
Also, complicated sound patterns of some words might be physically impossible due to physical differences in facial muscles and vocal cords.

edit: A small child usually uses less words, but usually can use them quite well. From a chimp, I expect long pauses, slow speech, and some varying degree of speech impediment due to less control.

[Ziggy Stardust]

You would need to do more than just change the vocal chords. The articulatory muscles of the throat and mouth, the size and shape of the air ducts, the strength and positioning of the diaphragm, etc. Hell, just because they are not strictly bipedal vocal communication will be radically different.

Certainly, this hypothetical chimp would be able to master language at least as much as Alex, the African Gray. But I find it pretty doubtful that you could achieve complex communication on the level of normal human-human interaction, or you could get it to understand and express very complex ideas (for example, "I would have had to do something;" it is trivial for a human to parse this kind of thought, but it is not clear yet whether animals are capable of this level of abstraction).

[Akkleptos]

It's been tried already. Rather unsuccessfully, I might add.

See here

Many years ago, attempts were made to teach chimpanzees spoken language. Hayes and Hayes (1952) gave Vicki, their young chimpanzee, intensive training in language, but she only learnt to produce four sounds and unfortunately none of them was used in a language-like way.

The Kelloggs (1968) raised a young chimpanzee called Gua together with their son Donald, but unlike the child the ape never learnt to speak.

It is now known that these projects were doomed from the beginning, in so far as that the chimpanzee's vocal tract is simply not adapted to produce human speech or anything like it.

And, yes, chimpanzees most likely lack the comprehension of abstract concepts such as verb tenses, conditionals and possibilities that abound in human language. We're probably better off trying with dolphins, AFAIK.

[Tolya]

--edit: nevermind, mixed up the science programs and names, got to do the research right for this post.

[Alyrium Denryle]

You would need to do more than just change the vocal chords. The articulatory muscles of the throat and mouth, the size and shape of the air ducts, the strength and positioning of the diaphragm, etc. Hell, just because they are not strictly bipedal vocal communication will be radically different.

Certainly, this hypothetical chimp would be able to master language at least as much as Alex, the African Gray. But I find it pretty doubtful that you could achieve complex communication on the level of normal human-human interaction, or you could get it to understand and express very complex ideas (for example, "I would have had to do something;" it is trivial for a human to parse this kind of thought, but it is not clear yet whether animals are capable of this level of abstraction).

Yeah, the subjunctive case is not something most species are capable of. Even if they have "imagination" and the ability to deal with future scenarios in that manner, it is not something that is easy to wrap the mind around linguistically. Future tense and certainly past might be possible with some species, but the subjunctive "wishful thinking" talking about possibilities that have been negated by something happening in the present ("I would have had to do X") is probably not a thing with most species.

[Ziggy Stardust]

Yeah, the subjunctive case is not something most species are capable of. Even if they have "imagination" and the ability to deal with future scenarios in that manner, it is not something that is easy to wrap the mind around linguistically. Future tense and certainly past might be possible with some species, but the subjunctive "wishful thinking" talking about possibilities that have been negated by something happening in the present ("I would have had to do X") is probably not a thing with most species.

Hell, depending on the context, *I* sometimes have a hard time figuring out what the fuck people are trying to say in those instances.

[Setesh]

You would need to do more than just change the vocal chords. The articulatory muscles of the throat and mouth, the size and shape of the air ducts, the strength and positioning of the diaphragm, etc. Hell, just because they are not strictly bipedal vocal communication will be radically different.

Certainly, this hypothetical chimp would be able to master language at least as much as Alex, the African Gray. But I find it pretty doubtful that you could achieve complex communication on the level of normal human-human interaction, or you could get it to understand and express very complex ideas (for example, "I would have had to do something;" it is trivial for a human to parse this kind of thought, but it is not clear yet whether animals are capable of this level of abstraction).

Yeah, the subjunctive case is not something most species are capable of. Even if they have "imagination" and the ability to deal with future scenarios in that manner, it is not something that is easy to wrap the mind around linguistically. Future tense and certainly past might be possible with some species, but the subjunctive "wishful thinking" talking about possibilities that have been negated by something happening in the present ("I would have had to do X") is probably not a thing with most species.

The only tested and proven case I can think of would be the bonobos at the Great Ape Trust. Kanzi especially as he was the first ape to have ever learned lexagram language naturally rather than through direct training. They had to add hundreds of new lexagrams over the years to allow him to express more complex ideas than any of the older trained bonobos ever attempted showing that among great apes at least this kind of thought is possible. Granted among apes Kanzi is a genius, even learning some sign language from watching Koko the gorilla on TV, but signs point to it being a matter of learning the language in the correct manner. Like growing up bilingual rather then learning a language in highschool, sure if you have the talent you can do it, but a raised bilingual will have an advantage.

Recently they discovered kanzi is attempting to say the words, every word in fact, but do to the differences between his vocal structure and human it comes out distorted and a lot of it beyond our hearing range. They have experimented with this by having him say certain things and his sister Panbanisha reacts appropriately. Based on this they think if he could talk it would be like talking to an over-exited eight year old most of the time, with moments of more mature behavior.

[Ziggy Stardust]

Honestly, I would take a lot of the Kanzi stuff with a grain of salt. While there certainly is some interesting work with Kanzi, you can't trust the results to be entirely empirical. Sue Savage-Rumbaugh, who did most of the work with Kanzi, had very sloppy methodology, and you can't rule out the Clever Hans effect. The famous story about Savage-Rumbaugh and Kanzi that should immediately raise eyebrows is retold here. The relevant part:

I’d told Savage-Rumbaugh about some of my adventures, and she invited me to perform a Maori war dance. I beat my chest, slapped my thighs and hollered. The bonobos sat quiet and motionless for a few seconds, then all but Kanzi snapped into a frenzy, the noise deafening as they screamed, bared their teeth and pounded on the walls and floor of their enclosure. Still calm, Kanzi waved an arm at Savage-Rumbaugh, as if asking her to come closer, then let loose with a stream of squeaks and squeals."Kanzi says he knows you're not threatening them," Savage-Rumbaugh said to me," and he'd like you to do it again just for him, in a room out back, so the others won't get upset.”

Essentially, Savage-Rumbaugh was the only one who could "translate" what Kanzi was "saying," and she also had a vested personal/professional interest in the results. Really, I don't trust a lot of the stuff with Kanzi at all.

The best example of animal language learning is Alex the African Gray Parrot, but even in that case it is hard to discern how much of what the animal learned was about the fundamentals of language and language structure, and how much was anthropomorphizing behavior on the part of the investigators themselves. It is an incredibly difficult thing to test for, in all honesty.

[Parricidium]

A chimp brain hasn't evolved that far.

Please don't put evolution on a linear scale, that's not how it works. You and I are the result of nearly 4 billion years of evolutionary success. So is a cockroach.

[Memnon]

A chimp brain hasn't evolved that far.

Please don't put evolution on a linear scale, that's not how it works. You and I are the result of nearly 4 billion years of evolutionary success. So is a cockroach.

Sure it is, if you're just measuring the single variable of language skills. Seems pretty reasonable to me that you could generate a robust language number.

[Ziggy Stardust]

Sure it is, if you're just measuring the single variable of language skills.

But language by definition is not a single variable. It is a complex interaction of biomechanics, sound processing, memory, and other higher cognitive functions. Using it as a linear variable to judge relative evolutionary sophistication is just as pointless as using "smart" or some other broad concept. What we conceive as language is a complex form of communication. However, bees also have very intricate forms of communication, using chemical cues and the famous "dances." How can you linearly compare the two?

Seems pretty reasonable to me that you could generate a robust language number.

Not entirely sure what you mean by this.

[Memnon]

Sure it is, if you're just measuring the single variable of language skills.

But language by definition is not a single variable. It is a complex interaction of biomechanics, sound processing, memory, and other higher cognitive functions. Using it as a linear variable to judge relative evolutionary sophistication is just as pointless as using "smart" or some other broad concept. What we conceive as language is a complex form of communication. However, bees also have very intricate forms of communication, using chemical cues and the famous "dances." How can you linearly compare the two?

Seems pretty reasonable to me that you could generate a robust language number.

Not entirely sure what you mean by this.

Using humans in different stages of development as a model, it seems reasonable to me that you could create a variable that maps to typical human language development as your single axis. It might not work too well outside of mammals, but it would be particularly relevant for our nearer relations. Of course language is complex, but that certainly doesn't mean you can't abstract it if you want to.

Of course, you can throw all of that linearization out the window and just look at where the other species lies on your whatever-dimensional graph of language variables. Of course, in doing so you'd be drawing a straight line.

The point is that everyone who hasn't flunked high school biology knows that of course all creatures have evolved for the same period of time, since as far as anyone can tell we all have the same LUCA. That doesn't mean that the shorthand of 'evolved as much' isn't still meaningful, even though it doesn't communicate it very well. Clearly we're talking about comparing within a couple of million years' evolutionary distance, so I think it's very reasonable to say that chimps are much closer to 'zero' on some kind of language variable or graph.

[LaCroix]

A chimp brain hasn't evolved that far.

Please don't put evolution on a linear scale, that's not how it works. You and I are the result of nearly 4 billion years of evolutionary success. So is a cockroach.

Just stumbled back into that discussion...

Seriously? You took that quote, which comes out of the discussion of "complex vocal language like humans use it needs a lot of brain power, which unlike a small child, a chimp doesn't have" and extrapolated that I meant that all evolution is happening on a line with humankind at the end?

I never even said that it hasn't evolved far enough in the human direction, just not far enough!

We were talking about the necessities to give speech to a chimp. And apart from several adaptions on throat, vocal chords, jaw, tongue and other facial muscles, the biggest problem is that a chimp does have a much smaller brain.

For example, a new-born human child has a brain the size of an adult chimp (~330-370cc, and within the first year, it brain grow to almost 1000cc - almost triple the size of a chimp. Coincidentally, a child needs at least half a year to one year to start recognizing and using speech - and for the next two years, it still is staggering with the sheer complexity of speech. Usually, a child is limited to less than hundred words until the age of 2, at which age it starts forming sentences for the first time...

Chimps simply don't have brain power to compute speech at more than a year old human's level - which is astounding, already, in my opinion giving the difference in size. Still, a chimp would need a better brain. Which means that it needs to improve, which means - "It hasn't evolved far enough."

[Ziggy Stardust]

Using humans in different stages of development as a model, it seems reasonable to me that you could create a variable that maps to typical human language development as your single axis. It might not work too well outside of mammals, but it would be particularly relevant for our nearer relations. Of course language is complex, but that certainly doesn't mean you can't abstract it if you want to.

I am just always wary of that sort of simplification, for two reasons. First, I am a cognitive scientist that studies language, so it's my job to pretend its more complex and important than it probably really is . Second, there are lots of people (Ray Kurzweil being a notable example) who use these oversimplistic linear models to make rather bold claims that don't hold any real water. I am not saying you CAN'T do it, but just that you should be wary of doing so and how you interpret the results.

Of course, you can throw all of that linearization out the window and just look at where the other species lies on your whatever-dimensional graph of language variables. Of course, in doing so you'd be drawing a straight line.

On a related note, there actually is a lot of work in computational psychology/linguistics in trying to develop linear, or low-dimensional, models of language. I am currently writing a paper on the subject, actually, in which (I hope) to demonstrate not only that early visual categorization in humans is shaped by semantic knowledge, but that the patterns of semantic categorization in humans correlates strongly with visual categorization in chimps. That is, conceptually chimps seem capable of processing the world in a way that aligns rather well with our own language-driven processing. The computational side of it involves lots of hundred-to-thousand dimensional matrices to which we apply singular value decomposition to "simplify" to more palatable low-dimensional matrices for analysis. The jury is still out on whether these decomposed linearizations are accurate or useful, depending on the study they either correlate well with standard human judgment or not at all.

[Ziggy Stardust]

Chimps simply don't have brain power to compute speech at more than a year old human's level - which is astounding, already, in my opinion giving the difference in size. Still, a chimp would need a better brain. Which means that it needs to improve, which means - "It hasn't evolved far enough."

I think a better way of phrasing this would be that the chimp's brain hasn't "evolved far enough in the right direction to conceptually manage human-like language abilities", or something like that. Saying stuff like the chimp needs a "better" brain is attaching value judgments, which may not be appropriate. The evolutionary path that chimpanzees have followed has not optimized their brains for language in the way that we understand it, because the pressures they faced historically were not selectively favoring that particular type of neural development. That is, they don't NEED language, and their brain has evolved just as "far" as ours, simply in a different "direction."

I realize this may seem rather pedantic and nitpicky, but ... well, I'm a scientist posting on an Internet forum, what do you expect?

[LaCroix]

I think a better way of phrasing this would be that the chimp's brain hasn't "evolved far enough in the right direction to conceptually manage human-like language abilities", or something like that. Saying stuff like the chimp needs a "better" brain is attaching value judgments, which may not be appropriate.

*snip*

I realize this may seem rather pedantic and nitpicky, but ... well, I'm a scientist posting on an Internet forum, what do you expect?

Given that chimps are far ahead of our abilities in rote memorization (e.g. their ability to instantly memorize number positions on a screen that are only shown for a second - humans simply can't match them in that test), I fully agree with you.

I simply viewed their evolutionary status under the premise of speech, in which they are found wanting. It was never supposed to be a judgement in general.

[Memnon]

I never even said that it hasn't evolved far enough in the human direction, just not far enough!

But what about dolphins, which may have comparable-ish powers of language? That isn't really in the human direction, after all. And yet you're using human traits as your basis of language skills. Considering also that the OP is about making them more physiologically like humans, I think it was a completely reasonable assumption to go from

I would almost expect it to be worse. A small child's brain is, after all, at least human. A chimp brain hasn't evolved that far.

to 'their language skills are not as evolved in the human direction'.

I am just always wary of that sort of simplification, for two reasons. First, I am a cognitive scientist that studies language, so it's my job to pretend its more complex and important than it probably really is :mrgreen: . Second, there are lots of people (Ray Kurzweil being a notable example) who use these oversimplistic linear models to make rather bold claims that don't hold any real water. I am not saying you CAN'T do it, but just that you should be wary of doing so and how you interpret the results.

Haha yes, it should be reasonable to assign biological meaning. Of course the meaning of such a variable would simply be 'how does a species compare, linguistically, to a human model of development' and not, say, 'can we boil language down to a single number that solves everything.' My bad if it seemed like I said that.

That is, conceptually chimps seem capable of processing the world in a way that aligns rather well with our own language-driven processing. The computational side of it involves lots of hundred-to-thousand dimensional matrices to which we apply singular value decomposition to "simplify" to more palatable low-dimensional matrices for analysis. The jury is still out on whether these decomposed linearizations are accurate or useful, depending on the study they either correlate well with standard human judgment or not at all.

Very cool. To be more specific, I was talking about using principal component analysis, the first couple of which are generally pretty useful. I've never heard of singular value decomposition but it looks really interesting!

[Ziggy Stardust]

Very cool. To be more specific, I was talking about using principal component analysis, the first couple of which are generally pretty useful. I've never heard of singular value decomposition but it looks really interesting!

Typically, SVD is used instead of PCA, even though the methods and results are broadly similar, because SVD maintains the sparseness of the original data, so for very large bodies of text PCA isn't as powerful, oddly enough. That said, recently there have been people using more advanced probabilistic models that use PCA within a maximum-likelihood framework. Then of course there is all the Bayesian model fitting stuff which my brain cannot properly handle yet.