Blondin-Massé, Alexandre; Harnad, Stevan; Picard, Olivier; and St-Louis, Bernard (2013) Symbol Grounding and the Origin of Language: From Show to Tell. In, Lefebvre, Claire; Cohen, Henri; and Comrie, Bernard (eds.) New Perspectives on the Origins of Language. Benjamin
Arbib, M. A. (2018). In support of the role of pantomime in language evolution. Journal of Language Evolution, 3(1), 41-44.
Vincent-Lamarre, Philippe., Blondin Massé, Alexandre, Lopes, Marcus, Lord, Mèlanie, Marcotte, Odile, & Harnad, Stevan (2016). The Latent Structure of Dictionaries. TopiCS in Cognitive Science 8(3) 625–659
Organisms’ adaptive success depends on being able to do the right thing with the right kind of thing. This is categorization. Most species can learn categories by direct experience (induction). Only human beings can acquire categories by word of mouth (instruction). Artificial-life simulations show the evolutionary advantage of instruction over induction, human electrophysiology experiments show that the two ways of acquiring categories still share some common features, and graph-theoretic analyses show that dictionaries consist of a core of more concrete words that are learned earlier, from direct experience, and the meanings of the rest of the dictionary can be learned from definition alone, by combining the core words into subject/predicate propositions with truth values. Language began when purposive miming became conventionalized into arbitrary sequences of shared category names describing and defining new categories via propositions.
Subscribe to:
Post Comments (Atom)
-
Harnad, S. (2003b) Categorical Perception . Encyclopedia of Cognitive Science . Nature Publishing Group. Macmillan. http://www.youtube.co...
-
Searle, John. R. (1980) Minds, brains, and programs . Behavioral and Brain Sciences 3 (3): 417-457 This article can be viewed as an atte...
-
Fodor, J. (1999) " Why, why, does everyone go on so about thebrain? " London Review of Books 21(19) 68-69. I once gave a (perfec...
If all the essential cognitive components for language seem to have been in place in hominins 250,000 years ago, why didn’t they learn to use this highly adaptive skill earlier? I don’t believe that language was invented; rather I think that it would have been discovered, in that at one point in time, we realized that we had the ability to do better, and we were motived to do so for our survival. I wonder whether at some point if chimps would come to the same realization as our species did which would motivate them to cultivate new and more efficient forms of communication as well. If language started off as pantomime, that would have been useful for some communication, but we couldn’t convey new categories with that technique alone. That is why sharing the distinguishing features of members of new categories and the non-members verbally is so efficient and easy, such that we never looked back; although, we do start with categories stemming direct sensorimotor interactions before we can learn from others.
ReplyDeleteAlexander, whether language was discovered or invented, or it evolved gradually out of genetic precursors in learning capacity or communication capacity, it is clear that it has a strong genetic component today. Human species can (and all do) learn language; and nonhuman species cannot, even if they are extensively trained. This is true even though all the essential components seem to be there already, at least in primates, and perhaps in many other species. Whether what is missing is a cognitive component or a motivational component, it is clearly missing in nonhumans. And even in our own species it could not have happened overnight. The genetic changes, whether cognitive or motivational, must have been progressive, somehow building on one another.
DeleteThe analogy between language and tool use is a very superficial one, but one similarity is that the manipulative skills and motivation underlying tool creation and use would also have evolved gradually. Baldwinian evolution is an example of the “laziness” of evolution, in that it is not the capacity to create and use a particular tool that is encoded in our genes, but the capacity and motivation to learn it (and many others). (What is “Baldwinian Evolution”?)
The above was a reply from me (SH). I'm using CATCOMCON to reply for a while because something seems to be wrong with my regular gmail. Has anyone gotten from blogger a warning that you have exceeded your comment limit for the day?
DeleteTo answer the question above, Baldwinian evolution is an evolutionary theory suggesting that selected offspring tend to have an increased capacity for learning new skills rather than being confined to genetically coded, relatively fixed abilities. Environmental changes lead to a need for adaptation, and selected offspring can learn new skills to deal with the changes. Those confined to the previously fixed abilities will be naturally extinct, leading to future generations having increasing capacities to learn the adapted behaviors, which eventually will be an instinct.
DeleteFrom my understanding, Baldwinian evolution is the process by which something that was once learned from direct sensorimotor experience has now become something that is innately programmed to learn. This is because those who were the best learners for a certain capacity were selected for based on environmental conditions, i.e., were able to survive and reproduce more effectively with this capacity to be "pre-disposed" to learning than those without this ability. Thus, through Darwinian natural selection, the trait for learning effectively became somewhat innate as an adaptive capacity as its selection over time made it become something that was passed down genetically.
Delete^ To clarify the first sentence: something that we would learn just through happenstance sensorimotor exposure is now something that we innately look to learn about (we are tailored to attend to learning about something / a certain capacity).
DeleteDarwinian evolution is a form of lazy evolution in which there is a genetic tendency to learn certain things. What is learned is learned, not innate, but the inclination or disposition or motivation to learn it is innate.
DeleteLanguage learning is a good example. We are not born knowing language, but the motivation to learn it is very strong in children.
More generally, all learning is in a sense Baldwinian, in that what is learned is learned, rather than built in genetically. The genetic part is just the capacity or the motivation to learn it,
Do hand-gestures come from miming? Since teachers/mothers were motivated to teach their learners/offspring new categories, and sometimes had to resort to miming in the absence of the objects being learnt, it seems that it would have been a very useful to have this skill in cases where we cannot provide direct sensorimotor interactions for the learner. Normally, the teacher could just use the object and perform an action on it to teach the category, but if there is no object, one can engage in a kind of “pretend” instruction. I believe that this miming skill would have been brought along with our verbal language ability to learn more abstract categories or categories with items that are not present at the time of instruction. It could be that hand gestures are the leftovers of that evolved function.
ReplyDeleteAlexander, I’m not sure what you are asking. Gestures and miming are not the same as gestural language (sign language), but they might have been among its precursors (how?). (And gestural language might have been the precursor of vocal language.)
DeleteImitation might help you learn a sensorimotor skill; and copying or building might help you learn the features of an object category (rather like unsupervised learning); but categorization still requires supervised trial/error/correction with members of the category and members of the categories it could be confused with, to detect and abstract their distinguishing features.
Questions for reflection:
1. What is iconicity? How is it an advantage and how is it a handicap?
2. How would pantomime have helped to ground symbols in their referents?
3. What is the difference between the role of the “teacher” in supervised learning and in verbal learning?
4. What are the advantages of vocal language over gestural language?
In response to the reply questions,
Delete1. What is iconicity? How is it an advantage and how is it a handicap?
Iconicity is the property of symbols (eg gestures) being similar to their meaning, as in the example of pantomime or in some written/spoken languages still today. It is advantageous to relate symbols with meaning, but language mostly moved away from iconicity towards arbitrary symbols in order to make the sensory modality of category names irrelevant, allowing the power of language as a tool for category acquisition to move from subcortical areas to the cortex. Additionally, iconicity may be a handicap on cognitive resources since arbitrary symbol shapes are more easily able to be compressed into more basic or short shapes.
2. How would pantomime have helped to ground symbols in their referents?
Pantomime is iconic, and helped with grounding by using symbols shapes (gestures) that resembled the shapes of their referents (things in the world). As such, it eases the connection of symbols with their referents by ‘grounding’ this link in physical similarity.
3. What is the difference between the role of the “teacher” in supervised learning and in verbal learning?
In supervised learning, the “teacher” can confirm/correct the learner’s attempts to categorize, by saying whether a particular categorization is right or wrong. This constitutes feedback in a ‘trial and error’ fashion, which allows the learner to adapt their categorizations. Alternatively, the “teacher” in verbal learning directly describes the invariant feature(s) of the category of interest (ie the feature(s) that allow to distinguish all members from nonmembers), using only words that are already grounded for the learner. As such, a supervised learning “teacher” only provides feedback to a learner who then must learn categories by induction, while a verbal learning “teacher” actively describes/defines a category to a learner who can understand it simply by combining its feature elements (instruction).
4. What are the advantages of vocal language over gestural language?
Vocal language enables people to have free hands for other actions while communicating. It also allows people to name things and communicate these names even when the teacher is obstructed from the learner due to their being at a distance or in the dark for example. This facilitates category learning by instruction (verbal learning), which is an advantage since induction&instruction learning is more efficient than just induction learning.
Amélie, Excellent.
DeleteIMPORTANT: I've been noticing that blogger sometimes loses postings. Please, after posting, always keep a copy of your posting and come back later to check that it's still there (and whether there are replies). If it's gone, post it again.
ReplyDelete(I'm not sure what the bug is, and it doesn't happen often: maybe when several people are posting at once. But take take this precaution to be sure yours is not lost.)
I'm re-posting Mathilda's skywriting which I managed to catch before it disappeared, below. But usually I will not detect lost postings,
Delete“As discussed in class, we learn categories through direct experience or through verbal learning, described in this paper as induction or instruction respectively. These two types of categorizing processes add to our discussion of the symbol grounding problem: grounded ‘core’ words are learned through induction, which we then use to learn the rest of the words by definition/instruction, combining grounded words into subject/predicate propositions with truth values (like the example in class with the Venn diagram: zebras are horse-like and striped). An important point to understand is why verbal instruction evolved in humans: according to this paper, it relies on the motivation of a ‘knower’ to intentionally help its kin learn new composite categories. Indeed, it seems that the lack of the development of language in other species, like apes, is not a problem of intelligence but rather a lack of this motivation/impulse to name and describe things to others. Once these categories are known and grounded by both the teacher and the learner, their names then start shortening, becoming less iconic and therefore more arbitrary. The advantages of this type of category learning, including saving time, energy, and avoiding the risk and uncertainty associated to direct sensorimotor induction, made it favorable and a now trademark trait of the human cognition.”
DeleteVerbal learning is much faster, easier, safer and more powerful than sensorimotor learning (unsupervised or supervised). But sensorimotor learning is essential too. I helps ground feature-names and verbal definitions are approximate. “A picture is worth a thousand words (of description).”
DeleteMelis, good points.
DeleteVerbal learning is powerful in that it allows one to quickly develop new and abstract categories (so long as our already grounded words can reach them so to speak), but also more powerful in that it allows communication and consensus, which would be very difficult if we only had sensorimotor learning (I can make propositions for others to agree or disagree with, but with sensorimotor learning I can only show what I have learned by using the other's mirror capacities as I do the right thing with the right kind of thing).
DeleteThe authors define (natural) language operationally as a symbol system in which one say can say anything that can be said in any other language. Languages have, in addition to syntactic rules, a "semantic constraint", in the sense that words are also manipulated according to their meanings.
ReplyDeleteThey then argue that we learn the meaning of content words (i.e., words that name categories) through direct sensorimotor experience and/or instruction, tracing the origins of language to our ancestors tapping into the power of acquiring and conveying new composite categories by instruction.
My question concerns the other side of this equation: how did the syntax of natural languages (esp. Universal Grammar) evolve? At some point the authors claim that " all the essential cognitive components [for language to develop] already seem to have been in place in other hominins 250,000 years ago". Does this also include UG? If so, why would UG have evolved before full-blown language? What function did it perform back then, and what evolutionary advantages did it provide?
Gabriel, good questions! Although there is only one "the hard problem," the evolutionary origin (and adaptive function" of UG is definitely "a" hard problem.
DeleteThis will be one of the (optional) readings next week:
Harnad, S. (2008). Why and how the problem of the evolution of Universal Grammar (UG) is hard Behavioral and Brain Sciences, 31(5), 524-525.
Gabriel's inquiry about how UG may have evolved before language was very intriguing to me. If I were to put forth a wild guess, I think it's possible the early purpose of UG was to be a flexible mental structure and rule set to help us form connections between concepts and describe features of categories (easing the feature detection process).
DeleteThe UG theory suggests that every language shares some core fundamental rules and these shared rules is what makes up UG. These rules are not clearly defined but (from what I understand), these rules tend to concern relating concepts. Having innate rules to connect concepts is still useful even if they aren’t being used to communicate with others. Having these internal set of rules could be used to understand similarities between different situations, helping us making a decision on what to do despite being in novel situations.
The ability to learn language is the greatest distinction between human and nonhuman species. Language is what allows us to communicate with others, and most importantly, is what allows us to use verbal learning to accelerate the process of learning categories. If UG was originally used to help us learn and connect different categories, its not hard to see how language could be a super powered extension of that ability.
Gabriel put the questions I had into words. The reading covers the evolutionary component of language as a subset of communication, as well as vocal communication, however the evolutionary origin of UG was not outlined, so I assumed that UG was something that goes beyond language in the evolutionary sense, and as Sophearah said, language can be seen as an extension or a result of UG. If it is inborn, how does it work?
DeleteI agree with Stevan that the origin of UG definitely seems to be '"a" hard problem,' but I thought I might share some more context from linguistics especially as the paper under discussion briefly references minimalism. Under the most "current" research program in generative linguistics, the Minimalist Program, syntax is autonomous (in the same sense as used in the paper), and produces a hierarchical structure of symbols which is then passed at an interface to the articulatory-phonetic system (which makes the utterance pronounceable) and the semantic system (which computes the meaning). These two systems impose constraints on what the syntactic module produces, so that if it doesn't turn out to be in a form that is both semantically computable and pronounceable in a particular way, the derivation will "crash" and the utterance will never be produced. So in light of the 8a discussion where we mentioned we never make UG errors (except consciously as linguists), we may actually make them all the time, it's just that they never see the light of day as we never actually pronounce them - we'll continue the derivation process until we actually generate a good sentence. The "Strong Minimalist Thesis" (SMT) is basically that UG is just the computational principles that allow for syntax (symbol manipulation) to meet the requirements of the articulatory-phonetic and semantic modules with the greatest computational efficiency. (The Wiki page for the Minimalist program goes into a bit more detail.) This of course poses more questions as to what these requirements are, some of which are being investigated and formalized in further linguistic inquiry. But this might help to consider why UG might have evolved: to leverage our capacity for categorization and our motivation for learning and expressing new categories in a systematic (thus universally understandable, subject to particular "parameter settings" of individual languages) and highly efficient manner. Since our mirror neurons likely had already evolved and were in use for these tasks already, UG at its origin likely leveraged those as our gestural and vocal capacities are relatively well suited to imitation.
DeleteMelis good question. By "inner language" do you mean talking to yourself in your head in words? Or thinking "in words"? Either way, I think external language would have to have started first. Without it there's nothing to talk about or think verbally about, because there are no words (why would there be?). Surely words are first invented as well as learned in an external communicative context. And if gestural language came before oral language, and pantomime came before gestural language (including naming), then all that is external too.
ReplyDeleteInteresting question: Do congenitally deaf people who have never learned oral language at all think in gestures? What if they have also learned an oral language?
Of course it's an old debate, whether we think in words, or do the words verbalizing the thought come after the thought? And many people say they don't, or hardly, think in words at all...
In response to your last point, I am of the belief that words verbalizing the thought come after the thought. I believe this relates to the hard problem and understanding how we are able to feel. Feelings go beyond language, and we are only able to gather the words to express our feelings after we have already experienced the emotion, rather than before or simultaneously. Other thought processes may follow this same order, in which we search for language in attempt to communicate what we are experiencing internally.
DeleteAs years go by, I am more and more aware of how much of what I do is really served for me by my brain on a plate. (The feeling of free will is a core part of the hard problem.)
DeleteMy swkywriting that disappeared was saying that this paper made me think of "internal language" both in terms of thinking in words and the voice in you hear head. I am wondering if internal language began with external language (for communication), would it just be a byproduct of language. Does internal language have an evolutionary goal?
DeleteHi Melis,
DeleteIt is a very interesting problem. Given that language may arise from gestures via a gestural protolanguage, how can our ancestors talk to themselves with gestures? I wonder is it possible. Or, ‘internal language’ might be something that is exclusively unique to speech language – after the turning to speech has happened.
Internal language may be a half point between thinking and speaking out loud. The fact that internal thinking can be semi-verbalized as internal language in humans’ minds demonstrates that humans’ thoughts can be verbalized as propositions with truth values, which languages represent. This could be evidence that humans’ cognition is propositional.
One obvious evolutionary benefit is that this could let humans be more aware of their every step of thinking, so decisions can be more carefully made, thus humans can survive better.
Han, it has been long debated whether thinking is language-like or image-like or both. (Images can be visual, auditory, tactual, gustatory, olfactory, motor, and even affective.) There’s no reason deaf signers can’t think in gestured words the same way speakers think in spoken words (when they are thinking in words). Musicians can read musical scores and hear them in their heads; composers who become deaf (like Beethoven – but not people born deaf) can compose even though they cannot hear. Pianists can rehearse their pieces in their heads. Athletes can and do rehearse their skills in their heads, especially when they are convalescing.
DeleteBut, as Chomsky says, thinking is not the same thing as speaking, or signing, or writing. These are all ways of expressing thoughts externally, to communicate them to others – yet another mirror capacity?
This discussion of “internal language” and what thinking looks like: if it’s image or language-based, makes me wonder about the tip of the tongue phenomenon. That is, when you can “see,” for lack of a better word, the word or phrase that you’re looking for in your mind’s eye, and know the definition and words that sound like it or are related to it, but are unable to recall the exact word you’re looking for. It’s interesting to me that one can be aware of the particular category they’re looking for but unable to recall the label attached to it.
DeleteI found this article very easy to understand, perhaps because we already touched on most, if not all, points written in this. The author defines that language is a formal symbol system, for which you can express any preposition with such symbols. Furthermore he gives a very good explanation of why natural language has a symbol grounding problem (as the question on the midterm). Lastly, I really liked the analogy made with verbal learning and ‘show and tell’. Steven explains that ‘show’ is like learning through our sensory system (unsupervised or supervised learning), such as the mushroom example or the rabbit vs hares. And ‘tell’ is simply verbal learning, where you have no consequences and instead is just told the features of such a category, like reading a definition of a word in the dictionary. As we learned in class, even though ‘tell’ is great and takes less time, perhaps the reason why we evolved such capability, life can not be lived by just ‘telling’ and instead some ‘showing’ needs to happen.
ReplyDeleteI particularly liked the way this article addresses the symbol grounding problem. The paper says "As with so-called protolanguages, it can't be definitions all the way down."
DeleteSo far, I don't think this exact definition of the symbol grounding problem has been mentioned, but I wish it had come up earlier, because even though it's extremely similar to what we've been discussing so far, I think this is the first time it's really clicked in my brain. I really like the visual of eliminating words in a dictionary which aren't used to define other words, and then thinking about how the meanings of this new set of words are acquired. Combining this with the edible/inedible mushroom example, it's interesting to think about the grounding of concepts that cannot be verbally explained.
Vitoria, even though, in principle, once you have language and you have learned a minimal directly grounding set (MinSet) with which every other category can be learned through definition or description (from a textbook or from someone who already knows the category and its features), that does not mean we stop grounding new categories directly too. But direct supervised learning of categories by trial, error and consequences (e.g., mushrooms on a desert island) is rather more than just a matter of “showing”!
DeleteViva a derrota do JB…
Teegan, remember the task of remembering the name of one’s 3rd grade schoolteacher (as an illustration of the fact that so much of what we can do, we don’t know how we do it, and are waiting for cogsci to reverse-engineer it and then tell us how)?
Well, when we learn new categories directly, by sensorimotor induction (unsup and sup), we are learning the category itself, therefore implicitly our brains are also learning to detect and abstract its distinguishing features. But that does not mean we know its distinguishing features explicitly (which usually means verbally, or verbalizably), especially if we have not learned and named the feature directly as a category (from its own distinguishing features).
What’s probably true, though, once we have at least one directly grounded Minset (plus probably a lot more), is that more and more of the features of the new categories we still learn directly will have features that are already named categories.
I found a similar point to Vitoria (above) which was that this reading was able to encompass many of the points that we have been touching upon the past couple of weeks. The distinction and relevance of both induction (direct experiential learning through sensorimotor experience) and instruction (verbal learning) was made very clear in this article, and despite (perhaps) the majority of our learning to categorize coming from instruction, induction lies at the foundation of our ability to categorize. The manner in which dictionaries were reduced to their “grounding kernel” (words that can define the rest of the words in a dictionary) and how that showed to be approximately 10% of words in the dictionary was very interesting to me, especially once we consider that only about 1500 categories would need to be learned through induction. However, as stated in this paper, reaction times for categories learned through induction were shorter than those through instruction, which leads me to believe there must be some sort of advantage to learning through direct sensorimotor experience (other than its necessity in regards to symbol grounding). My belief is that much more than the baseline limit (of 1500 categories) have been learned through induction, as not only is this means more practical, but would also account for a young child’s extensive categorical learning in a short time period.
ReplyDeleteKarina, I'm sure you're right. Even if in principle once you have directly grounded a MinSet directly it would be possible to learn everything else indirectly through verbal instruction, in practice we continue grounding new categories directly by sensorimotor induction too, lifelong.
DeleteHi Karina, I also believe that there must be some sort of advantage to learning through induction. I was surprised by the results of the artificial life simulation experiment, in which the virtual creatures that tended to learn by instruction outsurvived those that tended to learn by induction, within a few generations. These results showed the adaptive advantage of instruction compared to induction when learning categories. However, I am surprised by how fast the out-competing occurred, because I think that learning through direct sensorimotor experience provides learners with a deeper understanding of the category in question, which should result in some sort of advantage over instruction-tending learners. It seems like evolution chose quantity over quality; learning indirectly allows for a greater number of categories to be learned with less risk, although maybe not with as much understanding as those learned directly.
DeleteWhile people tending to learn by instruction outsurvive induction-tenders, they still require some categories (~1500) to be learned through direct sensorimotor experience. This reiterates the idea that T3 capacity is needed to reverse engineer cognition and explain how we can do everything we can do. T3, unlike T2, can directly ground a MinSet, allowing other categories to be learned through instruction and be more than just meaningless symbols.
I thought this article was well-structured and well-written (kid-sib friendly :). In order to affectively answer the larger question at hand, the authors first tackle the question of “what is language?” as a way to work backwards and first understand the rules of language, why it was created, and then how it could have possibly evolved. Through the explanation of symbol systems, the authors make it clear that our ability to categorize encompasses most of our cognitive capacities and is crucial to our ability to learn and in return, survive. The sharing of knowledge, postulated to first have been done through gesturing, is attributable to our social natures and our motivation to teach our offspring how to survive. The benefits of vocal language made the process of instruction much more efficient, allowing for the free-ing up of hands for other tasks and also to allow listeners to communicate when the instructor is out of visual or auditory range. Though this vocal instruction is highly beneficial and now considered crucial in the 21st century, it could only exist subsequent too and in cooperation with learning through induction.
ReplyDeleteLaura, be careful not to conflate the pantomime/proposition distinction with the gestural-language/vocal-language distinction. They’re not the same: What are this two pairs of distinctions?
DeleteThrough this article, I was able to clarify some of these distinctions that were previously fuzzy for me regarding the theories of the evolution of language. I believe that these are the distinctions Prof Harnad is asking about:
DeletePantomime/Proposition:
Pantomime is the imitation of the real world through iconicity, which means the gestures are similar to what they refer to. Pantomime is not language.
A proposition is any statement containing a subject plus a predicate with an assigned truth value (true or false). Propositions are part of what makes language special, constituting more arbitrary symbols and removing the sensory modality of category names (essential to pantomime).
Gestural-language/vocal-language:
Gestural language is sign language (ASL for example) and vocal language (dutch for example) are both symbol systems used to communicate each and any proposition (see above). The difference lies in what part of the organism is used: gestural language involves the hands and certain facial expressions whereas vocal language involves vocal cords and face movements. These are two different forms of language.
Kayla, that's it.
DeleteI think this paper summarized the evolutionary benefits of learning through instruction better than 7a. Though I still have trouble understanding how the advent of language came about, it is more clear how it spread and became evolutionarily advantageous to share knowledge through instruction rather than induction, which involves a lot of trial and error. Acquiring categories by instruction was way more efficient a method and enables further categorization. Through language, it became possible to give offspring a better chance at survival, by equipping them through verbal instructions rather than releasing them unequipped in the uncertain world. It also is an advantageous capacity in the context of globalization and migration, in which offspring and communities were more and more separated. However, if categorizing is intrinsically tied to doing, and if there really is a “grounding kernel,” then I struggle to comprehend how a word like “peace” is defined (I apologize for commenting again on the problem of abstract words). Sure, it is a content word and not a function word, but defining “peace” does depends on understanding a lot of other subcategories. Peace typically involves more than a single party, and involves stability, harmony, and non-disturbance, all of which are not really tied to doing properties. How does one come to understand this notion even through instruction? Is there ONLY induction and instruction or can there maybe be also an influence of innate feelings that goes into learning new categories and defining words?
ReplyDeleteTess, if you’re not sure how language began, welcome to the club! There are lots hypotheses – some just Just-So stories, some more plausible and perhaps closer to the truth – but it’s impossible to go back and test it; once it happened, it happened (although some theorists think it might have happened multiple times separately around the same time), and that was long ago (between 50,000 and 250,000 years ago). It was followed by a period in which many genetic adaptations occurred in our brains, first to language and then to speech. The result (also long ago) was that thereafter every homo sapiens was born with a brain specially prepared for language.
DeleteNeither the advent of language in (1) the developing child, nor in (2) congenitally deaf communities, like the one in Nicaragua , which brought together individual deaf children from speaking families who had not been taught sign language, and within 3 weeks the deaf community had jointly invented the beginnings of a sign language and was using it to talk to one another as an oral community would.
In both (1) and (2) all the children already have brains specialized for language, so they cannot reveal how language began, before the adaptation 100,000 years ago.
Senghas, A., & Coppola, M. (2001). Children creating language: How Nicaraguan Sign Language acquired a spatial grammar. Psychological science, 12(4), 323-328.
https://siegler.tc.columbia.edu/wp-content/uploads/2020/08/Children_Creating_Language_How_Nicaraguan_Sign_Lan.pdf
About “peace” and other abstract categories: You ask how we can learn them, either by induction or instruction. Here are some questions to ask yourself.
Do I know what “peace” means? (If not, then we are talking about an incoherent notion; so let’s assume we (or someone) do know what’s an example of “peace” and what is not.)
Knowing means to be able to distinguish peace from non-peace (e.g., war). There are distinguishing features. And it is crucial to have examples of members and of non-members (members of other categories with which “peace” could be confused).
Dictionaries define “peace.” The definitions – like all definitions, but also all categories learned directly by induction, through feature-detection and abstraction – are approximate. You can always tighten the approximation with more features. And if you can’t agree with someone, then there are ambiguous examples for which there is no correct/incorrect: Everything on earth is not either a member of “peace” or of “non-peace.” A mushroom, for example, is neither.
But if you don’t know how to distinguish examples of peace from non-peace, or you can’t define the distinguishing features, that does not mean there aren’t any (if there exists a correct or incorrect response for each relevant example). All it means is that you should keep trying (if learning directly by supervised induction) or find more features (if the verbal definition is too approximate).
What none of this implies is that “peace” is “innate.” That would be Jerry Fodor’s “Big-Bang” theory of the origin of categories.
It feels different to think of different categories, but category membership is a not a matter of feeling (except maybe for what we find beautiful). There has to be a correct or incorrect if there is a category (e.g., edible mushroom); and you have to sample both members and non-members, because the category is based on the features that distinguish them.
There are some innate categories, because we are born with evolved feature-detectors (e.g., colors). But look in a dictionary to estimate how many of our categories are likely to have innate feature-detectors. “Peace” is certainly not one of them.
SUGGESTION: It would be helpful to think aloud about the (hypothetical) transition
ReplyDelete(1) from learning categories through unsupervised and supervised learning
(1’) and purposive communication through pointing and pantomime
(2) to learning categories through propositions (grounded subject/predicate descriptions or definitions of the distinguishing features of new categories), given by someone who already knows the new category to someone who does not.
Some of the pieces to sort out are the transition
(a) from iconic gestures (imitation) to shared arbitrary gestures
(b) from showing to telling (i.e., “the propositional attitude”: describing rather than depicting)
(c) from pointing to referring
(d) from gesturing to speaking
What are the advantages of first starting to evolve language gesturally rather than vocally?
Hi Stevan! I won’t pretend to fully understand why it is advantageous for language to start with gestures then evolve to speaking, but I will take a stab at the transition from gesturing to speaking.
DeleteGesturing is limited in its faculties, for example you cannot hold things and communicate via gesturing, as you can with speaking. Therefore speaking proved to be a more efficient means of communication. It was also discussed in the article that category learning via instruction (someone telling you which category is which) is much more efficient then learning via induction (trial and error). Again, speaking allows for much faster instruction then gesturing, it allows you to discuss things that are not actually there. For example, you can teach me the difference between poisonous and nonpoisonous mushrooms by describing properties of it without me ever having to see the mushroom. Whereas with gesturing/showing, you would have to show me the actual mushroom in order for me to learn it myself. We can also discuss abstract categories with more ease than with gestures. Speaking allows for more effective / faster category learning than just gesturing, therefore proves advantageous to survival and would have been selected for.
((I will note that I’m not sure how sign language fits into all this… I know it has all the propositional attitudes etc. of spoken language, did it come after spoken language or evolve separately?))
Hi Sophie! I think that early sign language, simply a more iconic form of natural language without any vocal sounds, could have either developed after spoken language or concurrently. Like Stevan says, it's impossible to know exactly how natural languages began. Therefore, I believe it's possible that either:
Delete1) Sign language developed in populations where our brains were already genetically prepared for natural language, making it relatively easy for deaf people to develop, learn, and comprehend signs.
2) Sign language was born out of gestural communication, like any other natural language, concurrently with other natural languages, in a population that could not converse vocally for whatever reason.
Upon further thought, I believe that (2) is unlikely because neuroanatomical imaging demonstrates that language abilities migrated from a gestural to a vocal modality and this would not be a general principle of brain organization if language had remained purely gestural for certain populations.
Sophie, I think you are mixing up gestural communication (pantomime, imitation, “showing”) with gestural language (sign language, “telling”). These are two separate questions: the transition from pantomime to propositions is about the transition to language.
DeleteThe advantages of spoken language over gestural language are a separate question. Many of the features of spoken language that you described (which ones?) are also features of gestural language: they were advantages of language, not just speech.
Communication by gesture and pointing is already a big change from other nonhuman communication systems. There is controversy about whether nonhuman species can even point at all (though they probably can). Nonhuman species can imitate, but it is unclear whether they use imitation deliberately to communicate.
The big gap is between showing (pantomime) and telling (propositions), between gestural communication and gestural language. Once you have propositions, the advantages of expressing them orally rather than gesturally are many, but they are advantages of one medium of language over another medium of language.
The visual world is rich, and so is the scope for copying it for communicative purposes. The auditory world is impoverished, compared to that. How much that happens in the world can be mimicked vocally? (Start with mimicking daily facts vocally rather than miming them visually.)
Most species can learn to categorize (do the right thing with the right kind of thing) by unsupervised and supervised learning. So humans already had countless categories long before language. Once learned, their categories had no “names,” but many could be copied or imitated gesturally. How many categories could be copied or imitated vocally? Imitation (whether gestural or vocal) preserves the connection between the act of imitation and the object or action that is imitated. It is a mirror-capacity.
Once categories (food, eat, take, give, you, me, he, she, and specific individuals) are imitated (whether gesturally or vocally) many times, over and over, by the same small group of familiar people (extended family or tribe), the exactness of the imitation, its iconicity, becomes less and less important. The imitative gestures or vocalizations can become increasingly simplified and arbitrary while still retaining their familiar connection to the object/action/event categories of which they are the imitations.
But the most important thing is that if somehow (not yet explained) the imitation (object/action “showing”) becomes propositional (subject/predicate “telling”), the original iconic link of similarity to the category becomes the link to the original referent of the now increasingly arbitrary symbol, the word.
The question is whether this transition is more likely to begin in the (rich) visual-gestural medium of visual categories or the (sparse) auditory-vocal medium of auditory categories. Once the transition from iconic imitation to arbitrary propositions has taken place, however, the advantages of the vocal medium over the gestural one are obvious. “Arbitrariness” means it no longer matters whether the words are gestured, spoken, or written.
(Have a look at the Rizzolati mirror-neuron paper (Week 4) and the Arbib paper on gestural origins of language.
One of the many advantages of first starting to evolve language gesturally rather than vocally is that gestural is a more natural way to depict events. It also gives nonhuman primates the ability to communicate flexibly and intentionally. Also, gestures convey meaning globally, relying on visual and mimetic imagery. On the other hand, speech conveys meaning that relies on words and meaning. Thus, I believe vocal language is more constrained in some way compared to gesture. Although it is evident that language first evolved gesturally rather than vocally, it is still used daily. It remains an essential aspect of language, as it was said at the end: "After all, direct sensorimotor experience -- rather than just indirect verbal hearsay -- is, bottom, still what living is all about, even for Homo loquens. ".
DeletePolly, I think you are assuming that language came before sign language (gestural language). But we are now pondering how language itself began. And the gestural hypothesis is that it began in the transition -- from pantomime (which is communicative, imitative, iconic gesture but not language, because it is not propositional) -- to propositions.
DeleteThe gestural-origin hypothesis is that the transition occurred in the gestural modality, first by a gradual reduction in iconicity toward arbitrary gestures, shared by familiarity and habit (1), but preserving their originally iconic link to the categories they formerly resembled (2), then the adoption of the “propositional attitude” (whether that is cognitive or motivational, or both) (3), which was the first propositional (but still gestural) language (4), which, in turn, now freed of iconicity, could “migrate” to the much more efficient vocal modality. (This is the difference from miming (showing) myself eating a banana, to signing (telling) “I eat banana” – or, more likely, “I want to eat banana”…)
There are a lot of “Just-So” conjectures in that gestural-origins hypothesis, but the alternative vocal-origins hypothesis, that language began directly in the vocal modality, would require far more conjectures, because it lacks the precursors of the word/referent connection left over from the icon/object link in pantomime. You can’t get much of that from just vocal mimicry of sounds without the richness of visual similarity to objects.
(I think the reason we tend to believe that language had to have started with speech rather than gesture is just a cognitive (and Baldwinian) bias resulting from the genetic and neurological changes that implemented the migration from the original gestural modality where language began to the much more efficient vocal modality where it ended up. But read Rizzolati as well as Arbib and make up your own mind.)
Nadila, you are right about the richness of gestural imitation as the starting point, compared to vocal imitation, and you are certainly right that we still gesture (and point) when we speak, as a supplement, or complement. But our gesture is not propositional, as it is in the gestural language of the deaf.
If you want an example of propositional “gesturing,” it’s writing (and semaphore). And images (and videos), though we use them too, to supplement and complement our words (e.g., in a powerpoint presentation) are not language.
Yet it’s true that “a picture is worth a thousand words.” What does that actually mean?
In the article, the authors suggest that many cognitive components preceding the emergence of language were already present in hominins 250,000 years ago and in other intelligent social species. Somehow, humans were the only species that got to transition from gestural to vocal language. I wonder if the capacity to be self-aware also has to do with said cognitive components. For example, the capacity for pointing and having shared attention means that the addressee must understand that the addresser's signal (verbal or gestural) refers to a certain distal referent and that the referent is relevant to the addressee. The readings from week 4 (mirror neurons) also suggest that human speech is mostly a dyadic, person-to-person communication system. In contrast, animal calls are typically emitted without a well-identified receiver. In other words, in human speech, there is often an intended audience. This communication system suggests that the participants must understand their positions in relation to the signaler and the referent that the signaler refers to. This cannot be done unless a participant can regard himself or herself from a non-first-person perspective (the understanding that he or she is being addressed to). For instance, among humans, we can look at the object someone is pointing at by understanding that such an action is intended for us (i.e., when someone points in a direction in the hope of showing me something, I naturally look towards that direction); and that his action (gesture or vocal output) refers to something relevant to us (i.e., potential danger). In contrast, species that do not have sufficient cognitive components, instead of directing their attention to the object pointed at, their attention seems to fixate on the gestural signal itself rather than what it refers to. I wonder if this difference in cognitive ability also underlies the ability of some species to recognize themselves in a mirror.
ReplyDeleteYucenyes, shared attention, perspective taking and mirror-capacities are important precursors of language. But they are also important precursors of purposive gestural communication. And purposive gestural communication is not language (why/how not?); it is just pantomime.
DeleteIf by purposive gestural communication we mean any gestures that do not meet our criteria of being a (gestural) language, the criteria in question are the abilities to express anything that could be expressed in any language, and to be able to do so by way of propositions. Simple gestures without the UG-backed structural and representational power of "full" language (using arbitrarily-shaped symbols) cannot express everything we might want to say about things that are not present in front of us spatially or temporally.
DeleteIn section 5.1, the authors ask how come the chimps are not utilizing or developing their ability to systematically combine symbols and define new categories. The given answer to that is not that they lack the cognitive capacities to do it but that they do not have the motivation for it.
ReplyDeleteThis is an interesting aspect to me. If they don't have the motivational capacity for it, then how come humans do? Is it because we have fewer physical abilities than they do? Or is it something entirely different? Since I can hardly see why evolution would keep an unneeded trait (evolutionary spandrel)?
This is a good point on motivation with chimps and humans, however, I question whether we do indeed have fewer physical abilities than chimps.
DeleteFirstly, in the reading, it had mentioned that an aspect arising with the evolution combining symbols and defining new categories would be the idea of us being more social, cooperative, collaborative, and kin-dependent- nature which in turn may have caused some of us to find the potential of learning categories through instruction rather than just induction, at first passively, randomly, and without the aid of any genetic predisposition.
So with that in mind, I want to argue against the claim that “motivation is an ‘un-needed trait’”. Specifically, our actions are on the basis of motivation - one major theory being related to external motivation, i.e.., our beahviour is motivated by a desire for external rewards. In this sense, regarding motivation for humans, since we may encompass much more in relation to our behaviours on a daily basis compared to chimps then not only do we have more physical abilities but more or less also have more motivation going on in our lives. Thus, in the case of being more social, cooperative, etc. as mentioned in the reading it would make sense that the potential of learning categories did, in some ways, relate to motivation desires because it must have initially provided advantages in life to early human species, which eventually evolved to where we are today.
Alexei, the things we have the capacity to learn to do (but don’t) are not spandrels. They’re just the (huge) legacy of potential that lazy evolution has provided for us through the evolution of the capacity to learn. (These unlearned but learnable things are analogous to all the potential categories we could learn, but haven’t, at least not yet; also the infinite propositions we could speak and think, but haven’t, yet.)
DeleteBut you’re right that if the difference between humans and nonhuman species, as far as language is concerned, is not cognitive but motivational, then there still remains a Just-So Story to be told about how and why our species was the only one that started to use the cognitive capacity, so that the enormous advantages it confers could then go on to select, by Baldwinian Evolution (what is that?), the species-specific and powerful motivation (coded genetically and neurally) that we now have, and monkeys don’t, to learn and use language.
Maira, even if the difference was motivational rather than cognitive, that does not mean it was an “un-needed trait.” Maybe the right thing to do is to consider motivation a cognitive trait: Baby ducks are born with the motivation to follow the first moving form they encounter after hatching (normally, their mother). Baby snakes and turtles do not have that motivation, though they can both follow, if need-be. (But don’t conflate social motivations in general with the motivation to communicate – and to learn to communicate – propositionally: That seems to be unique to our species.)
Re: What are the advantages of first starting to evolve language gesturally rather than vocally?
ReplyDeleteSome advantages I can think of are:
1. Initially, the schedule for organisms functioning is active when there's lights and inactive (sleep) when it's dark.
2. Day-time imitation of certain things are more obvious to another human if he/she is exposed to these certain things before.
3. Gestures are relatively general, well-understood.
Additionally, through the sensorimotor interactions with the world, gesture-development sets the foundation for us to be able to ground meanings once we have language.
DeleteIn response to: what are the advantages of first starting to evolve language gesturally rather than vocally?
DeleteI agree, Monica! Gestural communication was an advantageous beginning for natural language because it enabled sensorimotor grounding through vivid iconicity. Below, I try my best to sort out some of the pieces that Professor Harnad mentioned earlier in the forum:
a) we transitioned from iconic gestures (imitation) to shared arbitrary gestures by way of our mirror capacities, which link visual perception and motor production, enabling others' gestures to be understood and imitated in experiential relation to our own gestural repertoire.
b) we transitioned from showing to telling because oral propositional statements allow more sophisticated categorization. Since cognition is mainly categorization, this allowed us to gain adaptive advantages over our ancestors and other species.
c) we transitioned from pointing to referring because, as our categories became more sophisticated with "telling", we could no longer communicate with each other effectively by merely pointing to the superficial qualities of a thing.
d) we transitioned from gesturing to speaking because, as others have mentioned above, communicating orally frees up our hands to do other tasks, permits communication in the dark, and enables communication at a distance. Thus, the neuroanatomical seat of language migrated accordingly (i.e., from the gestural to the vocal modality) in our brains over time.
I'd be very open to hearing if others agree because I'm not 100% sure about the accuracy of these distinctions.
Monica, please see other comments and replies. You are still not distinguishing purposive gestural communication (pantomime) from language (propositional communication).
DeletePolly:
a) Mirror capacities are needed for imitation and communication, but what (if anything) do they have to do with the loss of iconicity?
b) as you say, correctly) in d), vocal language is faster than gestural, it frees your hands, and it can be done in the dark, and from a big distance. That has nothing to do with “more sophisticated categorization.” The third way of learning categories, verbal instruction, is just as advantageous if the instruction is vocal or gestural (except if you need to have your hands free…)
c) Once the categories in the predicate defining the distinguishing features of a new category are grounded, there’s no need to point to anything (i.e., show); the grounded words already pick out their referents (the distinguishing features), so telling is enough.
Hi Stevan, thank you for pointing it out, I went back to the notes and other comments and I think I got it now!
DeleteThe circular dictionary exists when describing complex emotions (depends on whether we have felt similar emotions before), debatable colours (greenish blue)... I guess grounding by having some words bounded via a sensorimotor way, it breaks this circle.
ReplyDeleteMonica, I'm not sure what you are saying here. (Please say it clearly enough for kid-sib!).
DeleteThe dictionary is circular for all words unless you already have enough grounded words in your head to understand the definitions.
Words referring to emotion categories are special, because of the other-minds problem, but mirror-capacities and shared experiences can disambiguate them enough to be able to talk about them.
Among secondary colors (like crimson and scarlet) there is, in the greenish-blue range, “cyan,” and “teal,” but only artists have learned to pick them out reliably. They are combinations of primary colors, and their feature-detectors would be combinations of cones.
I found the part of the article that discusses the experiments with primates being taught how to communicate interesting. It highlighted that maybe the ability to use language in humans compared to chimps to communicate ideas isn’t necessarily a matter of degree of intelligence, but more a matter of motivation. That shows that maybe language exists because humans have a disposition to learn because initially, the human beings that had that disposition to learn were more adaptive in the process of natural selection. This is how our species ended-up with a “language-biased” brain, as said in the article.
ReplyDeleteI do wonder whether "motivation" can't be construed as something fundamentally evolutionary. Technically most organisms have the capability to perform other abilities like the species closest to it; but they don't do it out of what we could call motivation. Could the disposition towards using prelinguistic mechanisms and category usage not have arisen from genes as much as humans' modern throat physiology?
DeleteI took a class on human motivation last year, and although they didn’t mention natural selection specifically, it has been shown that motivation with animals is a significant determinant in terms of their survival. For example, animals that are motivated to seek resources, shelter and procreate more regularly have been shown to have increased odds of survival, which may seem obvious, but in some dynamics within ecosystems, this isn’t always that straight forward. Although we might think that animals spend all their time doing these things, they actually engage in other, less essential activities as well, similar to how humans might watch T.V or socialize with friends.
DeleteSee other comments on motivation. What is Baldwinian evolution, and how is that related to motivation?
DeleteI enjoyed this article and found it very clear. It discusses language acquisition throughout the years as well as categorization. It separates the way organism learn categories in two ways. First, direct experience or induction. From what I understood, these types of categories are feasible by every organism as they are acquired through interaction with one’s environment. On the other side, humans are able to learn categories by word of mouth. Thus, only humans are able to do this as it passes through verbal learning. This paper linked categorization and language to the symbol grounding problem. This problem essentially revolves around grounding abstract shapes into meaning and reality. However, I didn’t quite understand what a protolanguage was. What does the term mean? What does that comprehend in a type of language?
ReplyDeleteInes, try to make your comments more informative for kid-sib.
DeleteSome hypotheses about the origin and evolution of language refer to the notion of “protolanguages,” intermediate between nonverbal communication and language. But the notion is vague and possibly incoherent. There are precursors of language-relevant capacities in the behavioral interactions and communication of many species, capacities that are also part of the capacity for language, but that are not language.
One of the features of language (related to propositionality) is that every human language can say anything that any other language can say. That’s why, if you can express one proposition, you can express any proposition. Otherwise, it is not a proposition: How can you say and mean “the cat is on the mat” if that’s all you can say? Understanding “the cat is on the mat” entails the understanding of “mat is on the cat,” “the cat is not on the mat,” “the cat is not a mat,” “mat,” “cat” “on,” and so on, for any combination of (arbitrary) names of grounded categories.
So if someone wants to suggest that there can be a “protolanguage” in which you can say this>/i> proposition, but not that proposition, they have to show why not. Two trivial and erroneous answers are these (why?).
1. “The ‘protolanguage’ has the vocabulary to say this but not that.”
2. “The ‘protolanguage’ can’t say this with the same number of words as the ‘language’.”
There is no such thing as a “protolanguage.” If it can express propositions, it is a language. And it can express any proposition.
The same is true of pidgin languages, which are usually inter-languages which are simplified mixtures of two languages for communication between speakers of two different languages. Pidgins can become creole languages, which are independent languages spoken by the children of the pidgin speakers, who learn and use them as their full language. That’s the origin of Haitian creole.
The Symbol Grounding and the Origin of Language reading approaches complicated concepts in a simple perspective which made me better understand them. In the reading, it is believed that the origin of language does not correspond to vocal communication, however I was wondering whether newborns also first approach language with gestures or with verbal language. When I think of it, I believe newborn infants try to communicate with adults by making noises and they actually try to tell us something, but they just cannot word them out because they haven’t learned them yet. So, did our brains evolve in a way such that we first directly try to communicate with verbal language from birth? Or do we intuitionally first try gesturing and then switch to verbal communication because that’s what we observe from our environment and other people (i.e., our parents)?
ReplyDeleteYou bring up a very interesting point, and I wanted to share my thoughts on your questions. I feel like the fundamental difference between the noises a newborn make and gestural communication is that the latter involves some sort of ‘grounding.’ The iconicity of the action helps create a ‘grounded’ connection between the symbol and the referent– all because of the PHYSICAL similarity between the two. This is a theory for the evolution of language, but of course, language moved on to more arbitrary and abstract symbols. However, in the case of newborns, I’d argue that they also have some sort of gestural capacity– often when they cry or wail, they make grasping or reaching movements. We look at these gestures to understand what they are trying to communicate… the noises don’t matter as much because there is no form of ‘grounding’ between the sound and the referent.
DeleteAlara and Anaïs, distinguish babies born 250,000 years ago and the babies born today: The huge difference is that all babies today are born with a brain “prepared” for language, and in fact strongly biased toward speech. The ancient baby has neither.
DeleteBoth, however, are able to move and gesture, just as nonhuman primates can. But none of that is language.
Deaf babies today are also evolutionarily biased toward both language itself (propositionality) and spoken language. But there is evidence that they also retain their ancestral capacity for gestural language (not just pantomime), and learn it almost as readily as speech, especially if raised in a sign-language culture. (“Almost,” because deafness is a greater handicap than just a linguistic one.)
And, of course, hearing people, including the relatives of deaf children, can learn sign-language as a second language as readily as any other spoken language (and the earlier, the better). (I wish I had learned a sign language; and if it weren’t for the fact that all my remaining life is – and should be -- devoted to trying to save animals from humans, I would definitely learn a sign language now!)
Thank you for the answer, Professor. You have reminded me to consider the crucial difference between the babies of today and those from thousands of years ago. To follow up on your comment, I wanted to make sure I’ve understood your point and the overall message of this reading accurately.
DeleteI think the biggest confusion I was having in my original comment was between gestural communication (pantomime) and gestural language. Your response as well as other responses have clarified the fundamental difference between the two, but there is one aspect of it that I’m still unsure about. I understand that gestural communication is NOT language, and that gestural LANGUAGE refers to things like sign language. However, what I’ve concluded is that gestural communication (ie. pantomime) is LESS arbitrary than language and gestural language (ie. sign language)… and I’m not sure this is a correct conclusion. My thinking is that because gestural communication/pantomime bases itself in iconicity, it is essentially less arbitrary… the movements look like the thing that they are referring to. As language has evolved, it has become more and more arbitrary as the symbols referring to objects have moved further and further from iconicity.
Anaïs, yes, the fact that we don't have to call "dogs" "bow-wows" makes category-names more arbitrary, but propositions make language infinitely more powerful than either imitation (pantomime, whether gestural or vocal) or supervised category-learning.
DeleteTo me, it seems very logical that the reason why hominins transitioned from pantomime to proposition 250,000 years ago is because of the superiority of instruction learning versus induction-only learning, since evolutionarily, natural selection would have just dictated the survival of those hominids able to use language and their subsequent reproduction. If we are to believe that Darwin’s theory of evolution is correct, then this would in fact be the most likely explanation. In fact, even without a simulation, this isn’t difficult to imagine: trying to communicate with hand-signals, no matter how sophisticated they are, is more unclear than telling someone something in a language.
ReplyDeleteBrandon, fine, instruction is better than induction.
DeleteBut to get to instruction, you need language (propositions). "Hand signals" (pantomime) is not language. Sign language is language. So I don't know what you mean by "hand signals."
How do you get from gestural communication to propositions -- whether in sign language or in speech?
I suppose by hand-signals I'm referring to primitive forms of sign language, so the gestural communication that proceeded sign language in evolutionary history. In terms of how the gestural communications becomes sign language, symbol grounding would be necessary so that propositions can actually be used
DeleteI enjoyed reading this article as it unified many of the course concepts including symbol systems and symbol grounding into its explanation of language. It presented natural language as a symbol system who’s symbols are able to be grounded and categorized through sensorimotor and categorization capacities. Further integrated was the idea that with the evolution of mirror neurons we were able to move from simple iconic gestures to shared symbol systems and thus abstract person-person communication.
ReplyDeleteEmma, but you've still left out the crucial steps: icons --> words and pantomime to propositions.
DeleteOur language abilities developed due to Baldwinian evolution, which is a special case of the traditional view of evolution in which the trait that is evolved allows us to learn something more quickly, rather than something that directly affects our chances of survival. Language allows us to learn categories more quickly, thus drastically increasing our chances of survival. Beyond this, we are able to make it more likely that our genes will be inherited by teaching our children the same categories, so they are less likely to make mistakes that will put them in danger. However, language use requires at least a subset of its symbols to be grounded in reality, the minimal grounding set. The finding that people categorized objects faster when they were directly grounded is really interesting. However, I wonder whether this is a distinction between direct and indirect grounding or between the familiarity of the categories.
ReplyDeleteElena, learning to categorize is learning to DO the right thing with the right kind (i.e., category) of thing. The doing can be anything (eating, collecting manipulating, fleeing), including one special DOing: naming. If by “symbols” you mean category-names, why would humans name categories before they had language. Language (and verbal instruction) requires the capacity to produce and understand propositions, which are subject/predicate strings of category names. Imitating what you DO with a category is not naming it – but it can become the name of the category: how?
DeleteWhen a person imitates the association that another person makes between a verbalization/sound and a category, that verbalization becomes a shared name for that category.
DeleteI had a question I didn't get to ask during class. I wonder, for example, if communicating through messengers on the phone is considered gesturally. If so, are we going back to gestural (because we constantly text through our phones)?
ReplyDeleteNadila., texting is verbal, so it is derived from speech. Gestures can be pantomime (imitation, which is iconic, and resembles the shape of what you are imitating) or it can be propositions (which are language whether they are in signed language, speech or writing). Testing is not "going back" to gesturing. It is writing, just as typing and morse code is. That's true for deaf people who can only speak sign language -- when they are texting it; also when bling people type English in braille
DeleteSky 8b
ReplyDelete"Chimps have categories. We keep training them to “name” their categories (whether with gestures, symbolic objects, computer keyboards, or words) — even to combine those names into quasi-propositional strings. And the chimps oblige us if they are hungry or they feel like it. But what is striking is that they never really pick up the linguistic ball and run with it. They just don’t seem to be motivated to do so, even if they sometimes seem to “get it,” locally, for individual cases. " (page9)
In section 5, this paper discussed how simple propositions may be originated because of the motivation of teaching new categories to others kins of our human ancestors, and that chimps did not evolve this ability because of lack of motivation. I incline to agree with this view. But I think to "name" categories, other than the motivation, you might also need the memory capacity for it. Because the names of categories are learnt though trial and error, and then are stored in the memory. So even if chimps are motivated to produce propositions, will they be successful if they don’t have name association and memory capacity?
Unless, naming categories is just another doing capacity? For example, "say this word or use this sign when you want to say this", same as "drink water when you are thirsty”.
In many ways, the origin of laguage amounts to the transition from show to tell. Learning categories through sensorimotor experience is important because it allows us to connect symbols to their real-world counterparts. However, it is highly inefficient on its own and can have adverse consequences. Hypothetically, say we are trying to identify a spider that is dangerous versus a harmless one. Through trial and error, one would be putting one's self at risk in trying to identify whether it is dangerous (learning solely based on interaction with the real-world). However, if we are simply informed based on its characteristics, without direct interaction with the spider, we would reduce the risk of danger. The benefit of language is that it saves us the time and trouble of having to interact with every possible category in order to learn it.
ReplyDeleteI appreciate the summary. I think the purpose of this reading was to tie together a number of concepts that we have already discussed. While the last reading(s) actively avoided discussing UG, this reading took it as a given and that is why there were no dedicated arguments made for it. I think the reading put in practice the hypothesizing we read earlier to determine the development of language and the role of symbol grounding that allows language to exist.
DeleteAfter reading all the great comments and responses pertaining to the text “from show to tell” I understood that in order to make this transition from showing (pointing out, sensorimotor grounding learning with unsupervised and supervised learning) to telling (verbal learning, hearsay) you need language. To show, you use pantomime which are signs without language, and telling is done through language propositions.
ReplyDeleteThis got me thinking about non-human communication. I wonder at what stage of evolution of language are they still at? In other words, we know animals cannot learn through telling (hearsay) as they don’t have the gift of language but can they even perform the task of showing? How would a mother communicate to her offspring that it needs to go back to the nest, could she show him where the nest is? I believe she could because otherwise, it could never learn where to localize the nest and return to it on its own.
On another note, I wonder what the implications of animal language are. Indeed, dogs can bark, bears can roar, wolfs can howl… Are all of these not aspects of language to communicate with each other? They can produce sounds, if they are not to communicate some sort of information, what is their use and why would they do it? They must present an advantage, otherwise, evolution would have disregarded them.
Your comments about animal communication are really interesting. There are certainly animal species out there which demonstrate gestural communication. I wonder if there are other communication methods which animals use that are not gestural yet serve the same function. For example, ants use pheromones to communicate with one another and that is what allows them to walk in trails. To address your first question, I don't think any animals have evolved language capacities at all, but certainly still have complex communication methods. Dogs barking, bears roaring, and wolves howling, as you pointed out all must serve some evolutionary function. I'm no expert on the subject but I would imagine they are auditory communication methods, but not language.
DeleteRandom and possibly irrelevant to the larger discussion (and maybe I’ve confused myself a little on this small point) but in saying that all languages can be said by all other languages, I think the argument that language is the same across board in terms of purpose and effectiveness is being made. I still struggle to understand this, similar to my understanding of the Whorf Sapir hypothesis and how perception can be affected by language. Or I guess not be affected.
ReplyDeleteI think it's easier to understand the statement when you think of it as "anything I can say in language A, can be expressed in language B, given the freedom to use as many or as little words as it takes". I would say that across languages, they do serve the same purpose and that's been demonstrated time after time in providing a mode for communication as well as its contributions as being a central component to our cognitive development. Really, the proof for that is in the fact that if you took two people who speak two completely different languages, they would still be able to communicate just as effectively with others (who speak the same language) and would hold similar cognitive capacities.
DeleteIn terms of effectiveness, maybe the argument can be made that some languages are more "effective" than others but I don't know how much that would hold up especially since certain aspects of language develop in a context-dependent manner so I would say they are made to be effective in the geographical regions where they are localized to as well as to the situations that its users can face. For instance, some languages have an extensive vocabulary to describe different states of snow since its users live in an area where snow is very common thus having these different words helps them communicate ideas central to their survival whereas other languages could have evolved to only have one word for snow because its users live in areas where they do not encounter snow. All this to say that you could say all language is equal in purpose and effectiveness for the given user.
DeleteGestural communication methods, such as pantomime and imitation, have the benefits of relating symbols directly to their meanings, because the two are visually similar to one another. Since humans are visual creatures, our worlds are much richer in visual content than auditory content. Gestural communication like pantomime acted as a necessary stepping stone towards actual gestural language which is expressed orally but still rooted in and originating from a highly visual context. Moving away from iconicity towards the use of arbitrary symbols in language (which are grounded in a set of base level words directly attached to sensorimotor representations) is what has allowed humans to tell stories, create culture, and dominate all other species.
ReplyDeleteThe idea of “propositional attitude” is interesting because most species can understand and learn categories. Chimps were able to gain some level of understanding of propositions. The chimps' case study was also interesting because the behaviour of the chimps did not display motivation which brings to question how much cognitive capacity was required for propositions to be acquired. The reading discussed earlier that propositions moved from being incidental to intentional and less iconic to more arbitrary. Regarding the symbol grounding problem, the mutation for propositions was critical in understanding how motivation and intelligence play in language. The chimps did not want to start integrating propositions, but humans did.
ReplyDeleteThis paper attempts to explain the origins of categorization in natural languages. Indeed, Harnad, Blondin Massé et al. explain how, from an evolutionary perspective, it is beneficial to learn from instruction (being taught the category of something) rather than induction (figuring it out for yourself) because induction poses a much higher risk of exposing oneself to risky situations (such as eating poisonous mushrooms). Learning via instruction is also much more efficient, which likely explains why it stuck in humans. However, a certain knowledge base needs to be learned via induction to make sense of instruction (Symbol Grounding: to link arbitrary category names to real-world objects/ideas). Being able to make propositions, i.e. to develop a category from another, the basis for unlimited language began. This was possible with gestures alone, but language allows for much more efficient shaping and recombination of categories.
ReplyDelete