Identifying rotated letter-like symbols

Previous research has shown that the identification of rotated alphanumeric symbols seems to be performed via the extraction of critical features encoded invariant to the symbol’s orientation. The present research argued that the use of such feature extraction processes might be a function of, first, a subject’s familiarity with the symbols, and second, the number of symbols from which a presented symbol is sampled. Earlier research has used highly over learned alphanumerics, in sets of six symbols; this practice is argued here as being seemingly conducive to feature extraction. In two experiments, the generality of a feature extraction interpretation, in contrast to one of mental rotation, was tested by having subjects previously trained to relative high- vs. low-familiarity criteria identify novel symbols in conditions in which a presented symbol was 1 of either 5 or 20 possibilities. Identification response times were found to be constant across all nonstandard orientations of presented symbols, irrespective of symbol familiarity or symbol set size. The findings support the generalization of a feature extraction interpretation to varying numbers of novel symbols of varying familiarity.