Evidence that subliminal is not so 'sub'

In an article in the November 9, 2006, issue of the journal Neuron, published by Cell Press, Kimihiro Nakamura and colleagues report experiments with human volunteers demonstrating such “top-down” processing of subliminal information.

Their findings also shed light on the neural mechanism by which reading a printed word evokes the representation of the spoken form. This “lexical-phonological” linkage is critical to learning to read and is disrupted in some forms of reading disorders.

In the researchers' experiments, they showed volunteers either words or pronounceable nonwords and asked them to perform either a lexical task or a pronunciation task on the words. The lexical task was to identify whether the word they saw was a real word or a nonsense word.

However, unbeknownst to the subjects, they had been first presented with a subliminal word that either matched or didn't match the target word. Such subliminal words were “masked” with nonsense characters that would render the presentation subliminal. The researchers' initial experiments showed that presenting subliminal words identical to the target word produced a “priming” effect in which subjects responded faster on the lexical or pronunciation tasks.

The researchers next applied a harmless magnetic pulse–called transcranial magnetic stimulation (TMS)–to two key regions of the brain involved in such perception, before presenting the priming word. The two regions were known to be involved in either converting visually perceived words to phonological representations or to integrating perceived words across visual and auditory modes. TMS is known to transiently affect neural function in a target area.

Nakamura and colleagues found that TMS applied to one brain area or the other could selectively disrupt the priming effect for either the lexical or pronunciation task. The researchers concluded that the conscious task instruction for either of the tasks caused a different neural network to be engaged for generating the appropriate behavioral response.

They concluded that their results “provide direct evidence for the proposal that even the unconscious processing of incoming stimuli operates under the strong influence of the conscious task instructions.” They also concluded that “results further suggest that such top-down, strategic control modulates the bottom-up neural activation produced by unconsciously perceived words to set up a different neural circuit for generating the intended behavioral response.”

In a preview of the Neuron paper, Stanislas Dehaene and Lionel Naccache commented that “perhaps the most important implications of the Nakamura et al. study concern our concept of automaticity. Many theories of human cognition postulate that nonconscious cognitive processes are automatic and independent of attention. Recently, however, experimental reports using the masked priming paradigm have revealed that subliminal processing is affected by several top-down effects. By showing that repetition priming can be suppressed by applying TMS to distinct locations depending on the task, the present results strongly support this point of view.”

Dehaene and Naccache concluded that the results “support the idea that a whole chain of processing defined by the task, once prepared consciously, can be applied to nonconsciously perceived stimuli. Thus, 'subliminal' is not synonymous with 'automatic' or 'task-independent.' Our expectations shape our processing of subliminal stimuli.”