Free
Research Article  |   March 2007
Fear perception: Can objective and subjective awareness measures be dissociated?
Author Affiliations
Journal of Vision March 2007, Vol.7, 10. doi:https://doi.org/10.1167/7.4.10
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Remigiusz Szczepanowski, Luiz Pessoa; Fear perception: Can objective and subjective awareness measures be dissociated?. Journal of Vision 2007;7(4):10. https://doi.org/10.1167/7.4.10.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Whereas previous studies of fearful-face perception have probed visual awareness according to either objective or subjective criteria, in the present study, we probed the perception of briefly presented and masked fearful faces by assessing both types of perception within the same task. Both objective and subjective sensitivity measures were assessed within a common signal detection theory framework. To evaluate single-participant awareness, we employed a nonparametric receiver operating characteristic (ROC) analysis of the behavioral data, which involved collecting a large number of trials over multiple sessions. Our findings revealed that nearly all subjects could reliably detect 17-ms fearful-face targets, thus exhibiting above-chance objective perception at this target duration. Reliable subjective sensitivity was also observed for 33-ms fearful-face targets and, for some subjects, even for 17-ms targets. The analysis of single-session data suggests that previous experiments may have lacked sufficient statistical power to establish above-chance performance. Taken together, our findings are consistent with a dissociation of fear perception according to objective and subjective criteria, which could be assessed for each individual participant. The determination of such a dissociation zone may help in understanding the conditions linked to aware and unaware fear perception.

Introduction
Visual backward masking is a powerful technique that has been used in the study of visual awareness during the viewing of emotion-laden visual information (Esteves & Öhman, 1993). In backward masking, the processing of a briefly presented target stimulus is degraded by a subsequent mask stimulus. Stimulus visibility is manipulated by varying both target and mask strength and/or timing. Until recently, it was generally accepted that emotional faces are processed during “subliminal” conditions, namely, when subjects are unaware of the masked emotional expression (Dolan, 2002; Öhman, 2002). However, recent studies have challenged this view and claimed that visual awareness may be critical for the processing of such stimuli (Pessoa, 2005). 
One factor that may help explain the discrepancy of previous results concerns the use of different criteria to determine whether or not a subject is aware of perceiving a stimulus. On the one hand, behavioral and neuroimaging studies that report unaware perception of emotional faces have often evaluated awareness according to subjective criteria (e.g., Whalen et al., 1998). According to subjective criteria, unaware perception occurs when subjects report not having seen target stimuli or being unable to perform the task better than chance (independent of their actual performance). Subjective criteria hold that only the subjects themselves have access to their inner states and that their introspection is a reliable source of information about conscious experiences (Merikle, Smilek, & Eastwood, 2001). On the other hand, objective criteria have been used in studies that have suggested that awareness may be necessary for the processing of emotional faces (Pessoa, Japee, & Ungerleider, 2005; Pessoa, Japee, Sturman, & Ungerleider, 2006). According to objective criteria, unaware conditions occur when a subject's performance in a yes/no or forced-choice task is at chance, such as when subjects fail to detect alternative stimulus states (presence vs. absence of visual targets). Thus, the perception of visual stimuli is objectively unaware when the sensitivity measure of awareness (such as d′) is at the null level—that is, subjects exhibit null sensitivity. Under such conditions, behavioral effects of unaware stimuli (e.g., faster reaction time for undetected fearful faces), as well as associated physiological or neuroimaging signals, would constitute correlates of unaware perception. 
Unfortunately, no consensus exists regarding the “best” approach to measure and characterize awareness. In fact, there is a mounting tension between objective and subjective threshold approaches (Merikle et al., 2001; Snodgrass, Bernat, & Shevrin, 2004). Historically, an important concern with subjective procedures is that they can be quite sensitive to response bias (Eriksen, 1960). For instance, subjects may be reluctant to indicate having seen a stimulus when the available evidence is very weak. Such concern came into sharp focus with the development of signal detection theory (SDT; Green & Swets, 1966). As shown by SDT models, subjective threshold effects are highly vulnerable to alternative conscious perception explanations. At the same time, the importance of subjective measures of awareness has been that they resonate with the idea that an intuitively appealing definition of awareness should be based on introspective reports of individuals' inner states (James, 1890). In fact, from the subjective awareness perspective, at times, it has been stated that the utilization of objective measures logically precludes the existence of unconscious perception (Bowers, 1984; Öhman, 1999). 
As further addressed in the Discussion section, we argue that it is important to investigate both objective and subjective measures of perception. Accordingly, in the present study, to investigate the perception of minimally visible stimuli, our approach was to employ both types of measures. In the past, the comparison between objective and subjective measures has been hampered by the use of different tasks to gauge them. For example, although subjects may perform a forced task during studies measuring objective thresholds, they typically employ some form of nonforced, self-report during studies measuring subjective thresholds. Here, our goal was to obtain objective and subjective measures of perception within a single task. Initially, we will focus on characterizing the perception of minimally visible stimuli with both types of measures, without necessarily drawing conclusions concerning what they might mean in terms of aware versus unaware perception. We will discuss this issue in the Discussion section, where we address the potential implications of our findings. 
In the present study, in each trial, subjects viewed a target face that was fearful or nonfearful, which was followed by a mask stimulus ( Figure 1). Subjects were instructed to first provide a response that indicates whether or not a fearful face was present in the display at any time. Such component of the task allowed us to compute a standard sensitivity index (such as d′) that could be used to assess objective perception. In addition, subsequently, subjects indicated the level of confidence of their yes/no responses. As others in the past, we reasoned that we could use the subject's ability to discriminate between correct and incorrect responses based on their confidence ratings as an index of subjective perception (Kolb & Braun, 1995; Kunimoto, Miller, & Pashler, 2001). Thus, if subjects could access information about the stimulus, higher confidence trials should be linked with correct responses more often than with incorrect responses; likewise, lower confidence trials should be linked with incorrect responses more often than with correct responses. In other words, the relationship between response confidence and accuracy should provide an index of subjective perception. 
Figure 1
 
Backward masking paradigm. Fearful, happy, or neutral target faces were shown for 17, 25, 33, or 41 ms and were immediately followed by a neutral face mask that was shown such that the target plus the mask lasted 100 ms. Subjects were first asked whether fear appeared at any time in the display. They were then asked to rate the confidence of their initial response by using a 1-to-6 scale (from low to high confidence).
Figure 1
 
Backward masking paradigm. Fearful, happy, or neutral target faces were shown for 17, 25, 33, or 41 ms and were immediately followed by a neutral face mask that was shown such that the target plus the mask lasted 100 ms. Subjects were first asked whether fear appeared at any time in the display. They were then asked to rate the confidence of their initial response by using a 1-to-6 scale (from low to high confidence).
Figure 2 provides a representation of the two tasks performed by the subject. The top part depicts the standard detection task, namely, a yes/no choice between two types of event. The yes/no detection task can be used as an index of objective perception. The bottom part depicts the second task, which is known as a Type 2 task (Galvin, Podd, Drga, & Whitmore, 2003; Pollack & Decker, 1958), and can be viewed as an extension of the standard detection problem. One of two events, correct or incorrect, occurs when the observer makes the yes/no decision. The observer is then asked to decide “Yes, it was correct” or “No, it was not correct” during the decision interval. Alternatively, as in our experiment, the observer is asked to provide a confidence rating during the decision interval, which can be used to gauge subjective perception. 
Figure 2
 
Abstract representation of the tasks performed by the subject. The first task (top) corresponded to the standard detection task. During the observation interval, either a signal (S) or a noise (N) event occurs with probability p(S) and p(N), respectively. During the decision interval, the subject indicates “Yes, a signal occurred” or “No, a signal did not occur.” The observation and decision intervals of the detection task become the observation interval for the Type 2 task (bottom). In this case, either a correct (C) or incorrect (I) event occurs with probability p(C) and p(I), respectively. During the decision interval, the subject indicates “Yes, it was correct” or “No, it was not correct.” In the present experiment, detection signal trials comprised fear-containing trials (fearful–neutral target–mask pairs), and detection noise trials comprised non-target-containing trials (happy–neutral and neutral–neutral target–mask pairs). In addition, confidence ratings were used instead of “correct”/“incorrect” judgments (inspired by Figure 1 of Galvin et al., 2003).
Figure 2
 
Abstract representation of the tasks performed by the subject. The first task (top) corresponded to the standard detection task. During the observation interval, either a signal (S) or a noise (N) event occurs with probability p(S) and p(N), respectively. During the decision interval, the subject indicates “Yes, a signal occurred” or “No, a signal did not occur.” The observation and decision intervals of the detection task become the observation interval for the Type 2 task (bottom). In this case, either a correct (C) or incorrect (I) event occurs with probability p(C) and p(I), respectively. During the decision interval, the subject indicates “Yes, it was correct” or “No, it was not correct.” In the present experiment, detection signal trials comprised fear-containing trials (fearful–neutral target–mask pairs), and detection noise trials comprised non-target-containing trials (happy–neutral and neutral–neutral target–mask pairs). In addition, confidence ratings were used instead of “correct”/“incorrect” judgments (inspired by Figure 1 of Galvin et al., 2003).
A useful representation of the trial types of the rating task can also be formulated within the SDT framework: Discrimination accuracy (correct vs. incorrect) can be paired with response confidence (“low” vs. “high”) to yield a 2 × 2 table ( Table 1). Within this representation, a “hit” corresponds to a correct trial with high confidence; a “miss” corresponds to a correct trial with low confidence; a “false alarm” corresponds to an incorrect trial with high confidence; and a “correct rejection” corresponds to an incorrect trial with low confidence. Thus, correct trials for which the subject rates with “high” confidence would comprise “hit” trials, whether or not they actually contain a target stimulus. At the same time, incorrect trials for which the subject rates with “high” confidence would comprise “false alarm” trials, irrespective of the type of incorrect trial (i.e., incorrect trials containing or not containing a target stimulus). Given such trial categorization, it is then possible to compute an SDT-like measure of sensitivity that reflects the distance between the means of the distributions representing correct and incorrect responses ( Figure 3). We can then see that, in this context, a d′ index would indicate the subjects' ability to discriminate between their correct and incorrect responses. In this sense, it provides a (subjective) measure of the subjects' own assessment of their performance. 
Table 1
 
Trial type categorization for subjective perception sensitivity (based on Kunimoto et al., 2001).
Table 1
 
Trial type categorization for subjective perception sensitivity (based on Kunimoto et al., 2001).
Discrimination accuracy Confidence
High Low
Correct Hit Miss
Incorrect False alarm Correct rejection
Figure 3
 
Assessing subjective perception with SDT. Two event types corresponding to correct and incorrect trials are represented via Gaussian distributions. The criterion for reporting confidence is indicated by the vertical line and dictates when the subject reports “high” versus “low” confidence. A d′ value greater than zero reveals the subject's ability to discriminate between correct and incorrect responses and can be used to index subjective perception. In this article, we employed a nonparametric measure of sensitivity based on ROC curves (inspired by Figure 1 of Kunimoto et al., 2001).
Figure 3
 
Assessing subjective perception with SDT. Two event types corresponding to correct and incorrect trials are represented via Gaussian distributions. The criterion for reporting confidence is indicated by the vertical line and dictates when the subject reports “high” versus “low” confidence. A d′ value greater than zero reveals the subject's ability to discriminate between correct and incorrect responses and can be used to index subjective perception. In this article, we employed a nonparametric measure of sensitivity based on ROC curves (inspired by Figure 1 of Kunimoto et al., 2001).
Previous studies of threshold perception of emotional faces have evaluated the perception of minimally visible stimuli at the group level (Esteves & Öhman, 1993). Thus, for instance, an unaware condition is one in which, on average, subjects cannot detect emotional faces better than chance. One result that has emerged from this framework is that participants appear to be unaware of masked emotional faces shown for 30 ms or so. In practice, such duration has often been treated in the literature as a “hard threshold” that would apply universally (i.e., for all participants). Such approach is undesirable because it implicitly assumes that subjects are homogeneous in terms of emotional face perception. We suggest that a more fruitful strategy is to assess threshold vision at the single-participant level, which is especially important given recent evidence of intersubject differences in the ability to detect fearful-face targets (Pessoa et al., 2006). Thus, in the present study, we investigated threshold vision for all individual participants separately. 
In summary, in the present study, we investigated the perception of emotional faces by employing a backward masking paradigm. Both objective and subjective indexes of perception were computed based on the same fear-detection task. To investigate the minimal critical stimulus conditions for perceiving fearful faces according to our criteria, we parametrically varied the exposure duration of target stimuli. In addition, we examined potential changes in threshold vision by having participants undergo multiple behavioral sessions ( Experiment 1). Such approach allowed us to test whether or not performance was stable across time or, alternatively, whether subjects exhibited improvements in sensitivity. 
Experiment 1: Mixed design
Methods
Participants
Five volunteers (of whom three were males) participated in this study (29 ± 5 years old), which was approved by the Institutional Review Board of Brown University. Subjects were in good health with no past history of psychiatric and neurological diseases and gave informed consent. Subjects had normal or corrected-to-normal vision. One of the authors (R.S.) served as a subject. 
Stimuli and design
In each trial (see Figure 1), a white fixation cross was displayed for 300 ms on a black background, followed by a 50-ms blank screen, followed by a fearful, happy, or neutral target face, which was immediately followed by a neutral face that served as a mask. In a given trial, the identity of the face used as the target or the mask was always different (as well the identities used in successive trials). Face stimuli subtended 4° × 5° of visual angle. To manipulate stimulus visibility, we varied the duration of the target face parametrically: 17, 25, 33, and 41 ms. The total duration of the target–mask pair was fixed at 100 ms. After the presentation of each target–mask pair, subjects had 2 s to indicate “fear” or “no fear” via button press. They then had 2.5 s to rate the confidence in their response on a scale of 1 to 6 (low to high confidence). The total duration of each trial was 5 s. In each experimental session, subjects performed 160 trials shown in a random order, which followed a 3 (target–mask pairs: fearful–neutral, happy–neutral, and neutral–neutral) × 4 (target durations: 17, 25, 33, and 41 ms) design. Half of the trials contained a fearful-face target and half did not (1/4 happy–neutral and 1/4 neutral–neutral target–mask pairs). Subjects did not receive any information about stimulus durations or trial types. Subjects performed a total of five independent sessions; three subjects performed the sessions in consecutive days and two subjects performed the sessions within 4 days, with two sessions in 1 day (separated by at least 2 hr). Overall, all subjects performed 3,200 trials (800 trials for each duration). 
Face stimuli were obtained from the Ekman set (Ekman & Friesen, 1976), a second set elaborated by Öhman et al. (KDEF, Lundqvist, D., Flykt. A., and Öhman, A.; Karolinska Hospital, Stockholm, Sweden), and a third set validated by Alumit Ishai at NIMH (Bethesda, USA; Ishai, Pessoa, Bikle, & Ungerleider, 2004). Stimuli were displayed on a ViewSonic Professional Series P95f+ monitor (screen refresh rate: 120 Hz) fitted with a Radeon 9800 AGP graphics card. Measurements using a photodiode and a digital oscilloscope (Tektronix, TDS3000B series) confirmed that the targeted durations were kept constant in our setup. 
We employed, as targets, 40 instances each of fearful, happy, and neutral faces, and, separately, we used 80 neutral faces as masks. In most masking studies involving emotional faces, only neutral–neutral trials are employed as “noise” trials. In the present study, happy faces were included to more closely match fearful faces in terms of low-level features, such as brightness around the mouth and eye regions, because both fearful and happy faces tend to be brighter than neutral in these regions. Thus, the inclusion of happy faces reduced a subject's ability of detecting fearful faces by using such low-level cues. In addition, the inclusion of happy faces discouraged subjects from adopting a strategy of indicating fear whenever features deviated from those of a neutral face. Finally, to prevent the “detection” of fear based on subtle motion cues from the transition between fearful and nonfearful faces, mask stimuli were randomly displaced so as to not perfectly overlap the target face. Specifically, on approximately half of the trials (chosen randomly), the mask stimulus was shifted along one of the four diagonal directions by a small spatial offset of ∼0.5° of visual angle (Phillips et al., 2004). 
SDT for the analysis of objective and subjective perception: Nonparametric receiver operating characteristic approach
To evaluate individual objective and subjective measures, we employed a nonparametric receiver operating characteristic (ROC) analysis of the behavioral data. Such approach does not make normality assumptions about the signal and noise distributions (Green & Swets, 1966; Hanley & McNeil, 1982; Macmillan & Creelman, 2005). ROC curves of both objective and subjective measures were generated (see below), and the area under the curve (A′; Egan, 1975; Green & Swets, 1966) was employed as Aobjective′ and Asubjective′ values, respectively. 
ROC curves for objective perception were estimated based on yes/no decisions (“fear”/“no fear”) and the corresponding confidence ratings on a six-point scale (Macmillan & Creelman, 2005), which were used to obtain cumulative conditional probabilities (Green & Swets, 1966; Macmillan & Creelman, 2005). The hit rate was defined as the probability of reporting “fear” given that a fear-containing stimulus was shown (p(“fear”|fear-containing stimulus)), whereas the false alarm rate was defined as the probability of reporting “fear” given that a non-fear-containing stimulus was shown (p(“fear”|non-fear-containing stimulus)). 
ROC curves for subjective perception were estimated based on response correctness and confidence as motivated in the Introduction (see Table 1). Correct responses involved “fear” report for fear-containing trials and “no fear” report for non-fear-containing trials. Incorrect responses included “no fear” report for fear-containing trials and “fear” report for non-fear-containing trials. In the context of subjective awareness, we considered “hits” as correct trials associated with high-confidence ratings and “false alarms” as incorrect trials with high-confidence ratings (Kunimoto et al., 2001; Pollack & Decker, 1958). We considered high-confidence trials as those with ratings between 4 and 6 (inclusive); critically, other partitioning schemes, such as those based on the median (i.e., “median split”), yielded nearly indistinguishable results. The hit rate was defined as the probability of reporting “high confidence” given that the trial was correct (p(“high confidence”|correct)), whereas the false alarm rate was defined as the probability of reporting “high confidence” given that the trial was incorrect (p(“high confidence”|incorrect)). These definitions follow those employed in Type 2 SDT analyses (Galvin et al., 2003; Kunimoto et al., 2001; Macmillan & Creelman, 2005; Pollack & Decker, 1958). 
When plotting ROC curves ( Figure 4), pairs of hit and false alarm rates derived from 12 alternative yes/no decisions (6 levels for reporting “fear” and 6 levels for reporting “no fear”) were used to plot objective ROC curves; pairs of hits and false alarms derived from 6 confidence ratings (3 levels for reporting “high confidence” and 3 levels for reporting “low confidence”) were used to plot subjective ROC curves. All ROCs originated at (0,0) and ended at (1,1). 
Figure 4
 
ROC curves for two representative subjects. Top: Subject S3 exhibited above-chance objective and subjective sensitivity at all target durations. Bottom: Subject S4 exhibited above-chance objective sensitivity at all target durations and above-chance subjective sensitivity for the 25-, 33-, and 41-ms durations. For 17-ms targets, Subject S4 could detect fearful faces better than chance but failed to reliably discriminate between correct and incorrect responses. Such pattern of results is suggestive of a dissociation between objective and subjective perception (see Figure 10).
Figure 4
 
ROC curves for two representative subjects. Top: Subject S3 exhibited above-chance objective and subjective sensitivity at all target durations. Bottom: Subject S4 exhibited above-chance objective sensitivity at all target durations and above-chance subjective sensitivity for the 25-, 33-, and 41-ms durations. For 17-ms targets, Subject S4 could detect fearful faces better than chance but failed to reliably discriminate between correct and incorrect responses. Such pattern of results is suggestive of a dissociation between objective and subjective perception (see Figure 10).
To assess the statistical significance of the area under the ROC curve, we employed two methods. First, we employed bootstrap resampling methods (Efron & Tibshirani, 1993) for the estimation of confidence intervals around A′ values, as implemented in the AccuROC software package (Accumetric Corporation, Montreal, Canada). We deemed subjects as objectively or subjectively above chance when A′ measures exceeded 0.5 (i.e., the area under the y = x line when the hit and false alarm rates are equivalent) and the 95% confidence interval did not overlap with 0.5; in figures, error bars indicate 95% bootstrap confidence intervals based on 2,000 iterations. In addition, to report p values, we employed the method of DeLong, DeLong, and Clarke-Pearson (1988). In practice, the two methods provided equivalent results. 
Because multiple statistical tests were applied simultaneously in the context of Tables 2, 3, and 4, without an explicit correction for multiple comparisons, the probability of a Type I error may have attained unacceptable levels. Statistical significance was thus corrected for multiple comparisons according to the Bonferroni method. For Table 2, the α level was .0013 (.05/40, where .05 is the standard α level and 40 corresponded to the number of simultaneous tests [5 subjects × 4 durations × 2 perception measures]). For Table 3, the α level was .0003 (.05/150, where .05 is the standard α level and 150 corresponded to the number of simultaneous tests [5 subjects × 3 durations × 2 perception measures × 5 sessions]); in Table 3, we only considered the 17-, 25-, and 33-ms target durations, as 41-ms targets were essentially suprathreshold. For Table 4, the α level was .0021 (.05/24, where .05 is the standard α level and 24 corresponded to the number of simultaneous tests [6 subjects × 2 durations × 2 perception measures]). For completeness, uncorrected p values are provided in Tables 2, 3, and 4, and color coding is used to indicate when values were significant when multiple comparison were employed (green), when they were significant when multiple comparisons were not employed (pink), and when they were not significant (red). 
Table 2
 
Experiment 1: Objective and subjective sensitivity values ( A′): pooled data.
Table 2
 
Experiment 1: Objective and subjective sensitivity values ( A′): pooled data.
 

Note: p values are enclosed in parentheses.

 

Green: statistically significant, corrected for multiple comparisons; pink: statistically significant, not corrected for multiple comparisons; red: not statistically significant.

 

p values of .000 indicate values less than .001.

Table 3
 
Experiment 1: Objective and subjective sensitivity values ( A′): single-session data.
Table 3
 
Experiment 1: Objective and subjective sensitivity values ( A′): single-session data.
 

Note: p values are enclosed in parentheses.

 

Green: statistically significant, corrected for multiple comparisons; pink: statistically significant, not corrected for multiple comparisons; red: not statistically significant.

 

p values of .000 indicate values less than .001.

Table 4
 
Experiment 2: Objective and subjective sensitivity values ( A′): pooled data.
Table 4
 
Experiment 2: Objective and subjective sensitivity values ( A′): pooled data.
 

Note: p values are enclosed in parentheses.

 

Green: statistically significant, corrected for multiple comparisons; red: not statistically significant.

 

p values of .000 indicate values less than .001.

In an important paper, Pollack and Hsieh (1969) showed that the sampling variability of the area under the ROC curve is well approximated by a binomial probability of a mean proportion—this is especially the case when values are not near the extremes of the distribution (e.g., when A′ does not approach 1). To investigate the issue of sample size, we performed a power analysis under the binomial assumption (Figure 7). 
Results
Results for trials pooled across sessions
We initially investigated our results by pooling trials across the five sessions. To determine individual objective and subjective measures during the fear-detection task, we computed sensitivity values ( A′) for the following target durations: 17, 25, 33, and 41 ms. Examples of both objective and subjective ROC curves for two representative subjects are presented in Figure 4. Subject S3 could reliably detect fearful faces, as well as reliably discriminate between correct and incorrect responses, for all stimulus durations. On the other hand, Subject S4 could detect fearful faces at all stimulus durations but was unable to discriminate correct/incorrect responses for 17-ms targets (the ROC curve for 17-ms targets fell along the diagonal). Thus, for 17-ms targets, Subject S4 could detect fear but did not exhibit evidence of accessing information concerning his or her performance. Table 2 displays A′ values for all subjects and conditions. 
Somewhat surprisingly, our results revealed that A objective′ values reliably exceeded 0.5 (the value expected by chance) for all subjects even during the 17-ms condition; thus, all subjects could detect 17-ms targets. In terms of A subjective′ values, all subjects discriminated correct/incorrect responses for target durations of 33 and 41 ms; all subjects, except for Subject S1, were above chance for the 25-ms duration; Subjects S2 and S3 exhibited above-chance performance even for very brief 17-ms targets (see the Null sensitivity and statistical power section). 
The above results indicate potential dissociations between objective and subjective values that can be further visualized by considering Figures 5A and 5B. When error bars of subjective measures (red) cross the 0.5 line and error bars of objective measures (blue) do not cross this line, evidence of a dissociation occurs (see arrows). Note that the error bars do not take into account any correction for multiple comparisons; thus, although they are useful measures of variability for an individual subject, they are too liberal when one considers the series of results shown in Table 2 (which take into account multiple comparisons). 
Figure 5
 
Objective and subjective measures of perception. Values of A objective′ (blue) and A subjective′ (red) are plotted as a function of subject for Experiment 1 (A and B) and Experiment 2 (C and D). (A and C) Perception measures based on the nonparametric analysis. (B and D): Perception measures based on the parametric analysis. The arrows indicate conditions for which there was evidence for a dissociation between objective and subjective perception. See text for further details. Error bars indicate 95% bootstrap confidence intervals.
Figure 5
 
Objective and subjective measures of perception. Values of A objective′ (blue) and A subjective′ (red) are plotted as a function of subject for Experiment 1 (A and B) and Experiment 2 (C and D). (A and C) Perception measures based on the nonparametric analysis. (B and D): Perception measures based on the parametric analysis. The arrows indicate conditions for which there was evidence for a dissociation between objective and subjective perception. See text for further details. Error bars indicate 95% bootstrap confidence intervals.
To rule out the possibility that the potential dissociations between objective and subjective measures stemmed from our method of computing A′, we also computed concurrent d′ sensitivity measures (which assume that signal and noise distributions are normally distributed; see 1). As seen by considering Figures 5A and 5B, both methods made relatively consistent assessments of dissociations (for the 17-ms condition of Subject S1, the d subjective′ value also crossed the chance line, consistent with the results of Table 2). Combined, the two analyses were consistent with an objective/subjective dissociation for 17-ms targets for Subjects S4 and S5 and a dissociation for 25-ms targets for Subject S1 (here, we refer to the dissociations suggested by both methods). 
Objective and subjective measures as a function of noise trials
In the present study, “signal” trials involved fearful–neutral target–mask pairs, whereas noise trials involved two types of trials: happy–neutral and neutral–neutral pairs. In the preceding analyses, these two trial types were pooled together in our estimates of A objective′ and A subjective′ values. It is instructive to compute these measures as a function of noise type (i.e., after sorting trials based on noise type). The results are shown in Figure 6; only the 17- and 25-ms target durations are shown given that they were “perithreshold” conditions. First, let us consider A objective′ values (blue) for the 17-ms duration (top). It is interesting to note that all values were suprathreshold (i.e., >0.5) when neutral–neutral trials (right) comprised the noise. When happy–neutral trials were considered (left), Subjects S4 and S5 did not exceed 0.5 and Subject S1 barely did so (as before, the error bars may be thought as “liberal” given that they do not incorporate a correction for multiple comparisons). These results suggest that these subjects could reliably detect fearful targets only insofar as these differed from neutral stimuli; in other words, they may have solved the task by adopting a strategy of detecting deviations from neutral and not necessarily by detecting fear per se. However, Subjects S2 and S3 were able to reliably detect fearful targets even in the context of happy–neutral noise, suggesting that they were able to detect fear (i.e., not only deviations from neutral). 
Figure 6
 
Objective and subjective perception measures as a function of noise trials. Objective measures are shown in blue, and subjective measures are shown in red. Top: Results for 17-ms targets based on happy–neutral noise trials (left) or based on neutral–neutral noise trials (right). Bottom: Results for 25-ms targets based on happy–neutral noise trials (left) or based on neutral–neutral noise trials (right). Error bars indicate 95% bootstrap confidence intervals.
Figure 6
 
Objective and subjective perception measures as a function of noise trials. Objective measures are shown in blue, and subjective measures are shown in red. Top: Results for 17-ms targets based on happy–neutral noise trials (left) or based on neutral–neutral noise trials (right). Bottom: Results for 25-ms targets based on happy–neutral noise trials (left) or based on neutral–neutral noise trials (right). Error bars indicate 95% bootstrap confidence intervals.
Now, let us consider A subjective′ values for the 17-ms duration ( Figure 6; top, red). Three subjects (S1, S4, and S5) did not exhibit above-chance behavior in the context of happy–neutral noise trials (left), and two subjects (S3 and S4) did not exhibit above-chance behavior in the context of neutral–neutral noise trials (right; see Table 2). In the former case, the elimination of subjective perception was pretty complete (the A subjective′ values were 0.5, 0.47, and 0.51, respectively), suggesting that the correct versus incorrect judgment was more difficult during happy–neutral trials (left) than during neutral–neutral trials (right). 
Finally, as shown in Figure 6, objective/subjective dissociations appeared to occur when happy or neutral faces were considered as noise targets. For instance, for 17-ms targets, dissociations were observed for Subjects S3 and S4 (and possibly for Subject S5) when neutral–neutral trials (top, right) were considered as noise. For 25-ms targets, dissociations appeared to occur for Subject S1 (and possibly Subject S4) when happy–neutral trials (bottom, left) were considered as noise. 
It should be noted that because subjects performed trials in a mixed fashion (all trial types were randomized), it is conceivable that participants found it difficult to maintain stable response criteria for both yes/no and confidence rating responses across the experiment (see Experiment 2). Thus, some of the differences between happy and neutral noise trials reported here may be due to the adoption of different response criteria for these trial types. 
Perception measures as a function of experimental session
Given that subjects participated in five 1-hr sessions, we probed how objective and subjective sensitivity values varied as a function of experimental session. 
Objective perception
First, let us discuss the 17-ms duration ( Table 3). On the one hand, Subjects S2 and S3 consistently exhibited suprathreshold A objective′ values for single sessions (for Subject S3, this was the case when a correction for multiple comparisons was not considered). On the other hand, Subjects S1, S4, and S5 often failed to exhibit suprathreshold A objective′ values for single sessions. These results, taken together with those obtained when all sessions were pooled together ( Table 2), were evaluated in the context of both (a) statistical power and (b) learning effects (i.e., changes in performance across sessions). 
During single sessions, for 17-ms targets, for Subjects S1, S4, and S5, A objective′ values ( Table 3) that exceeded 0.5 were relatively modest (with the exception of Session 5 for Subject S1 for which A objective′ was 0.68). Consequently, single sessions may have been underpowered to reveal suprathreshold A objective′ values (even when corrections for multiple comparisons are not considered). To further investigate this issue, we performed a power analysis ( Methods section), the results of which are shown in Figure 7. For an individual session, the critical value for detecting a true deviation from 0.5 with 80% power ( α = .05) was slightly more than 0.6. Such results reveal that individual sessions are probably underpowered to detect true suprathreshold perception at short durations, which are typically associated with modest sensitivity values (i.e., <0.6). At the same time, the critical value of A objective′ was approximately 0.55 when trials from the five experimental sessions were pooled together (80% power; α = .05), suggesting that the present experimental design had adequate statistical power for the detection of true objective sensitivity at short target durations (see values in Table 2). 
Figure 7
 
Statistical power curves. The number of trials is plotted as a function of the critical value of A′ at several levels of statistical power ( α = .05, two-tailed test). Thus, for example, to detect a true sensitivity of A′ = 0.6 with 80% power, approximately 300 (∼2 sessions) trials would be required. To detect a true sensitivity of A′ = 0.54 (vertical dashed line) with 80% power, approximately 1,300 (∼8 sessions) trials would be required.
Figure 7
 
Statistical power curves. The number of trials is plotted as a function of the critical value of A′ at several levels of statistical power ( α = .05, two-tailed test). Thus, for example, to detect a true sensitivity of A′ = 0.6 with 80% power, approximately 300 (∼2 sessions) trials would be required. To detect a true sensitivity of A′ = 0.54 (vertical dashed line) with 80% power, approximately 1,300 (∼8 sessions) trials would be required.
To investigate potential changes in performance, we plotted objective sensitivity values as a function of session number (relative to first-session values). Figure 8 illustrates that, on average, objective measures increased across sessions; improvements were observed for all stimulus durations. More formally, a significant linear trend was observed ( F = 68.9, p = .001). 
Figure 8
 
Learning effects. Sensitivity measures are plotted as a function of session number (as a ratio relative to Session 1 values). Top: Objective values increased across sections (linear trend, p < .01). Bottom: Subjective values showed some evidence of increase across sessions, but the results were not significant (linear trend, p = .07).
Figure 8
 
Learning effects. Sensitivity measures are plotted as a function of session number (as a ratio relative to Session 1 values). Top: Objective values increased across sections (linear trend, p < .01). Bottom: Subjective values showed some evidence of increase across sessions, but the results were not significant (linear trend, p = .07).
Subjective perception
For 17-ms targets, few individual sessions exhibited A subjective′ values significantly greater than chance ( Table 3). However, for at least Subjects S2 and S3, results were significant when the data from all five sessions were pooled together. See the Discussion section for further elaboration on statistical power. 
We also investigated learning effects for subjective measures. An interesting pattern emerged when we considered the evolution of A subjective′ values across sessions. Although values were relatively flat for the 17- and 25-ms durations, an increase in sensitivity was evident for 33- and 41-ms targets ( Figure 8). Overall, the linear trend only approached significance ( F = 6.088, p = .07), possibly due to differences in the trends for the two shorter and two longer durations. 
Experiment 2: Blocked design
In Experiment 1, participants performed trials in a random (intermixed) order, with each of the 12 different stimuli (3 target–mask pairs × 4 durations) occurring with equal probability. Thus, it is conceivable that participants did not adopt stable response criteria for both yes/no and confidence rating responses across the experiment. Such scenario complicates the interpretation of A objective′ and A subjective′ values in general, as well as potential dissociations between these measures, in particular. Accordingly, in Experiment 2, we adopted a “blocked” design for which stimulus conditions were held constant throughout blocks, maximizing the likelihood that participants would adopt a stable criterion. 
Methods
Participants
Six volunteers (of whom one was male) participated in this study (29 ± 7 years old), which was approved by the Institutional Review Board of Indiana University, Bloomington. Subjects were in good health with no past history of psychiatric and neurological diseases and gave informed consent. Subjects had normal or corrected-to-normal vision. 
Stimuli and design
Stimuli and design were very much like those in Experiment 1; here, we focus on the differences between the two experiments. Target faces were either fearful or neutral and were shown for 17 or 25 ms. Thus, Experiment 2 employed a 2 × 2 factorial design. During each session, subjects performed 800 trials, which occurred in a blocked fashion. During each block, a single target duration was employed, which was indicated via an initial “instruction screen” at the beginning of the block. Blocks contained 80 trials presented in a random order, with half of them containing fearful–neutral target–mask pairs and the other half containing neutral–neutral target–mask pairs. Block type was randomized across the session. Participants performed a total of two independent sessions (separated by 1–2 days for five subjects; one subject performed sessions during the same day, which were separated by 2 hr). Overall, subjects performed 1,600 trials (800 trials at each target duration). 
Comparing observed and predicted objective and subjective values
To further explore potential dissociations between objective and subjective measures, we compared observed values to predicted values, as computed via computer simulations. Following an ideal-observer framework, samples were drawn from pairs of overlapping Gaussian distributions with d′ separations from 0.0 to 2.5 (in steps of 0.1). A total of 400 independent random samples were drawn (200 noise samples and 200 signal + noise samples). As done in the past, we assumed that the distributions had a fixed standard deviation ratio of 1:1 (Hajian-Tilaki, Hanley, Joseph, & Collet, 1997; Obuchowski, 1994). A six-point confidence scale was simulated by considering equidistant cutoffs. Note that this procedure is resistant to departures from the normality assumption of the sensory distributions. Thus, although the ideal-observer model does not explicitly take into account different configurations of pairs of overlapping noise and signal + noise distributions, ROC-based measures are robust to departures from normality and potential biases introduced in their estimation are negligible (Hajian-Tilaki et al., 1997). Both objective and subjective ROC curves were then computed, and the area was evaluated (A′ values). For subjective measures, we considered high-confidence trials as those with ratings between 4 and 6 (inclusive), as done with actual data (see above). For both objective and subjective measures, the procedure was repeated 100 times to obtain average sensitivity values, which are plotted in Figure 9. A separate set of simulations was performed following methods suggested by Pollack and Hsieh (1969) and yielded very similar results. 
Figure 9
 
Simulated and observed objective and subjective values. (A) Experiment 1. (B) Experiment 2. Simulated results are shown in blue, whereas specific data points are shown in red (17 ms) and green (25 ms). Evidence of stronger dissociation between objective and subjective measures is given by points that do not overlap with the prediction line but cross the horizontal “chance” line. Error bars indicate 95% bootstrap confidence intervals.
Figure 9
 
Simulated and observed objective and subjective values. (A) Experiment 1. (B) Experiment 2. Simulated results are shown in blue, whereas specific data points are shown in red (17 ms) and green (25 ms). Evidence of stronger dissociation between objective and subjective measures is given by points that do not overlap with the prediction line but cross the horizontal “chance” line. Error bars indicate 95% bootstrap confidence intervals.
Results
The results of Experiment 2 ( Table 4; Figures 5C and 5D), which employed a blocked design, were largely consistent with the findings of Experiment 1, which employed a mixed design. Five of the six subjects could reliably detect 17- and 25-ms fearful faces (objective measure). At the same time, Subjects S2, S4, and S5 were not able to accurately discriminate between correct and incorrect responses for 17-ms targets; Subjects S4 and S5 also did not exhibit above-chance subjective values for 25-ms targets. The results of Experiment 2 also revealed objective/subjective dissociations as observed in Experiment 1, such as, for instance, those exhibited by Subjects S2 and S5 during 17-ms targets. Taken together, the replication in Experiment 2 of the general pattern observed in Experiment 1 suggests that the assessment of objective and subjective measures in the first experiment was not strongly contaminated by unstable response criteria. 
Dissociation between objective and subjective measures in Experiments 1 and 2
The results of Experiments 1 and 2 suggested a potential dissociation between objective and subjective measures. In general, correct/incorrect discrimination scores are expected to be lower than yes/no sensitivity values—for instance, subjects may typically find the latter task harder than the first (Kunimoto et al., 2001; Galvin et al., 2003). This pattern of results was clearly observed in our data (Figures 5A and 5B). It is important then to investigate the observed differences between objective and subjective measures and how these relate to expected differences between the two measures, as shown in Figure 9. In this figure, the diagonal lines represent the case in which objective and subjective values for a given condition are expected to be equal; the horizontal lines represent the subjective chance line; finally, the blue lines provide predicted values based on simulations (Methods section). In Experiment 1 (Figure 9A), the error bars of several observed values overlapped with the predicted line. At the same time, for Subject S1, the 25-ms target duration did not overlap with the prediction line but crossed the chance line, consistent with a stronger dissociation between objective and subjective measures. In Experiment 2 (Figure 9B), most observed values were actually lower than predicted ones. In particular, for Subjects S2 (17 ms) and S5 (17 and 25 ms), observed values did not overlap with the prediction line but crossed the chance line, again consistent with a stronger dissociation. If we consider the results of Experiments 1 and 2 collectively, 15 of 20 points fell below the predicted values, a deviation that is statistically significant (binomial test, p < .05; here, we excluded Subject S4 of Experiment 2, who was unable to reliably perform the task). Thus, overall, the comparison of observed and predicted data is consistent with a dissociation between objective and subjective measures. 
Naturally, the relationship between observed and simulated data critically depends on the assumptions of the computational method (see the Methods section for a discussion of the procedure and assumptions). For instance, like other similar models in the past (Hajian-Tilaki et al., 1997; Obuchowski, 1994), we did not assume any loss of information. In particular, no memory trace for the stimulus was assumed (Gold, Murray, Sekuler, Bennett, & Sekuler, 2005). However, because confidence ratings were made after the detection of the yes/no response, the existence of some form of memory trace for the stimulus event could lead to confidence ratings being made on noisier representations. Thus, in future work, it would be desirable to investigate a broader range of models. 
Discussion
In the present experiment, subjects performed a single detection task for which they were required to make two evaluations: an evaluation of whether or not a fearful face was present and an evaluation of their response confidence. As discussed in the Introduction, the first evaluation corresponds to a yes/no choice during a standard detection task. The second evaluation can be thought of as a discrimination between correct and incorrect responses (by providing “low” and “high” confidence ratings). Whereas the detection task explicitly probed subjects' perception of fear, confidence ratings provided an indirect assessment of the accessibility of information about fear. Thus, our experiment allowed us to test for potential dissociations between these two measures of fear perception, for instance, whether the successful detection of fearful faces would be accompanied by random discrimination of response correctness. 
Summary of findings
Objective perception
Until recently, it was largely assumed that, under masking conditions similar to those employed here, participants were unable to detect emotional target faces for durations of ∼30 ms. In a recent report (Pessoa et al., 2005), we challenged this view given that, in our masking study, 7 of 11 subjects were able to reliably detect 33-ms fearful-face targets; in addition, 2 of 11 subjects were able to reliably detect 17-ms-target faces (see also Maxwell & Davidson, 2004). The present findings confirm and extend our previous results. Indeed, when the data from multiple experimental sessions were pooled together (Experiments 1 and 2), all but one subject exhibited above-chance detection of fearful faces for all stimulus durations, including 17- and 25-ms targets. For 17-ms targets, it is important to note, however, that when single sessions of Experiment 1 were considered in isolation, Aobjective′ values often failed to differ significantly from 0.5 (see further discussion in the Null sensitivity and statistical power section). 
Subjective perception
Our results revealed that subjects not only could objectively detect fearful faces at short stimulus durations but also demonstrated above-chance subjective performance. Specifically, all subjects in Experiment 1 exhibited reliable, correct versus incorrect discrimination for durations of 33 and 41 ms. In addition, most subjects exhibited above-chance discrimination for 25-ms targets and several subjects exhibited subjective sensitivity even for very brief 17-ms targets ( Experiments 1 and 2). Taken together, our results demonstrate that sensitivity for briefly presented fearful faces is not limited to detection tasks but that it can be demonstrated for subjective perception as well. 
Individual differences
Although our experiment was not designed to investigate individual differences, our results revealed that subjects varied considerably in their ability to detect fearful faces and to perform correct/incorrect discriminations. For example, in Experiment 1, Subjects S2 and S3 were better at detecting fearful faces than the remaining subjects. In fact, objective measures from all individual sessions of Subject S2 were statistically significant even after correcting for multiple comparisons; the same is true for the last two sessions of Subject S3. 
Null sensitivity and statistical power
Naturally, establishing null sensitivity for any measure of perception faces extreme methodological difficulties (Macmillan, 1986; Reingold & Merikle, 1988). In the present study, we attempted to mitigate this problem by having subjects perform a relatively large number of trials across multiple experimental sessions (i.e., much larger than related studies). In the present section, we focus on the results of Experiment 1 for which single- and multiple-session results were investigated more closely; comparable points can be made in the context of Experiment 2. For objective perception, we did not find evidence for null sensitivity even for the shortest duration of 17 ms. However, in four cases, we obtained evidence for subjective null sensitivity: 17 ms for Subjects S1 (when a correction for multiple comparisons was employed), S4, and S5; 25 ms for Subject S1. 
How should these results be interpreted? We tentatively suggest that the A subjective′ values of 0.51 (Subject S4, 17 ms) and 0.52 (Subject S1, 25 ms) do, in fact, reflect null sensitivity ( Experiment 1). In these cases, it is extremely unlikely that performing additional sessions (for instance, doubling the number of sessions) would have allowed us to detect true sensitivity at a reasonable level of statistical power (see Figure 7). At the same time, the value of 0.54 (Subjects S1 and S5, 17 ms) is less straightforward to interpret. The power analysis suggested that true sensitivity at a level of 0.54 would require approximately eight sessions (which is not too far from what we employed) at an 80% level of power (see dashed line in Figure 7). Thus, we may have failed to reject null sensitivity because of relatively low statistical power (from our plot, we seemed to have had approximately 60% statistical power). 
The above power considerations assumed that a standard α level of .05 was employed. However, in our experiment, we deemed necessary to correct p values given that a sizeable number of simultaneous comparisons was made. In this context, the probability of making at least one Type I error would be [1 − (1 − α) C], where C is the number of comparisons. Such probability would assume unacceptable values when, for instance, the 40 comparisons of Table 2 were made ( p = .87); that is, at least one Type I error would have occurred with “near certainty.” These considerations suggest that using an α level of .05 for power calculations may be “too liberal” and that the number of trials indicated by the power analysis may be an “underestimate.” 
In summary, as extensively discussed in the past, establishing null sensitivity is surrounded by great methodological difficulties (Macmillan, 1986; Reingold & Merikle, 1988). The present results suggest that subjective null sensitivity occurred for two subjects (Subject S1, 25 ms; Subject S4, 17 ms) and, possibly, for a third subject (Subject S5, 17 ms). Our results also suggest that single-session measures from the present study, or from other studies in the literature, need to be interpreted carefully because they may have been underpowered to detect true sensitivity. 
Learning effects
The five-session format of Experiment 1 allowed us to investigate potential changes in task performance as a function of session. Such “learning effects” were observed in a more reliable manner for objective measures; for subjective measures, only a statistical trend was found. Because we characterized performance in terms of SDT instead of, for instance, a percent-correct index, it is likely that increases in sensitivity reflected enhancements in the ability to detect fear per se. Such learning may be conceptualized as increasing the distance between the signal and noise distributions as subjects performed more trials. Although further experiments are needed to investigate this question, it is possible that performance improvements relied on the ability to focus on, or more effectively employ, discriminative facial features involving the eyes, the mouth, or both, which are known to be important features for the perception of emotional expressions (Adolphs et al., 2005; Gold, Sekuler, & Bennett, 2004; Schyns, Bonnar, & Gosselin, 2002; Schyns & Oliva, 1999; Spezio, Adolphs, Hurley, & Piven, 2007). 
Experimental logic revisited: What do potential dissociations mean in terms of visual awareness?
A common experimental logic adopted in awareness studies has been to seek to demonstrate a dissociation between two measures of perception (Merikle et al., 2001; Snodgrass, 2004; Snodgrass et al., 2004). In our experiment, subjects performed a single task for which they were required to make detection and discrimination evaluations. Our findings were consistent with a dissociation between these two measures of fear perception such that, for some subjects, the successful detection of fearful faces was accompanied by random discrimination of response correctness. Such dissociation was supported by converging analyses of our data, including power analysis (Figure 7) and simulation results (Figure 9). Figure 10 illustrates the potential relationship between the critical durations associated with objective and subjective measures. The shorter duration represents the critical duration for reliably detecting a fearful-face target, and the longer duration represents the critical duration for reliably discriminating between correct and incorrect performance. The shaded area indicates a potential dissociation zone in which the subject is objectively above threshold but subjectively below threshold. The critical question that we next turn to is the following: What does such dissociation zone imply? There are at least three answers to this question. 
Figure 10
 
Dissociation between objective and subjective perception measures. Illustration of perception dissociation in which the objective threshold (OT) is smaller than the subjective threshold (ST). The dissociation zone refers to stimulus durations for which the subject would be able to reliably detect a fearful-face target but for which the subject could not reliably discriminate between correct and incorrect responses. The arrows around the duration values suggest that, when a dissociation occurs, the values will vary for different subjects. In addition, the arrows call attention to the fact that the OT and the ST were not determined in the present experiment given that a fixed set of durations was employed. The situation depicted in the graph could represent, for example, objective and subjective perception for Subject S4 in Experiment 1 (see Figure 4 and Table 2). A possible interpretation of our results is that a dissociation zone reveals that the subject would be subjectively unaware but objectively aware of the stimulus. However, other interpretations that do not imply awareness dissociations are also possible (see text). Abbreviations: OU = objectively unaware; SU = subjectively unaware.
Figure 10
 
Dissociation between objective and subjective perception measures. Illustration of perception dissociation in which the objective threshold (OT) is smaller than the subjective threshold (ST). The dissociation zone refers to stimulus durations for which the subject would be able to reliably detect a fearful-face target but for which the subject could not reliably discriminate between correct and incorrect responses. The arrows around the duration values suggest that, when a dissociation occurs, the values will vary for different subjects. In addition, the arrows call attention to the fact that the OT and the ST were not determined in the present experiment given that a fixed set of durations was employed. The situation depicted in the graph could represent, for example, objective and subjective perception for Subject S4 in Experiment 1 (see Figure 4 and Table 2). A possible interpretation of our results is that a dissociation zone reveals that the subject would be subjectively unaware but objectively aware of the stimulus. However, other interpretations that do not imply awareness dissociations are also possible (see text). Abbreviations: OU = objectively unaware; SU = subjectively unaware.
Subjective unawareness with objective awareness
In the past, proponents of the importance of subjective measures of awareness have suggested that such dissociations be interpreted as evidence for unaware perception. In the present context, because participants could successfully detect fearful faces but were unable to reliably discriminate response correctness, they would be subjectively unaware of stimuli that were linked to above-chance detection performance (i.e., objectively aware). 
Historically, the importance of subjective measure of awareness has been that it resonates with the idea that an intuitively appealing definition of awareness should be based on introspective reports of individuals' inner states (James, 1890). As argued by Kunimoto et al. (2001), the present measure of subjective perception would be phenomenologically valid because it depends crucially on the subjects' own introspective assessment of their performance. According to this view, the subjects' confidence cannot reflect accuracy unless they are partially aware of the information on which they based their discriminative responses, that is, “high” vs. “low” confidence (Kunimoto et al., 2001). At the same time, under conditions in which subjects have no awareness of the stimuli, their confidence ratings should be unrelated to accuracy. 
It is well known that when awareness is probed via verbal reports, both detection- and discrimination-type tasks are heavily influenced by individual strategies (Eriksen, 1960). It is not surprising, therefore, that many previous studies of awareness based on verbal reports have been strongly criticized in the past (Eriksen, 1960; Holender, 1986). In this article, we sought to analyze subjective perception in a manner that would be less susceptible to individual strategies than verbal reports. Our measure was based on early proposals of discrimination between correct and incorrect responses, which was originally named the Type 2 task (Clarke, Birdsall, & Tanner, 1959; Pollack & Decker, 1958), with the corresponding Type 2 SDT analysis. Such approach allowed us to define a perception measure that can be analyzed within the SDT framework, thereby reducing the response bias problem. 
Subjective measures likely reflect weakly conscious processes
Although the notion of a subjective threshold has been persuasive to many (Öhman, 1999; Merikle et al., 2001), others have argued that it may be problematic (e.g., Snodgrass & Shevrin, 2006). In particular, the SDT critique suggests that, in general, subjective threshold effects can be caused by weakly conscious perception (although below the subjective criterion). Strictly speaking, the SDT framework is “agnostic” regarding the awareness question because consciousness (or lack thereof) has no role in the theory. Accordingly, in the present context, our findings should not be considered as strong evidence for unaware processing because a subjective approach is inherently problematic. 
Objective and subjective measures index qualitatively different processes
In a sense, the interpretation of the present findings is largely dependent on the overall conceptual framework of the study. On the one hand, the objective SDT approach provides a parsimonious account of a large body of data without resorting to a “threshold” that provides a dividing line between consciousness and its absence (Macmillan & Creelman, 2005). Instead, the response criterion describes how subjects trade off sensitivity and specificity during a specific task (including task instructions and “inherent” biases). On the other hand, the subjective threshold approach argues that measures that index participants' “introspective” states are required if one wishes to investigate consciousness. 
In general, we propose that, instead of deciding which approach is “superior,” a more fruitful approach is to measure both objective and subjective measures of perception in trying to understand the nature of visual awareness. Indeed, regardless of the ultimate status of how objective and subjective measures map to awareness, we argue that both are important because they can index two distinct processes. In the context of the present findings, we suggest that the observed dissociation may reveal a qualitative dissociation in behavior. 
In addition, objective awareness may constitute a more elementary form of awareness than subjective awareness (Pessoa, 2005; Snodgrass & Shevrin, 2006). For instance, objective awareness may index phenomenal consciousness (experiential contents), whereas subjective awareness may index reflective or access consciousness (metacognitive processes involving the evaluation of phenomenal contents; see also Block, 1995, 2005). On a more experimental fashion, these two forms of awareness may index processes that differ with respect to control—for instance, objective awareness effects may not be amenable to response strategies, whereas subjective awareness effects may be (as in attempting to exclude certain types of responses; Snodgrass & Shevrin, 2006; see also Yonelinas, 2001). 
In concluding this section, we need to emphasize that the above discussion concerning a potential dissociation zone is limited to the specific context of our study. In other words, while we have illustrated the dissociation in terms of the duration of the target stimulus ( Figure 10), in general, several other factors need to be taken into consideration. In particular, stimulus visibility will strongly depend on parameters such as contrast, size, and spatial frequency of the target stimulus, in addition to spatial and temporal characteristics of the mask stimulus. Furthermore, while we treated our face stimuli set as uniform, there may have been variability among faces in terms of the expressiveness of fear that may have affected the results. Thus, further studies are needed to address how these and other factors contribute to dissociations such as those illustrated in Figure 10
Neuroimaging studies
Our study is also relevant in the context of neuroimaging studies that have sought to probe the neural correlates of unaware perception (Liddell et al., 2005; Morris, Öhman, & Dolan, 1998; Pessoa et al., 2006; Phillips et al., 2004; Whalen et al., 1998). Some of these studies determined brain responses while subjects passively viewed briefly presented (30 ms), masked fearful faces. Because subjects reported not seeing the fearful faces upon debriefing, they were deemed unaware of the fearful faces (at least in a subjective sense). The present study revealed that all participants exhibited subjective sensitivity for 33-ms targets when perception was gauged in terms of their ability to discriminate between correct and incorrect responses (via their response confidence ratings). In addition, whereas previous studies did not control for differences in individual response strategies, our study attempted to mitigate the response bias problem by employing an index of subjective perception that could be analyzed within the SDT framework. In fact, we believe that our study is the first to probe subjective perception of masked fearful faces in a manner that attempts to more directly confront the response bias problem. 
Summary and implications
Our findings demonstrated that when the perception of minimally visible stimuli is assessed by employing a relatively large number of trials, subjects reliably exhibit objective sensitivity for fearful-face targets shown as briefly as 17 ms. Reliable subjective sensitivity is also obtained for 33-ms targets and, for some subjects, even for 17-ms targets. Our paradigm revealed a potential dissociation of fear perception according to objective and subjective criteria. By characterizing perception via ROC analyses, we showed that such dissociation can be assessed for each individual participant. The determination of a dissociation zone offers the possibility of investigating the physiological correlates (e.g., fMRI responses) of such parameter range, which may offer new insights into the neural bases of aware and unaware perception. 
Appendix A: SDT for the analysis of objective and subjective perception: Parametric ROC approach
For comparison, we also analyzed our results in terms of a parametric SDT approach ( Figures 5B and 5D), which assumed normal distributions for both signal and noise (Green & Swets, 1966; Macmillan & Creelman, 2005). Participants were deemed above chance when d′-type measures were significantly greater than 0. The p value for statistical significance was .05. 
For objective awareness, we employed the standard d′ that assesses the distance between the hit rate and the false alarm rate in standard-deviation units. The variance of d′ was approximated with the Gourevitch and Galanter (1967) formula. 
Subjective perception was also evaluated in terms of a parametric analysis (Clark et al., 1959; Macmillan & Creelman, 2005). Under the common assumption that the initial “fear”/“no fear” decision is unbiased, we can partition the decision axis in such as way that the subject would respond “high confidence” if x ≤ −k or x > k and respond “low confidence” if −kx < k, where k determines the cutoff points (Macmillan & Creelman, 2005). The subjective d′ index and the criterion value k are then determined by solving the following equations for the conditional probabilities of “hits” (p“hit”) and “false alarms” (p“false alarm”): 
p``hit''=p(|x|>k|x>0)=p(x>k)p(x>0)=Φ(d/2k)Φ(d/2)p``falsealarm''=p(|x|>k|x<0)=p(x<k)p(x<0)=Φ(d/2k)Φ(d/2),
(A1)
where Φ(z) is the normal cumulative distribution function evaluated at the z score (Macmillan & Creelman, 2005). As discussed in the Introduction, the definition of “hits” and “false alarms” in the context of subjective perception followed the partitioning of Table 1
The variance of d subjective′ was derived by employing a Taylor expansion with two variables, similar to derivations of the same–difference discrimination method (Bi, 2002): 
var(d)=var(p``hit'')b12(a1b1a2b2)2+var(p``falsealarm'')b22(a1b1a2b2)2,
(A2)
where a1 =
p``hit''d|k0,d0
, a2 =
p``falsealarm''d|k0,d0
, b1 =
p``hit''k|k0,d0
, and b2 =
p``falsealarm''k|k0,d0
are the partial derivatives with respect to the d′ index and the cutoff k
Acknowledgments
We thank the National Institutes of Mental Health for its support for this work (Award 1R01 MH071589 to L.P.). We also thank the reviewers for a number of suggestions that have improved this article. 
Commercial relationships: none. 
Corresponding author: Luiz Pessoa. 
Address: Department of Psychological and Brain Sciences, Indiana University, 1101 East Tenth Street, Bloomington, IN 47405, USA. 
References
Adolphs, R. Gosselin, F. Buchanan, T. W. Tranel, D. Schyns, P. Damasio, A. R. (2005). A mechanism for impaired fear recognition after amygdala damage,, Nature, 433, 68–72. [PubMed] [CrossRef] [PubMed]
Bi, J. (2002). Variance of d′ for the same–different method. Behavior Research Methods, Instruments, and Computers, 34, 37–45. [PubMed] [CrossRef]
Block, N. (1995). On a confusion about a function of consciousness. Behavioral and Brain Sciences, 18, 227–247. [Article] [CrossRef]
Block, N. (2005). Two neural correlates of consciousness. Trends in Cognitive Sciences, 9, 46–52. [PubMed] [CrossRef] [PubMed]
Bowers, R. M. Bowers, K. S. Meichelbaum, D. (1984). On being unconsciously influenced and informed. The unconscious reconsidered. (pp. 227–272). New York: John Wiley & Sons.
Clarke, F. R. Birdsall, T. G. Tanner, Jr., W. P. (1959). Two types of ROC curves and definition of parameters. Journal of the Acoustical Society of America, 31, 629–630. [CrossRef]
DeLong, E. R. DeLong, D. M. Clarke-Pearson, D. L. (1988). Comparing the areas under two or more correlated receiver operating characteristic curves: A nonparametric approach. Biometrics, 44, 837–845. [PubMed] [CrossRef] [PubMed]
Dolan, R. (2002). Emotion, cognition, and behavior. Science, 298, 1191–1194. [PubMed] [CrossRef] [PubMed]
Efron, B. Tibshirani, R. J. (1993). An introduction to the bootstrap. New York: Chapman & Hall.
Egan, J. P. (1975). Signal detection theory and ROC analysis. New York: Academic Press.
Ekman, P. Friesen, W. V. (1976). Pictures of facial affect. Palo Alto, CA: Consulting Psychologists Press.
Eriksen, C. W. (1960). Discrimination and learning without awareness: A methodological survey and evaluation. Psychological Review, 67, 279–300. [PubMed] [CrossRef] [PubMed]
Esteves, F. Öhman, A. (1993). Masking the face: Recognition of emotional facial expressions as a function of the parameters of backward masking. Scandinavian Journal of Psychology, 34, 1–18. [PubMed] [CrossRef] [PubMed]
Galvin, S. J. Podd, J. V. Drga, V. Whitmore, J. (2003). Type 2 tasks in the theory of signal detectability: Discrimination between correct and incorrect decisions. Psychonomic Bulletin & Review, 10, 843–876. [PubMed] [CrossRef] [PubMed]
Gold, J. M. Sekuler, A. B. Bennett, P. J. (2004). Characterizing perceptual learning with external noise. Cognitive Science, 28, 167–207. [CrossRef]
Gold, J. M. Murray, R. F. Sekuler, A. B. Bennett, P. J. Sekuler, R. (2005). Visual memory decay is deterministic. Psychological Science, 16, 769–774. [PubMed] [CrossRef] [PubMed]
Gourevitch, V. Galanter, E. (1967). A significance test for one parameter isosensitivity functions. Psychometrika, 32, 25–33. [PubMed] [CrossRef] [PubMed]
Green, D. M. Swets, J. A. (1966). Signal detection theory and psychophysics. New York: Wiley.
Hajian-Tilaki, K. O. Hanley, J. A. Joseph, L. Collet, J. P. (1997). A comparison of parametric and nonparametric approaches to ROC analysis of quantitative diagnostic tests. Medical Decision Making, 17, 94–102. [PubMed] [CrossRef] [PubMed]
Hanley, J. A. McNeil, B. J. (1982). The meaning and use of the area under a receiver operating characteristic (ROC curve. Radiology, 143, 29–36. [PubMed] [CrossRef] [PubMed]
Holender, D. (1986). Semantic activation without conscious awareness in dichotic listening, parafoveal vision, and visual masking: A survey and appraisal. Behavioral and Brain Sciences, 9, 1–23. [CrossRef]
Ishai, A. Pessoa, L. Bikle, P. C. Ungerleider, L. G. (2004). Repetition suppression of faces is modulated by emotion. Proceedings of the National Academy of Sciences of the United States of America, 101, 9827–9832. [PubMed] [Article] [CrossRef] [PubMed]
James, W. (1890). Principles of psychology. New York: Holt.
Kolb, F. C. Braun, J. (1995). Blindsight in normal observers. Nature, 377, 336–338. [PubMed] [CrossRef] [PubMed]
Kunimoto, C. Miller, J. Pashler, H. (2001). Confidence and accuracy of near-threshold discrimination responses. Consciousness and Cognition, 10, 294–340. [PubMed] [CrossRef] [PubMed]
Liddell, B. J. Brown, K. J. Kemp, A. H. Barton, M. J. Das, P. Peduto, A. (2005). A direct brainstem–amygdala–cortical ‘alarm’ system for subliminal signals of fear. Neuroimage, 24, 235–243. [PubMed] [CrossRef] [PubMed]
Macmillan, N. A. (1986). The psychophysics of subliminal perception. Behavioral and Brain Sciences, 9, 38–39. [CrossRef]
Macmillan, N. A. Creelman, C. D. (2005). Detection theory: A user's guide. New York: Cambridge University Press.
Maxwell, J. S. Davidson, R. J. (2004). Unequally masked: Indexing differences in the perceptual salience of “unseen” facial expressions. Cognition and Emotion, 18, 1009–1026. [CrossRef]
Merikle, P. M. Smilek, D. Eastwood, J. D. (2001). Perception without awareness: Perspectives from cognitive psychology. Cognition, 79, 115–134. [PubMed] [CrossRef] [PubMed]
Morris, J. S. Öhman, A. Dolan, R. J. (1998). Conscious and unconscious emotional learning in the human amygdala. Nature, 393, 467–470. [PubMed] [CrossRef] [PubMed]
Obuchowski, N. A. (1994). Computing sample size for receiver operating characteristic studies. Investigative Radiology, 29, 238–243. [PubMed] [CrossRef] [PubMed]
Öhman, A. Dalgleish, T. Power, M. (1999). Distinguishing unconscious from conscious emotional processes: Methodological considerations and theoretical implications. Handbook of cognition and emotion. (pp. 321–352). Chichester: John Wiley & Sons.
Öhman, A. (2002). Automaticity and the amygdala: Nonconscious responses to emotional faces. Current Directions in Psychological Science, 11, 62–66. [CrossRef]
Pessoa, L. (2005). To what extent are emotional visual stimuli processed without attention and awareness? Current Opinion in Neurobiology, 15, 188–196. [PubMed] [CrossRef] [PubMed]
Pessoa, L. Japee, S. Ungerleider, L. G. (2005). Visual awareness and the detection of fearful faces. Emotion, 5, 243–247. [PubMed] [CrossRef] [PubMed]
Pessoa, L. Japee, S. Sturman, D. Ungerleider, L. G. (2006). Target visibility and visual awareness modulate amygdala responses to fearful faces. Cerebral Cortex, 16, 366–375. [PubMed] [Article] [CrossRef] [PubMed]
Phillips, M. L. Williams, L. M. Heining, M. Herba, C. M. Russell, T. Andrew, C. (2004). Differential neural responses to overt and covert presentations of facial expressions of fear and disgust. Neuroimage, 21, 1484–1496. [PubMed] [CrossRef] [PubMed]
Pollack, I. Decker, L. R. (1958). Confidence ratings, message reception, and the receiver operating characteristic. Journal of the Acoustical Society of America, 30, 286–292. [CrossRef]
Pollack, I. Hsieh, R. (1969). Sampling variability of the area under the ROC-curve and of d′e. Psychological Bulletin, 71, 161–173. [CrossRef]
Reingold, E. M. Merikle, P. M. (1988). Using direct and indirect measures to study perception without awareness. Perception & Psychophysics, 44, 563–575. [PubMed] [CrossRef] [PubMed]
Schyns, P. G. Bonnar, L. Gosselin, F. (2002). Show me the features! Understanding recognition from the use of visual information. Psychological Science, 13, 402–409. [PubMed] [CrossRef] [PubMed]
Schyns, P. G. Oliva, A. (1999). Dr Angry and Mr Smile: When categorization flexibly modifies the perception of faces in rapid visual presentations. Cognition, 69, 243–265. [PubMed] [CrossRef] [PubMed]
Snodgrass, M. (2004). The dissociation paradigm and its discontents: How can unconscious perception or memory be inferred? Consciousness and Cognition, 13, 107–116. [PubMed] [CrossRef] [PubMed]
Snodgrass, M. Bernat, E. Shevrin, H. (2004). Unconscious perception: A model-based approach to method and evidence. Perception & Psychophysics, 66, 846–867. [PubMed] [CrossRef] [PubMed]
Snodgrass, M. Shevrin, H. (2006). Unconscious inhibition and facilitation at the objective detection threshold: Replicable and qualitatively different unconscious perceptual effects. Cognition, 101, 43–79. [PubMed] [CrossRef] [PubMed]
Spezio, M. L. Adolphs, R. Hurley, R. S. Piven, J. (2007). Analysis of face gaze in autism using “Bubbles”. Neuropsychologia, 45, 144–151. [PubMed] [CrossRef] [PubMed]
Whalen, P. J. Rauch, S. L. Etcoff, N. L. McInerney, S. C. Lee, M. B. Jenike, M. A. (1998). Masked presentations of emotional facial expressions modulate amygdala activity without explicit knowledge. Journal of Neuroscience, 18, 411–418. [PubMed] [Article] [PubMed]
Yonelinas, A. P. (2001). Consciousness, control, and confidence: The 3 Cs of recognition memory. Journal of Experimental Psychology: General, 130, 361–379. [PubMed] [CrossRef] [PubMed]
Figure 1
 
Backward masking paradigm. Fearful, happy, or neutral target faces were shown for 17, 25, 33, or 41 ms and were immediately followed by a neutral face mask that was shown such that the target plus the mask lasted 100 ms. Subjects were first asked whether fear appeared at any time in the display. They were then asked to rate the confidence of their initial response by using a 1-to-6 scale (from low to high confidence).
Figure 1
 
Backward masking paradigm. Fearful, happy, or neutral target faces were shown for 17, 25, 33, or 41 ms and were immediately followed by a neutral face mask that was shown such that the target plus the mask lasted 100 ms. Subjects were first asked whether fear appeared at any time in the display. They were then asked to rate the confidence of their initial response by using a 1-to-6 scale (from low to high confidence).
Figure 2
 
Abstract representation of the tasks performed by the subject. The first task (top) corresponded to the standard detection task. During the observation interval, either a signal (S) or a noise (N) event occurs with probability p(S) and p(N), respectively. During the decision interval, the subject indicates “Yes, a signal occurred” or “No, a signal did not occur.” The observation and decision intervals of the detection task become the observation interval for the Type 2 task (bottom). In this case, either a correct (C) or incorrect (I) event occurs with probability p(C) and p(I), respectively. During the decision interval, the subject indicates “Yes, it was correct” or “No, it was not correct.” In the present experiment, detection signal trials comprised fear-containing trials (fearful–neutral target–mask pairs), and detection noise trials comprised non-target-containing trials (happy–neutral and neutral–neutral target–mask pairs). In addition, confidence ratings were used instead of “correct”/“incorrect” judgments (inspired by Figure 1 of Galvin et al., 2003).
Figure 2
 
Abstract representation of the tasks performed by the subject. The first task (top) corresponded to the standard detection task. During the observation interval, either a signal (S) or a noise (N) event occurs with probability p(S) and p(N), respectively. During the decision interval, the subject indicates “Yes, a signal occurred” or “No, a signal did not occur.” The observation and decision intervals of the detection task become the observation interval for the Type 2 task (bottom). In this case, either a correct (C) or incorrect (I) event occurs with probability p(C) and p(I), respectively. During the decision interval, the subject indicates “Yes, it was correct” or “No, it was not correct.” In the present experiment, detection signal trials comprised fear-containing trials (fearful–neutral target–mask pairs), and detection noise trials comprised non-target-containing trials (happy–neutral and neutral–neutral target–mask pairs). In addition, confidence ratings were used instead of “correct”/“incorrect” judgments (inspired by Figure 1 of Galvin et al., 2003).
Figure 3
 
Assessing subjective perception with SDT. Two event types corresponding to correct and incorrect trials are represented via Gaussian distributions. The criterion for reporting confidence is indicated by the vertical line and dictates when the subject reports “high” versus “low” confidence. A d′ value greater than zero reveals the subject's ability to discriminate between correct and incorrect responses and can be used to index subjective perception. In this article, we employed a nonparametric measure of sensitivity based on ROC curves (inspired by Figure 1 of Kunimoto et al., 2001).
Figure 3
 
Assessing subjective perception with SDT. Two event types corresponding to correct and incorrect trials are represented via Gaussian distributions. The criterion for reporting confidence is indicated by the vertical line and dictates when the subject reports “high” versus “low” confidence. A d′ value greater than zero reveals the subject's ability to discriminate between correct and incorrect responses and can be used to index subjective perception. In this article, we employed a nonparametric measure of sensitivity based on ROC curves (inspired by Figure 1 of Kunimoto et al., 2001).
Figure 4
 
ROC curves for two representative subjects. Top: Subject S3 exhibited above-chance objective and subjective sensitivity at all target durations. Bottom: Subject S4 exhibited above-chance objective sensitivity at all target durations and above-chance subjective sensitivity for the 25-, 33-, and 41-ms durations. For 17-ms targets, Subject S4 could detect fearful faces better than chance but failed to reliably discriminate between correct and incorrect responses. Such pattern of results is suggestive of a dissociation between objective and subjective perception (see Figure 10).
Figure 4
 
ROC curves for two representative subjects. Top: Subject S3 exhibited above-chance objective and subjective sensitivity at all target durations. Bottom: Subject S4 exhibited above-chance objective sensitivity at all target durations and above-chance subjective sensitivity for the 25-, 33-, and 41-ms durations. For 17-ms targets, Subject S4 could detect fearful faces better than chance but failed to reliably discriminate between correct and incorrect responses. Such pattern of results is suggestive of a dissociation between objective and subjective perception (see Figure 10).
Figure 5
 
Objective and subjective measures of perception. Values of A objective′ (blue) and A subjective′ (red) are plotted as a function of subject for Experiment 1 (A and B) and Experiment 2 (C and D). (A and C) Perception measures based on the nonparametric analysis. (B and D): Perception measures based on the parametric analysis. The arrows indicate conditions for which there was evidence for a dissociation between objective and subjective perception. See text for further details. Error bars indicate 95% bootstrap confidence intervals.
Figure 5
 
Objective and subjective measures of perception. Values of A objective′ (blue) and A subjective′ (red) are plotted as a function of subject for Experiment 1 (A and B) and Experiment 2 (C and D). (A and C) Perception measures based on the nonparametric analysis. (B and D): Perception measures based on the parametric analysis. The arrows indicate conditions for which there was evidence for a dissociation between objective and subjective perception. See text for further details. Error bars indicate 95% bootstrap confidence intervals.
Figure 6
 
Objective and subjective perception measures as a function of noise trials. Objective measures are shown in blue, and subjective measures are shown in red. Top: Results for 17-ms targets based on happy–neutral noise trials (left) or based on neutral–neutral noise trials (right). Bottom: Results for 25-ms targets based on happy–neutral noise trials (left) or based on neutral–neutral noise trials (right). Error bars indicate 95% bootstrap confidence intervals.
Figure 6
 
Objective and subjective perception measures as a function of noise trials. Objective measures are shown in blue, and subjective measures are shown in red. Top: Results for 17-ms targets based on happy–neutral noise trials (left) or based on neutral–neutral noise trials (right). Bottom: Results for 25-ms targets based on happy–neutral noise trials (left) or based on neutral–neutral noise trials (right). Error bars indicate 95% bootstrap confidence intervals.
Figure 7
 
Statistical power curves. The number of trials is plotted as a function of the critical value of A′ at several levels of statistical power ( α = .05, two-tailed test). Thus, for example, to detect a true sensitivity of A′ = 0.6 with 80% power, approximately 300 (∼2 sessions) trials would be required. To detect a true sensitivity of A′ = 0.54 (vertical dashed line) with 80% power, approximately 1,300 (∼8 sessions) trials would be required.
Figure 7
 
Statistical power curves. The number of trials is plotted as a function of the critical value of A′ at several levels of statistical power ( α = .05, two-tailed test). Thus, for example, to detect a true sensitivity of A′ = 0.6 with 80% power, approximately 300 (∼2 sessions) trials would be required. To detect a true sensitivity of A′ = 0.54 (vertical dashed line) with 80% power, approximately 1,300 (∼8 sessions) trials would be required.
Figure 8
 
Learning effects. Sensitivity measures are plotted as a function of session number (as a ratio relative to Session 1 values). Top: Objective values increased across sections (linear trend, p < .01). Bottom: Subjective values showed some evidence of increase across sessions, but the results were not significant (linear trend, p = .07).
Figure 8
 
Learning effects. Sensitivity measures are plotted as a function of session number (as a ratio relative to Session 1 values). Top: Objective values increased across sections (linear trend, p < .01). Bottom: Subjective values showed some evidence of increase across sessions, but the results were not significant (linear trend, p = .07).
Figure 9
 
Simulated and observed objective and subjective values. (A) Experiment 1. (B) Experiment 2. Simulated results are shown in blue, whereas specific data points are shown in red (17 ms) and green (25 ms). Evidence of stronger dissociation between objective and subjective measures is given by points that do not overlap with the prediction line but cross the horizontal “chance” line. Error bars indicate 95% bootstrap confidence intervals.
Figure 9
 
Simulated and observed objective and subjective values. (A) Experiment 1. (B) Experiment 2. Simulated results are shown in blue, whereas specific data points are shown in red (17 ms) and green (25 ms). Evidence of stronger dissociation between objective and subjective measures is given by points that do not overlap with the prediction line but cross the horizontal “chance” line. Error bars indicate 95% bootstrap confidence intervals.
Figure 10
 
Dissociation between objective and subjective perception measures. Illustration of perception dissociation in which the objective threshold (OT) is smaller than the subjective threshold (ST). The dissociation zone refers to stimulus durations for which the subject would be able to reliably detect a fearful-face target but for which the subject could not reliably discriminate between correct and incorrect responses. The arrows around the duration values suggest that, when a dissociation occurs, the values will vary for different subjects. In addition, the arrows call attention to the fact that the OT and the ST were not determined in the present experiment given that a fixed set of durations was employed. The situation depicted in the graph could represent, for example, objective and subjective perception for Subject S4 in Experiment 1 (see Figure 4 and Table 2). A possible interpretation of our results is that a dissociation zone reveals that the subject would be subjectively unaware but objectively aware of the stimulus. However, other interpretations that do not imply awareness dissociations are also possible (see text). Abbreviations: OU = objectively unaware; SU = subjectively unaware.
Figure 10
 
Dissociation between objective and subjective perception measures. Illustration of perception dissociation in which the objective threshold (OT) is smaller than the subjective threshold (ST). The dissociation zone refers to stimulus durations for which the subject would be able to reliably detect a fearful-face target but for which the subject could not reliably discriminate between correct and incorrect responses. The arrows around the duration values suggest that, when a dissociation occurs, the values will vary for different subjects. In addition, the arrows call attention to the fact that the OT and the ST were not determined in the present experiment given that a fixed set of durations was employed. The situation depicted in the graph could represent, for example, objective and subjective perception for Subject S4 in Experiment 1 (see Figure 4 and Table 2). A possible interpretation of our results is that a dissociation zone reveals that the subject would be subjectively unaware but objectively aware of the stimulus. However, other interpretations that do not imply awareness dissociations are also possible (see text). Abbreviations: OU = objectively unaware; SU = subjectively unaware.
Table 1
 
Trial type categorization for subjective perception sensitivity (based on Kunimoto et al., 2001).
Table 1
 
Trial type categorization for subjective perception sensitivity (based on Kunimoto et al., 2001).
Discrimination accuracy Confidence
High Low
Correct Hit Miss
Incorrect False alarm Correct rejection
Table 2
 
Experiment 1: Objective and subjective sensitivity values ( A′): pooled data.
Table 2
 
Experiment 1: Objective and subjective sensitivity values ( A′): pooled data.
 

Note: p values are enclosed in parentheses.

 

Green: statistically significant, corrected for multiple comparisons; pink: statistically significant, not corrected for multiple comparisons; red: not statistically significant.

 

p values of .000 indicate values less than .001.

Table 3
 
Experiment 1: Objective and subjective sensitivity values ( A′): single-session data.
Table 3
 
Experiment 1: Objective and subjective sensitivity values ( A′): single-session data.
 

Note: p values are enclosed in parentheses.

 

Green: statistically significant, corrected for multiple comparisons; pink: statistically significant, not corrected for multiple comparisons; red: not statistically significant.

 

p values of .000 indicate values less than .001.

Table 4
 
Experiment 2: Objective and subjective sensitivity values ( A′): pooled data.
Table 4
 
Experiment 2: Objective and subjective sensitivity values ( A′): pooled data.
 

Note: p values are enclosed in parentheses.

 

Green: statistically significant, corrected for multiple comparisons; red: not statistically significant.

 

p values of .000 indicate values less than .001.

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×