A Cognitive Model of the Same-Different Task Based on the Inhibition of "Different" Answers

Abstract

“[The] sense of sameness is the very keel and backbone of our thinking” (James, 1890). To make sense of the ever-shifting information in our environment, we constantly assess whether the world around us changes or not, if objects are the “same” or if they are “different”. This basic decision-making process is found from the lowest level of cognition (e.g. when contrasts are encoded by the retina), to the highest (e.g. when comparing concepts), and anywhere in between. In an experimental context, this process is studied with the “same-different” task, where subjects are asked if two stimuli presented sequentially are strictly identical or not. This experiment has been documented since the 1960s and its results have been replicated with diverse stimuli types (letters, shapes, faces, words, etc.). However, every attempt to model the subjects’ accuracy and response times on correct and incorrect answers simultaneously was unsuccessful so far. Part of the challenge in explaining this task is that “same” answers are faster than expected compared to “different” answers, a phenomenon called the “fast-same effect”. This thesis aims to assess whether a formal model based on the inhibition of “different” answers is plausible, effectively changing the problem from “fast-same” to “slow-different”. In the first chapter, I review the previous theories and models of the same-different task to learn why they failed. By elimination process, I identify the only cognitive architecture that seems congruent with the data. I then propose a model prototype based on the inhibition of “different” answers that implements this architecture. In the second chapter, I test this prototype with an experimental paradigm designed specifically to assess its plausibility. I conclude that resources should be spent in developing a formal model based on the inhibition of “different” answers, as the prototype’s qualitative predictions are confirmed by both the typical same-different data and the newly acquired data.