Reliability of phase-encode mapping in the presence of spatial non-stationarity of response latency.

Phase-encode mapping has been used as the primary functional magnetic resonance imaging (fMRI) technique to reveal spatial variation of response properties in both visual and auditory cortex (e.g., retinotopy and tonotopy). Spatial variation in the time of response produces a map of cortical sensitivity to a time-varying property of the presented stimulus. Inherently, this technique assumes that response latency is invariant across cortex. However, the latency of the fMRI response has been found to exhibit both temporal and spatial non-stationarity, with variance across cortex increasing with longer stimulus duration--a critical concern given the typically long duration (e.g., 48-64 s) of stimuli presented in phase-encode mapping experiments. This study addresses two issues--assessment of error rates caused by latency variance and reliability of maps identified using phase-encode mapping. Our findings suggest that the latency variance was found to have a greater effect on the type II error (missed detection) rate than on the type I error (false alarm) rate, with the size of the cortical map controlling the type I error rate. Moreover, empirical determination of false alarm rates provides the first approximation of the statistical significance (i.e., P values) of observed maps in both retinotopic and tonotopic studies, with maps as short as three voxels in length achieving the P < 0.05 significance level for a single subject.