Generativist accounts

Many generativist accounts of the development of verb inflection argue for early knowledge of inflection on the basis of the low frequency of inflectional errors in children's early speech (e.g. Hoekstra & Hyams, Reference Hoekstra and Hyams1998; Wexler, Reference Wexler1998). According to this view, young children's largely correct use of verb morphology reflects underlying knowledge of inflection, and those errors that do occur reflect a tendency to produce Optional Infinitive (OI) or Root Infinitive (RI) errors (i.e. non-finite forms in contexts in which a finite form is required in the adult language; see also Rizzi, Reference Rizzi1993/1994).

Some theories of the RI stage attempt to explain why RI errors occur in some languages and not in others. For example, according to Wexler (Reference Wexler1998), RI errors reflect a Unique Checking Constraint (UCC), which interacts with the type of language being learned to result in RI errors in obligatory subject languages such as English, Dutch, French and Swedish, but not in INFL-licensed null subject languages such as Spanish and Italian. However, others have attempted to extend the idea that children make errors involving the incorrect use of a ‘tenseless’ verb form to languages in which RI errors do not occur. For example, Salustri and Hyams (Reference Salustri, Hyams, Beachley, Brown and Conlin2003) argue that, in Italian, the imperative is an RI analogue, and Grinstead, De la Mora, Vega-Mendoza, and Flores (Reference Grinstead, De la Mora, Vega-Mendoza, Flores, Crawford, Otaki and Takahashi2009) argue that, in Spanish, the 3sg present tense form is an RI analogue. The concept of an RI analogue has also been extended to non-Indo-European languages. For example, Kim and Phillips (Reference Kim, Phillips, Greenhill, Hughes, Littlefield and Walsh1998) argue that, in early child Korean, children's overuse of a particular mood marker is analogous to the use of RIs, and Murasugi, Nakatani, and Fuji (Reference Murasugi, Nakatani and Fuji2009) argue that, in early child Japanese, the past tense form can be viewed as an RI analogue (see also Murasugi, Reference Murasugi and Nakayama2015). The key idea in all of these analyses is that there is an early stage in development, during which TENSE can be underspecified, or fail to merge with V, in the underlying representation of the sentence, and that, during this stage, children learning non-RI languages use a particular form of the verb as a ‘tenseless’ form in the same way that children learning RI languages use the infinitive.

The RI-analogue approach has the potential to provide a unified account of data across a wide range of different languages. However, it is subject to two potential problems. The first is that it assumes that early child language exhibits ‘inflectional imperialism’ (Slobin, Reference Slobin and Slobin1985, p. 1216): the tendency to make errors that involve ‘defaulting’ to a single dominant pattern in the language. However, as Dressler (Reference Dressler, Booij, Guevara, Ralli, Sgroi and Scalise2005) points out, there is wide typological variation in the extent to which languages exhibit this property. Thus, although a common feature of weakly inflecting fusional languages such as English, it is much less common in agglutinating languages such as Turkish and Hungarian. This observation raises doubts about how far it is possible to extend the RI-analogue approach – and, in particular, about the extent to which the approach can be extended to agglutinative languages.

The second is that it is often difficult to distinguish empirically between an RI-analogue account of children's early use of verb morphology and a constructivist account that emphasises the role of input frequency (e.g. Tomasello, Reference Tomasello2000). This is because the hypothetical RI analogue is often the most frequent form in the input. For example, Grinstead et al. (Reference Grinstead, De la Mora, Vega-Mendoza, Flores, Crawford, Otaki and Takahashi2009) argue that, in Spanish, the 3sg present tense form is an RI analogue. However, in Spanish, the 3sg present tense form is the highest-frequency form in the input (Aguado-Orea & Pine, Reference Aguado-Orea and Pine2015), and is also homophonous with the imperative for regular verbs. This makes it impossible to tell whether the overuse of the 3sg present tense form by Spanish children reflects its linguistic status as an RI analogue or a tendency to default to the most common form of the verb when the child is unable to generate the correct form in a particular morphological context.

In view of these considerations, the first aim of the present study is to evaluate the plausibility of the RI-analogue approach by focusing on the early use of verb inflection in Japanese – an agglutinative language which shows substantial by-verb variation in the frequency with which verbs occur in different forms. This will be done by testing Murasugi et al.'s (Reference Murasugi, Nakatani and Fuji2009) claim that the past tense form is an RI analogue for the Japanese-learning child. More specifically, we will investigate: (i) whether there is any evidence of a special preference for past tense forms in Japanese children's early verb vocabularies; and (ii) whether the errors produced by Japanese children tend to involve the overuse of past tense forms.

Constructivist accounts

Constructivist accounts of the development of verb inflection have tended to emphasize the limited productivity of children's early use of verb morphology and the fact that children's early use of particular inflections tends to mirror the frequency statistics of the input (e.g. Pizzuto & Caselli, Reference Pizzuto and Caselli1992; Tomasello, Reference Tomasello2000). According to this view, young children's correct use of verb morphology reflects knowledge that is initially embedded within particular lexically restricted patterns, and those errors that do occur reflect a tendency to use the form of the verb that occurs most frequently in the input language.

There is already considerable evidence that children's early verb learning and use is related to the frequency with which particular verbs occur in the input. For example, Smiley and Huttenlocher (Reference Smiley, Huttenlocher, Tomasello and Merriman1995), Naigles and Hoff-Ginsberg (Reference Naigles and Hoff-Ginsberg1998), and Theakston, Lieven, Pine, and Rowland (Reference Theakston, Lieven, Pine and Rowland2004) all report significant relationships between the frequency with which particular verbs occur in the input language and the order in which they emerge in the child's speech (see Ellis, Reference Ellis2002, and Ambridge, Rowland, Theakston & Kidd, Reference Ambridge, Rowland, Theakston and Kidd2015, for reviews of the data on frequency effects in language learning). Evidence has also recently emerged that it is possible to explain the relative frequency with which young children produce particular forms of the verb in terms of the relative frequency of particular forms in the input language. For example, Freudenthal, Pine, and Gobet (Reference Freudenthal, Pine and Gobet2010) have shown that it is possible to explain by-verb differences in the rate at which children make RI errors in terms of the relative frequency of infinitive versus finite forms of the verb in English, Dutch, French, German, and Spanish, and Räsänen, Ambridge, and Pine (Reference Räsänen, Ambridge and Pine2014) have shown that it is possible to explain by-verb differences in the rate at which children produce RI errors in English in terms of the relative frequency of bare stem versus 3sg forms in the input.

However, a particular challenge facing constructivist analyses is how to disentangle effects of input frequency on children's learning and effects of sampling and/or of the semantic–distributional properties of the language being learned. With respect to the first of these issues, Tomasello and Stahl (Reference Tomasello and Stahl2004) point out that analyses based on naturalistic speech samples tend to confound order of acquisition with frequency in the language, such that forms that occur with high frequency in the language are likely to be sampled earlier than forms that occur with lower frequency in the language, even if both forms were acquired by the child at the same time. This problem makes it difficult to distinguish between true effects of input frequency on learning and spurious effects that actually reflect differences in the likelihood that particular items will be sampled in naturalistic data.

A similar point can be made about measures of the relative frequency with which verbs occur in particular forms in the children's speech. Thus, it may be tempting to take correlations between the relative frequency with which particular verbs are used in particular forms by the child and by the caregiver as evidence of input-driven learning. However, as Yang (Reference Yang2013) points out, some words are more likely to be used in some contexts rather than others because of the semantic–distributional properties of the language. For example, in English, one is more likely to use the word bath with the indefinite article a than the definite article the because baths are things that one has or takes, and more likely to use the word bathroom with the definite article than the indefinite article, because bathrooms are locations which one tends to go to or visit. This kind of patterning makes it difficult to distinguish between real effects of input frequency on learning and spurious effects that simply reflect the fact that all speakers of the language are more likely to use some words in one way rather than another.

In view of these considerations, the second aim of the study is to test an input-driven account of Japanese children's use of past tense forms while controlling for the effects of sampling and/or the semantic–distributional properties of the language being learned. This will be done by looking for dyad-specific effects of the relative frequency of past tense and other inflectional forms in the input, while controlling for the effects of relative frequency in data averaged across the other caregivers in the sample. The rationale for this approach is that it has the potential to identify effects of the child's input over and above effects of input from the average mother – effects which therefore cannot be explained in terms of general differences in the likelihood of particular verbs occurring in the past tense as opposed to other inflectional forms in the input language. Note that Japanese is particularly well suited to this kind of analysis because both the past and non-past verb forms occur frequently in both Japanese child-directed speech and in Japanese children's early production data (cf. Clancy Reference Clancy and Slobin1985; Shirai, Reference Shirai1998). This makes it possible, at least in principle, to distinguish between by-verb variation that reflects the general properties of child-directed Japanese and by-verb variation that reflects the specific properties of the speech to which particular children are exposed. We now provide a brief sketch of Japanese syntax and morphology.

A brief sketch of Japanese syntax and morphology

Japanese is a typical head-final language, with features such as verb-final word order, attributive–noun word order, and the use of postpositional case markers. The basic word order is SOV, but the order of nominal arguments is pragmatically conditioned and relatively free in spoken discourse – a phenomenon often referred to as ‘scrambling’ (see Shibatani, Reference Shibatani1990). Nominal arguments are also frequently left unexpressed. This can make it difficult for the researcher to infer speakers’ communicative intentions from spoken discourse, especially in the case of young language-learning children. It also means that it is more difficult to identify obligatory contexts for particular verb inflections in Japanese than it is in English.

Japanese verbs have agglutinative morphology with extensive use of suffixation to mark inflectional distinctions. Finite verbs are always inflected for tense. However, unlike many Indo-European languages, Japanese does not have subject–verb agreement. A basic form consists of a stem and a tense marker as in tabe-ru ‘eat-NON-PAST' and tabe-ta ‘eat-PAST’. However, verb forms can also be more complex with morphological processes such as derivation, inflection, and concatenation. Note that past versus non-past is the only tense distinction in Japanese, with past tense forms being used to refer to past events and non-past forms being used to refer to both present and future events. Bare verb roots or stems are not well formed as words in Japanese, and are never produced as such. Nor does Japanese have infinitive forms like those in European languages such as German, French, and Spanish. Table 1 shows some basic inflectional distinctions and the suffixes that encode them for two major morphological classes: verbs where the stem ends in a consonant and verbs where the stem ends in a vowel. The former show stem alternation, which is dependent on whether or not they are followed by an auxiliary. Note that Table 1 does not provide a comprehensive description of Japanese verb inflection. It simply provides the reader with some information about the most common forms that Japanese speakers produce. Children are therefore typically exposed to a wider range of forms than those in Table 1 (see Shibatani, Reference Shibatani1990, for a more detailed description of Japanese verb morphology).

Table 1.

Some basic inflectional forms in Japanese

Previous studies of verb inflection in early child Japanese have reported that Japanese-speaking children typically use a number of different inflections by the age of two years, although some of these inflections may not be handled productively (e.g. Clancy, Reference Clancy and Slobin1985; Rispoli, Reference Rispoli1981). Clancy lists the inflectional forms that children tend to acquire early. These are: V-te imperative, V-ta past, V-teru non-past progressive, V-ru non-past, V-chatta completed past, V-nai negative non-past, and V-tai desiderative non-past (Clancy, Reference Clancy and Slobin1985, p. 426, with glosses modified by the current authors). Shirai (Reference Shirai1998) and Shirai and Miyata (Reference Shirai and Miyata2006) report frequent use of both the past and the non-past inflection during the early stages. However, they also report individual variation in acquisition profiles, which is consistent with the idea that it may be possible to find dyad-specific input effects on Japanese children's early use of verb morphology. Murasugi et al. (Reference Murasugi, Nakatani and Fuji2009), on the other hand, argue that there is an initial stage in early child Japanese in which children only use past tense verb forms in their speech.

The present study

The aim of the present study is to investigate children's early use of verb inflection in Japanese by comparing the idea that the past tense form has a special default-like status for the Japanese-learning child with the idea that Japanese children's use of past tense forms reflects the relative frequency with which particular verbs occur in past tense form in their input. The RI-analogue account will be tested by looking for evidence that Japanese-speaking children show a particular preference for past tense forms during the early stages, and a specific tendency to overuse past tense as opposed to other inflectional forms of the verb in their early speech. The input-driven account will be tested by looking for by-verb effects of the relative frequency of past tense and other inflectional forms in the child's input on the relative frequency with which they use past tense forms in their speech, while controlling for general differences in the relative frequency with which verbs occur in past tense form in Japanese child-directed speech.