A systematic exploration of temporal bisection models across sub- and supra-second duration ranges


Journal Article

© 2019 Elsevier Inc. An integral component to the validity of timing models is their ability to accurately fit behavioral data from detection and discrimination tasks such as the temporal bisection procedure. Two of the most prominent timing models are the Sample Known Exactly (SKE), based on scalar timing theory, and the pseudo-logistic model (PLM). Recently, evidence accumulation models based on drift–diffusionprocesses (DDM) have been utilized for modeling temporal bisection data. Currently, there is no standard by which timing behavioral data are fit, resulting in the implementation of both theoretical and atheoretical models, such as generalized logistic functions (L4P). As differences in timing behavior have been shown across sub- and supra-second durations, a comparative evaluation of these 4 different types of models (SKE, PLM, L4P, and DDM) was conducted to assess each model's ability to capture these differences using timing data from the temporal bisection procedure. Psychometric functions from rats, trained on a bisection task using sub-sec (200 ms vs. 800 ms.) and supra-sec (2 s vs. 8 s) conditions, were fit with each of the four models. Using Akaike Information Criterion (AIC) we demonstrate that theoretical models vastly outperformed the L4P across both duration ranges, Furthermore, significant differences existed in key parameters between L4P and PLM. Three DDM models were analyzed with varying degrees of freedom, which showed that allowing for non-decision time, drift rate, and starting point to vary across signal durations outperform simpler models (Δi≥ 10) that only allowed for variation in drift rate or drift rate and starting point. These models explained variance in both choice and reaction time data. While we were unable to directly compare timing and temporal decision models the results from both demonstrate the need for a shift toward theoretically based models such as the SKE, PLM, or DDM which provide greater parsimony as well as provide for greater qualitative analysis and interpretation of the data.

Full Text

Cited Authors

  • Lusk, NA; Petter, EA; Meck, WH

Published Date

  • February 1, 2020

Published In

Volume / Issue

  • 94 /

Electronic International Standard Serial Number (EISSN)

  • 1096-0880

International Standard Serial Number (ISSN)

  • 0022-2496

Digital Object Identifier (DOI)

  • 10.1016/j.jmp.2019.102311

Citation Source

  • Scopus