Andrew Leber

Department of Psychology
University of New Hampshire

In our approaches to understanding cognitive processing, we are often reminded of the pitfalls
of group-level analyses, which can present the picture of an "average subject" that does
not exist in reality.  That is, if some subjects exhibit a pattern of data at one extreme and the
remaining subjects show another extreme, no individual exhibits a pattern resembling the
group average.  Less frequently considered, however, are the related pitfalls of averaging
single-subject data, which can present the picture of an individual who does not exist in
reality. That is, if the subject shows a pattern of data at one extreme for some proportion of the
time and shows another extreme for the remaining time, at no moment in time does the
individual resemble his/her own session average.  As a result, a key tool for investigating
virtually any phenomenon of interest is to directly examine within-subject variability in
behavior over time.

In my talk, I will describe recent examinations of within-subject variability, using fMRI, to
illuminate mechanisms of cognitive control.  Subjects were scanned while participating in
task switching or attention capture procedures.  To examine within-subject variability, each
voxel in the brain was independently treated as a potential covariate of behavioral
fluctuations.  Results from this approach provided 1) confirmation that single subjects indeed
show wide variation in their patterns of behavior over time, 2) the ability to predict in advance
which pattern of behavior the subjects would display from moment to moment, 3) description
of which brain regions underlie the fluctuations in behavior, and 4) a clearer picture of how
the brain enables shifts in cognitive states over time.  Additional discussion of the advantages
and limitations of the current approach, as well as potential future directions, will be included.