Duration differences between rare and common words and their implications for the interpretation of word frequency effects

The spoken durations of rare and common words were compared in three experiments. Depending upon the task and the amount of practice, rare words (with frequencies less than 3 per million) were spoken as much as 24% slower than common words (with frequencies greater than 100 per million), when the words were equated for number of letters. This difference was observed even when the memory and lexical access demands of the task were minimized, and it can be explained by differences in the phonetic constituents of the two classes of words. The existence of these phonetic differences has been previously reported by Landauer and Streeter (1973) and by Zipf (1935). These findings argue against the view that all effects on the processing of accurately perceived words that are correlated with frequency may be unambiguously ascribed to operations that involve secondary memory (although this may be true of frequency effects per se). One recent experiment (Watkins, 1.977) is examined in light of these findings.