Beyond phonotactic frequency: Presentation frequency effects word productions in specific language impairment

Abstract

Phonotactic frequency effects on word production are thought to reflect accumulated experience with a language. Here we demonstrate that frequency effects can also be obtained through short-term manipulations of the input to children. We presented children with nonwords in an experiment that systematically manipulated English phonotactic frequency and the frequency of presentation within the experiment. Both of these manipulations affected the accuracy and time-to-response for nonword production both for typically developing and children with specific language impairment. Children with SLI were less accurate in their productions overall, but still exhibited an effect of the short-term frequency manipulation. Children with SLI differed significantly from their typical peers in terms of time-to-response only when both English and Experimental frequency were low. The results indicate that simple manipulations of the input can affect children's representation of word forms, and this can facilitate word production without the need for long term exposure or articulatory practice.

Learning outcomes: The reader will learn that sound frequency affects the production of new words. This includes not only the frequency with which sound sequences are represented in the speaker's native language, but the frequency with which they are heard within a single session.

Introduction

The fact that children are more accurate and more consistent at producing forms that are more frequent in their language is well established (Alt and Plante, 2006, Edwards et al., 2004, Mainela-Arnold and Evans, 2005, Munson, 2001, Munson et al., 2005a, Munson et al., 2005b, Munson et al., 2005c, Storkel, 2001a, Storkel, 2001b, Storkel, 2004, Zamuner, 2009). However, what remains a mystery is the mechanism by which more frequently occurring forms yield better production. One possibility is that more frequent forms in a language are produced by children more often and they benefit over time from the greater articulatory practice. Alternatively, simply hearing frequently occurring forms may lead to stronger internal representations of these forms. Consequently, the stronger internal representation assists with production of similar phonological forms in production tasks.

Articulatory complexity may contribute directly or indirectly to children's ability to produce novel forms (Gathercole and Adams, 1993, Gathercole and Adams, 1994, Graf Estes et al., 2007). Gathercole and colleagues have argued that articulatory skills may constrain performance on nonword repetition tasks (Gathercole and Adams, 1993), but that articulatory proficiency alone does not fully account for children's ability to add words to their lexicon (Gathercole, Service, Hitch, Adams, & Martin, 1999). However, it is logically possible that greater articulatory practice with frequent phonotactic sequences could lead children to a stronger sublexical representation of these sound sequences. This, in turn, could improve subvocal rehearsal during the operation of the phonological loop (Baddeley, 2003). A strong phonological loop, in turn, might then support vocabulary growth (Gathercole and Badeley, 1989, Gathercole et al., 1999, but see also Gathercole, Adams, & Hitch, 1994, concerning the issue of subvocal rehearsal by young children).

In contrast to this notion, others have argued that growth in vocabulary drives the frequency effect observed in nonword repetition tasks (Edwards et al., 2004, Munson et al., 2005a, Munson et al., 2005b, Munson et al., 2005c). This perspective posits that the development of the lexicon assists children in developing an abstract representation of the sublexical aspects of word forms. This abstract representation in turn influences children's performance on nonword tasks. The mental representation is influenced by the frequency with which specific phonological forms are common to the words in the child's lexicon, independent of articulatory ability. Gathercole, Willis, Emslie, and Baddeley (1991) also suggested that high frequency phonotactic forms in the child's lexicon may facilitate encoding into phonological working memory because encoding is supported by “phonological frames” that are abstracted from frequently occurring forms. The difference in these two perspectives relates to whether vocabulary development is thought to drive nonword repetition performance or the enhanced phonological working memory accounts for both vocabulary and nonword repetition results. Both positions are supported by associations between vocabulary test scores, particularly in the expressive domain, and frequency effects on nonword repetition tasks (Edwards et al., 2004, Munson et al., 2005a, Munson et al., 2005c). However, children with phonological disorders exhibit the phonotactic frequency effect, even when the effect of vocabulary size is partialed out (Munson, Edwards, et al., 2005). This leaves open the possibility that factors other than the size of a child's lexicon can influence production.

What is not in dispute in any of the above scenarios is the idea that children must represent the sounds of their language at a sufficient level of abstraction to serve speech production. Learners must convert the acoustic patterns in their input to some more abstract sequence of segments, which in turn can be used in their own productions. This view is partially supported by data from infants, who distinguish more frequent from less frequent phoneme sequences in their language before they produce their first word (Jusczyk, Luce, & Charles-Luce, 1994). Infants are also able to learn new phonotactic sequences in the laboratory, suggesting that they are highly sensitive to this kind of information (Chambers et al., 2003, Saffran and Thiessen, 2003). The infant perception findings argue against the idea of articulatory practice as the sole driving force behind frequency effects in production. However, it is possible that infant sensitivity to what is frequent in their input has little to do with the production effects seen in older children, for whom greater experience with language might support alternate learning strategies. Children's experience with their native language is thought to change their sensitivity to various types of language and language-like input that they receive (e.g, Cristia et al., 2010, Gerken and Bollt, 2008, Werker and Tees, 1984). Data from infant studies, therefore, is not adequate to address the issue of whether articulatory practice, lexical or sublexical internal representations, or any combination of these may account for frequency effects in word production of older children.

Recently, Richtsmeier, Gerken, Goffman, and Hogan (2009) addressed this theoretical issue in a study of typically developing four-year-old children. These investigators designed a study that narrowed the range of potential explanations for frequency effects. They used CVCCVC nonwords for which the medial biphone was either relatively frequent or infrequent in English. This contrast was crossed with frequency of presentation within the context of their experiment by presenting some words 10 times and other words just once. Children listened to presentations of nonwords for a brief period and then were given an opportunity to pronounce each nonword. Children heard multiple presentations of these nonwords before being asked to pronounce them.

Two findings from that study are particularly relevant for the present discussion. First, when a single talker presented each of the CVCCVC nonwords, the only effect on children's nonword production was that of frequency of occurrence in English, regardless of the frequency of presentation in the experiment. However, when each word was spoken by different talkers, then frequency of occurrence within the experiment had a greater effect than frequency of occurrence in English. This effect occurred not only for accuracy of the production of nonwords, but also for the time it takes for the child to respond (thought to reflect planning time). This pattern of results suggested that the talker-related variation in word production may have facilitated the development of an abstract internal representation in a way that high frequency presentation of the same nonword token from a single speaker had not. This interpretation is consistent with other work suggesting that multiple input tokens are required for learners to move from item-specific representations to general representations that allow for generalization beyond the input (Gómez, 2002).

The purpose of this study was to determine whether the frequency of speech input to children with specific language impairment (SLI) and their typically developing peers is sufficient to effect word production. We used the nonword stimuli of Richtsmeier et al. (2009) to in order to replicate the effects in typically developing children and to determine whether the results can be extended to include children whose language development is impaired. Although children with SLI are known to have difficulty with nonword repetition, they are also known to benefit in nonword production tasks when nonwords contain sound sequences that occur frequently in English (see Graf Estes et al., 2007 for a meta-analysis). Therefore, we expect an effect of English phonotactic frequency on their nonword productions. However, it is unknown whether children with SLI would be sensitive to a short-term manipulation of presentation frequency. An important benefit for these children for word acquisition might be derived if they could also be shown to be sensitive to frequency manipulations within a single experimental (or therapeutic) session.

The Richtsmeier et al. (2009) study demonstrated that the use of multiple talkers was critical for producing the type of perceptual learning that could drive Experimental frequency effects. We will likewise employ multiple talkers when words are presented frequently within the experiment.