Monotonicity in graph theoretic summaries of fMRI data acquired during human learning.

Introduction: Behavioral performance during associative learning typically improves monotonically; performance on each successive iteration of the task is no worse (and typically better) than on the previous one. It is unclear whether connectomic measures of brain function (from fMRI data acquired during learning) also increase monotonically. We used a well-established associative learning paradigm to test for the possible co-observance of monotonicity in behavior and connectomics. Methods: fMRI data were summarized using two distinct connectomic (i.e., graph theoretic) measures: (a) Betweenness Centrality (of nodes) and (b) Average Shortest Path Length (i.e., a measure of network efficiency) across the graph. To broaden our study's breath, in addition to healthy controls ( Results: We observed robust evidence for monotonic changes in behavior at the group level, and in most participants regardless of group. Evidence for monotonic changes in graph theoretic summaries of the co-acquired fMRI data was less widespread and was in general, more evident in group level summaries (regardless of group). Discussion: This modest co-observance of monotonicity in behavior and fMRI-based connectomics re-emphasizes what has long been suspected: the relationship between overt measures of behavioral competence and the co-acquired imaging signals is complex. This may be because psychological events (whether in the healthy brain, or in clinical populations like schizophrenia) emerge not from local activity in circumscribed brain regions, but rather from widely distributed activity across the brain. While well-defined mathematical concepts like monotonicity can anchor attempts to co-observe properties of change in overt behavior, and underlying brain signals, we suggest that the search for such relationships will remain a challenge.