Features for the identification of single letters and words

The determination of the visual features mediating letter identification has a long-standing history in cognitive science. Researchers have proposed many sets of letter features as important for letter identification, but no such sets have yet been derived directly from empirical data. In this talk, we will present a series of experiments in which we used Bubbles (Gosselin & Schyns, 2001) to reveal directly the visual features important for single letter (Fiset et al., 2008; Fiset et al., 2009) and word identification (Blais et al., 2008). The underlying logic of Bubbles experiments is relatively simple. If specific visual information is important for the task at hand, depriving the observer of this information will strongly impair his or her performance; in contrast, depriving the observer of unimportant information will not substantially alter his or her performance. We have shown that line terminations are the most important features for letter identification (Fiset et al., 2008; Fiset et al., 2009). We also replicated the finding reported by Huey more than one hundred years ago (Huey, 1908) that the upper parts of words are more important for word identification than the lower parts (Blais et al., 2008). Interestingly, we did not observe this upper parts bias for single letters (Fiset et al., 2008; Fiset et al., 2009). We will discuss implications of these findings—as well as other findings from our lab—for the design of highly legible types.

Back to the programme