One of the characteristics of the process of language acquisition is the phenomenon of overregularization. In one of the periods in language acquisition, children tend to extend regular grammatical patterns to irregular words, which results in overregularizations like feeled, goed, mans and mouses. These errors, during this period of overregularization, appear systematically. For any systematic error, three descriptive questions arise: Why do children make this error, what is the time-course of this error, and what makes them stop? The aim of this paper is to give a brief overview of current linguistic research in this field, with the special accentuation on the answer to the third question. Having in mind the extent of the subject, the paper is focusing on the past-tense and plural morphemes, as these have been at the centre of overviewed debates. This paper is organized as follows. Section 2 presents the notion of overregularization and its phases. Section 3 discusses one of the most debated problems in the study of language acquisition œ the ”argument from poverty of stimulus‘. Section 4 discusses the possible solutions to the problem, dividing them into two basic groups œ basic and alternative approach. Section 5 offers concluding remarks.

Most verbs in English form the past tense by adding the suffix œed, as in walk-walked, pass­passed, dance-danced, turn-turned, and so on. These are called —regular“ verbs, and contrast with the approximately 180 —irregular“ verbs where the past tense relates to the stem in some other way. Some irregular past-tense forms differ entirely from their stems (go-went), others are identical (beat-beat), and others involve different sorts of vowel or consonant changes while preserving most of the phonological structure of the stem (break-broke, blow-blew, feel-felt) (Pinker, 1989). Adults know the past-tense forms of these irregular verbs (e.g., went is the past tense of go), and also know that the regular -ed suffix cannot apply to these verbs (e.g., goed is not an acceptable past tense of go). It is similar with the production of noun plurals in English. Most nouns form the noun plurals by adding the suffix œs, as in bird-birds, rat-rats, hand-hands, duck-ducks, and so on. Some nouns form the plural irregularly. The most common types of irregular plurals are when the noun changes f to v and then adds œ(e)s (knife-knives, wolf-wolves), when the singular and plural are the same (sheep-sheep), and when they change the vowel (man-men), the word (person-people), or add a different ending (child-children).

The regular rule is productive. Adults can use it to create past-tense forms that they have never heard before. For example, after hearing absolutely new present-tense verb such as kermit, an adult can use it in the past-tense form, saying I kermitted the file to my computer yesterday (Bloom, 1996). Berko (1958) found evidence for the same sort of productivity with young children. If they were shown a picture of a man doing an activity and they were told: —There is a man who likes to rick. Yesterday he did the same thing. Yesterday he…,“ children would apply regular rule and say ricked. This finding disputed the claim that children are limited to imitating the forms that they hear, while acquiring language. In this task and others similar to it (Marcus, 1992), children seem to be capable of using a word to which they have never been exposed before. Ervin (1964) argued in favour of the same productivity, based on his research of children‘s spontaneous speech. Children will spontaneously say sentences like I goed to the store or He breaked the glass. This phenomenon of overapplying the regular rule is called overregularization or overgeneralization.

The conventional characterization of the time-course of these errors (Ervin, 1964) is a U­shaped curve, where the —U“ is designed as the percentage of correct performance over time. The left part of the —U“ is the earliest stage of morphological production, in which the child is performing perfectly, using both regulars (walked) and irregulars (broke) correctly. This is followed by a long period of overregularization (the round middle of the —U“), where children overregularize, producing forms such as goed and breaked. Finally, on the right part of the —U“, where children stop overregularizing and return to correct competence.

Therefore, it can be concluded that there can be conditionally singled out three phases of overregularization. In initial stage of the process, children first acquire all of the past-tense or plural forms through rote memory. They learn by heart that the past tense of walk is walked and past tense of break is broke, and that the plural of man is men and plural of bird is birds. At this early stage there is no psychological difference between the two classes of past-tense verb, nor between the two classes of noun plurals. They do not distinguish “regular“ and —irregular“ verbs or noun plurals. The outbreak of the second stage of the process of overregularization is when children determine that there is a productive rule at work (the rule: add the suffix œed to form the past tense or add the suffix œs to form the plural). Children start overapplying the rules, producing forms like breaked, goed, mans, and mouses. The third stage is the recovery from overregularizing. In this stage, child becomes aware of —irregularity“ of particular verbs and nouns, and stops overgeneralizing. What causes children to stop overregularizing and return to correct competence? The contemporary scholars argued on possible answers to this question in last 35 years.

The problem of recovery from overregularization is, in contemporary linguistics, also known as the ”logical problem of language acquisition‘ (Baker, 1979), ”Plato‘s Problem‘, ”Chomsky‘s Problem‘ or ”Baker‘s Paradox‘, but it is usually referred to as the ”argument from poverty of stimulus‘. In the form of the ”argument from poverty of stimulus‘, it appeared in 1980, and was formulated by Noam Chomsky. According to Chomsky, the child‘s acquisition of grammar is indeterminate because of two reasons. The first source of indeterminacy is seen in —degenerate nature of the input“ (Chomsky, 1980). The child is not provided of clear indication of the possible sentences of the language, because the sentences that are heard by the child are full of errors and inconsistencies. The second source originates in the problem of unavailability of negative evidence. It is impossible for the child to decide which form is grammatical (acceptable) and which is not, because the adults do not provide consistent and distinct evidence regarding the ungrammaticality of unacceptable forms. And even, when they provide such evidence, it is often ignored by the child.

(1)       Child: I goed to shop. Mother: You went to the shop? When did you go? You went with whom? Child: Yesterday, I goed with granny. Mother: So, you went with your granny to the shop yesterday.

And even when this feedback is available, children may ignore it, as illustrated in example (2):

(2)       Child: I goed to shop. Mother: No, say ”I went to the shop‘. Child: I goed to shop. Mother: Now listen carefully, say ”I went to the shop‘. Child: Oh! I goed to the shop.

The possible solutions to the problems exposed above, can be divided into two main groups. The first group is based on certain premises œ children possess a rule, they go through a stage where they believe that certain irregulars have œed past tenses and œs plurals, and they need to recover from these errors. The second group challenge these assumptions and suggest extremely different explanations for children‘s errors.

Since the earliest discussion of overgeneralization, the competence involved has been described as an implicit rule œ add œed to form the past tense (or, add œs to form the noun plural) œ and the children‘s errors are assumed to result from overapplication of this rule. In the group of solutions that have this premise in their core are limiting the class of grammars, conservatism, and mechanism of competition and blocking.

This solution of the logical problem relies on the assessment of indirect negative evidence. In this case, child behaves like a computer. At first, it needs to track the frequency of the past tense as marked by the regular œed. Then it needs to assess regular œed as a percentage of all verbs. Next a child needs to track the frequency of the irregular verb (e.g., feel) in all of its uses and the frequency of its overregularized form (e.g., feeled). Then a child should also compute the frequency of a regular verb (e.g., walk) and the frequency of its past-tense form (e.g., walked). At this point, with all of these calculations, a child can compare the frequency of feeled and walked, and conclude that there are definitely fewer cases of feeled that they should be expected, if the regular rule is applied. A child could then deduce that feeled is ungrammatical. According to MacWhenney (2004), a child can do it without receiving corrective feedback.

Another solution to the logical problem is the emphasis of the conservative nature of child‘s language acquisition. The child will reject to use a specific form as ungrammatical, unless it hears an adult use the same form. The basic idea of this approach is to avoid formulating overly generated grammars in a first place. If the child never overgeneralizes, the problem of recovery doesn‘t exist and there is no need for negative evidence or corrective feedback. But, according to Berko‘s (1958) study, this is demonstrably false; because child is willing to use past-tense and plural forms that it has never heard before.

Next solution is based on the mechanism of competition. The Competition Model finds the two types of pressures that produce overregularizations. The first pressure is the —underlying analogical force“ (MacWhenney, 2004, 900) that produces the overregularization. The second pressure is the growth in the rote auditory representation of a correct form. These two forces fight for the control of the production. The competition can operate directly during conversation, as in example (3) (MacWhenney, 2004, 902).

1984). Such a condition prohibits the application of a rule that would produce an item that already exists in the lexicon. In example (3), past-tense felt is blocked by feeled. Assuming that children possess such a principle, the solution may go as follows (Pinker, 1984): Children begin learning all past-tense or plural forms by rote, and then induce the œed or œs rule. Overapplication of these rules leads to overgeneralization. But as irregular forms such as felt and men are reintroduced into the lexicon, they block the application of he past-tense or plural rules and the child stops overgeneralizing.

This is one particularly elegant solution for the argument from the poverty of stimulus. Unfortunatelly, it doesn‘t mean that it is without any weak points. Marcus (1992) pointed out at the first place œ given that the child initially has the word felt encoded as the past tense of feel, we would expect blocking or contrast (competition) to keep the child from using feeled œ but it does not. Therefore, why would a child overgeneralize at all? Secondly, the blocking theory doesn‘t allow a period of simultaneous existence of feeled and felt in child‘s lexicon. According to Cazden‘s study from 1968 (Bloom, 1996), this is false œ irregular past tenses and plurals coexist with overregularizations for years in children‘s lexicons.

The —blocking“ theory is really wounded with the first argument. But, the second argument can be disputed. The child might not know that felt is the past-form of feel. In this way, child‘s productive use of feeled as the past tense of feel would not violate the principle of blocking (or competition). In one moment, a child realizes that the adult form felt is synonymous with their rule-generated form feeled, and then the child will cast out feeled from it‘s lexicon œ and this process might take years (Clark, 1987).

This attempt to solve the logical problem is based on the child‘s ability to monitor and detect its own errors. By itself, competition does not fully satisfy the dynamics of language processing in real social interactions. MacWhenney (2004) differentiates two types of monitoring œ ”expressive monitoring‘ in which the child listens to its own output, compares the correct weak rote form with the incorrect overgeneralization, and tries to block the output of the incorrect form (e.g., I gived, uh, gave my friend a peach.), and ”receptive monitoring‘ œ when the child follows closely input structures, it will be able to notice many disparities between his own productive system and the forms it hears.

The alternative to the sort of theories given above is so called ”connectionists model‘ or parallel distributed processing. Rumelhart and McClelland (Bloom, 1996) created a special computer model that succeeded in capturing certain core properties of children‘s behaviour. According to their connectionist analysis, the capacity to determine the past-tense or plural form, a stem is acquired through an associative network which can extract statistical regularities from the environment. They concluded:

We have shown that a reasonable account of the acquisition of past tense can be provided without recourse to the notion of a rule as anything more then a description of the language (Rumelhart&McClelland, 1986:267)

This theory is the subject for a vastly critic that comes mostly from Pinker. One of the criticisms is based on Pinker's study of children‘s spontaneous speech. On the contrary to the connectionists‘ results, in Pinker‘s analysis the ratio of irregular to regular verbs does not change at the period when children start to overgeneralize. In that sense, the connectionist model would not adequately predict the outcome of the actual conditions of language development, having in mind that they fed their computer model with higher proportion of irregular verbs.

In the section above, a brief review of the solutions to the logical problem was given. It would be a mistake to think that any one of these solutions would operate by itself. For example, it would be unreasonable to suppose that child might rely only on monitoring or only on indirect negative evidence to recover from overgeneralization. Nor should it be imagined that child would rely only on conservatism to avoid errors and create the perfect grammar. Instead of that, the solution is found in conjunctive work of all proposed solutions in order to create a whole picture of the process of the language acquisition. While examining the interaction of the solutions, the thing that all of us must agree upon is the central role of the input. The most recent studies (Marcus, 1992) suggested the importance of child‘s exposure to good positive data. It was concluded that informants (adults) are inconsistent in their providing of negative evidence to the learner (child). But it is positive data that is crucial for learning. There are a lot of methods of exposing a child to positive evidence. For example, recasting, elicited repetition, choral recitation of stories, interaction with family, or games, and all these methods that emphasize shared attention and understanding can guide a child toward the control of literate expression (or, advances monitoring). At the end, although the differences in the way of providing the reliable positive evidence must be taken into consideration, there is no possibility that the child acquires language without positive input.

Avram, L. An introduction to Language Acquisition from a Generative Perspective. (Last updated: October 2005) Baker, C.L. (1979). Syntactic theory and the projection problem. Linguistic Inquiry, 10, 533-581.

Baker, C.L. & McCarthy, J.J. (Eds.) (1981). The logical problem of language acquisition. Cambridge, Mass.: MIT Press

Berko, J. (1958). The child‘s learning of English morphology. Word, 14, 150-177. In J. Berko (1993). The Development of Language (3. ed.). New York: Macmillan Publishing Company

Berko, J. (1993). The Development of Language (3. ed.). New York: Macmillan Publishing Company

Bittner, D., Dressler, W. & Kilani-Schoch, M. (2003). Development of Verb Inflection in First Language Acquisition: A Cross-Linguistic Perspective. Berlin: Mouton de Gruyter

Chomsky, N. (1980). Rules and representations. Cambridge, Mass.: MIT Press

Chomsky, N. (2002). On Nature and Language. Cambridge, Cambridge University Press

Clark, R. (1987). Language teaching techniques. Brattleboro, VT: Pro Lingua Associates

Ervin, S. (1964). Imitation and structural change in children‘s language. In P. Bloom (ed.). (1996). Language Acquisition: Core Readings. Cambridge, Mass.: MIT Press

Gold, E. (1067). Language identification in the limit. Information and Control 10, 447-474. In S. Pinker (1989). Learnability and Cognition: The Acquisition of Argument Structure. Cambridge, Mass.: MIT Press

Goodluck, H. (1991). Language Acquisition: A Linguistic Introduction.Oxford, UK: Blackwell Guasti, M.T. (2002). Language Acquisition: The Growth of Grammar. Cambridge, Mass.: MIT Press

Ingram, D. (1989). First Language Acquisition: Method, Description and Explanation. Cambridge, Mass.: MIT Press

MacWhinney, B. (2004). A multiple process solution to the logical problem of language acquisition. Journal of Child Language 31, 883-914.

Marcus, G., Pinker, S., Ullman, M., Hollander, M., Rosen, T.J. & Xu, F. (1992). Overregularization in Language Acquisition. Monographs of the Society for Research in Child Development 57, serial no. 228.

Marcus, G. (1996). Why do children say —breaked“? Current Directions in Psychological Science, 5, 81-85.

Pinker, S. (1989). Learnability and Cognition: The Acquisition of Argument Structure. Cambridge, Mass.: MIT Press

Pinker, S. (1991). Rules of Language. Science, 253, 530-535

Pinker, S. (1998). Words and Rules. Lingua, 106, 219-242.