Assessment of Receptive Verb Vocabulary Using Dynamic Scene Stimuli: Evidence from ASD and Late Talkers

• In Event: 2-093 - Beyond Vocabulary Size: New Considerations for Receptive Vocabulary Assessment in the Preschool Years

Integrative Statement

We discuss a new approach for assessing receptive verb vocabulary in preschoolers, which is a critical issue for research and clinical practice. Receptive vocabulary predicts later language abilities (Marchman & Fernald, 2008), and verb knowledge, particularly, better predicts grammatical outcomes than does noun knowledge (Hadley, Rispoli, & Hsu, 2016). However, standardized assessments (e.g., PPVT) are limited in at least two ways. First, verbs denote events, which unfold over time and may look different across time (e.g., throwing changes between the windup and release); however, current assessments depict events using static images. Second, children must indicate knowledge through overt behavioral responses (e.g., pointing), which may be difficult for children in special populations (Muller & Brady, 2016).

Our design addresses both limitations. First, we use dynamic scenes so that children may watch events unfold. Second, we use eye gaze as an indicator of knowledge, which is well-established for studying noun knowledge (e.g. Fernald, Perfors, & Marchman, 2006). Eye gaze can also be used to identify individual processing differences not revealed by binary correct/incorrect measurement. We measure verb knowledge (preference for the target scene at test) and processing (time it takes to show this preference).

We present results from two populations of clinical interest. First, children with ASD (N = 20, M(age) = 42 months) saw 18 trials featuring two dynamic scenes (from Konishi et al., 2016) in which a recorded audio prompt directed them to find one (Figure 1). As expected, participants preferred the target over the distractor scene at test (M(accuracy) = 60%). However, there was variability in the time children took to exhibit this preference, suggesting individual differences in processing speed. When we used the same time window of analysis for all children, preference for the target did not strongly correlate with concurrent MCDI-2 scores (r = 0.41, p = 0.08), but when we analyzed each child’s data within their own best time window, the correlation was significant (r = 0.67, p = 0.006),

Second, we compared typically-developing children (TDs) and late talkers (LTs), who are 2-year-olds with atypically small expressive vocabularies. In prior research with nouns, LTs showed longer latencies to look to familiar noun referents than TDs (Fernald & Marchman, 2012). We predict that in our verb task, LTs will know fewer verbs and have longer latencies than TDs. Results of ongoing data collection (N(LTs) = 9, N(TDs) = 30, M(age) = 28 months) thus far support our hypotheses: LTs know fewer verbs (M(LTs) = 21%, M(TDs) = 38%, t = 2.8, p = 0.01) and have longer latencies (M(LTs) = 2183 msec, M(TDs) = 1552 msec, t = 2.4, p = 0.04) than TDs.

Together, these two studies indicate the feasibility of 1) assessing receptive verb vocabulary through eye-gaze and 2) using dynamic scene stimuli for verb vocabulary. We argue that there is tremendous potential in using eye-tracking methods to understand verb knowledge and processing in clinical or at-risk populations, whose knowledge may not be fully evident in existing standardized tests.

Authors

• Sabrina Horvath, Boston University
  Presenting Author

• Sudha Arunachalam, New York University
  Non-Presenting Author