Experiment 1 demonstrates that, while the outline shapes of bilaterally symmetrical dot patterns play a role in symmetry detection, the removal of the outline by a surrounding random-dot annulus merely reduces performance by a fixed amount. It does not affect the relative salience of different symmetry-axis orientations. The converse is also true: performance is almost as good when the symmetrical dot pattern is confined to the surrounding annulus but is disrupted similarly when the central area is filled with random dots (Experiment 2). In Experiment 3, any one or more of three areas — a central circular area and two surrounding annuli — could be filled with vertically symmetrical or random dots and symmetry was detected reliably only when the central circular area contained the symmetrical dots. A new paradigm was explored in Experiment 4: subjects judged the orientation — left or right of vertical — of 20-dot symmetrical patterns oriented 1-4 deg left or right of vertical and with or without surrounding random-outline masks. Surprisingly, the orientation judgments were extremely precise, with JNDs of the order of 1 deg or less, and outline masking dots made no difference. Experiment 5 showed that performance was even better with just two dots and it made no difference whether these were separated by just under 4 deg or just over 20 deg. It was concluded that while pattern outline plays some role in dot symmetry detection, the major mechanisms are located near the fovea. It is suggested that the relative salience of vertical and horizontal symmetry only in central vision is related to the preponderance of foveal cells tuned to those orientations (Mansfield, Science 186, 1133-1135, 1974; Mansfield and Ronner, Brain Res. 149, 229-234, 1978) but that attentional factors and scanning strategies also boost the salience of oblique symmetry and the salience of vertical over horizontal symmetry (Wenderoth, Perception 23, 221-236, 1994).