In typical categorization tasks, humans are presented with a sequence of instances and report whether each instance is a member of a given category or not. In the current study, we examine the relationship between the reaction times (RTs) of human participants and the position of the instance in the conceptual space. Our main hypothesis is that instances closer to the boundary of the two categories, which are harder to be categorized, will require longer cognitive processing, resulting in longer RTs. Human subjects categorized images of novel objects to one of two given categories (represented by images of their prototypes); the selected category, RT and confidence rating for each trial were recorded. For trials with longer RTs people responded with less confidence and were more prone to making errors than for trials with shorter RTs. Moreover, people responded faster to stimuli with high similarity to at least one of the prototypes of the given categories than to stimuli that were distant from both prototypes, and hence closer to the boundary of the two categories, confirming our main hypothesis.
The above results, though preliminary, are very promising. First, they replicate previous findings exploring the meaning of RTs in categorization tasks, while limiting potential effects arising from the nature of previous experimental designs. We consider that replicating previous results even with the use of novel images that form unspecified concepts, indicates that our basic hypothesis is primitive w.r.t the basic processes of human categorization. Second, the experimental design we used, combined with the findings of the present study, uncover many hidden aspects of previous studies, opening the way to future work towards multiple directions. We are currently investigating possible bias effects arising from the position of the prototypes (top / bottom) or by any other presentation effects. Eye-tracking techniques can also be used to better interpret findings from RTs, as a quantitative method of the cognitive processes involved in the task, as well as a tool for exploring other possibe effects and revealing biases. Future work could also involve experimentation with more familiar stimuli, such as (i) images of familiar objects, (ii) images depicting more than one objects, or (iii) excerpts of text, which could be characterized by multiple labels, etc.