A hypothese about the learning and memory of symmetrical patterns in graphic design was tested in the present study. The hypothese, which states that symmetrical patterns are easier to be remembered, was investigated with a pattern-recognition experiment. In the experiment, graphic patterns were displayed for 1.5 second during the learning episode. The subjects were then asked to pick up the patterns that they just saw with a 4AFC task to test their pattern-recognition, and the rates of correct answer were recorded. The results showed a better general performance for the symmetrical patterns over the non-symmetrical patterns in the recognition task. Symmetrical patterns that produced with different structure principles also have different levels of correct rates. Both the compositional elements used to build the experimental patterns and the styles of symmetry affected the recognition task performance. Patterns composed by△○□elements had lower recognition rates than those composed by a single unit. Symmetrical patterns with both mirrored and radial structures showed the best recognition performance over other symmetrical patterns tested in the present study.