Campana F, Rebello I, Urai AE, Wyart V & Tallon-Baudry C. (2016) Visual consciousness proceeds from global to local content in goal-directed tasks and spontaneous vision. Journal of Neuroscience 36(19).
During my MSc at the ENS in Paris, I did my internship in the laboratory of Dr. Catherine Tallon-Baudry, where I worked on a project about the hierarchical nature of conscious perception. This project, led by Dr. Florence Campana, aimed to experimentally address several predictions made by the Reverse Hierarchy Theory (Hochstein & Ahissar, 2002). It has long been known that visual information is passed through the cortical hierarchy in a bottom-up fashion, with cells in early visual cortex responding to simple visual information such as oriented lines, and downstream cortical areas showing increasingly complex response profiles to eg. faces and scenes. The RHT states that conscious perception of that visual information does not arise with the processing of local details, but rather emerges in higher cortical areas where global information about the outside world is represented. Early behavioural and psychophysics work has shown promising evidence for this idea, but previously used stimuli were never truly hierarchical.
This study thus used specially designed images containing local information (oriented lines) and global orientation (shapes defined by clusters of similarly oriented lines). Importantly, local and global information in those images can be manipulated separately, but the global shape is only defined We first confirmed that these stimuli can only be processed in a bottom-up fashion: the global information can only be accessed by first processing all the local elements (and not, for example, through the low-frequency information contained in the image). We then found that even with this constraint, people consistently report the global information first, either when asked to describe what they see (‘spontaneous vision’) or when forced to choose between different response options (‘goal-directed vision’).
People also did this task in the MEG scanner. We measured their brain activity to identical images, but asked them to respond to either local or global features of those stimuli. We then used a decoding technique (linear support vector machine) to estimate whether specific information is represented in patterns of brain activity. Consistent with the predictions of the RHT, global information could be decoded from neural activity regardless of which feature people were attending to. This suggests that the brain automatically processed global visual information, leading to our coarse but vivid perception of the world (Campana & Tallon-Baudry, 2013). Local information about details in the visual world can only be retrieved with additional attention to those details, possibly instantiated in recurrent loops of neural activity throughout the cortical hierarchy.
In short, we found support for the idea that while visual information is processed from local to global, our conscious perception arises at the level of global, integrated information. This has several implications for theories of consciousness, and for finding out why the world appears to us the way it does.