Attempts to demonstrate unconscious processing of visual stimuli are very old and riddled in controversies, but the controversy does not so much concern the existence of unconscious processing (most researchers seem to be convinced of this), but rather the question of how to demonstrate unconscious processing in a given experiment. If one needs to demonstrate it, one has to make sure that a critical stimulus was completely outside of awareness (the so called zero-awareness criterion). Schmidt (2007) proposes two lines of attack for establishing unconscious processing beyond the zero-awareness criterion. The paper deals with different types of dissociation between measures of awareness and measures of processing.
To demonstrate that a critical stimulus was processed unconsciously, one usually produces some dissociation between different behavioral measures of performance. This is done by comparing two measures obtained from different tasks. One measure (called the direct measure, D) signals the observer's awareness of a critical stimulus. For example, consider a visual perception experiment, where the task is to detect the offset of a vernier (a vernier is simply a set of two vertical lines, one below the other, where the lower line can be offset either to the right or left of the upper line). However, before the vernier is shown (let's say for 25 msec), a prime is flashed for a short period of time (say 15 msec). This prime could be, for example, arrows pointing to the right or left. A forced-choice prime discrimination task ("Was the arrow to the right or left" ) would measure if the observer was aware of the stimulus, and hence would comprise the direct measure. The second measure (called the indirect measure, I) would indicate that the primes themselves are not consciously detected, but they are involved in a priming effect, and hence affect the reaction times of responding to whether the target vernier is offset to the right or left. For example, if the prime and vernier are both congruent (i.e. the arrow points to the right and the vernier is also offset to the right) and if the subject cannot consciously detect the direction of the prime but the reaction times during the congruent cases are always shorter than the reaction times during the incongruent cases (prime pointing to the right but vernier offset to the left), then this would comprise the indirect measure.
Schmidt and Vorberg (2006) examined the assumptions required by the zero-awareness criterion and other approaches. They start by assuming that the direct and indirect measures may depend on two sources of information labeled conscious (c)and unconscious (u). In other words, D = D(c,u) and I = I(c,u), where information is non-negative. The dependency is weakly monotonic, which means that if any type of information increases, the measures can only increase or remain constant. An additional constraint is that if D and I are to be modeled as functions of c and u, then both D and I must be functions of the same underlying conscious and unconscious information. Thus the direct and indirect tasks must be designed to use identical stimuli, identical responses and identical stimulus-response mappings. Schmidt gives an example of this mismatch in the study by Dehaene et. al (1998), where the indirect task was to determine as quickly as possible if a target digit was numerically larger or smaller than 5. The target was preceded by a masked prime digit. The response times were much shorter if the prime was consistent with the target (i.e. both were less than 5 or greater than 5) than when the two were inconsistent (one of them was larger than 5 and the other was smaller). The optimal direct task in this experiment would have been to ask the subject "Was the prime greater than or lesser than 5" because then, both measures would've been tapping into the same source of information (conscious or unconscious). Instead, Dehaene et. al chose two different direct tasks where the subject was asked to detect primes against an empty background, and the second where the subject was asked to discriminate primes from random strings of letter, none of which addressed the critical question of whether the prime was larger or smaller than 5.
Given that D-I mismatch can be avoided, how can the null model of only conscious processing be disproved. The traditional way is the zero-awareness criterion, which produces what is called a simple dissociation of direct and indirect measure (zero D in presence of non-zero I, see Fig. 1a). In other words, I(c,u)> 0, implies that either c > 0, or u > 0, or c and u > 0. But, does, D(c,u) = 0, imply c = 0 and hence I(c,u) > 0 only because u > 0 ? No, because, D is a weakly monotonic function of c. It is possible that under this assumption, c did change but D was not sensitive enough to pick up this change. To resolve this problem, one must make the stronger assumption (see Fig. 1b) that D is a strictly monotonic function of c, which means that D can detect any changes in c, no matter how small. Hence, D(c,u) = 0 implies c = 0, irrespective of whether u is zero or not, since D is weakly monotonic on u anyway. And now, if I(c,u) > o, then, I(0,u) > 0 which finally implies u > 0.
The exhaustiveness assumption is important. In it's absence, one can always argue that it is only conscious processing c, that is occurring and while I is sensitive enough to detect it (I > 0), D is not (D=0). Finally, there is an alternative set of assumptions that abolishes the need for an direct measure altogether. This is when the indirect measure can be assumed to be an exclusive monotonic function of u and is unaffected by c. In this case, I(c,u) = I(u) > 0 implies u > 0 directly (see Fig. 1c)
In the next post, we shall look at double dissociation and beyond
Schmidt, T. (2007). Measuring unconscious cognition: Beyond the zero-awareness criterion Advances in Cognitive Psychology, 3 (1), 275-287 DOI: 10.2478/v10053-008-0030-3