An analysis of previous studies of bilateral symmetry detection in dot patterns revealed what appeared to be an almost arbitrary choice of pattern parameters and constraints with no systemic examination of the effects of these parameters and constraints on observer performance. In Expt 1, 100-dot patterns either had no constraints on how they were plotted or had one or both of two constraints: either no dot was permitted to be plotted within a fixed distance of any other dot; or randomly selected dot radii were transformed to make the dot distribution more uniform. While a large vertical symmetry salience effect was obtained, both in number correct and reaction time, only marginal differences occurred between the various constraint conditions. However, when number of dots in the pattern was varied in Expt 2, increasing dot number from 10 to 80 had no effect at all on vertical symmetry detection but linearly decreased performance for other axis orientations. Experiments 3 and 4 together suggested that the critical variable producing the performance decrease was number of dots per se, not increasing dot density (which would tend to give all patterns a more circular outline) and not decreasing the distance between neighbouring dot. Thus, the relative salience of vertical over other symmetries is critically dependent on number of dots in the patterns and it is suggested that vertical symmetry is processed globally so that dot pairs are compared in parallel, whereas at other axis orientations symmetry is processed locally so that dot pairs are compared in serial fashion. Possible neurophysiological and cognitive factors are discussed which might account for the relative performances with different symmetry axis orientations.