The influence of local and global attributes of symmetric patterns on the perceptual salience of symmetry was investigated. After tachistoscopic viewing, subjects discriminated between symmetric and either random patterns (experiment 1) or their perturbed counterparts (experiment 2) created by replacing one third of the mirror element-pairs of symmetric stimuli with 'random' elements. In general, it was found that perceptibility of symmetry, measured by response time and detection accuracy, was not influenced in a consistent way by type of pattern element (dots or line segments oriented vertically, horizontally, obliquely, or in all three orientations about the symmetry axis). Nor did axis orientation (vertical, horizontal, oblique), advance knowledge of axis orientation, practice effects, or subject sophistication differentially affect detection. A highly salient global percept of symmetry emerged, on the other hand, when elements were clustered together within a pattern, or grouped in symmetric pairs along a single symmetry axis or two orthogonal axes. Results suggest that mirror symmetry is detected preattentively, presumably by some kind of integral code which emerges from the interaction between display elements and the way they are organized spatially. It is proposed that symmetry is coded and signalled by the same spatial grouping processes as those responsible for construction of the full primal sketch.