Human Category Learning in Reversal Tasks: Dynamic Learning Rates Help Overcome Catastrophic Interference in Connectionist Networks
Abstract
During category learning tasks participants often group together, i.e., “commonly code,” within-category exemplars. In two experiments, human participants learned to classify fractal exemplars into two arbitrarily-defined categories. After an initial learning phase, we reversed the exemplar-category assignments in two ways: a Total-Reversal (all assignments were switched) and a Partial-Reversal (only half were switched). If common coding occurs, participants should learn the Total-Reversal faster than the Partial-Reversal and should show within-category interference in the Partial-Reversal task (between reversed and nonreversed exemplars). Our behavioral results supported these predictions and in both experiments we found that participants reach higher accuracy in the Total-Reversal task. To understand the mechanisms behind these findings, we modeled category learning with a 3-layered artificial neural network (ANN). We observed that such models can readily form category representations, but suffer from catastrophic interference during reversal training. To overcome this problem, we modified the networks by applying different dynamic learning rates rules to their different layers. Inspired by associative learning theories (Mackintosh, 1975; Pearce and Hall, 1980), we incorporated a mechanism that promotes slower weight changes in the input-hidden layer and faster changes in the hidden-output layer as the ANN learns the task. This modification allows the ANN in the Total-Reversal task to maintain the category representations in the input-hidden layer simulating our empirical findings. This demonstrates that applying differential dynamic learning rates can help prevent catastrophic interference problems and provide a mechanistic explanation for common coding effects. Our study highlights the importance of associative mechanisms in category learning.
Related
- Connectionist models of human language processing : Progress and prospects
- Shedding Light on COVIS: Evidence from Optical Neuroimaging for Dissociable Category Learning Systems
- Orbitofrontal noradrenaline supports adaptive learning-rate adjustment in probabilistic reversal learning
- The ‘stickiness’ of disgust: Testing the hypothesis of disgust intractability in the context of instrumental reversal learning.
- Locus coeruleus–hippocampal pathway activation promotes spatial memory updating during reversal learning
BibTeX
@misc{obeso2026human,
title = {{Human Category Learning in Reversal Tasks: Dynamic Learning Rates Help Overcome Catastrophic Interference in Connectionist Networks}},
author = {Santiago Castiello de Obeso and Andrew R Delamater},
howpublished = {PsyArXiv},
year = {2026},
month = nov,
doi = {10.31234/osf.io/sbncd_v2},
url = {https://osf.io/preprints/psyarxiv/sbncd_v2/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Human Category Learning in Reversal Tasks: Dynamic Learning Rates Help Overcome Catastrophic Interference in Connectionist Networks
AU - Santiago Castiello de Obeso
AU - Andrew R Delamater
PY - 2026
DA - 2026/11/02
PB - PsyArXiv
DO - 10.31234/osf.io/sbncd_v2
UR - https://osf.io/preprints/psyarxiv/sbncd_v2/
AB - During category learning tasks participants often group together, i.e., “commonly code,” within-category exemplars. In two experiments, human participants learned to classify fractal exemplars into two arbitrarily-defined categories. After an initial learning phase, we reversed the exemplar-category assignments in two ways: a Total-Reversal (all assignments were switched) and a Partial-Reversal (only half were switched). If common coding occurs, participants should learn the Total-Reversal faster than the Partial-Reversal and should show within-category interference in the Partial-Reversal task (between reversed and nonreversed exemplars). Our behavioral results supported these predictions and in both experiments we found that participants reach higher accuracy in the Total-Reversal task. To understand the mechanisms behind these findings, we modeled category learning with a 3-layered artificial neural network (ANN). We observed that such models can readily form category representations, but suffer from catastrophic interference during reversal training. To overcome this problem, we modified the networks by applying different dynamic learning rates rules to their different layers. Inspired by associative learning theories (Mackintosh, 1975; Pearce and Hall, 1980), we incorporated a mechanism that promotes slower weight changes in the input-hidden layer and faster changes in the hidden-output layer as the ANN learns the task. This modification allows the ANN in the Total-Reversal task to maintain the category representations in the input-hidden layer simulating our empirical findings. This demonstrates that applying differential dynamic learning rates can help prevent catastrophic interference problems and provide a mechanistic explanation for common coding effects. Our study highlights the importance of associative mechanisms in category learning.
ER -