Connectionist Modelling of Subliminal Priming

Connectionist Modelling of Subliminal Priming

Howard Bowman

This research was performed in collaboration with Friederike Schlaghecken of the University of Warwick and Martin Eimer of Birkbeck College University of London.

One of the biggest objectives of cognitive science is to try to understand consciousness, which is of course at the heart of the mind-body problem. There are many questions to be answered. For example, "what is consciousness for and what advantages (in an evolutionary sense) does it afford?"; "how does conscious (phenomenological) experience emerge from the brain?"; and "how much conscious (volitional) control do we have over our actions?".

With regard to the latter of these, there is considerable evidence that we in fact have less control than we might have thought. For example, a classic study by Benjamin Libet [Libet,89] has shown that the conscious experience of intending to act occurs after the preparations to act (in terms of neural activity) have already been set in motion. In addition, a number of studies have demonstrated that visual stimuli that are presented below the threshold of conscious perception, i.e. subliminal stimuli, can still elicit motor response preparation. Typically these studies have been based upon visual masking.

It is well known that a target stimulus that is visible when presented on its own, can be made totally invisible by the subsequent presentation of a non-target stimulus (the mask). Furthermore, it is not essential that the mask stimulus is presented in the same spatial location as the target stimulus. In particular, the presentation of a surrounding ring (an annulus) can prevent perception of a target stimulus. This style of masking is called metacontrast masking. A number of researchers have used these techniques in subliminal priming experiments.

For example, Fehrer and Raab [Fehrer,62], Neumann [Neumann,94] and Vorberg [Vorberg,02] have all collected metacontrast masked priming results, in which response speeds are modulated by stimuli that are metacontrast masked. The results clearly show that motor responses can be initiated by priming stimuli that are not "consciously perceived". This has prompted Neumann's theory of direct parameter specification, which postulates a fast low-level route from perception to action which is not subject to volitional control.

A further twist to the story arises from work by Eimer and Schlaghecken [Eimer,98]. Rather than metacontrast masking, they used a pattern masking paradigm in which subjects respond according to the direction of a target stimulus (either a left or right pointing arrow). However, this stimulus is preceded by a masked prime, which itself is a left or right pointing arrow. In accordance with the direct parameter specification theory, they also discovered that the subliminal primes affected response speeds. However, the direction in which response speeds were affected was somewhat surprising. Specifically, subjects responded more slowly when the prime and target stimuli were compatible (i.e. arrows in the same direction) and more quickly when the prime and target were incompatible. Such negative compatibility effects suggest a role for inhibition in direct parameter specification and could arise from a low-level "emergency breaking mechanism". Thus, once the evidence for a response is removed (as accrues from mask presentation), the corresponding motor action is suppressed.

These results prompt consideration of the computational mechanism that underlies such an inhibitory reversal. Our proposal is that the effect arises from the interplay of response competition (as implemented by lateral inhibition) and a threshold gated direct suppression of strongly activated response nodes. The latter of these is implemented using a mechanism called an opponent circuit, in which each response node is interconnected with an associated opponent node in such a way that the response node excites the opponent node, which in turn relays inhibition back onto the response node. In order to confirm this hypothesis, we developed neural network implementations of this mechanism.

The following articles present early versions of our model,

"A connectionist model of inhibitory processes in motor control and its application to a masked priming task."
H. Bowman, A. Aron, E. Eimer, and F. Schlaghecken.
Technical Report 14-01, Computing Laboratory, University of Kent at Canterbury, November 2001.


"Neural Network Modelling of Inhibition in Visuo-Motor Control."
H. Bowman, A. Aron, E. Eimer and F. Schlaghecken.
Seventh Neural Computation and Psychology Workshop: Connectionist Models of Cognition and Perception.
University of Sussex, Brighton, UK
John A. Bullinaria & Will Lowe (editors)
pages 209-222,
Singapore: World Scientific, 2002.

However, the fullest presentation of our work can be found in the following article,

"A neural network model of inhibitory processing in subliminal priming". H. Bowman, F. Schlaghecken, and M. Eimer. Visual Cognition, 13(4):401-480, February 2006.

I have also given a number of talks on this topic, of which the most complete version is this presentation.


[Eimer,98] Eimer, M. and F. Schlaghecken (1998). "Effects of Masked Stimuli on Motor Activation: Behavioural Electrophysiological Evidence." Journal of Experimental Psychology, Human Perception and Performance, 1998. 24: p. 1737-1747.
[Fehrer,62] Fehrer, E. and D. Raab (1962). "Reaction Time to Stimuli Masked by Metacontrast." Journal of Experimental Psychology 63(2): 143-147.
[Libet,89] Libet, B. (1989). "The timing of a subjective experience" Behavioural and Brain Sciences, 12, pp 183-185.
[Neumann,94] Neumann, O. and W. Klotz (1994). "Motor Response to Nonreportable, Masked Stimuli: Where is the Limit of Direct Parameter Specification." Attention and Performance, XV. C. Umilta and M. Moskovitch. Cambridge, MA, MIT Press: 123-150.
[Vorberg,02] Vorberg, D., et al., (2002) "Invariant Time-course of Priming with and without Awareness." in Psychophysics beyond sensation: Laws and Invariants of Human Cognition, C. Kaernbach, E. Schröger, and H. Müller, Editors. 2002, Erlbaum: Hillsdale, N.J.

Back to Howard Bowman's home page.

Last modified April 2012.