Verb Acquisition in Young Children
From PsychWiki - A Collaborative Psychology Wiki
Theories of Verb Acquisition in Young Children:
Semantic Bootstrapping Versus Distributional Information
[Carlyn Carter http://www.psychwiki.com/wiki/Carlyn_Carter]
At birth, we are unaware that other people have thoughts or emotions, and we have no concept of having a mind of our own. We are unable to lie because we assume our thoughts are public knowledge, and we are unable to fully understand what will become our first language. All of this is true for the first few years of our lives. Language is an especially impressive skill for a developing mind because it is very new in our evolution and is seemingly more of a learned ability than an innate one. As our brains develop and our language environments richen, we are able to hear, speak and write the most important elements of a complex language (or two or three) by the time we enter grade school. Theorists to this day argue about the ways in which we acquire language and its many parts. This paper will focus on the various theories of verb acquisition, the most grammatically complex syntactic category or word type.
Contents |
The Typical Child Development Process
The typical two-year-old child has a vocabulary of a few hundred words and composes sentences of two to three words in length. By contrast, the typical five-year-old child has a several-thousand-word vocabulary and nearly perfect grammar. The question most researchers try to understand is: what mechanisms influence such rapid learning? Fast-mapping, an observed phenomenon that explains how exposure to unchanging words are mapped in the brain, explains to some extent that nouns and adjectives are just memorized based on exposure. When it comes to words that change, such as with the conjugation of verb tenses, the conceptual model grows in complexity.
The question of what occurs in verb acquisition was answered long ago. It is well known that young children use un- as a prefix to negate verbs before they learn verbs of opposite meanings. For example, a child will turn the word come into uncome, instead of saying leave and will change hate to unhate instead of using a word such as like (Whorf, 1956). Bowerman (1982) created situations in the lab where children were asked to express a command to fix a situation. This would increase the urgency, making children respond as quickly as possible, not giving them much time to think. For example, a researcher would show a picture of a stranger holding onto a child and the participant would be asked what he or she would say to the stranger. The most common response was Uncapture me!.
The use of suffixes and helping verbs are also commonly found during the refinement of verb use. Words or words endings used to make a sentence grammatical are called grammatical morphemes. For example, in the sentence Holly is going to the store, both is and the ending –ing are grammatical morphemes. Adding these bound morphemes to a base word is appropriately called affixation. Around ages two and three the suffixes –s and –ed are overused. Sometimes this overuse is in the form of redundant verbs, where the old method of adding an –ed ending (assuming regularity) is used in conjunction with the irregular conjugation. For example, a child of this age might say he wented or he camed. Clark (1993) found that around two years old, children use conversion, turning nouns into verbs. In this case, a child might say I’m souping or I noised, both of which have obvious meanings but are grammatically incorrect.
Verb Tenses
The acquisition of different verb tenses has been developmentally charted. Brown (1973) explored the order of acquisition of verb tenses and found the following. First, the progressive tense (i.e. –ing) is always mastered before the past tense (i.e. –ed). Second, there are time differences in the stages of verb acquisition. For example, some children may have a four-month gap between a mastery of the progressive tense and the past tense, while others may experience a gap of about a year. Clark (1992) introduced three principles of verb acquisition. Following the Transparency Principle a young language learner uses elements to form new words (e.g. -er added because it is consistent and has a transparent meaning). Under the Simplicity Principle the young language learner uses as few changes as possible: 1st - Conversion, 2nd –Affixation. Finally, following the Productivity Principle, the young language learner prefers lexical rules once learned (e.g. the subject in a sentence usually comes before verb) and overuses them at first.
Context
The issue of what occurs in verb acquisition is obvious and undeniable. Still, the question of how verbs are acquired and what underlying mechanisms are at work is still under investigation. Familiarity of context seems to be an obvious way in which children learn verbs. Some feel that the meaning and function of a verb within a sentence provides enough semantic information to make the verb’s meaning and function transparent and, therefore, the verb is quickly learned (Pinker, 1984). Others feel that a smaller unit of the sentence (specifically the word preceding the verb and the word following the verb, which forms a frame) will allow a child to understand a verb’s function and meaning after repeated exposures (Mintz, 2006).
Brown (1957) showed that given a scene and a novel word, three- to five-year-old children’s interpretation of how the word relates to the scene changed depending on the word’s morphosyntactic environment. When shown the action of kneading something, children who heard the scene using the word sib as a verb thought that it referred to the kneading action; whereas if they heard a sib, they thought sib referred to the bowl. Therefore, the grammatical category of the word determines the assumptions children make about word meaning. Children pay attention to the syntactic privileges of words to determine whether a word is a noun or verb.
Distributional Analysis
Maratsos and Chalkley (1980) introduced the idea that children perform distributional analysis, assessing the order of words in a sentence, on their input to identify grammatical form classes. They found that children tracked the range of environments in which words occurred, including co-occurrence with other words and affixes, and grouped words together that occurred in overlapping environments.
In his recent experiment with 14-month-old infants, Mintz (2003) explored how frequent frames function in teaching what words in the English language are verbs. This type of distributional information, which is the patterning of words in a sentence, is specifically three respective words with the verb as the middle word of the three. He believes these frequent frames provide a bootstrap, or an initial basis, for categorizing and identifying verbs. His sample consisted of twenty-four infants averaging 12 months, which were randomly assigned to two groups. Each infant heard two nonce (nonsense) verbs and two nonce nouns in a total of 12 familiarization contexts. For the verbs specifically, there were only 6 familiarization contexts. The first four pair groups introduced each consisted of two identical sentences with a different nonce verb in each sentence. This provided the distributional information. Then, each infant was presented with two new pairs of sentences that use two formerly seen nonce verb pairs, with one novel grammatical and one novel ungrammatical test sentence. Frequent frames surrounding the verbs were the most common frames found to surround verbs in the English language (e.g. you_____the, you_____it).
In order to determine whether the infant recognized that the novel ungrammatical sentence was indeed ungrammatical, Mintz used the Headturn Preference Procedure (or HPP), which is a validated indication of recognition of improbable information for infants. For Simply, something that does not fit a learned rule, such as an object defying gravity, elicits a longer headturn toward that object than something that fits the rule because it requires more time to process this new information. Jusczyk and Aslin (1995) validated this procedure using nonce words as stimuli.
Overall, Mintz found that infants listened longer to ungrammatical over grammatical strings. He interpreted these findings to mean that infants’ sensitivity to the difference means they categorized the nonce words based on distributional information in the familiarization of sentences. This point was strengthened given that the only systematic difference between the two sentence types was the distributional category of nonce words. When comparing the results of the noun frames to those of the verb frames, Mintz concluded that there is an advantage for the verb categorization over the noun categorization. Something more than just distributional information contained within the familiarization material is giving these infants an advantage.
In a study with slightly older, 15-month-olds who were learning German, Hohle, Weissenborn, Kiefer, Schulz and Schmitz (2004) found evidence that infants use categorical information to categorize novel nouns, but not novel verbs. Hohle et al used bigrams as an alternative type of distributional information to frames. A bigram is any two word sequence in an utterance, and typically, one word provides the distributional context for the other. They showed that the novel words following a determiner were categorized as nouns but novel words following a subject pronoun were not categorized as verbs. They attributed this difference to the fact that determiners are more predictive of nouns than pronouns are of verbs in child-directed German. It seems that Hohle et al also linked categorization to distributional properties in the input. Both this study and that of Mintz suggest that categorization is driven by the informativeness of distributional contexts.
If a grammaticality effect had resulted for both verb frames and noun frames, one could not with certainty conclude that categorization was due to the frames being frames in English. Meaning, the patterning of the nonce words in the set of familiarization sentences could provide evidence to a distributional learner who lacked experience with English that certain words simply belong together (Mintz, 2002). Evidence of strictly experiment-internal distributional learning would certainly be interesting, but the present pattern of results suggests that this is not the critical process at work. The immediate distributional environments of the nonce nouns and the nonce verbs differ in the amount of prior exposure they have had, which can confound the results in these studies. Additionally, frames are not guaranteed to be the only categorization information children used in Mintz’s study, nor are the bigrams (the word before the verb plus the verb or the verb plus the following word) that Hohle et al used.
Although distributional information as the sole indicator for categorization for children is an appealing theory, it is not without problems. Some of these problems are so serious that the approach is sometimes not considered viable (Pinker, 1984, 1987). One possible problem is regarding how a learner identifies the appropriate distributional contexts on which to base an analysis. For example, the absolute position of a word in a sentence (e.g. 1st word, 3rd word, etc.), while a perfectly reasonable piece of distributional information, is not a sufficient basis for categorizing words (Pinker, 1987). Words are constantly changing position. Verbs can occur at the end or the beginning of a sentence, ruling out a frame or bigram as distributional information. Thus, distributional environments appropriate for categorization must be defined relative to other words and morphemes in an utterance.
Semantic Bootstrapping & Arguments against distributional analysis
Most arguments against distributional analysis as the sole means of categorizing words posit that learners would already have to have a considerable amount of linguistic knowledge. The argument further argues that this linguistic knowledge is imperative in knowing how to treat the distributional information in a linguistically meaningful way. Only once a considerable amount of structure is already fixed in the child’s developing grammar (and that categories had already been assigned to many words and affixes by non-distributional means) can distributional information play a role in determining the category of an unknown word. Pinker uses the term structure-dependent distributional learning (Pinker, 1984, p. 40-42) to refer to distributional analyses that are guided by some knowledge of phrase structures, and knowledge of the category membership of some other words or inflections in a sentence. In short, although a relatively advanced language learner could use distributional information to categorize a novel word, many theorists believe that other sources aid in categorization as well. These vary from biological predispositions to neural pathways to prior (perhaps unidentified) environmental influences.
Semantic information is often thought to be a categorization strategy competing with or working in conjunction with distributional information. Given the link between grammatical form and meaning, some researchers believe that learners initially determine the grammatical category of a new word by observing what kind of entity it refers to. If it refers to an object, or a substance, then it is a noun; if it refers to an action, then it is a verb. Pinker (1984) most clearly articulated this idea and termed it the Semantic Bootstrapping Hypothesis. According to Pinker, aspects of the meaning of an utterance (who and what are being talked about) are transparent to learners, even before they have acquired much knowledge of the vocabulary and structure particular to their language. This allows learners to identify the semantic category of a word (e.g., “action word”) by observing the referential contingency of the word’s use (e.g. referring to an action). Innate linking rules then allow the child to classify the word syntactically (e.g., as a verb). This kind of classification involves the ability to note morphemes that are combined in the same order as they would be if they were separate words in a corresponding construction. The newly categorized word can then be fit into the developing grammar, and early on, may be used as a source of information for determining certain language-specific aspects of the grammar.
Pinker (1991) studied how semantic bootstrapping is utilized by presenting his participants with nonsense verbs in various contexts, so that distributional information alone could not help them categorize the verbs. He found that semantic information must have been used to determine the verbs were action words, a type of semantic information. Similarly, Peccei (1991) studied semantic information, also using nonsense words. Peccei presented participants with novel, nonsense words and asked which of the two words was to receive a certain suffix. For example, the nonce words sugarize and saltize were presented to four-year-olds and adults. Both cohorts answered that the –ize suffix was appropriately attached to sugarize but not correct at the end of the base word salt. Though both of these words are made up, these results show that children and adults alike have a grasp on semantic rules in English. Here, the rule that –ize is added to two syllable bases and not to one syllable bases is exemplified.
While semantic bootstrapping is defined as using semantics to cue what a verb means, syntactic bootstrapping is a newer term in the literature defined as identifying morphemes commonly found together to understand verb meaning. Both follow Pinker’s ideas that distributional information alone does not make for categorization but rather that context can serve as bootstrap or initial basis for verb learning. Studies under the heading of syntactic bootstrapping all demonstrate that syntax guides young children’s interpretations during verb learning but there are two distinct theories about how this happens. The first perspective, the “universalist” view, holds that syntactic bootstrapping falls out from universal properties of the mapping between lexical (words and morphemes) and syntactic semantics (Glietman, 1990; Pinker, 1989). This is a kind of top down conceptualization, asserting that children categorize verbs based on multiple types of semantic information. The second perspective, the “emergentist” view, holds that argument structure patterns emerge from a process of categorization and generalization over the input. This is also a top down conceptualization, asserting that semantic information follows the categorization of these verbs. In this view, there are no constraints on what a word means from human nature. Most researchers are interested in the first perspective, which also must assume that there are not many cross-linguistic problems concerning semantic bootstrapping and syntactic bootstrapping.
This creates a challenge because in addition to words not having a clear definition determined by nature, syntax differs so much that it is seemingly a miracle that children are able to map the correct pair of verb and its meaning. Verb-world parings support numerous verb-meaning mappings, so syntactic bootstrapping has emerged as one explanation of how children categorize verbs. In syntactic bootstrapping, children use the syntactic frames in which verbs are placed and associated morphemes to help discover the verbs’ meanings in English (Glietman, 1990). In certain languages (e.g., Mandarin Chinese) noun phrases, such as the direct object, are not always present after verbs because this information is included in other words. The effectiveness of syntactic bootstrapping seems to be slight or it may not even apply (Bowerman & Brown, 2006; Rispoli, 1995).
Other Languages
The presence or absence of the direct object distinguishes transitive from intransitive verbs in English because English requires the overt expression of a verb’s argument structure. But in languages that do not have this requirement, language learners may not make such information about verb meaning. In Mandarin Chinese, one can simply describe a child carrying a puppy to her grandmother with the verb alone, without any need to mention what is to be carried or who is to carry it (C. Li & Thompson, 1981). The same scene can be described with transitive (e.g., carry) and intransitive (e.g., walk) verbs in practically the same frame. For example, “carry it to Grandma who is over there” can be said in four words in Mandarin Chinese.
This four-word sentence in Mandarin presents a challenge to the use of syntactic bootstrapping because this theory is based on the notion that verbs that vary in their meanings also vary in their syntactic expressions. However, the mere existence of these utterances need not be regarded as proof that syntactic bootstrapping could not operate in Mandarin, because syntactic bootstrapping has always been presented as probabilistic, rather than a categorical procedure (Gillette et al., 1999; Glietman, 1990). It may not be the case that transitive verbs must appear one-hundred percent of the time with their direct objects but instead that the presence of the direct object must be a reliable cue to transitivity. An example of a reliable cue to transitivity is the fact that postverbal noun phrases appear more frequently with transitive verbs than with intransitive ones (Naigles & Hoff-Ginsberg, 1995). So how frequently or reliably do transitive verbs appear with their direct objects in languages in which noun phrases are not always present after verbs?
Although several researchers have found null results when trying to connect syntactic bootstrapping to languages where the noun phrase following a verb is not always present (e.g., Bowerman & Brown, 2006; Rispoli, 1995), these studies all used small numbers of utterances as their stimuli and may have had more sentences without a noun phrase following the verb than is typical of these languages overall. More recent studies have yielded higher percentages of utterances that included the verb’s direct object. For example, Tao (1996) analyzed over 600 utterances from Mandarin speakers engaged in adult-adult conversations and found that over fifty percent of the utterances with transitive verbs included the direct object. D. Cheung and Brooks (2002) examined over 2,000 child-directed utterances of Cantonese (very similar to Mandarin) and reported that over fifty percent of these utterances containing transitive verbs included a postverbal object.
Lee and Naigles (2005), claim that syntactic bootstrapping works for Mandarin Chinese just as well as in English and were able to replicate results of Naigles and Hoff-Ginsberg (1995), who found that reliable cue to transitivity is the fact that postverbal noun phrases appear more frequently with transitive verbs than with intransitive ones. In their study, although transitive verbs in Mandarin did not appear with postverbal noun phrases as frequently as they did in English, the strength of the positive postverbal noun phrase cue was equivalent in the two languages, as over eighty percent of utterances with this frame appeared with transitive verbs. Both Naigles and Hoff-Ginsberg and Lee and Naigles found that more than three quarters of the verbs in each of their sets of utterances were presented in multiple frames, supporting the potential use of multiple frames in verb acquisition.
Issues in Semantic Bootstrapping
Semantic bootstrapping theorists, just like distributional information theorists who measure frames and bigrams, also run into problems. One problem is that semantic bootstrapping has to do with the difficulty of identifying the meaning, and thus the semantic category, of unknown verbs (Gillette et al., 1999; Gleitman, 1990). The problem is not that verb meaning cannot be recovered, because children clearly do learn verbs eventually. Rather, because of the ambiguities found in verb-to-world mappings, learners apparently rely on structural information in the carrier utterance to focus them on the relevant aspects of the world. For example, when confronted with scenes in which a causative and non-causative action simultaneously occur (e.g., one character feeds the other character; the latter character eats what is fed to her), two-year-olds interpret novel verbs in transitive constructions (affecting something else) as referring to causative actions (feeding) and novel verbs in intransitive constructions as referring to non-causative action (eating) (Fisher et al., 1994; Naigles, 1990). The latter character’s action could be interpreted as an adjective (e.g., “he is eating” could easily mean “he is overweight” or “he is pleased”). But semantic bootstrapping as the sole explanation for mapping is limited, because that structural information (transitive/intransitive) would not yet be available in early stages of verb learning, as that is what is to be deduced after the category is determined. Therefore, this information would not yet be available to act as a cue of what the verb refers to.
When syntactic cues are available, accounts of semantic bootstrapping assume that a learner can deduce the referent of (what is referring to) a novel word by observing many situations in which the same unknown word is used to describe different scenes. Then, the learner parses out the elements that are common across all uses to arrive at the correct interpretation. For example, the child may hear the word feed used in situations in which there is no eating (e.g., someone can feed a dog without the dog actually eating), and may hear the word eat used in a situation where there is no feeding (e.g., later when the dog actually eats the food).
Some very valid arguments have been made against this, contending that cross-situational comparison cannot solve some of the logical problems in identifying a word’s referent, without resorting to syntax (Fisher et al., 1994; Gillette et al., 1999; Gleitman, 1990). The events of chasing and fleeing are found commonly in the literature to exemplify this. Any situation in which there is a chasing event, there is also a fleeing event. Therefore, no amount of exposure to the fleeing/chasing concept would clear up the mapping ambiguity of whether it is a noun or verb. Fleeing may be stated when chasing is not (although, logically, chasing is also always happening) and vice-versa. In the ‘dog eating’ example above, two grouped verbs being stated alone seemingly establishes that the two are not conceptually bound, that one does not have to occur if the other. However, in the fleeing/chasing example, the two are often stated alone but are never conceptually separate (Fisher et al., 1994; Gillette et al., 1999; Gleitman, 1990). Either cross-situation comparison leads verb learners to make false conclusions about concepts like fleeing and chasing not being related or syntactic cues are not the only way a learner determines the meaning of verbs.
Semantic driven categorization has another problem. Even if the learner could reliably recover the semantic type of a word, the links between semantic and grammatical categories are not one-to-one but are instead many to many. In part, this is because there are words for which the semantic antecedent conditions (e.g., action implies verb) do not come into play. For example, ‘know’ and ‘love’ are not actions but they are verbs. Linking rules would not be relevant for categorization, as many semantic types map to the verb category other than action. Semantic-bootstrapping theories have proposed structure-dependent distributional learning to approach this problem. Another part of the problem is that semantic-to syntactic linking rules are subject to one-to-many mappings as well; that is, one semantic type can be associated with several syntactic types (Maratsos & Chalkley, 1980). Nouns and adjectives can have action-like semantics just as verbs do, so learners could apply action-word-to-verb linking rules to words other than verbs. For example, the words ‘action’ and ‘noisy’ are not verbs, but Maratsos & Chalkley argue that they are enough like verbs in their semantics that they would be mapped onto the verb category. Distributional analyses, such as frames and bigrams, do not have these problems because the initial categorization is not dependent on accessing word meanings and the unreliability of semantic linking is not a factor.
Recent research
In recent years, researchers have started to re-examine the usefulness of distributional methods in providing an initial classification of words in child directed speech (Mintz, 2003; Mintz & Bever, 2002; Redington, Chater, & Finch, 1998), perhaps because of all the problems surrounding semantic driven categorization. By analyzing transcripts of input children actually receive, researchers have set out to uncover whether the arguments raised against distributional approaches in principle are indeed problematic in practice. A new trend in this literature is to take a statistical approach by measuring frequency when grouping words with other words in the input. If the problematic aspects of distributional information are relatively rare, then it may be possible to filter out the problematic cases via sensitivity to frequency. Rare cases can be important in research, but it is not relevant to those who are seeking to understand the initial categorization of verbs.
Researchers are providing the field with more and more evidence that infants are sensitive to patterns of regularity in their linguistic output. For example, infants notice patterns in syllable sequences that are highly predictable (Saffran, 2001). These kinds of patterns involved computing relations between adjacent languages; however, there patterns in languages that involve non-adjacent elements. For example, present progressive sentences in English involve the morpheme is preceding the verb and the suffix –ing added to the end of the verb, with intervening material, as in the sentence She is reading the book or She is slowly reading the book. Eighteen-month-olds are sensitive to these non-adjacent dependencies, as they notice when an ungrammatical word (e.g., can) occurs in the place of is in the example above (Santlemann & Jusczyk, 1998). It seems that infants store information about elements that occur both at a distance and surrounding other material, which is the type of pattern that constitutes frames in the sentence She is reading the book example but not necessarily in the She is slowly reading the book example.
More evidence that young learners attend to non-adjacent dependencies (support for frames) comes from work by Childers & Tomasello (2001). In their study, two-and-a-half-year-olds were instructed to produce novel action-verbs in transitive frames. Children performed better when the task was modeled with verbs that were embedded in frequently occurring pronoun frames, meaning the verb category was more readily accessible when the verbs fell within the frames that occur more frequently in English.
There are some arguments that frames are structurally more restrictive than bigrams because they involve a relationship between the context elements themselves (the framing words), in addition to the relationship between context and target word. Also, bigrams contexts have an advantage over frames, because frames used for categorization are required to be frequent, which constrains the context (frame) even more. Evidence for using frames for research in verb acquisition deal with the nature of verbs. In most simple transitive sentences, the verb’s arguments form frames around it. These verb frames include, on average, contain many more pronouns than full nouns and the set of pronouns in these subject and object positions are very small and frequent. Childers & Tomasello (2001), as mentioned above, found learning advantages with pronoun frames, which also happened to be the most frequent frames. Whether frames or bigrams help young learners acquire verbs is an academic argument that continues to pervade word acquisition literature but mostly within the distributional information paradigm.
The larger remaining debate concerns whether distributional information or semantic bootstrapping is the initial means of categorization and the discovery of verb meaning. Either way, the language environment undoubtedly teaches young learners the meaning of words. Children learn by pulling information from context. It would be ridiculous for a parent to define words before speaking them in sentences. The child needs the parent to speak so that the words they learn have significance. Some researchers argue that the meaning and function of a verb within a sentence provides enough semantic information to make the verb’s meaning and function transparent and, therefore, the verb is quickly learned (Pinker, 1984). Others feel that a smaller unit of the sentence, such as a frame or bigram, will allow a child to understand a verb’s function and meaning after repeated exposures (Mintz, 2006). Both approaches have advantages and problems. It is possible that both could distributional information and semantic bootstrapping; however, one of the two is still thought to be the initial means of categorization. It is also possible that there are individual differences in the methods young learners use to learn meaning and function of verbs. It is certainly plausible that one child could use distributional information, while another uses semantic bootstrapping. More studies exploring inborn mechanisms may aid in the examination of individual differences.
There is relatively little research regarding the inborn mechanisms of language, compared to the empirical learning studies. Because language has only manifested itself recently in human evolution, some feel that language areas in the brain are not very refined and not very consistent within our species. Only humans learn grammar readily, as compared to Chimpanzees and other close relatives to humans. Chimpanzees require many more exposures to words and sentences, end up learning less overall, and have poorer short- and long-term memory for words (Savage-Rumbaugh et al., 1993; Seyfarth & Cheney, 1996).
Nativist Theory
Chomsky’s (1959) Nativist theory posited that the typical language-learning environment (what he calls the “stimulus”) is so poor that an innate ability must aid in the acquisition of language’s complexities. This has provided that basis for many biological assessments of the language-learning brain. Broca’s and Wernike’s have been identified, using fMRI scans, as brain areas involved in language comprehension. There is a critical period for language. After 12 years of age it is more difficult to learn a second language and the synapses created for a secondary language learned after this time form around the language areas, not within them. After 12 years of age, it is near impossible to learn grammar (verbs, etc.) of first language, as in the famous case of Genie, who could only learn nouns and a few verbs but was never able to speak a full grammatical sentence.
Although innate language mechanisms are important, the scope of this paper has limited itself to empirical learning and the two theories about how young language learners acquire verbs. Establishing whether distributional information or semantic bootstrapping is the initial means of verb categorization has practical implications for parents and teachers. Still, it seems that the information from all the aforementioned studies could be applied. Parents could alter the content of child-directed speech to cover all types of semantic information and distributional information. Teachers could emphasize what words hang together and elaborate on variations in word clusters and order. These researchers can all agree on one thing: infants (even as young as one year old) are able to detect grammatical and ungrammatical verbs. Perhaps how exactly they make these distinctions is less important than implementing these findings by simply talking with an infant.
References
Brown, R. (1973). A First Language: the Early Stages. Cambridge, Mass.: Harvard University Press.
Chomsky, N. (1957) Syntactic Structures. The Hague: Mouton.
Clark, H. H. (1992). Arenas of Language Use. Chicago, IL: The University of Chicago Press.
Fisher, C., Hall, D. G., Rakowitz, S., & Gleitman, L. R. (1994). When it is better to receive than to give – Syntactic and conceptual constraints on vocabulary growth. Lingua, 92, 333-375.
Gillette, J., Gleitman, H., Glietman, L. R., & Lederer, A. (1999). Human simulations of vocabulary learning. Cognition, 73, 135-176.
Gleitman, L. (1990) The structural sources of verb meanings. Language Acquisition, 1, 3- 55.
Gropen, J. Pinker, S., Hollander, M., & Goldberg, R. (1991). Affectedness and direct objects: The role of lexical semantics in the acquisition of verb argument structure. Cognition: Special Issue: Lexical and conceptual semantics, 41, 153- 195.
Hohle, B., Weissenborn, E., Kiefer, D., Schulz, A. &Schmitz, M. (2004). Functional elements in infants’ speech processing: The role of determiners in the syntactic categorization of lexical elements. Infancy, 5, 341-353.
Lee & Naigles (2005). The input to verb learning in Mandarin Chilnese: A role for syntactic bootstrapping. Developmental Psychology, 41, 529-540.
Li., C., & Thompson, S. (1981). Mandarin Chinese: A functional reference grammar. Berkeley: University of California Press.
Maratsos, M.P., & Chalkley, M.A. (1980). The Internal Language of Children’s Syntax: The Ontogenesis and Representation of Syntactic Categories. In K.E. Nelson (Ed.), Children’s Language, Vol. 2 (pp.127-214): Gardner, New York.
Marchman, V.A. (1997). Children's productivity in the English past tense: The role of frequency, phonology, and neighborhood structure. Cognitive Science: A Multidisciplinary Journal, 21, 283-304.
Mintz, T.H. (2003). Frequent frames as a cue for grammatical categories in child directed speech. Cognition, 90, 91-117.
Naigles, L., & Hoff-Ginsberg, E. (1995). Input to verb learning: Evidence for the plausibility of syntactic bootstrapping. Developmental Psychology, 31, 827-837.
Hirsh-Pasek, K. & Golinkoff, R.M. (Eds., 2006). Finding the verbs: Distributional cues to categories available to young learners. Action Meets Word: How Children Learn Verbs. Oxford: Oxford University Press.
Peccei, J. S. (2006). Child Language: A Resource Book for Students. New York, NY: Routledge.
Pinker, S. (1984). Visual cognition: An introduction. Cognition. Special Issue: Visual cognition, 18, 1-63.
Pinker, S. (1987). The bootstrapping problem in language acquisition. In B. MacWhinney (Ed), Mechanisms of language acquisition. Hillsdale, NJ: Lawrence Erlbaum Associates.
Redington, M., Chater, N., & Finch, S. (1998). Distributional information: A powerful cue for acquiring syntactic categories. Cognitive Science, 22, 425-469.
Rispolo, M. (1995). Missing arguments and the acquisition of predicate meanings. In M. Tomasello & W.E. Merriman (Eds.), Beyond names for things: Young children’s acquisition of verbs (pp.331-352). Hillsdale, NJ: Erlbaum.
Saffran, J.R. (2001). Words in a sea of sounds: the output of infant statistical learning. Cognition, 81, 149-169.
Savage-Rumbaugh, E.S., Murphy, J., Sevcik, R.A., Brakke, K.E., Williams, S.L. & Rumbaugh, D.M. (1993). Language comprehension in age and child. Monographs of the Society for Research in Child Development, 58, (3-4, Serial No.233).
Seyfarth, R., & Cheney, D. (1996). Inside the mind of a monkey. In M. Bekoff & D. Jamieson (Eds), Readings in animal cognition. Cambridge, MA: MIT Press.
Tao, H.Y. (1996). Units in Mandarin conversations: Prosody, discourse, and grammar (Studies in Discourse and Grammar, Vol. 5). Philadelphia: John Benjamins.
Whorf, B. L. (1956). Language, thought and reality. Cambridge, MA: MIT Press. (Original work published 1940
