Memory & Cognition

The normative frequency of words is empirically related to performance of a variety of memory tasks, including recognition (Schulman, 1967; Shepard, 1967). Low-frequency (LF) words are better recognized than high-frequency (HF) words, and the word frequency effect (WFE) is an example of a mirror effect (Glanzer & Adams, 1985): For studied words, LF words are more likely than HF words to be recognized as being old (i.e., a hit), and for unstudied words, LF words are less likely than HF words to be recognized as old (i.e., a false alarm).1 We refer to these components of the WFE as the LF hit rate (HR) ad-vantage and the LF false alarm rate (FAR) advantage. In the following experiments, we investigated the effect of at-tentional strain on the LF HR advantage, and we tested a hypothesis for the allocation of attentional resources dur-ing learning. In some theories, the allocation of attentional resources plays a critical role in producing the WFE. This account, referred to here as the elevated-attention hypothesis, is based on two premises: (1) LF words attract more atten-tion than do HF words, and (2) the strength of an encoded episodic trace is a positive function of the amount of at-tentional resources allocated during encoding (e.g., Glanze