Abstract Past studies concerning iconic memory have reported greater accuracy of report for items from the middle row of tachistoscopically presented arrays. This effect has been attributed to nonselective readout from the middle row. The current study provides evidence that this effect is actually the result of selective readout guided by the cues inherent in the fixation stimulus used to focus attention.

Sperling (1960) introduced the partial report method as a means of demonstrating the existence of a high capacity, short duration visual store. Recent research showed that participant’s accuracy of report in the partial report paradigm varies with the row cued. Sanford, Daily, Anderson, and Peabody (1994) demonstrated that accuracy of recall differed significantly for the three lines in their arrays, with recall best for the middle row, intermediate for the top row, and worst for the bottom row. In a second experiment, Sanford, et al. varied the array location relative to the fixation point so that participants were focused with equal probability on each of the lines in a three line array. In all cases accuracy was highest for the focal line. Gegenfurtner and Sperling (1993) also reported better recall for the middle line of an array.

These studies presented the differential attention paid to the focal row as a characteristic of the partial report task. It is possible, however, that the nature of the fixation point provided participants an attentional focus. Participants are typically given a fixation point (a small cross or dot), the location of which corresponds to the center of the array. They are told to focus on that point and, when focused, to perform some action that initiates the array presentation. It seems likely that the small, compact nature of the fixation stimulus serves to narrow the focus of attention and that different methods of fixating attention on the location of the array, ones that encourage a broader focus of attention, may lead to different patterns of accuracy. The current study is designed to assess this hypothesis.

Participants were 42 students enrolled in undergraduate courses in Memory and Cognition. All had participated in a replication of Sperling (1960) in the laboratory section of the course.

IBM compatible PC’s were used to present the stimulus arrays and record responses. To avoid only partial display of the arrays, the software was written to synchronize array presentations with the vertical synchronization pulse of the computers’ video cards.

Participants were randomly assigned to three groups that differed only in the nature of the fixation point provided. One group was given the traditional fixation cross (+) presented at the center of the array location. The second group had the corners of the array location marked with appropriate ASCII characters. The third group received no fixation stimulus at all, in other words each trial began with a blank screen. Figure 1 illustrates the relationship between the various fixation stimuli and a superimposed stimulus array.

Figure 1. Examples of fixation stimuli used in the 3 conditions.

Each participant completed a total of 240 trials; 120 of the trials were practice trials and 120 were experimental trials. Within each 120 trials were 4 blocks of 30 trials. One of these blocks was whole report; participants were instructed to report as many letters as possible from the whole array. The remaining blocks were partial report; participants were to report only the cued line of the array. In one of the partial report blocks the cue delay was 0, in the others delays of 150 and 500 msec were used. For the partial report conditions, tones of 250, 650, and 2500 Hz were used to cue, respectively, the bottom, middle, and top rows. The order of the blocks was randomized for each participant and within the partial report blocks each line was cued in a random order with the constraint that each row was cued with equal frequency.

On each trial participants were presented with the appropriate fixation stimulus and, when focused, pressed a key to initiate the trial. The stimulus array was then presented for 100 msec. Stimulus arrays contained 12 randomly selected consonants and were presented in 3 rows of 4 letters in the center of the screen in the default PC font. After the appropriate cue delay (0 for whole report) participants typed their response on the computer keyboard. After recall the participants pressed the enter key and the fixation stimulus for the next trial was presented.

The mean number of letters correctly recalled in the partial report conditions as a function of row and cue delay is shown in Figure 2 for the fixation cross condition and in Figures 3 and 4 for the corners and no fixation conditions, respectively. Visual inspection suggests a main effect of cue delay, a possible main effect of row, and an interaction of

Figure 2. Mean number of letters recalled as a function of row and cue delay in the fixation cross condition.	Figure 3. Mean number of letters recalled as a function of row and cue delay in the corner fixation condition.

row and fixation condition. These intuitions were confirmed by a 3 (Fixation Condition) X 3 (Cue Delay) X 3 (Row) mixed analysis of variance with Fixation Condition as a between subjects factor and Cue Delay and Row as within subjects factors. The main effect of Cue Delay was significant, F(2, 78) = 28.09, p < .001, MS_e = .23. The number of letters correctly recalled declined as cue delay increased. The main effect of row was also significant, F(2, 78) = 22.15, p < .001, MS_e = .48.

Participants’ accuracy varied as a function of row, but this effect varied across fixation conditions as indicated by the significant Fixation Condition by Row interaction, F(4, 78) = 3.47, p = .012, MS_e = .48. In the fixation cross condition accuracy of report is highest for the middle row, followed by the top and then bottom rows. Accuracy for the top and middle rows is equivalent in the corners condition and both are better than the bottom row. Finally, participants’ accuracy does not differ significantly by row in the no fixation condition. No other main effects or interactions were significant.

Figure 4. Mean number of letters recalled as a function
of row and cue delay in the no fixation condition.

These results would be suspect had we failed to find the typical partial report advantage over whole report. Accordingly, we entered the partial report estimate at each of the 3 cue delays and whole report accuracy into a 3 (Fixation Condition) X 4 (Type of Report) mixed analysis of variance with Fixation Condition as a between subjects factor and Type of Report as a within subjects factor. As there was no significant effect of Fixation Condition, F(2, 39) = 1.10, p = .342, MS_e = 6.07, and no interaction of Fixation Condition and Type of Report, F(6, 117) = .42, p = .863, MS_e = .84, we collapsed across the Fixation Condition factor (see Figure 5 ). Post hoc analysis showed that the partial report estimate was significantly greater than whole report performance at the 0 and 150 msec cue delays, but not at the 500 msec cue delay. These results are consistent with past findings using the partial report procedure.

Figure 5. Mean number of letters recalled as a function of cue delay
collapsed across fixation condition.

We were successful in replicating the Sanford, et al. (1994) and Gegenfurtner and Sperling (1993) findings that accuracy in a partial report task is highest for recall of the middle row. We also demonstrated, however, that this result only holds when participants’ visual attention is focused prior to the presentation of the array using a single location focal point and that patterns of accuracy differ with different types of fixation stimuli. These results have implications for models of iconic memory. Averbach and Coriell (1961) suggested the notion of "nonselective readout" to explain why partial report performance does not decay to zero. Between the array presentation and the presentation of the cue, participants begin to transfer information about the array to a more durable storage. When the cue is presented, participants focus attention on the cued row and begin transferring information from that row to durable storage, a process that Gegenfurtner and Sperling (1993) call selective transfer. Gegenfurtner and Sperling incorporate these ideas into their model and suggest that transfer from the middle row of an array is nonselective. Our results suggest, however, that the part of the array that is involved in nonselective transfer depends on the nature of the fixation stimulus.

With a fixation cross, attention is tightly focused on the middle row and accuracy of recall is highest for that row. With the corners of the array offered as a fixation stimulus, attention is less tightly focused, allowing participants more leeway in selection of a location. Some may use the corners to locate the center of the array leading to a preference for that row, while others may choose to focus on the top row. Dick (1974) reported that participants do in some circumstances choose to process briefly presented arrays in a top-to-bottom, left-to-right fashion. Consistent with this interpretation is our finding that there is no difference in recall for the top and middle rows in the corners condition. Finally, participants in the no fixation condition could be presumed to have a great deal of difficulty focusing attention on a particular portion of the array since there is no clear indication of the array’s location and we found, in fact, no preference for one row over another in this condition. We suggest, therefore, that there is no essential difference between nonselective and selective transfer. What appeared to be nonselective transfer is actually selective transfer guided by cues inherent in the fixation stimulus used to focus attention.

Averbach, E. & Coriell, A. S. (1961). Short-term memory in vision. Bell System Technical Journal, 40, 309-328.

Dick, A. O. (1974). Iconic memory and its relation to perceptual processing and other memory mechanisms. Perception & Psychophysics, 16, 575-596

Gegenfurtner, K. R.& Sperling, G. (1993). Information transfer in iconic memory experiments. Journal of Experimental Psychology: Human Perception and Performance, 19, 845-866.

Sanford, J. F., Daily, L. Z., Anderson, M. A., & Peabody, A. Y. (1994, April). Differential attention within the icon: Implications for visual sensory memory. Paper presented at the meeting of the Eastern Psychological Association, Providence, RI.

Sperling, G. (1960). The information available in brief visual presentations. Psychological Monographs, 74, (Whole No. 498).