Pullum, G.K. & Scholz BC (2002) Empirical assessment of stimulus poverty arguments. Linguistic Review 19: 9-50
This article examines a type of argument for linguistic nativism that takes the following form: (i) a fact about some natural language is exhibited that al- legedly could not be learned from experience without access to a certain kind of (positive) data; (ii) it is claimed that data of the type in question are not found in normal linguistic experience; hence (iii) it is concluded that people cannot be learning the language from mere exposure to language use. We ana- lyze the components of this sort of argument carefully, and examine four exem- plars, none of which hold up. We conclude that linguists have some additional work to do if they wish to sustain their claims about having provided support for linguistic nativism, and we offer some reasons for thinking that the relevant kind of future work on this issue is likely to further undermine the linguistic nativist position.
Subscribe to:
Post Comments (Atom)
-
Harnad, S. (2003b) Categorical Perception . Encyclopedia of Cognitive Science . Nature Publishing Group. Macmillan. http://www.youtube.com...
-
(10a. Comment Overflow) (50+)
-
Psychology PSYC 538, Winter 2018: Categorization, Communication and Consciousness Time : Tuesdays 2:35-5:25 Place : STBIO S3/3 ...
"'How does the child construct her grammar? In other words, why is the adult output grammar the one that it is?'
ReplyDeleteSurely, how a grammar is constructed is not the same question as why that particular grammar was constructed, yet Wexler appears to conflate the two questions."
Maybe it's just the way i'm reading it, but to me it doesn't seem that Wexler is necessarily wrong in conflating how and why in this case. His argument is referring to the specific output of one particular individual, not the sum of all native speakers of a language. So the question "Why is the output of an English speaking adult head initial" is uninteresting, it's head initial because the child learned it that way, which just brings us to "How did they learn it was head initial?".
This reading supported some of my skepticism about some evidence of linguistic nativism presented in the Pinker article. Pullum and Scholz argue that the argument for the poverty of the stimulus is insufficient to validate the claims given by linguistic nativism. P&S believe a more fruitful avenue of research lies in the data-driven hypothesis of language acquisition. The POS argument is difficult to gather empirical evidence for because most (if not all) study designs would be unethical (depriving a child of an environment that most children are given when acquiring a language).
ReplyDeleteBoth Pinker papers we have read kind of skirt around the problem of UG. Prof. Harnad posits that this is because Pinker doesn’t fully understand the distinction between OG and UG. This article illuminates how this could be the case for a lot of people: “each successive writer on this topic shakes together an idiosyncratic cocktail of claims about children’s learning of language, and concludes that nativism is thereby supported. Most of the frequently encountered claims are about children’s observable accomplishments or aspects of the child’s environment.”
While I agree with the tack you are taking and I believe that Pinker certainly skirts UG more than he should, my question to you is: why does data acquisition seem like a more fruitful theory for you than POS? Despite the lack of proof (at least as of now) that POS is really correct, does it seem more plausible that a child is exposed to a wide enough array of unique linguistic utterances to learn the more esoteric rules of grammar? I by no means wish to suggest that the theory is implausible... I think the best method might be a hybrid model where much of UG acquisition can be explained through data acquisition and the rest requires a more nativist view.
DeletePullum, too, fails to distinguish OG and UG.
DeleteI think it is important to note that acknowledging the short-comings of explanations for linguistic nativism, and more specifically the questioning of arguments for the poverty of the stimulus is not the same as saying data acquisition seem like a more fruitful theory. P&S specifically say:
DeleteAlthough this article will no doubt be misread as a defense of empiricism, that is not an accurate characterization of it. We are agnostic at this point, perfectly prepared to accept that perhaps at some time in the future data-driven algorithmic techniques for language learning will hit a serious roadblock.
What’s being said in the article is that there is not adequate evidence to back up claims of poverty of the stimulus and that data-driven acquisition needs to be studied and understood better as well if it is to be dismissed.
Pullum and Scholtz are arguing that linguists still need to provide evidence that there are elements in natural languages that are unknowable based on input alone but I am concerned with those structures that are clearly knowable on the basis of input but that aren’t. Why is it that some elements that we can find in the oral linguistic environment aren’t learned as quickly as other elements and are overgeneralized? Take for example the past tense of the verb to go which is “went”. This is an irregular verb form for past tense but unlike other irregular forms it is highly frequent in everyday speech and certainly a child learning English will have heard this form quite a bit. Yet children say “goed” instead of “went” despite the mistaken form never appearing in speech. Why is it that early child language learning favors “goed” over “went”? I don’t see why it ever would unless some internal mechanism is favoring a generalized form based off of the rules it acquires in the input over isolated forms. This is a highly efficient way of learning there is no doubt about that so then why does the child eventually retreat from the overgeneralization? Why and how does the preference change? Though unknowable forms would provide evidence of the poverty of the stimulus argument perhaps “knowable” forms that aren’t so easily acquired can perhaps shed light on the issue as well.
ReplyDeletePast tens is OG. General regularity (rule) plus exceptions (memorized). So the rule is overgeneralized till the exceptions are memorized.
Delete“The argument type in question turns on the claim that during the language acquisition process, children often come to know things about the language they are acquiring despite not having access to the crucial evidence that shows those things to be true of the language”
ReplyDeleteThroughout the paper, the authors take issues with the reasoning and evidence people are using for the poverty of stimulus argument. But I think this may be a big case of the pygmies. Chomsky’s point about UG, isn’t what is on trial here (which is what really should be if we really want to argue about poverty of stimulus). And this is evident because as the authors of this paper make clear from the start that they are arguing against the idea that “children come to know something about the language **they are acquiring**”. Anyone arguing that children learn English vocab/grammar without being exposed to the requisite positive/negative evidence is therefore surely not talking about what Chomsky was talking about, right?
A lot of OG can be learned via unsupervised learning, but not UG.
DeleteDoes it really matter to make the distinction about “the language they are acquiring” – because yes, UG is talking about something innate to human beings to have this innate recognition of grammar, but the examples given to in favour of UG, the examples of poverty of stimulus are rooted in the specific languages as well. What I am saying is that just because they are using examples from specific languages does not mean they are not still demonstrating that there is a lack of proof for the statement that there is **poverty of stimulus**. I guess in response I’d wonder what the alternative to challenge would be?
DeletePullum and Scholz make some very interesting points about the poverty of stimulus argument. It is impossible for us to empirically test exactly how much and of what type of input children are receiving because we cannot monitor enough children every second of everyday for enough time to get solid results. For this reason, linguists may have come to this conclusion too early without enough data. I think the absence of stimulus argument is also interesting and I had not heard it before. Because children only learn what is permitted grammatically and don’t have negative examples of what is not permitted, this relates back again to the idea of Universal Grammar. Do they need to some extent these negative examples or is it mostly innate that they can learn proper grammar without these negative examples?
ReplyDeleteA man named Deb Roy set up cameras throughout his house to record everything his son ever said which equated to 90,000 hours of footage. (Stevan says that Deb will be sued by his son at a later date). Although the tedd talk focused on OG instead of UG, it would be interesting to use the data and see which words his son never made mistakes with or had no negative feedback for (show UG).
Deletehttps://www.ted.com/talks/deb_roy_the_birth_of_a_word/up-next
There is no way to make sense of this without distinguishing OG and UG. OG is learned (and differs from language to language, and changes). Much of OG can be learned via unsupervised learning alone (but need not be, as there is plenty of negative evidence and feedback available, if needed).
DeleteUG cannot be learned from what the child hears and says; no UG errors, no corrections. (Deb Roy will not find UG errors or corrections in his experiment. He will be sued for nothing!)
In stating the Argument from Poverty of the Stimulus, Pullum includes “But infants do in fact learn things for which they lack crucial evidence. Thus human infants do not learn their first languages by means of data-driven learning.”
ReplyDeleteThis, similarly to Pinker, fails to differentiate UG from OG. Yes, children learn things for which they lack crucial evidence. However, their learning of languages is combined with data-driven learning, i.e. OG. There are many things bothering me in this article, especially the fact that Pullum misses the point of the APS: given the fact that he doesn’t address the issue of negative evidence, he can’t properly look at APS, as this is its very problem! Above that, he fails to confine the APS to UG, and mixes everything up… I’m don’t know much about Linguistics, so I might be wrong about all of this, but it seems to me that the issue that this paper is criticizing is not completely grasped by those who wrote it, and is (purposely or not) concealed by bringing into it OG, which (from what I have understood) has nothing to do with it in this respect.
Children do make OG (but not UG) errors, though they may not be corrected. But they keep hearing correct OG too, and eventually figure it out. They have nothing to figure out with UG, because they never make UG errors. So they "know" UG already.
DeleteThis was a really refreshing read, to see such a widely popular and accepted argument in linguistics be dissected. To my understanding, this article is trying to answer the question of whether the lack of negative evidence is enough to validate a nativist account. Although there may not be a lot of direct negative evidence from the parents telling a child whether their utterances are right or wrong, I think the child can still get instruction and feedback from other methods. For example, it feels like something to be understood, and that can be a form of instruction/feedback. Our brains also take auditory and sensorimotor data and can either cause a synapse to strengthen (reward) or weaken (punishment), which we can compare to reinforcement learning. We can get instruction from prosody and semantics as well. However, I do think that this type of negative evidence is necessary but perhaps not sufficient for language understanding.
ReplyDeleteAll true for OG. But there are no UG errors...
Delete“Of course, in order to demonstrate that there is no relevant experience with respect to some property of language, we really would have to have a complete record of a person’s experience - a job that would be totally boring; there is no empirical problem in getting this information, but nobody in his right mind would try to do it. So what we can do is to find properties for which it is very implausible to assume that everyone has had relevant experience.”
ReplyDelete“A construction inaccessible to infants during the language acquisition process must be rare enough that it will be almost entirely absent from corpora of text quite generally.”
I like how Pullum and Scholz highlight what is wrong when you are considering what a child’s linguistic input constitutes of. To assume the PoS based on utterances that seem rare is not a sound belief because we underestimate the frequency of the utterance in the spoken language if we only look at written sources. Pullum and Scholz point out that some sentences that we say in colloquial language are not consistent with “journalistic prose” and can be edited out. It would then be flawed to base a hypothesis using text corpora. Gordon’s study was based on his analysis of the Brown corpus, which is a corpus based on American English texts. It seems odd to me that Pullum and Scholz would use Gordon’s study as an example of how his APS was flawed when his method of formulating his hypothesis was one that was criticized earlier in the paper.
Nonetheless, I think Pullum and Scholz do a great job debunking what is needed for the argument of the PoS to be supported, which is to show how/why children learn with lack of negative evidence instead of when they do. The four studies that they provide in their paper all have a component of their APS wrong, and I believe that this stems from the confusion of UG with OG.
What do P & S show, if they fail to distinguish OG and UG? The APS applies only to UG. And there are no UG errors, either in what the child says or what the child hears.
Delete"How does Angela know that F is a fact about L? If the answer to this involves giving evidence from expressions of L, then Angela has conceded that such evidence is available, which means that S could in principle have learned F from that evidence. That contradicts what was to be shown. If, on the other hand, Angela knows L natively, and claims to know F in virtue of having come to know it during her own first-language acquisition pe- riod with the aid of innate linguistically-specific information, then Angela has presupposed nativism in an argument for nativism. That is viciously circular. And there are no other cases to consider, so the argument refutes itself."
ReplyDeleteI think this is a poorly structured and confusing argument that tries to refute language nativism. The question about language nativism is not on how we think we learn it, but rather why under poverty to negative stimulation. It is the still the case. Universal grammar is something intuitive, but it does not mean it cannot be discovered by deliberate contemplation.The epiphany of universal grammar also came from the reflection of many linguistics on language, so why does it matter for somebody to know, and how do they know the knowledge of F about L?
The paper also claims that there are different environment children growing up to acquire language, therefore considering this diversity as a whole is oversimplification. This idea clearly shows the lack of understanding of universal grammar, since the variability of learning exists only in the boundary of OG, and UG is something that despite all those variations, still remain consistent and unanswered.
Correct
DeleteIt would be interesting to research a model that aims both at the nativist views of UG and data acquisition. However, it's important that we research in order to understand more about the phenomenon of language rather than find additional proof that supports whatever theory we have in mind.
ReplyDeletePersonally, I think that children do have an innate capacity of language rooted in UG but I also think that language happens as a product of the environment in the sense that there are many aspects of our everyday life that affects the way we communicate.
Maybe scholars tend to argue against the theory of UG because they can't wrap their head around the fact that such a uniquely human phenomenon (language) is rooted in the innate capacity of forming syntactically correct phrases. Sure, it doesn't suffice in explaining everything about the origin of language, but I believe it is a stable grounding set for the elements of language to build on.
The problem is not that UG is innate, but how and why it evolved.
DeleteI have a few issues with Pullum and Scholz’s objections to the APS:
ReplyDelete1. Their critique largely takes aim at proponents of linguistic nativism who themselves don’t understand the APS or the OG/UG distinction. The fact that degeneracy, convergence, ingratitude and idiosyncrasy so often come up is a clear indication of this. These things have nothing to do with the APS. They are unrelated to negative evidence, which is the essence of the APS.
2. They dismiss the for need negative evidence as a special case which isn’t entirely relevant to the APS, when in fact it is central to it. They write:
“For reasons of space, we do not attempt to give further details here. There is of course much more to be said about whether negative data are crucially necessary for learning natural languages from it.”
By this point we’ve CLEARLY established that negative evidence is the linchpin of the learning process. The “laylek” thought experiment is a great example of this. You can’t acquire categories without negative evidence. Pullum and Scholz betray their fundamental misunderstanding of what it means to learn, and by extension what the APS argues, in this passage.
3. This is related to the previous point, but, they think that the APS is about positive evidence (and positive evidence of OG, as opposed to UG, at that). The APS as they understand it is that parents don’t speak to their children in a way that is sophisticated enough to get them to learn the rules of grammar:
Delete“At least one researcher who has spent time looking at the way adults speak to children reports that “the input to young children is neither so depleted nor so uniform as some have suggested” (Berman 1990: 1160). Berman also notes that speech addressed to the learner is not the only evidence about the language that the learner has; children learn from at least some utterances that are not addressed to them. And indeed, Weddell and Copeland (1997) found that even children between ages 2 and 5 pick up and understand much more (e.g., about news events) from the language they hear on television than their parents typically realize.”
Again, it’s well established that the APS isn’t about children hearing baby-speak but learning to speak a language anyway–obviously children hear adults say grammatical things to one another and on television. It’s about them never hearing or producing UG-violating statements and then hearing or receiving corrective feedback for those statements. Searching through various genres of literature for colloquialisms won’t produce UG violations because they are so rarely produced (except by the linguists who go looking for them)! Again, if they properly understood this, they would realize that the following requirement they posit is actually met for UG (regardless of genre):
“a construction inaccessible to infants during the language acquisition process must be rare enough that it will be almost entirely absent from corpora of text quite generally”
4. The idea that all arguments in favour of linguistic nativism presently given are circular:
“Consider the position of a linguist – let us call her Angela – who claims that some grammatical fact F about a language L has been learned by some speaker S who was provided with no evidence for the truth of F. Sampson raises this question: How does Angela know that F is a fact about L? If the answer to this involves giving evidence from expressions of L, then Angela has conceded that such evidence is available, which means that S could in principle have learned F from that evidence.”
They are right that UG-violating sentences exist. But the evidence is never heard by the child as they learn a language. This isn’t “presuppos[ing] nativism in an argument for nativism”, as they claim. While it’s true that the fact F was deduced intuitively, and that we usually provide examples of violations of F to illustrate that it’s true, the evidence for its nativism is that children know F before the environment gives them an opportunity to have learned it.
Broadly, my issues with this article stem from the fact that Pullum and Scholz don’t properly understand the APS or the OG/UG distinction.
I have no worries about your grasp of it all, Willem!
DeletePullum places emphasis on language as an innate feature of humans. Stimulus poverty is when learning a language is made more difficult by an environment lacking in positive evidence (as defined by Pinker). Nonetheless, language acquisition occurs, which gives evidence to the innate theory of language.
ReplyDeleteA question is brought up surrounding how children learn that a sentence or word structure is ungrammatical in an environment that exclusively offers information on what is grammatical. For example, children learn to differentiate special cases of tense changes such as saying “bought” instead of “buyed.” This is called overcoming a generalization of a rule, which is learned without much instruction. Universal Grammar, as proposed by Chomsky, supports that the environment shapes an innate collection of language properties. It would be interesting to see how a child’s language would develop if feedback from parents or guardians was offered, but only incorrect corrections. Would a child still be able to decipher the grammatical from the ungrammatical regardless of feedback correctness?
I think that the child certainly would not decipher correct grammar from incorrect grammar, assuming that the child was receiving the majority of their language exposure from their parents. Children learn incorrect grammar from their parents all the time. For example, if parents use double negatives in sentences “I didn’t do nothing, I won’t go nowhere” the child will certainly use similar grammatical structures. Similarly if the parent incorrectly corrects “bought” to “buyed” the child will learn this too. It doesn’t matter what the correct ordinary grammar of the language is; if some grammar is used consistently in the child’s environment, the child will learn it. I think your question is more compelling if you phrase it in terms of universal grammar. If the parents spoke in sentences and made grammatical corrections that did not adhere to the rules of universal grammar, would the child learn them?
DeletePullum and Scholz’s article evaluates the notion of linguistic nativism by considering the argument from the poverty of the stimulus (APS). The argument suggests that our environment provides insufficient evidence for language learning to be entirely data-driven and hence we must possess innate linguistic knowledge as infants. P+S make a convincing argument that before accepting APS to be true we must ensure we have an adequate background of empirical evidence to display that the environment is indeed too impoverished to enable such acquisition (without a pre-existing innate capacity for language). They therefore conclude that until data-driven learning is investigated in greater detail, linguists can only speculate on the validity of the APS.
ReplyDeleteIf, however, the necessary empirical evidence is revealed that renders our environment sufficiently linguistically rich and hence deems APS unprovable, does this automatically disprove linguistic nativism as well? To me, the advanced nature of language acquisition (i.e., infants’ abilities to master such complex linguistic constructions so early on), even with a sufficient rich environment, still suggests that some aspect of our linguistic capacity is likely pre-programmed and innate.
I agree with your point that there must be some amount of linguistic capabilities that are innate and found in all human infants. Based on the fact that infants even a few days after birth are able to distinguish between languages with different stress patterns (syllable based, stress based, mora timed), there must be some innate language processing system that would allow for this to occur. Babies also respond preferentially to the language their parents speak versus another language as early as a few days after birth. Furthermore, we see that babies have the ability to segment streams of linguistic input based on stress pattern, prosodic information, and relative frequency (of the syllables). Based on these abilities there must be some sort of language faculty that babies rely on to perform these tasks
DeleteOne issue I have with P&S's argument that we need empirical evidence of an impoverished argument is that they fail to distinguish between learning language specific grammatical complexities (OG) from the UG that all languages share. When it comes to UG, it's extremely easy to show that the environment is impoverished, as people do not violate UG in their speech. So the APS is essentially that there is only positive evidence of UG, never broken UG, so it must be innate.
DeleteIn obtaining further empirical evidence regarding the linguistic availabilities, we may learn how infants learn idioms or complex grammar for specific languages but it won't speak at all to the underlying structure of all languages.
I think this is why you have the intuition that no matter how linguistically rich empirical evidence finds our language to be, it wouldn't rule out innate language structures.
Gordon states that compound nouns with a regular plural as a non-head like rats-infested should not be possible while Pullum and Scholz find a number of counter arguments like chemicals-maker or forms-reader. According to their hypothesis, these examples refute Gordon’s claim because children receive elements which should not make them more likely to prefer rat-infested over rats-infested because they receive evidence of both types. It is possible both of them are aiming at a too large rule to apply on compounds and that if they chose to more carefully look at a specific type of compound nouns, they would find evidence that agreed with one author or the other. It is clear to me that this discussion achieves nothing since it is possible to have many types of compound nouns and so far none of the authors have been able to come up with a rule that fit both mice-infested, rat-infested but also chemicals-maker. Perhaps it matters in whether the plural in the non-head can refer to different sorts of the noun, the way chemicals-maker refers to a maker of many chemicals while rats-infested does not make sense because one does not usually differentiate between different types of rat and so would not say rats-infested to mean infested by different species of rats. This is only a poor attempt at showing how much more specific the authors should be about the grammatical rules they are discussing, in Gordon’s case, otherwise others can find counter-arguments quite easily like Pullum and Scholz did. This section of their paper didn’t even refuse the APS, they just found counter-arguments to a too broad grammatical rule.
ReplyDeletePullum and Scholz thoroughly evaluate common arguments for linguisitic nativism in their paper. They note that the statement that is most important to evaluate is “can infants learn things for which they lack crucial evidence?” They go through some examples of phrases that could be candidates for something that a child would be able to infer the form without hearing evidence of the form around them. For example, they consider the phrase “by and large”. They note that this phrase is incredibly unique in the English language and could never be learned from observation of expressions that contain it. They note that this example is of course silly because it could never be assumed or generalized without being learned. They note that supporters of APS claim that there are other instances of this sort of form that are not quite so ridiculous which could be learned without evidence. They proceed to evaluate this point. However, what was interesting for me about the example given, was the similarities between “by and large” and the phrase that Pinker claims to be a violation of UG: Darwinsism. I contend with Pinker’s claim based on similar rational: I believe that the knowledge that Darwinism is the proper grammar seems to also be an example of something that would have to be heard to be learned, i.e. not something innate.
ReplyDelete“It is necessary for the APS advocate to tell us how many examples it is supposed to take before a child is likely to pay attention, and then compare that with how many such examples turn up in actual language use.”
ReplyDeletePullum & Scholz look at auxiliary sequences, which Kimball stated are rare (ex. “must have been thinking”), through the APS specifications they provided. According to the inaccessibility evidence schema, these sequences aren’t readily available to the learner during the acquisition process, and without hearing them it’s not possible to learn that they’re grammatical unless it’s innate. P & S show that these kinds of phrases are actually quite common, so it can’t be said that children aren’t exposed to positive evidence of such utterances. HOWEVER, I thought the poverty of the stimulus argument was that they don’t have access to starred (WRONG) examples. Knowing that “must have been thinking” is a grammatical sentence does not entail that “*[UG-violating auxiliary phrase]” is not, and THAT’S what APS (I believe) is getting at. They’re saying that kids have many opportunities to hear what is correct, but they miss the point that they don’t have any opportunities to hear what’s not.
Although Pullum and Scholz are adamant this is not a defense of empiricism, by stating multiple times they are not out right declaring that innate learning, and thus linguistic nativism as illegitimate, but yet they claim to have casted doubt on APS. Firstly, I make no assumptions on whether they are empiricists but they have a clear agenda to promote research into data-driven learning, and it often creates a bias, that makes their arguments challenging APS weaker. Secondly, there is a prominent lack of distinction between OG/UG in the article. Thirdly, their understanding of APS seems to be based on positive evidence which caused me quite a bit of confusion during my first read of the article as I was not quite sure how they brazenly confuse the issue at hand.
ReplyDelete"That is, some aspects of languages are known to speakers despite the fact that the relevant positive evidence, although it does exist, is not accessible to learners during the acquisition process, because of its rarity: linguists can in principle discover it, but children will not."
"But the idea seems clear: there could be positive evidence for the correctness of some rule or principle, evidence that could be found in language use by an adult linguist doing a really intensive search, but would not emerge in conversational data anywhere near often enough to guarantee that any particular child would ever encounter it. This, we think, is the basis for the argument that most clearly deserves to be referred to as the argument from the poverty of the stimulus"
The confusion is evident in the above quotes, where their focus is on the access to positive evidence. APS is simply the ability of children to produce grammatically well-formed sentences and thoughts, without ever having heard a UG ill-formed sentence (negative evidence) to provide an example of what not to say to know what is right to say. To frame the PoS argument correctly would be to say: its about the absence of negative evidence NOT how much negative evidence is necessary for learning. IF they truly understood it, they would've used their effort not to fight whether UG is innate (which evidently was the point of their article although they don’t explicitly approach it this way), but attempted to find out how and why UG evolved the way it did.
Here the authors try and dispute the absolutism of the poverty of stimulus (which is always fun to see). In doing so they bring up they point that when trying to explain that learning language does not depend on experience. The issue there is that when trying to explain the nativist argument for this issues, one presupposes nativism, creating circularity. I think the issue of circularity is the one that I have against the APS. Generating rules with experts around a table and then saying Eureka these rules are not broken in language acquisition ever. It's the definitive nature of this principle that relies on negative evidence that bothers me. Then again Chomsky was a giant and i'm not even close to a gnat, so maybe the UG he discovered is such a fundamental thing that the lack of evidence is evidence for it.
ReplyDeleteAnother way to consider negative feedback in relation to to the PoS argument is that since children couldn’t possibly learn language only by listening to others talk, they must already have some knowledge of the language’s structure. This is because people around us do not speak with perfect grammar most of the time. Children, thus, don’t receive enough negative feedback and thereby aren’t explicitly exposed to ordinary rules of grammar, but somehow still know them. I think in this case, I would disagree with your claim that there is lack of evidence for UG; I believe the scenario I just explained would provide sufficient evidence for UG.
DeleteSacha, I also have difficulty getting around the APS. The phenomenon this concept points out (i.e. lack of positive evidence) is mouthful. Indeed, it is amazing and hard to imagine that children could learn only from experience since they do not get experience for every sentence they will be able to produce throughout their life. However, jumping from an idea (lack of positive evidence) to a conclusion (language must be innate) with thought-generated examples seem like a bit of a stretch to me as well! Moreover, there are studies increasingly showing infants can learn to discriminate new words from a string of sound, just by being exposed to it. For example, an infant would hear blablo-ba-by-plu and plubla-ba-by-blo, and recognize that statistically “ba” is likely to precede “by”, and so “baby” must be a word. That is pretty amazing and I think makes the point that we might want to continue looking for evidence of language acquisition before all agreeing with nativism. Lastly, I would like to make a link between the topic of language and computationalism. I know that computation does not explain ALL cognition… but could it explain language? It would certainly not explain the “feeling” of understanding a language, but what about the “learning” of the language?
DeleteI feel like the learning of a language itself, beyond syntax and grammar, has something to do with the part of cognition that computation can't explain (feeling). This kind of brings us back to the chinese room thought experiment, can language have meaning without being grounded in feeling?
DeleteNot to get too meta, but how do we know that humans are the only ones with an LAD? Yes, for the only languages we understand and produce. However, how do we know that other mammals such as chimpanzees don't have their own LAD used to pick up their own language and learn other dialects of monkeys? This is probably a stretch, but it's something to think about. Just because we can't understand how they communicate, doesn't mean we can assume that we are the only species with LADs.
ReplyDeleteI understand the argument that Pullum and Scholtz are making in their article -- suggesting that there has been no convincing research done that suggests that children do not in fact acquire language from data-driven learning, and until there is, we cannot just assume that the APS argument is true. Indeed, given the lack of data, their argument has some validity in it. But to me it seems they are speaking strictly of OG -- the errors, the feedback, the environmental input received ,and so forth. Children do not experience the same sort of feedback when it comes to UG. Because of this, the distinction between OG and UG is crucial. Although children may experience both positive and negative examples of speech when it comes to OG, they never experience feedback in relation to UG, because they do not make mistakes in this realm. Therefore, after reading this article I feel a tad uninspired as the question of UG and how/why children develop this still remains to be unanswered.
ReplyDeletePullum and Scholz quote Wexler, saying, “How does the child construct her grammar? In other words, why is the adult ouput grammar the one that it is?”. Pullum and Scholz follow this but criticizing Wexler about equating the “how” and “why” here. However, I’m a little bit confused to why they do this. To me, the “why” here is kind of arbitrary based on the language. For instance, why do we drop the pronouns in English, but don’t do so in Greek or Spanish? It seems kind of pointless to ask questions like this, so I’m sure that Wexler is saying “how” are we able to know “how” to do this in the first place, right?
ReplyDeleteI liked that Pullum and Scholz included the counter-example (Sampson's linguist) to the "strong" poverty of stimulus argument posited by Hornstein and Lightfoot. I hadn't thought to make the distinction between "no evidence" and "evidence that is simply not available to the child," but I'm now inclined to agree with the revised (and "weaker") APS. "Correct" evidence is theoretically there, but despite not accessible to the child in their everyday interactions, the child is still able to grasp these rules anyways. However, one thing that Pullum and Scholz never quite address is whether this evidence is for UG or OG, which I think is an important distinction that must be made. Otherwise, it would be easy to conflate the two, which would be incorrect as UG is innate and OG is learned.
ReplyDeletePullum's main question strikes as if there is enough negative to define a rule, while the poverty of the stimulus argument is the absence of negative evidence in a child's exposure. My main question after reading this article is how this theory could account for the variance of language that children hear. For example, what would happen if a child's parent did not have native like attainment of the language? Many immigrant parents may not have complete mastery over the language of their new country and thusly may give more examples of negative evidence to their children. Especially if their parent's negative evidence is all the child hears in early ages before school where language development is crucial (ages 0-3years old), wouldn't this make a large impact on their language? Additionally, it is known that parents with a higher socioeconomic status speak more to their children, giving almost assuredly exclusive positive evidence in much higher quantities. Thusly, my issue for the poverty of the stimulus argument is that since the input children receive can be so varied (in quantity and quality) how can this theory singularly explain language acquisition?
ReplyDeleteHere, Pullum and Scholz make the point that the Poverty of the Stimulus argument has been inappropriately presented in the past. What is crucial to support this argument is that people gain knowledge of the structure of their language without evidence of which in the data to which they are exposed. In other words, children come to know things about their language in the complete absence of negative evidence. They know sentences are ungrammatical without having ever been told that they are impossible in the language. I think we can draw a parallel between UG and feeling, as we have not yet figured out how or why either of these concepts exist. Evolutionarily, UG has no adaptive advantages (that have been discovered so far), and we have no idea what the causal mechanism is in the brain that allows children to learn ungrammaticality without negative evidence. Similarly, we have yet to solve, and may never solve, the hard problem. Perhaps attempts at solving one will provide insight for solving the other, and what makes studying UG so crucial in the realm of cognitive science. If we can reverse engineer the mechanism that allows UG principles to function, maybe we can reverse engineer the mechanism that allows organisms to feel.
ReplyDeleteI agree with you in the sense that uncovering feeling and UG are parallel. Pullum and Scholz are trying to come to the conclusion of the poverty of the stimulus argument that "it is supposed to show that human in- fants are equipped with innate mental mechanisms with specific linguistic con- tent that assist the language acquisition process". But how do we get there and how are we supposed to know how and why UG has adapted towards this.
Delete"How does a child construct her grammar? In other words, why is the adult output grammar the one that it is? Chomsky's answer notes that the attained grammar goes orders of magnitude beyond information provided by the input data and concludes that much linguistic knowledge must therefore be innate."
ReplyDeleteIn a super ironic twist, the poverty of stimulus argument lacks a true argument. Stating that children's ability to learn grammar is "orders of magnitude beyond information provide by the adult" is not really an argument, since no one proposes an alternative hypothesis as to where the information is coming from.
Once again, we encounter the easy and hard problems, here having everything to do with language acquisition. Though most linguists seem to be in agreement that the poverty of stimulus argument s true, it explains the "when and where" of the short comings in the language acquisition process, but does not attempt the "how and why" of the process falling short of the achieved results.
"This example is silly, because by and large is a unique asyntactic idiom, and no one would be tempted to imagine it could be learned by someone who had never heard it; it would be expected to occur in English on the basis of the rest of English syntax. However, linguistic who accept the APS claim there exist other instances of the same general form that are not silly, instances in which the claim of learning without evidence is actually true"
Perhaps I'm missing something here, but are phrases like "by and large" to me, are learned by children based on the context in which they are presented. It is impossible to examine a phrase like "by and large" in isolation because it is impossible to grasp the meaning of a phrase that is so context-dependent, without context. For example, if a child heard a parent say, "by and large, everything is fine," it is possible, if not likely, that the child could extrapolate the meaning of "by and large" for the sentence as a whole as well as the conditions under which the words were said. For this reason, I don't think examples of cases like "by and large" are good arguments in favour of APS.
Pullum and Scholz criticize both linguistic nativism and empiricist language learning, and argue that future researchers should focus on providing more empirical evidence of APS in order to support the argument of linguistic nativism. Despite discussing two ways of language learning in childhood, innate primed learning and data-driven learning, the paper still lacks a clear distinction between Universal Grammar and Ordinary Grammar, leading to many invalid and conflicting points in the paper. Also, as stated in the conclusion, “the advocates of the APS must shoulder the burden of participating in empirical work to support instances of the argument, and thus to begin to vindicate Fodor’s belief that cognitive science is possible.” We understand the importance of the APS argument to Cognitive Science, however, without the distinction of OG and UG, and a clear understanding of the relation between APS and UG, not only will we not able to understand how language acquisition works, but also leading to more and more confusion and controversies, thus we will never be able to reach the hard problem of UG at all.
ReplyDeletePullum and Scholz seek to examine the empirical evidence supporting the argument from poverty of the stimulus (APS), an argument for UG that states that children cannot acquire a language from exposure to it alone, since they master UG despite never being exposed to UG-non-compliant utterances and never producing UG-non-compliant utterances and getting corrected. However, Pullum and Scholz seem to be confusing UG and OG, because the examples of language acquisition that they use have do with how children learn OG, which is not the subject of the APS. Therefore, while they present interesting cases of OG learning, they do not successfully argue against the APS.
ReplyDeleteI am confused again about UG. I seem to understand it theoretically, but something keeps on bugging out in my head. I am not a linguist and have very little background in the field.
ReplyDeleteDo we know what the rules of UG are ? How could we potentially know ? Especially if we make no UG errors. There must be some "wrong" or incorrect UG sentences that exist, ones that violate or contradict the rules, otherwise, they would not be rules, would they ? But since the evidence that this particular sentence is wrong or incorrect is never heard... it makes me confused.
+ still doesn't explain how and why UG evolved...
Delete