Abstract
The proposed paper focuses on the input properties in the bilingual and monolingual acquisition of two Russian expressions - “aa” and “mm”. These expressions can be used as pause fillers, i.e., non-silent hesitations similar to the English “uh” and “um,” but also have other functions. The data for our study come from an audio-aligned corpus of monolingual and bilingual child and child-directed Russian spoken in Russia, Germany, and the US which the authors are constructing. This is the first corpus-based study exploring the pragmatic functions of these expressions in Russian. In deciding whether pause fillers are a symptom or a signal of a speaker’s mental state, lexemic fillers such as English “like”, Russian “èto”/ “this” are traditionally thought to be signals, and their pause filler use to be just one of several word meanings. There is a question, however, of whether non-lexemic fillers are different. We note that, like lexemic fillers, which can carry their regular meaning, as well as the functions served by filling pauses, expressions “aa” and “mm” can carry other functions as well.
Based on the conducted analysis, we suggest that “aa” and “mm” are words signaling several different functions and having distinct distribution patterns, and present evidence of significant differences between monolingual and bilingual speakers in their use of these words as pause fillers, but not in their other functions. We then explore correlations between these discourse
markers and false starts in parental speech, and between their use in parents’ and children’s speech. We analyzed 85,754 words of (mostly child-directed) adult speech, from which all instances of “aa” and “mm”, the most frequent pause fillers in our data, were selected using an automated search. This was supplemented by data from the spontaneous informal spoken sub-corpus (https://ruscorpora.ru/new/search-spoken.html) in the Russian National Corpus (RNC). Each token, based on context and intonation, was annotated as expressing one of three broad discourse functions:
(i) pause fillers, including markers of hesitation, search, etc. (1);
(ii) a call for addressee engagement (2); and
(iii) other uses, such as signaling information receipt or commitment (3).
(Translations for discourse markers (notoriously ineffable) are approximate)
(1) Salfetku nužno položit’ v aa v musor
napkin need put.INF in aa in garbage
‘The napkin goes into uh into the garbage’
(2) Ma, ty golodnaja, aa? - Mm? - Ty golodnaja?
ma, you hungry, eh? mm? you hungry?
‘Ma, are you hungry, eh?’ - ‘Eh?’ - ‘Are you hungry?’
(3) Mm! Belka idёt -Aa, ponjatno!
mm squirrel goes.3SG Aa clear
‘Oh, a squirrel!’ -‘Oh, I got it!’
Mixed effects regression analysis using speakers as random effect showed that when comparing pause fillers to other uses of “aa” and “mm”, the country is the significant predictor (p<<0.01); i.e., bilingual parents use these expressions as pause fillers significantly more often than monolinguals. However, filler type (“aa” vs. “mm”) does not have significance (p=0.07): when
speakers are filling pauses, it does not matter what pause fillers they use. Analysis of examples from the RNC showed that in adult-directed speech, “aa” and “mm” are predominantly used as pause fillers, whereas in monolingual parents’ child-directed speech in our corpus there were almost no such uses. Based on this contrast, we hypothesize that bilingual parents may experience lexical access or speech planning difficulties which are similar to those that are experienced by
monolinguals when they address adults.
When we consider the two non-pause-filling functions of “aa” and “mm”, (ii) call on addressee and (iii) information signal, filler type comes up as a significant predictor (p<<0.01). While both are used for both functions, they appear with each of them at different rates: “aa” is used more often as information signal (58%), whereas “mm” more often functions as the call on addressee (again 58%). We take this as evidence that “aa” and “mm” behave like words which have distinct distributions and discourse effects in their non-pause-filling uses. Unlike with the pause-filling function, country is not a significant predictor (p=0.05) of such uses; i.e., bilingual and monolingual parents use these words in fairly similar ways.
We are following up on the analysis of “aa” and “mm” in parental speech by correlating pausefilling with the amount of false starts in the parents’ speech, since we take false-starts to be another indicator of cognitive load. At the same time, we are correlating parental use with the distribution of “aa” and “mm” in the children’s speech to study the influence of input properties on bilingual and monolingual acquisition