We've recently complete a study an event-related fMRI study using a spatial VSTM task as part of a larger study. We expected to replicate several previous reports that VSTM capacity (as measured by K) would correlate with BOLD response in posterior IPS. We found the correlation, except it is negative - the larger the BOLD response to four objects, the lower K was. Subjects did complete training on the task prior to scanning but relying on the argument that the negative correlation reflects "neural efficiency" is a weak argument and hard to defend. Are there any reports, published or otherwise with similar negative correlation between capacity and BOLD response?