Psychology of Consciousness (Psyc. 358)
A.W. Kaszniak, Professor
Class Notes for 1/31/02
How well can we know ourselves? Further Exploration of Introspection
Video-tape interview of Robert Ornstein, Ph.D.:
Ornstein asserts that what we think about ourselves (i.e., our self-concept and our introspective self-awareness) is actually an illusion, and reflects the output of just another processing module that has access to the outputs of some (but by no means all) of the other processing modules. I has, however, no access to the actually processes of these modules. Thus, according to Ornstein, introspection does not provide direct access to mental processes. This conclusion is similar to the theoretic position concerning introspection that was put forward by Nisbett & Wilson, as described in your reading for today.
Nisbett & Wilson's anti-introspectionist theory:
In an influential 1977 paper entitled "Telling More than We Can Know," they argued that higher-order mental processes (those involved in judgments and decisions leading to voluntary actions) are non-conscious. According to their perspective, when people try to give introspective reports on the causes of their behavior, what they are really doing is making reasonable inferences about what the causes must have been.
Nisbett & Wilson reported on a series of experiments designed to determine whether people can report the causes of their behavior. Subjects were tested under two or more stimulus conditions that were known to produce measurable differences in behavior, where they had to make inference-based responses. Afterward, they were asked to report why they responded the way they did. These reports were analyzed to determine whether the subjects reported the stimuli that the experimenters manipulated and knew to have causal influence on the subjects' responses.
The results were interpreted by Nisbett & Wilson to support three main conclusions: (1) People do not have introspective access to the causal relationship between stimuli and their responses. They cannot accurately report from introspection which stimuli affected their responses, and/or they cannot report how the stimuli affected their responses; (2) Rather, reports of effects of stimuli on responses are based on a priori theories (prior beliefs) about the causal connections between the stimuli and responses; (3) When subjects' reports on stimulus-response relationships are correct, it is because their a priori theories happen to be correct, not because of correct introspection.
It was argued that people make post hoc (after-the-fact) inferences about the causes of their responses, based on some combination of reasoning, observations, and prior knowledge or beliefs, rather than on observations alone. Nisbett & Wilson suggested several possible origins for the a priori theories: (1) There are socially-learned rules that govern much of behavior; (2) The culture provides theories about the causes of behavior and feelings (i.e., "folk psychology"); (3) A causal theory may be based on personal observation of covariation between types of stimuli and types of responses; (4) People may generate causal hypotheses based on judgments of the connotative similarity between the stimuli and responses (i.e., the stimuli and responses seem to go together, such that one seems to imply the other).
The Nisbett & Wilson position has two important implications: (1) People may sometimes report accurately on the causes of their behavior. However, since their reports are based on a priori theories, their reports will be correct only if their a priori theories are correct. (2) People's reports on the causes of their own behavior will be no more accurate than the reports of observers who use the same a priori theories.
Based on these implications, Nisbett & Wilson derived the following criterion for introspective access to the causes of behavior: Introspective access would be demonstrated if it were shown that real subjects who respond in a particular experimental situation (e.g., judgments of the likability, sympathy, flexibility & intelligence of job applicants) can subsequently make verbal reports on the causes of their responses that are more accurate than the reports obtained from observers. Across several experiments, they failed to find clear evidence for introspective access, based on this criterion.
Nisbett & Wilson suggested that people have an illusion of introspective access because they confuse awareness of mental contents with awareness of mental processes. Mental processes make various computations and judgments on pertinent information to produce particular decisions, which in turn guide behavior. They argued that we do not have introspective access to these mental processes. Rather, we have introspective access to the results (or products) of those processes, which are mental contents (e.g., decision outcomes). We may also have access to intermediate results at various stages of the mental processes that lead to decisions.
Critique of Nisbett & Wilson theory:
Several critics have argued that Nisbett & Wilson did not clearly distinguish mental processes from mental contents. They did not provide definitions that were clear enough to enable one reliably to judge whether subjects were describing mental processes or mental contents.
Mental process can be described at several different levels. People may be able to make introspective reports on the higher-level mental processes (e.g., strategies or rules for problem solving or social behavior), but not to intermediate or lower-level processes.
Bowers (1984) argued that while people may sometimes have introspective access to the stimuli that cause their behavior, it is logically impossible to have introspective access to the causal connection between the stimuli and their behavioral effects. Rather, understanding of the causal connections between events is always and necessarily a theoretical inference. For example, you might observe that almost every time that you feel tired you also are irritable, and so you infer that tiredness causes irritability. But you do not directly observe the process that links tiredness with irritability (in fact, tiredness might not cause irritability; they might both be caused by another factor that happens sometimes to be correlated with both of them).
Other critics noted problems of research method: (1) subjects reports on causal influences on their behavior were all made retrospectively, several minutes or hours after the critical behavior occurred. They may have forgotten pertinent information that they would have known if they had been questioned sooner; (2) Most of the experiments involved between-subjects designs. In order to understand how a stimulus influences one's behavior, it is important to have an opportunity to experience covariation between the stimulus and the response. In the between-subjects experiment it might be unrealistic to expect subjects to know that a stimulus influenced their behavior, when they experienced only one stimulus condition and made only one response. Within-subjects designs would be preferable.
Evidence supporting introspective access:
Since the time of the Nisbett & Wilson (1977) review, several studies have employed within-subjects designs to determine whether there is evidence for introspective access to the causes of behavior. In each of these studies, real subjects (actors) made judgments of each of several different stimuli (e.g., different people), in situations where the stimuli varied on several different attributes.
Most of these within-subjects design studies found evidence for introspective access, by showing that the actor's reports of the effects of the different stimulus attributes on their judgments (e.g., of the friendliness or intelligence of people) were somewhat more accurate than observer's estimates. However, the actors' and observers' judgments were correlated, suggesting that they are using the same a priori theories to explain their judgments. None-the-less, the fact that actors' estimates were somewhat more accurate than those of observers suggest that actors are using some privileged information- possibly from introspection - that is not available to the observers to explain the actors' judgments.
Taking into account results of the more recent within-subjects design studies, Wilson & Stone (1985) suggested that people likely use several types of information when explaining the causes of their own behavior: (1) Shared theories (theories about the causes of behavior known by both actors and observers in the same culture or subculture); (2) Privileged information (known only to actors - the people whose behavior is in question - and not by observers), which can be either (a) introspective data (i.e., the "workings of one's own mind"), (b) covariation data (accumulated knowledge of one's different responses to different stimuli, that can be used to infer a cause-effect relationship), or (c) idiosyncratic theories (theories unique to the individual, that may be wrong due to faulty data analysis or faulty inference).
When observers are provided with stimulus-response covariation data, they are more accurate (in estimating the impact of the different stimulus attributes on actors' judgments) than observers without covariation data, and almost as accurate as the actors themselves. This suggests that actors also used covariation data. However, ther accuracy of actors' reports cannot be attributed entirely to covariation information (i.e., actors appear to use some other privileged information, not known to the covariation observers). Actors' reports thus appear to be based partly on the introspections of individual judgments, in addition to their after-the-fact covariation analysis of the relationships between stimulus attributes and their responses.
In summary, although a priori theories are an important component of people's causal explanations, they are not the sole influence, as originally hypothesized by Nisbett & Wilson. Actors also have privileged information access that includes some degree of introspective access to pertinent causal stimuli and thought processes, as well as better access (than observers) to stimulus-response covariation data about their own behavior.