First-Person Data, Publicity, and Self-Measurement[1]
Gualtiero Piccinini
3/9/2009
DRAFT – DO NOT QUOTE WITHOUT PERMISSION
Abstract:
First-person data have been both condemned and hailed because of their alleged privacy. Critics argue that science must be based on public evidence: since first-person data are private, they should be banned from science. Apologists reply that first-person data are necessary for understanding the mind: since first-person data are private, scientists must be allowed to use private evidence. I argue that both views rest on a false premise. In psychology and neuroscience, the subjects issuing first-person reports and other sources of first-person data play the epistemic role of a (self-) measuring instrument. Data from measuring instruments are public and can be validated by public methods. Therefore, first-person data are as public as other scientific data: their use in science is legitimate, in accordance with standard scientific methodology.
This paper is about the proper scientific use of first-person data. Paradigmatic examples are data about mental states and events obtained from first-person reports, such as “I see purple here” and “I’m sad.” Many experimenters ask their subjects about their mind, then use what they hear as a source of data.
First-person data are used in many fields, including psychophysics (Snyder, Fast and Bartoshuk 2004), cognitive psychology (Ericsson and Simon 1993, Schooler and Schreiber 2004), psychiatry and personality psychology (Hurlburt and Heavey 2006), neuropsychology and clinical psychology (Cytowic 2004, Samsonovich and Nadel 2005), cognitive neuroscience (Leopold, Maier and Logothetis 2003), subjective wellbeing (Alexandrova 2005, Angner ms.), pain (Aydede 2005), and consciousness studies (Jack and Shallice 2001, Baars 2003, Hohwy and Frith 2004).
Are first-person data legitimate scientific evidence? In spite of over a century of debate, the answer remains unsettled. The most fundamental concern is privacy: according to standard scientific methodology, scientific data should be public. But first-person data appear to contain an element of privacy.
Critics argue that science must be based on public evidence: since first-person data are private, they should be banned from science. Apologists reply that first-person data are necessary for understanding the mind: since first-person data are private, scientists must be allowed to use private evidence. I will argue that both views rest on a false premise. In psychology and neuroscience, the subjects issuing first-person reports and other sources of first-person data play the epistemic role of a (self-) measuring instrument. Data from measuring instruments are public and can be validated by public methods. Therefore, first-person data are as public as other scientific data: their use in science is legitimate, in accordance with standard scientific methodology.
Most discussions of first-person data characterize them as data from first-person verbal reports. But there is more to first-person data than data from first-person reports. To better demarcate our topic, a broader characterization of first-person data would be helpful.
First-person data need not be based on first-person reports, or even on verbal reports of any kind. If a subject presses one of two buttons depending on which of two ways she perceives a Necker cube at a time, her button pressing is a source of first-person data. For this reason, the term ‘first-person report’ is too restrictive to capture our target. Instead, I will use ‘first-person behavior’ to denote any behavior that is the source of first-person data. What counts as a (non-verbal) first-person behavior?
Most behaviors are not first-person. If a subject presses one of two buttons depending on whether a triangle or a circle is presented to her, there is nothing especially first-personal about her behavior. True, the two possible behaviors (pressing one button, pressing the other button) discriminate between two possible mental states (perception of a triangle, perception of a circle). But they also discriminate between two physically different stimuli (triangle, circle). Stimulus discrimination is enough to infer that the subject is in different mental states under different conditions.
By contrast, in the Necker cube case, the stimulus is physically the same throughout. Yet at some times, the subject perceives it as facing towards her; at other times, she perceives it as facing away. Since the stimulus is the same, we cannot infer that the subject is in different mental states at different times on the basis of stimulus discrimination. The behavior provides information about the subject’s mind that cannot be inferred from stimulus discrimination alone.
One suggestion worth exploring is that first-person behaviors are behaviors produced with the goal of revealing something about one’s own mental states.[2] This characterization covers ordinary first-person reports as well as the Necker cube case. In both cases, we may suppose that the subject intends to inform others about her mental states. Unfortunately, this characterization is still too restrictive. Intending to reveal something about one’s mind is not necessary for first-person behavior. For one thing, someone might produce first-person reports in spite of lacking the goal to reveal something about her mind. She may wrongly think no one is listening, she may speak her mind without considering what she is saying, or she may be tricked into revealing her secrets by a skilled interrogator. In all these cases, her behavior is still first-person.
It is also plausible that children and some non-human animals can behave in first-person ways, even though they lack the sophistication to formulate the goal of informing others about their mind. As a quick survey of the literature reveals, scientists treat some behaviors of children and monkeys as first-person behaviors. For instance, they might train monkeys to press one of two buttons depending on how they see a Necker cube, or something closely analogous (Leopold, Maier, and Logothetis 2003). Scientists treat such behaviors as sources of first-person data, even though there is scarcely any evidence that the monkeys intended their behaviors to be so used. We should not rule out these scientific practices by definitional fiat. Thus, we should not assume that the subject of a first-person behavior is always attempting to reveal something about her mind.
In looking for a better characterization, we need not look for necessary and sufficient conditions. Most concepts cannot be defined in this way, and even when they can, necessary and sufficient conditions are hard to find. What we can hope for is to identify one or more nontrivial properties that are important for a (non-verbal) behavior to be first-person. To avoid trivializing the notion of first-person behavior, a good characterization should rule out obvious cases of non-first-person behaviors as such, but it need not establish a sharp boundary.
Let’s go back to our examples. On one hand, we have cases like the Necker cube, with or without the goal of informing an external observer about one’s mind. On the other hand, we have simple stimulus discrimination. The contrast between these cases suggests that a (non-verbal) behavior is first-person only if it can provide information about the subject’s mind beyond what can be inferred from stimulus discrimination alone.
By this, I do not mean that for a (non-verbal) behavior to be first-person, the subject must respond multiple times to the same stimuli – as in the Necker cube example. Most data from first-person behaviors are not based on multiple trials. In many cases, it is not practically possible to conduct multiple trials under the same conditions. But if such multiple trials were conducted, then the only way to differentiate between the relevant mental states would be the (first-person) behavior in question – stimulus discrimination would not be enough. Thus, a (non-verbal) first-person behavior should give us more information about a subject’s mind than stimulus discriminations alone.
This condition is still insufficient for our purposes. The selection of one stimulus over another made by human and animal subjects may be used to discriminate between mental states (i.e., the subject’s preferences). Yet selection of one stimulus over another is not first-person behavior. To take this into account, we should add that a (non-verbal) first-person behavior must differentiate mental states while the subject’s choices remain the same. Consider a subject who generally chooses hazelnuts over walnuts. This alone is not first-person behavior. But suppose the same subject also presses one of two buttons: one button indicates that she chose based on taste, while the other indicates that she chose based on texture. The button pressing is a first-person behavior.
I tentatively conclude that a (non-verbal) behavior is first-person only if it can provide information about the subject’s mind that cannot be inferred from stimulus discrimination and selection alone. Much more could be said about first-person behaviors, but this should be enough to demarcate our topic. Since it takes more than stimulus discrimination and selection for a behavior to be first-person, there is a nontrivial distinction between behaviors that are and those that aren’t first-person. This condition does not mark a precise boundary between first-person and non-first-person behavior. Like most natural boundaries, it is vague, and its exploration should be left to empirical investigation.
Before assessing the validity of first-person data, we can avoid some confusion if we briefly disentangle the present question from some other, largely orthogonal questions.
Much of the recent discussion of first-person data has taken place within consciousness studies. Some authors see the methodology of first-person data as a first step towards a science of consciousness (e.g., Dennett 1991, Varela 1996, D. Chalmers 2004). Some also see the use of first-person data as a way of closing the explanatory gap or solving the “hard problem” (e.g., Varela 1996, Roy 2003, D. Chalmers 2004). But we should not assume that first-person data are all and only about consciousness, or that legitimizing first-person data is enough to construct a science of consciousness. First-person data are used well beyond the study of consciousness, and the science of consciousness faces many problems besides first-person data’s legitimacy. Thus, the problem of first-person data is not the problem of how to construct a science of consciousness, although there is overlap between the two problems. Furthermore, closing the explanatory gap or solving the hard problem requires finding the appropriate explanations. Such explanations cannot be found simply by solving a methodological problem. Thus, the problem of first-person data is not the problem of how to close the explanatory gap or how to solve the hard problem.
First-person data are sometimes called ‘introspective data’ or ‘data from introspective reports’. I used the latter term in a previous paper. I now find this usage misleading, because it invites a conflation between the question of first-person data’s legitimacy and the question of what introspection is and how it works. On the latter question, there is a wide range of opinions (cf. Robbins 2006). In addition, it’s unlikely that there is a single process responsible for producing all first-person behaviors that human beings are capable of (cf. Prinz 2004). Of course, a complete theory of first-person data must include a theory of how the data are generated, which must say whether there is such a process as introspection and how it works. But the question of whether first-person data are legitimate is not the same as, and may be answered largely independently of, questions about introspection. Because of this, I will avoid appealing to introspection or talking about ‘introspective data’ as much as possible. My topic is not the process by which subjects generate first-person behaviors—be it introspection or something else. My topic is the legitimacy and validity of first-person data.
In recent years, a number of authors have revived a view of first-person data that can be traced to some versions of classical introspectionist psychology and Husserlian phenomenology. I will call it ‘privatism’. Privatists maintain that first-person data are private but nevertheless scientifically legitimate. Since first-person data are private, their use results in a break from ordinary science. Some privatists call the use of first-person data ‘first-person methods’—they contend that first-person methods are qualitatively different from ordinary, third-person scientific methods (e.g., D. Chalmers 1999, Varela and Shear 1999).
Privatism may be stated more explicitly as follows:
(1) Subjects producing first-person behaviors are (scientific) observers of the mind.
(2) Their data are their mental states.
(3) Their mental states, and hence their data, are private.
From these three assumptions, a conclusion follows:
(4) First-person data cannot be validated by public means.
The conclusion follows because if first-person data are truly private, i.e., directly accessible only to the subjects, then there is no way for anyone else to either validate or invalidate them (more on this below).
Variants of privatism may be found in the writings of several authors. A candid defense is given by Goldman (1997), although later Goldman moved away from privatism (2004). A more recent example is by Chalmers:
When a conscious system is observed from the third-person point of view, a range of specific behavioral and neural phenomena present themselves. When a conscious system is observed from the first-person point of view, a range of specific subjective phenomena present themselves. Both sorts of phenomena have the status of data for a science of consciousness (D. Chalmers 2004, p. 1111, emphasis added).
Chalmers seems to suggest that introspecting subjects observe their mind analogously to how experimental scientists observe phenomena, and both sets of observations must play a role in science (assumption (1)). In the last sentence, he seems to identify first-person data and the subjective experiences themselves (assumption (2)). Later in the same paper, he explicitly says that first-person data are private (assumption (3)) and draws the conclusion that the reliability of introspection is untestable (assumption (4)).[3]
Some privatists attempt to do without (1). In an article often cited as the founding document of ‘neurophenomenology’—a label advocated by a growing number of authors—Varela argues that the proper source of first-person data is something like Husserl’s “phenomenological reduction”. According to Varela, the phenomenological reduction is different from introspection—it is not based on observing one’s mind:
[The phenomenological reduction] is not a 'seeing inside', but a tolerance concerning the suspension of conclusions that allows a new aspect or insight into the phenomenon to unfold. In consequence this move does not sustain the basic subject-object duality but opens into a field of phenomena where it becomes less and less obvious how to distinguish between subject and object (this is what Husserl called the 'fundamental correlation') (Varela 1996, p. 339).
I find this passage mystifying. At a minimum, science is a form of theorizing constrained by empirical data. For there to be data to constrain theory, someone must produce them. If the “subject-object duality” is not “sustained”, who is producing the data? Some fusion of subject and object? If so, then the fusion of subject and object counts as the scientific observer, and we are back to (1). From now on, I will assume that for present purposes, Varela’s view collapses into (1).
Privatism is based on the platitude that mental states and events are not shared among subjects. I have no quarrel with that: mental states and events are private in the sense that we each undergo all and only our own. As a consequence, we can only be aware of (some of) our own mental states. In this sense, we have a special epistemic access to (some of) our (and only our) mental states.
In addition to these truisms, privatism construes mental states as scientific data, directly accessible only to the subject undergoing the mental states. This is where the trouble comes from. If mental states, which are private, are scientific data, it follows that the science of mind is grounded (in part) on private data:
In most areas of science, data are intersubjectively available: they are equally accessible to a wide range of observers. But in the case of consciousness, first-person data concerning subjective experiences are directly available only to the subject having those experiences (D. Chalmers 2004, p. 1117, cf. also Goldman 1997, p. 533).
Lack of publicity has two considerable costs. First, private data cannot be reproduced by independent observers. Second, private data cannot be validated (or invalidated) by public means. That both of these are consequences of privatism, and how serious they are, is not always appreciated.
Privatism runs directly contrary to one of the most basic principles of scientific methodology: scientific methods must be public. A method is public just in case by applying it, any competent observer can generate the same data on the same questions. (For a defense of method publicity, see Piccinini 2003b.) By following public methods, scientists can insure two desirable results. First, any competent scientist can reproduce other scientists’ data, so as to obtain evidence that she is in the presence of the same phenomena. Second, scientists can look for correlations between different types of data, thereby validating them on independent grounds. Data obtained by one method can be correlated with data obtained by other (independently established) methods, thereby validating the first set of data.
If privatism is correct, none of this can be done with first-person data. They are immune to intersubjective validation:
Our access to [first-person data] depends on our making certain assumptions: in particular, the assumption that other subjects really are having conscious experiences, and that by and large their verbal reports reflect these conscious experiences. We cannot directly test this assumption: instead, it serves as a sort of background assumption for research in the field (D. Chalmers 2004, 1117; cf. Goldman 1997, 539).
The important point to notice is that if privatism is correct, the reliability of verbal reports can never be independently established, for principled reasons. Any attempt to establish it would have to find a correlation between the private first-person data and some type of public data, such as the subject’s verbal reports. But the private data are by definition directly accessible only to the subject, so no one else is in a position to establish a correlation between them and anything else. The identification of first-person data with (private) mental states entails that publicly validating first-person data requires other observers to undergo the same token mental states. This, of course, can’t be done. At most, the subject herself can swear that her reports are truthful. That adds no additional warrant to the reports themselves.
There is another way to see the gravity of the situation. Reliance on private methods of investigation leads to epistemic divergence (Piccinini 2003b). Epistemic divergence occurs when different investigators are unable, as a matter of principle, to settle disputes. Consider what happens when two investigators A and B apply private (i.e., non-public) methods. Since the methods are private, they may generate different (inconsistent) data as answers to the same question. A and B’s methods need not be introspection; they may be the reading of tea leaves or any other private method. Let’s suppose that by applying private methods, A and B deliver answers to the same question. Specifically, suppose that A obtains data to the effect that p, while B obtains data to the effect that not p. A and B disagree: based on their empirical investigations, A maintains that p; B denies it. Due to the privacy of their methods, and absent other methods to answer the same question, there is nothing that A and B can do to find out who is right. They are both competent investigators who, by the light of privatism, applied their methods correctly. And since their methods are private, the methods need not yield the same data when applied by different investigators to answer the same question.
The obvious and sensible solution—the way out of epistemic divergence—is to jettison A and B’s methods of investigation and find better ones—ones that do not yield contradictory results when applied by different investigators to answer the same questions. But this is the same as rejecting private methods of investigation in favor of public ones. That is, the solution to A and B’s impasse is to reject privatism.
A privatist might reply that at least in the case of first-person data, there is no evidence that we are in a state of epistemic divergence. As long as our science of mind seems to be doing ok, we should allow private data into it. In fact, a common defense of first-person data by privatists is that although the deliverances of introspection are private, they can still be used as scientific evidence because introspection is reliable.[4] Or at least, it should be taken to be reliable in the absence of evidence to the contrary (Goldman 1997, D. Chalmers 2004). This response is wrong on three counts.
To begin with, the caveat that introspection should be considered reliable unless there is evidence to the contrary is a red herring. If first-person data are truly private, what evidence could there be that introspection is unreliable? Internal inconsistency, perhaps. In fact, internal inconsistency is the only suggestion made by Goldman in this context. If a subject delivered contradictory reports as answers to the same question, at least one of the reports must be false. If this internal inconsistency were noticed systematically across conditions, Goldman argues (1997, 543), unreliability would be established. But a process may be unreliable without producing contradictory data. As Goldman notes, testing the reliability of introspection requires bringing independent evidence to bear. If the deliverances of introspection are private, Goldman continues, no independent evidence can be brought to bear. Thus, under the assumption of privacy, the degree to which and the conditions under which introspection is reliable cannot be established by public means. And if they cannot be publicly established, I add, nothing warrants the suggestion that introspection is reliable (to any degree).[5]
If privatism is right, our science of mind is even worse off than A and B. At least A and B can generate data on the same question and see if their data agree. But no two subjects can answer the same question by the privatist’s method, because each subject can only introspect her own mind. Thus, if privatism is right, we can never bring any public evidence (besides self-contradiction) to bear on the reliability of first-person data. And if so, we are not warranted to rely on them.
A committed privatist might still reply as follows. We must assume that introspection is reliable, on pain of skepticism about the mind. The reliability assumption is needed to get our science of mind started. This reply assumes that introspection is necessary to study the mind. Some would dispute this assumption, but I will grant it for the sake of the argument. The reply also assumes that introspection is a basic source of evidence – a source whose reliability need not be established by means of an independent source (cf. Goldman 2004, 3). Examples of putative basic sources include perception, memory, inductive inference, and deductive inference. Finally, the reply assumes that being a basic source of evidence is enough for being scientifically legitimate. On the contrary, being basic is not enough for being a legitimate source of (scientific) evidence. What is also needed is the possibility of establishing reliability on public grounds. Since the reliability of introspection (as conceived by the privatist) cannot be publicly established, introspection remains illegitimate.[6]
To illustrate, consider our reliance on perception in physics. Chalmers suggests that relying on the reliability of introspection in our science of mind is analogous to relying on the reliability of perception in physics (cf. D. Chalmers 2004, 1117). Chalmers is right that the reliability of perception can only be established using some form of perception – with or without the aid of recording instruments. Thus, perception is a basic source of evidence. But if privatism is correct, the case of introspection is not analogous to that of perception: unlike the degree of reliability of introspection (according to privatism), the degree of reliability of perception can be established by public methods, by correlating the reports of perceivers with the objects and events they perceive. The latter are public objects and events that can be perceived by other competent observers under similar conditions as well as recorded and measured by (public) instruments. In fact, sensory perception ceased long ago to be the primary means of observation in physics, because it (i) can access less physical information, (ii) is more subjective (less public), and (iii) is less reliable than the artificial detectors physicists employ in most of their observations (cf. Shapere 1982, esp. Section III).
The same point applies to memory, inductive inference, and deductive inference. Basic as they may be, they deliver public conclusions about public objects and events. Such conclusions can be replicated by others and publicly tested. Thus, their degree of reliability can be publicly established.
Sometimes we may assume that a method is reliable on the grounds of limited evidence. This is ok so long as we can eventually find public evidence of such reliability. Even if there are cases where we simply assume the reliability of a method to start an investigation, it doesn’t follow that we may do so in the case of introspection. We may if the reliability of introspection can eventually be established by public means. If the privatist is right, this cannot be done. By entailing that first-person data cannot publicly validated, privatism poses a principled obstacle to the science of mind.
This obstacle is due to the alleged privacy of first-person data. This is the second mistake in the reliability-based defense of privatism: it presupposes that first-person data stand or fall with their privacy. But I am not arguing against first-person data. I am arguing that they are legitimate precisely because, when properly understood, they are as public as other scientific data. The error of privatism is not in defending first-person data; the error is in defending first-person data as private.
Finally, science is rarely as clear-cut as my artificial example of A and B, in which I stipulated that by answering the same question following private methods, they get different answers. Things are murkier in real life. Yet scientific data are often disputed. In actual science, disputes over data can generally be resolved by taking appropriate measures. Such measures include controlling experimental conditions more carefully, making more explicit assumptions, establishing more rigorous procedures, and instituting more precise standards—all of which may be roughly summarized as striving for a higher degree of publicity in methods and data. An especially important way to resolve disputes over data is for a third party to make a breakthrough – using a new technique or instrument – by which new (public) data are obtained.[7] The new data explain why the original parties were getting conflicting results using their original techniques. After the new data and techniques are available, everyone can obtain mutually consistent results. None of this would be possible if the methods being used or data being produced were private.
In practice, publicity comes in degrees. It’s not so much something to establish (or eschew, as the privatist does) by stipulation, but to conquer by painstaking attention to the details of our scientific methods, by seeking to avoid the sources of bias and error that can infect our data, by comparing data with our colleagues, and by learning more about the subject matter. By these lights, privatism is an obscurantist methodology. It excuses the many epistemic-divergence-inducing tendencies that are naturally present in science—the tendency to idiosyncratic jargon, to implicit assumptions, to unclear protocols, to biased encoding and processing of data—while being a drag on the way forward: methods (and consequently, data) that are as public as we can muster.
So far, I have rejected a defense of privatism based on the alleged-though-untestable reliability of introspection. Another defense of privatism may be that (private) first-person data constitute legitimate scientific evidence insofar as they are purely factual—untainted by theory or interpretation. The claim that such untainted data are obtainable may be found in the writings of some classical introspectionists and phenomenologists.[8] In recent years, a similar claim has been resurrected by some neo-introspectionists and neo-phenomenologists (e.g., Varela 1996, Price and Aydede 2005). But this defense of privatism is no better than the previous one.
For starters, even the most “pre-theoretical” of observations are affected by our expectations, interests, and skills. Furthermore, observations must be described in a language, which comes with its own set of distinctions, presuppositions, implications, etc. There is no such thing as an unbiased observation described in a purely observational language. This theory-ladenness of observation has been a traditional source of criticism of first-person data.[9] Nevertheless, theory-ladenness is not enough to impugn privatism.
Biases are bad only if they thwart your epistemic purposes. In many cases, they don’t. As Bogen and Woodward have argued, observation is generally not theory-laden in a harmful way. To serve its epistemic purpose, observation must be reliable, that is, it must produce data that lead to detecting phenomena and discriminating correctly between the relevant hypotheses about the phenomena (Bogen and Woodward 1988, 1992). As long as observation is reliable in this sense, it doesn’t matter much how biased or theory-laden it is. Of course, there are cases in which observation biases do interfere. For instance, in the classic case of N-rays, some physicists claimed to observe the effects of N-rays when, in fact, there were no such effects to observe, because N-rays don’t exist (Klotz 1980, Nye 1980). But pernicious biases can eventually be exposed and removed.
The rub is, establishing the degree to which observation is reliable requires public scrutiny. The N-ray affair was resolved by publicly demonstrating that the observations in questions could be obtained even under conditions in which N-rays were unquestionably absent by the standards of those who purported to observe their effects. Once we bring in reliability, the second defense of privatism collapses into the first. As we saw, we cannot assume that introspection is reliable in the absence of public evidence—or at least the possibility to collect public evidence—to that effect. And that requires our data to be public.
Until now, I have argued that the proposed justifications for using private first-person data as evidence are untenable. Privatism leads to methodologically unacceptable consequences, forcing the rejection of first-person methods. It doesn’t follow that we should reject first-person data. We would have to do so only if the assumptions behind privatism were correct. It’s time to briefly examine their plausibility.
Let’s begin with (1), the premise that subjects producing first-person behaviors are the scientific observers of their mind. Although this is how some classical introspectionists and phenomenologists saw the matter, their approach is now theoretically and methodologically discredited. In other scientific psychological and neuroscientific traditions, ranging from psychophysics to cognitivism, the subjects are treated as the objects of study, not the observers. This is how it should be. Contrary to (1), the scientific observers of the mind are the psychologists and neuroscientists studying the subjects, not the subjects themselves.
This says nothing on whether first-person behaviors, or at least some of them, are the product of some kind of inner observation. They may be. Some people like the idea that first-person behaviors are the product of introspection, construed as inner observation. They point to some positive analogies: like the observation of external events, first-person reports involve detection and discrimination (of internal events), are under voluntary control, are subject to epistemic biases such as distortion by preconception and deliberate deceit, and in some cases carry a strong epistemic authority. The inner-observation model also satisfies the needs of the dualist: if the inner events are non-physical and thus can’t be detected physically, the dualist might yearn for a way to observe them. Inner observation fits this bill.
As far as this essay is concerned, I can remain neutral on inner observation. What I am arguing is simply that with respect to the methodology of science, subjects producing first-person behaviors do not play, and should not be seen as playing, the role of scientific observers. In the next section, I will argue that subjects producing first-person behaviors play the role of self-measuring instruments. Thus, for the purpose of a sound scientific methodology, we do not need to assume that first-person behaviors involve some kind of inner observation.
This result will please those who are uneasy about the inner-observation model. The main cluster of objections goes back at least to Auguste Comte (Heyd 1989, Wilson 1991): how can the same thing be the observer as well as the observed? What are the organs of such observation? In the case of public events, observation is a relation between an agent, her sensory and perceptual organs, and the observed event: the agent observes the event through her sensory and perceptual organs. But in the case of “inner observation”, it’s not clear what the sensory and perceptual organs are. And if there aren’t any, it’s not clear what it means to say that something is being observed. There is a large literature discussing this and other problems with inner observation (cf. Hatfield 2005, Gertler 2008).
One way around such problems is to deny that observation requires sensory organs. For example, Armstrong does so on the grounds that “proprioceptors, stimulation of which gives rise to bodily perception, are not [sensory] organs in the fullest sense because their operation is not under the direct control of the will” (1968, 325). Proprioception is a kind of perception, on which observation is grounded. Armstrong maintains that proprioception requires no sensory organs, because sensory organs are organs that can be moved voluntarily and proprioceptive organs cannot be so moved. If Armstrong is right, proprioception shows that observation does not require sensory organs. Armstrong’s point is persuasive only if you agree that all sensory organs can be moved voluntarily whereas proprioceptive organs cannot be moved voluntarily. It is more plausible to maintain that proprioceptive organs can be moved voluntarily – as we voluntarily move our body, we can move our proprioceptive organs with it – or, if that is somehow unacceptable, to use proprioception as a counterexample to the claim that all perceptual organs can be moved voluntarily. Finally, even granting Armstrong that proprioception requires no sensory organs, it surely requires organs. Comte’s objection can then be reformulated in terms of organs tout court.
More recently, Hatfield simply weakens the notion of observation, so that it doesn’t require sensory and perceptual organs. He construes inner observation as “the application of introspective concepts within everyday perceptual experience” (2005, 292). Since introspection applies to conscious thoughts and intentions as well as perceptions, presumably we can generalize Hatfield’s definition by omitting the term ‘perceptual’. If all there is to inner observation is the application of introspective concepts within experience, it’s hard to deny that many first-person behaviors are the product of inner observations. It’s also hard to see why this should be called ‘observation’. At any rate, whether first-person behaviors are the product of observation is orthogonal to the validity of first-person data.
Leaving inner observation aside, aren’t first-person data the mental states themselves, as (2) declares? This appears to be a category mistake, perhaps due to confusion between scientific data and sense data. Sense data are a kind of mental entity postulated by a mostly abandoned theory of perception. They were intended to constitute the foundation of empirical knowledge, but they shouldn’t be confused with scientific data.[10] By contrast, scientific data are observation statements, measurement records, and the like. Mental states are none of the above; they are not the right kind of thing to be scientific data.
Consider an analogy with other experimental sciences. In astronomy, the data are not the distant physical events under investigation. The data are the photographs and other recordings collected by the experimenters through their apparatus. Ditto in paleontology. The data are not the evolutionary events under investigation, which for the most part are long gone. The data are not even the fossils or geological strata. Rather, the data are the drawings, photographs, and other records of fossils, rocks, and geological strata. By the same token, mental states are not psychological data. Psychological data are records collected by psychologists and neuroscientists about experimental subjects and their mental states.
What about first-person data? Aren’t they at least private, as per (3)? Once we abandon (2), we no longer have reason to believe so. Ordinary psychological data are records of subjects’ behavior or neural activity. By the same token, first-person data are records of subjects’ first-person behaviors. Mental states remain private, in the innocuous sense that no one other than the subject undergoes them. But data from first-person behaviors are no more private than any other scientific data.[11]
Once we drop the assumption that first-person data are private, we no longer have reason to accept (4)—the conclusion that the validity of first-person data is untestable by public means. In fact, there is plenty of relevant public evidence.[12]
So far, I have rejected privatism about first-person data. There remains to paint a positive picture of what first-person data are and why they are public.
In a previous paper (Piccinini 2003a), I compared the process of generating first-person data to ordinary experimental processes in other sciences, processes from which experimenters collect data. Particle physicists set up accelerators to smash particles and then collect data from them. Biologists set up model organisms to generate biochemical cascades and then collect data from them. Analogously, psychologists instruct subjects to perform certain tasks, which may include generating first-person behaviors, and then collect data from them.
This approach shows that (1) the scientific observers are the experimenters, not the subjects, (2) the data are public records of first-person behaviors, not mental states, and (3) the empirically intractable question of the reliability of private data is replaced by the empirically tractable question of the validity of public (first-person) data. Like other scientific data, first-person data are useful and legitimate as scientific evidence largely to the extent that they have not been empirically invalidated, or even better, to the extent that they have been empirically validated. In sum, the use of first-person data is a form of legitimate, third-person science.[13] This approach goes in the right direction, but the analogy between the source of first-person data and ordinary experimental processes can use some refinement.
The analogy holds best when the object of investigation is the process of generating first-person behaviors itself. A large number of psychological studies focus on how subjects generate various kinds of first-person behaviors. Theories of the relevant mental processes range from Nisbett and Wilson’s (1977) theory that (at least some) first-person behaviors are the result of a priori theories that subjects hold (as opposed to empirical knowledge subjects gain about their mental states) to Ericsson and Simon’s (1993) theory that (at least some) first-person behaviors are the result of encoding the contents of the subjects’ short-term memory into verbal reports. In these and other studies, the production of first-person behaviors and its properties is the object of research.
For the most part, however, first-person behaviors are used to obtain data about psychological states and processes other than those responsible for generating first-person behaviors. Such states and processes include perception, problem solving, and many others. In the general case, it is these other processes that are the object of investigation. In order for first-person data to be relevant to the study of these other processes, the generation of first-person behaviors must be shown or assumed or at least hoped to satisfy two conditions. First, it must not interfere too much with the process under investigation. Second, it must yield information about the process under investigation. In the majority of cases, then, production of first-person behaviors plays a different epistemic role than that of being the process under investigation. It plays the role of measurement.[14]
A subject generating first-person behaviors to fulfill the purposes of a scientific observer is a self-measuring instrument. When a subject generates first-person behaviors, she implements not only (part of) the experimental materials but also (part of) the measurement apparatus. The psychological or neural system whose activity is under investigation is (part of) the experimental materials. The process of generating the first-person behaviors is (part of) the measurement process. By looking at first-person behaviors in this way, we can shed light on the epistemology and methodology of first-person data by drawing from the literature on measuring instruments (e.g., Hacking 1983, chap. 11, A. Chalmers 2003, Chang 2004). I will now briefly list some pertinent morals.[15]
Instrument users, not the instruments themselves, are the observers. If we look at first-person data as the outcome of self-measurement, we need not assume any notion of “inner observation”.
Instruments are limited in what they can measure. Like every measurement apparatus, the processes responsible for producing first-person behaviors can measure only some variables and not others. They can measure some aspects of a subject’s internal states and events but not others. Of those they can measure, they may measure some better than others. And they measure what they measure only to some degree of approximation.
Like other measurement apparatuses, subjects generating first-person behaviors need to be prepared carefully before they can yield valid results. This is the role played by the experimental set up and the instructions given by the experimenter to her subjects.
Like other measurement apparatuses, subjects generating first-person behaviors may need to be calibrated. This is the role of training. Contrary to what privatism suggests, we have plentiful public ways – independent of their specific first-person report – to tell when others are inaccurate about their mind, and sometimes we correct them.[16] This is especially obvious with children, when the training begins. After they learn how to point, babies express their desires by whining while pointing at objects. At a later stage, they may replace some of the pointing and whining with the name of the desired objects—perhaps pronounced in a recognizably desirous tone of voice. Finally, they may settle on saying something like, “I need that”. In some cases, this may be true—they do need “that”. But most of the times, they don’t need “that” at all. The grown-ups’ job is to teach them further: they just want it, they don’t actually need it. After they master this and other distinctions, children’ utterances about themselves begin to acquire the status of first-person reports. By teaching children to recognize and deploy such distinctions, adults train children to express their sensations, thoughts, and feelings. Friends, teachers, and counselors train people further, to produce more accurate and informative first-person reports. This is how many adolescents spend their time: debating how they feel about X, W, and Z, thereby enriching and refining their first-person reports. As they do this, they often call each other on their first-person bullshit—for instance, when they deny feelings that they patently have. Human subjects can self-measure their mental states, to the extent that they can, in large part thanks to years of implicit training by their community. Nevertheless, it may be useful in some experimental settings to add additional, explicit training to experimental subjects, making sure they master the relevant concepts and apply them reliably (cf. Nahmias 2002, Schwitzgebel 2004). Training the subjects is especially crucial when the subjects are non-verbal animals, such as monkeys (cf. Leopold, Maier and Logothetis 2003).
Like other measurements, first-person behaviors need to be processed via appropriate procedures and interpreted under appropriate theoretical assumptions before they can yield useful data. Different assumptions or procedures, applied to the same raw data, may yield different final data, which may support different conclusions about the phenomena.
Measuring instruments produce artifacts and aberrations, which need to be discovered and either prevented or filtered out while processing and interpreting the data. There is no failsafe procedure for discovering aberrations and artifacts. Consider microscopes. We can’t observe microscopic events without microscopes, so we have no direct method of validating or invalidating our observations through microscopes. Nevertheless, microscopists found noncircular ways to discount artifacts and aberrations and validate their data (cf. Hacking 1983, A. Chalmers 2003). Analogously, subjects engaged in first-person behaviors may be prone to illusion, delusion, forgetfulness, confabulation, wishful thinking, lying, and other confounders. To obtain valid first-person data, the effects of these factors must be minimized.
Instrument users may obtain valid data without knowing the correct theory of their instrument’s operation. Users possess practical skills and ingenuity that often allow them to distinguish observations of artifacts and aberrations from observations of actual facts. Microscopes and thermometers are examples of measuring instruments that were used quite effectively—though not without mistakes—centuries before good theories of their operation were developed (Hacking 1983, Chang 2004). By the same token, most of us are already quite adept at learning from each other’s first-person behaviors without being misled too much, even though we probably lack the correct theory of the process by which first-person behaviors are generated. With some specialized training, such as the training scientists go through, we can improve. In the long run, we should still aim for a full scientific understanding of the underlying processes.
One of the greatest benefits of the measurement framework here proposed is that it allows us to make progress on the validity of first-person data. Traditionally, both supporters and critics of first-person data focus their attention on first-person verbal reports. They see first-person data as the semantic content of verbal reports. If the problem of validity is construed as the problem of determining the truth value of first-person reports, it is tempting to conclude that only the subject (if anyone) is in a position to verify or falsify her reports.
To make matters worse, first-person reports are commonly seen as observational. This makes the validity problem especially intractable, because no one other than the subject can perform the same putative observations. This is one route towards the conclusion that first-person data are private data, whose validity cannot be established by public means. The problem becomes somewhat more tractable if we abandon the notion that first-person reports are observational, or at least that first-person observation play the role of observations within a science of mind. But it remains difficult to eliminate the suspicion that at least some kinds of report, qua reports, are based on private evidence inaccessible to anyone other than the subject.
By renouncing the conception of first-person data as the semantic content of first-person reports, the measurement approach side-steps the problem of the reliability of first-person reports. This is because in the first instance, measurement instruments do not produce reports. They produce physical outputs. To be sure, the outputs of measurement instruments need to be interpreted, and when interpreted, they give rise to reports or other forms of representation, such as diagrams and equations. But the production of the relevant reports or other representations is done by the experimenters, not by the measurement instruments.
Typically, at least some aspects of the interpreting and reporting is automated and built into the instruments. Many instruments have labeled marks that help read their outputs. Some instruments feed into computers that process the information and generate linguistic outputs. But the interpreting and reporting is still done on the basis of assumptions built into the measurement process by those who design, build, calibrate, and use the instruments. So even in such cases, there is no intractable issue of privacy.
The important point is not that first-person behaviors should never be construed as reports, or even as observational reports. In fact, in many cases, it may be appropriate and helpful to call first-person behaviors 'reports'. The important point is that whether first-person behaviors are construed as reports is of little consequence to whether first-person data are valid. First-person data are scientifically valuable insofar as they are indicative of mental states and events. The epistemic burden of establishing how accurate, reliable, and precise the data are falls on the scientists, not the subjects.
Of course, human beings are different from other measurement instruments in several respects. They have both special capacities, which aid self-measurement, and special sources of interference. On one hand, most human beings come already equipped with the ability to understand instructions, direct their attention, assess evidence, and speak their mind. On the other hand, they may be prone to cognitive deficits, self-serving bias, dishonesty, prejudice, and other conditions that confound self-measurement. In short, human beings are epistemic agents. This explains why it is so tempting to think of first-person data as the private outcome of introspection. We like to think of ourselves as knowing our minds better than anyone, scientists included. Be that as it may, the scientific validity of first-person data has a different basis. First-person data are legitimate because they are public recordings of public behaviors. They are useful insofar as investigators instruct the subjects appropriately, eliminate potential confounders, record behaviors carefully, and interpret the results soundly.
According to the measurement framework here proposed, first-person verbal reports are a subclass of first-person behaviors. All such behaviors are, in the first instance, physical outputs epistemologically on a par with the outputs of other measurement instruments. Instrument users take their instruments’ outputs and interpret them to yield data. Gathering data from first-person reports becomes a special case of gathering data from first-person behaviors, which becomes a special case of gathering data from a measuring instrument. The resulting data are no less public than the data obtained from any other measurement instrument.
First-person data can be fruitfully seen as the outcome of a process of self-measurement. When seen under this light, the problems of purported privacy and lack of public validation of first-person data evaporate. Data from measuring instruments, including first-person data, are public, and the degree to which they are valid can be established by public methods.
References
Alexandrova, A. (2005). "Subjective Well-Being and Kahneman's 'Objective Happiness." Journal of Happiness Studies 6: 301-324.
Alston, W. (1972). "Can Psychology Do Without Private Data?" Behaviorism 1(1): 71-102.
Angner, E. ms. "The Measurement-Theoretic Argument against Subjective Measures of Wellbeing."
Armstrong, D. M. (1968). A Materialist Theory of the Mind. London, Routledge & Kegan Paul.
Aydede, M., Ed. (2005). Pain: New Essays on Its Nature and the Methodology of Its Study. Cambridge, MA, MIT Press.
Baars, B. J. (2003). "How Brain Reveals Mind: Neural Studies Support the Fundamental Role of Conscious Experience." Journal of Consciousness Studies 10(9-10): 100-114.
Bogen, J. and J. Woodward (1988). "Saving the Phenomena." Philosophical Review XCVII(3): 303-352.
Bogen, J. and J. Woodward (1992). "Observations, Theories and the Evolution of the Human Spirit." Philosophy of Science 59(4): 590-611.
Brewer, W. F. and B. L. Lambert (2001). "The Theory-Ladenness of Observation and the Theory-Ladenness of the Rest of the Scientific Process." Philosophy of Science 68(3): 176-186.
Bucklew, J. (1955). "The Subjective Tradition in Phenomenological Psychology." Philosophy of Science 22: 289-299.
Chalmers, A. (2003). "The Theory-Dependence of the Use of Instruments in Science." Philosophy of Science 70(3): 493-509.
Chalmers, D. (1999). "First-person Methods in the Science of Consciousness." Consciousness Bulletin Fall 1999.
Chalmers, D. (2004). How Can We Construct a Science of Consciousness? The Cognitive Neurosciences III. M. S. Gazzaniga. Cambridge, MA, MIT Press.
Chang, H. (2004). Inventing Temperature: Measurement and Scientific Progress. Oxford, Oxford University Press.
Cytowic, R. E. (2003). "The Clinician's Paradox: Believing Those You Must Not Trust." Journal of Consciousness Studies 10(9-10): 157-166.
Dennett, D. C. (1991). Consciousness Explained. Boston, Little, Brown & Co.
Dennett, D. C. (2003). "Who's on First? Heterophenomenology Explained." Journal of Consciousness Studies 10(9-10): 19-30.
Dennett, D. C. (2007). "Heterophenomenology Reconsidered." Phenomenology and the Cognitive Sciences.
Dickson, M. (1999). "The Light at the End of the Tunneling: Observation and Underdetermination." Philosophy of Science 66(3): 47-58.
Ericsson, K. A. (2003). "Valid and Non-Reactive Verbalization of Thoughts During Performance of Tasks: Toward a Solution to the Central Problems of Introspection as a Source of Scientific Data." Journal of Consciousness Studies 10(9-10): 1-18.
Ericsson, K. A. a. S., H. A. (1993). Protocol Analysis: Verbal Reports as Data. Cambridge, MA, MIT Press.
Estany, A. (2001). "The Thesis of Theory-Laden Observation in the Light of Cognitive Psychology." Philosophy of Science 68(2): 203-217.
Feigl, H. (1958). The 'Mental' and the 'Physical'. Minnesota Studies in the Philosophy of Science, II. H. Feigl, M. Scriven and G. Maxwell. Minneapolis, University of Minnesota Press: 370-497.
Gertler, B. (2008). Self-Knowledge. The Stanford Encyclopedia of Philosophy (Winter 2008 Edition). E. Zalta. URL = <http://plato.stanford.edu/archives/win2008/entries/self-knowledge/>. .
Gertler, B. (forthcoming). Introspection. The Oxford Companion to Consciousness. T. Bayne, A. Cleeremans and P. Wilken. Oxford, Oxford University Press.
Goldman, A. (1997). "Science, Publicity, and Consciousness." Philosophy of Science 64: 525-545.
Goldman, A. (2000). "Can Science Know When You're Conscious? Epistemological Foundations of Cognitive Research." Journal of Consciousness Studies 7(5): 3-22.
Goldman, A. (2004). "Epistemology and the Evidential Status of Introspective Reports." Journal of Consciousness Studies 11(7-8): 1-16.
Hacking, I. (1983). Representing and Intervening. Cambridge, Cambridge University Press.
Hatfield, G. (2005). Instrospective Evidence in Psychology. Scientific Evidence: Philosophical Theories and Applications. P. Achinstein. Baltimore, Johns Hopkins University Press: 259-286.
Haybron, D. (2007). "Do We Know How Happy We Are? On Some Limits of Affective Introspection and Recall." Nous 41(3): 394 - 428.
Heyd, T. (1989). "Mill and Comte on Psychology." Journal of the History of the Behavioral Sciences 25: 125-138.
Hohwy, J. and C. Frith (2004). "Can Neuroscience Explain Consciousness?" Journal of Consciousness Studies 11(7-8): 180-198.
Horst, S. (2005). "Phenomenology and Psychophysics." Phenomenology and the Cognitive Sciences 4(1): 1-21.
Hurlburt, R. T. (1993). Sampling Inner Experience in Disturbed Affect. New York, Plenum.
Hurlburt, R. T. and E. Schwitzgebel (2007). Describing Inner Experience? Proponent Meets Skeptic. Cambridge, MA, MIT Press.
Klotz, I. M. (1980). "The N-Ray Affair." Scientific American 242(5): 168-175.
Leopold, D. A., A. Maier, and L. Logothetis (2003). "Measuring Subjective Visual Perception in the Nonhuman Primate." Journal of Consciousness Studies 10(9-10): 115-130.
Levine, J. (1994). "Out of the Closet: A Qualophile Confronts Qualophobia." Philosophical Topics 22: 107-126.
Nahmias, E. A. (2002). "Verbal Reports on the Contents of Consciousness: Reconsidering Introspectionist Methodology." Psyche 8(21): http://psyche.cs.monash.edu.au/v8/psyche-8-21-nahmias.html.
Nisbett, R. E. and T. D. Wilson (1977). "Telling More than We Can Know: Verbal Reports on Mental Processes." Psychological Review 84: 231-359.
Nye, M. J. (1980). "N-rays: An Episode in the History and Psychology of Science." Historical Studies in the Physical and Biological Sciences 11(1): 125-156.
Piccinini, G. (2003a). "Data from Introspective Reports: Upgrading from Commonsense to Science." Journal of Consciousness Studies 10(9-10): 141-156.
Piccinini, G. (2003b). "Epistemic Divergence and the Publicity of Scientific Methods." Studies in the History and Philosophy of Science 34(3): 597-612.
Price, D. D. and M. Aydede (2005). The Experimental Use of Introspection in the Scientific Study of Pain and Its Integration with Third-Person Methodologies: The Experiential-Phenomenological Approach. Pain: New Essays on the Nature of Pain and the Methodology of its Study. M. Aydede. Cambridge, MA, MIT Press: 243-273.
Prinz, J. J. (2004). "The Fractionation of Introspection." Journal of Consciousness Studies 11(7-8): 40-57.
Robbins, P. (2006). "The Ins and Outs of Introspection." Philosophy Compass 1(6): 617–630.
Roy, J.-M. (2003). Phenomenological Claims and the Myth of the Given. The Problem of Consciousness: New Essays in Phenomenological Philosophy of Mind. E. Thompson. Calgary, University of Calgary Press: 1-32.
Samsonovich, A. V. and L. Nadel (2005). "Fundamental Principles and Mechanisms of the Conscious Self." Cortex 41: 669-689.
Schooler, J. W. and C. Schreiber (2004). "Experience, Meta-consciousness, and the Paradox of Introspection." Journal of Consciousness Studies 11(7-8): 17-39.
Schwitzgebel, E. (2004). "Introspective Training: Reflections on Titchener's Lab Manual." Journal of Consciousness Studies 11(7-8): 58-76.
Schwitzgebel, E. (2007). "No Unchallengeable Epistemic Authority, of Any Sort, Regarding Our Own Conscious Experience – Contra Dennett?" Phenomenology and the Cognitive Sciences 6(1-2): 107-113.
Schwitzgebel, E. (2008). "The Unreliability of Naive Introspection." Philosophical Review 117 245-273.
Sellars, W. (1956). Empiricism and the Philosophy of Mind. Minnesota Studies in the Philosophy of Science, Vol. I, The Foundations of Science and the Concepts of Psychology and Psychoanalysis. H. Feigl and M. Scriven. Minneapolis, University of Minnesota Press.
Shapere, D. (1982). "The Concept of Observation in Science and Philosophy." Philosophy of Science 49(4): 485-525.
Snyder, D. J., K. Fast, and L. Bartoshuk (2004). "Valid Comparisons of Suprathreshold Sensations." Journal of Consciousness Studies 11(7-8): 96-112.
Titchener, E. B. (1902). An Outline of Psychology. New York, Macmillan.
Varela, F. (1996). "Neurophenomenology: A Methodological Remedy for the Hard Problem." Journal of Consciousness Studies 3(4): 330-349.
Varela, F. J. and J. Shear (1999). "First Person Methodologies: What, Why, How?" Journal of Consciousness Studies 6(2-3): 1-14.
Vermersch, P. (1999). "Introspection as Practice." Journal of Consciousness Studies 6(2-3): 17-42.
Wilson, F. (1991). "Mill and Comte on the Method of Introspection." Journal of the History of the Behavioral Sciences 27: 107-128.
Wilson, T. D. (1994). "The Proper Protocol: Validity and Completeness of Verbal Reports." Psychological Science 5(5): 249-252.
Wilson, T. D. (2003). "Knowing When to Ask: Introspection and the Adaptive Unconscious." Journal of Consciousness Studies 10(9-10): 131-140.
Wood, L. (1940). "Inspection and Introspection." Philosophy of Science 7(2): 220-228.
[1] Thanks to Erik Angner, Jim Bogen, David Chalmers, Jordan Dodd, Robert Gordon, Marcin Miłkowski, Philip Robbins, Martin Roth, Anna-Mari Rusanen, and Eric Schwitzgebel for comments and discussion. Thanks to audiences at University of Missouri – St. Louis, Washington University in St. Louis, the 2006 PSA Meeting, and the Consciousness Online conference. The writing of this paper was supported in part by two University of Missouri Research Grants.
[2] This suggestion was made by an anonymous referee.
[3] Other authors agree to varying degrees. For instance, Hatfield 2005 and Gertler forthcoming appear to endorse something very close to (1)-(4), Price and Aydede 2005 appear to endorse (1)-(3), Horst 2005 and Levine 1994 appear to endorse (2), and Roy 2003 appears to endorse at least (1). A view superficially similar to privatism is defended in a penetrating article by Alston (1972). But while Alston forcefully defends theses that sound exactly like (1)-(3), he equally forcefully rejects (4). In fact, he defends the legitimacy of first-person data precisely on the grounds that they can be publicly validated. As the considerations to follow should make clear, (4) follows from (1)-(3), so Alston’s rejection of (4) turns out to be inconsistent with privatism as I define it. If first-person data can be validated by public procedures, as Alston maintains, they can’t be truly private after all (in the relevant sense).
[4] As I said before, I prefer to discuss the validity of first-person data without bringing in the theoretically problematic notion of introspection. Since the authors I’m discussing address the matter in terms of introspection reliability, I will temporarily go along with them for the sake of the argument.
[5] In another paper, Goldman adds another test of reliability, similar to suggestions made by Alston (1972):
We can also demand that introspection should pass the test of not yielding too many false beliefs ... when combined with other presumptively reliable processes or procedures. Here is an example of how introspection can pass such a test. I now have an introspectively formed belief that I currently intend to snap my fingers in the next two minutes. I have a background belief that intentions usually stick around for a while and get acted upon if they are not overridden. I therefore predict that I shall indeed snap my fingers in the next two minutes. If that action is then observed within the specified time interval, this partly ‘corroborates’ the reliability of introspection. That is, it corroborates introspection’s reliability in the weak sense of not defeating or not challenging introspection’s reliability. There was a chance that introspection (together with other processes) should have yielded a false belief, but, assuming I can trust my visual observations, this did not occur. So introspection survives a test of sorts (Goldman 2000, 19).
But the test described here is inconsistent with privatism. If the deliverances of introspection are truly private, then they can’t be challenged by their inconsistency with putative generalizations about the mind (such as the generalization that people tend to act upon their intentions); at most, the deliverances of introspection can be used to challenge those generalizations. I take the above test as implying the abandonment of privatism; as such, I commend it.
[6] It goes to Goldman’s credit that he changed his mind on this point, from maintaining that introspection cannot be independently corroborated by other sources of evidence (1997) to maintaining the opposite (2004, 6).
[7] I owe the present point to Heather Tienson.
[8] Cf. Titchener:
When we introspect, we must be absolutely impartial and unprejudiced. We must not let ourselves be biassed by any preconceived idea. We are likely to think that, in all probability, a certain thing will happen, or we may actually want to obtain a given result, to confirm some view which we have already formed. In either case, we are in danger of mistaken observation. We ought to be ready to take the facts precisely as they are (Titchener 1902, 45).
[9] For some criticisms of first-person methods on grounds of theory-ladenness or similar notions, see Bucklew 1955 and Dennett 1991. In spite of these critiques, the claim that first-person data can be theory-free continues to be found in the literature. For some recent discussions of the theory-ladenness of observation, see Dickson 1999, Brewer and Bruce 2001, Estany 2001. Among proponents of first-person methods who discuss the theory-ladenness of observation explicitly, Roy 2003, insofar as I understand him, seems to admit that observation is theory-laden even in the case of self-observation. So does Vermensch 1999, 39.
[10] Herbert Feigl used the expression ‘first-person data’ to mean something like sense data: “[T]he first person data of direct experience are, in the ultimate epistemological analysis, the confirmation basis of all types of factual knowledge claims. This is simply the core of the empiricist thesis over again.” (Feigl 1958, p. 437, emphasis original). See also Wood 1940 for a similar notion, which he calls ‘inspective data’. Clearly Feigl was in the grips of a view of evidence according to which all empirical evidence is reducible to something like sense data. After we update our philosophy of evidence (following Sellars 1956, Shapere 1982, Bogen and Woodward 1988, and others), we no longer postulate sense data, let alone follow Feigl in believing that they constitute the confirmation basis for all factual knowledge claims. At any rate, I am not suggesting that Feigl himself confused first-person data in his sense with scientific data (in the present sense), only that some recent supporters of privatism appear to.
[11] David Chalmers has replied that the above points about ‘data’ are merely terminological. But Chalmers’s alternate use of ‘data’ - by ‘datum’, he does not mean observation record but “directly observable phenomenon or fact” (personal communication) – distorts the methodological question under discussion. If we follow Chalmers in using ‘data’ to mean directly observable phenomena or facts and claiming that mental phenomena or facts are private – i.e., directly observable only by the subject – we equivocate between two senses of ‘directly observable’. In one sense, a mental event is directly observable just in case a subject can be aware of experiencing it. In this sense, mental events are surely private. But this is irrelevant to scientific methodology. In the sense that pertains to scientific methodology, a phenomenon or fact is directly observable just in case it can be detected by a reliable (public) procedure (cf. Shapere 1982). In this latter sense, legitimate objects of scientific investigation – including mental events – are public, i.e., detectable by all competent investigators using the same reliable procedures. So even if we decide to use ‘data’ to mean directly observable phenomena or facts, first-person data are either public (in the relevant sense) or scientifically illegitimate.
[12] There is no room to review the evidence here. For evidence that first-person data are invalid or questionable under certain conditions, see Nisbett and Wilson 1977, Wilson 1994, Haybron 2007, Hurlburt and Schwitzgebel 2007, 2008. For empirical evidence that first-person data are valid under certain conditions, see Ericsson and Simon 1993, Ericsson 2003, Baars 2003, Leopold, Maier and Logothetis 2003, Wilson 2003, Cytowic 2003, Schooler and Schreiber 2004, Hohwy and Frith 2004, Snyder, Fast and Bartoshuk 2004, Hurlburt and Heavey 2006. Even Chalmers 2004 says that empirical investigations can distinguish reliable from unreliable first-person reports, in apparent conflict with his claim (in the same essay) that the reliability of introspection is untestable because we don’t have a “consciousness meter”.
[13] In a recent paper, Dennett appears to endorse my proposal while attempting to assimilate it to his “heterophenomenology” (2007, fn. 1). Heterophenomenology has virtues, including taking first-person data to be public. But I also think heterophenomenology can be improved upon. I disagree with, among other things, Dennett’s claims that (i) scientists should be agnostic about the truth value of first-person reports and (ii) they should interpret all first-person reports as expressions of beliefs (e.g., Dennett 2003). I also disagree that the primary explananda of a science of consciousness are the first-person behaviors; in my view, the primary explananda are the mental phenomena. Unfortunately, I lack the space for the detailed discussion that heterophenomenology deserves. Views that I find more congenial than Dennett’s include Nahmias 2002, Goldman 2004, Haybron 2007, and Angner ms. For an insightful discussion of some problems with heterophenomenology, see also Schwitzgebel 2007.
[14] I intend this point and the rest of this paper to supplement (as opposed to repudiate) what I wrote in Piccinini 2003a.
[15] Notice that although measurement is mentioned by many scientists while discussing first-person data, it is usually invoked to characterize the outcome of the process of data collection by scientists, not the role of the subject producing the first-person behaviors. The role of the subject is characterized by most commentators—even those who eschew privatism—as that of performing “inner observations”.
[16] For more on our independent evidence about people’s mind and how it comes about, see Piccinini 2003a, Section II.