Scientific research composed and compiled by Shaun Gallagher and Morten Overgaard
We have little to say about pain itself. It is clear that although pain is distinguishable from other experiential states, Price and Aydede do not isolate it from those other experiential dimensions (especially the affective). Thus they write: “First, pain, unlike most conscious experiential states such as visual, auditory, tactile experiences, have an immediate affective and emotional aspect to it, which underlies its intimate personal as well as clinical urgency”. Our ordinary experience of pain, then, is not of some pure painfulness and we should even say that pain infects our visual, auditory, tactile experiences. When I am in pain, my experience of the world via these various modalities is affected and indeed, these perceptual experiences do have affective and emotional aspects to them that may be the result of the pain itself. So we agree that pain does not function as an object of our perceptual experience, but infiltrates that experience itself. “Notice that pain here is not the object of our perceptual experience, but rather, it is the experience itself”. But this also motivates the following question: to what extent does the pain that one tries to introspect actually affect the introspection? This is just the opposite of what might be the more expected methodological question: to what extent does introspection affect (change) the experience of pain (or any experience)?
The authors argue that introspection is indispensable and that it is already a part of third-person studies. But what is the nature of the introspection that is already practiced in science? There does seem to be a very basic agreement among scientists interested in introspection that it involves a direct attending to the subject’s own consciousness, no matter how one would go about defining introspection from there.
To quote just a few of the “first-generation introspectionists,” Knight Dunlap writes “Introspection’ is usually defined in terms which are equivalent to the expression consciousness scrutinizing itself.” (1912). Angell writes: “It [introspection] consists simply in the direct examination of one’s own mental processes” (1908). Stratton considers it to be the “direct acquaintance with the state of our minds which all of us to some extent possess” (1914). Finally, William James characterizes it as “the looking into our own minds” (James 1890). Price and Aydede, however, suggest that it is beyond the scope of their paper to examine the nature of introspection, and they are satisfied with some combination of HOP and HOT models. They suggest that nothing crucial about their analysis depends on this issue. We disagree, especially in regard to the first part of their paper. For purposes of discussion we will stay with the model they propose, but we note that there is certainly much more to say about whether HOP or HOT models genuinely capture the concept of introspection. What’s important here is that whether one appeals to HOP, HOT, or some alternative phenomenological model of introspection, introspection is understood as a reflective second-order cognitive act that thematizes first-order phenomenal experience, and makes that experience the object of reflection. Price and Aydede claim that, concerning this kind of introspection, “there isbabsolutely no reason to think that the use of such a first-person approach is scientifically and methodologically suspect.” Especially with respect to what they characterize as the introspection already practiced in experimental settings, however, there may be some justified suspicion. Indeed, and on a more basic level, before we get to the issue of scientific reliability, we think that Price and Aydede’s analysis may reflect a not uncommon confusion about the very nature of introspection and how it is practiced. Indeed, in controlled scientific experiments that require verbal reports, it is not clear at all that introspection is used in any strict methodological fashion. There are really two points here. First, whether the practices that Price and Aydede call introspective are really introspective; second, whether those practices are above suspicion.
According to Price and Aydede, introspection is already practiced in experimental science because “the subjects’ verbal reports [or nonverbal behaviors like button pushes] about their own cognitive states have routinely been taken as evidence for the cognitive models postulated”. First, one might argue that all reports given by subjects are, at least indirectly, about their own cognitive (mental, emotional, experiential) states. If one instructs a subject to push a button, or say “now” when she sees the light come on, then the subject is reporting about the light, but also about her visual experience. Even if one instructs the subject in a way that carefully avoids mention of an experiential state: “Push the button when the light comes on,” the only access that the subject has to the fact of the light coming on is by way of her experience of the light coming on. In this sense the first-person perspective is inherent in all experiments that depend on subjects’ reports. One might follow this to its logical conclusion, that even scientific observations made by the experimenter, usually considered as third-person data, presuppose and are tied to first-person reports of the scientist. The scientist might say “The subject’s premotor cortex was activated 300msecs before the subject raised her arm.” But that could easily be a report on the scientist’s experienced perception of the instrument that measured the timing of the activation. It would be odd, however, to say that this third-person fact was based on introspection, although it is a first-person report of the scientist. More generally, however, and less extremely, it is odd to say that the first-person reports of a subject are necessarily introspective, although this is precisely what Price and Aydede claim. For example, I may ask the subject to say “now” when she sees the light come on.
How precisely does the subject know when she sees the light come on? Does she reflectively introspect her experience looking for the particular visual state of seeing the light come on? Or does she simply see the light come on and report that? One might ask, “How could she possibly report that she sees the light come on if she doesn’t introspectively see that she sees the light come on? Is it possible that we can report on what we experience without employing introspection?” There is a long tradition in philosophical phenomenology (specifically the tradition that follows Husserl) that answers in the affirmative. We can report on what we experience without using introspection because we have an implicit, nonintrospective, prereflective self-awareness of our own experience. At the same time that I see the light, I know that I see the light. This knowledge of seeing the light is not based on reflectively or introspectively turning our attention to our own experience. It is rather built into our experience as an essential part of it, and it is precisely that which defines our experience as conscious experience. On this view, I consciously experience the light coming on just as I see the light coming on. I don’t have to verify through introspection that I have just seen the light come on, since my first-order phenomenal experience
First-person reports of this kind, then, are not introspective reports. They are prereflective experiential reports. So it is not correct to say that from a first-person perspective “conscious experiences seem accessible only through introspection” or “introspection seems to be the only available method of access to qualia” This applies to pain as well. It is not the case that our access to pain phenomena is “only through the special epistemic faculty of introspection”. Indeed, introspection on pain is usually motivated only because we already know that we are in pain—and we have that knowledge prereflectively, and can report it on that basis, without introspecting it. A stimulus is applied. The experimenter asks, “Is that painful?” I do not have to introspect to say “yes.” I do not have to “observe/believe that such and such sensations” are happening to me (ibid., p. 252).
I can report directly and immediately on my experience of pain because my access is directly and immediately in the pain experience.
In addition, not all reports are about consciously experienced states. In philosophy and especially in cognitive science, there seems to be agreement, at least to a large extent, that not all mental states are conscious (Marcel 1983). Examples such as blind-sight and subliminal perception serve to illustrate that subjects may perform a number of tasks that we normally do not hesitate to call “mental” even though subjects report no conscious awareness of them. Therefore, it seems logical to conclude that not all reports about mental states are introspective, that is, not all of them are about consciousness.
The second point concerns how reliable or how methodological such experiential reports are. In general and for many cases, these kinds of reports do seem very reliable. If an experimenter applies a stimulus that causes a relatively high degree of pain, or a sensory stimulus that is well above threshold, the subjects’ reports that they experience the stimulus as painful or as clearly present seem above suspicion. Reliability may decrease, however, when the stimulus is closer to threshold, and may depend on the mode of report, or other subjective factors that qualify the report. Marcel (1993), for example, has shown that requests for quick reports of close-to-threshold stimuli using different modes of report (verbal, eye blink, button push) elicit contradictory responses. At the appearance of a just noticeable light stimulus, subjects will report with a button push that they did see the light and then contradict that with a verbal report. This kind of data, and more generally, uneven or inconsistent data can motivate two different strategies. Most often, following established scientific procedure,vdata are averaged out across trials or subjects, and the inconsistencies are washed out. Less often, scientists are motivated to take this first-person data seriously and to employ introspective methods to investigate it. The second part of Price and Aydede’s paper turns to what is genuinely the use of introspection in such contexts.
Methodologies of Introspection
So far we have argued that not all verbal reports on experience are introspective reports, and introspection is not the only access we have to experience. Thus we take issue with Price and Aydede’s claim that “Introspection is a way—apparently the only way—of coming to know about our experiences and their qualities directly”. Even if, however, many first-person reports found in scientific experiments are first-person, nonintrospective experiential reports, we do not mean to rule out the usefulness of introspection. The use of introspective reports, that is, self-reports that thematize experience, can certainly provide more information about the subject’s experience. In regard to pain, for instance, introspective reports can specify the qualities and subjective measurements of pain. In addition, if done in a methodically controlled way, introspection can address issues pertaining to the reliability of some
nonintrospective experiential reports.
It will be fruitful to compare and contrast three different models of how to employ a methodical introspection in experimental situations: (1) Price and Aydede’s experiential model; (2) what has been called a “new introspectionism,” developed by Overgaard, his colleagues, and a number of other researchers (e.g., Marcel 1983, 2003; Jack and Roepstorff 2002; see comments by Frith 2002; Gallagher 2002); and (3) the method of neurophenomenology developed by Varela (1996), Lutz, and their colleagues (Lutz et al. 2002). All three share the same ambition to improve techniques for subjective reporting in order to gain more insights into the “subjective pole” inb comparison between objective neural states and subjective conscious states.
The Experiential Model
Price and Aydede suggest an approach consisting of two stages. First, one is to use a “horizontal approach” in which an investigator or subject introspectively examines what some or other subjective state feels like. Price and Aydede advise us to avoid speculations about why something was experienced and to focus specifically on how it feels instead, thus avoiding interpreting or judging one’s own experiences. To do this, they argue that one should notice experiences passively without controlling attention so that one observes ongoing thoughts, emotions, or perceptions as if they were “seen in the periphery of one’s visual field.” Price and Aydede suggest not only “simple” kinds of experiential states as objects for the introspective examination, but, apart from their main example of pain, they suggest performance anxiety as a kind of mental state that can be studied with their approach.
The introspective examination consists not only of “inner observation” but also of a description or verbalization of the observations. Of course, aside from the possibilities presented by those who are poetically blessed, one cannot describe, say, the sensation of coldness with many words. Our ordinary, prosaic linguistic practices have not sufficiently evolved to describe a subjective state in such a way that someone who never experienced coldness himself would get an idea of what that sensation is like simply through the description. To address this scarcity of words, Price and Aydede want their subjects to describe associations and thoughts that may arise in connection with the relevant experience: “Intense burning and throbbing in my hand. Feel bothered by this and slightly annoyed. Is it going to get stronger? Feeling of concern. Hope my hand isn’t going to be scalded.”
Price and Aydede suggest that scientists should use themselves as subjects. The example of a description of associations that arise when lowering one’s hand into cold water was in fact given by one of the authors of the paper. The argument for using oneself, as investigator, as a subject, seems to be that the reports of the investigators are as “subjective” as are the reports of naive subjects, and, in this sense, just as valid as experimental data. However, one should not forget that investigators must be assumed to always have certain hypotheses and results they hope to find, and thus they are likely more biased as subjects. Using the same argument as Price and Aydede, one could say that given that investigators and naive subjects have the same status in terms of validity, one could reduce the possibility of the confounding effects of interpretations and judgments by using naive (though probably trained) subjects only.
Price and Aydede believe that their approach is compatible with experimental methods found in psychophysics. To integrate their rather open method of describing experiences, however, they find it necessary to ask subjects to scale the presence of, say, “a throbbing sensation of pain,” “fear of bodily harm,” or some other state described by the subjects themselves during the “open description.” This would in essence make possible a quantification of the descriptions, and, as such, it would make the reports commensurable with cognitive neuroscience. This is the second, “vertical,” aspect of their approach. In classical cognitive neuroscience, one uses stimulus input as an experimental variable. Subjects would be presented with two or more different kinds of stimuli and in order to find the neural activations for perceiving one kind of stimulus, the neural activations caused by the other kinds of stimuli would be subtracted from the first. Thus, the reasoning goes, one will find the essential features involved with perceiving the first kind of stimulus. When using neuroscientific techniques there seems no way around using such a subtractive method even though it has been severely criticized (Friston et al. 1996; Overgaard 2004). However, one does not need to define one’s variables based upon different stimuli. One could keep stimulus features constant and only vary the instructions as now seen in an increasing number of studies, or one could define the experimental conditions based on the subjective reports themselves. This latter strategy is suggested by Price and Aydede, and it points to a way of integrating the open, subjective reporting strategy with the methodology of cognitive neuroscience. Price and Aydede do not stand alone defending this kind of view on subjective reports and their integration with neuroscience. On several occasions, the authors appeal to introspectionism and phenomenology as research directions with similar goals. Within both traditions, in the last decade, new developments have occurred that deserve a comparison with the suggestions of Price and Aydede.
A “new” introspectionist approach to subjective reporting has been put forward in Marcel 1983, 2003; Jack and Roepstorff 2002; Overgaard 2003a; Overgaard, Nielsen, and Fuglsang-Frederiksen 2004; and is further developed in Overgaard and Sørensen 2004; and Ramsøy and Overgaard 2004.
In Overgaard 2003, an outline for introspective reporting is described. It is suggested that one should perform consciousness studies on a metaphysically neutral ground, and that an important reason for the shortcomings of classical introspectionism was the commitment to certain metaphysical claims, for example, the belief in “elementarism.” It is suggested that one should use introspective reporting in experiments on consciousness, given that nonintrospective reporting may reflect non-conscious processing. Furthermore, it is suggested that one should reconsider the use of stimulus conditions as the only variable in experiments in cognitive neuroscience, and instead use differences in instructions or in the subjects’ own reporting as the categories of analysis. This line of thinking corresponds very well with Price and Aydede’s reflections on the use of subjective reporting in experiments as well as in analysis of data. Ramsøy and Overgaard (2004) presented subjects with a visual identification task using varied durations, and asked subjects (1) to guess what was shown on the computer screen, and (2) to scale how clearly they experienced the image. The steps of the scale, including their definitions, were made by the subjects themselves with the instruction that there should be a 1:1 correspondence between experienced differences and reported differences. After a pilot experiment, in which the subjects developed the scale and became accustomed to using it, the actual experiment was run. The subjects ended up using more or less the same scale (named the Phenomenal Awareness Scale or PAS), and for reasons of analysis, the investigators decided to merge the scales to include only the points of the scale that were shared by the subjects. This strategy corresponds almost completely to the proposals of Price and Aydede in the “horizontal stage” of analysis, and in the transformation from the horizontal to the vertical strategy.
In Overgaard, Nielsen, and Fuglsang-Frederiksen 2004, PAS was used with different subjects when using a similar visual display, and coupled with transcranial magnetic stimulation and EEG. This study aimed to identify the involvement of the ventral projection streams from primary visual cortex in visual consciousness.
In effect, the approach of “new introspectionism” (described in more detail in Overgaard 2003b) shares all important features with the approach of Price and Aydede. The subjects started out using their own terminology, which then was used for the purpose of scaling the subjective reports, and finally, it was integrated with neuroscience to search for neural correlates of consciousness.
There are some minor differences between the two approaches as well. Price and Aydede seem more optimistic about which mental phenomena one can study with their suggested approach. Yet, during the horizontal stage, a phenomenon like performance anxiety would give rise to many different associations and thoughts, with the result that it would be almost impossible to tell whether the very different subjective reports basically reflect identical conscious states. Even in the example of the experience of coldness, as mentioned above, the “spontaneous utterings” of the subjects are so relatively different that the investigators must perform some amount of interpretation of the reports in order to create categories suitable for quantitative analysis. Such a post hoc analysis, of course, shares all the problems of the creation of reporting categories in advance of an experiment.
A further line of research using introspective reports aims to identify differences between reports that are specifically about “how” something is experienced and reports about “what” is experienced. In Overgaard and Sørensen 2004, subjects were presented with a simple design for visual stimuli. On one of three possible locations, a simple figure followed by a mask would appear. The figure was either a triangle, a square, or a circle, or some variation of one these figures (e.g., a half-circle, an upside-down triangle, or a combination of two figures). The colour was either blue, green, or red, although the hue varied. The subjects were to identify the displayed figure by pointing at the corresponding figure drawn on three scales: one for stimulus shape, one for color, and one for the location. The scale of shapes consisted of a display of thirty-four different figures, some of which were included in the data material. The scale of colors consisted of eight different levels of hue for the colors. As with the shapes, only some of the colors were actually included as stimuli. The scale of positions displayed the fixation cross in the middle and the three different locations where the stimulus could occur. The responses of the subjects were treated as being either “correct,” “incorrect,” or “near correct.” “Near correct” responses partially matched stimulus in a manner that was only partially correct (e.g., when they pointed at the same color as the one presented, but in a brighter or darker tone). The results of the data analysis showed that subjects in the nonintrospective condition had significantly more correct and incorrect responses, whereas the introspective subjects most often were “near correct.” In addition, subjects in the introspective condition tended to be more liberal about their reports of, say, color, while the subjects in the nonintrospective condition tended to show a more conservative style conforming to specific color categories.
These results open up questions such as to what degree and how precisely introspection might change (visual) experiences. It seems necessary to address the issue if or how this knowledge should change our way of using introspective reports.
A third approach, neurophenomenology, as espoused by Francisco Varela (1996), follows the phenomenological tradition initiated by Husserl. This view involves training both the scientists and the experimental subjects in phenomenological method, including use of the phenomenological reduction, that is, the setting aside or “bracketing” of opinions or theories that a subject may have about his experience. This method involves shifting our attention from what we experience to how we experience. This correlates well with Price and Aydede’s advice to avoid speculations about why something was experienced and to focus specifically on how it feels. Lutz et al. (2002) employ the neurophenomenological method to study the highly variable successive brain responses to repeated and identical stimulations in many empirical testing situations that target specified cognitive tasks. Their hypothesis is that this variability is generated in mental fluctuations due to the subject’s attentive state, spontaneous thought processes, strategy decisions for carrying out the task, and so on. These subjective parameters include distractions, cognitive interference, and so on. To control for such subjective processes is difficult and they are usually averaged out across a series of trials and across subjects. Lutz and his colleagues decided to take these subjective parameters more seriously. They combined a process of trained phenomenological reflection with the dynamical analysis of neural processes measured by EEG in a paradigm involving a 3-D perceptual illusion. Importantly, Lutz and his colleagues used the introspective first-person data not simply as more data for analysis, but as contributing to their analytic framework.
Phenomenological training in this experiment consisted in training subjects to deliver consistent and clear reports of their experience through a reflective introspection. The goal of phenomenological reflection is to gain intuitions of the structural invariants of an experience, not to average them out. Phenomenological reflection can be either self-induced by subjects familiar with it (not unlike Price and Aydede’s proposal that the scientist use herself as the subject), or guided by the experimenter through open questions…questions directed not at opinions or theories, but at experience. Again, this resembles the “open description” discussed by Price and Aydede.
Rather than employing predefined categories, and asking “Do you think this experience is like X or Y or Z?” the open question asks simply “How would you describe your experience?” Open questions posed immediately after the task help the subject to redirect his or her attention toward the implicit strategy or degree of attention he or she implemented during the task. Subjects can be reexposed to the stimuli until they find “their own stable experiential invariants” to describe the specific elements of their experiences.
In a series of preliminary or practice trials, the subjects developed descriptions (refined verbal reports) of the subjective parameters while engaged in a depth perception task. Subjects thus became knowledgeable about their own experience and developed descriptions of experiential invariants on the basis of open questions, reporting on the presence or absence or degree of distractions, inattentive moments, cognitive strategies, and so on. On the basis of these first-person introspective descriptions, descriptive categories were formulated a posteriori to create phenomenologically based clusters that are then used as analytic tools in the main trials. For example, with regard to the subject’s experienced readiness for the stimulus, the results specified three readiness states: steady readiness (SR), in which subjects reported that they were alert and well prepared as the task began; fragmented readiness (FR) in which subjects reported that they were prepared less “sharply” (due to a momentary tiredness) or less “focally” (due to small distractions, etc.); and unreadiness (SU) in which subjects reported that they were unprepared as the task began. Subjects then used these categories to report their readiness state during the main trials as the experimenters recorded the electrical brain activity. The first-person reports were correlated with both behavioral measures (reaction times) and dynamic descriptions of the transient patterns of local and long-distance synchrony occurring between oscillating neural populations. Using these correlations, Lutz et al. were able to show that distinct subjective parameters correlate to specific dynamic brain patterns just prior to presentation of the stimulus. The results were significantly different relative to results based on averaging across trials.
The experimental protocol used in Lutz et al. 2002 thus employs a practical phenomenological method. The subjects are asked to provide a description of their own experience using an open-question format, and thus without the imposition of predetermined theoretical categories. They are trained to gain introspective intimacy with their own experience. Their first-person introspective reports are then intersubjectively and scientifically validated both in setting up the phenomenological clusters and in using those clusters to interpret results that correlated with objective measurements of behavior and brain activity.
These three approaches share a number of common features.
1. Use of preliminary trials or pilot experiments to train subjects and to develop introspective or phenomenological descriptions of experience or subject-developed scales
2. A pushing aside of theories or speculations in favor of attending to how experience is happening (Price and Aydede’s avoidance of speculation; Varela’s phenomenological reduction).
3. The use of open questions to develop a description of the experience (Price and Aydede’s “open description”; Lutz’s open questions).
4. The formulation of common categories that transform first-person introspective descriptions into intersubjectively verified and commonly understood reports (Lutz’s phenomenological clusters; Overgaard)
5. The use of these phenomenologically generated categories not just as data, but also as part of the analytical instrument (Price and Aydede; Overgaard; Lutz).
6. The integration of first-person data with third-person behavioral, psychophysical, and neurological measurement (EEG, TMS, PET, fMRI) in search of correlations among experiences, brain activity, and behavioral responses (Price and Aydede; Overgaard; Lutz). Putting all of these elements together may provide a fuller and more detailed conception of how a methodical introspection could work than found in any one of the models. Perhaps a more integrated model that recognizes the precise difference between introspection and first-person, prereflexive, experiential reports is now called for.