Among the Two Kinds of Metacognitive Evaluation,Only One Is Predictive of Illusory Object Perception

Abstract

The relationship between expectation-induced hallucination proneness and self-confidence in performance was studied in a visual perception task. Participants were prompted either to recognize briefly shown faces as male or female or to rate the subjective vividness of a square surrounding the face. Importantly, in a few critical trials, the square was absent. Upon completion, participants rated their performance in the face recognition task; they were also asked whether they were sure that their estimation was correct. Out of 35 participants, 33 “hallucinated” on at least one trial, rating the square as visible when it was actually absent. Negative correlation between hallucination proneness and self-confidence in performance (metacognitive rating) was found: The more hallucinations a participant experienced, the less confident he/she was in his/her performance in the face recognition task. Most subjects underestimated their performance; higher ratings were also more accurate. Thus, higher hallucination proneness was associated with more inaccurate ratings of one’s own perception. However, confidence in self-ratings as measured by the second follow-up question was unrelated to both, hallucination proneness and self-confidence in performance, supporting the view that there is no unitary mechanism of metacognitive evaluations and extending this view to the domain of visual hallucinatory perception.

Keywords

visual perception metacognition self-evaluation illusion hallucination confidence expectation

Behavior is regulated not only by how environmental cues are perceived but also by how the perceiver self-evaluates his/her perception in terms of reliability. Cognitive processes pertaining to this domain are subsumed under the concept of metacognition (Bang et al., 2018; Flavell, 1979; Koriat, 2012; Metcalfe & Shimamura, 1994; Yeung & Summerfield, 2012). Generic descriptions of metacognitive processes refer to “thinking about thinking,” “thinking about awareness” (e.g., higher-order thought theory of consciousness; Rosenthal, 2005), “being aware of one’s awareness,” “knowing about knowing,” and so forth.

In cognitive psychology and psychophysics of perception and memory, an important distinction separates sensitivity (measured by objective correctness of detection, discrimination, or recognition as corrected for guessing) and metacognitive evaluation of how confident a person is about the veridicality of his/her behavioral reports concerning the stimuli (de Gardelle & Mamassian, 2014; Rahnev & Fleming, 2019; Shea & Frith, 2019; Spence et al., 2016; Zehetleitner & Rausch, 2013). It has been shown that the connection between objective behavioral sensitivity and subjective self-evaluation of it as measured by confidence ratings is by no means straightforwardly monotonic and/or linear and that our knowledge about this connection is far from definite (Fleming et al., 2010; Hanks & Summerfield, 2017; Koriat, 2012, 2018; Hauser et al., 2017; Rahnev & Denison, 2018; Rosenthal, 2019; Spence et al., 2018; Summerfield & Tsetsos, 2015). Nevertheless, understanding the dynamics of different types of metacognition is relevant to all aspects of perception research, as a metacognitive component is proposed to always be included in conscious perception (Shea & Frith, 2019).

Self-evaluations of performance operationalized by confidence measurements can be grouped into two general subtypes: evaluations of the efficacy or correctness of the ongoing or just-completed tasks and evaluation of one’s own capability for some cognitive function, including the ability to self-evaluate (Burns et al., 2016; Kleitman & Stankov, 2007; Stankov et al., 2012; see also Rosenberg et al., 1995, on global vs. specific self-esteem, and Fleming & Daw, 2017, on first-order and second-order models of self-evaluation). However, virtually all studies devoted to investigation of these varieties of confidence evaluation have used tasks of perception, memory, or problem-solving where the subjective experience refers to real, actually presented stimulation. We do not know of metacognitive confidence studies focused on so-called hallucinatory experiences where people perceive objects that are not actually present (e.g., Aru & Bachmann, 2017a; Aru et al., 2018; McKenzie et al., 2010; Partos et al., 2016; Powers et al., 2017; Tompkins et al., 2016). It is both theoretically interesting and practically relevant to find measurable correlates of such hallucinatory experiences; this might help to estimate the reliability of a given percept. Metacognitive confidence is one such correlate. There is a principal difference between the perception of actual stimuli (a process instigated and constrained by bottom-up sensory signals) and the perception-like experience of stimuli that are actually absent (an endogenous process). While recent research has gained knowledge about the interrelations between sensory and metacognitive aspects of perceptual behavior, we do not know whether the same regularities also similarly characterize hallucinatory experience.

This direction of research seems timely as hallucinations—including when observed in the neurotypical population—are purportedly explained in the context of the currently prevailing predictive coding approach. Predictive coding models of perception suggest that perception is determined by the relative weighting of sensory data and expectations (Clark, 2015; Friston, 2005; Hohwy, 2013; O’Callaghan et al., 2017). Depending on this relative weighting, the subjective perception depends more on actual sensory evidence or on prior knowledge; prior knowledge dominates perception when sensory data are degraded (Aru et al., 2012, 2016; Powers et al., 2017) or ambiguous (Sterzer et al., 2008; Weilnhammer et al., 2018) or when there is a strong prior (Kuhn & Rensink, 2016; Tompkins et al., 2016). Hallucinatory experiences speak about a disbalance between the effects of priors and actual sensory inflow data processing. While subjective clarity of perception-like hallucinations has been studied (Aru & Bachmann, 2017a; Aru et al., 2018) by subjective scales such as Perceptual Awareness Scale (PAS; Overgaard et al., 2006), metacognitive confidence in perceptual experiences, occasionally intermingled with hallucinatory experiences, has not been studied. However, metacognition can serve as a valuable marker of precision, and it is important to understand how internal precision estimates relate to metacognitive precision consciously awarded to perceptual decisions.

We formulated three hypotheses. The first hypothesis assumes that illusory perception exemplified by expectation based hallucinations has no correlation with confidence. (This is despite the fact that objective correctness of perception of actually presented stimuli is positively correlated with confidence in the clearly above-threshold range of objective performance—e.g., Bonder & Gopher, 2019; Fleming et al., 2010; Rahnev & Fleming, 2019; Rausch & Zehetleitner, 2018; Shen & Ma, 2019; however, see Koriat, 2018; McIntosh et al., 2019.) This hypothesis has two sources. First, as the research on metacognitive beliefs in patients prone to positive symptoms has produced inconsistent results (see Mintz & Alpert, 1972; Moritz et al., 2010), we opted for a null hypothesis. Second, there is some data showing that metacognitive differences (including self-certainty) do not correlate with positive symptoms such as hallucination proneness in psychosis prone samples (Cotter et al., 2017; Moritz et al., 2016; Wright et al., 2020).

The second hypothesis assumed that the higher one’s self-evaluation of his/her performance in the perceptual task, the higher one’s confidence in the correctness of this self-evaluation. This hypothesis was put forth based on common intuition and some data about a general self-confidence factor predictable by accuracy of performance in a variety of perceptual and other tasks (Bonder & Gopher, 2019; Burns et al., 2016; Faivre et al., 2018; Kleitman & Stankov, 2007). We predicted a connection between the subjective estimate of the level of correctness in the perceptual task and the level of general self-confidence about how reliable these former estimates were: We assumed that a person is more likely to be unsure of his/her performance estimation when the estimated level of performance was low, whereas confidence in self-evaluation increases with the estimated level of performance in the perception task.

Third, we aimed to extend previous research on the link between objective perceptual performance and subjective confidence in performance onto perceptual tasks involving hallucinatory experiences. In a 1981 study by Woodhead and Baddeley (cited on pages 184–185 of Baddeley, 2013), 100 participants were shown a mix of new and already presented faces and were asked to recognize which faces had been presented earlier, as well as rate how confident they were in the correctness of their responses. There was no relationship between performance and self-rating. We hypothesize that a similar pattern would emerge in relation to illusory perception.

Method

Apparatus and Participants

For the computer-controlled experimental display, a SUN CM751U CRT monitor with 1,024 × 768 pixels resolution and 100-Hz refresh rate was used. The experiments were programmed and run using custom-made Python scripts. In all experiments, the observers viewed the arrays from a distance of 80 cm.

Thirty-five participants (25 F, 10 M) took part in the experiment; their ages ranging from 18 to 62 (M = 25.9). The sample size was determined by practical constraints. All participants had normal or corrected-to-normal vision. The study was approved by the ethics committee of University of Tartu, and the experiment was carried out in compliance with national legislation and the Declaration of Helsinki.

Perception Task

The main experiment consisted of 4 blocks of 80 trials (altogether 320 trials). Each trial started with a fixation cross presented in the center of a gray screen for 1 second, followed by an endogenous-type arrow cue pointing to the left or right of the screen for 150 ms. The cue was valid for 80% of the trials and invalid for 20%. The stimulus was a peripherally presented male or female face with a neutral expression (1.8 degrees of visual angle), surrounded by the lines of a square-shaped figure (4 degrees of visual angle) slightly darker than the background (see Figure 1). The stimulus was presented on the left or right side of the screen for 100 ms, followed by an empty screen for 700 ms. Although eye movements were not directly controlled for, the participants were instructed to fixate on the small cross in the center of the screen and direct their attention (but not gaze) to the cued location. Also, presentation time equal to 100 ms is short enough to prevent typical saccades during stimulus on time. The average luminance of the background of visual stimuli was 15.8 cd/m².

Figure 1.

Illustration of the Stimulus and Response Screens Used in the Task.

On 90% of the trials, the task was to discriminate the gender of the face (male or female). The response keys were S (male) and K (female). On 10% of the trials, the participants’ task was to rate how clearly they perceived the square on a 4-point PAS (Overgaard et al., 2006). Here, the response keys were A, D, J, L, where “1” corresponded to no experience of the stimulus, “2” corresponded to brief glimpse of the stimulus but could not recognize what it was, “3” corresponded to an almost clear impression of the stimulus, and “4” corresponded to a clear impression of the stimulus.

Before starting the main experiment, participants practiced both tasks separately for 30 trials each. The contrast of the square was varied only during the practice trials of the visibility task to ensure that participants were using the subjective visibility scale as intended. In the main task, the luminosity of the background was 14.9 cd/m² and that of the square was 12.5 cd/m² (Michelson contrast 0.138).

On six critical trials (three in the valid and three in the invalid cue condition), no square was presented, although participants were asked to give a perceptual clarity rating. No critical trials were included for the first 80 trials to ensure that the participants were already used to the task (i.e., that they had, by default, built up an expectation about the stimuli appearing on the screen).

Metacognitive Confidence

Following the experiment, participants were given two multiple-choice questions. First, they were asked to rate their performance in the face recognition task by indicating on a 10-point scale (from 0%–10% to 90%–100%) roughly what percentage of the faces they believed to have identified correctly. Second, they were asked whether they were certain in their self-rating, to which they could answer either “yes” or “no.”

Data Analyses

In the current study, the tendency for illusory perception was operationalized as the clarity with which participants perceived the missing squares in the critical trials. For each participant, average clarity rating across the six critical trials was calculated and used in data analyses as the illusory perception score. We also conducted the analyses using the number of trials in which the participant rated the square as clearly or almost clearly visible as an indicator of illusory perception and obtained very similar results.

Participants’ self-rating of performance was used as the metacognitive confidence score for their perceptual task performance. The self-rating was measured by a range of 10% steps (0%–10%, 10%–20%, etc.); for the metacognitive confidence score, the midpoint of a given range was used (e.g., for 20%–30%, it was 25%).

For each participant, we also calculated a score reflecting the accuracy of his or her self-rating, by subtracting the metacognitive confidence score from the actual percentage of correct answers obtained in the perception task. The larger the score, the bigger the difference between self-rating and actual performance; thus, the less accurate the self-rating.

Data were analyzed using R software. The main analyses were performed using Pearson’s correlation analyses and paired t tests, supplemented by repeated measures analysis of variance.

Here, a remark is necessary for why signal detection theory (SDT, e.g., Macmillan & Creelman, 2004) and/or type-2 SDT (e.g., Morales et al., 2019)-based analyses of sensitivity and bias were not used. Essentially, this is predetermined by the ambiguity of what must be considered signal-in-noise if we are measuring subjective experiences of a nonpresented stimulus or feature (Bachmann, 2004; see also Abid, 2019; Rausch et al., 2018, on related matters). Technically, this can be subsumed either under the category of misses (the sensory reality with a missing stimulus in it is missed) or category of false alarms (reporting a stimulus where there was none). Similarly, a “detected” (i.e., “hallucinated”), but actually nonpresent signal could be categorized as noise that makes an obstacle for correct detection of nonpresence of the signal or rather as illusory signals generated by expectation. We propose that illusory perception as generated by expectations can be better understood as analogous to hallucinations rather than signal or noise from a strictly SDT perspective. For example, Luccio (2019) argues that SDT as a well-established methodology is highly pertinent for the “primary” perceptual processes, contrasting these with secondary processes taking place in difficult perceptual conditions and when judgments are required (the latter having been tried to understand in the context of the Bayesian approaches). As our experimental paradigm exactly corresponds to the latter variety, SDT may not be the main approach here.

Results

Confirming the suitability of the procedure for obtaining illusory perception of actually absent visual stimulation, 33 out of the 35 participants rated the square as visible in at least one of the critical trials, giving it a clarity rating of 2 or more. Seventeen participants reported illusory perception in more than 3 trials, of whom 7 experienced illusory perception in all six critical trials. Detailed information about the use of clarity ratings can be found in Table 1.

Table 1.

Numerical and proportional description of the use of different clarity ratings.

Clarity rating	Number of participants using the rating	Number of times the rating was used	Frequency of the rating (%)
1 (No experience of the stimulus)	28	89	42.4
2 (Brief glimpse of the stimulus but could not recognize what it was)	24	40	19
3 (Almost clear impression of the stimulus)	19	35	16.7
4 (Clear impression of the stimulus)	18	46	21.9

We analyzed whether the validity of the cue had an effect on the subjective clarity rating of the square in critical and noncritical trials. We observed a significant main effect of trial type (critical/noncritical) on visibility ratings, F(1, 34) = 34.52, p < .001, ges = 0.407, but neither an effect of cue condition, F(1, 34) = 0.02, p = .89, ges < 0.001, nor an interaction, F(1, 34) = 0.82, p = 0.37, ges = 0.004.

Reaction time (RT) did not differ significantly between the critical and noncritical trials, although there was a trend showing faster responding in critical trials, t(64) = –1.822, p = .07. RT in critical trials was unrelated to hallucination proneness (r = –0.03, p = .87).

We observed a significant positive correlation between the average clarity rating in critical trials and the average clarity rating in noncritical trials (r = 0.6, p < .001). In other words, participants who perceived the square more clearly when it was present also gave higher clarity ratings when it was absent. The average clarity rating in critical trials was lower than that in the noncritical trials, t(67) = –3.74, p < .001.

There was a significant negative correlation between hallucination proneness, as measured by the average clarity rating in the six critical trials, and self-confidence in performance, as measured by the self-rating (r = –0.57, p < .001; Figure 2). The more clearly a subject “perceived” nonexisting squares, the lower he/she rated his/her performance in the face recognition task. (Whether this may be a result of more attention divided to the square surrounding the face to the detriment of perception of face itself remains out of the scope of the present study.)

Figure 2.

Scatterplot of the Correlation Between Clarity Ratings of the Actually Absent Stimulus and Self-Ratings of the Correctness of One’s Own Performance in the Face Discrimination Task. Subjects more prone to experience hallucinations were less confident in their perception.

Most subjects underestimated their performance, making higher ratings more accurate. A significant positive correlation was found between hallucination proneness and the inaccuracy of self-rating (r = 0.55, p < .001): People with higher hallucination proneness had more inaccurate ratings of their own conscious perception (Figure 3).

Figure 3.

Scatterplot of the Correlation Between Clarity Ratings of the Actually Absent Stimulus and Inaccuracy of Self-Rating as Reflected by the Difference Between Self-Rating of Performance and Actual Percentage of Correct Answers in the Perception Task. Subjects more prone to experience hallucinations were less accurate in their self-ratings of performance.

The actual percentage of correct answers was weakly correlated with the self-rating of performance (r = 0.38, p = .02). No association was found between confidence in self-rating, as measured by the second follow-up question, and hallucination proneness, t(33) = –0.13, p = .9. There was also no association between the two questions measuring self-confidence, t(33) = 0.08, p = .94.

Discussion

Our results show that the veridicality of perception depends not only on how accurately one perceives actually presented visual objects but also on whether perception of objects not actually present—that is, a hallucinatory experience—has occurred. This means that when conscious perception is measured by some visibility scales and/or confidence ratings together with measures of objective performance, subjective evaluations and objective measures may depart not only because actual stimuli were misperceived but also because absent stimulation was mistakenly experienced as involving some “phantom” object(s). The overwhelming majority of our participants experienced at least some hallucinations as assessed by the PAS scale. Lack of behavioral sensitivity of a scale at its near-threshold end may stem from such illusory experiences. We recommend that experiments in this domain always include catch trials with absent stimuli to tease out the contribution of illusory/hallucinatory experiences in the results (Aru & Bachmann, 2017a; Aru et al., 2018; Bachmann, 2018). For models of consciousness mechanisms to be realistic, they must also model the emergence of hallucinations.

While almost all subjects experienced illusory perception, the critical trials differed from noncritical trials in some ways. First, although below the significance margin, RTs tended to be faster in critical trials. This weakens the possibility that evaluations of the contents of these trials were based on some delayed working-memory processes possibly prone to confabulation. Of course, specific experiment(s) are needed to check this conjecture, including the need to compare RTs of no-illusion and illusion trials. Second, actual stimuli were reported as having been perceived more clearly than were the clarity ratings in trials where subjects reported illusory perception. Therefore, we can infer that when illusory perception did occur, it was a different qualitative or quantitative experience compared with the perception of an actual stimulus. Here, bear in mind that the term clarity rating allows ambiguous use: In terms of different values of ratings, it is “quantitative,” but in terms of what are the criterion contents founding the ratings, different-level “quantitative” ratings can and likely are based on different qualitative contents of experience. One possibility to explain this observation stems from the microgenetic approach (Aru & Bachmann, 2017b; Bachmann, 2000). According to the microgenetic stance, conscious experience emerges gradually, whereby the clarity and detailedness of the experienced object increases over time within about 100 to 300 ms. Faster reactions and lower clarity in hallucinated trials may mean that when an expected object is hallucinated, its microgenetic cycle is “frozen” at an early stage with vague representational content (e.g., proto-object or “blob” stage without fine detail). This allows also fast responses. In trials with real objects experienced, microgenetic process will go further from the proto-object (indistinct) stage up to the fine-grain representation. As this process takes longer and ends with more detailed experience, the clarity rating is higher, and in some cases, the response also has longer latency.

We formulated three hypotheses about illusory (i.e., hallucinatory) and veridical experiences for the present study. The first hypothesis assumed that perception of actually absent stimuli has no correlation with confidence in perception. However, our research revealed a strong correlation, and the first hypothesis was overthrown—illusory perception was negatively correlated with confidence relating to task performance (type-1 confidence). In other words, participants who were less likely to perceive absent stimuli gave more confident self-ratings of performance. According to the microgenetic theory, this result can be expected if we assume that less distinct experiences in relatively frequent hallucinatory trials (due to the immature end of the microgenetic episode) offer more instances of underarticulated criterion contents for metacognitive evaluation. This leads to less confidence. On the other hand, for perceivers who rarely hallucinate, the comparative quality of perceptual experience of the actual stimulus and experience of the “empty” display location where the stimulus could be expected but is not present, is better distinguishable (i.e., without vague intermediate clarity stages). This in turn feeds confidence in visual discrimination. The latter assumption is supported by the well-known regularity whereby behavioral perceptual accuracy positively correlates with metacognitive confidence about task performance (Bonder & Gopher, 2019; Fleming et al., 2010; Rahnev & Fleming, 2019; Rausch & Zehetleitner, 2018; Shen & Ma, 2019).

There was no relationship between RT and level of illusory perception, meaning that the participants who gave higher clarity ratings did not differ from the participants who gave lower clarity ratings in terms of response speed. However, it cannot be ruled out that other types of response biases may have influenced the results. Subjects who tended to give higher clarity ratings in critical trials did so in noncritical trials as well. The relationship between high clarity ratings of illusory experiences and a low confidence rating might thus reflect an extreme response bias—that is, the tendency to answer in the extreme—of some subjects.

On the other hand, the consistency of clarity ratings could also mean that people who perceive existing objects more intensely experience more intense illusory perception as well. Research does seem to point to a strong association between mental imagery and perception, at least in the visual domain. It has been shown that visual imagery can have a functional effect on sensory processing akin to a weak form of visual perception, for example, facilitating the perception of a relevant stimulus (Pearson, 2019; Tartaglia et al., 2009). In addition, a lot of brain-imaging evidence demonstrates considerable overlap between brain areas involved in perception and imagery (Kosslyn & Thompson, 2003; St-Laurent et al., 2015; Todd et al., 2013, 2015), and experienced imagery vividness correlates with neural overlap between imagery and perception in the entire visual system (Dijkstra et al., 2019). If we consider expectation-based hallucinations as a form of imagery, it is natural to expect that a trait of relatively stronger involvement of imagery mechanisms also predicts stronger involvement of direct perception serving mechanisms. (However, see Fulford et al., 2018, on the possibility that perception and imagery mechanisms may in some ways work in opposite direction.)

In the current experiment, more confident ratings were also more accurate. Thus, the relationship between confidence and hallucination proneness could in fact be explained by the relationship between accuracy and hallucination proneness. Mintz and Alpert (1972) found that subjects who experienced auditory hallucinations at a clinical level were less accurate in their self-ratings of auditory perception than healthy subjects or schizophrenic subjects who did not experience hallucinations, though they tended to overestimate their performance as opposed to the current study. To tease out the individual contributions of accuracy and confidence, future studies should use tasks that are subjectively perceived as relatively easier compared with the task in the current study.

The second hypothesis assumed that the higher the self-rating of task performance (type-1 confidence), the greater one’s certainty in the self-rating (type-2 confidence). This hypothesis was also not confirmed. There was no association between the self-rating of task performance and confidence in one’s evaluative performance. This shows that a person can be very certain that only a small part of his/her evaluation in a perceptual task is true. Conversely, but similarly, a person can be uncertain whether his/her high self-rating of perception is accurate. In our research, the putative unique factor of self-evaluation as manifesting similarly in type-1 and type-2 confidence estimations was not supported. Metacognitive confidence seems to have different facets or subtypes depending on the self-dependent action for which its reliability or quality has to be evaluated. To put it in a different way—metacognitive evaluations of a subordinate metacognitive evaluation are not mutually predictive in terms of their level of confidence. On the other hand, if looked at from within the same level of metacognitive processes as directed at different tasks, commonality has been found. For example, Faivre et al. (2018) examined whether metacognition operates through shared, domain-general mechanisms, or through idiosyncratic, domain-specific mechanisms, but remained within the set of unimodal and intermodal perceptual tasks. They found that (a) metacognitive efficiency correlated between auditory, tactile, visual, and audiovisual tasks, (b) confidence in an intermodal task was best modeled by supramodal formats assuming integrated intermodal representations, and (c) confidence in correct responses involved similar electrophysiological markers for different modal tasks. The common denominator was associated with motor preparation preceding the perceptual judgment. Faivre et al. (2018) concluded that metacognitive domain-generality is based on supramodal confidence estimates and decisional signals shared across sensory modalities.

The aim of the third hypothesis was to see whether our results could extend the findings that there is no correlation between a person’s confidence in a memory and the objective accuracy of a memory (Baddeley, 2013) to the case of direct perception (involving hallucinations). This hypothesis was not confirmed—performance in the perception task was positively, though weakly, related to the self-rating of performance.

For an interim summary, it appears that in subjects more prone to hallucinations, we find less confidence in performance in a perception task involving the presentation of brief visual stimuli and that metacognitive evaluations of the accuracy of one’s own perception are not significantly related to metacognitive evaluations of one’s capacity for reliable type-1 evaluations.

The relationship between confidence and objectively measurable capability to give truthful reports about external stimulation is complicated and multifaceted (Abid, 2019; Burns et al., 2016; de Gardelle & Mamassian, 2014; Fleming et al., 2010; Fleming & Daw, 2017; Kleitman & Stankov, 2007; Koriat, 2012, 2018; McIntosh et al., 2019; Rahnev & Denison, 2018; Rahnev & Fleming, 2019; Rausch et al., 2018; Rosenthal, 2019; Shea & Frith, 2019; Spence et al., 2016, 2018; Stankov et al., 2012; Summerfield & Tsetsos, 2015; Zehetleitner & Rausch, 2013). The results of the present study add more aspects to consider when examining relations between objective task performance, metacognitive self-evaluation of task performance, and metacognitive self-evaluation of the capacity for reliable self-evaluation. More vivid nonveridical perceptual experiences (“hallucinations”) caused by associatively formed task-specific contextual expectations may be predictive of lowered metacognitive confidence levels in a neurotypical sample, but only when it comes to metacognitive evaluation of task performance. Precision of metacognitive evaluation of the preciseness of first-order, task directed evaluation appears independent of the propensity to hallucinatory experiences. The relative weight of expectation-related priors compared with actual sensory evidence (possibly carrying the error signal) in percept formation is hypothetically higher in hallucination-prone subjects. Correctness of responses in a perceptual task, on the contrary, requires better discrimination of actual stimulation and therefore a relatively weaker effect of prior expectations on each specific encounter with objects of perception. These considerations explain why propensity to hallucinate is related to objective perceptual task performance and its metacognitive evaluation. However, as the higher order metacognitive evaluation is not exclusively targeted at perceptual experience, but focuses on a nonperceptual aspect of behavior associated with self-referential and personality traits-related dispositions, the link to first-order evaluation is lost. As a bold theoretical proposition, we could say that first-order metacognitive evaluations in perceptual decision making are dependent primarily on the access to phenomenal consciousness (Block, 1995) and on the learned associations of perceptual experience with objectively evidenced behavioral success, whereas second-order metacognitive evaluations depend on reference to self-consciousness in its personality-trait aspect.

Our results also bear some applied aspects. For example, in a forensic context, exploring the confidence of a witness and the evidence in which a person is more or less confident can help evaluate the testimony and determine how to proceed with the investigation. Namely, ascertaining higher level metacognitive predispositions of an eyewitness may not be a good strategy, but assessing the metacognitive confidence with regard to a specific perceptual task performance could be of practical value. Of course, all conclusions made on the basis of the present research are obtained by examining a small group of people. To make generally sound conclusions, there is a need to continue similar research with larger samples and a highly relevant variety of perceptual objects.

The biggest limitation of the current study was the small number of critical trials per participant. However, this was necessary to build a strong prior of expectation, leading to more frequent reports of illusory perception. Second, we did not directly control the potential effect of eye movements. As the performance in the main task was well above chance in all participants, this indicates that participants were overall successful in maintaining gaze at fixation. Nevertheless, we cannot be confident that during critical trials participants did not exert more overt attention to the stimuli locations when the stimulus was no longer visible. Finally, we only included two measures of metacognitive confidence ratings at the end of the experiment. Future studies should aim to include trial-by-trial confidence measures as well as a general confidence questionnaire to explore in more detail the link between perceptual decisions and subjective confidence.

Footnotes

Acknowledgements

The authors are grateful to Renate Rutiku and Carolina Murd for their assistance at some stages of this work.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The present research benefited from institutional research support grant IUT20-40 (sub-account TSVPH14140I) from Estonian Science Agency.

ORCID iDs

Sandra Vetik

Talis Bachmann

References

Abid

(2019). Deflating inflation: The connection (or lack thereof) between decisional and metacognitive processes and visual phenomenology. Neuroscience of Consciousness, 2019, niz015.

Aru

Axmacher

Do Lam

A. T.

Fell

Elger

C. E.

Singer

Melloni

(2012). Local category-specific gamma band responses in the visual cortex do not reflect conscious perception. Journal of Neuroscience, 32, 14909–14914.

Aru

Bachmann

(2017a). Expectation creates something out of nothing: The role of attention in iconic memory reconsidered. Consciousness and Cognition, 53, 203–210.

Aru

Bachmann

(2017b). In and out of consciousness: How does conscious processing (d)evolve over time? Frontiers in Psychology, 8, 128. https://doi.org/10.3389/fpsyg.2017.00128

Aru

Rutiku

Wibral

Singer

Melloni

(2016). Early effects of previous experience on conscious perception. Neuroscience of Consciousness, 2016, niw004.

Aru

Tulver

Bachmann

(2018). It’s all in your head: Expectations create illusory perception in a dual-task setup. Consciousness and Cognition, 65, 197–208.

Bachmann

(2000). Microgenetic approach to the conscious mind. John Benjamins.

Bachmann

(2004). Inaptitude of the signal detection theory, useful vexation from the microgenetic view, and inevitability of neurobiological signatures in understanding perceptual (un)awareness. Consciousness and Cognition, 13, 101–106.

Bachmann

(2018). Excess of individual variability of priors prevents successful development of general models. Behavioral and Brain Sciences, 41, E224.

10.

Baddeley

(2013). Essentials of human memory (classic edition). Taylor & Francis/Psychology Press.

11.

Bang

J. W.

Shekhar

Rahnev

(2018). Sensory noise increases metacognitive efficiency. Journal of Experimental Psychology: General, 148, 437–452.

12.

Block

(1995). On a confusion about a function of consciousness. Behavioral and Brain Sciences, 18, 227–247.

13.

Bonder

Gopher

(2019). The effect of confidence rating on a primary visual task. Frontiers in Psychology, 10, 2674.

14.

Burns

K. M.

Burns

N. R.

Ward

(2016). Confidence—More a personality or ability trait? It depends on how it is measured: A comparison of young and older adults. Frontiers in Psychology, 7, 518. https://doi.org/10.3389/fpsyg.2016.00518

15.

Clark

(2015). Surfing uncertainty: Prediction, action, and the embodied mind. Oxford University Press.

16.

Cotter

Yung

A. R.

Carney

Drake

R. J.

(2017). Metacognitive beliefs in the at-risk mental state: A systematic review and meta-analysis. Behaviour Research and Therapy, 90, 25–31.

17.

de Gardelle, V., & Mamassian, P. (2014). Does confidence use a common currency across two visual tasks? Psychological Science, 25, 1286--1288.

18.

Dijkstra

Bosch

S. E.

van Gerven

M. A. J.

(2019). Shared neural mechanisms of visual perception and imagery. Trends in Cognitive Sciences, 23, 423–434.

19.

Faivre

Filevich

Solovey

Kühn

Blanke

(2018). Behavioral, modeling, and electrophysiological evidence for supramodality in human metacognition. The Journal of Neuroscience: The Official Journal of the Society for Neuroscience, 38, 263–277.

20.

Flavell, J. (1979). Metacognition and cognitive monitoring: A new area of cognitive-developmental inquiry. American Psychologist, 34, 906--911.

21.

Fleming

S. M.

Daw

N. D.

(2017). Self-evaluation of decision-making: A general Bayesian framework for metacognitive computation. Psychological Review, 124, 91–114.

22.

Fleming

S. M.

Whiteley

Hulme

O. J.

Sahani

Dolan

R. J.

(2010). Effects of category-specific costs on neural systems for perceptual decision-making. Journal of Neurophysiology, 103, 3238–3247.

23.

Friston

(2005). A theory of cortical responses. Philosophical Transactions of the Royal Society B: Biological Sciences, 360, 815–836.

24.

Fulford

Milton

Salas

Smith

Simler

Winlove

Zeman

(2018). The neural correlates of visual imagery vividness–An fMRI study and literature review. Cortex, 105, 26–40.

25.

Hanks

T. D.

Summerfield

(2017). Perceptual decision making in rodents, monkeys, and humans. Neuron, 93, 15–31.

26.

Hauser

T. U.

Allen

Rees

Dolan

R. J.

(2017). Metacognitive impairments extend perceptual decision making weaknesses in compulsivity. Scientific Reports, 7, 6614.

27.

Hohwy

(2013). The predictive mind. Oxford University Press.

28.

Kleitman

Stankov

(2007). Self-confidence and metacognitive processes. Learning and Individual Differences, 17, 161–173.

29.

Koriat

(2012). The self-consistency model of subjective confidence. Psychological Review, 119, 80–113.

30.

Koriat

(2018). When reality is out of focus: Can people tell whether their beliefs and judgments are correct or wrong? Journal of Experimental Psychology: General, 147, 613–631.

31.

Kosslyn

Thompson

W. L.

(2003). When is early visual cortex activated during visual mental imagery? Psychological Bulletin, 129, 723–746.

32.

Kuhn

Rensink

R. A.

(2016). The vanishing ball illusion: A new perspective on the perception of dynamic events. Cognition, 148, 64–70.

33.

Luccio

(2019). Limits of the application of Bayesian modeling to perception. Perception, 48, 901–917.

34.

Macmillan

N. A.

Creelman

C. D.

(2004). Detection theory: A user's guide. Lawrence Erlbaum.

35.

McIntosh

R. D.

Fowler

E. A.

Lyu

Della Sala

(2019). Wise up: Clarifying the role of metacognition in the Dunning-Kruger effect. Journal of Experimental Psychology. General, 148, 1882–1897.

36.

McKenzie

K. J.

Poliakoff

Brown

R. J.

Lloyd

D. M.

(2010). Now you feel it, now you don't: How robust is the phenomenon of illusory tactile experience? Perception, 39, 839–850.

37.

Metcalfe

Shimamura

A. P.

(Eds.). (1994). Metacognition: Knowing about knowing. MIT Press.

38.

Mintz

Alpert

(1972). Imagery vividness, reality testing, and schizophrenic hallucinations. Journal of Abnormal Psychology, 79, 310–316.

39.

Morales

Odegaard

Maniscalco

(2019). The neural substrates of conscious perception without performance confounds. In de Brigard

Sinnott-Armstrong

(Eds.), Anthology of neuroscience & philosophy. https://philpapers.org/rec/MORTNS-4

40.

Moritz

Balzan

R. P.

Bohn

Veckenstedt

Kolbeck

Bierbrodt

Dietrichkeit

(2016). Subjective versus objective cognition: Evidence for poor metacognitive monitoring in schizophrenia. Schizophrenia Research, 178, 74–79.

41.

Moritz

Peters

M. J.

Larøi

Lincoln

T. M.

(2010). Metacognitive beliefs in obsessive-compulsive patients: A comparison with healthy and schizophrenia participants. Cognitive Neuropsychiatry, 15, 531–548.

42.

O’Callaghan

Kveraga

Shine

J. M.

Adams

R. B.

Jr. Bar

(2017). Predictions penetrate perception: Converging insights from brain, behaviour and disorder. Consciousness and Cognition, 47, 63–74.

43.

Overgaard

Rote

Mouridsen

Ramsøy

T. Z.

(2006). Is conscious perception gradual or dichotomous? A comparison of report methodologies during a visual task. Consciousness and Cognition, 15, 700–708.

44.

Partos

T. R.

Cropper

S. J.

Rawlings

(2016). You don’t see what I see: Individual differences in the perception of meaning from visual stimuli. PLoS One, 11, e0150615.

45.

Pearson

(2019). The human imagination: The cognitive neuroscience of visual mental imagery. Nature Reviews Neuroscience, 20, 624–634. https://doi.org/10.1038/s41583-019-0202-9

46.

Powers

A. R.

Mathys

Corlett

P. R.

(2017). Pavlovian conditioning-induced hallucinations result from overweighting of perceptual priors. Science, 357, 596–600.

47.

Rahnev

Denison

R. N.

(2018). Suboptimality in perceptual decision making. Behavioral and Brain Sciences, 41, 1–66.

48.

Rahnev

Fleming

S. M.

(2019). How experimental procedures influence estimates of metacognitive ability. Neuroscience of Consciousness, 2019, niz009.

49.

Rausch

Hellmann

Zehetleitner

(2018). Confidence in masked orientation judgments is informed by both evidence and visibility. Attention, Perception, & Psychophysics, 80, 134–154.

50.

Rausch

Zehetleitner

(2018). Multiple statistical signatures of human confidence. bioRxiv. https://doi.org/10.1101/426635

51.

Rosenberg

Schooler

Schoenbach

Rosenberg

(1995). Global self-esteem and specific self-esteem: Different concepts, different outcomes. American Sociological Review, 60, 141–156.

52.

Rosenthal

(2005). Consciousness and mind. Clarendon Press.

53.

Rosenthal

(2019). Consciousness and confidence. Neuropsychologia, 128, 255–265.

54.

Shea

Frith

C. D.

(2019). The global workspace needs metacognition. Trends in Cognitive Sciences, 23, 560–571.

55.

Shen

W. J.

(2019). Variable precision in visual perception. Psychological Review, 126, 89–132.

56.

Spence

Mattingley

Dux

(2018). Uncertainty information that is irrelevant for report impacts confidence judgments. Journal of Experimental Psychology: Human Perception and Performance, 44, 1981–1994.

57.

Spence

M. L.

Dux

P. E.

Arnold

D. H.

(2016). Computations underlying confidence in visual perception. Journal of Experimental Psychology: Human Perception and Performance, 42, 671–682.

58.

St-Laurent

Abdi

Buchsbaum

B. R.

(2015). Distributed patterns of reactivation predict vividness of recollection. Journal of Cognitive Neuroscience, 27, 2000–2018.

59.

Stankov

Pallier

Danthiir

Morony

(2012). Perceptual underconfidence: A conceptual illusion? European Journal of Psychological Assessment, 28, 190–200.

60.

Sterzer

Frith

Petrovic

(2008). Believing is seeing: Expectations alter visual awareness. Current Biology, 18, R697–R698.

61.

Summerfield

Tsetsos

(2015). Do humans make good decisions? Trends in Cognitive Sciences, 19, 27–34.

62.

Tartaglia

E. M.

Bamert

Mast

F. W.

Herzog

M. H.

(2009). Human perceptual learning by mental imagery. Current Biology, 19, 2081–2085.

63.

Todd

R. M.

Ehlers

M. R.

Müller

D. J.

Robertson

Palombo

D. J.

Freeman

Levine

Anderson

A. K.

(2015). Neurogenetic variations in norepinephrine availability enhance perceptual vividness. Journal of Neuroscience, 35, 6506–6516.

64.

Todd

R. M.

Schmitz

T. W.

Susskind

Anderson

A. K.

(2013). Shared neural substrates of emotionally enhanced perceptual and mnemonic vividness. Frontiers in Behavioral Neuroscience, 7, 40. https://doi.org/10.3389/fnbeh.2013.00040

65.

Tompkins

M. L.

Woods

A. T.

Aimola

A. D.

(2016). The phantom vanish magic trick: Investigating the disappearance of a non-existent object in a dynamic scene. Frontiers in Psychology, 7, 950–950.

66.

Weilnhammer

V. A.

Stuke

Sterzer

Schmack

(2018). The neural correlates of hierarchical predictions for perceptual decisions. Journal of Neuroscience, 38, 5008–5021.

67.

Wright

Nelson

Fowler

Greenwood

(2020). Perceptual biases and metacognition and their association with anomalous self experiences in first episode psychosis. Consciousness and Cognition, 77, 102847.

68.

Yeung

Summerfield

(2012). Metacognition in human decision making: Confidence and error monitoring. Philosophical Transactions of the Royal Society of London Series B, Biological Sciences, 367, 1310–1321.

69.

Zehetleitner

Rausch

(2013). Being confident without seeing: What subjective measures of consciousness are about? Attention, Perception, & Psychophysics, 75, 1406–1426.