Haber does not dispute the basic facts of iconic memory first discovered by Sperling. The two basic facts of iconic memory are as follows. First, subjects can recall a greater proportion of a stimulus display when given partial report instructions. And second, partial report superiority declines as cue delay increases. Recall performance in partial- and full-report is about the same with cue delays of longer than .25 seconds, although the icon persists somewhat longer under special circumstances (e.g., dark pre-exposure field).
Methodological and empirical debates have refined the concept of iconic storage. But, according to Haber, the validity of the construct in the context of information processing models is seldom questioned. He states, "What has been surprising is the largely uncritical acceptance of the concept of the icon itself as an initial store of information early in the chain of information processing" (p. 3). Furthermore, those who have questioned the utility of the concept "have largely been ignored" (p. 3).
Its unquestioned validity notwithstanding, Haber contends that iconic memory has little relevance to actual visual functioning because normal perception is "not made of brief discrete flashes, singly or in combination" (p. 3). Haber bases his argument on a consideration of the "natural ecology of vison." For example, he asks why it is necessary to posit an internal visual storage mechanism when objects in the real visual world provide continuous stimulation. That is, "the stimulus maintains itself, and without fading" (p. 4). Storage, he argues, is provided by the continuous visual scene. Further, when moving, continuous visual transformation is valuable information. Frozen images of the visual scene do not make sense in this context. Indeed, Haber suggests that only in extreme circumstances, such as reading in a lightening storm, does the notion of iconic persistence have any relevance. But what of the experimental evidence for iconic storage?
Parallel information encoding is one property of iconic storage. That is, visual stimulation results in an instantaneous imprint or snapshot, which fades after about a quarter of a second. But according to Haber, this supposition is not consistent with the evidence. He states, "for a stimulus spread out over the retinal surface, its information is available to more central processes in a temporally distributed manner, not simultaneously" (p. 6). This view implies that, while visual stimulation from an object is continuous, the nature of information extracted from that object changes with time. This view is inconsistent with a picture conception of iconic memory. Haber notes that Coltheart’s demonstration of the dissociation between iconic memory and visible persistence has this same implication. He suggests, however, that Coltheart fails to grasp this implication: "It takes but a small leap to conclude that something must be fundamentally wrong with the iconic concept itself—a leap Coltheart does not contemplate at all" (p. 6).
Masking is said to occur in iconic memory when a post-exposure distractor stimulus "overwrites" the to-be-recalled stimulus. Averbach and Coriell demonstrated such an effect by cueing recall with a circle surrounding the spatial location where the stimulus had appeared. Extended to normal vision, successive fixations mask each other, and therefore "Maskability of persisting excitation reduced the potential usefulness of an icon by limiting its duration, perhaps to nothing" (p. 7).
Do moving stimuli produce icons? Haber argues that, although studies of iconic memory are generally limited to static presentations of visual stimuli, the partial report technique is in principle extendable to moving stimuli. An experiment by Treisman (1975) illustrates. Subjects were shown a display consisting of dots which moved around separate visual pathways for 100 ms. The task for subjects was to report the directions in which the dots were moving. Whole and partial report were contrasted. Treisman found evidence for partial report superiority. Her argument was that a dynamic icon has to be represented dynamically, since information about vision is processed centrally. A static icon is therefore still implicated.
Where is the icon located?
Haber claims that a retinal locus for the visual icon is more damaging to the notion of iconic memory than is a neural locus because "it is specifically this freezing that cannot be helpful in normal viewing of the visual world" (p. 8). A related question is whether iconic memory is centered in space or on the retina. Haber concludes the latter because in spite of movements and displacement of images on the retina, "what we perceive is an integrated, continuous, clear panorama of the scene" (p. 8). Another experiment by Jonides et al. provides evidence for a spatially centered icon, as well. In successive presentations, subjects were shown patterns of dots which, if superimposed, filled 24 of 25 cells in a 5 x 5 matrix. The task was to report the coordinates of the empty cell. In one condition, the successive stimuli appeared in separate locations on the screen. Performance levels were low. In another condition, subjects were instructed to move their eyes to the first stimuli as soon as it appeared. Then, the second stimulus appeared. Therefore, the stimuli appeared in different retinal locations, but in the same spatial location. Performance improved. Haber interprets this finding as follows: "each flash of dots was represented from the very beginning in a set of coordinates defined by the screen, irrespective of where the subject’s eyes may have been directed at the instant the flash occurred" (p. 9).
Was pursuit of the icon wrong?
Haber remarks that, while the facts of iconic memory stand, too much has been made of them: "the icon is not merely a descriptive term for a particular phenomenon but a theoretical notion bristling with implied hypotheses" (p. 9). Too much significance has been attached to an interesting phenomenon. He contends that normal vision can be understood only by simulation normal vision conditions. That is, "when building theories or models about ecologically valid behavior (the behavior of typical people performing their natural tasks in their normal environments), we have to scrutinize carefully the generality and applicability of the data used to create the theory" (p. 10).
Why, then, has so much significance been attached to the icon? First, the notion of echoic memory—which may well be valid given the temporal nature of auditory stimulation—was extended to vision. However, this extension is misguided given that auditory stimulation is distributed temporally, while spatial information is distributed spatially. Second, vision researchers have relied to strongly on tachistoscopic stimulus presentation, and not enough on continuous visual stimulation. Third, theorists have attempted to explain visual processing in terms of stage models. The concept of iconic memory fit nicely into this framework. Finally, the notion of an icon is rooted in the "insidious eye-as-a-camera metaphor." Haber concludes that "The icon as a theoretical entity has been a 20-year mistake" (p. 11) and that theorists have tossed the term around without realizing that it has little relevance to information processing.
Haber dismisses iconic storage as the first stage of visual information processing by appealing to a description of vision in the natural ecology. For example, why is a brief storage of visual information necessary when visual stimulation is continuous? Furthermore, perception of motion is dependent on changing visual stimulation. The idea of a static visual image is antithetical to this requirement.
Haber’s argument was countered from a number of angles. For example, Coltheart’s analysis, reviewed earlier, showed that stimulus duration has no effect on iconic storage. The implication of this is that continuous visual stimulation is irrelevant for visual functioning. Similarly, Inhoff et al. (cited in Ashcraft) found that after subjects had fixated on a word for 50 ms, replacing it with another word had no effect on performance. This suggests that visual information may, in fact, be gathered through quick snapshots, as proponents of iconic storage suggest. Ashcraft concludes, "Thus Haber’s point about the environment—that we can continuously sample information from it—may in fact be irrelevant to the way the eye actually extracts visual information" (p. 98).