Cognitive Control

The stop signal paradigm is a popular tool to study response inhibition. Participants perform a response time task (go task) and, occasionally, the go stimulus is followed by a stop signal after a variable delay indicating subjects should withhold their response (stop task). The main interest for modeling is in estimating the unobservable stop-signal processing time, that is, the covert latency of the stopping process as a characterization of the level of response inhibition mechanism. In the dominant model performance is hypothesized as a race between two stochastically independent random variables representing go and stop signal processing (independent race model, IRM). Different of versions of the IRM including parameter estimation methods have been proposed, in particular classic non-parametric ones by G. D. Logan and colleagues and parametric ones by D. Matzke and colleagues. An important prediction of all independent race models is that the distribution of reaction times to the go signal, without a stop signal being present, lies below the go signal distribution of when a stop signal is presented after a certain time interval (stop signal delay, SSD). On the other hand, consistent violations of this prediction have been observed for certain SSD values (e.g., P.G. Bissett and colleagues). Here we propose non-independent versions of the race model based on the statistical concept of copula. Copulas allow one to study multivariate dependency separately from assuming specific marginal distributions. We investigate under what conditions these new race models are consistent with violations of the distribution inequality stated above.

This is an in-person presentation on July 19, 2023 (11:20 ~ 11:40 UTC).

The stop-signal paradigm is a cornerstone of research on response inhibition and there is a rich history of formal cognitive models that explain individuals’ behavior on the task. It is therefore not surprising that this task has been included as the primary measure of response inhibition in the Adolescent Brain Cognitive Development (ABCD) Study, a longitudinal neuroimaging study of unprecedented scale that is in the process of following over 11,000 youth from middle childhood though age 20. However, the ABCD Study's unique task design involves a visual stop-signal that replaces the choice stimulus, creating a masking effect that impedes information processing on trials with short stop-signal delays. As this design feature violates the critical "context independence" assumption shared by most current methods for estimating stop-signal reaction time (SSRT), some experts have called for the task to be changed or for previously collected ABCD data to be used with caution. We present a cognitive process modeling framework that provides a parsimonious explanation for the impact of this design feature by combining prior insights on the effects of visual masking on choice evidence accumulation with recent “hybrid” racing-diffusion ex-Gaussian (RDEX) approaches to modeling the stop-signal task. We demonstrate that the resulting model, RDEX-ABCD, successfully accounts for key behavioral trends in ABCD data, including the inhibition function and the impact of context independence violations on choice accuracy rates. Simulation studies using this model suggest that failing to account for context independence violations in the ABCD design can lead to erroneous inferences in several realistic scenarios. However, RDEX-ABCD effectively addresses these violations and can be used to accurately measure the timing of response inhibition processes and additional mechanistic parameters of interest. More broadly, results demonstrate the feasibility of addressing context independence violations by building process-based explanations for them into models of the stop-signal task.

This is an in-person presentation on July 19, 2023 (11:40 ~ 12:00 UTC).

The stop-signal reaction time (SSRT) is often interpreted as capturing people’s cognitive ability for rapidly and flexibly inhibiting prepotent actions when not desirable anymore. Its calculation is designed to circumvent the effect of strategy on baseline speed in the hope to specifically isolate the speed of signals conceived as reactive, top-down and inhibitory. However, the ability to act upon any signal, be it by producing or withholding a response, also always depends on i) when the information becomes available within decision areas, i.e. sensory delay, and ii) how long actions take to be executed (because only actions still in the planning stage can be withheld, and only executed actions are observable behaviourally). In line with (i), SSRT is clearly influenced by stop-signal contrast. In line with (ii), manual SSRT is also clearly longer than saccadic SSRT. In recent work (https://www.biorxiv.org/content/10.1101/2023.02.20.529290v1), we showed that the sum of (i) and (ii), i.e. non-decision time, is directly measured by the dip onset time (the earliest time point where the signal-absent and signal-present RT distributions depart). Importantly, this measure is immune to top-down factors, including whether the task is to stop or ignore the signal. Predictably, we show that individual differences in manual and saccadic SSRT are correlated with dip onset time (all r > 0.5 across datasets). This result is consistent with individual differences in SSRT largely reflecting sensory and/or motor influences, rather than being a pure measure of top-down inhibition speed. Alternatively, there may also be genuine correlations within the population between, on the one hand, the speed of bottom-up and/or motor signals and, on the other hand, the speed of top-down signals. Either way, this result calls for a reconsideration of past conclusions drawn from the use of SSRT (554 articles according to pubmed on the 29th of March, 2023), in particular studies reporting changes in SSRT in relation to clinical conditions, drugs, age, personality traits, general intelligence, brain structure, brain function, genes or performance on other tasks. Our results open the possibility that within and between-individual differences in SSRT, so far interpreted as differences in top-down control abilities, largely reflect variations in sensory delay and/or motor output time.

This is an in-person presentation on July 19, 2023 (12:00 ~ 12:20 UTC).

While the relation between congruent and incongruent conditions in conflict tasks has been the primary focus of cognitive control studies, the expectation of neutral condition behavior is oft ignored, or set as directly between the two conditions. However, empirical evidence suggests that average neutral reaction time (RT) contradicts this assumption. The present studies, thus, sought to, first, reinforce the informative nature of the neutral condition and, second, to highlight how it can be useful for modeling. To do this, we first explored how RT in the neutral condition of conflict tasks (Flanker, Stroop, and Simon Tasks) deviated from the predictions of current diffusion models. Many of the deterministic versions of cognitive conflict models predicted a neutral RT that is the average of the congruent and incongruent RT, called the midpoint assumption. This assumption is maintained over the time course, resulting in the parallel assumption. To investigate these assumptions, we first conducted a limited literature search that recorded the average RTs of conflict tasks with neutral conditions. Upon finding evidence of a midpoint assumption violation with smaller RT differences between the congruent and neutral conditions, we tested the prior mentioned conflict tasks with two different, distinct sets of stimuli. The results suggested that a violation of the midpoint assumption is present in different manners depending on the conflict task and the stimuli. Then, a follow up study was performed in order to test the parallel assumption via the Speed-Accuracy Tradeoff paradigm. From the results, clear violations of the parallel assumption were observed in all three conflict tasks. Due to the implications of these violations, the authors then suggested possible elaborations of the Diffusion Model of Conflict to account for these phenomena.

This is an in-person presentation on July 19, 2023 (12:20 ~ 12:40 UTC).