Talk:Poverty of the stimulus

From Wikipedia, the free encyclopedia
Jump to: navigation, search
WikiProject Linguistics / Applied Linguistics  (Rated C-class)
WikiProject icon This article is within the scope of WikiProject Linguistics, a collaborative effort to improve the coverage of linguistics on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
C-Class article C  This article has been rated as C-Class on the project's quality scale.
 ???  This article has not yet received a rating on the project's importance scale.
Taskforce icon
This article is supported by Applied Linguistics Task Force.


I removed the statement, `... Chomsky himself has never advocated an argument based on Gold's proof.'. This seems to conflict with "Hauser, M. D., Chomsky, N., and Fitch, W. T. (2002) The Faculty of Language: What Is It, Who Has It, and How Did It Evolve? Science, 298:1569--1579." where they claim `... A version of the problem has been formalized by Gold ... No known “general learning mechanism” can acquire a natural language solely on the basis of positive or negative evidence, and the prospects for finding any such domain-independent device seem rather dim.' -- Cagri (talk) 13:07, 9 September 2008 (UTC)

I cleaned up this article a bit. I think the language flows a bit better now. Tyrell turing 19:34, 17 May 2006 (UTC)

JA: Yes, it does. Jon Awbrey 20:01, 17 May 2006 (UTC)

Poverty of the Stimulus and negative evidence[edit]

I think we should try to keep a clear distinction between the POS argument and negative evidence. The stimulus is "poor" in a number of different respects (e.g. utterances may be misheard, may be spoken by non-native speakers, may be ungrammatical) and the lack of negative evidence is only one of these. Currently, the article implies that there is a single POS argument which crucially relies on the lack of negative evidence as a premise, but this is a bit misleading -- especially since the POS argument has only been set on in such an explicit deductive form by it's critics (e.g. Pullum & Scholz). If there are no objections I will be making some changes to the article to reflect these facts. Cadr 18:50, 23 June 2006 (UTC)

I think you should bring up those issues up in a subsection, something like, "Aleternative forms of the argument", rather than altering the main portion of the article. The reason is that although you're right that the poverty of stimulus argument has inspired discussion of other arguments about "impoverished" aspects of linguistic stimulus, the technical term "The Poverty of Stimulus Argument" does refer to a single argument, i.e. to the original argument surrounding learning from positive evidence. That's why people like Pullum & Scholtz concentrate on it. So, I think that it is important to leave the original argument as the main portion of the article. Nonetheless, I think that a subsection would be good, because there are indeed many variations that have spawned off from the original argument.Tyrell turing 18:56, 25 June 2006 (UTC)

I basically agree with you, and a subsection sounds like a perfectly good compromise, but I'm not sure if it's so clear that there is an "original argument". I don't know when the term was first coined (if you do it would be very helpful if you could give a reference). Pullum & Scholz show that the term is understood very broadly; their narrow definition of a POS argument is simply given to make life easier for them (so they have 1 form of argument to challenge rather than 20). Cadr 23:14, 25 June 2006 (UTC)
You know, I had thought that Chomsky originally focused on the negative evidence issue in his work because of what one old teacher taught me. But, after you brought it up I went and looked at some other articles on the issue and it appears that you are right, and there isn't really a single "original" argument. (Note to self: check references before editing...) However, for whatever weird historical reason the negative evidence issue appears to have garnered a lot of attention, not just from Pullum & Scholtz, but in my experience many critics in computational linguistics and developmental psychology bring it up (e.g. Reich 1970, Bohannon & Stanowicz, 1988, Developmental Psychology (24) 684-689, and Onnis et al, (2002) here: [1]). So, I now recognize that your first inclination was correct, and I was wrong about any "original" argument. But I would recommend a fairly solid section on the negative evidence issue because it has garnered a lot of discussion. Tyrell turing 23:07, 26 June 2006 (UTC)
I agree with you that there has been a lot of focus on the negative evidence issue in the literature, and that this should be reflected in the article. The history of negative evidence seems to be quite interesting (and I wish I had a better first-hand knowledge of the relevant literature). According to Steven Pinker [2] the negative evidence issue was first raised in detail by an empiricist. Anyway, I'll make some (relatively minor) edits in the next few days to try to explain that there are various kinds of POS argument, while keeping negative evidence the focus of the article. (Of course, feel free to get in there first and make the changes yourself.) Cadr 23:46, 26 June 2006 (UTC)

Added a response to Pullum and Scholz[edit]

I added Legate and Yang 2002 as an example of the responses to the article by Pullum and Scholz. I believe their response is representative and particularly interesting because it challenges Pullum and Scholz's central claim (this claim being the one that in all the cases they discuss, the data needed by the language learner is present in the input in a sufficient amount) by pointing out that the relevant data are, in fact, rare. (Actually, there is a huge discussion in the literature and these two papers stand for the two sides but are by no means the only ones, though they are representative of both sides of this particular debate. I think that the article might become too long by including more than the two though.) (talk) 05:47, 13 March 2008 (UTC)

Gold's Theorem overapplied?[edit]

I've tagged "a proof by E. Mark Gold showed that any formal language which has hierarchical structure capable of infinite recursion is unlearnable from positive evidence alone" as not very well supported. From reading the K Johnson reference, it looks like Gold's work has been misinterpreted by some scholars of natural language acquisition; Johnson mentions Chomsky as disputing even the assumptions required for Gold's Theorem to apply to human language acquisition. Change "showed that" to "is often cited to support the claim that"? Then footnote this to show where GT has been cited? Johnson's paper has useful references for the latter. Yakushima (talk) 05:24, 30 May 2008 (UTC)

This statement "a proof by E. Mark Gold showed that any formal language which has hierarchical structure capable of infinite recursion is unlearnable from positive evidence alone" is completely false. Learnability in the Gold sense is a property of classes of languages rather than languages. And hieracrhical structure has nothing to do with it -- presumable this means CF languages. For sure, the class of context free languages is not Identifiable in the limit, but there are classes of context free languages that are identifiable in the limit. —Preceding unsigned comment added by (talk) 15:56, 6 December 2008 (UTC)

Infinite recursion: real or not?[edit]

The article now states:

"As for the argument based on Gold's proof, it's not clear that human languages are truly capable of infinite recursion. Clearly, no speaker can ever in fact produce a sentence with an infinite recursive structure, and in certain cases (for example, center embedding), people are unable to comprehend sentences with only a few levels of recursion."

This is missing the point (and referring to Chomsky's response on working memory only deals with the comprehension part of the issue). The fact that a speaker cannot produce an infinitely recursing sentence does not show the invalidity of the assumption about infinite recursion. In English it is always possible to add one more embedding:

(i) John said that he loves Mary (ii) Bill said that John said that he loves Mary (iii) etc.

The number of embeddings that a speaker of English can produce depends on the time (s)he is able to devote to this task without eating, sleeping, etc. The relevant point, however, is that there is no speaker of English who will say at some point (the n-th embedding) that adding another embedding suddenly makes the sentence ungrammatical. Recognizing this is crucial for evaluating the assumption that human language is infinitely recursive. Not accepting it means claiming that such long sentences are ungrammatical. Onc70 (talk) 13:15, 28 August 2008 (UTC)Onc70

What is the evidence for the statement "there is no speaker of English who will say at some point..."? And, when you say "only deals with the comprehension part of the issue", what other parts of the issue are you implying exist? Unless you believe that the brain has the potential to process infinitely complex structures, then it seems to me that you have to concede that there is an upper bound on embedding. Gavril09 (talk) 04:37, 10 November 2009 (UTC)

God awful[edit]

Completely discounts the substantial linguistics studies undertaken by socionists. Representative of substantial ignorance in light of those studies. See the journals of the Socionics Institute, Kiev for more. Tcaudilllg (talk) 17:06, 7 November 2008 (UTC)

Regardless of whether socionics is at all valid or not, it doesn't seem to have much at all to do with the topic. (For reference, socionics is a borderline-quackery method of personality typing, and hasn't got anything relevant to say about language acquisition.) Saltwater Rat (talk) 04:29, 23 February 2009 (UTC)


From reading this article, one would get the impression that the poverty of the stimulus argument is highly controversial. Indeed it even says so right in the introduction! Maybe 30 years ago, but nowadays UG is undeniably dominant. This article should obviously have a criticism section, since a few critics still remain, but not one longer than the evidence section! LSD (talk) 01:09, 16 January 2009 (UTC)

  • The argument is highliy controversial; nowadays surely more so than 10 or 20 years ago. Moreover, there are large geographical differences; UG has a very low level of support in Europe, as opposed to the US. Piramidon (talk) 14:34, 30 August 2009 (UTC)
  • Indeed, the whole POTS argument is based on the assumption that indications that is neither absolute positive evidence nor absolute negative evidence is useless for learning, ignoring the neurological fact that real brains, unlike conventional computers, can handle degrees of maybe and grade correlations in different strengths of correlation. That some sentences heard by the child is linguistically wrong in some way does not change the fact that each single error is rare, so they statistically cancel out in a way that any non-binary-code brain can notice. When the POTS argument was created, it was still believed that real brains used binary code and that binary computers could become truly intelligent. Now that modern research on organic brains and experiments with non-binary computers have shown the importance of non-binary statistical correlation grading for any learning appliceable to practical, non-laboratoric situations, the POTS argument, while still a good illustration of the monumental stupidity of binary computers, is not appliceable to non-binary brains. Any notion of POTS being appliceable to organic brains is based on the old theory of real brains being binary, and should by any scientific right have demised with it. Thus, there is a good reason for the criticism section to be longer than the "evidence" section. —Preceding unsigned comment added by (talk) 10:15, 22 May 2011 (UTC)
  • The claim that "UG is undeniably dominant" is false and part of the manufacturing of consent that pro-Chomsky linguists are constantly engaged in. The non-dominance of UG is explained by Newmeyer (2003, p. 683) who states that "my impression is that more linguists around the world do cognitive linguistics than do generative grammar." Newmeyer, who is very knowledgeable about the history of linguistics, includes the following types of researchers who do not use and support UG: functional linguists, some syntacticians working in the generative tradition (he lists Bresnan and Wasow), the great majority of psycholinguists, and the majority in the field of natural language processing. To Newmeyer's list, we can probably add most phoneticians, most sociolinguists, most anthropologists, and most language educators and applied linguists. Reference: Newmeyer, F. (2003). Grammar is grammar and usage is usage. Language, 79, 682-707. The following are other examples of the manufacturing of consent by pro-Chomsky linguists: at MIT and many other "generative" linguistics departments, syntax classes usually only discuss Chomskyan theory and do not cover any of the other 31 theories in the Oxford Handbook of Linguistic Analysis; the theoretical literature review in doctoral theses about UG normally does not discuss any alternative theories; and many linguistics textbooks are covertly pro-UG and do not mention any other theories.Occa123 (talk) 03:16, 18 September 2009 (UTC)
  • idkbro. This reads like an extremely ideological POTS-sucks high school paper more than anything else IMO. —Preceding unsigned comment added by (talk) 01:19, 13 November 2010 (UTC)

I'm a neuroscientist and have no particular bias either way. However, I find the article to be clear, and balanced. One has only to read this Discussion to see that the issue is rather controversial, and in the circumstances the article is first-rate. The debate stems largely from our continuing ignorance about the contributions of development and learning to neural circuitry (which presumably ultimately generates language!), and is just a more elaborate, higher-stakes, and much less solvable, version of debates about the origin of receptive field circuitry e.g. in low level vision. Perhaps the article could point this out.Paulhummerman (talk) 16:36, 10 December 2010 (UTC)

  • The poverty of the stimulus argument is controversial, it is rejected as a fallacy by most non-generative linguists. And contrary to the statement above UG is not "dominant" except perhaps on certain American Universities. Occa123 is right.·ʍaunus·snunɐw· 19:33, 3 September 2012 (UTC)

Bad examples and incorrectly described[edit]

The article discusses the learning of how to invert word order using bad examples, one of which is incorrectly described. The first example is the sentence "You are happy," in which the word "are" is incorrectly called an auxiliary verb. It is not the auxiliary verb, it is the main verb, linking to a predicate adjective. In a discussion about inverting of auxiliary verbs, the example should be a sentence with an auxiliary verb, for example, "You are bleeding."

The second example is the sentence "Anyone who is interested can see me later." This is a bad example because it is not a statement about the world, it is an expression of willingness on the part of the speaker to help the listeners or readers. One might turn an offer of assistance into a question (I am willing to help you -> Am I willing to help you?) — but it's not the best example. "Can anyone who is interested see me later?" isn't really a question; it's structured that way only for politeness. It really means something like, "My allotted minute is up, so if you are interested in hearing more about this, please see me later." The only responses expected are silence or "Yes"; is not intended to generate a "No" response, as in "No, Professor, I'm interested, but I can't see you later." A better example would be a sentence that inverts to a real question to which one could really answer Yes or No.

A further problem is that a normal question (as opposed to a sentence structured as a question for the purpose of politeness) can be answered literally without conducting a poll. For example, the question "Is the girl wearing the pink dress late for supper?" is answerable by checking to see if supper has started and if the girl is at the table. But the question "Can anyone who is interested see me later?", taken as a literal yes-or-no question, cannot be answered until the entire population of potentially interested people is polled to find out which ones are interested, and whether any of those interested are able to see the questioner later. The absurdity of conducting such a poll to answer the question illustrates that the question is not intended to be take as a literal yes-or-no question. So, it's a bad example; a good example would invert a statement into a question that is intended to be taken literally.

I leave it for more knowledgeable people to improve the examples, ideally based on examples in the literature (to avoid original research), but, unless someone either changes the examples or replies here and explains why these two examples really are good ones after all, eventually I will change the examples in the article. Anomalocaris (talk) 09:40, 13 December 2009 (UTC)

Poverty of the genome.[edit]

When chomsky claimed that "we cannot seriously propose that a child learns the values of 10 to the 9 parameters in a childhood lasting only 10 to the 8 seconds", he ignored the fact that the amount of genetic information humans have evolved since the human/chimpanzee split is less than 6 times 10 to the 7 base pairs, placing nativism even further from being able to explain language acquisition than classical behaviorism was (classical behaviorism is as dead as nativism ought to be, modern behaviorism is based on emergents). This highlights the necessity of something like spontaneous emergence of abstraction from statistical sampling of empirical data. — Preceding unsigned comment added by (talk) 12:30, 30 July 2011 (UTC)

The theory makes absurd predictions.[edit]

The nativism called "universal grammar" actually predicts that adaptation to different languages should, by natural selection, have produced groups of humans genetically incapable of learning foreign languages. That racist prediction have been conclusively disproved in lots of studies. Avoiding falsifiability by avoiding extrapolation of theories to their logical extremes is not scientific at all. Furthermore, there is no way to explain why a vast range of redundant linguistic capacities obviously not needed to build a complex language (no language uses the whole worldwide range and some languages only use a very small fraction of it) should have evolved in the first place. This is explained in more detail on the pages "Brain" and "Origin of language" (and to some extent "Piraha"), all on Pure science Wiki, a wiki devoted to the pure scientific method unaffected by academic obsession with status and prestige. (talk) 13:38, 8 January 2013 (UTC)Martin J Sallberg

Biased article[edit]

This article is just fraught with subtle allusions to try and convey the writers' contempt for Chomsky's theory. It shouldn't be too hard for anyone with an open mind to find them and pick them out, but here's one that's particularly egregious:

"That many linguists accept all three of the premises is testimony to Chomsky's influence in the discipline"

This is just pointless and childish—asserting that the main reason for this theory is accepted by many linguists is because of an irrational dogma. That's just not the case. There are also errors in the article pertaining to some of the basic arguments, for example, the notion that recursion has to be infinite, or that language can't be recursive because of limitations of human mental faculties. Those are statements that stem from an incomplete understanding of what recursion is.

I would in fact argue that the opposite of what I quoted is true: because the popular thing to do in linguistics is to try and disprove UG, particularly anything Chomsky has ever written, there are many of those who write articles like this one, fraught with basic errors and irrelevancies. (talk) 20:28, 30 May 2013 (UTC)

Oh, I agree that that sentence should be removed, and I have removed it, and the uninformative adjacent paragraph. I think it should go because I think very few linguists accept the premises of the poverty of stimulus argument anymore.
It's not a good argument and the number of different approaches to falsifying it attests to how it may be the least supported notion still extant in linguistics. Gold's theorem is a demonstrably unsound argument on purely formal grounds - as Horning showed as early as 1969 - and as far as I can tell, even the surviving advocates of the poverty of stimulus no longer make it. The absence of negative evidence is demonstrably false - not only does poorly formulated language elicit visible negative reinforcement under ordinary circumstance, something every student of a second language knows well, but also empirical study of the language children actually encounter shows a high level of explicit correction of negative use.
Really, this article should be a lot less neutral, more in line with the way we treat caloric theory, luminiferous aether, and the monoamine hypothesis. (talk) 09:06, 10 December 2013 (UTC)

the POS rests on shaky math ?[edit]

the POS is often presented as a mathematically proven idea, but doesn't the detailed math make assumptions that are unrealistic ? (sorta like economics sometimes ?)

also, having young children, maybe I'm more verbal then most parents, but the idea that children don't get a lot of neg criticism - is there any empirical data there ? — Preceding unsigned comment added by 2601:6:3880:CC8:DCF5:EA4D:171A:DD3F (talk) 00:43, 1 June 2015 (UTC)


(part about gold ) there does seem to be a very large literature on empirical studies;perhaps someone can add it to the article — Preceding unsigned comment added by 2601:6:3880:CC8:DCF5:EA4D:171A:DD3F (talk) 00:55, 1 June 2015 (UTC)

Gold's theorem is true and the proof is sound. Not all linguists understand this very well -- some fairly big names still claim conspicuous absence counts as negative evidence for the purpose of Gold's theorem, but it does not. However: If you loosen your definition of language learning from "always recognizes all and only grammatically correct productions" to "recognizes correct productions with an arbitrarily small rate of error", then Gold's theorem simply no longer applies, and a proof of learnability in the abstract has existed since the late 1960s. — Preceding unsigned comment added by (talk) 13:06, 15 April 2016 (UTC)

Weird syntax and inconsistency.[edit]

Original sentence ( before my first edit):

It has been proposed that if a language pattern is never encountered, but its probability of being encountered would be very high were it acceptable, then the language learner might be right in considering absence of the pattern as negative evidence.

The problematic construction can be distilled into the following sentence:

Something would be very high were it acceptable.

I can't understand this - seemingly simple - sentence. Something would be very high if what? What is the meaning of "were it acceptable"? I can't find it in any grammar, neither descriptive nor prescriptive.

Maybe the author made a typo and the intended sentence was like:

Something would be very high where it's acceptable.

How about rephrasing our sentence into:

It has been proposed that if a language pattern is never encountered, but its probability of being encountered is very high, where it is acceptable, then the language learner might be right in considering absence of the pattern as negative evidence.

I've replaced "would be" with "is" for obvious reasons. On the other hand, the whole phrase "where it is acceptable" is redundant because acceptability is no criterion here.

What's more - the original sentence is inconsistent because if something is never encountered then its probability of being encountered is exactly zero, not very high.

So our sentence can be changed to:

It has been proposed that if a language pattern is never encountered, although seems to be intuitive, then the language learner might be right in considering absence of the pattern as negative evidence. (talk) 18:25, 1 February 2017 (UTC) I'm not entirely sure, but I always took "were it acceptable" to mean "if it were acceptable" (using the subjunctive mood), so the sentence would be saying ", but its probability of being encountered would be high if it were acceptable (to encounter it?)". The original sentence is indeed hopelessly involved. I'll let someone more knowledgeable on the topic chime in. Me, Myself & I (☮) (talk) 23:22, 1 February 2017 (UTC)

"were it acceptable" as "if it were acceptable"? Oh my God. This is exactly what I have been waiting for. This is so obvious to me now. Yes - subjunctive mood, and inversion (which is archaic here, but sometimes used). How could I not noticed that? It was an enlightening experience for me. Thank you. (talk) 06:03, 3 February 2017 (UTC)

I can get behind that last proposed variant. It greatly simplifies the presentation concept being forwarded; honestly, that sentence as it was written was almost impossible to parse even for someone with background in this area; I can't imagine how inapproachable it was for the average reader looking for a primer on the the topic. In fact, we might go a step farther to:
It has been proposed that if a language pattern is never encountered, although seems to be intuitive, then the language learner might regard the absence of this language pattern as negative evidence.
Afterall, what we are talking about is what the learner intuitively perceives from the null element, not whether this perception is, in any abstract sense, "right". Snow let's rap 03:47, 2 February 2017 (UTC)

You are absolutely right. There is a huge difference between "might be right in considering X as Y" and "might regard X as Y" in any context. And only the latter has any sense in the context of our sentence. But we can go even further, and make our sentence easier to read:
It has been proposed that if a language pattern seems to be intuitive, but is never encountered, then the language learner might regard the absence of this pattern as negative evidence.
The next step might be:
It has been proposed that if a language pattern seems to be intuitive, but is never encountered, then the absence of this pattern might be regarded as negative evidence. (talk) 06:03, 3 February 2017 (UTC)

Yes. Though bear in mind that by switching to the passive voice and losing the grammatical actor, you make it ambiguous as to who is doing the "regarding" of this theoretical mental association. And since we are talking about the learner exclusively as the focus of this observation (not the outside observer), I would suggest that the second to last option is preferable. Snow let's rap 06:26, 3 February 2017 (UTC)
Good catch. I was afraid of that passive voice too, but now your sobering thought has removed any doubt. I think that "Me, Myself & I" agrees with us too, so I have made the suggested changes to the article. (talk) 09:28, 3 February 2017 (UTC)

Good humour[edit]

The argument has long been controversial within the field of linguistics[citation needed],

I nearly spewed. That's the funniest use of "citation needed" I've seen yet.

controversy (n.) 
see Noam Chomsky.

MaxEnt 15:41, 17 March 2017 (UTC)

Citations added.·maunus · snunɐɯ· 16:09, 17 March 2017 (UTC)