1001 ways to make information non-conscious

Listen to audio

It is likely that "in nature, non-conscious thinking [is] the rule rather than the exception" (Bargh and Morsella, 2008). The psychologist wishing to study non-conscious operations is spoilt for choice: we are unaware of the causes of our behavior, the architecture of our cognitive system, the algorithms we employ, certain perceptual attributes, and even the very presence of certain subliminal or unexpected stimuli.

The 2009 lecture reviewed the major experimental conditions on which current research into non-conscious operations is focusing. Visual illusions offer a vast field in which our perception is vigorously influenced by various cues (contrast, perspective, binocularity, etc.) to which we have no conscious access, although they sometimes conflict with our actions. The bistability of perception, illustrated by Necker's cube or the vase-and-face illusion, underlines the fact that our consciousness has access to only a fraction of the interpretations that our visual system nevertheless calculates in parallel. Among bistability situations, binocular rivalry, in which distinct images are presented to the right and left eyes, is the subject of intensive study. The unperceived image induces adaptation effects and early visual brain activations, demonstrating that it undergoes intense non-conscious processing. Other related paradigms, flash suppression (Wilke, Logothetis & Leopold, 2003) and its continuous variant (Tsuchiya & Koch, 2005) have the great advantage of allowing prolonged presentation of stimuli which, however, remain non-conscious.

Masking is probably the most widely used form of non-conscious visual presentation. It is defined as the reduction in visibility of a briefly presented stimulus (the target) by a second stimulus presented at an adjacent point in space and time (the mask). We distinguish between masking by metacontrast, in which the mask closely follows the contours of the target; masking bypattern, composed of strokes presented in the same place as the target; or masking by substitution, in which the mask is minimal (four points around the target) but has the particularity of extending beyond the target. Different types of masking are also distinguished by the effect of the temporal interval between the target and the mask (known in the English-speaking literature as SOA, for stimulus onset asynchrony). In type A masking, the effect of the mask is maximal when the target and mask appear simultaneously, suggesting an early visual effect. In type B masking, more surprisingly, the effect peaks at a positive interval, i.e. when the mask appears well after the target (up to 50 ms or even 100 ms; backward masking). It is this latter condition that is the focus of many cognitive psychology studies, since it maximizes cognitive processing in the absence of awareness, by giving the masked target a long temporal lead.

In masking, a stimulus is rendered invisible even as the subject devotes his or her full attention to it. We have suggested reserving the use of the term "subliminal" for this situation, where non-consciousness is not linked to attention, but to the limitations of bottom-up information from visual stimulation (Dehaene, Changeux, Naccache, Sackur & Sergent, 2006). Conversely, other experimental conditions rely on the detour of the subject's attention, even though the stimulus itself has sufficient duration and energy to be perceptible - a situation for which we propose to reserve the term "preconscious". In the "inattentionalblindness" paradigm, for example, the subject's attention is diverted from the target by a challenging concurrent task involving other stimuli presented at the same time. It is then possible to present visual stimuli in full fovea, sometimes for several seconds, without the subject being aware of it.

Attentionalblink "( AB) is a fine-grained temporal variant of this competition for attentional resources. During the processing of a first, brief and slightly masked visual target, a second, similar target may not be perceived. Finally, in change blindness, massive changes in an image element go undetected by the subject if the automatic mechanisms for attracting attention are short-circuited, either by a brief flash of all or part of the image, or by the introduction of a very gradual change.

Towards a hierarchy of non-conscious operations

Are all these conditions of non-conscious visual stimulation equivalent? Some experiments suggest not: the invisibility caused by binocular rivalry seems to occur at an earlier stage of processing than that caused by masking. Indeed, binocular rivalry, when affecting the mask, can cancel out the masking effect (Breitmeyer, Koc, Ogmen & Ziegler, 2008). Moreover, binocular rivalry eliminates any subliminal priming effect for animal images, assumed to be processed by the ventral visual pathway, but leaves intact the subliminal priming effect for tools, assumed to be processed by the dorsal visual pathway; whereas masking induces priming effects in both pathways (Almeida, Mahon, Nakayama & Caramazza, 2008).

In a future hierarchy of non-conscious visual operations, which remains largely undefined, the invisibility caused by diverting attention seems to have the most delayed effects. Indeed, analysis of the potentials evoked by a stimulus rendered invisible by attentional blinking shows that all early visual processing, and even the semantic processing of words, remains rigorously intact (Luck, Vogel & Shapiro, 1996; Sergent, Baillet, & Dehaene, 2005).

Based on our theory of the global neural workspace, to which we'll return in the 2010 lecture, my colleagues and I have proposed a preliminary taxonomy of non-conscious processes. This distinguishes between different types of non-conscious mental operations, depending on the way in which information is encoded and the nature of the mechanism that makes this information inaccessible to consciousness (see Dehaene et al., 2006):