Within-subject effects were first estimated in each individual subject using a fixed-effects model in which both the placebo and alcohol sessions were modeled. Linear drifts were corrected, but temporal smoothing or autoregression was not used given the long SOA (Della-Maggiore et al., 2002). The discrete 2 second periods of odorant (or sham) valve openings (light gray tracing, Fig. 1) were convolved with SPM’s canonical hemodynamic reference function, which initial analyses showed to provide better olfactory sensory system responses (activation in piriform and orbitofrontal cortex) than convolution with the entire 36-second blocks of odor valve openings (black dashed tracing, Fig. 1). A binary mask was applied to constrain within-subject analyses to gray matter by using each subject’s own smoothed (8 mm FWHM) gray matter mask, as derived by SPM2’s segmentation algorithm. This fixed-effects model was then used to create contrast images representing the average activation effects of AO, NApO, and, ApCO (each of these odor types compared to sniffing non-odorized air) in alcohol and placebo sessions.
