Should we wave, ask, run, steal? Categorizing aliens by teens with typical development or ASD

• In Event: Poster Session 13
  In Poster Session: PS 13 Topic - Cognitive Processes

Abstract

Word meanings and semantic categorization are clearly related during language development, with TD school-age children with larger vocabularies and stronger word learning skills showing better categorization abilities (Ellawadi et al., 2017; Tecoulesco et al., 2018). What is still unclear is the extent to which lexical acquisition specifically promotes category knowledge. Children with ASD provide a way to address this question because their lexical acquisition has been found to be dissociated from their category knowledge (Potrzeba et al., 2015); however, most tests of categorization with this population thus far have relied on word comprehension (e.g., ‘this is a rabbit; does it eat grass?’). The current study assesses categorization abilities teens with ASD or TD (Naigles & Fein, 2017), without asking them to label the entities they are categorizing (Tylen et al., 2016).
Participants included six teens with ASD and seven with typical development (M(ASD)=15.67 years, M(TD)=14.57). The TD group’s CELF vocabulary (raw) scores were significantly higher than those of the ASD group (M(ASD)=20.15, M(TD)=35.20). The categorization task was instantiated as a parent-teen video game played on a laptop. Across 40 trials, dyads were presented with images of aliens and asked to decide among four action responses (Wave, Ask, Steal, Run), depending on whether they determined that the alien was nice or not/possessed gems or not. Dyads received feedback after each trial.
Results: Accuracy (0 or 1) was modeled as a function of diagnostic group and (monotonic) trial in a Bayesian multilevel logistic model. As Figure 1 shows, teens with ASD showed only moderate evidence of learning (ER = 3.24), reaching an average accuracy of 46% (95% CIs 26-66%), against a chance baseline of 25%. In contrast, TD teens showed fairly strong evidence of learning (ER=73.07), reaching an average accuracy of 77% (CIs 50-96%). There is moderate evidence that the TD teens learned faster than those with ASD (ER=12.27).
Reaction time (RT) was modeled with a Bayesian shifted log normal multilevel model; the other parameters were the same as for accuracy. As Figure 2 shows, both groups of teens show strong evidence for a decrease in RT to find the correct answer (ERs > 1000), with teens with ASD going from 4.2s (95% CIs: 2.4-5.8) to 3s (95% CIs: 1.25-5.21) and TD teens going from 3.5s (95% CIs 2.8-4.6) to 2.2s (95% CIs: 1.8-2.8). There is only moderate evidence that the TD teens decreased more quickly than those with ASD (ER=3.28).
We observe a preliminary pattern of the TD teens performing both more accurately and more quickly than the (older) teens with ASD. No significant correlations emerged within groups, between vocabulary scores and categorization scores. We speculate that these teens continue to demonstrate a dissociation between lexicon and categorization, even when the latter is tested without labels. Data collection was suspended (due to covid-19) but has recommenced; moreover, future analyses will examine the underlying categorization processes through consideration of the relation between properties of the stimuli and responses, as well as the language used by the dyads as they played the game.

Authors

• Juandiego Carmona, University of Connecticut
  Presenting Author

• Riccardo Fusaroli, Aarhus University
  Non-Presenting Author

• Ethan Weed
  Non-Presenting Author

• Khanhvy Nguyen, University of Connecticut
  Non-Presenting Author

• Letitia R Naigles, University of Connecticut
  Non-Presenting Author