Abstract
We examined how word length affects performance in three recognition memory experiments to resolve discrepant results in the literature for which there are theoretical implications. Shorter and longer words were equated on frequency, orthographic similarity, age of acquisition, and imageability. In Experiments 1 and 2, orthographic length (i.e., the number of letters in a word) was negatively related to hits minus false alarms. In Experiment 3, recognition performance did not differ between one- and two-syllable words that were equated on orthographic length. These results are compatible with single-process item-noise models that represent orthography in terms of features and in which memory representation strength is a product of the probabilities that the individual features have been stored. Longer words are associated with noisier representations than shorter words.