A cognitive model of response omissions in distraction paradigms
DOI: 10.3758/s13421-021-01265-z
archive: archived pipeline: cataloged verified
Summary
Cognitive-modeling study (Damaso, Castro, Todd, Strayer, Provost, Matzke, Heathcote; U. Newcastle + U. Utah + U. Amsterdam; Memory & Cognition 50, 2022, doi:10.3758/s13421-021-01265-z) developing the LBA-Omission (LBAO) model — an extension of Brown & Heathcote's (2008) Linear Ballistic Accumulator that decomposes response omissions into three causes: intrinsic omissions (accumulation rates that fail to reach threshold), design omissions (slow responses outside response window), and contaminant omissions (process unrelated to the task). The model is fit to two distraction-paradigm datasets: (1) Castro et al. (2019) DRT data, where 20 participants completed a 2 (stimulus intensity) × 2 (count-back-by-3 secondary-task load) within-subjects design; (2) a new choice-response paradigm with 34 younger (M=23 yr, U. Newcastle) and 23 older (M=67 yr) adults completing assessment plus session-2 task. The LBAO captured load-driven omission rate increases (4.3% no-load → 6% load) and stimulus-driven RT differences while ignoring stimulus-intensity effects on omissions; design omissions dominated over intrinsic omissions.
Key finding
Response omissions in distraction paradigms are not noise — they carry recoverable information about cognitive processes. The LBA-Omission model shows that most distraction-driven omissions are 'design omissions' (slow valid responses cut off by response windows) rather than intrinsic accumulation failures, reframing how omission rates should be reported and analyzed in DRT, choice-RT, and other distraction paradigms.
Methodology
Cognitive-modeling reanalysis + new experiment. Dataset 1: reanalysis of Castro et al. (2019) DRT data (N=20, 2×2 within-subjects). Dataset 2: new choice-response distraction experiment, 34 younger (18-38 yr, U. Newcastle) + 23 older (59-74 yr) adults. LBA-Omission model (extension of LBA with intrinsic, design, and contaminant omission mechanisms) fit via Bayesian methods (Heathcote et al. 2019) with non-informative priors; model selection by DIC plus posterior-predictive checks.
Sample size: DS1: N=20 (Castro et al. 2019 reanalysis). DS2: N=57 (34 younger 18-38yr M=23, 23 older 59-74yr M=67).
Quality score: 5 / 5