Seeing the parts versus seeing the whole: How attention affects perception and learning

Attention, perception, and learning

What’s the difference between what we see and what we perceive? Turk-Browne’s research suggests that we switch between two distinct but similar processes depending on where our attention is. One process involves understanding relationships among individual parts (statistical learning), while the other involves understanding the whole as a collection of parts (statistical summary perception). It is possible that there are common cognitive systems or networks underlying both processes, which would mean that these systems are responsible for processing the same information in different ways.

On the surface, statistical learning and statistical summary perception are quite different: statistical learning involves extracting regularities over repeated experience, while statistical summary perception involves extracting statistics from a single display; and statistical learning involves acquiring stimulus-specific relationships (i.e. that this particular object co-occurs with another particular object), while statistical summary perception (by definition) involves representing the general properties of a collection.

Despite these surface differences, however, these two processes are both inherently statistical: they involve aggregating a sample, and distilling this sample to statistics… Thus, statistical learning and statistical summary perception may interact in meaningful ways. We tested what happens to statistical learning when objects are attended during a statistical summary task….
statistical summary perception
Statistical learning was robust in the passive viewing and control task conditions, but was weaker — actually, non-existent — in the summary task condition. This suggests that computing summary statistics may interfere with statistical learning, possibly because of a reliance on shared statistical computations, or because of a reliance on different spatial scales of attention. That is, global attention helps summary performance, while local attention may be necessary for stimulus-specific learning.

Statistical learning and its consequences | Nicholas B. Turk-Browne


The surprising connection between two types of perception | Association for Psychological Science