Evidence for different types of errors being associated with different types of post-error changes

Abstract

Errors in simple choice tasks result in systematic changes in the response time and accuracy of subsequent trials. We propose that there are at least two different causes of choice errorsresponse speed and evidence quality, which result in different types of post-error changes. We explore these differences in types of errors and post-error changes in two recognition memory experiments with speed versus accuracy emphasis conditions that differentially produce response-speed and evidence-quality errors. Under conditions that give rise to more response-speed errors, we find evidence of traditional post-error slowing. Under conditions that give rise to evidencequality errors, we find evidence of post-error speeding. We propose a broadening of theories of cognitive control to encompass maladaptive as well as adaptive strategies, and discuss implications for the use of post-error changes to measure cognitive control.

A direct and conceptual replication of post-loss speeding when gambling

“…Our findings are also in line with findings of post-error speeding in difficult tasks (Damaso, Williams & Heathcote, 2020;Williams, Heathcote, Nesbitt & Eidels, 2016).…”

“…These findings are aligned with those of Dyson and colleagues (Dyson, Sundvall, Forder & Douglas, 2018) who found that participants only slowed down after a loss when getting more cautious led to better performance, for example when they could successfully apply a strategy to guide their choice. However, when slowing down does not lead to performance improvement, participants seem not to slow down but even speed up (for similar findings see Damaso et al, 2020).…”

Subthreshold error corrections predict adaptive post‐error compensations

“…We also note that for the present study, participants were instructed to weigh the speed and accuracy of their response equally and were not provided any instruction regarding whether they should (or should not) correct their errors. A recent study by Damaso et al., (2020) illustrated that when errors are committed during trials in which accuracy is prioritized over speed, performance on that trial would benefit from spending more time evaluating the incoming information and would consequently lead to post‐error slowing. However, during trials in which accuracy and response time are equally prioritized, performance on that trial would not benefit from additional information and would consequently lead to post‐error speeding.…”

Inclusion Bayes Factors for Mixed Hierarchical Diffusion Decision Models

“…Building on the work of Laming (1968) and Ratcliff (1978), Ratcliff and Rouder (1998) pointed out that this pattern could be accommodated by Gaussian across-trial variability in the rate of accumulation, with mean v and standard deviation s v , and uniformly distributed across-trial variability in the starting point, with mean z and range s z . Rate variability causes slow "stimulus-quality" errors that cannot be entirely eliminated by increasing a, and starting-point variability causes fast "response-caution" errors that can be eliminated by increasing a (see Damaso et al, 2020, for further discussion of these two types of errors in evidence-accumulation models).…”