Introspection, psychology of

DOI: 10.4324/9780415249126-W019-1
Version: v1,  Published online: 1998
Retrieved June 09, 2023, from

4. Epistemic status and scientific role

Many philosophers have taken issue with the Cartesian view that introspection is infallible. For example, James held that because introspection involves conceptualization, it suffers from fallibility just as any mode of observation does. Armstrong agrees that introspective awareness is fallible, but for a different reason, namely, that the state being introspected and the introspecting state are ‘distinct existences’, and, hence, connected only contingently.

Psychologists have focused not so much on the epistemic status of introspective awareness per se but on whether introspection can constitute a reliable method for the scientific study of the human mind. There were considerable efforts in the nineteenth century to found psychology on careful introspective reports, and the work of Weber, Fechner and Helmholtz gave rise to important work on the ‘psychophysics’ of sensation (or the relation of introspected to physical magnitudes of, for example, light and sound) that continues to the present day (see Stevens 1951; Sekuler and Blake 1994). In contrast, the ‘structuralist’ programme of Wundt and Titchener, which sought to discover the basic elements of thought and the laws by which they combine into more complex mental experiences, was far less successful.

In the early nineteenth century, Comte (1830) issued two challenges to a general introspective psychology. First, he asserted that introspection of intellectual (as opposed to emotive) activity is impossible. The conscious self is, in some sense, essentially unified. Hence, it cannot ‘split’ itself into two, one portion engaged in first-order intellectual activity with a second portion looking on. Second, even supposing that introspection is possible, it is worthless, presumably, because it is unreliable (see Consciousness §6).

Psychologists and philosophers interested in scientific psychology have responded to these challenges in a variety of ways. For example, James (1890) agreed with Comte that introspective awareness of ‘on-line’ intellectual activity is impossible. However, since we clearly are often aware of what passes in our minds, we must ask how this can be. His view, following Mill (1865), was that introspection is actually retrospection. We are not aware of our occurrent states as we have them but, rather, of our immediate memories of those states. Since these memories are not part of the stream of consciousness, the impossible split in the conscious self is avoided (see Consciousness §10).

Comte’s second challenge is, perhaps, the more serious of the two. For it might well be the case that although introspection is possible, the process of introspecting distorts the first-order mental activity being introspected. The threat of such distortion was a serious problem for the introspectionist psychologists of the late nineteenth and early twentieth century. In fact, some have argued that Titchener’s failure to take the problem of unreliability sufficiently to heart contributed not only to the downfall of introspectionist psychology in particular, but also to the radical behaviourist ‘revolution’ and the eclipse of any mentalistic psychology for much of the twentieth century (see Behaviourism, methodological and scientific).

Behaviourism has itself now been overthrown in favour of an approach that, once again, takes the mental seriously. As a consequence, psychologists are struggling with the question of whether and to what extent introspective awareness can give us information about our mental states and processes. A contemporary challenge to the reliability of introspection comes from the work of Nisbett and Wilson (1977) who considered the psychological causes of our judgments and behaviour. They argued that people have little or no privileged introspective access to such causal information. Instead, their reports are based on implicit, cultural ‘theories’ that they apply to themselves as they do to others.

In one classic study cited by Nisbett and Wilson, subjects were asked to tie together the ends of two cords hung from the ceiling that were placed far enough apart so that one cord could not be reached while holding on to the other. To solve the problem, subjects were permitted to use any of the various objects strewn about the room. There were several solutions, one of which involved attaching a weight to the end of one cord, and then swinging it so that it could be caught while holding on to the second cord. Most subjects could not discover this solution on their own. However, when cued by the experimenter casually putting one of the cords in motion in the periphery of the subjects’ vision, they typically solved the problem within 45 seconds. Subsequently, though, when subjects were asked how they arrived at this solution, the experimenter’s hint was mentioned by less than a third of the subjects.

The most powerful defence of introspection in recent years comes from two cognitive scientists, Ericsson and Simon (1980). They argued that verbal reports of mental activities constitute important data, but that we can only assess the epistemic value of this data if we understand the mechanisms by which the reports are generated. They then proposed a model according to which cognitive information can be stored in various memories, including short-term memory (STM), which has a limited capacity and a short duration, and long-term memory (LTM), which has a very large capacity and is relatively permanent. Information that is attended to is kept in STM and is directly accessible for verbal report. Information stored in LTM must be retrieved before it can be reported (see Memory).

There are many different kinds of verbal report. First, we can distinguish reports that are concurrent with the execution of the primary task (such as solving a problem in mathematics) from reports that are retrospective. Second, some verbal reports (‘level I’) require no additional processing. This is so when information used in the primary task is already in verbal form (in STM) and simply requires articulation. In contrast, there are verbal reports that do require additional processing, either recoding, say, from visual to verbal form (‘level II’), or filtering and providing information (such as the reasons for certain behaviour or choices) to which attention would not ordinarily be paid (‘level III’).

The Ericsson and Simon model predicts what a host of experimental studies have borne out; that each of these different types of verbal report will differ in accuracy. In particular, concurrent level I reports neither change nor slow the course or structure of the primary-task cognitive process. Concurrent level II reports may slow the primary-task cognitive processes, but also do not change them. However, for level III reports, there may well be substantial effects on task performance. Finally, the accuracy of retrospective reports depends not only on the level involved but also on the duration of the task and the lag between completion and the report. As both the duration and lag time increases, accuracy will decrease.

Ericsson and Simon claim that their model is quite consistent with Nisbett and Wilson’s findings. In particular, in the studies reviewed by Nisbett and Wilson, the retrospective procedures used to elicit the actors’ reports were not those the model identifies as eliciting valid reports. For example, in the cord-tying task, subjects do not report the causal role of the hint because by the time they are queried, this information is no longer in short-term memory.

Citing this article:
Von Eckhardt, Barbara. Epistemic status and scientific role. Introspection, psychology of, 1998, doi:10.4324/9780415249126-W019-1. Routledge Encyclopedia of Philosophy, Taylor and Francis,
Copyright © 1998-2023 Routledge.

Related Articles