When searching for the neural correlates of contents of consciousness (NCCCs), there are several complicating factors. One might a priori have an assumption of modularity, meaning that one feature is encoded in one dedicated NCCC area. The idea that such single, maximally-informative regions exist for different features, however, is no more than an assumption. It might turn out that perceptual coding—even for single features—inherently involves processes in multiple brain regions.
Empirically, it is known that information about objects is distributed across multiple regions. One question is whether one brain region can have information about more than one content (e.g., Haxby et al. 2001; Cichy et al. 2013). In a study inspired by Haxby et al. (2001) we investigated whether object-selective brain regions have information only about objects from their preferred category (Cichy et al. 2013). Participants viewed images from four different categories: objects, visual scenes, body parts, and faces. These categories were chosen because faces, body parts, and places are believed to be processed by highly selective brain regions. We found that a classifier not only contained information about a region’s preferred category: take the example of the face-selective region FFA. It was not only possible to classify faces from this region, it was also possible to classify the difference between other, non-face-related objects, say between a chair and a window. The flipside of this finding that individual regions encode multiple contents is that individual perceptual contents were found in multiple regions. For example, information about faces was also found in supposedly “non-face-selective” brain regions (e.g., in the PPA). This presents a challenge to the idea that each content is represented in one region only.
However, the problem might not be as severe as it first appears. It is actually expected that multiple regions will contain information about each type of content. Different brain regions do not exist in isolation, but are densely causally interconnected (Felleman & van Essen 1991). Furthermore, in the visual pathway, stimulus-related information will reach higher-level brain regions by way of low-level regions. Even if the FFA is the visual region that responds most (albeit not fully) selectively to faces, the presence of a face could also be inferred from the discharge pattern of ganglion cells in the retina. Thus, vertical and horizontal propagation of information is expected. One crucial criterion, which has not received much attention, is whether the information in different regions is redundant or whether it is independent with respect to a person’s perceptual experience. If one hypothetical brain region, say the uniform unicorn area (UUA), is directly responsible for visual experiences of unicorns, it should have more information about a person’s unicorn experiences than any other region.
The relationship between information in the UUA and other areas will reveal a lot about the nature of representation. If other regions also have information about unicorn experiences, and they receive their information about unicorn experiences via the UUA, then the unicorn-related information in the other regions should be partially redundant to that in the UUA. A classifier should not be able to extract more information about a person’s unicorn experiences by additionally taking other regions into account, over and above the information available from the UUA. If, in contrast, other regions have information that goes beyond that in the UUA that allow the system to improve the classification of unicorn experiences, it is likely that the representation itself is distributed across multiple brain regions. Another way to put it is to distinguish between representational and causal entanglement. A change in neural activity in one region will typically be propagated to any neighbouring regions with which it is connected. This causal entanglement, however, does not directly implicate representational entanglement. Only if it were not possible to find an individual region where neural activity patterns is not fully informative of a specific feature, and if taking into account the joint activity of this region and another region did provide full information, would this provide evidence for representational entanglement.