Glanzer and Adams (1990) described what they called the mirror effect. This is the strong regularity in recognition memory that if targets are recognized more accurately in a condition of the experiment the foils tend also to be rejected more accurately. This can be shown with respect to a number of factors but one of the most researched factors is word frequency. If subjects are shown high-frequency words and are tested with these mixed in with high-frequency foils they tend to do worse than if they are shown low-frequency words and tested with these mixed in with low-frequency foils. This effect shows up both in increased recognition of low-frequency targets and increased rejection of low-frequency foils. The mirror effect with respect to frequency has been recently addressed by McClelland and Chappell (1994) and Shiffrin (1995) who proposed similar models which in effect say that low-frequency words have more discriminative feature representations which increase distinctiveness of the stimuli.
We ran a simulation of one of the experiments by Glanzer and Adams (1990; Exp. 2) which illustrates the mirror effect with respect to word frequency. Subjects in the Glanzer and Adams experiment studied a list of 148 items of which the first and last 24 were buffer items. The middle 100 were a random mix of 50 low-frequency words (mean log frequency from the Kucera and Francis norms (1967) of 2.5) and 50 high-frequency words (mean log frequency of 5.1). The words were presented at a 1 second rate. Then subjects were tested on a list of 200 words which consisted of the original 100 plus 50 low frequency distracters and 50 high frequency distracters.