Poster #4 - Exploring Relations Between Action Production and Novel Measures of Action Observation

• In Event: 3-159 - Poster Session 12
  In Poster Session: PS 12 Section - Attention, Learning, Memory

Integrative Statement

Research on action understanding suggests there is a bidirectional relationship between the neural or motor representations underlying goal-directed action production and the ability to represent others’ goals during action observation (Longo & Bertenthal, 2006). The guidance of visual attention to others’ actions is also inextricably linked to these underlying motor representations (Flanagan & Johannson, 2003), making the study of eye movements particularly useful in quantifying infants’ action understanding abilities and developmental change thereof (Hunnius & Bekkering, 2014). Moreover, given that action production and observation are supported by shared representations, it is plausible that individual differences in infants’ ability to imitate novel goal-directed actions are associated with differences in their dynamic visual engagement of the same actions.
The goal of the present study is to examine the extent to which infants’ imitation abilities can predict individual differences in distinct measures of visual attention during action observation—action outcome prediction and sequential fixation patterns (or “scanpaths”, Noton & Stark, 1971). We tested 20- to 36-month-old infants using two behavioral assessments: an imitation task and subsequent action observation test. In the imitation task, infants interacted with two novel containers that could only be opened by manipulating three distinct components in a specific order (Figure 1). After a baseline assessment—used to ensure they could not solve the task without a demonstration—the experimenter modeled the actions required to open each apparatus. Participants were then given 60 seconds to imitate the actions as demonstrated. Two trained coders used a standardized scoring system to independently rate the similarity between infants’ behavioral strategies and the modeled action sequences. All infants then completed an eye-tracking protocol where they watched video recordings of the same demonstrations shown previously.
We calculated an anticipatory dwell time measure to quantify infants’ action predictions during the eye tracking protocol. Multiple regression analyses show significant positive correlations between infants’ imitation scores and outcomes on action prediction measures, even when accounting for age (Figure 2). Further analyses confirm anticipatory dwell times are not better explained by infants’ overall proclivity to fixate for longer durations (Figure 3). For the scanpath analysis, we will utilize a novel technique called Successor Representation Scanpath Analysis (SRSA; Hayes, Petrov, & Sederberg, 2011). Previous scanpath analysis methods only examine pairs of fixations, which can be limiting when applied to situations where multiple actions occur continuously over a temporally extended period of time. By extending this event horizon, SRSA can capture individual regularities in how viewers select several subsequent visual targets after fixating specific action outcomes. This information may better discern the ways in which variation in imitation performance relate to individual differences in infants’ visual attention strategies when observing goal-directed actions. In addition, because action prediction results ignore the many eye movements that occur between the offsets and onsets of each step in the action sequence, scanpath analyses confer specific advantages by utilizing this unused data source.

Authors

• Eugene Kim, Indiana University--Bloomington
  Presenting Author

• Bennett I. Bertenthal, Indiana University
  Non-Presenting Author