3 General Discussion
3.1 Where Do the IDV Forms Come from?
A critical issue we need to address is where these forms come from. We will restrict our discussion to RSR (HL) words here. The question arises from the fact that most of the IDV forms are, at least on the surface, phonologically unre-lated to the adult form—/buHbu/ (car) is not phonologically reunre-lated to the adult form /kuruma/. Japanese infants, who learn these forms initially, would have to unlearn it and relearn the adult form later. The presence of word learning con-straints, such as “mutual exclusivity” (c.f., Markman 1994), could mean that the use of IDV may hinder Japanese children’s word learning. In other words, the RSR words are doubly dissociated from the target language—phonologically dissociated from the target word at the level of individual words, and distribu-tionally dissociated from adult Japanese vocabulary. Then, why do Japanese mothers use such forms? In the literature, several explanations have been put forward. We will consider three of them below.
3.1.1 Exaggerated Form of Adult Japanese
It is possible that the prosodic form of the IDV words are refl ecting general proper-ties of the Japanese language, even though the individual IDV items may be phono-logically unrelated to their target words. To address this issue, we analyzed adult Japanese in a number of formats; two kinds of spoken corpora, a Word Familiarity Database, and a newspaper corpus. In the analysis of the Word Familiarity Database, RSR forms were not found to occur frequently among highly familiar words. As we discussed above, our analysis of the Word Familiarity database underestimates the occurrence of special morae, as diphthongs were not counted as special morae.
However, even taking into account the underestimation of diphthongs, the results showed that RSR forms are not cited very frequently among “highly familiar” words.
In terms of frequency of occurrence in written and spoken corpora, we saw that the same pattern was found. RRR words occurred more frequently than RSR words in R-JMICC AD corpus and CSJ, and RSR forms occurred approximately as often as RRR or other forms in the Asahi newspaper corpus. In no case did we fi nd RSR types more dominant than RRR or other types. Taken together, we have no data that suggest that the frequent occurrence of RSR type words among 3-mora words is the general pattern found in adult Japanese.
3.1.2 Default Form for New Word Formation
Itô and Mester (1992) discussed that when new words are created by shortening, or combining loan words, the resulting form often takes the form of RSR (HL).
For example, when “paNhureQto” (pamphlet) is shortened, it becomes “paN.
hu”. Similarly, in a word game where the fi rst half and second half of a word are inverted, the resulting word often takes the RSR or RSRS form. For example,
“jazu” (jazz) will become “zuHja” (RSR) instead of simply juxtaposing the two syllables, i.e., “zuja”. Kubozono (1995; 2002; 2006) argues that the RSR form in IDV may be derived by the same principle. It is possible that for Japanese mothers, producing IDV is like creating new words to fi t the need of communi-cating with their children. Consequently, IDV words produced by mothers would share the same forms as the new word formation.
Data from the present paper are consistent with this interpretation. That is, the default word forms for creating new words and IDV forms share the same prosodic forms. It does not mean, however, that this hypothesis implies a causal relation between the two—just because the typical IDV forms share the same prosodic forms with templates for creating new words does not mean one is the cause of the other. It is very possible that the two types of word forms are con-strained by the same principle. But we are still left with explaining why these specifi c forms are preferred for IDV (and for new word formation), because they are not the general pattern in adult Japanese vocabulary.
3.1.3 Imitation of Early Production by Children
Perhaps the most frequent explanation that is offered for the specifi c form of IDV is that it is attributable to children’s early production. For example, Murata (1960; 1968) argued that the specifi c forms of the infant-directed vocabulary come from children’s early production. That is, these are the forms children produce at an early stage of word production, and the mothers are adopting them in their own production. Murase et al. (1992), and Ogura et al. (1993) found that the use of the specialized form vocabulary decreases as children’s production of adult-form vocabulary increases, suggesting that there is an interaction between the mother’s use of these forms and children’s vocabulary development.
4We thank Mitsuhiko Ota for allowing us to access his original coding of the three chil-dren’s utterances.
In order to evaluate this explanation, we analyzed the production data from Ota (2006) in terms of their syllable weights.4 The original data was taken from the Miyata corpus (Miyata 1992; 1995; 2000) of Japanese CHILDES (Miyata 2004).
We used the three children’s natural utterances at age 2;1. Figure 5 shows the proportion of children’s utterances in terms of mora length. Calculations are based on token count for each child. At this age, children’s utterances were already mostly 2, 3 or 4 mora long, which is comparable to the word length in adult spoken corpora shown in Figure 2. We then classifi ed 3 and 4-mora words according to their syllable weights. As shown in Figure 6, the pattern of their production is very similar to the IDV survey results, that is among 3-mora words, a very high propor-tion of them were of RSR (HL) form for all three children, and for 4-mora words also, RSRS (HH) forms were produced much more frequently than RRRR type words for all three children. Children produced higher proportion of “other” types of 4-mora words than the IDV survey. Thus, children’s early productions indeed show the predominant frequency of RSR and RSRS patterns.
0% 25% 50% 75% 100%
Fig. 5. Children’s production at 2;1 (Reanalysis of data from Ota, 2006.) Proportion of words in terms of mora-length (# of tokens analyzed).
0
RSR RRR RRS RSR RRR RRS RSR RRR RRS RSRS RRRR other RSRS RRRR other RSRS RRRR other
Aki Ryo Tai Aki Ryo Tai
Proportion for each word type
Fig. 6. Children’s productions of 3 and 4-mora words.
At the same time, the fact that the IDV forms are similar to the early form of children’s production does not necessarily support a claim that mothers (and other adults) are “adjusting” their speech imitating children’s production. If so, we predict that mothers who had more experience with children at an early production stage should be using more of these forms than non-mothers or mothers of younger infants who are not producing words yet. As shown above, however, our data showed that non-mothers with little experience with children produced a remark-ably similar list of IDV as mothers did. In addition, the mothers in our survey have not had much experience listening to their own child’s production. Except for four mothers who also had older children, mothers in our survey had 8 to 12 month old infants as their only child. Since infants at this age do not have much productive vocabulary, it is not likely that the mothers’ list in the survey was inspired by their own children’s production. Our data does not support a direct version of learning-hypothesis where individual mothers are assumed to learn to produce the RSR forms of IDV by listening to the actual utterances of the children. It does not rule out the possibility that more indirect forms of learning can occur through media or other cultural experiences about how young children’s utterances would sound like, and it could infl uence the IDV form indirectly.