Extract from the syllabus of P. Bertelson "The treatment of information in humans" (1983-1984),pp.149-151
We recommend that you first go to the experimental situation as a naive subject.
To make things easier you can also make the applet in a pop-up window appear which remains on screen while you read the text.

Demonstration

The discovery of iconic memory is due to George Sperling (1960). The starting point for Sperling's work is the classical experiment on the determination of the memory-span. If a certain number of elements, letters, for example are presented using a tachistoscope it can be seen that the number of correct responses does not exceed 4 to 5 regardless of the number of letters presented. Provided that the number of elements presented remains below this limit, all the elements are reproduced correctly; if more than this are presented performance remains at this limit however many are presented. This limit is known as the memory-span.

The notion of whole report procedure excited a lot of interest at the end of the last century after Javal discovered the phenomenon of fixations and saccades : presentation by means of the tachistoscope became a way of simulating what happens during a fixation. Whole report procedure suggests the existence of a fixed limit to the sensory information which can be processed during a fixation. The idea that whole report procedure reveals a limitation to the quantity of sensory information processed in a presentation did not always seem to agree with subjects reporting the impression hat they had seen more elements than they could remember. This impression would suggest another interpretation of the procedure which was that it was a limitation of mnemonic order.

Sperling's main contribution has been to find a technique which has supplied a clear answer to the question we have just posed: this is the partial report procedure. In one of his experiments he presents a matrix of 3 lines of four letters each to his subject for 50msec. If the subject is asked to reproduce what he has seen, on average four answers are correct. Under partial report procedure a high-, medium- or low- pitched tone is produced at the same time as the presentation is over and the task given to the subject is to reproduce only the first, second or third line according to the pitch of the tone. A performance in the order of three correct responses was obtained under these conditions. Since the subject was unable to foretell what line he would be asked for to report, it must be acknowledged that immediately after the end of the presentation, the information necessary to recall the letters had to be available somewhere.

This led Sperling to the hypothesis that one form of presentation of the visual stimulation remains accessible for a short time after the presentation. It was Neisser in 1967 who proposed that this representation should be called an icon.

The reason why it is not possible to reproduce nine elements in the case of whole report procedure is that the iconic trace deteriorates while the first elements are reproduced and by the time it takes to have reproduced four or five elements there is no longer any iconic memory available.

The temporary decline of iconic memory has been studied by Sperling in another experiment where the index of selection was not present immediately after the disappearance of the matrix but at different intervals. The results show that in about one second the performance falls to the level which corresponds with whole report procedure.

Nature of coding in iconic memory

1. It is a "pre-categorical" trace, i.e. a relatively literal copy of the pattern of the visual stimulation which has undergone very little perceptual treatment. In one of his experiments Sperling compared two selection criteria for partial recall : selection on the basis of position (one of the three lines) as in the previous experiments, selection on the basis of a semantic category : give the letters or give the figures (in this experiment there was a mixture of letters and figures). With the latter criterion there was no observable superiority of partial over total recall. This result would suggest that information on placement is immediately available in the icon while information on belonging to one of the categories is not and probably will not become so until after perceptual treatment which, having to be carried out on all the elements, cancels the advantage of partial recall.

2. It is a "visual" trace. This statement is based on the finding that the factors which have an effect on the efficiency of iconic storage are generally events limited to the visual domain. Iconic stocking is influenced by factors such as intensity, duration, figure-ground contrast of the images presented, stimulations which precede and follow the stimulation-target. On the other hand, stimulation in other sensory modalities, or mental activities, would have little effect on the characteristics of the icon.