Attention enables human observers to report and to remember visual information. However, in our experiments, the formation of the memory trace that gives rise to the partial report benefit takes place without focal attention being directed to the items while they are in view. After the items have disappeared and the cue appears, focal attention shifts to the cued item in memory and enables observers to make a large amount of information available for report. This suggests that much more information is processed beyond the scope of focal attention.
The properties of the iconic representation indicate that the items in the scene have undergone more extensive processing than one might expect given that no focal attention was directed at them. We have shown that a cue allows for recovery of multiple features of the same object, indicating that feature binding has taken place, and objects are at least processed up to the level of figure-ground segregation (Landman et al. Reference Landman, Spekreijse and Lamme2003; Reference Landman, Spekreijse and Lamme2004); characteristics that historically were only attributed to attentional – that is, reportable – stages of visual perception. In addition, iconic representations can last up to several seconds, and they are not a mere after-effect (Sligte et al. Reference Sligte, Scholte and Lamme2008). Also, several older studies have already indicated that iconic memory has a spatiotopic component, suggesting that it has undergone more processing than if it were strictly retinotopic (e.g., Breitmeyer et al. Reference Breitmeyer, Kropfl and Julesz1982; McRae et al. Reference McRae, Butler and Popiel1987; for a review, see Cowan Reference Cowan1995).
In daily vision, and in most experiments, when new information enters the visual cortex, the processing of that information gets priority over keeping information about stimuli that are no longer there. In the iconic memory experiment, we limit the entry of new information by showing a blank screen after the stimuli presentation. This allows us to test how much information about the stimuli is available in the visual system, by cueing one of them. If there is a distinction between phenomenal awareness and access awareness, then iconic storage can be seen as a window into the contents of phenomenal awareness. This is difficult to achieve in another way. If we test while the stimuli are still in view, subjects can just get information by looking, or switching attention, whereas if we test once new stimuli have already appeared, the processing of those new stimuli interferes with information we were interested in in the first place.
Iconic memory may be so vulnerable to interference that even the subjects' own response interferes with it. Many classic iconic memory experiments required a verbal response and identification of items. It could be that while the subject was reporting items, the report itself interfered with memory. Our experiments, however, merely required a “yes/no” button press, and change detection instead of identification of the items. Thus, the “report” requirement was much smaller, while we could still get an estimate of the amount of information available. That could be a reason why cues in our experiments are effective up to several seconds after stimulus offset, longer than in a classic iconic memory experiment. Another major difference is that in standard partial report designs, subjects are shown a brief sample display just once before reporting about them, whereas we employed a match-to-sample design. Thus, in our design all items are shown twice; once during encoding and once during report. It seems that memory in general (including other forms of memory, such as long term memory) does better on recognition than on recall.
In favor of Block's proposal, studies so far indicate that iconic memory does not have the key properties of an unconscious process, while it does have properties typical of a conscious process. Unconscious processes (such as masked representations) typically remain inaccessible even when attention is focused on the specific object of interest, whereas iconic memory is accessible when cued. Also, in unconscious processing, even basic features may not be processed to the level of perception. Recent evidence shows that unconscious color-priming effects are more dependent on physical stimulus properties than on perceptual properties (Breitmeyer et al. Reference Breitmeyer, Ro and Singhal2004). In iconic memory, however, not only have features been resolved, but also figure-ground segregation and feature binding have already taken place, properties that are more associated with conscious processes.
Additional ways to find out whether iconic memory has properties in common with conscious processes include neurophysiological studies. In neurophysiology, it is possible to distinguish between feed-forward processing, and recurrent processing (Lamme & Roelfsema Reference Lamme and Roelfsema2000). By selectively disrupting recurrent processing (RP), but leaving feed-forward processing intact, it has been observed that visual awareness does not arise. This was shown by backward masking (Lamme et al. Reference Lamme, Zipser and Spekreijse2002), by applying transcranial magnetic stimulation to the primary visual cortex (Jolij & Lamme Reference Jolij and Lamme2005; Pascual-Leone & Walsh Reference Pascual-Leone and Walsh2001), and by inactivating higher visual areas (Hupe et al. Reference Hupe, James, Payne, Lomber, Girard and Bullier1998; Lamme et al. Reference Lamme, Zipser and Spekreijse1998). Even when there are sudden lapses in awareness, it is observed that RP is absent, whereas feed-forward processing is intact (Supèr et al. Reference Supèr, Spekreijse and Lamme2001b). Many scholars agree that RP is likely to be involved in conscious perception. Current and future experiments are necessary to address what the exact mechanism behind iconic memory is, in which areas of the brain it occurs, and whether RP is involved. However, if there is RP, depending on how widespread it is, views will still differ on whether it looks more like a conscious process or more like an unconscious process. Some argue that RP within the visual cortices is sufficient for conscious perception to arise (Block Reference Block2005; Lamme Reference Lamme2003; Reference Lamme2006), while others maintain that consciousness requires more widespread recurrency, including areas involved in cognitive access and control, such as the prefrontal cortex (Dehaene et al. Reference Dehaene, Changeux, Naccache, Sackur and Sergant2006).
It is important to test whether RP is involved in iconic memory. Preliminary data provides some indirect evidence for this, by showing a cortical origin and a long-lasting, reverberating nature. Still, more direct evidence is required to determine whether RP takes place and whether iconic processing is necessary and sufficient for visual awareness to occur.
