Seamon (1972) has found that reaction times (RTs) to memory probes do not increase with memory set size (M) for words encoded in mental images in a memory scanning task. Rothstein and Atkinson (1975) have failed to replicate Seamon's results. Experiment 1 investigated this discrepancy by manipulating one methodological difference (whether images were or were not described) between the two investigations. Results revealed that described images produced typical linear increases in RT with M. Undescribed images, however, revealed no change in RT with M for positive probes (i.e., the word presented was contained in the current memory set).Experiment 2 manipulated to which visual field probes were presented (RVF = right visual field; LVF = left visual field). Results showed that the imagery group showed no relation between RT and M for probes presented to the LVF, but a linear relation for probes presented to the RVF. Described-image and repetition groups revealed a linear relation between RT and M no matter which visual field received the probe.Since Sternberg (1966) introduced the memory scanning task, hundreds of studies have been performed to investigate the type of processing that occurs in item recognition. In this paradigm, a small number of stimuli (the memory set) is presented to the subject. A probe or test stimulus is then presented to the subject. The sub· ject's task is to make a "yes" response if the probe is a member of the memory set and a "no" response if it is not. Sternberg (1966) found a linear function for the relationship between reaction time (RT) and memory set size (M). He concluded that subjects engaged in a serial search process in which subjects sequentially compared the encoded test stimulus against the memorial representations of each of the items in the memory set. In addition, he concluded that the serial search process was exhaustive because the slopes of "yes" and "no" RTs in relation to memory set were parallel. In general, a monotonic increase in RT with M has typically been found (e.g., J. A. Anderson, 1973;Briggs, 1974;Nickerson, 1972). Seamon (1972) has presented an interesting exception to the generalization that RT is linearly related to M. He found that a group of subjects instructed to use relational imagery to remember the words in each memory set (varying in size from one to three words) showed no increase in RT with M. A second group in Seamon's investigation using a repetition strategy to remember words in the memory sets exhibited the more common linear increase of RT with M. Seamon suggested that memory codes for repetitively rehearsed memory sets required a sequential comparison process, whereas words represented in a visually imagined scene could be processed in parallel. 1 Similarly, Paivio (1971) has suggested that imagery leads to a visual memory code that he believes is a parallel processing system, in contrast to verbal processing, which is thought to operate in a sequential manner. Rothstein and Atkinson (1975), however, attempted and failed to replicate Se...