Studies examining unconscious perception usually rely on the experimental logic known as the dissociation paradigm (Reingold & Merikle, 1990), whereby the presence of subliminal perception is supported by a dissociation between two measures of perception. One measure is assumed to provide an index of subliminal perception whereas the other is assumed to index conscious awareness. The presence of subliminal processing is supported when changes in a critical stimulus effect the measure of subliminal processing, but the observer is unaware of either the presence or the changes in the critical stimulus.
For example, in the widely used masked priming paradigm, a lower case word is briefly presented (40 miliseconds) and replaced by a target word in capitals. Although the first word is rarely consciously perceived (measure of consciousness), recognition of the second word is faster when the two words are identical (e.g., star -> STAR), compared to when they are different (cope -> STAR). This pattern of results suggests that the first word is processed subliminally, and that this processing produces savings in how long it takes to process the target word.
Measures of conscious awareness are divided broadly into subjective and objective measures. Subjective measures are typically verbal reports of the observers' perceptual awareness during or after the task. Subjective reports can either be obtained with a direct question, or through a series of graded questions that gradually probe more specific aspects of an observers awareness. An experimenter might ask participants at the end of an experiment whether they were aware of any briefly presented stimuli, whether they noticed any patterns in the stimuli, or whether they developed particular strategies throughout the task. Objective measures are obtained through tasks and provide a numerical estimate of whether an observer was aware of a stimulus or implicit structure of a task. For example, observers might be asked to make a forced choice decision about a briefly presented stimulus to assess how visible the materials are.
Subjective measures are generally regarded to be unreliable, as participants may indicate that they are aware of a stimulus only when they are highly confident. This means that participants may neglect to mention instances where awareness was only partial. Another problem faced with subjective measures is that direct questioning during an experiment may draw participants attention toward the subliminally presented materials. Alternatively, participants may forget that they were aware of the critical stimulus if consciousness is assessed at the end of the testing session.
Although seen as more reliable, objective measures are associated with a number of limitations. Reingold and Merikle (1990) argue that a good measure of consciousness should be both exhaustive and exclusive. That is, the measure should index all components of conscious awareness (exhaustive), and should not index unconscious processes (exclusive). It is unclear whether objective measures are able to satisfy both of these criteria, as objective measures may only index a subset of the contents of consciousness (depending on the task instructions), and may also tap into unconscious processing. Another practical issue concerns how a lack of consciousness is estimated. Support for a lack of consciousness is obtained by showing that objective performance is at chance (e.g., guessing). This criterion is undesirable as this involves accepting the null hypothesis that participants were not aware, instead of rejecting the null hypothesis that participants were aware. Moreover, it cannot be guaranteed that participants will maintain motivation during a difficult task that consists mostly of guessing. In these cases participants may 'give up' and simply press buttons without considering the stimuli.
E. Reingold, P. Merikle. (1990). Using direct and indirect measures to study perception without awareness, Perception & Psychophysics, 44, pp. 563–575