“You Say What I Say”: The Value of Studying Elicited Verbal Imitation in Toddlers

Abstract

Copying what another says is a quintessential human ability. As toddlers learn to talk, they imitate sounds, words, and sentences. The frequency of verbal imitation typically accelerates during the second year of life. However, there are certain clinical populations, such as late talking toddlers (LTs), for whom imitation can be challenging. This thesis studies elicited verbal imitation in 2-year-olds, addressing three central themes. The first theme concerns the connection between elicited verbal imitation and new word learning overtime. How important is it for toddlers to imitate new words when they are first exposed to them? This was examined retrospectively in an investigation of the word learning abilities of 48 typically-developing (TD) 2-year-olds. It was found that a toddler’s ability to accurately imitate a nonword linked to a referent (i.e., a new word) during training significantly predicted their ability to name the referent at 1-minute and 5-minutes later, along with their extant vocabulary abilities. However, at a later time-point (1-7 days later), only extant vocabulary was a significant predictor of naming. These findings suggested that the ability to accurately imitate a new word may be important during the initial cusp of word learning and as such, important in the process of learning to talk. In clinical populations, verbal imitation has widespread relevance. It is employed in assessment and intervention across language and speech domains to identify, understand, and help children who struggle with talking. For toddlers, who are just beginning to learn to talk by building a lexicon, elicited verbal imitation has potential for studying whether the characteristics of new words may influence production. This idea formed the second theme of this thesis. To pursue the idea, a test was needed that allowed investigation of the influence of stimulus characteristics on imitation accuracy, while simultaneously maximising the likelihood that toddlers would attempt all the stimuli in the test. In the absence of such a test, the Monosyllable Imitation Test for Toddlers (MITT) was designed as part of this thesis. The MITT represents a type of online processing assessment. It comprises two toddler-friendly computer animations. The animations are short, interactive stories that contain pragmatic reasons to imitate eight carefully designed new words/stimuli. Stimuli are manipulated by both neighbourhood density and consonant complexity resulting in two stimuli in each of the following categories: dense neighbourhood/early-developing consonant (dense-early), sparse neighbourhood/early-developing consonant (sparse-early), dense neighbourhood/latedeveloping consonant (dense-late), and sparse neighbourhood/late-developing consonant (sparse-late). The independent and convergent influence/s of these stimulus characteristics was studied in age-matched TDs and LTs who attempted to imitate every stimuli item (i.e., TDs N = 26 / 26 and LTs N = 16 / 26). Both the TDs and LTs were sensitive to stimulus characteristics but exhibited differing patterns. Specifically, for TDs, both neighbourhood density and consonant complexity had significant independent influences—dense were more accurate than sparse, and early were marginally more accurate than late. The LTs only showed a significant independent influence for consonant complexity—early were more accurate than late. For convergent influences, TDs imitated all stimuli with equally high accuracy except for sparse-late, which were significantly less accurate. The LTs imitated dense-early stimuli with significantly greater accuracy than stimuli in the other three categories. The clinical utility of the MITT relative to compliance rates and diagnostic accuracy was also examined and compared to an existing toddler test of nonword imitation that uses 1- 4 syllable stimuli. Non-compliance was defined as refusal to start the test or ≥4 consecutive no responses. Thus, although 10 from 26 LTs did not provide a verbal response for every stimulus item on the MITT, none refused to start the test and only two failed to respond to ≥4 consecutive items. All TDs responded to every item. This resulted in a non-compliance rate for the MITT of just 4%. The existing nonword imitation test had a non-compliance rate of 13%. Both the MITT and the existing test were promising diagnostically. It is posited that the diagnostic value of the elicited verbal imitation tests studied may have reflected expressive phonology difficulties in the LT group. Although the MITT successfully permitted examination of the influence of stimulus characteristics on imitation accuracy in TDs and LTs, there was more to be learnt from the MITT imitation data. For instance, many LTs provided verbal responses in the MITT that bore little resemblance to the target stimuli, and therefore, examining their responses via accuracy alone seemed remiss. What could be learned from the diversity of responses made by toddlers, particularly LTs? Exploration of the breadth of toddlers’ responses formed the third theme of this thesis. Using both the MITT and a toddler-specific expressive phonology assessment, six types of responses were identified: no response, protoword-like response, different verbal response, correct phoneme, common phonological error and uncommon phonological error. Toddlers’ responses tended to correlate across the two sampling contexts, and responses used by toddlers also correlated with a variety of direct/parent-report assessment measures. Differences between the LTs and TDs in their proportion use of responses were found. The LTs responses were further investigated. Using cluster analysis, five subgroups of LTs were identified based on toddlers’ proportional use of the six response types. It was also evident that each subgroup had a unique profile of expressive vocabulary and phonology abilities. In the final chapter of this thesis, a conceptual framework for what might be involved when toddlers first hear, process, and try to produce words guides the discussion and interpretation of the findings from the three papers documented in this thesis. It is suggested that both TDs and LTs draw upon their existing phonological and/or articulatory-phonetic knowledge when processing and producing new words. Moreover, it is posited that late talking may be a manifestation of varied underlying difficulties with developing knowledge and processing abilities. For at least some LTs, their difficulties seem to be related to building phonological and/or articulatory-phonetic representations of words. Clinically, this implies that LTs may require goals and intervention approaches tailored to their specific area(s) of difficulty. It also has potential implications for longitudinal LT research. In conclusion, verbal imitation appears to be a strategy that toddlers may call on as they learn to talk. Elicited verbal imitation offers a window into studying new word production in toddlers. For LTs, the use of elicited verbal imitation together with other assessment procedures, allows for rich insights about the nature of their difficulties with learning to talk.