Researchers adhering to constructivist accounts of language development have objected to the idea that very young children may have adult-like command of inflectional morphology (Pizzuto & Caselli, 1992, 1993, 1994; Rubino & Pine, 1998; Gathercole, Sebastián & Soto, 1999, 2000, 2002; Serratrice 2001). The argument put forward is that early morphological productivity is a mere artefact of the non-stringent criteria employed to analyse children’s spontaneous data (e.g., Pizzuto & Caselli, 1992). It is suggested that children acquiring morphologically rich languages have been credited with productive use of inflectional morphology simply on account of their error-less speech production. When the analytical tools of children’s spontaneous speech samples are adequately tightened, the alleged early morphological productivity of children acquiring languages with large arrays of inflections would disappear.
Pizzuto and Caselli (1992) claim that a rigorous investigation of productive use of grammatical morphology in a richly inflected language like Italian should be based on the same criteria proposed by Brown (1973) and Cadzen (1968) for the study of grammatical development in first language learners of English. Following Brown, they consider that mastery of a given inflectional morpheme is achieved if the morpheme in question appears in at least 90% of contexts in which it is required. Following Cadzen, they define point of acquisition as the first of three consecutive speech samples in which a given inflectional morpheme is used in at least 90% of obligatory contexts. In addition, they claim that children’s performance cannot be adequately evaluated in cases where only one or two obligatory contexts arise. Therefore, following De Villiers and De Villiers (1973), the researchers impose a further restriction: that with respect to each specific morpheme, a given speech sample can be considered for analysis only if it contains at least five obligatory contexts of use. They also note that one serious methodological problem with the assessment of the productivity of verb inflections in the early stages of speech production is that, regardless of the number of verb tokens produced, children tend to use certain inflectional suffixes exclusively with one specific verb type. Hence, they argue that in order to disentangle productive use of verb inflections from error-free production, some “measure of variety” (1992, p. 517) should also be taken into account.
In order to have some measure of variety of instantiation, we noted the number of different verbs with which each person inflection was used. As a general rule, we estimated that any given person inflection was beginning to be used productively by each child when it was used with at least two different verbs (e.g. when in the same sample the child produced the 2nd-person inflection not only with api [for apri ‘open-2SG] but also with edi for vedi ‘see-2SG’).
(Pizzuto & Caselli, 1992, p. 517) They conducted a longitudinal study of the acquisition of Italian verbs, articles and pronouns by three children, Claudia, Francesco and Marco, who were audio-taped between the ages of 1;3-2;9, 1;4-3;9 and 1;5-3;0 respectively. The three children were recorded in their homes, mostly in dyadic interaction with their mothers. Claudia and Francesco’s data were collected at fortnightly intervals, while Marco’s play sessions took place approximately once a month. Their results show that agreement errors in the use of articles and verbs were very rare, ranging between 3% and 4% for articles and 1% and 4% for verbs. They found that the three children displayed similar patterns with respect to the appearance of different verb forms: plural forms appeared later than singular forms; simple- tense forms appeared earlier than compound-tense forms; the different inflections encoding first, second and third person did not appear simultaneously but at time intervals lasting from one to several months. The authors also point out that an impressive number of verb inflections appeared from the earliest ages in the children’s overall production (a total of 25, 23 and 17 for Claudia, Francesca and Marco, respectively). For example, between ages 1;4 and 1;9, all three children used present-tense first-, second- and third-person-singular forms. Claudia and Francesco also used the present-tense, third-person-singular copula form. The authors argue that: “if one examined a fragment of the corpora, without imposing acquisition criteria, it would be tempting to conclude that the Italian system of inflectional morphology is mastered very early” (1992, p. 525). However, they point out how a deeper analysis reveals that none of the major inflectional paradigms investigated was fully mastered by age 3;0. In all three children’s samples, only some singular, present-tense inflections met their acquisition criteria. In Claudia and Francesco’s data, these included the first and third person of the present indicative, the second person of the present indicative/imperative and the third person copula. In Marco’s samples, only the third person
of the present indicative was acquired by the end of the study. Finally, mastery of these verb forms was not achieved instantaneously, but a time interval ranging from one month to twelve months elapsed between first appearance and acquisition. Similar patterns were found in relation to first appearance and acquisition of pronouns and articles. Since the three children only achieved mastery of some singular inflectional forms and because a time lapse of several months occurred between first appearance of a particular inflection and the point at which it met their acquisition criteria, the authors conclude that agreement is not an early acquisition in Italian.
Gathercole, Sebastián and Soto (2002) present data on the acquisition of person, tense and number by three monolingual Spanish children, Maria, Juan and Miguel, who were videotaped and audiotaped, from ages 1;3.6 to 2;6.16, 1;8.5 to 2;1.13 and 1;6.7 to 2;2.1 respectively. The children were recorded in naturalistic settings for half an hour at monthly intervals. The researchers followed Pizzuto and Caselli’s (1992) criteria with respect to productive use of inflections and contrastive use of verb types. They also established a further measure of productivity: a given tense, person or number was considered productive only if the child’s speech contained at least two different persons, two different tenses and two different numbers. Their reasoning behind this measure is that in Spanish, each (finite) verb form encodes person, number and tense, hence, if only one person-number-tense combination is productive, i.e., appears with at least two verb types, it cannot be ruled out that that particular form is some sort of “default form” (Gathercole et al., 2002, p. 688). They found that Juan only used the first person singular and the third person singular of the present tense. Hence, he had no number or tense contrast by the end of the sampling period. In Maria’s speech, the first type of contrast that became productive was a tense contrast, because she used the third person singular in the present indicative, the imperfect and the present perfect. Miguel acquired person contrast first, with the use of the first person singular and the first person plural in the present indicative. Then, he also acquired contrast of tense, with the use of the third person singular and the third person plural in both the present indicative and the present perfect. The authors claim that their results suggest that the three children exhibited individual differences in the development of person, number and tense.
Rubino and Pine (1998) investigated the acquisition of the subject-verb agreement system in one child acquiring Brazilian Portuguese. The child was recorded once a week, for a period of two months from 3;2.7 to 3;4.8 in his home, in spontaneous interaction with his mother or another member of his family. Their analysis shows a discrepancy in the production of correct singular and plural verb inflections. The researchers point out that the overall error rate is as low as 3%, but that when one takes into account the frequency with which different forms are produced and the target deviations concerning each one of such forms, the picture changes considerably. Firstly, the sheer number of utterances containing plural verb forms is very small, in comparison with the number of utterances containing singular verb forms. Out of a total of 1464 utterances 1414 utterances show singular inflections and 50 plural inflections. Secondly, the rate of agreement errors in utterances containing plural verb forms is 28%, whereas the error rate in utterances containing singular inflections is 2.2%. Furthermore, the researchers identify a considerable degree of variation among the error rate associated with each person-number combination. When the person/number variables are also considered, it appears that most errors concern the first person singular, the first person plural-a gente and the third person plural, whose error rates are 8.4%, 23.5 and 43.5% respectively. The second person plural never appears in the speech samples and no target deviations are produced with respect to the second person singular and the first person plural-nos. The authors also point out that agreement errors with the first person singular tend to occur with specific verb types: out of the 25 instances of deviations seven were produced with the verb querer (to want) and five with the verb ir (to go), which were often produced by the adult speaker in questions and imperative utterances addressed to the child.
Third person plural forms only emerged in the fifth session, when the child was 3;3.4. Target deviations consist of the use of a third person singular form with a third person plural subject. The authors point out that, from the point at which the child begins to produce third person plural inflections until the end of the study, the error rate for the person-number combination in question is 25%. Thus, even after third person plural forms appear in the child’s speech, the error rate associated with them only gradually approximates to the error rate associated with other person-number combinations over the same developmental period. They claim that this finding is not congruent with the view that
the child has innate knowledge of subject-verb agreement, but fails to use it because he has yet to learn the relevant plural verb inflections. Additionally, seven out of twelve utterances exhibiting correct subject-verb agreement for the third person plural result from the production of two specific subject-verb strings: esses são (these are) and eles vão (they are going). For this reason, they could be accounted for by the learning of two unanalysed subject-verb strings. With regard to the first person plural-nos, except for one instance, all the utterances contain the verb form vamos (to go – 1pl) and occur within the syntactic frame vamos X (let’s X). They argue that this type of finding has clear theoretical implications, because it clearly shows that the child has not yet performed the morphological analysis of segmenting verbs into stem + inflection. They claim that inflections are linked, in his mental lexicon, to specific verb forms.
[These errors] seem to be intrinsically related to the piecemeal nature of the child’s verb morphological development, reflecting a process whereby verb inflections are initially produced by the child as part of specific lexical items, and only later reanalysed and organised into a more general verb morphological system. Certainly, the finding that 12 out of 25 of this child’s 1 SG agreement errors (i.e., 48%) are related to two particular verb forms which occur with high frequency in the mother’s speech suggests that the initial lexical specificity of particular verb inflections does not disappear abruptly, but only gradually gives way to the flexibility which characterises the adult verb morphological system.
(Rubino & Pine, 1998, p. 46)
Following Pizzuto and Caselli’s productivity criteria, Serratrice (2001) investigated the acquisition of the subject-verb agreement system by an Italian-English BFLA child, Carlo, born in Scotland to an American father and an Italian mother. Naturalistic speech samples were collected for both languages every second week for a period of 15 months, from when the child was 1;10. In her analysis, she distinguished between copula, auxiliaries and lexical verbs. With respect to copula be, she found that only two person- number combinations were used, the third person singular and the third person plural, and that Carlo provided the required copula form in 92% of third person plural obligatory contexts, but only in 43% of singular contexts. A discrepancy that she argues can be
accounted for by the fact that third person plural tokens only appeared in Carlo’s speech six months later than singular forms, at a time when he had begun to acquire knowledge of the obligatory nature of the copula. In addition, the copula only appeared in lexically specific constructions, in which the subject was invariably represented by that, it or they and the predicate by an adjective. As for the auxiliary to be, the author points out that Carlo’s knowledge was limited to the progressive construction, which was mostly used in third person contexts with either do or go in the –ing form. In relation to lexical verbs, she only considered present indicative third person singular forms, because they are the only ones that bear an inflectional suffix. However, forms with the periphrastic does were also
included. Only two verb types, come and go, were correctly inflected and bare forms were
produced in 33 out of 43 contexts.
Serratrice concludes that, by the age of 3;1.25, the BFLA child in her study had no knowledge of the English subject-verb agreement system. In Italian, Carlo’s use of the copula was considerable by 2;1.23, but a certain degree of lexical specificity was evidenced in the use of the existential construction c’è/ci sono. In seven cases, third person singular forms were used in third person plural contexts. However, as the author highlights, the overall error rate was still very low (1.7%). Equally low were the error rates of present indicative inflections on lexical verbs, ranging from 10% for the third person plural to 0% for the first person plural. However, the author points out that, until the child was 2;2 years of age, there was not more than one verb form per verb type per session. The emergence of new grammatical contrasts in the child’s speech proceeded slowly throughout the recording period, and a more accurate inspection of the distribution of different inflections across verb types reveals that only 17 out of a total of 82 verb types, (some of which, she argues, are highly irregular), occurred with four or more different inflectional suffixes across the 20 Italian files. The remaining 65 verb types appeared with only one or two inflections. She concludes that the child’s almost errorless production of Italian inflected verb forms can be attributed to the fact that the number of different person-number combinations is limited and the use of verb forms is tied to lexically specific constructions, rather than to adult-like mastery in the use of inflections.
It is worth noting that, in studies like Gathercole et al.’s (2002) and Rubino and Pine’s (1998), no distinction is made between bound morphemes and function words (i.e.,
copula and auxiliaries) and acquisition of person-number combinations is investigated across different tenses’ paradigms. These circumstances can account for Gathercole et al.’s conclusion that the three Spanish children do not exhibit across-the-board acquisition for any person-number combination and the high error rates that Rubino and Pine report in their study. Spanish and Brazilian Portuguese are richly inflected languages and inflections corresponding to each person-number combination are numerous (over 40 according to Ambridge and Lieven, 2011). In Pizzuto and Caselli’s (1992) study, the acquisition of different paradigms is investigated separately and indeed the error rate that the authors report is much lower. Their conclusion that agreement is not an early achievement for Italian acquiring children is based on the observation that, even within the major paradigms, children do not meet their acquisition criteria for all the different inflections, and when acquisition does take place, this happens quite slowly. For example, the different inflections for first, second and third person do not appear simultaneously, but at time intervals lasting from one to several months. Serratrice (2001) points out that only 21% of the verb types that appear in her Italian data are used with four or more different inflections. As pointed out by Hyams (1992), the acquisition of inflections is a form of lexical learning, because inflectional affixes are lexical items and as such they must be learned individually. Hence, the learning of an inflectional paradigm is not “triggered in an all-or-none fashion” (Pizzuto & Caselli, 1992, p. 506), but takes place gradually. In Hyams’ words:
Everyone agrees that children must learn individual lexical items, whether they are free or bound, and that this learning depends on a number of factors, including semantic, syntactic and phonological ones. Thus, the learning of an inflectional paradigm, like learning within other grammatical categories, e.g. prepositions, articles, pronouns, etc. is likely to be gradual in the sense that each form within the category will be acquired individually. We would not expect children to acquire all the inflectional affixes at once any more than we would expect them to acquire all the prepositions or pronouns of a language at once. […] In fact, I would add that such a development would be close to miraculous. (Hyams, 1992, p. 697)
3.6 Finiteness studies
Finiteness deviations have been the focus of considerable generativist-nativist theoretical explanatory efforts since the early 1990s and, more recently, have also received attention from constructivist researchers. This type of target deviation is common in the speech of young children but, cross-linguistically, it appears to be considerably more common in the speech of children acquiring languages that lack an overt subject-verb agreement system and, therefore, do not allow for subject dropping (e.g., English: Wexler, 1994; German: Poeppel & Wexler, 1993; Swedish: Platzack, 1990, 1992; Dutch: Haegeman, 1995; Icelandic: Thordardottir & Ellis Weismer, 1998 and French: Pierce, 1992). At the same time, within each language, this type of deviation appears to be more frequent in subject- less utterances. Finiteness errors involve the use of non-finite forms in contexts where a finite form is required; thus, they also include participle and gerund errors. In the literature, finiteness errors are often referred to as Optional Infinitive errors (OIs) (term due to Wexler, 1994) or Root Infinitives errors (RIs) (term due to Rizzi, 1994). Freudenthal et al. (2009) use the term Root Participle to refer to finiteness deviations consisting of past- participle forms lacking the auxiliary.
3.6.1 Generativist-nativist studies
As mentioned above, several different theories have been proposed by generativist-nativist researchers to account for the occurrence of non-finite forms in finite contexts in child speech. Some authors (e.g., Guilfoyle & Noonan, 1994, Radford 1996) hold that finiteness deviations result from the lack of functional categories such as Agreement and Tense in children’s early grammars. Under this theory, functional categories develop through