The limits of subliminal treatment

Listen to audio

As the work just reviewed shows, subliminal priming paradigms highlight both the reality of subliminal processing and its limitations. The "proper nature of consciousness" is beginning to emerge from these initial results: as a general rule, subliminal effects are only obtained under conditions of significant automation, they diminish as one progresses up the hierarchy of cognitive operations, and they fade rapidly over time without leading to rapid and lasting changes in behavior.

Is it therefore possible to associate consciousness with one or more cognitive functions that would be specific to it, or even give it a selective advantage that would explain its appearance in the course of species evolution? Anthony Jack and Tim Shallice (2001) propose that a cognitive process "that can only process information if the participant reports being aware of that information" should be called a Type C process. Lionel Naccache and I (2001) have proposed three main sets of candidate functions: the explicit and durable maintenance of information in working memory; intentional and voluntary behavior; and cognitive flexibility in the choice and execution of novel combinations of operations, going against routine strategies.

There is still too little empirical research on these issues. Jacoby and colleagues (1991; Debner & Jacoby, 1994) studied the ability to inhibit a routine response. Their dissociation method compares performance in an "inclusion" task, where the requested response is in the same direction as an automatic mental operation, and in an "exclusion" task, where it runs counter to it. Quantitative analysis shows that, under conditions of non-consciousness or inattention, the exclusion instruction is not followed correctly, and that responses are dominated by non-conscious processing that seems to have little or no controllability. This is one of the few situations, along with the betting method mentioned above (2007), where it has been shown empirically that a seemingly simple task - inhibiting a routine behavior - does not seem feasible when the subject reports being unaware of a piece of information.

Another interesting condition, requiring the concatenation of several cognitive operations, has recently been linked to conscious processing (Sackur & Dehaene, 2009). We have shown that several arithmetic operations, taken individually, can be performed, at least partially, without conscious awareness: even when a masked digit is deemed invisible by participants, addition, subtraction, comparison or naming tasks are performed non-consciously with much better performance than the chance level. However, the sequence of addition followed by comparison (judging whether n + 2 is greater than 5) brings performance down to almost chance level. It could therefore be that consciousness arises at a very specific level of cognitive architecture, where information relevant to the organism's current goals is amplified and temporarily stored in memory for reuse in subsequent calculations. According to this hypothesis, the selection, amplification, retention and use of information, with a view to its integration into a global, non-routine plan of action, would therefore characterize the level of conscious mental operations. This question will be revisited in 2010 when we study the cognitive and cerebral architectures underlying the access of information to consciousness.