User:Dynwrighter/sandbox/Poverty of the Poverty of the Stimulus/Poverty of the Poverty of the Poverty of the Stimulus

Poverty of the Stimulus (POS) is the argument from linguistics that children are not exposed to rich enough data within their linguistic environments to acquire every feature of their language. This is considered evidence contrary to the empiricist idea that language is learned solely through experience. It is supposed that the sentences children hear while in their period of learning a language do not have rich enough data for them to correctly achieve an adult grammar comparable to other speakers of their language.

The POS is closely related to the concept of universal grammar. This is the idea that there is a definable and limited set of grammatical rules that humans can acquire.

Both poverty of the stimulus and universal grammar are terms that can be accredited to Noam Chomsky. Chomsky coined the term "poverty of the stimulus" in 1980, however he argued for the idea since 1965 in his book Aspects of the Theory of Syntax.

Background
An argument from the poverty of the stimulus generally takes the following structure:
 * 1) Human children learn some grammatical feature of their native language.
 * 2) Speech input that children experience is consistent with numerous possible grammatical structures that could produce such input.
 * 3) Input that children experience does not have evidence that distinguishes the right grammatical structure from all the other possible structures.
 * 4) Children nonetheless learn the right grammatical structure.
 * 5) Therefore, the right grammatical structure arises due to some (possibly linguistic) property of the child.

Innateness and Linguistic Nativism
Linguistic nativism is the theory that humans are born with some knowledge of language. One acquires a language not entirely through learning. According to Noam Chomsky, "The speed and precision of vocabulary acquisition leaves no real alternative to the conclusion that the child somehow has the concepts available before experience with language and is basically learning labels for concepts that are already a part of his or her conceptual apparatus." Chomsky's view that the human faculty of language is innate is also affirmed by Steven Pinker. One of the most significant arguments generative grammarians had for language nativism is the poverty of the stimulus argument. In this argument, Noam Chomsky proposed that the input a child receives during language acquisition is insufficient to account for the grammatical structures they acquire. To be exact, he said that, "the native speaker has acquired a grammar on the basis of very restricted and degenerate evidence". Similarly, Pinker concludes that humans have a system that is more sophisticated than what they are exposed to.

Pullum and Scholz summarized the properties of a child's environment. They identified properties of positivity, degeneracy, incompleteness and idiosyncrasy. Under positivity, they assert that children are only exposed to positive linguistic data. Moreover, there is lack in negative data that aids a child in identifying ungrammatical sentences that are unacceptable in the language. Lasnik and Lidz identify the degeneracy in the input that children receive in terms of scope and quality. The input can neither provide evidence for all types of sentences that children will encounter nor contain information that tells them how to represent the grammatical sentences they hear. Furthermore, the linguistic data each child is exposed to is different (i.e. idiosyncrasy) and there are many utterances, compatible with the adult grammar, that a child might not have heard (i.e. incompleteness). However, despite the properties mentioned above, children would eventually be able to deliver a linguistic output that is similar to the target language within a relatively short amount of time. In contrast, when placed in certain environments, other organisms are unable to attain the language mastery humans have reached. From the nativists' point of view, all of these highlight that humans are hard-wired with a UG and thus support the innateness hypothesis.

However, it is important to note that the argument that the poverty of stimulus supports the innateness hypothesis remains very controversial. For example, Fiona Cowie claims that the Poverty of Stimulus argument fails "on both empirical and conceptual grounds to support nativism".

History
Although Chomsky officially coined the "poverty of the stimulus" theory in 1980, the concept is closely related to another Chomskyan approach named Plato's Problem. He outlined this philosophical approach in the first chapter of the "Aspects of the Theory of Syntax" in 1965. Chomsky asserts that there is a physiological component in the brain that develops in children, and thus, they are able to acquire language universally. Plato's Problem traces back to "Meno", a Socratic Dialogue. In Meno, Socrates unearths knowledge of geometry concepts from a servant who was never explicitly taught them. Plato's Problem directly parallels the idea of the innateness of language, universal grammar, and more specifically the poverty of the stimulus argument because it reveals that people's knowledge is richer than what they are exposed to. Chomsky illustrates that humans are not exposed to all structures of their language, yet they fully achieve knowledge of these structures.

Evidence
An overarching theme in examples provided as evidence for the poverty of the stimulus is that children learn to form language following certain rules, even though the speech that they hear is compatible with multiple generalizations. Children are not directly explained the grammar of their language, so if they cannot learn all aspects of the grammar from speech exposure, the gap must be filled in by properties of the learner.

Binding theory - Principle C
Children will not interpret sentences such as these in a way that violates Binding Theory. For example, they do not accept that in 2, 'He' and 'the Ninja Turtle' are the same character, even though they accept such an interpretation in 1. Children exposed to pronouns such as 'he' will see that their reference can vary based on discourse context. However, in contexts such as 2, an additional constraint, Principle C, further restricts possible antecedents for a pronoun. Children show this constraint even though there is insufficient evidence in the speech they hear for them to have learned Principle C of Binding Theory. That is, Principle C rules out certain interpretations of a sentence, but since negative evidence does not exist in language acquisition, there is no way that children could have learned the contexts in which such an interpretation is impossible. Children are not exposed to meanings that a sentence cannot have, thus it is argued that this constraint is innately acquired.
 * 1) While he was dancing, the Ninja Turtle ate pizza.
 * 2) He ate pizza while the Ninja Turtle was dancing.

Passives
In the above examples, the first sentence is being turned into a passive construction. Between 2 and 3, different noun phrases are moved to form the passive, either the smaller noun phrase: 'the dog', or the larger phrase: 'the dog's owner'. This movement could be accomplished by a linear rule, such as ‘move the first NP (the dog) after the verb to the subject position,’ or by a structural rule such as ‘move the highest NP (the dog’s owner) after the verb to the subject position.’ These two possibilities are distinguished between 2 and 3, as under the linear rule both would be grammatical, but under the structural rule only 2 is grammatical. In adult speakers, we see that only 2 is grammatical, however, children may not be exposed to enough of these distinguishing sentences to guarantee that they will learn the structural rule. As such, the fact that all adult speakers agree that 2 is grammatical and 3 is not suggests that the linear rule was never even considered and children are predisposed to a structure based movement system.
 * 1) I believe the dog's owner to be hungry.
 * 2) The dog's owner is believed to be hungry.
 * 3) The dog is believed's owner to be hungry.

Anaphoric "One"
Say a child is presented with two balls, one red, one blue. His parent might say to him: “Look! I have a red ball. Can you point to another one?" In this instance, it is ambiguous whether the speaker is referring to another ball or to another red ball by using the word ‘one’. This is the anaphoric ‘one’, the ability in English to replace noun phrases with the word "one." Even in situations where the speaker means ‘another red ball’, the speech that the learner hears could also mean simply ‘another ball’. This is because when a speaker refers to a red ball, they are inherently referring to a ball since the set of red balls is a subset of balls in general. Despite this ambiguity, the children show they believe ‘one’ to refer to ‘red ball’ and not ‘ball’, since they do not show evidence of believing ‘blue ball’ is a possible referent. As can be seen from the following examples, "one" can be used to replace different sections of a noun phrase.


 * 1) The girl has a striped water bottle.
 * 2) Who else has one?
 * 3) Who else has a striped one?

In item 2., it is evident that "one" can refer to either "striped water bottle" or "water bottle." However, 18-month-olds have been shown to think it means "striped water bottle," replacing the entire noun phrase instead of just the noun itself. The evidence available is systematically ambiguous, which means that this sentence and other possible sentences do not show the infant that "one" means "striped water bottle" and not "water bottle." Despite this ambiguity, children appear to have learned the more narrow interpretation.

Wh-Movement/Structural Dependence of language
The speech that children produce is always dependent on a structure, even though the evidence that they see is not explicitly presented as structural. When syntactic movement occurs, the surface structure of the sentence is open to multiple interpretations which cannot be narrowed down by what children hear alone. Since sufficient evidence for children to learn the structure does not exist, and yet they display a pattern reflecting that they know it, an argument from the poverty of the stimulus is formed. For example, the movement of question words from the verb with which they are associated, known as wh-movement, is restricted to certain syntactic structures. If the question word moved from a forbidden area, a syntactic island, the sentence will be ungrammatical, as in the second example below: This suggests that the grammaticality of the sentence is based on the it’s syntax. They could theoretically choose to move certain words without taking into account the underlying grammar of the sentence, but even when making errors in production, they do not do so. This ability, despite a lack of material from which it could be learned, forms an argument from the poverty of the stimulus and has been used as evidence that children are innately wired to build a hierarchical structure when learning language.
 * 1) What did you claim that Jack bought?
 * 2) What did you make the claim that Jack bought?

Phonology
Phonology refers to the mental representations of speech sounds, both in production and perception. The Poverty of the Stimulus argues that humans are predisposed to be accepting of certain phonological rules and patterns. According to this theory, humans are born with a sense of which phonological rules and patterns are possible, and use that sense to make specific generalizations about the language input they receive, building their grammars accordingly. The Poverty of the Stimulus argument does not state that all phonological knowledge is innate; however, it does suggest that a large portion of theoretically possible generalizations have already been eliminated at birth.

In Bergelson-Idsardi (2009), people were tested to determine their acceptance of novel constructed "words" with varying Consonant-Vowel stress patterns. English-speaking adults (tested through computer software) accepted more “words” whose stress patterns exist in non-English languages than they did of “words” whose stress patterns are not found in any human languages. Eighth-month-old children (tested via the Conditioned Headturn Procedure) were found to have the same preference as adults. The only distinctions between the two sets of "words" were regarding whether or not they existed in any human language, and this was arbitrary enough that the researchers could not identify another plausible reason for the preferences among the young children. Because of this, they concluded that human language acquisition mechanisms are "hardwired" to lead infants towards certain generalizations, supporting the argument for the Poverty of the Stimulus.

On the general subject of untaught knowledge, Halle (1978) points out that native speakers can identify and pronounce English words they've never seen before, and that they can also consistently apply rules to sounds that do not exist in English. Halle writes that English speakers are likely to be accepting of words like thole and flitch, while rejecting words like ptak and rtut, even though none of them are common enough to have been seen before. Halle also writes that English speakers consistently pluralize the German name Bach (pronounced /bax/) as /baxs/, despite not having any experience with the /x/ sound, which is nonexistent in English. Since there is "no indication" that speakers could have acquired this knowledge, Halle points to this being innate, unlearned, and untaught.

Semantics
The Poverty of the stimulus argues that there is something besides input that determines the generalization of how grammar works. UG is an indirect argument of the poverty of the stimulus, and is a guide to develop a grammar. Howard proposes that this is done with the use of semantic constraints as a way to force grammars to have some properties and not others. In talking about poverty of the stimulus and how the mind acquires and processes language, universal grammar comes up as an indirect argument. The Oxford Handbook of Universal Grammar describes the innate knowledge of language as something that "makes it possible for learners to acquire the complex system of knowledge that undergirds the ability to produce and understand novel sentences."[21] It is the constraints of Universal Grammar that makes it possible for learners to navigate their experiences in the identification of a grammar for the language they are exposed to. Cowie on innateness and language also talks about the importance of these constraints, the constraints in the universal grammar is what allows the understanding of a language[22].

An example of how the constraints involved in learning how a grammar works is the word learning process. When the learner hears the word in the environment, they can only see particular referential contexts from the environment. (e.g., for the word "dog" we see particular dogs). we never get information on what the category of that word is, or if it has a category at all from the environment. Neither do we get what the specific thing in the environment that we see counts as that word. it could be any part of what we see. The input/environment alone does not help determine this.

Gleitman in her work on "hard words", gives an explanation for the understanding of credal verbs (e.g., think and know), she explains that a "multiple-cue processing machinery" is what people use to people produce speech and parse the speech that they hear[23].

Word-learning increases significantly around 18 months[24]. This drastic change in acquisition signifies a change in constraints directly related to word-learning. Around this time, children use mutual exclusivity to constrain the possible meanings of objects. In an experiment, children were familiarized with an object and then presented with two objects (one familiar and one unfamiliar). When given a novel object label, the children always applied it to the novel object[25].

Another argument under semantics is the Principle C argument, which states that when a child hears a sentence with reflexives nothing in the environment tells them that the reflexives make reference to one person and not the other. For example, in the sentence "He doesn't think John came." nothing tells them that John is not "he", especially since John is not present. Nothing in the environment tells them that they can't produce a sentence like that and mean John is "he".

Other Arguments
There are some arguments that suggest the Poverty of the Stimulus argument, a nativist theory, is incomplete and lacks enough evidence to refute the empirical view that posits that language is learned primarily from external sources (as opposed to nativist internal innate hypotheses). The Poverty of the Stimulus argument is not a single hypothesis, but a number of hypotheses all supporting the nativist notion that there is something inherent about the language learning process. Because it is a multifaceted hypothesis, there are many areas of criticism. Some argue that from a logical, academic standpoint, "absent evidence" (such as whatever is "innate") cannot be proven by the methods so far used in experiments showing evidence of "absent evidence." Other areas of criticism range from discussions about what evidence or lack thereof belongs to a cognitive linguistic theory of learning or another cognitive or psychological phenomenon that isn't per se linked to linguistics.

The empirical views suggest that language can be learned with mental processes originally meant for other modes of cognition, and that there need not be a concept of innateness in order to account for the difference between the input a child receives versus the language they develop. Natural languages are composed of different statistical cues that children are able to learn. Children have different statistical learning mechanisms that they seem to employ when learning a language. This is shown when they are attempting to learn different aspects of semantics. The children observe a consistent pattern of a certain word referring to a certain object in order to learn the meanings of words. The also use similar processes in phonology by observing statistical patterns in speech in order to determine the phonemes of a particular language.

Another argument that represents this view is that input experienced by the infant is not as transparent as the poverty of the stimulus claims, and thus infants do not depend on an innate mechanism to learn language. This argument claims that the assumption that children do not learn language in the very early stages of life (e.g. in utero) is inaccurate. This inaccurate assumption has led the Nativist school of thought to claim that any sort of language learning that occurs in these early stages of life must be a product of innate mechanisms found within the mind of the child, and thus the input that the child is presented with is too transparent to acquire any sort of language. However, there has extensive research done that suggests that children are able to perceive language in utero. For instance, newborns are able to distinguish between their mother’s voice and other women’s voices. This may suggest that children may not have to rely on innate mechanisms in order to learn a language at a young age (i.e. the early stages of life where language acquisition was commonly believed to be impossible). The information that the child learns even before it is born can be used when attempting to learn other aspects of the language later in life. Therefore, it is possible that children used the information they gain from experience in order to learn language.

A computer simulation was composed to test if different computational models (i.e. synthetic models), that are similar to human computation models, could acquire a grammar based on positive evidence alone. This synthetic model was given common input that an infant would receive. The computer models were able to extract statistical information from the input alone, allowing the model to build a basic grammar. Given that these synthetic models, which were constructed based off human computation models, were capable of acquiring a basic grammar based on the positive evidence alone, it may suggest that children are able to do the same. Even though this does not expo-facto show that only positive evidence is needed for language acquisition, this does suggest that it is possible. Thus, this argument attempts to relate synthetic computational models to the computational models that children employ, in order to suggest that language acquisition is possible via experience of the input alone.