Previously (SFN, 2002) we showed that processing of vertical-axis mirror symmetry (V) is not consistent with either a pure parallel or serial model, and is more efficient along the vertical midline than off-axis. Here we extend the analysis to horizontal (H) symmetry, to determine the generality and basis of these observations. Each stimulus consisted of four 8x8 arrays of black and white checks. Three of the arrays (distractors) were colored at random; in the fourth array (the target), bilateral symmetry was introduced in a graded fashion. We measured fraction correct (FC) and reaction time in a 4-AFC task. Trained observers (N=4) viewed these stimuli in single-symmetry blocks (V or H), and in mixed blocks (V and H). The four arrays were positioned 4 deg from fixation along the cardinal axes, and presented for durations of either 100 or 400 ms. In single-symmetry blocks, an RSVP mode (SFN, 2002) was also employed (100 ms per stimulus, 50 ms ISIs). In many respects, V and H symmetry results were similar: F...