A Reexamination of the Effect of Word Length on Recognition Thresholds

Two studies were done to reassess the influence of word length on recognition thresholds. In the first, length varied from 3 to 11 letters while word frequency varied over three levels. The main effect of length and the interaction of length by frequency were not significant. The second study used only low-frequency words with length varied over four levels. No statistically significant effect of length was found. The present results are integrated with previous work via a two-process theory of the effect of length: word length influences recognition thresholds if the words are unfamiliar, but not if they are familiar. Early in the study of word recognition, Cattell (1886) found that visual recognition thresholds are lower for words than letters and that words in a prose passage can be read almost as fast as a list of letters. In their original paper on the effect of word frequency, Howes and Solomon (1951) reported that word length did not affect recognition thresholds in their experiment. However, they did find that words with many letters per syllable had higher recognition thresholds than words with short syllables. Three papers are commonly cited as evidence that (a) increasing word length linearly increases thresholds for English words and (b) the effect of word length decreases as a function of word frequency. That is, length has its greatest effect for low-frequency words. The studies of McGinnies, Comer, and Lacey (1952) and of Postman and Adis-Castro (1957) both used the same list of English words. Both studies employed four levels of word length and five levels of frequency, but only 20 words - only 1 word for each combination of length and frequency. Part of speech was not controlled in assembling this word list. No matter how random the sampling scheme employed to obtain these words, a conclusion based on single words in each cell of a 4 X 5 matrix is open to suspicion. These two studies were totally within-subject designs. Both reported a significant main effect of word length and a significant interaction of length by frequency. In this type of experimental design, if the interaction term is