Block's threefold strategy is as follows: (1) He seeks to separate phenomenology from what he calls “access” on the basis of empirical psychology. (2) Shifting to neuroscience, he locates phenomenology in posterior brain regions and access in frontal brain regions. (3) He then concludes that mechanisms of access are not constitutive of mechanisms of phenomenology and claims that this conclusion meshes with both the psychology (1) and the neuroscience (2). However, we contend that the separation of phenomenology and access attempted in (1) fails because it relies on too crude a notion of access. Adopting the more precise notion of broadcast (Shanahan & Baars Reference Shanahan and Baars2005) allows us to preserve the intimate relationship between phenomenology and the global workspace posited by Global Workspace theory (GWT). And we believe the attempt in (2) to isolate “mechanisms of access” from “mechanisms of phenomenology” is also flawed, because it relies on a faulty identification of the global workspace with working memory.

In contemporary elaborations of GWT, the emphasis is on discovering a plausible neurological basis for the hypothesised global workspace – the global neuronal workspace as Dehaene and his colleagues call it (Dehaene et al. Reference Dehaene, Kerszberg and Changeux1998). Dehaene and a number of others suggest that phenomenal consciousness of the visual world requires large-scale re-entrant interaction between posterior visual cortex and frontoparietal regions (see also Baars Reference Baars2002). According to the leading proposals, the neural substrate for the workspace is a network of long-range pathways that interconnects distant cortical sites (Dehaene et al. Reference Dehaene, Changeux, Naccache, Sackur and Sergant2006; Shanahan, in press). A pattern of activation is globally disseminated, or broadcast, when it is propagated by this white matter infrastructure, and thereby comes to exercise widespread influence in the brain. Because the great majority of such pathways are bidirectional, the anatomy suggests widespread re-entrant activity triggered by the initial broadcast. The overall pattern of information flow alternates episodes of broadcast with bursts of competition for workspace access, and the typical duration of discrete episodes of broadcast is on the 100 msec scale.

Using the idea that broadcast implements access, we can supply a more fine-grained account of what might go on in the sorts of experiments that, according to Block, demonstrate “phenomenology without access” (Sperling Reference Sperling1960; Landman et al. Reference Landman, Spekreijse and Lamme2003). It is worth noting that subjects in such experiments believe they are simply reporting the contents of their recent visual consciousness, even after the visual stimulus has disappeared from view. Therefore, subjects' reports do not support a distinction between access and phenomenology.

The theoretical account might go something like this (Shanahan Reference Shanahan2005). Information pertaining to the initial visual object gestalt in its entirety is broadcast at the time of presentation for one or more 100 msec broadcast cycles. As a consequence of broadcast, this information can influence widespread local brain processes involved in memory – both short-term and long-term. Subsequently, following some retrieval cue, a detailed memory of a portion of the original scene is recalled from the relevant part of short-term memory and is broadcast, enabling (with the help of many other brain processes) a verbal report. Thanks to this process of retrieval and report, or perhaps because of subsequent activity, or merely due to the time that has elapsed, the fragile contents of the short-term memory are degraded and a good deal of information about the original visual stimulus is lost. So no exhaustive subsequent reports detailing the shapes and orientations of specific objects in the scene are possible. But post-hoc partial reports can be retrieved with very high accuracy, which is of course why the Sperling procedure is called the “partial report procedure.” Over multiple trials the entire 12-object Sperling array can be accurately reported, always with a reported phenomenal experience of the retrieved item. Moreover, a general impression of the original visual stimulus is retained in short-term memory. If later evoked, this information can still be broadcast, facilitating a verbal report about the overall arrangement of objects in the scene.

The very possibility of such an account shows that, even if they did supply evidence for phenomenology without “access.” in Block's temporally coarse-grained sense of the term, experiments such as those by Sperling and Landman et al. offer no evidence of phenomenology without broadcast, to use the more precise and temporally fine-grained technical term found in the contemporary GWT literature. Admittedly, the theoretical concept of broadcast stands in need of further empirical justification. But it is, in this respect, in no worse a position than that of access.

Now, the above explanation makes little sense unless a clear distinction is maintained between (conscious) presence in the global workspace and (unconscious) retention in short-term memory. But Block incorrectly identifies the global workspace of GWT with working memory, leading to further difficulties with his argument. In humans, presence in the global workspace enables entry to working memory (Baars & Franklin Reference Baars and Franklin2003). But items can then be held in working memory unconsciously. They only become conscious again when they are retrieved from working memory and re-enter the global workspace.

In short, the global workspace and unconscious working memory are quite distinct, and for this reason it does no harm to GWT if it turns out that, as Block suggests, “activation of working memory circuits are not part of the neural basis of phenomenology” (sect. 14, para. 1). What counts for phenomenology, according to contemporary GWT, is broadcast via the long-distance axons alluded to earlier on, and this may or may not necessitate activation of the prefrontal areas associated with working memory.

To summarise, Block's challenge to GWT rests on a faulty view of the concept of the global workspace. GWT is an empirical theory, and as such it is open to both experimental refutation and conceptual critique. So challenges such as this are a welcome opportunity to test the theory, and a chance to revise or discard it if it is found wanting. But to mount a more successful attack on the theory, Block needs to consider more detailed GWT models of broadcast, and clearly to dissociate global workspace activity from unconscious aspects of working memory.