Draft of: Merikle, P. M., & Reingold, E. M. (1998). On demonstrating unconscious perception. Journal of Experimental Psychology: General, 127, 304-310.
On Demonstrating Unconscious Perception
Philip M. Merikle & Eyal M. Reingold
Draine and Greenwald (in press) describe a methodology based on regression analysis for demonstrating unconscious perception. They suggest that their methodology represents a major improvement over existing methodologies. An analysis of their methodology reveals that it is closely related to the classic dissociation paradigm. As such, interpretation of their results is compromised by the same issues concerning the measurement of awareness that have plagued all previous attempts to use the dissociation paradigm to demonstrate unconscious perception in the complete absence of conscious perception.
On Demonstrating Unconscious Perception
The experiments described by Draine and Greenwald (in press) address one of the oldest research questions in experimental psychology. Namely, do we perceive meaningful information even when we have no awareness of perceiving? This question has been the subject of hundreds of empirical studies, dating from the very beginning of experimental psychology in North America (e.g., Peirce & Jastrow, 1884; Sidis, 1898). On the basis of these studies, there is now a growing consensus that perception in the absence of the subjective experience of perceiving is a genuine phenomenon (e.g., Hirst, 1995; Kihlstrom, 1987; Merikle, 1992). However, despite this growing consensus, Draine and Greenwald's paper is proof that it is still possible to question the very existence of unconscious perception as a valid psychological phenomenon.
The primary question that we address in this commentary is does the methodology described by Draine and Greenwald (in press) represent a major improvement over other methodologies used to study perception without awareness? Their methodology is based on a mixture of ideas taken from both the classic dissociation paradigm and the relative sensitivity approach described previously by Reingold and Merikle (1988). Draine and Greenwald suggest that their methodology is better than existing methodologies. We disagree. Rather, we suggest that their methodology is closely related to the classic dissociation paradigm. As such, it has many of the weaknesses of the classic dissociation paradigm and none of the advantages of the relative sensitivity approach. For these reasons, we do not believe that Draine and Greenwald's methodology represents a significant advance over the other methods that are available for studying unconscious perception. In fact, we suggest that rigorous application of Draine and Greenwald's methodology could lead to misleading conclusions.
The Classic Dissociation Paradigm
The vast majority of studies directed at demonstrating perception without awareness have used a variant of the dissociation paradigm. The primary goal of the experiments based on the dissociation paradigm has been to demonstrate that it is possible to perceive stimuli in the complete absence of any conscious awareness of the these stimuli. To demonstrate perception in the absence of awareness, it is necessary 1) to identify a measure of conscious perception, 2) to show that this measure of perception indicates that the stimuli were not perceived, and 3) to find a second measure of perception which indicates that the stimuli were perceived under the same conditions that the measure of conscious perception indicated that the stimuli were not perceived.
One of the best known applications of the dissociation paradigm is Marcel's (1983) classic experiment in which he demonstrated semantic priming in the absence of stimulus detection. Marcel's experiment illustrates the three major components of any demonstration of perception without awareness based on the dissociation paradigm. First, he used stimulus detection as the measure of conscious perception. Second, he established the experimental conditions that made it impossible for the participants to discriminate at a better than chance level of accuracy between the presence or absence of the stimuli. Third, he used a measure of semantic priming based on reaction time to show that despite the participants inability to detect the stimuli, it was still possible for the stimuli to prime subsequent decisions regarding other stimuli.
Findings such as those reported by Marcel (1983) have considerable intuitive appeal. The fact that the participants could not distinguish between the presence or absence of the stimuli certainly suggests that they were completely unaware of these stimuli. Given that the same stimuli that the participants were unable to detect could nevertheless influence how rapidly they could make subsequent decisions regarding other stimuli, it is not surprising that Marcel and many other investigators reached the conclusion that the stimuli were perceived without awareness.
However, even though the logic underlying the dissociation paradigm is straightforward, interpretation of the results of the experiments based on the dissociation paradigm has not been straightforward. There are two interrelated issues that have clouded interpretation. Both issues concern the measure of conscious perception, which must be adequate before any dissociation between measures can be considered a demonstration of unconscious perception.
Is the measure of conscious perception valid? In any study of unconscious perception based on the dissociation paradigm, the most important issue concerns whether the measure of conscious perception has assessed ALL relevant consciously perceived information. Before the dissociation paradigm can be used successfully, a direct measure of conscious perception that assesses ALL relevant conscious perceptual experiences must be identified. Is it possible to find such a measure? We do not think so. There just doesn't seem to be any basis to assume a priori that any particular direct measure of perception is an exhaustive measure of ALL relevant conscious experiences. No matter what direct measure of perception is selected, there is just no way to demonstrate conclusively that this measure is a completely satisfactory measure of ALL relevant conscious experience. Although measures such as stimulus detection have considerable intuitive appeal, in the final analysis, the argument favoring a particular direct measure always comes down to a plausibility argument. As indicated by Draine and Greenwald (in press) in reference to the direct measures that they used to assess conscious perception, "although the assumption of exhaustive sensitivity to conscious processing is plausibly valid for the direct measures, the present experiments provide no empirical confirmation of this assumption" (p. 41).
Does the measure of conscious perception show null sensitivity? A second issue that must be addressed whenever the dissociation paradigm is used to demonstrate unconscious perception concerns whether the procedures were adequate to demonstrate that the measure of conscious perception exhibits a true null sensitivity (cf., Macmillan, 1986). Before any dissociation between measures can be considered as evidence for unconscious perception, the measure of conscious perception must indicate that the participants did not perceive the stimuli. Otherwise, it is always possible to argue that conscious perception, albeit minimal conscious perception, may have contributed to the observed findings. But how is it possible to prove conclusively that any measure of perception exhibits a true null sensitivity? Whenever the results of an experiment suggests that a particular measure exhibits no sensitivity, it is always possible to argue that the methodology may have been inadequate to reveal the true sensitivity of the measure. If null effects are always open to such alternative interpretations, then it may be impossible to ever demonstrate in a completely convincing manner that a measure of conscious perception exhibits a true null sensitivity.
The Relative Sensitivity Approach
Several years ago, we suggested an alternative to the dissociation paradigm for demonstrating unconscious influences (e.g., Reingold & Merikle, 1988, 1990). This alternative approach is based on comparisons of the relative sensitivity of comparable direct and indirect measures of perception or memory. Our goal in suggesting this alternative approach was to describe a method for demonstrating unconscious processes in perception, learning, and memory that did not require either a) a measure of ALL relevant conscious perceptual experiences or b) a demonstration that the measure of conscious perception exhibited a true null sensitivity.
Before defining what is meant by the phrase comparable direct and indirect measures or discussing the conceptual and methodological advantages of the relative sensitivity approach, it is useful to describe an experiment that illustrates the approach. An excellent example is the now classic study by Kunst-Wilson and Zajonc (1980). In this study, participants were initially shown 10 irregular geometric shapes. Each shape was presented five times for a 1 ms duration, and no participant ever reported seeing any of the shapes. Following these initial presentations, memory for the shapes was assessed by both a recognition test and a preference test. For both tests, the participants were shown 10 pairs of shapes, with each pair consisting of one old and one new shape. For the recognition test, the participants were explicitly instructed to select the member of each pair that had been presented previously; for the preference test, the participants were simply told to select the shape that they preferred, with no reference being made to the distinction between old ands new shapes. The results of the study indicated that the participants selected the previously presented shape in approximately 60% of the pairs when they selected the shape that they preferred, but when they selected the shape that they thought had been presented previously, their performance did not differ from chance (i.e., 50%).
The Kunst-Wilson and Zajonc study provides compelling evidence of unconscious memory for briefly presented stimuli. In fact, intuitively it seems that the Kunst-Wilson and Zajonc findings are a more convincing demonstration of an unconscious influence than are the results of many of the experiments based on the dissociation paradigm. What makes their findings so compelling is that Kunst-Wilson and Zajonc used comparable direct (recognition) and indirect (preference) measures. Both measures were based on 1) the same study conditions, 2) the same test conditions, 3) the same pairs of stimuli, and 4) the same response metric. In fact, the only difference between measures was the instructions. For the recognition test, the participants were given direct instructions and explicitly told to select the one shape in each pair that had been presented previously. In contrast, for the preference test, the participants were given indirect instructions and explicitly told to select the shape in each pair that they preferred, with no reference being made in the instructions to the fact that one shape in each pair had been presented previously. Despite the absence of any reference in the indirect instructions to the old/new discrimination, the indirect measure (i.e., preference) was a more sensitive measure of the old/new discrimination than the direct measure (recognition).
Comparisons of the relative sensitivity of comparable direct and indirect measures have an important conceptual advantage over approaches based on the dissociation paradigm. Namely, a much more minimal assumption is required in order to interpret greater sensitivity for an indirect than a direct measure as evidence for unconscious perception. All that needs to be assumed is that the sensitivity of the direct measure to consciously-perceived, task-relevant information is equal to or greater than the sensitivity of the indirect measure to the same information (Reingold & Merikle, 1988, 1990). The rationale underlying this assumption is that consciously perceived information relevant to a particular discrimination (e.g., an old/new judgment), if it exists, should be used equally or more efficiently when participants are instructed explicitly to make the discrimination (i.e., a direct measure) than when participants are instructed to judge the stimuli on some other basis (i.e., an indirect measure).
Based on this minimal assumption, an unconscious influence is implicated whenever an indirect measure is more sensitive than a comparable direct measure to the same perceptual discrimination. This is the case because the underlying assumption rules out the possibility that superior performance on the indirect task is attributable to the perception of conscious, task-relevant information. Therefore, by default, whenever an indirect measure is more sensitive than a comparable direct measure, it must reflect the influence of unconscious, task-relevant information. Of course, this interpretation is only warranted if the direct and indirect measures are truly comparable except for the reference in the direct-task instructions to the discrimination of interest. If the measures are not comparable, then greater sensitivity for an indirect than a direct measure may simply reflect a methodological difference between the direct and indirect measures rather than an unconscious influence.
The minimal assumption underlying the relative sensitivity approach gives it an important methodological advantage over approaches based on the dissociation paradigm. Given that inferences regarding the presence or absence of an unconscious process are based on comparisons of the sensitivity of comparable direct and indirect measures, there is no need to demonstrate that a direct measure exhibits null sensitivity, as is the case for an approach based on the dissociation paradigm. The only critical experimental finding that is necessary to infer the presence of an unconscious process is greater sensitivity for an indirect than for a comparable direct measure. Thus, with the relative sensitivity approach, the presence of an unconscious process is implied whenever the indirect measure is more sensitive than a comparable direct measure, and this is true even in situations where the overall accuracy for the direct measure may be above chance.
The Indirect-Without-Direct Data Pattern
Draine and Greenwald's (in press) primary goal is to demonstrate ". . . semantic activation in the absence of conscious cognition." (P. 6). They believe that the ". . . empirical evidence for unconscious perception remains controversial" (p. 3) because it has been difficult to develop convincing experimental demonstrations ". . . showing statistically significant indirect effects for stimuli that produce no evidence for direct effects" (p. 4). Thus, their primary goal is to show that an indirect measure of perception can indicate that a stimulus was perceived even when a direct measure of perception shows absolutely no sensitivity to the same stimulus. This goal is precisely the goal of all studies based on the classic dissociation paradigm. It is not the goal of studies based on the relative sensitivity approach because such studies only require an indirect-greater-than-direct rather than an indirect-without-direct data pattern to demonstrate unconscious perception.
Draine and Greenwald state that it has proven difficult over the years to develop completely convincing demonstrations of the indirect-without-direct data pattern. This is only true if their assumption regarding what constitutes an acceptable measure of awareness is accepted. Draine and Greenwald suggest that the only completely acceptable way to measure awareness is to use objective measures of perceptual discriminations. If Draine and Greenwald's position is accepted, then it certainly has proven difficult to demonstrate the indirect-without-direct data pattern. However, if it is assumed that measures of the subjective experience of perceiving are appropriate direct measures of awareness (e.g., Chalmers, 1996; Merikle, 1992; Weiskrantz, 1986), then there is a considerable body of evidence that has been accumulating for more than 100 years that demonstrates the critical indirect-without-direct data pattern (see Adams, 1957; Miller, 1942 for reviews of the early studies).
Given Draine and Greenwald's belief that a convincing demonstration of the indirect-without-direct data pattern must be based on studies involving objective, direct measures of perceptual discriminations, their preferred method for assessing the presence or absence of awareness is a measure of perceptual sensitivity (i.e., d') based on a two-alternative, force-choice task. The strong assumption implied by their approach is that there is a complete absence of ALL relevant conscious perceptual experiences whenever d' = 0. It is unclear whether they also believe that a d' greater than zero is an indication that relevant information was consciously perceived, or whether they believe that a d' greater than zero can at times be an indication that the relevant information was unconsciously perceived.
Draine and Greenwald recognize that two issues must be addressed and resolved before it is possible to demonstrate that an indirect-without-direct data pattern provides convincing evidence for perception in the complete absence of awareness. First, the direct and indirect measures must be comparable. Second, there must be a complete dissociation between the direct and indirect measures; the direct measures must have zero or null sensitivity to the same perceptual discrimination that the indirect measures are sensitive to. In the following sections, we discuss the success of Draine and Greenwald's experiments in dealing with these issues.
Are the direct and indirect measures comparable? In justifying the comparability of their direct and indirect measures, Draine and Greenwald (in press) pay considerable attention to Reingold and Merikle's (1988) previous analysis of direct and indirect measures. According to Reingold and Merikle, three criteria must be satisfied before either the indirect-without-direct data pattern or the indirect-greater-than-direct data pattern provide strong evidence for unconscious perception. These criteria are that the direct and the indirect measures 1) must be based on the same response metric; 2) must index the same perceptual discrimination, and 3) must be administered under the same study and test conditions.
In Experiments 1, 2, and 3, the direct and indirect measures satisfy the first criterion in that both measures are based on the same response metric (i.e., d'). However, the measures used in Experiments 1, 2, and 3 do not satisfy the second and third criteria. With reference to the second criterion, the indirect and direct measures indexed different perceptual discriminations; the indirect measures indexed the participants' ability to discriminate between semantic categories (i.e., male/female or pleasant/unpleasant), whereas the direct measures indexed the participants ability to discriminate between words and a string of alternating Xs and Gs (i.e., the XGvWd task). With reference to the third criterion, the indirect and direct measures were assessed under considerably different conditions. The indirect measures involved the response-window procedure which required the participants to respond within a specific, experimenter-defined, temporal interval. In contrast, with the direct measures, the participants were more or less free to respond anytime following the onset of the target stimulus.
Draine and Greenwald recognize the interpretive problems that arise whenever the direct and indirect measures are not comparable. They suggest that ". . . subjects might have approached the XGvWd task with the strategy of attending to physical information of the primes in order to identify Xs or Gs (rather than trying to read words)" (p. 26). If the participants did adopt this strategy when performing the direct task, then "this strategy could have reduced conscious perception of semantic information during the direct measure task compared to the strategy required by the indirect measure task of attending to the meaning of the target stimuli" (p. 26). In other words, given that the direct and indirect tasks used in Experiments 1, 2, and 3 were not comparable, Draine and Greenwald suggest that the results of these experiments may not necessarily provide evidence for unconscious perception.
To overcome the interpretive problems that compromised the results of Experiments 1, 2, and 3, Draine and Greenwald introduced a second direct measure in Experiment 4. Their goal was to have a direct measure that was more similar to the indirect measure. This second direct measure, the direct semantic classification (DSC) task, required participants to classify the primes in terms of their semantic category (i.e., pleasant vs. unpleasant or male vs. female). Given that the DSC task indexed the same perceptual discrimination as the indirect task (Criterion 2) and was based on the same response metric as the indirect task (Criterion 1, it was certainly better matched to the indirect task than any of the other direct measures.
However, there are still some very important differences in the testing conditions used for the DSC and the indirect tasks (Criterion 3). Once again, the indirect task was administered using the response-window procedure, whereas the DSC task was administered under conditions that did not require the participants to respond within an experimenter-defined, temporal interval. In addition, with the indirect task, the participants were required to classify clearly-visible target words, whereas with the DSC task, the participants were required to classify difficult-to-perceive primes in the presence of a clearly-visible, but irrelevant target words. Draine and Greenwald assumed that the participants were able to follow the DSC task instructions to ignore the clearly visible target words and to direct attention completely to the task of trying to classify the primes.
The results of Experiment 4 show that the participants could not perform the DSC task. Their performance on the task did not differ from chance (i.e., d' = 0) whether the primes were presented for 17 ms or 33 ms (see Figure 6). If this inability of the participants to classify the primes indicates that the primes is assumed to indicate an absences of conscious perception, then the observed priming revealed by the indirect measure would reflect unconscious semantic priming. However, it is also possible that other factors may have made it impossible for the participants to discriminate the semantic category of the primes when they performed the DSC task. For example, given that the presentation of each prime was followed immediately by the presentation of a clearly visible target word, it is very possible that the target word captured attention so that it was impossible to attend to semantic category of the primes. In fact, the experimental conditions used for the DSC task are very similar to the experimental conditions that are known to lead to conceptual masking (e.g., Intraub, 1984; Loftus, Hanna, & Lester, 1988). If conceptual masking did occur, then the participants would have found it difficult if not impossible to attend to the prime classification task once their attention was captured by a clearly visible target word.
Another reason why the DSC task may have failed to reveal any evidence that the primes were perceived is that the primes were ineffective by the time that the participants responded. From Greenwald, Draine, and Abrams' (1996) recent experiments, we know that evidence for semantic priming is only revealed by the response-window procedure when the temporal interval separating the onsets of the primes and targets is less than 100 ms. If the effectiveness of primes dissipates within 100 ms, then it is not surprising that the DSC task revealed no evidence that the primes were perceived. Given that the participants were not required to respond within a short response window when they performed this task, their more leisurely responses may have been made beyond the temporal interval when the primes were effective.
The fact that there are alternative interpretations of the participants inability to perform the DSC task illustrates the types of interpretive issues that can arise whenever the experimental conditions for administering direct and indirect tasks are not completely matched. If the direct and indirect tasks had been completely matched in terms of all three criteria described by Reingold and Merikle (1988), then these interpretive ambiguities would not have arisen and any indirect-greater-than-direct data pattern would have been strong evidence for unconscious perception. However, given the interpretive ambiguities, the results of all four experiments reported by Draine and Greenwald seem consistent with any of the following interpretations.
1) Both the direct and the indirect measures assess unconscious perception. The differential sensitivity of the measures indicates that each measure is maximally sensitive to different characteristics of the primes.
2) Both the direct and the indirect measures assess conscious perception. The differential sensitivity of the measures indicates that each measure is maximally sensitive to different characteristics of the primes.
3) The direct and indirect measures are sensitive to both consciously and unconsciously perceived information. The differential sensitivity of the measures indicates that each measure is maximally sensitive to different characteristics of the primes and that each measure indexes a somewhat different combination of consciously and unconsciously perceived information.
4) The direct measures index ALL relevant consciously perceived information regarding the primes. The indirect measures index unconsciously perceived information regarding the primes.
Given the methodology used by Draine and Greenwald, we do not see any straightforward way to disentangle these alternative interpretations of the indirect-without-direct data patterns found in their experiments.
Does the regression analysis demonstrate perception in the complete absence of conscious awareness? To establish the critical indirect-without-direct data pattern, Draine and Greenwald (in press) used a regression analysis in each experiment to determine if there was a statistically significant intercept for the indirect measure when the direct measure was used as the predictor variable. In effect, each regression analysis created a statistical participant for purposes of evaluating the critical indirect-without-direct data pattern. The logic underlying the regression strategy is straightforward; a significant intercept indicates that the sensitivity of the indirect measure is greater than zero when the sensitivity for the direct measure is equal to zero. Thus, a significant intercept for the indirect measure constitutes a demonstration of the critical indirect-without-direct data pattern. As such, it is considered to be strong evidence that the stimuli were perceived in the complete absence of any conscious awareness (cf., Greenwald, Klinger, & Schuh, 1995). However, before the regression strategy can be informative, two conditions must be met.
First, as for all other applications of the dissociation paradigm, the direct measures chosen to index conscious perception must be valid measures of ALL relevant consciously perceived information. If the direct measures are not completely satisfactory, then the regression strategy, even though it is a rather clever method for demonstrating the indirect-without-direct data pattern, will not be any more informative than any other approach based on the dissociation paradigm. We have already discussed a number of reasons why the direct tasks used by Draine and Greenwald may not be completely satisfactory. Given the questions that can be raised regarding these tasks, we do not believe that Draine and Greenwald's evidence for unconscious perception is anymore compelling than the findings from many other studies that have established dissociations between direct and indirect measures of perception
The second condition that must be met before the regression strategy can be informative is that there must be some degree of association or correlation between the direct and indirect measures. In fact, if the direct and indirect measures are not correlated to any great extent, then there is really no reason to compute a regression analysis. In the absence of a correlation between the direct and indirect measures, a regression analysis will simply reveal that the intercept for the indirect measure approximates the overall sensitivity of the indirect measure.
A review of Draine and Greenwald's figures shows that in every experiment the overall sensitivity of the indirect measure (i.e., the mean in the upper left of each figure) is approximately the same size as the intercept in the regression equation. Given that the slopes in 17 of the 18 regression analyses shown in the figures were not significant, it is not surprising that the overall sensitivity of the indirect measures is an excellent predictor of the intercepts. The small and mostly nonsignificant slopes found in the regression analyses indicate that the correlations between the direct and indirect measures are also small and most likely nonsignificant. Thus, in the absence of significant correlations between the direct and indirect measures, the regression analyses do not add any new information. Under these conditions, the overall sensitivity of an indirect measure will always be a good predictor of the intercept for the indirect measure.
Another consequence of the small and nonsignificant correlations between the direct and indirect measures is that the statistical participants created by the regression analyses are not representative of the majority of the participants who were tested in each experiment. A review of the scatterplots presented in the figures reveals that there is no consistent relation across participants between performance on the direct measures and performance on the indirect measures. In fact, in each experiment, there is only a small subgroup of participants who show the critical indirect-without-direct data pattern. The remaining participants show either a direct-greater-than-indirect data pattern or roughly equivalent performance on the two measures. Thus, the majority of participants in each experiment did not show the critical indirect-without-direct data pattern. In other words, the majority of participants did not show evidence of perception when the direct measure of conscious perception exhibited null sensitivity. Rather, in each experiment, the majority of the participants showed data patterns that are inconclusive with regard to the question of whether the primes were perceived with or without awareness.
Draine and Greenwald's (in press) experiments follow a long tradition of studies that have attempted to demonstrate perception in the complete absence of awareness. Do Draine and Greenwald's experiments provide more compelling evidence for unconscious perception than is provided by many previous studies? We do not think so. Their regression strategy is subject to the same criticisms that can be directed at any experiment based on the classic dissociation paradigm. Namely, questions can be raised regarding 1) whether the measures that they used to assess conscious perception are adequate measures of ALL relevant conscious experiences, and 2) whether their methodology was adequate to demonstrate that the measures of conscious perception exhibited null sensitivity for the majority of the participants. Given these questions, we do not believe that their findings necessarily provide better evidence for unconscious perception than is provided by results of many previous experiments.
Draine and Greenwald's experiments represent what is probably the most rigorous application of the dissociation paradigm to the study of unconscious perception that has ever been reported. Their experiments have pushed the dissociation paradigm to its limits in trying to establish that selected measures of conscious perception assess all relevant conscious experiences. However, in our view, Draine and Greenwald were unsuccessful in providing a completely convincing demonstration of unconscious perception. This outcome illustrates how futile it is to use the dissociation paradigm to demonstrate unconscious perception in the complete absence of conscious perception. It may just be impossible to design an experiment that satisfies the logical requirements of the dissociation paradigm. Perhaps, after more than 100 years of inconclusive experiments, it is time to abandon the research strategy of attempting to establish dissociations to demonstrate unconscious perception.
Another reason it is probably wise to abandon the dissociation paradigm in studies of unconscious perception is that rigorous application of the dissociation paradigm can result in misleading conclusions. As noted by Draine and Greenwald (in press), an important requirement that must be satisfied whenever the dissociation paradigm is used to demonstrate unconscious perception is that the selected measure of conscious perception must be ". . . maximally sensitive to conscious stimulus effects" (Footnote 3). It is for this reason that they attempted to find the most sensitive direct measure of perception and then to establish the experimental conditions in which this direct measure exhibited zero or null sensitivity. However, this research strategy can only be successful if the direct measure of perception is ONLY sensitive to consciously perceived information (cf., Reingold & Merikle, 1988, 1990). If the direct measure of perception happens to be sensitive to both consciously and unconsciously perceived information, then the strategy of establishing null sensitivity for the direct measure will lead to a severe underestimation of unconscious perception.
Previously, we suggested that a reasonable a priori expectation regarding any measure of perception, either direct or indirect, is that it is potentially sensitive to both consciously and unconsciously perceived information (Reingold & Merikle, 1988, 1990; see also Reingold & Toth, 1996). If this suggestion has any merit, then whenever experimental conditions are arranged so that the sensitivity of the direct measure is significantly reduced, the sensitivity of any measure assumed to assess unconscious perception will also be reduced. Such a state of affairs will lead to a serious underestimation of the influence of unconscious perception. In fact, it may even account for the findings reported by Greenwald et al. (1996) showing that unconsciously perceived information lasts no longer than 100 ms. Given that Greenwald et al. used the regression strategy to assess semantic priming under experimental conditions that minimized the sensitivity of a direct measure of perception, their findings are not unexpected, if the direct measure of perception was a sensitive indicator of both conscious and unconscious perception. By establishing the experimental conditions in which the direct measure showed minimal sensitivity, they may have also established experimental conditions in which the indirect measure showed minimal sensitivity.
One of the primary reasons that there has been so much interest over the years in demonstrating unconscious perception in the complete absence of conscious perception is that such a demonstration would be a compelling proof of the existence of unconscious perception. We doubt that it will ever be possible to prove the existence of unconscious perception. The distinction between conscious and unconscious processes provides a way of categorizing or thinking about cognitive processes that has been part of our intellectual history from at least a time shortly following Descartes' distinction between the body and the soul (Whyte, 1960). It is not at all clear what types of experiments would ever prove or disprove the validity of this conceptual distinction. For this reason, a better, more productive research strategy is to accept the distinction between conscious and unconscious cognitive processes as a working assumption and then to develop research strategies based on this assumption. In this way, the usefulness of the distinction for understanding human cognition can be assessed.
In recent years, this alternative research strategy has proven successful. For example, we have used the relative sensitivity approach to identify situations in which unconscious influences play a significant role (Merikle & Reingold, 1991), Debner and Jacoby (1993) have used the process-dissociation procedure to estimate the separate contributions of conscious and unconscious perceptual processes, and Merikle, Joordens, and Stolz (1995) have used estimates of the relative magnitude of unconscious perceptual processes to predict qualitative differences in the consequences of perceived stimuli. Although none of these experiments necessarily "prove" the existence of unconscious perception, the results of these experiments have provided new insights and understandings regarding human cognition. The success of these experiments demonstrates the value of thinking about cognition in terms of the distinction between conscious and unconscious processes.
Adams, J. K. (1957). Laboratory studies of behavior without awareness. Psychological Bulletin, 54, 383-405.
Chalmers, D. J. (1996). The conscious mind. New York: Oxford University Press.
Debner, J. A., & Jacoby, L. L. (1994). Unconscious perception: Attention, awareness, and control. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 304-317.
Draine, S. C., & Greenwald, A. G. (in press). Replicable unconscious semantic priming. Journal of Experimental Psychology: General.
Greenwald, A. G., Draine, S. C., & Abrams, R. H. (1996). Three markers of unconscious semantic activation. Science, 273, 1699-1702.
Greenwald, A. G., Klinger, M. R., & Schuh, E. S. (1995). Activation by marginally perceptible ("subliminal") stimuli: Dissociations of unconscious from conscious cognition. Journal of Experimental Psychology: General, 124, 22-42.
Hirst, W. (1995). Cognitive aspects of consciousness. In M. Gazzaniga (Ed.), The Cognitive Neurosciences (pp. 1307-1319). Cambridge, MA: MIT Press.
Intraub, H. (1984). Conceptual masking: The effects of subsequent visual events on memory for pictures. Journal of Experimental Psychology: Learning, Memory, and Cognition, 10, 115-125.
Kihlstrom, J. F. (1987). The cognitive unconscious. Science, 237, 1445-1452.
Kuntz-Wilson, W. R., & Zajonc, R. B. (1980). Affective discrimination of stimuli that cannot be recognized. Science, 207, 557-558.
Loftus, G. R., Hanna, A. M., & Lester, L. (1988). Conceptual masking: How one picture captures attention from another picture. Cognitive Psychology, 20, 237-282.
Marcel, A. J. (1983). Conscious and unconscious perception: Experiments on visual masking and word recognition. Cognitive Psychology, 15, 197-237.
Merikle, P. M. (1992). Perception without awareness: Critical issues. American Psychologist, 47, 792-795.
Merikle, P. M., Joordens, S., & Stolz, J. A. (1995). Measuring the relative magnitude of unconscious influences. Consciousness and Cognition, 4, 422-439.
Merikle, P. M., & Reingold, E. M. (1991). Comparing direct (explicit) and indirect (implicit) measures to study unconscious memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 17, 224-233.
Miller, J. G. (1942). Unconsciousness. New York: Wiley.
Peirce, C. S., & Jastrow, J. (1884). On small differences in sensation. Memoirs of the National Academy of Sciences, 3, 73-83.
Reingold, E. M., & Merikle, P. M. (1988). Using direct and indirect measures to study perception without awareness. Perception & Psychophysics, 44, 563-575.
Reingold, E. M., & Merikle, P. M. (1990). On the inter-relatedness of theory and measurement in the study of unconscious processes. Mind & Language, 5, 9-28.
Reingold, E. M., & Toth, J. P. (1996). Process dissociations versus task dissociations: A controversy in progress. In G. Underwood (Ed.), Implicit cognition (pp. 159-202). Oxford: Oxford University Press.
Sidis, B. (1898). The psychology of suggestion. New York: Appleton.
Weiskrantz, L. (1986). Blindsight. New York: Oxford University Press.
Whyte, L. L. (1960). The unconscious before Freud. New York: Basic Books.
Preparation of this article was facilitated by research grants from the Natural Sciences and Engineering Research Council of Canada to Philip M. Merikle and Eyal M. Reingold.