Anthony,
With statistical consideration aside, I'm more troubled with the approach of group analysis of zscores in terms of hypothesis testing. usually when we run group analysis, we want to know whether there are any main effects across factor levels or/and differences between some specific factor levels. Such tests make sense in the perspective of percent signal change of hemodynamic response.
On the other hand, the conversion of t value to zscore is equivalent to scale the percent signal change by individual subject's variance (intra-subject variance), and the main effects and contrasts would be regarding the scaled percent signal change, not pecent signal change per se.
Another difference between the two approaches basically boils down to the underlying normality assumption: one about percent signal change, the other about the percent signal change adjusted by intra-subject variance.
Gang