Pinker, S. & Bloom, P. (1990). Natural language and natural selection. Behavioral and Brain Sciences13(4): 707-784.
Many people have argued that the evolution of the human language faculty cannot be explained by Darwinian natural selection. Chomsky and Gould have suggested that language may have evolved as the by‐product of selection for other abilities or as a consequence of as‐yet unknown laws of growth and form. Others have argued that a biological specialization for grammar is incompatible with every tenet of Darwinian theory ‐‐ that it shows no genetic variation, could not exist in any intermediate forms, confers no selective advantage, and would require more evolutionary time and genomic space than is available. We examine these arguments and show that they depend on inaccurate assumptions about biology or language or both. Evolutionary theory offers clear criteria for when a trait should be attributed to natural selection: complex design for some function, and the absence of alternative processes capable of explaining such complexity. Human language meets this criterion: grammar is a complex mechanism tailored to the transmission of propositional structures through a serial interface. Autonomous and arbitrary grammatical phenomena have been offered as counterexamples to the position that language is an adaptation, but this reasoning is unsound: communication protocols depend on arbitrary conventions that are adaptive as long as they are shared. Consequently, language acquisition in the child should systematically differ from language evolution in the species and attempts to analogize them are misleading. Reviewing other arguments and data, we conclude that there is every reason to believe that a specialization for grammar evolved by a conventional neo‐Darwinian process.
Tomasello, M., & Call, J. (2018). Thirty years of great ape gestures. Animal Cognition, 1-9.
Graham, Kirsty E; Catherine Hobaiter, James Ounsley, Takeshi Furuichi, Richard W. Byrne (2018) Bonobo and chimpanzee gestures overlap extensively in meaning. PLoS Biology
Pinker provides a compelling case for why we should think language, like our other abilities, is a product of natural selection. It fits the criteria of evolutionary theory; it's complexity eliminates some nonselectionsit explanations; it can have evolved gradually and be based on our genetics; the diversity we see is not just alright, it's expected. Overall, the argument seems sound.
ReplyDeleteFor a lack of criticism, I offer another avenue of thought that is a touch tangential but certainly related to the paper. In the last section on phyletic continuity, Pinker mentions that apes don't need to be able to understand language for us to say that it should have come from a common ancestor. However, the same way the eye looks simpler in a bacterium but more complex in an insect, so should we expect to see variants on language in species farther away from us if language were so beneficial.
This much we know. The same way we can imagine to be different if like the bee, our eye were a kaleidescope that could see UV light, so can we see how our language is to other animals. For the most part, animals do communicate with each other about predators and food sources, as well as their internal states. However, one species that is worth noting is that of the sperm whale. Consider the other minds problem: we cannot ever be sure another experiences qualia by communication alone. However, it seems like sperm whales can do something that is closer to communicating qualia.
Enjoy :https://www.wnycstudios.org/podcasts/radiolab/articles/worlds-smartest-animal
Full segment on sperm whales: 16:50-22:50
Segment on sperm-whale language: 19:12-20:46
Transcript available if listening's not an option - search for: [AUDIO CLIP: Sperm whale clicks.]
Things about language and language-evolution to think critically about:
DeleteDistinguish what is genetically encoded from what is learned and learnable.
Distinguish communication and language (set, subset). Language is a special kind of communication.
Communication conveys information, reduces uncertainty about what to do and what not to do, among options that matter.
Language does that too, in a special way. What is special about that way?
There is no doubt that communication capacities have evolved (genetically). The capacity for language evolved, but don't confuse that with changes in what we do with language, once we have it: Vocabulary "evolves," but not genetically: through learning, and through language itself.
It is hard (but not impossible) to determine when, why and how language capacity evolved. And there is no doubt that it evolved, because other species don't have it, and ours does. They have communication capacity, and lots of other cognitive capacities, but not language. What is it that they don't have?
And what is controversial about language evolution? There is one thing in particular. And it is syntactic (grammatical) not semantic: What is it, and why is it controversial?
Is sperm whale "language" language? What is language?
DeleteHi Professor Harnad,
DeleteI’ll attempt to respond to a few of your points..
You asked whether sperm whale language is language. I’ll say no, given that whales don’t communicate via spoken word. I think this is the distinguishing feature of language, in relation to communication. Whales and many other species communicate, but only humans have language — a very specific mode of communication.
You asked what is special about language.. I think one of the reasons might be that there are SO MANY different languages — such a wide variety of ways to say the same thing. Languages can be translated, so that any human being on the planet can understand. In addition, language can be either written or spoken - there are different modes of transmission.
You asked what’s controversial about language evolution. Is it universal grammar?
“There is no psychologically realistic multipurpose learning program that can acquire language as a special case, because the kinds of generalizations that must be made to acquire a grammar are at cross-purposes with those that are useful in acquiring other systems of knowledge from examples” (p. 25 of 50)
ReplyDeleteI think I am struggling a bit to identify exactly what the authors mean by the means by which one acquires language being at a cross-purpose with the means of acquiring other knowledge. Are they referring to acquiring other knowledge through categorization as we discussed in previous weeks? And therefore suggesting that language acquisition is fundamentally distinct from any sort of general purpose mechanism that we discussed like categorization? I remember that we discussed that there could be some innate categorical perception predispositions for aspects of language like hearing the different sounds, so I am a bit confused how suddenly they are at a “cross-purpose.”
Mostly what my questions address is the idea of having a specialized language module (separate from other cognitive capacities) which also is discussed frequently in the article. In terms of actually reverse-engineering, I don’t think that knowing there could be a module really helps us out because whether we designate it as a module or not we still have to figure out how and why it works the way it does. After reading this paper (although honestly I think I felt this way even prior to reading the paper), it seems clear to me that language cannot just be a spandrel or some chance occurrence that happened as our brains grew throughout evolution. The “why” of having language seems very advantageous in a group social setting. As far as the “why” different languages are organized in different ways, I think the paper makes it clear that these can be arbitrary (e.g. why we call dogs dogs, or the ordering of adjectives and nouns etc.), but nevertheless meaningful because as long as a community shares the same rules, communication is facilitated.
"Language" capacity is made up of a lot of components, some of them requiring the evolution of genetically evolved causal mechanisms, others just requiring learning capacity (already involved), or the prior presence of language itself.
DeleteWhat are these components? And which ones require genetically coded causal mechanisms?
And among those that require genetically coded causal mechanisms, which ones are (1) easy to determine, (2) hard to determine, or (3) controversial, and why?
What is "Baldwinian" Evolution, and how is it related to Darwinian Evolution?
When I think about components of language I think about the different courses that McGill offers in linguistics. Language is made up of phonetics, phonology, morphology, syntax, and semantics, just to name some. As far as why determining if some take advantage of genetically coded causal mechanisms could be controversial, I wonder if it has to do with the process of gathering empirical evidence. If something is innate, infants with limited/no environmental experience should be able to show some above-chance tendency or skill. From my child psychology class I remember that experimental set-ups used with infants had to be very creative given infants cannot communicate with us via words. So perhaps there is some controversy over if infants cannot show a certain language skill, or if we simply have not designed a proper experiment so that they could show us this.
Delete“The process whereby environmentally-induced responses set up selection pressures for such responses to become innate, triggering conventional Darwinian evolution that superficially mimics a Lamarckian sequence, is sometimes known as the Baldwin Effect.” (page 29 of 50)
Delete“Therefore, the Baldwin effect is a combination of learning and the genetic assimilation of a learned trait” Podlipniak P. (2017). The Role of the Baldwin Effect in the Evolution of Human Musicality. Frontiers in neuroscience, 11, 542. https://doi.org/10.3389/fnins.2017.00542
From this quote as well as some additional reading (Podlipniak, 2017), I think that Baldwinian evolution is the transition from our behaviour needing to be learned from scratch by every individual, to eventually becoming instinctual or genetically innate. Darwinian evolution is then the application of natural selection to favour those individuals who have this genetic predisposition to acquire such a behaviour faster, or better, or more efficiently. So as far as language, this would be the idea that the components of language that people initially had to learn from scratch could eventually become an innate ability. Maybe an innate ability like how we have categorical perception for speech sounds that exist along a continuum but we hear as discrete categories. It would be really inefficient and probably require a lot of trials to learn where that boundary between the sounds exists. I suppose this could be where the poverty of stimulus enters in that infants can distinguish these categories with minimal environmental input, so therefore some aspect of this ability must be innate.
Baldwinian Evolution. All spot-on, Stephanie. Saved me the need to explain Baldwinian Evolution. It's yet another talent of lazy evolution: If something that is learned (already a lazy effect) turns out to be very adaptive, select for a stronger and stronger disposition to learn it faster and faster.
DeletePhoneme boundaries -- ba/da/ga, ba/pa, and ra/la CP -- have been shown to be innate in preverbal infants. using repetition/novelty dis-habituation techniques. You can lose them if you don't use them -- in production or perception -- during their early critical period.
With Universal Grammar (UG) the question is more complicated, because of the Poverty of the Stimulus (POS): no negative examples (remember the "Laylek" example?).
The "lazy evolution" concept seems to describe micro and macro evolutionary processes as taking the path(s) of least resistance, as all organisms interact with and multiply or go extinct within both their social and physical environments.
DeleteBaldwinian Evolution as the marriage of environmental influences and social factors and organization also seems to mesh with "cognitive ecology" approach described by Cauchoix and Chaine in their discussion of evolutionary psychology, which brings together socially and environmentally influenced trait selection. The selection process is not necessarily for innate, already-developed traits which are more conducive to survival in certain ecosystems, but rather for innate, dynamical-causal mechanisms which happen to be the most ideal for developing and learning certain beneficial traits such as linguistic communication and CP, which have become more and more prevalent over tremendously long periods of time.
I think the paper makes some great points as to why language could be the result of evolution. They clearly show that language is complex enough to have to be selected for specifically as opposed to being a side product of some other function and that there could be reproductive advantages that select for mutations that lead to better grammars (although I don't necessarily see how this would work in cases where the grammar rule is arbitrary (as defined in the paper). Arbitrary rules lead to an improved understanding now, when we all have them, but I'm not necessairly convinved they would lead to a communicative advantage between someone that naturally follows the rule and someone who doesn't). I am, however, unconvinced by their argument regarding Phyletic Continuity.
ReplyDeleteThey seem to be arguing that the reason we have the innate ability to learn this complex and extremely useful tool that no other animal species to our knowledge has is that we developed it in the last 3.5-5 million years which is a valid argument. However, I don't think they address the more important issue of why no other animal evolved language in the way we did. If it is really a graudal change, as it might be, that's as useful to us as it has been, then what prevents other animals from developing something similar?
This is not to say that any of the other theories of language development address or solve this problem any better, which is the point that Pinker and Bloom make in the paper to dismiss the argument. But I don't think we can definitively argue for language being developed by natural selection until a theory accounts for this.
Good points. Here's some more food for thought:
Delete(1) What are the reasons to believe that language evolved rather than just being invented (like chess)?
(2) If language evolved rather than being invented, why might explaining how and why it evolved be difficult? Why did Pinker & Bloom write this paper?
(3) What is Universal Grammar (UG)? and what is the difference between UG and Ordinary Grammar (OG)? (See other Replies for both readings, 8a and 8b)
(4) What is "Poverty of the Stimulus"? How is it related to the problem of explaining the origin and evolution of language?
(5) What is the difference between semantics and syntax?
(6) What is language? And what is it that no other species has?
Since it seems that most of the crucial points and takeaways from this reading have been mentioned in one post or another, I'll try to answer a few of the "food for thought" questions to wrap my head around what we are discussing here.
Delete(1) According to P&B, language meets the necessary criteria for it to be an evolved capacity. Namely, it has "complex design for some function, and the absence of alternative processes capable of explaining such complexity". I think it is pretty clear that language has tremendous adaptive advantages. The question really isn't whether language evolved, but rather how and why. I suspect that in order to invent something like ches you might need a language. The same can be said of language: in order to invent one (as we do with formal languages), you need to have language in the first place.
(2) Explaining how and why language evolved is difficult because of UG. Darwinian evolution is subject to gradualism. This implies that UG should have been selected for gradually. That cannot be the case since UG is what gives us the capacity to learn a language and with only part of it, there is no language at all. In other words, language is all-or-none in the sense that you either have a symbol system that can produce all propositions or you don't. Does this allow us to ask whether UG is a spandrel/exaptation? I'm not sure about that but this seems to be the point: no one really knows how to explain UG.
The question is: what features of language are evolved (genetically) and what features are learned (or invented)? It's probably also misleading to treat language as a single capacity. Some features might have preceded others.
DeleteYes, UG is special; it does not seem to be a bunch of independent or semi-independent features. Rather it is itself one of the features of language. Propositional and predicate logic does not seem to need UG. And then there's the question of whether UG (syntax) is independent of semantics.
Maybe you need UG to produce and understand propositions of complexity much greater than "the cat is on the mat." But did our ancestors need that, at the advent of the proposition?
“No single mutation or recombination could have led to an entire universal grammar, but it could have led a parent with an n-rule grammar to have an offspring with an n+1 rule grammar, or a parent with an m-symbol rule to have an offspring with an m+1 symbol rule. It could also lead to a parent with no grammatical rules at all and just rote associations to have an offspring with a single rule.” (p.30)
ReplyDeleteI find this passage to be somewhat contradictory to the support of Baldwinian evolution also present in this paper. Would it be enough for one person who acquires a linguistic or grammatical mutation to disseminate it within a community through genetics alone? Or is this suggesting that a combination of mutations increasing grammatical ability could then be passed on both through learning and through genetics?
Another thing that I keep returning to: language is a nuclear weapon for categorization and learning because we can create and exchange categories through agreed upon symbols. But in a context when symbols and innate grammar are minimal, how could one individual explain their grammatical rules and categories (which are purely abstract) to others? I find that this explanation of grammar transmission was lacking in the paper considering it is a necessary step which could lead to universal grammar.
Yes, the P&B passage is vague, speculative, and probably incoherent. The problem with the evolution (and evolvability) of UG is that it does not make sense rule-by-rule. What is a "partial language"?
DeleteBut distinguish Ordinary Grammar (OG) -- which is learnable, changeable, and not universal -- from Universal Grammar (UG), which is unlearnable (because of the Poverty of the Stimulus, POS), fixed, and universal.
And, yes, the adaptive value of language has to be mutual, between at least two individuals. No simplistic mutation theories make sense.
(OG can be learned passively, by unsupervised learning, or actively, by either supervised learning or by instruction. But UG cannot be learned by either supervised or unsupervised learning, and the rules were not even known before Chomsky, so it was certainly never learned by instruction!)
I found the claim that frontal lobe development in humans could have resulted from a "cognitive arms race" between cooperative groups of early humans and cheaters (p.35) interesting. This reminds me of the in-class discussion about whether lying spurred the creation of language, and Professor Harnad's view that truth must have come before lies, and therefore a language must have come before.
ReplyDeleteIt seems to me there is a chicken and the egg kind of problem at hand. Did the development of language, spurred on by increasingly complex societies which required increasingly complex grammar, lead to an increased cognitive development, or was it that the development of cognition lead to the development of complex grammar? It seems to me that it could very well be that both go hand in hand - you can't develop language if you don't have the cognitive skills for it. However, we've seen in class that basic abilities that are required in cognition, like categorization, exist in non-human animals. So what is this something "extra" that we have?
All open questions since the events occurred 2-300K years ago.
DeleteI think the authors did a very good job at explaining how evolution could be a reason as to why we have language, however, I don't think they stressed enough about why it needed to evolve in the first place.
ReplyDeleteThere could be many reasons as to why -- I think a convincing one is that an evolutionary pressure to have language is because we as a species needed to communicate effectively in order to meet the demands of the group (this was spoken about briefly in p.35). Language is the best form of communication that humans have, and considering the harsh environmental demands a hunter-gatherer would experience, developing an effective form of communication seems almost as important as eating. The authors evaluated language as a tool that arose out of evolutionary pressures, but didn't really provide a context in which the tool was utilized the most (e.g. a group, and not just with one person).
While I don't think this exactly helps solve how language was really acquired, or how we have universal grammar, I think evaluating how language could circulate in a group and how it evolves generationally provides a better answer or direction versus simply evaluating it through the individual. Because it's not as if the person who has the strongest linguistic skills will necessarily be main person who passes down their genes with superior capacities for language; in order for the person to be integrated into the group he/she must also be understood by others and be understood as well (discussed in the reading). This means that despite their superior linguistic abilities, they would have teach it others, and the rest of the group. Assuming that most of the group members reproduce successfully and all have sufficient language capacities, they would then pass their genes with the capability of producing and learning language. Thus, this would facilitate the propagation of language throughout the generations much easier than just one individual being blessed with superior genes than the rest. Though, I am not ruling out that an individual with vastly superior language capacities could not aid in this propagation, I just think that it can't be relied on as the best explanation for the genetic propagation of language.
No doubt that if you have a social species, say, chimpanzees, it would be a great benefit to them if they could communicate through language. But language is not a unitary trait, like blue eyes; it has many complex features, so is unthinkable as a point mutation. Many of its features are learnable and learned, but their adaptive advantages are all based on already having language capacity, and language.
DeleteLanguage is probably the most difficult biological trait to explain evolutionarily (second only to the hard problem of the evolution of feeling). Simple gradualism simply does not work. Think about it.
P&B oversimplify the problem -- and on UG they beg the question altogether.
It does seem undeniable that language would have had to develop through natural selection, given that it is so adaptive and complex. According to Pinker and Bloom, all specialized biological systems are shaped by natural selection, and therefore language is too. It seems like this would need to be especially true of universal grammar, given its innate quality. How could we acquire something innate if not through evolution?
DeleteDespite this, it is true that language is uniquely hard to explain through natural selection processes, notably due to its strictly interpersonal value. As Professor Harnad said above, the advantages are based on already having language. This is something that I had never considered before this class, but it is indeed a very tricky question
The solution you describe does not seem to solve the problem. First, as Professor Harnad also pointed out, language could not have arisen from a single mutation, so there would likely never be a situation where only a single person in an environment could suddenly speak. Moreover, your description seems to suggest that having the linguistically gifted speaker teach others in the group would transmit genetic capabilities to them, which they could spread by reproducing. However, we do not genetically encode things by learning them; their children would be back at square one. In either case, the “learners” would require specific capabilities for it to be possible to teach them language in the first place. I may have misunderstood what you were describing, but that is what seemed to be implied.
All of this seems to suggest that the development of capacities necessary for language would somehow have had to be synchronous. The tipping point would have been the advent of propositions, but it’s difficult to imagine how exactly this occurred.
@Prof and @Julia, I'm going to have to separate my reply in two parts, as it is too long to be posted in one reply.
DeletePart 1:
Prof: Thank you for clarifying where I went wrong with my proposition. I was seeing language too much as a unitary trait and not recognizing the issue at hand being: Language acquisition is predicated on already having language capacity, and language itself. This definitely puts a hole in my gradualistic argument. Though, I must say that I don't think it would be an all-or-none evolutionary trait that would engender language capacities. It seems to me to that assuming our closest ancestor (the one that directly precedes us, not our common ancestor) did not have language capacities seems a bit far-fetched.
@Julia: What do you mean by "strictly interpersonal value"?
You're right that my comment may have been a bit unclear -- I'll explain what I meant in more detail and provide a better context. My argument is centered around Gene Expression that is elicited in individuals and how that can be one way that language may have been acquired and/or evolved.
I hope I am not butchering the beautiful science that is genetics in my explanation, but we do genetically encode information. One way in which we do this is with Immediate Early Gene expression (IEG). IEGs occur as a transient and immediate response to cellular stimuli. It is the first line of response that is encoded before protein synthesis occurs, and this will affect gene-expression in the individual. IEGs are responsible for synaptic plasticity, memory encoding & consolidation, learning, and long-term potentiation in neurons. This is a continuous process that occurs all the time, and genes can turn on and off all the time.
Part 2:
DeleteI tie this into my group argument because if there is one individual who is linguistically gifted, he still carries the same genes as the rest of his group (just in a different combination that makes him gifted), and thus he could transmit his knowledge through eliciting IEGs in others. Let's say that he possesses 100% of the linguistic skills, and the group is at 75% compared to him, this may leave some room for the group to learn from him and "elevate" their linguistic skills. If he teaches what he knows about language in a manner that the group understands and can properly integrate into their knowledge, this would most certainly cause IEGs for the neurons responsible for encoding language. Therefore, if this is done repeatedly, and to a point where the individuals in the group reach their maximum potential (doesn't have to be 100%, but even a 10% jump would be a lot), he would have effectively altered the genes the non-gifted individuals carry (This would be permanent as the individual would use the skills given to him for the rest of his life, which would solidify the plasticity that they underwent when initially learning). This alteration would not be grandiose, as we presumably have thousands of genes being affected here, and they all have tiny effect sizes individually, but when working together, they produce those "skills". So, when it comes time to pass on one's genes, they would have passed on their genes that have been altered by the gifted-speaker.
This isn't to say that he is helping them increase their capacities, per se, but simply improving the ways in which they utilize their capacities. As the Professor has pointed out, we need to have those language capacities and language in the first place for this to even possibly occur, so there is definitely a huge gap in my proposition.
Lastly, I think propositions are very interesting and rather convincing, but I get the sense that it may need a sensorimotor explanation for its potential origins, versus it exclusively being a product of language. What I mean by this is that if propositions are sentences/statements that are either inherently true or false, we must have an innate understanding of what is true and false. Obviously truth is an extremely complex topic, but finding out what our sensorimotor capacities interpret as "true" could be seen as a foundation of what a proposition is. One example I could think of is knowing that pain is something that hurts, we have innate reactions to many forms of pain, and that in order to survive, it is best to avoid pain. This would be understood implicitly by our bodies (and indeed, we do have such understandings. We retract our limbs when touching a hot surface before we even think about retracting it), and I think that could serve as a form of proposition.
"There is no psychologically realistic multipurpose learning program that can acquire language as a special case, because the kinds of generalizations that must be made to acquire a grammar are at cross-purposes with those that are useful in acquiring other systems of knowledge from examples (Chomsky, 1982; Pinker, 1979, 1984; Wexler and Culicover, 1980). The gross facts about the dissociability of language and other learned cultural systems, listed in the first paragraph of this paper, also belie the suggestion that language is a spandrel of any general cognitive learning ability."
ReplyDeleteI’m not sure if I agree with the authors on this point. Based on what we know from past lectures, language learning is a type of categorical learning. We can learn what is grammatical or not through a combination of supervised learning (someone explicitly correcting us if we make a mistake) and unsupervised learning (mere exposure to speakers). If learning grammar is a form of categorical learning, then why would it be at a cross-purpose with other systems that acquire knowledge through examples? Is it because the ratio between supervised:unsupervised learning is radically different with grammar than with, for example, learning how to play baseball?
Follow up question/concern: if we figure out the ratio between supervised:unsupervised learning, doesn't that mean we could technically engineer a T3 robot to learn grammar? T3 would also have an edge to grounding language in nonverbal cues.
If language is not a spandrel of any general cognitive learning ability, then by what chance in the natural selection process did it become such a nuclear weapon? Going back to last week's reading by Lewis et al., it is very difficult to narrow down the when/where/how of a certain environment to aid in the evolution (or is it invention?) of language. I'm tempted to say that something has occurred through a sexual encounter somewhere in the past, but we’re most definitely not only only species who reproduce sexually. I still can't explain why we have this specific type of language, and chimpanzees don't.
Hi Wendy! I just wanted to offer some commentary to some of the points you made. If we follow some of what we learned about language in class (in particular, in regards to Chomsky), grammar is a special category because it is arguably one of the few that are innate to some degree. What makes people think that it’s unique is that the input we receive when we learn our first language is all (or virtually all – sometimes your parents could utter an improper sentence by accident) correct. Recall that in the lase lecture, having a negative instance of a category was necessary to learn the category. So, if there are no negative instances to be had, it follows that grammar is not learned but innate within the baby and ‘crystallises’ depending on what language they learn. I would posit to say that in a T3 robot, if you wanted to have a grammar, you would have to establish some general rules of syntax that could apply to all languages. Then they would learn a specific language according to, for a lack of a better word, unsupervised learning.
DeleteAs for the chimpanzee part of your post, I think that what the authors say, that it’s plausible that an ancestor of humans that is not shared in common with chimpanzees, originated language. But this poses a dead-end to the argument really since you can’t observe an extinct species and see what language they do. The authors point out to a case where they found a well-developed Broca’s area in a hominid fossil, but as seen many times in the course, you cannot causally link brain structure to cognitive function, especially in an instance as tenuous as this.
William's reply to Wendy is correct. Language has the special problem of UG, which, unlike OG, is unlearnbable. And yes, P&B beg the question.
DeletePinker & Bloom express that “engineering demands are simply too complex” for language not to have been the result of a complex mechanism devoted to facilitating the communication of “exquisitely complex and subtle messages” (p.15). But what if it really was the result of an auspicious combination of other abilities and their spandrels? Might that explain why we haven’t (yet) seen language be developed in other species? That they don't have the full list of ingredients necessary whereas we did (or do)? This asks for a reverse-engineering of all the necessary components for the formation of language, just like it is deemed necessary for cognition, in a machine.
ReplyDeleteEli, I think you bring up a super interesting point! I think another (but very simple) question I have as a follow-up is... what is so special about propositions? Last class, we concluded that language is special because it has the capability to form propositions. Maybe other species just haven't had a need to use propositions... so even if the full ingredients were there, they didn't need to use them cause they're not interested in creating propositions?
Delete(Kind of building off of the argument that natural selection acts as a "designer" to certain basic plans, parts, and materials already present as spandrels (2.2))
DeleteE and E: If we wanted an answer to the question of where, when, and especially why and how language evolved, we need something a lot more specific and testable than "an auspicious combination of other abilities and their spandrels."
DeleteAnd things don't evolve because you need them; they evolve because those individuals who had genes that coded for them survived and reproduced better than those who didn't. Propositions sound simple, but apparently the capacities needed to produce and use them are not that simple. (Ants and bees would also be even better off if they could talk, but the "parts" are apparently not in place: What are they, in us?)
My father and I once had a discussion about language. I thought that since my dog knew the words “walk” and “supper” and even my own name, surely he understands language to some degree. My father disagreed and argued that there was something different about the way I understood these words than how Rusty the dog did. Rusty doesn’t manipulate words in his head, and he doesn’t come up with unique ideas using the words he knows. (Is manipulation a requirement of language?) Still, I believe the words were grounded for Rusty. If I told him to get a stick, he would fetch one. If he came trotting back with a tree trunk, I could say, “too big”, and he would go find a smaller stick. So then what is language? And what makes humans so special to be the only species to have mastered it?
ReplyDelete(My father also argued that if the dog has language, then so does the smoke alarm which yells "fire" when it detects smoke. Unfortunately, this was years ago so I couldn't explain to him that the smoke alarm's words were ungrounded, but the dog’s were.)
Bloom and Pinker claim “There must have been a series of steps leading from no language at all to language as we now find it.” Are dogs at some stage between no language at all and some language?
Or, did language simply emerge from some other neural processes? If this is the case, the distinction between human language and animal communication becomes more obvious. Every animal can communicate to some degree. Language is something completely different that humans have been lucky enough to have obtain (along with more general communication). But then we don’t believe language emerged like spandrels. So does this mean that the evolution hypothesis is correct? And if language did emerge as a response to evolutionary pressures, why? Language is incredibly helpful, but I don’t think such a complex thing as language would have saved enough lives to have been created through natural selection. If I am wrong and language did evolve, then I want to believe that my dog Rusty did have some sort of capacity for language, just to a lesser degree than humans.
Yes, language is not just associating an arbitrary name with a category you have (i.e., a kind of thing for which you know how to do the right thing). It is the capacity to combine and recombine category names to define and describe further categories.
DeleteYour father was right. And that re-casts the problem: Since Rusty obviously knows what "stick" and "fetch" and "too big" calls to mind, and can do the right thing that will make you happy (and maybe give him a reward of a hug or a treat), how come he cannot understand (or produce -- which is something that is far more important in the reciprocal relation of saying and understanding something) a subject/predicate proposition simply stating "the stick is big" -- without requests/commands or rewards?
Dogs are undeniably little geniuses (so are pigs and other mammals, and many kinds of birds, and even octopuses): have we just not found a way to prove that they don't understand propositions? (They certainly don't make them.)
I think it's obvious that language is a form of communication, that other species communicate too, and that some of the features of prelinguistic communication were drawn upon in the evolution of language. The trick is identifying them, and putting them together in a credible sequence in which the adaptive advantages are evident.
Hi Katherine! I kept nodding my head while reading your post because, I too, thought that if a dog or any other little genius paid attention and responded accordingly to us, that meant it understood our language. After reading Professor Harnad's comment, I now know that there are two requirements for language: categorization (the capacity to "do the right thing with the right kind of thing") and manipulation ("the capacity to create combinations of categories to explain further categories").
DeleteHowever, I am still not convinced that we can draw the line here and say that no other species can have a language similar to ours. A few months ago, I found out about a dog named Bunny. She can communicate with her parents with buttons on a soundboard! According to her owners, the soundboard is similar to the AAC (augmentative and alternative communication) device Stephen Hawking used. AAC is a communication technique used to supplement or replace verbal speech for individuals with speech-production or speech-comprehension disorders. Each button on the soundboard is associated with a word or phrase. Through a combination of reinforcement learning, supervised learning and learning by instruction, Bunny was able to ground English words and connect them to their referents in the world. For example, she can show affection by pressing on the "love you" button or let her parents know of an injury by pressing on two buttons: "hmm" and "ouch." Although she cannot form propositions with both a subject and predicate, they do have a truth value, and she is understandable. In this case, hasn't Bunny created her own language?
Here's a (super cute) video of Bunny "talking": https://www.youtube.com/watch?v=FeiHw2oxYJM
Not to mention, there are many other videos of dogs and cats "talking" with their owners too!
I love that video of Bunny communicating. And she's certainly communicating, about categories of things sh wants. That's been done with chimps too. But it's not propositions, which are not just statements of wanting, or of affection ("Love ya"). (And the "translations" are misleading. They make it sound as if she was saying the equivalent of those English words.)
DeleteThe mystery stands: why is Bunny not having conversations about philosophy -- or at least about toys -- by now? What's going on in her mind (which is undeniably brilliant) and why is it not being expressed in words?
By the way, I think it would work better and faster if they were soft buttons on a wall that she could press with her nose rather than buttons on the floor to press with her feet. Possibly morse-like bark sequences would be good too. And there could maybe be a better choice of basic buttons (though I don't know what they all are).
Notice that, like children, she understand a lot more than what she can say.
Such a neat video!! Could it be argued that Bunny is computing rather than cognizing? It's clear that Bunny has been able to ground words with real world concepts, and associate them with buttons. She effectively manipulates symbols (interacts with buttons) and provides the correct outputs in response to inputs (owner's questions). However, is this cognizing? Perhaps this is a completely different question and it goes beyond the scope of this topic (it's not necessarily related to language), but it came to mind as I was reading this thread and watching the video.
DeleteOf course what Bunny does is cognizing; the question is whether it is language (if language means the capacity to make and understand propositions).
DeleteI have seen these Bunny videos before, but I feel like they only drive home the point that dogs don’t have language. Bunny does seem to know the meanings of these some of the words. She does seem to be categorizing. But the manipulating of words seems flawed.
DeleteBunny reminded me of another intelligent dog I had seen on the internet a couple years ago. Chaser the dog has an incredible memory. She knows the names of one thousand of her toys. Her owner can instruct her to fetch any toy by name from a huge pile, and she will get it right every time. (The video I reference also shows Chaser fetch a toy she has never seen before, just by inferring it wasn’t one of her normal toys). It is clear that dogs are very, very intelligent. But this ability to remember which toy has which name greatly resembles what Bunny is doing. Bunny knows what “play” and “tug” and “mom” mean, but only in the way that Chaser knows her toys’ names. It seems to be a labeling ability. And sure, Bunny can communicate these labels back to her owner and appear to be using language, but all she is doing is labeling concepts in her head with a specific button press. It is very clever, but it isn’t language.
Video of Chaser: https://www.youtube.com/watch?v=omaHv5sxiFI&ab_channel=NOVAPBSOfficial
When we discussed Bunny, and her communicative capacities in class, one of the major things that came up was the question of whether what she was doing is really language. Professor Harnad contested this idea on the basis that requests - such as the one made by bunny that her paw hurt - do not constitute propositions.
DeleteAs Blondin Masse et. al defines them, propositions are statements about categories than have truth values (ex. "the sky is blue" puts the sky in the category of "blue things" and is false if sky in fact belongs to the mutually exclusive category of "red things"). While Bunny's request about the thorn in her paw seems like language, it does not actually have the shape of a proposition and it certainly does not have a truth value (although, in her defense, one can ask whether any evokation or communication of a painstate (unavailable to confirmation via the problem of other minds) can be checked for truth value). If it doesn't have a truth value, it isn't a proposition and therefore isn't language.
Further, it is difficult to know if the symbols (button-words) Bunny is using are sufficiently grounded: paw seems easy enough, I can readily believe that she knows which parts of her body she is referring to, but a belief that she understands more complex and abstract ideas like 'stranger' is more difficult to produce. After all, I don't think most English speakers would describe an splinter as a 'stranger'.
As was evident in class, we kind of want to believe this, because Bunny is clearly very smart and very cute - further many people see their pets as 'family' and therefore pets often seem more 'human' to us than other animals because they have been placed in the same category as some humans (CP effect perhaps?).As such, we are all ready to make the leaps for her - to say that "stranger" is like "foreigner" and "foreigner" is like "foreign object" - an appropriate word for splinter. Yet, these are our leaps, and our cognition, not Bunny's. As much as we would all like her to really 'talk', I'm not sure she does.
The authors in this paper provide strong arguments for why language may have evolved from natural selection, by highlighting the fact that, so far "natural selection is the only scientific explanation of adaptive complexity" (2.2). When I first read this passage, I didn't really understood what that meant, but after looking at it a couple times, I think that they're saying that so far, only natural selection can provide a causal mechanism for language. But does this help us reverse-engineer cognition/language? Natural selection provides a possible causal mechanism for why language exists... can we use this to also find a causal mechanism for how language works in our minds? I think the authors are exploring this a bit in section 3 (Design in language) when they propose that language shows "signs of design for the communication of propositional structures over a serial channel" (3.0). But I'm not entirely sure what they meant by a "serial channel".
ReplyDeleteTo be very honest, I found this paper really hard to digest, so I'm just writing this skywriting as a cry for help haha.
(and I'm responding to the paper by Pinker & Bloom!)
DeleteA "serial channel" just means that the bits come out one after the other, not all at once, as in a complex sound or complex shape.
DeleteAnd yes, T2-T4 all require a causal mechanism for producing and understanding language, and an evolutionary explanation of how and why language evolved might give some help in designing the causal mechanism.
I think most of us agree that language is too complex to be a biproduct (I forgot the fancier word for it) but are still unclear as to why we would have evolved this ability. Language consists of a subject and predicate with a truth value, all together called a proposition. Our ability to make propositions is unique to us. Other animals have communication but not language. Some of the skywritings above mentioned that maybe we faced a specific evolutionary pressure that other animals did not which led to the evolution of language, others suggest that maybe other animals simply haven’t evolved language yet. We know that for something to be evolutionary selected for, those with that trait/ability were better able to survive and pass down their genetics. So, if language were evolutionarily evolved, there was something about it that enabled the language haver to live and reproduce in a way that non-language havers could not. The best explanation I could think of was that perhaps at some point, humans needed to lie to survive. I imagined this in a situation where one caveman was about to kill the other for stealing his food, and then the threatened caveman lied and said he did not steal his food. I know this is a super oversimplified way of thinking, but I wanted to explain how I thought lying might actually be evolutionarily useful for survival.
ReplyDeleteTo be able to lie and say you did not steal the food, you need to have language (hence, presumably true propositions) first. How/why do those evolve?
DeleteMaybe language evolved for more efficient and detailed communication of information. Or possibly for the transmission of ideas — something unique to humans(?) I think language is an indication that humans are creative and imaginative, but I’m not sure how this might have come about. What made humans more likely than animals to develop culture, art and literature? I also wonder what developed first — written or spoken language. My guess would be written but I haven’t looked into it.. I’ve been thinking about cave drawings and hieroglyphics. Perhaps the initial basis for language development, followed by the verbalization of these symbols, to make communication more efficient?
DeleteIf language began with writing and writing began with drawing images, that's similar to language starting from pantomime (icon to symbol to proposition). But how and why?
DeleteI quite liked this article for its thoroughness and organized structure, although I have to say (as the professor mentioned above) that it mainly deals with topics that are 'uncontroversial' such as: language is important for human social interaction and the development of culture. However, as cognitive scientists, the crucial part of language for us is decidedly controversial.
ReplyDeleteWe spoke in the last class about what makes language different and 'special' compared to other forms of communication that we see throughout the animal kingdom, and the answer was that it allows its users to form propositions. This is profound not only because it allows us to convey complex ideas but because language speakers and hearers can learn categories through language alone, without any direct interaction from the senses. As Harnad says in a previous paper on categorization, this the "the full and unique category-conveying power of language" (Harnad, 2003). However, the fact that we can identify that propositions make the difference between language and other forms of communication doesn't help us reverse engineer the capacity and it does not tell us in the least why we have developed this ability when others have not. As someone said above, how and why is it that dogs and other animals can ground the sound of an arbitrary string of symbols in a real-world object and yet they cannot form/understand propositions? If we want to have any hope of creating a T3 (and probably T2) we need to be able to answer these questions.
There's also the question of Universal Grammar (UG) which someone brought up in an earlier comment. Chomsky would argue that a huge amount of our language abilities is innate. This is based on the poverty of the stimulus argument (again, expanded upon above) which says that we do not have nearly enough stimulus exposure to acquire our language proficiency from learning alone— therefor much of our capacity must be hardwired and according to Chomsky "is shaped by neither learning nor evolution" (Harnad, 2003). Pinker and Bloom are right in that many things about language fit into the Darwinian natural selection mould, but if I'm understanding correctly, what they refer to is OG and not UG. UG it seems, still has many mysterious and controversial elements to it.
1. Pinker & Bloom skipped over the controversial part (the Poverty of the Stimulus, POS).
Delete2. POS does not apply to all of "language" (phonology, vocabulary, semantics, even Ordinary Grammar, OG). It applies only to Universal Grammar, UG).
3. Propositions are already there in ordinary propositional logic (If proposition "P" is true, and proposition "P implies Q" is true, then proposition "Q" is true) as well as in first-order predicate logic (If All humans are Mortal, and Socrates is a human, then Socrates is mortal). If you can produce and understand propositions, you can produce and understand all possible propositions.
So it looks as if you don't need UG (or even OG) to do formal propositional logic or predicate logic. Just the simple -- and learnable -- syntax of the first-order logics. To make it into propositional language all you need is that its symbols should be grounded (i.e., its words must be connected somehow to their referents).
If the adaptive advantages of language are already there with that, then it's unclear what we need UG for (except maybe some abstract distinctions, such as the fact that we can say "John is eager to please" and "John is easy to please" and John is eager to please Mary" but not *"John is easy to please Mary." The learnable rules of OG cannot explain that (nor explain an infinity of other no-no's we all know that are explainable by unlearnable UG but not by learnable OG).
These present puzzles for the evolution of language, and also the nature of language. But Pinker & Bloom pass over them in silence.
Overall I enjoyed the paper, I hadn’t thought about how more specific biological systems are more likely to be direct products of natural selection, but it makes sense. It also clearly supports the authors’ argument that a complex system like the capacity for language is a result of natural selection.
ReplyDeleteProf. Harnad replied to another post saying that language is, “the capacity to combine and recombine category names to define and describe further categories.” Are the categories he’s referring to syntactic or semantic?
If they are semantic, does this quote accurately describe the process he’s referring to?:
“Similarly, a verb like ‘hit’ is made into a verb phrase by marking it for tense and aspect and adding an object, thus enabling it to describe an event.”
I wonder in what ways learning language impacts our cognition, because it seems like at least your conscious thoughts filter through it; how would it differ if we grew up without being taught language?
See below.
DeleteLanguage makes it possible to define or describe any category.
ReplyDeleteIf you remember just one proposition for the power of language, I recommend "zebra = horse + stripes"
For an example of what happens to a species with evolved, language-prepared brains when deprived of language, see:
Senghas, A., Kita, S., & Özyürek, A. (2004). Children creating core properties of language: Evidence from an emerging sign language in Nicaragua. Science, 305(5691), 1779-1782.
Exercise: Why is the emergence of language in the Nicaraguan deaf community not a glimpse of how language originally evolved?
DeleteIn response to your question and the paper on the Nicaraguan deaf community:
DeleteAfter reading the paper on the Nicaraguan deaf community I could not help but think that the environment of the deaf children is very different from the original environment of human ancestors that began to evolve language capacities. The paper mentions that many deaf people without peers develop “home signs” that they use with family members. As a result, it is clear that simply by living with family members that can speak means there is a different learning environment than our ancestors. The paper also highlights how people who can speak gesture differently than those who use NSL, again offering another example of a different learning environment.
Furthermore, the children who are deaf would already possess an advanced evolutionary benefit to language acquisition. This means that they cannot provide a glimpse of how language originally evolved because they are starting their language formation with a head start compared to the original language developers. This sounds similar to the concept of Baldwinian Evolution that has been defined and discussed in earlier posts.
Yes. And of the things you mentioned, being born with a language-prepared brain is the most important. But gestural language is interesting and important in its own right. Home sign is not a language; it's more like a pidgin; but its users (both the hearing, speaking family members and the non-hearing one) have language-prepared brains. It's interesting that it required a gesturing community to turn home-sign into a language (or to beginning to be used as a language). The obstacle clearly had not been the capacity to make propositions. Nor did the demands of family life drive the family members into expanding home sign into a language. And the role played by the sensory handicap (and the mistaken bias that the nonhearing child was not cognitively capable of language, or of more language use) is worth some more reflection.
DeleteIn some of the previous readings, we have learned that evolution is “lazy” - it is much easier to code into our genome the ability to learn something instead of the thing itself. Pinker and Bloom argue that for this reason, language must have evolved, and only the mechanism by which we can produce and learn language is hard-coded into our genome. What my colleagues have pointed out is the paper has avoided the question of Universal Grammar as described by Chomsky. The mystery of universal Grammar (UG) is that it seems to be something that is innate, because humans only hear grammatically correct sentences, yet somehow understand which sentences are grammatically incorrect. This is known as the Poverty of Stimulus - we do not have to hear sentences that are incorrect to know that they are incorrect. Yet if evolution is lazy, then the capacity to learn UG must somehow be evolved and not innate, but we cannot learn UG because the only data we have are grammatically correct sentences. Is it even possible that UG is an evolved trait?
ReplyDeleteYou put your finger on it. Pinker & Bloom are wrong in their message that there is "no problem" about the evolution of language. But it's a problem specific to the evolution of UG, which is the one question they passed over in silence. (This, by the way, is the reason Chomsky declined to be a commentator on their paper.)
DeleteFor the intrepid, here is the P&B target article plus all the open peer commentary (skywriting) it elicited at the time, plus P&B's Response.
DeleteI can certainly appreciate and agree that evolutionary explanations can be heavy handed in their usage. However it is a bit odd to me that Pinker admits to our language acquisition being important and impactful enough that it was developed to the point of being a detriment to other essential capacities (such as breathing). Many other animals do not share a breathing/eating tube, and are in no danger of choking, as food lodged in the throat does not block their breathing systems whatsoever. This is a massive disadvantage that humans are at in terms of survival, and does show that language must have been highly beneficial to continue in the face of this adversity. I do of course see the point that it could have originated as a byproduct of some other element in our development, /and then/ was selected for after presenting itself. It just seems an odd point to make. Were it something such as two breasted humans, where one versus two does not make much of a difference I would appreciate the argument more. However the fact that language is so relevant and important it feels like less of a relevant point to make.
ReplyDeleteAnd if language first evolved as gesture -- and then migrated to speech because of the obvious advantages of gesture over speech -- then it would have been those advantages, and not the advantages of language itself, that were worth the risk of choking! (But there's an awful lot of speculation in all of this... It's the usual risk of telling Just-So stories inherent in evolutionary theorizing, especially about behavioral capacities. The problem of the evolution of UG, however, is special, and even more difficult.)
DeleteMy understanding from this reading and my peers' comments is that Pinker and Bloom's argument is that human language evolved by natural selection. The reason why this claim is not valid is because P&B did not address the question of UG. Evolution is lazy, meaning that we encode the ability to learn things rather than the thing itself. In the case of UG, the issue is that we cannot learn UG (because of POS)- it is one of the few things which is innate. So, UG cannot be evolved.
ReplyDeleteMy question is how this fits in in the broader context of this course. We have learned how to assess whether we have successfully reverse-engineered the brain. Searle proved that cognition is not all computation. We talked about the symbol grounding problem, that is the causal mechanism that gives us the ability to categorize (do the right thing with the right kind of thing). Understanding categorization and symbol grounding is synonymous to understanding cognition. Next, we saw evopsych, and how it is answering the why of the easy problem without addressing the how. In this reading, we see why evolution cannot adequately explain language - is language one of the few innate things in the brain that needs to be understood and encoded as a base in order for the rest of cognition to be possible in a machine? Is the key to symbol grounding and categorization in language, and hence once we understand the innate properties of language that exist in our brains, we would be able to reverse engineer the rest of cognition?
After reading "symbol grounding and the origin of language: from show to tell" (8b), here is my take on the question I posed after reading the paper for 8a:
DeleteCategorization is something that all species (at least those with similar brains as ours) do. Chimps do so, and are even able to combine the names of their categories into propositional strings. However, they do not produce language. Propositions was the result of the teacher attempting to teach new categories to students observing them. So, language evolved as a result of this teaching. Humans who were able to communicate via language outlived those who did not because they could learn new categories more effectively (via instruction) without needing to go through risky trial and error (induction). This is how it evolved.
However, I am still curious as to how UG fits into all of this: is UG what you refer to in the paper when you talk about our brains being "language-biased"? Is this language bias our seemingly innate ability to distinguish a correct sentence from an incorrect one despite POS?
It’s not that Pinker’s wrong that language evolved by natural selection. Of course it did. But he does not state the special problem of how (and why) UG would have evolved.
DeleteNo one knows; but his article suggests that it was thought there is a problem with how language evolved, whereas the only problem is with how (and why) UG evolved. (I think you have understood this.) All the things Pinker discusses (to show they are not a problem after all) really aren’t a problem; but UG is, and he does not address it.
The questions you raise are among the ones that make the evolution of UG a special problem (but it has a much better chance of being solved than “the hard problem”).
About propositions, It’s not at all clear whether the apes that have been taught sign language or other symbolic codes are able to make subject/predicate propositions (with truth values). They can string category-names together, but that’s not a proposition like “the cat is on the mat.” (Same question arises for the dog “Bunny” whom we discussed in class and skywritings last week.)
So the problem for an evolutionary explanation is UG: how did it evolve; why did it evolve?
Pinker, S. & Bloom, P. (1990). Natural language and natural selection
ReplyDeleteIn this paper, Pinker and Bloom argue that human language should be attributed to natural selection as it meets the criteria of having a complex design and a function (ex: grammar is a complex mechanism with function) and absence of alternative explanations. Their argument convincingly demonstrate that language is a distinct evolution as opposed to a by-product of evolution (or something that is historically invented by humans). There could be reproductive advantages to developing more complex or better language. Communication did develop progressively, mutations by mutations, but the question remains what makes language develop in us and not in other animals? The authors don’t seem to state the problem of how UG evolved. Pinker and Bloom talk about a problem with the evolution of language while the problem is really the problem of how UG evolved because some language capacity can be explained.
This may be a bit away from the main thesis of the Pinker & Bloom paper, but one issue I had with this paper was the tendency for circular conclusions, as exemplified by the early argument about "human artifacts" and the adaptive complexity of the eye. Pinker and Bloom write:
ReplyDelete"It is impossible to make sense of the structure of the eye without noting that it
appears as if it was designed for the purpose of seeing -- if for no other reason
that the man-made tool for image formation, the camera, displays an uncanny
resemblance to the eye."
and later state again that:
"resemblance to human artifacts fulfilling the same putative function give independent criteria for design"
While I agree with the authors - natural selection surely shaped our eyes - adding the resemblance of eyes to cameras as supporting evidence genuinely makes their argument less convincing. Of course our eyes seem/work like cameras; it would be more accurate to say that cameras seem like eyes - because we created them through reverse engineering, based on our own analog vision. Because eyes are analog sensors, it would make no sense to reinvent the wheel when we tried to make cameras (other analog sensors - at least the old ones were analog). We can't apply the constructive work of humans to imitate eyes to create visual sensors (cameras) as justification for eyes being shaped specifically to see just because we based camera's designs on seeing.
As described in many of comments above it seems, that while Pinker & Bloom brought back the evolution of language in the spotlights, they silently glossed over the evolution of Chomsky’s universal grammar (UG).
ReplyDeleteThey defend that language, and more specifically, “generative grammar” is a product of evolution just as our opposable thumbs are a product of natural selection.
The argument follows:
“language shows signs of complex design for the communication of propositional structures, and the only explanation for the origin of organs with complex design is the process of natural selection”.
They early on stop on the notions of exaptation, by which new uses are made of traits that were originally adapted to some other function, and gradualism, whereby evolution advances through many very small and slow adaptations. These notions are explained early on the paper almost as if expecting a retort on behalf of UG defenders and the paradoxes in evolution UG entails. How to explain that symbols, grammar, and more specifically UG evolve if not in an all-none fashion (I.e. symbols are discrete while the biological substrate of evolving beings are gradient)? To this, (vulgarly) they answer that language and grammar simply evolved like “other mental systems”. But where this may be enough to explain, although speculatively, many notions relevant to language (e.g. vocabulary or ordinary grammar) it cannot explain UG precisely because children are never taught the complex set of rules composing UG, nor can children really learned UG by themselves through unsupervised learning because they are never actually exposed to example phrases outside UG. Then UG must be innate, and more precisely, as explored in my comment in 7.a., UG must either hold of some universal law (because there is no feedback involved in its learning), or there must be as Chomsky suspects, aspects of its learning we are still blind to.
P & B’s vagueness concerning UG leaves us begging the question. I was however wondering if avenues had been explored around the relation between a “language of thought” (or pre-language) and POS - more specifically, if language of thought could provide enough basic stimulus to overcome the POS dilemma in order for him to make his first propositions within UG. Neuroscience has suggested, as we’ve seen in exploring neurons, that an infant would in fact be exposed to enough observation of causal phenomena as well as behaviour from his peers that he would be able to constitue some initial rule of UG. Since UG seems to concern only with syntactic rules this language of thought would have to echo some of these rule with forms of “pre-propositions”. Let me explain myself. Suppose we have a child not yet able of forming propositions. We cannot deny that although he is not yet able to formulate proposition he experiences and is aware of certain forms of “classification” and relationships. He will not confuse light and sound nor will he personify them into his mother to expect food, nor will he expect that one entails the other.
ReplyDeleteLanguage of thought would allow him to get a basic understanding of initial categorical boundaries and relations, where and if to differentiating A from B, of understanding sometimes that A causes B and other times that A cannot cause B (or even be in any way related to it) because A and B are not of the same type. He would then by the time he starts making propositions be exposed to enough events that some types of propositions could not be formed.
But again perhaps my limited understanding of UG is makes this question nonsense.
As many people have pointed out in this thread, while Pinker and Bloom argue that the ability to learn language is acquired through evolution, they do not address the issue of Universal Grammar (UG), which is the innate, underlying structure of all languages and which cannot be learned because of the poverty of the stimulus argument. If I am remembering my linguistics right, UG has to do with the internal structure of sentences -- and how every language has roughly the same internal structure (with certain elements switched around). If you draw a syntax tree for a sentence, for example, you will see that no matter the language, the trees looks similar, which is a reflection of UG. (Although there is some variation -- for example, some languages like English have verb then complement (e.g. "I go to the park") and some languages have complement then verb (e.g. "I to the park go")).
ReplyDeleteWe have been arguing that since evolution is lazy, it is easier to encode the capacity to learn things than it is to encode the raw capacity. But UG cannot be learned and must be innate. However, if UG represents a basic, underlying structure of all languages, whose structures remain the same but for a few small variations like the examples I have given above, couldn't it be argued that UG is in fact a kind of mechanism for learning languages, and thereby consistent with the notion of lazy evolution? In other words, UG is a structure which makes it possible to learn languages, and lazy evolution encoded UG to allow us to learn any language. We are all born with the innate structure which is UG -- this means you can put a baby anywhere in the world and it will learn the language of the place it grows up. Hence, UG allows us to learn any language, and this seems to me fully consistent with lazy evolution.