2.1 Subjective and objective measures

Let us begin by mentioning, from amongst the subjective measures most commonly discussed in this ongoing debate, the following three procedures (cf. Timmermans and Cleeremans Reference Timmermans, Cleeremans and Overgaard2015): Perceptual-Awareness Scale (PAS) (Ramsøy and Overgaard Reference Ramsøy and Overgaard2004), Confidence Ratings (CR) (Dienes et al. Reference Dienes, Altmann, Kwan and Goode1995), and Post-Decision Wagering (PDW) (Persaud et al. Reference Persaud, McLeod and Cowey2007).

PAS is based on an explicitly introspective report as regards quality of experience. One of the possible scenarios for the procedure (based on Wierzchoń et al. Reference Wierzchoń, Paulewicz, Asanowicz, Timmermans and Cleeremans2014) runs as follows: After a fixation cross is flashed up on a laptop monitor, a target stimulus is presented to the participant over near-threshold time durations (e.g., between 16 and 192 ms), followed by a 200 ms presentation of a mask. The target stimulus may be either quite simple (e.g., geometrical shapes, numbers, letters) or a more complex image (e.g., male or female faces), while the mask will be an unspecified image (e.g., various meaningless patterns). Immediately after the presentation of the stimulus and the mask, the participant is asked to identify the object (e.g., as male or female) and, after that, to rate the quality of his or her visual experience (e.g., using a scale structured into four consecutive grades: clear experience, almost clear experience, vague experience, and no experience). In most cases, reports are given by pressing the relevant buttons on the laptop keyboard: for example, in the context of the identification task, pressing “right arrow” may correspond to “female face” and “left arrow” to “male face,” whilst when it comes to quality-related ratings pressing number “1” may correspond to “clear experience,” “2” to “almost clear,” and so on.

Turning to the CR procedure, we may note that this can be performed under an identical scenario to that for PAS as described in the preceding text (Wierzchoń et al. Reference Wierzchoń, Paulewicz, Asanowicz, Timmermans and Cleeremans2014). However, the most important difference is that after completing an identification task, a participant is asked to rate their confidence in their own (identification-related) performance by choosing confidence levels (on a scale running between total guessing and knowing [outlined in Wierzchoń et al. Reference Wierzchoń, Paulewicz, Asanowicz, Timmermans and Cleeremans2014] in four consecutive steps, these being guessing, not confident, quite confident, and very confident). Although CRs are based on one’s judgment of one’s own performance, and aim to avoid explicit introspection, they may implicitly involve a relation to the quality of one’s own experience.

Meanwhile, PDW has, in turn, a great deal in common with the procedure for CR, the major difference being that, after completion of the identification task, instead of deciding about confidence the participant is asked to wager on how they will eventually assess their own performance with a specified amount of money (e.g., by opting for a 20, 40, 60, or 80 PLN stake). The initial aim of the procedure was to encourage participants—by appealing to their desire to gamble and maximize earnings—to give answers in cases where confidence was too low, and consequently to increase the sensitivity and exhaustiveness of the measure (especially for near-conscious perceptions). However, the results here have proved to be of debatable validity (due to the influence of the phenomenon of risk aversion). Most importantly, for the purposes of this article, the initial claim that PDW is a direct and yet also an objective measure of awareness has been subjected to powerful criticism (Seth Reference Seth2008a, Reference Seth2008b).

Currently, the most commonly encountered examples of so-called objective measures are furnished by different kinds of forced-choice discrimination or identification tasks. Again, the procedure may be performed under an identical scenario to that of the aforementioned cases: Coming after the fixation cross, a given target stimulus is presented to a participant over some near-threshold duration, followed by a mask and a forced-choice identification task (e.g., choosing between alternative options: male or female). Based on the objective number of correct and incorrect identifications, the measure may be calibrated to boundary conditions (e.g., using varied presentation times), such that performance is just above the level of chance. (Such boundaries with respect to objective sensitivity determine the parameter known as d’, as defined in Signal Detection Theory, a theoretical foundation for objective measures [Macmillan and Creelman Reference Macmillan and Creelman1991; Phillips Reference Phillips2016].) The original assumption behind objective measures of awareness is the so-called Worldly Discrimination Theory (Gaillard et al. Reference Gaillard, Vandenberghe, Destrebecqz and Cleeremans2006; Fu et al. Reference Fu, Fu and Dienes2008), according to which, if a person is able to discriminate between two stimuli (presented within sensitivity thresholds), they are aware of those stimuli. So, the objective approach implies that sensitivity to certain stimuli (as shown in the relevant behavior) correlates with awareness of those stimuli (or even may involve the very same process), which is quite debatable. (These issues have been described more extensively elsewhere [see Timmermans and Cleeremans Reference Timmermans, Cleeremans and Overgaard2015].)

It is worth noting that subjective and objective procedures currently coexist fruitfully in practice, as they make inspecting and correlating their results straightforward. Most often, the above-chance (objective) sensitivity obtained in the discrimination task (in the form of a sufficient number of correct identifications) is compared to some different subjective measures. Meanwhile, the most interesting cases are those that show a dissociation between subjective and objective measures: the so-called guessing criterion, where a participant reports “guessing” or “no experience” even while exhibiting a high level of accuracy in respect of the relevant discrimination task (potentially implying access to certain unconscious yet relevant information), and the zero-correlation criterion, where confidence or awareness ratings do not relate to accuracy (potentially implying exposure to certain conscious yet nonrelevant or biased information) (Dienes Reference Dienes2008; Dienes et al. Reference Dienes, Altmann, Kwan and Goode1995). A similar idea of applying signal-detection theory to subjective awareness ratings is explored in the so-called meta-d’ approach. (Just very roughly, as d’ is obtained by quantifying correct and incorrect discriminations, the meta-d’ parameter is obtained by quantifying correct and incorrect discriminations of one’s own correctness or incorrectness [see Maniscalco and Lau Reference Maniscalco and Lau2012, Reference Maniscalco, Lau, Fleming and Frith2014].)

2.2 Behavioral and phenomenological objections

With reference to subjective measures, Persuh (Reference Persuh2018) has claimed in a recent article that “[the] ‘subjective’ character of [such] measures is illusory and … subjective measures, like objective measures, estimate only performance on a discrimination task.” He relates this declaration explicitly to introspective report procedures, such as PAS, and also argues that “[n]o matter how the question is posed [for the hypothetical participant, John—author’s note], language does not give John any special powers to make his report more subjective.” Persuh justifies this from the perspective of a very broadly specified behavioral assumption to the effect that there is no relationship between someone’s report about their own experience and their actual experience other than certain behavioral correlations because “an organism can only report correlations and we can only measure task performance” and, “[b]y the same token, a parrot can learn a correlation between its visual experience and some arbitrary motor output.”

If Persuh’s position on this amounts to a “behavioral stance,” then we might easily move in the opposite direction and argue, from what might be said to be a “phenomenological stance,” that in the case of objective measures their objective character is illusory, because objective measures, like subjective ones, estimate only with reference to a (subjective) response criterion—and, moreover, that no matter how the question is posed, the language employed does not itself have any special power to make the answer more objective. In fact, such considerations have already been highlighted by several researchers, who have pointed out that even in the context of forced-choice decisions or identification tasks, a participant must, to a certain extent, use their own response criteria to make up their mind and arrive at an answer (Timmermans and Cleeremans Reference Timmermans, Cleeremans and Overgaard2015; Phillips Reference Phillips2016). The nature of such a decision process—whether it amounts to an unconscious or preconscious gathering of evidence or build-up in confidence—remains a matter for investigation.

Both behavioral and phenomenological objections to the subjective versus objective distinction as it figured in the context just outlined are, in this author’s view, misleading. Indeed, it will be argued in the following text that they rest on a misunderstanding of the very foundations of that which distinction amounts. Yet, at the same time, it must be admitted that such a line of argumentation points in the direction of some things that hold true for the measures in question. First, it indicates that both subjective and objective measures are in fact behavioral because they measure (directly) specific behaviors (e.g., button-pressing on a laptop keyboard, or verbal activity), and second, it reveals that both measuring procedures do require decisions on the part of participants, these being based on their processing/perception of stimuli presented during the experiment. These simple facts not only make the previously mentioned objections potentially plausible but also make the measures vulnerable to other well-known problems. The latter are exemplified in such questions as whether the behavior measured correlates with any form of consciousness, and whether, if so, it correlates only with consciousness (and not, e.g., with certain unconsciously learned response strategies), as well as whether the (subjective) response criteria are free of bias—or, at least, less susceptible to being biased, and so on (Phillips Reference Phillips2016). In general, questions about the validity, exhaustiveness, and exclusiveness of the measures in question (Timmermans and Cleeremans Reference Timmermans, Cleeremans and Overgaard2015) are very hard to answer unambiguously, and quite apart from all the methodological issues involved, this will anyway be the case for a comparatively more fundamental reason: Namely, that the very notion of consciousness is far from unambiguous. (The science of consciousness has not fixed any definite threshold or thresholds, and neither has it managed to construct any conception or theory of consciousness that has gained universal acceptance; thus, we remain uncertain as to which states should be counted as conscious and which as unconscious or less conscious. Ultimately, then, all one can do is seek to relate the preceding questions to some sort of operational definition of consciousness [cf. Jonkisz Reference Jonkisz2012, Reference Jonkisz2015].)

2.3 The methodological distinction

The present author’s position is that both of the preceding lines of argumentation (i.e., the behavioral and phenomenological objections) misconstrue the distinction between subjective and objective measures, and that they do so by overestimating both the character of the task involved in the procedure and the language used by participants there. It is not the difference in the tasks involved—between ratings of experience, on the one hand, and delivering discriminative answers based on sensory content, on the other—that is key to that distinction. Moreover, it is also not the difference in the language used when describing qualitative features of experience, on the one hand, and objective features of the stimulus, on the other. (Regarding linguistic influences on such measures, see also the work of Lyyra [Reference Lyyra2019].) At the same time, what we may note is that the (subjective/objective) character of the criterion used as an experimental threshold for consciousness is being underestimated or even totally neglected here, whereas it will be argued in the text that follows that this should in fact count as crucial where this distinction is concerned.

The subjectivity of putatively subjective measures is unequivocally justified by the very criteria according to which some given behavior (e.g., reporting on the quality of one’s own experience or on one’s own performance using button-pressing) is correlated with conscious or unconscious stimulus-processing (in the case of a binary measure), or with certain intermediate grades of consciousness (in the case of graded and continuous measuring scales). The criteria are subjective simply because all that counts here is the participant’s biased decision about whether their visual experience has been clear, almost clear, vague, or nil (in the PAS procedure), as revealed in the corresponding behavior. (The same goes for subjective performance judgments within the CR and PDW procedures, where subjectively biased decisions about confidence or wager stake are all that counts.)

Similarly, the criteria employed in what are claimed to be objective measures qualify as objective inasmuch as what counts there is just sheer performance or accuracy in respect of some discriminatory task: that is, the (objective) proportion of valid discernments to false alarms. Of course, the objection coming out of the “phenomenological stance” (described in the preceding text) might be that in this case even sheer performance is based on a subjectively biased response criterion (formed with reference to the perception of the stimulus by some participant/subject or other). The answer to this, though, is straightforward: One grants that this is true, but in objective measures both the sensitivity levels (with d’>0) and the personal (biased) response criteria (with criterion C) are controlled (in that no matter what the response criteria happen to be, the performance should anyway be above the level of chance). Therefore, what the objective procedure considers when deciding whether or not the perception of the stimulus was conscious is whether or not a participant was sensitive to the stimulus and whether or not he or she performed accurately.

Just as in the case of objective measures the phenomenological objection can be refuted, so in the case of subjective measures the behavioral objection can be so. According to the latter, the subjective character of subjective measures is illusory, as (like objective measures) they only estimate performance in certain discriminatory or forced-choice tasks presented in behavior. The answer, once again, is quite simple: One grants this, but the behavior presented in subjective measures (such as PAS, CR, or PDW) is based on personal, biased criteria (e.g., a choice between “clear experience” and “no experience,” or between “guessing” and “confident”), which are not controlled. Therefore, it may also be said that whereas in subjective measures a subjective threshold is being utilized, in objective measures it is an objective threshold (see Timmermans and Cleeremans Reference Timmermans, Cleeremans and Overgaard2015). The thresholds in question were already defined by Cheesman and Merkile (Reference Cheesman and Merikle1984), when they wrote that “[a] subjective threshold may be defined as the detection level at which subjects claim not to be able to discriminate perceptual information at better than a chance level, whereas an objective threshold is the detection level at which perceptual information is actually discriminated at a chance level.”

Consequently, what makes some measure subjective or objective is not the specific task, with some special form of reporting or language-use being involved alongside this (as the previously mentioned lines of argumentations seem to suppose). Rather, it is the criterion or threshold according to which some given behavior is correlated with conscious, almost conscious, barely conscious, or unconscious stimulus processing (depending on the procedure used). The subjectivity or objectivity of the criterion, or the threshold for consciousness implemented, cannot be questioned in the ways outlined in the preceding text, and so may serve as an unambiguous characterization of the measures in question. Thus:

• Measures will be objective when they utilize an objective threshold (a bias-free criterion for consciousness).

• Measures will be subjective when they utilize a subjective threshold (a biased criterion for consciousness).

Surprisingly, such a methodological or pragmatic formulation of the distinction is not commonly encountered in the current debate—a fact that is no doubt bound up with the existence of other more frequently adopted ones (see the following text).

2.4 The semantic distinction

Another way of defining the distinction between subjective and objective measures goes in the direction of differentiating between so-called outer states (stimulus features, worldly discriminations, etc.) and inner states (mental, experiential, perceptual). Here are a few examples:

In principle, such characterizations specify what the report required or requested in connection with some task outlined in relation to a given measure refers to (i.e., whether it should be directed toward inner or outer states). Hence, the distinction, when constructed along such lines, may be said to possess a semantic character, taking the following form:

• Measures are objective when they ask one to discriminate (and report) states of the world (e.g., features of the stimulus presented).

• Measures are subjective when they ask participants to introspect (and report) their perceptual states (e.g., experiential features).

Intuitively, the distinction seems fine. However, for at least two reasons, it is not as efficacious as the methodological one presented earlier. First, this is because it is very difficult to be sure about whether a participant is introspectively reporting inner states or, instead, discriminating outer states without introspection: We cannot even be certain whether these two processes are going on in ways that are separate or intermixed. (In this context, see also the so-called transparency thesis, according to which isolated awareness of features of one’s own experience is virtually impossible [e.g., Tye Reference Tye2014].) Admittedly, it has recently been proposed that the semantics of the PAS should be examined, and that only “experiential terms” such as clarity or intensity should figure in the task description (Lyyra Reference Lyyra2019). However, even with such a refined PAS, doubts may well remain as to whether taking due care over the semantics will suffice to ensure adequate receptivity to the participant’s phenomenology. Second, it is because when all subjective measures are taken into account, such a definition better fits the procedures involved when using a PAS, while being problematic in relation to both CR and PDW. By asking people to judge their own performance, either in terms of their own confidence in the latter or through their willingness to wager on it, such measures explicitly ask for neither the stimulus features nor the inner perceptual states to be reported—yet, implicitly, both are most likely involved. (Even so, opinions are divided on this: see, e.g., Seth [Reference Seth2008a, Reference Seth2008b], who includes PDW in objective measures, and Timmermans and Cleeremans [Reference Timmermans, Cleeremans and Overgaard2015], who treat both measures as subjective.) In fact, both CR and PDW may implicitly involve metacognitive judgments about both the identification process and one’s own performance within it. (See also the next part of this article.) In the present context, it would be more accurate to refer to them as subjective performance measures, and to PAS and similar procedures as subjective experience measures.

There is also a more fundamental reason for the semantic distinction’s being problematic. Whereas the criterion or threshold applied in the context of a given measure can be unambiguously identified as subjective or objective (as shown in the previous section), the task’s reference to inner or outer states can in both cases be interpreted in either direction. That is because the features of the presented stimulus alone can be both objective (e.g., location and shape of the object, its category or name) and, at least partly, subjective (e.g., color, appearance in the sense of seeming a certain way)—something that, of course, pertains to the classical (Lockean) distinction between primary and secondary qualities. Similarly, features of inner perceptual states can also be analyzed from both an objective and a subjective angle: The referential (or intentional) object of a given perceptual state (e.g., a human face or a certain geometrical shape flashed up on a screen) is more or less objective, whereas the qualitative features of perceptions are subjective (e.g., intensity or clarity of the image involved). Such comments must suffice here, as a fuller analysis of this issue would require one to delve into different philosophical theories of perception—something not essential to the aims of this text.

3.1 Metaphysical directness

The subjective character of conscious experience, in the sense of its privacy or first-person-based accessibility, is undeniable; moreover, it is commonly held to be the aspect that differentiates consciousness from other phenomena known to science. As a basis for such theoretical concepts as phenomenal consciousness (Block Reference Block1995) and the hard problem of consciousness (Chalmers Reference Chalmers1995, Reference Chalmers1996), the subjectivity of experience has ultimately pushed many contemporary philosophers of mind in the direction of endorsing some otherwise quite unpopular theoretical and/or metaphysical outcomes, such as dualism or panpsychism, often combined with antireductionism or mysterianism (Jackson Reference Jackson1982; McGinn Reference McGinn1989; Block Reference Block1995; Chalmers Reference Chalmers1995, Reference Chalmers1996; Searle Reference Searle2000). It seems that the issue also arises in discussions about the directness of awareness measures, where some prominent scientists are inclined to follow Searle’s declaration regarding the subjective ontology of consciousness:

If, as Seth claims, behavioral measures cannot be direct because the contents of consciousness are ontologically subjective, what kind of directness is being invoked here? He is quite scrupulous on this point, and states that a “‘direct measure’ is one that transparently reflects its target property, as a ruler directly measures length” (Seth Reference Seth2008b). If this is supposed to mean that a direct measure of consciousness would measure consciousness as if the subjectivity barrier had disappeared, with full access to conscious contents “from the outside” as if “from the inside,” then it has to be admitted that such an idealized scenario is of course impossible, and Seth is right. Yet such a barrier-crossing, metaphysical understanding of directness may be also accused of fallaciousness. Such an accusation would be very similar to those leveled against the majority of idealistic advocates of the so-called knowledge argument (formulated in many versions by, e.g., Broad, Feigl, Nagel, and Jackson) or explanatory gap postulate (coined by Levine), insisting that science should provide direct knowledge of other subjects’ experiences. The belief that science has only resolved the hard problem of consciousness when it permits us to know (directly) what it is like to see in color, or “what it is like to be a bat” (or any other being different from ourselves), is fallacious when construed literally, in that it involves, at the very least, a category error: Scientific (objective) explanations just are categorically different from (subjective) conscious experiences (Pigliucci Reference Pigliucci2013; Jonkisz Reference Jonkisz2016). And if indeed such a formulation of the hard problem “does not require a solution, but rather, a cure” (Edelman et al. Reference Edelman, Gally and Baars2011, 5), then the same goes for metaphysically construed directness. Measures just are located in a different category from conscious contents, so claiming that “the fact that consciousness is ontologically subjective precludes direct behavioral access to conscious content” (Seth Reference Seth2008b) is either tautological or (indirectly) fallacious. In this context, one may even reasonably ask whether there is anything at all arising within psychology or the philosophy of mind that could be directly measured in behavior.

As an aside, it is perhaps worth adding that when it comes to discussing measures of consciousness, a more conservative approach to the entertaining of metaphysical assumptions about consciousness would seem to be the safer option. Claims about the subjective ontology of consciousness, adopted from Searle (Reference Searle2000) and other philosophers, are very strong, and straightforwardly entail all the difficult metaphysical consequences mentioned in the preceding text, which may turn out to be needlessly problematic. Trying (once again) not to delve into wider philosophical discussions any more than is absolutely necessary, it may hopefully suffice just to point out that to justify the thesis of the privacy of experience one is not obliged to endorse a subjective ontology of consciousness, as a weaker position positing only a subjective epistemology will be sufficient for this—while still enabling one to avoid those perplexing consequences (Northoff and Musholt Reference Northoff and Musholt2006). According to epistemological subjectivism, the contents of consciousness are accessible or cognizable only from the first-person-based point of view (i.e., only as experienced). This position does not automatically entail consciousness’s existing only as experienced (see Northoff and Musholt Reference Northoff and Musholt2006). It therefore enables one to furnish a justificatory basis for explaining the mechanisms of consciousness—or even for their reduction to neurophysiology. Such a position also fits well with the multidimensional character of consciousness recently emphasized by some researchers (Bayne et al. Reference Bayne, Hohwy and Owen2016; Jonkisz et al. Reference Jonkisz, Wierzchoń and Binder2017), as a multidimensional consciousness may be both epistemologically subjective (in respect of content accessible from a first-person-based standpoint) and ontologically objective (in that it exists as a complex biological phenomenon). Moreover, epistemological subjectivity may be naturalized—in terms that make reference to individuating differences. (For the hypothesis of biological individuation, see Jonkisz [Reference Jonkisz2015, Reference Jonkisz2016].)

3.2 Methodological directness

One specific sense for the notion of directness we are concerned with is defined in Reingold and Merikle (Reference Reingold and Merikle1988), an article that targets unconscious perception using discrimination tasks with objective performance thresholds. In that approach, direct measures are those that explicitly define a discriminative answer in the task description (e.g., a word, a letter, a color), whereas indirect measures measure the effect on some given discrimination that a certain perceiving of information not defined explicitly in the task, but nevertheless implicitly present, has. As in their example (familiar from the Stroop experiments), when presentations of color-word stimuli are followed by descriptions (e.g. asking “Which word have you seen?”), the performances in such a task “constitute a direct measure of word perception,” while, when the task is formulated in converse fashion by asking “What color have you seen?,” it is judged that “any effect that the words may have on color-naming performance would constitute an indirect measure of word perception” (Reingold and Merikle Reference Reingold and Merikle1988, 564). The authors stress that they are abstracting from any implications about the sort of underlying process that might influence performance (e.g., whether it is priming or subliminal perception). At the same time, they assume that under normal conditions direct measures should result in higher or equal performance rates (in respect of sensitivity) compared to indirect measures, whilst arguing that in other cases one may justifiably posit some sort of influence of unconscious processes on participants’ performance. As one can clearly see, directness defined in these, methodological terms is straightforwardly applicable to objective measures of performance. Hence, instances of the procedure should probably be referred to as direct performance measures rather than direct awareness measures—the more so as the relationship between performance and awareness is a matter of debate, as indeed is the validity of such measures (Timmermans and Cleeremans Reference Timmermans, Cleeremans and Overgaard2015).

At the same time, a characterization of directness of this sort, as formulated by Reingold and Merikle, represents the only unambiguous sense attached to the term. As was already mentioned, this definition has not been explicitly applied within the current discussion concerning subjective measures, though it would be interesting to analyze the latter in the light of methodological directness. This may be done only very briefly here, as a preliminary step toward a proper investigation.

In the case of subjective experience measures such as PAS, what is explicitly defined in the task is the quality of (previous) stimulus experience (which has to be assessed). To be more literal in the application of the definition, if the task captured by the question “Which word have you seen?” constituted a direct measure of “word perception” (as in the original scenario from Reingold and Merikle’s example), then the task captured by the question “What was the quality of your experience?” would constitute a direct measure of “quality perception.” (In the PAS scenario analyzed here, it was “clarity” that was in fact explicitly defined in the task as a qualitative feature of experience.) In this sense, subjective performance measures, such as CR and PDW, will measure experience quality only indirectly because they do not to involve any experiential features being defined in the tasks (neither clarity, nor intensity or energy, etc.), being oriented instead toward the assessment of one’s own performance (by means of confidence or wagering). Consequently, it may be said that out of all the subjective measures analyzed here only PAS represents a methodologically direct measure of certain qualitative features of experience (such as clarity).

3.3 Semantic directness, and other senses of the term

It is fairly common to see proposed the idea that so-called worldly discriminations correlate with lower-order contents than do reports of one’s own perceptual experiences; the former are therefore also called first-order states, the latter second-, higher-order, or metacognitive ones. (Other terms relevant to this distinction are, for instance, primary vs. secondary consciousness, and sensory consciousness vs. introspective consciousness [Edelman Reference Edelman2003; Seth et al. Reference Seth, Edelman and Baars2005; Seth Reference Seth2008a].) Such a hierarchy is sometimes expanded to encompass three, four, or even five consecutive orders (Morin Reference Morin2006; Jonkisz Reference Jonkisz2012). With the aim in mind of seeking to analyze measures of consciousness, the following three orders of visual awareness will be discussed here:

• Awareness of the stimulus (object experience)—enabling one to discriminate different features pertaining to the perception of the stimulus (e.g., its location, shape, color or category);

• Awareness of seeing (seeing experience)—enabling one to discriminate different features of a visual experience (e.g., its clarity, intensity);

• Awareness of (one’s) visual experiencing (experience of oneself as a seer)—enabling one to judge one’s own visual experience (e.g., in terms of object-identification performance).

The process is likely to be a great deal more complex, in fact, as even first-order awareness (object experience) may be split into awareness of some lower-level features pertaining to the stimulus (such as shape, location, hue, or color) and awareness of higher-level features (like the identity of the object, or its semantic category). (In this context, see the level-of-processing hypothesis [Windey et al. Reference Windey, Gevers and Cleeremans2013; Anzulewicz et al. Reference Anzulewicz, Asanowicz, Windey, Paulewicz, Wierzchoń and Cleeremans2015; Jimenez et al. Reference Jimenez, Hinojosa and Montoro2020].) Moreover, taking the neurophysiology of visual perception as a basis, even the lower level features listed here may be ordered because it is well known that information about shape and location is processed earlier than information about color (see Lamme et al. Reference Lamme, Supèr, Landman, Roelfsema and Spekreijse2000, as well as the various articles collected in Spillmann and Werner Reference Spillmann and Werner2012). Even so, the nature of the three coarse-grained orders described in the preceding text is semantic rather than physiological. This is because the hierarchy reflects a referential ordering in which higher orders assume lower orders (in the sense of being about them; see Jonkisz et al. Reference Jonkisz, Wierzchoń and Binder2017 and Jonkisz Reference Jonkisz2012): Without first-order sensory content (stimulus experience), it is impossible to perform a second-order assessment of, say, clarity of visual experience, and without any (second-order) awareness of the experience of seeing (e.g., its degree of clarity), it is practically impossible to be aware of one’s being engaged in visual perception oneself and, for example, assess one’s own discriminative performance. (In this context, see also Seth [Reference Seth2008a].)

Why might such orders of consciousness be relevant when analyzing possible senses in which measures could be said to be “direct”? When a participant reports no signs at all of higher-order awareness (e.g., “vague experience” or “no experience” in PAS, or “guessing” in CR), this is still not enough to declare a total absence of any sensorily experienced content. (As the suggestive dictum goes, “absence of evidence is not evidence of absence” [Seth Reference Seth2008a; Timmermans and Cleeremans Reference Timmermans, Cleeremans and Overgaard2015].) Higher orders assume lower orders, but not vice versa. The lowest orders will therefore be in some sense basic or primary, and it may be hypothesized on this basis that:

• A measure may be considered semantically more direct (than another) when it targets a lower-order awareness.

If that is the case, then the most semantically direct measures will be those that target one’s first-order awareness of the stimulus. Objective performance measures are the most obvious examples here, the original assumption behind such measures (i.e., so-called World Discrimination Theory [Fu et al. Reference Fu, Fu and Dienes2008; Gaillard et al. Reference Gaillard, Vandenberghe, Destrebecqz and Cleeremans2006]) being that awareness of the stimulus correlates with behavioral sensitivity to the latter, disclosed in the form of an ability to discriminate its features with statistically significant accuracy. Meanwhile, PAS, despite its subjective threshold, also directly measures behavior (e.g., pressing the relevant buttons on a laptop keyboard). However, because a participant in a PAS experiment is asked to rate their own (first-order) visual experience (e.g., as clear, almost clear, vague, or nil), they have at the very least to be aware of seeing, so that the behavior correlates here with (at least) a second-order awareness of seeing. Although PAS does not seem to be semantically the most direct measure in this scenario, it is still likely to be more direct than other subjective measures. (At least, that was the initial claim put forward [Sandberg et al. Reference Sandberg, Timmermans, Overgaard and Cleeremans2010].)

According to the most common procedure applied in CR and PDW, participants have to assess their own performance in some prior identification task by determining a degree of confidence, or an amount of money in a wager, on a structured scale (binary, fourfold, or continuous), with their decision ultimately then being revealed in the corresponding behavior. (Behavioral responses may share an identical scenario across all subjective measures [Wierzchoń et al. Reference Wierzchoń, Paulewicz, Asanowicz, Timmermans and Cleeremans2014].) It seems that judgments about one’s own performance require, even if only implicitly, at least a rudimentary consciousness of oneself as visually experiencing something with a certain quality (i.e., clearly or not clearly enough). Hence, the task requirements of the CR and PDW procedures seem to result in highly metacognitive processes (Seth Reference Seth2008a, Reference Seth2008b) and require the most abstract orders-of-reference (Jonkisz et al. Reference Jonkisz, Wierzchoń and Binder2017). Ultimately, contra Persaud et al. (Reference Persaud, McLeod and Cowey2007), such measures will be the least direct of all those appearing in the scenario.

However, as some researchers have pointed out, to the extent that introspection is not explicitly called for in CR or PDW, one’s own performance judgments might then be based on more intuitive knowledge or instances of so-called gut feeling (see Timmermans and Cleeremans Reference Timmermans, Cleeremans and Overgaard2015, 34). If that were so, then subjective performance measures could in fact require lower-level contents—and possibly even ones lower than PAS (as originally claimed by Persaud et al. [Reference Persaud, McLeod and Cowey2007] and more recently by Lyyra [Reference Lyyra2019]). Consequently, it can be said that in principle CR and PDW may be considered—and probably also designed as—semantically more or less direct. They will be more direct when participants judge their choices using “intuition” or “gut feeling”—with metacognitive judgments being in this case inhibited somehow (e.g., by limiting decision times). And they will be less direct when participants are allowed, or even encouraged, to scrutinize their visual experience more carefully—with “intuitive judgments” having somehow to be inhibited here (e.g., by removing limitations on decision times and/or asking that answers be subsequently confirmed). Hence, just as objective performance measures may be designed to be direct or indirect (see the preceding text, and also Reingold and Merikle [Reference Reingold and Merikle1988]), subjective performance measures (CR and PDW) may also be carried out in ways that will make them more or less so, at least in the semantic sense of “directness.”

Are there any other senses in which talk of the “directness” of such measures merits our interest here? One interesting possibility has been pointed out by an anonymous reviewer of this very article. It concerns the idea of “causal directness.” Intuitively, a measure of awareness will be causally direct if there is no intervening cause between the presence of consciousness and the behavior recorded. Consequently, measures with less intervening causes would count as causally more direct, and these are construable as measures with fewer and/or stronger causal connections obtaining between the presence of consciousness and the recorded behavior. The issue certainly calls for more careful investigation, but we can sketch out a few brief thoughts even at this stage. It may be said, to begin with, that all sorts of objective methods seem more direct in this sense, if only because the time interval between actual experience and related/measured behavior may well be shorter here (with basic sensory detections or discriminations occurring much more rapidly than metacognition). At the same time, in relation to other senses of “directness” one can reasonably claim that semantically more direct measures also seem more causally direct (e.g., in the sense that measures of sensory content are more direct, both semantically and causally, than measures of higher-order and metacognitive contents), while methodologically indirect measures (e.g., priming methods) appear causally more direct than methodologically direct discriminations. It also seems that no measure could be absolutely causally direct, as this would represent a similarly idealized scenario to that of metaphysical directness. It would also appear to be the case that the idea of causal directness in some sense overlaps with that of metaphysical directness; however, this issue, as well as other possible senses of “directness,” will need to be further investigated elsewhere.