Category learning in infants and adults

The human brain has a remarkable ability to organise knowledge into a structured system. Categorisation represents one of the main mechanisms from which this system of structured knowledge is derived. The aim of this thesis is to provide a better understanding of novel category learning in infants and in adults. In Part I of the thesis, I explore how various parameters of the learning situation shape category learning in infants. In Experiments 1 – 3, I investigate the role of working memory demands and perceptual load in category learning. The number of items simultaneously available during learning was systematically varied in these experiments. Using a standard familiarisation-novelty preference procedure, infants were first presented with a set of training items, followed by the test phase in which category formation was assessed with the presentation of two items. Preference for one of the two objects was used as an index of category formation. Familiarisation was interleaved with test blocks in order to explore category formation at different levels of familiarity and to evaluate the dynamics of the learning process. To determine how an opportunity to compare items affects category learning, infants were familiarised with one item at a time (Experiments 1 and 1a), with a pair of items (Experiment 2), and with four items simultaneously (Experiment 3). The results revealed that paired presentation provides an optimal learning context, as infants showed evidence of category learning already after one familiarisation block. In Experiments 4–7, I explore the role of labels and motion in category learning with a gaze-contingent adaptation of the standard familiarisation-novelty preference procedure. The results indicate that adding motion as a category feature hinders visual category learning. In contrast, labels facilitate category learning, but only when both categories are accompanied by a category-specific label. When only one category is labelled, no learning can be observed. In Part II of the thesis, I investigate whether the same kind of implicit unsupervised category learning seen in infants is also present in adults, and how category learning shapes the visual processing of objects at the neural level. In Experiment 8, I develop a novel paradigm to explore implicit, unsupervised category learning in adults. There was no explicit mention of categories, and participants were not aware of the main aim of the task. The results revealed that adults, in the same way as infants, use statistical regularities to spontaneously organise a stream of visually presented items into categories. In Experiment 9, I investigated neural signatures of novel category learning in adults. By comparing neural signatures of visual object processing at the very beginning of the category learning process and after an extensive categorisation practice, I tested how categorisation experience alters the way in which exemplars are processed. The results revealed that categorisation experience is linked to a modulation of several event-related potentials components suggesting that categorisation may affect visual object processing at an early stage. In summary, this thesis demonstrates that category learning is a dynamic process shaped by different factors, which determine outcomes of the learning situation. In addition, the findings suggest that similar kinds of statistical category learning are present in infancy and adulthood.