It has been found that the 3-Gabor alignment task is performed equally well whether the orientations of the patches are collinear, parallel, or orthogonal. But does the orientation of the individual micropatterns have any effect on performance? The ability (threshold) of subjects to determine whether the middle Gabor patch of a 3-patch display was displaced to the right or the left was measured in a variety of conditions. When the orientation of the Gabors was random, performance was poorer than in the three standard conditions mentioned above. If the outer Gabors had orientations of 45° and 135° and the centre patch was vertical, then thresholds were even higher, and the point of subjective equality for the task was displaced from centroid alignment towards a position where the patches form a head-to-tail contour. However, phase randomisation did not affect performance appreciably compared to the standard conditions, and, if Gaussian patches with normally distributed standard deviations were employed, the uncertainty on the Gaussian patch size did not impoverish the ability of observers to do the task. The implication of these results is that orientational factors can affect performance, and that the visual system attempts to form contours out of plausibly located oriented segments, thus masking centroid cues. In the absence of orientational disruption it does indeed appear that centroid-like cues are used, as the randomisation of bar positions or Gaussian blob size did not degrade thresholds. The universal result that orientation does not improve performance indicates that its role in this task is that of a mask rather than a cue.