Poster #40 - Toddlers’ Uncertainty in a Perceptual Decision Task: Insight From Drift Diffusion Modeling

• In Event: 3-053 - Poster Session 10
  In Poster Session: PS 10 Section - Cognitive Processes

Integrative Statement

In order to make a well informed decision, one must successfully introspect on their knowledge of the subject. A key component to this ability is uncertainty monitoring, which is the ability to reflect on the level of certainty about a response and has been operationally defined in experiments as reporting higher confidence judgements for correct compared to incorrect responses. It has been shown that children as young as 3-years-old can report lower confidence for incorrect perceptual discrimination decisions (Coughlin et al., 2015). Even before children can explicitly report on their decision uncertainty, they may show spontaneous behaviors in response to uncertain situations that could be precursors for later uncertainty monitoring abilities. Response times may capture early signals of uncertainty because they might signal hesitation (and uncertainty) during decisions, but also because they can be used to fit drift diffusion models allowing to differentiate decision processes (i.e., evidence accumulation and bias parameters) from other processes (i.e., non-decision processes).

Across two experiments, we examined 2-year-olds’ response times in a perceptual discrimination task administered on a touchscreen computer. Toddlers were asked to point to a target image in a pair of black and white, partially occluded items that were either perceptually and semantically similar or dissimilar; similar trials were expected to generate higher uncertainty than dissimilar trials.

In study 1 (N=80), toddlers showed higher accuracy for the dissimilar (M=.86, SD=.14) compared to the similar trials (M=.71, SD=.17), t(69) = 8.31, p <.001. Because of the high accuracy in the task, we collapsed across similar and dissimilar trials for inaccurate trials. Toddlers showed slower response times for inaccurate responses (M=4.36, SD=2.27) than accurate-dissimilar (M=3.73, SD=1.66; p =.002) and similar (M=3.62, SD=1.38; p=.01), F(2,134) = 6.54, p = .002; the latter did not differ. Results of drift parameters revealed that with dissimilar trials, compared to similar, toddlers showed faster accumulation of evidence rates (t(66) = 7.37, p <.001), more caution (t(66) = 3.85, p <.001), and lower level of non-decision processes (t(69) = -2.12, p =.04). Study 2 (N=80), was a pre-registered replication of study 1. Toddlers again showed higher accuracy for the dissimilar trials (M=.82, SD=.15) compared to the similar trials (M=.71, SD=.15), t(66) = 5.43, p <.001. Toddlers also showed slower response times for inaccurate responses (M=5.33, SD=4.07) than accurate-dissimilar (M=4.25, SD=1.80; p =.02) and similar (M=4.56, SD=2.31; p =.03), F(2,126) = 4.90, p = .01; with the latter not differing. Drift parameter results revealed that toddlers showed faster accumulation of evidence rates for dissimilar compared to similar trials (t(63) = 6.44, p <.001), but no significant difference between the other parameters.

These results suggest that toddlers appropriately respond more slowly when they are incorrect suggesting judicious hesitation. Critically, the examination of drift diffusion parameters associated with our manipulation suggest that decision-related processes, including evidence accumulation and caution influence response times. This indicates that modeling response times can provide new insight on implicit signals of uncertainty in very young children who are not yet able to explicitly report on their uncertainty.

Authors

• Sarah Leckey, University of California, Davis
  Presenting Author

• Diana Selmeczy, University of California, Davis
  Non-Presenting Author

• Elliott Gray Johnson, University of California, Davis
  Non-Presenting Author

• Alireza Kazemi, University of California, Davis
  Non-Presenting Author

• Emily Hembacher, Stanford University
  Non-Presenting Author

• Simona Ghetti, University of California, Davis
  Non-Presenting Author