Fig. 5
From: The interplay of uncertainty, relevance and learning influences auditory categorization
Schematic diagram of the 3 decision-making models; psychometric curves and model fits shown for both representative participants as well as averaged across all participants. (A) A schematic of all 3 decision-making models. (top) The full Bayesian model: the participant considers that any tone can probabilistically either be ‘signal’ or ‘distractor’ (see main text for more information). Using a hypothetical trial, we illustrate that deciding that a tone is ‘signal’ versus ‘distractor’ shifts participants’ category choice probabilities. (center) The no-distractor model: the participant considers all tones to be ‘signal’ tones, and thus a tone that the full Bayesian model interprets as a distractor might instead by interpreted as a high signal tone. (bottom) The random-guess model: the participant considers all tones to be ‘distractor’ tones and thus makes their decision randomly. \(\widehat{L}\): \(\widehat{Low}\)-, \(\widehat{H}\): \(\widehat{High}\)—category decision. (B, C, D) Psychometric curves (black lines) and respective model fits (full Bayesian model: orange lines, no-distractor model: teal lines, random-guess model: magenta lines) for 3 example participants (for posteriors of the full Bayesian model, and for GLM analysis, see Fig S3; for more participants see Fig S4). The x-axes of the psychometric curves span the frequency range of the stimulus tones, and the y-axes denote the mean probability of a \(\widehat{High}\) category choice response. Error bars: standard error of the mean. (B) Example participant with high task accuracy on trials with either one or two distractor tones. (C) Example participant with high task accuracy on trials with one distractor, and poor accuracy on trials with two distractors. (D) Example participant with poor accuracy on trials with both one distractor and two distractor trials. (E) Psychometric curve and model fits averaged across all participants.
