[ CogSci Summaries home | UP | email ]

Gilbert, D. T. (1991). How Mental Systems Believe. American Psychologist, 46(2), 107-119.

author = {Daniel T. Gilbert},
title = {How Mental Systems Believe},
journal = {American Psychologist},
year = {1991},
volume = {46},
number = {2},
pages = {577-609}

Author of the summary: Stephen Jones, 2011, push_pause@yahoo.com

Cite this paper for:

The Process of Belief

Propositions may be automatically assessed as true upon comprehension. This proposal runs contrary to the commonsense concept of belief which depicts the 'true of false' evaluation of a proposition as occurring in a stage separate from and subsequent to comprehension.

(Descartes 1644/1984): The Cartesian view of the belief procedure posits an initial creation of a mental representation (the ‘passive’ comprehension stage) followed by a subsequent deliberate (‘active’) assessment stage, during which the comprehended proposition is assessed as either true (resulting in belief) or false (resulting in disbelief). [108]

(Spinoza 1677/1982): Spinoza rejected the Cartesian model, suggesting that comprehension and acceptance were identical operations. A subsequent ‘active’ assessment stage would be performed only if a realization were made that the automatically believed proposition was in conflict with another belief. [108]

                  Initial Representation                    Assessment
:  Comprehension                        ---> Acceptance / Rejection
Spinozan:   Comprehension & Acceptance ---> Rejection

In both models, every newly comprehended proposition is initially represented as ‘untagged’. In the Cartesian model, this is so only until it is subsequently tagged true or tagged false. A Spinozan model would evaluate untagged representations as true, only needing to apply a ‘false tag’ if the proposition is rejected during the subsequent evaluation stage. [109]

Development of Belief

Supporting the Spinozan model, observations in child development show that acceptance and denial of propositions are not equally effortful operations, and that related behaviours do not emerge at the same stages of development.
(Ceci, Ross & Toglia, 1984): Studies of young children show high ‘suggestibility’ when compared to older children. The ability to 'doubt' develops in later stages, along with other more complex behaviours.[111]
(Bloom 1970; Pea, 1980): Linguistically, the use of ‘no’ to deny the truth of a proposition is the last of the word's uses to develop.[110]

Breakdown of Belief

The two models make predictions about what sort of output behaviour results when heavy cognitive load causes a premature 'early output’ from early processing 'modules' (stages). Under adequate cognitive resource depletion, Cartesian systems should fail to evaluate comprehended propositions while Spinozan systems should fail to disbelieve comprehended / believed propositions. [110]

(Lifton, 1961; Bem, 1966): Studies of the effects of cognitive resource depletion show predictable decreases in subjects’ ability to reject doubtful propositions. ‘Brainwashing’ or forced confession programs rely on sleep deprivation tactics to accelerate subjects’ acceptance of the program. [111]

(Zuckerman, Depaulo & Rosenthal, 1981): Hearers will tend to evaluate a speaker’s autobiographical statement as true, even if it is known that the speech act is likely to be incorrect or dictated by an authority. Spinozan systems would also demonstrate this ‘truthfulness bias’, only rejecting the truth of the attribution at the later evaluation stage, if and when time and cognitive resources are available.[112]

(Greenberg 1966): Psycholinguistic research assumes that ‘unmarked words’ (eg. ‘happy’), generally stand for more basic concepts vs. their 'marked' counterparts (eg. ‘unhappy’). In English and Latin, when speaking about belief, the acceptance condition generally uses the unmarked verb. (eg. We describe believed statements as ‘true’, not ‘unfalse’). This tendency may reveal that acceptance is a more basic behaviour than rejection. [112]

(Clark & Clark 1977): One paramorphic* model which successfully replicates human reaction times when verifying the truth of sentences assumes that all incoming propositions are initially tagged as ‘true’, and evaluated against pre-existing knowledge at a later evaluation stage. [113]

(Clark & Clark, 1978): Negative statements use ‘not’ to invalidate an affirmative proposition. “Negative statements are about positive statements, while affirmatives are directly about the world”. (pg.3).
(Wegner, Wenzlaff, Kerker & Beattie, 1981): Negation functions as a secondary instruction to reject a proposition comprehended in its affirmative form. Under cognitive load, participants will remember negated statements as true, but never affirmative statements as false. The latter case should occasionally occur, according to the prediction made by the Cartesian model. [113]

Suspension of Belief

Two claims follow the Spinozan model: 1) The Asymmetry Hypothesis – that the belief stage precedes the disbelief stage temporally and 2) The Unity Hypothesis – that comprehension and belief are the same operation. [113]

Though the Cartesian model disagrees with both of the Spinozan claims, a conceivable hybridized third model (‘Cartozan’) could agree with the Spinozan Hypothesis of temporal precedence while maintaining that acceptance, though prior to rejection, is a discrete operation. The Cartozan model would behave exactly as a Spinozan model in the studies mentioned so far. Additional studies shed light on the likelihood of the unity hypothesis. [114]

                      Initial Representation                       Assessment
Cartesian:     Comprehension                            --->Acceptance /Rejection
:      Comprehension & Acceptance      ---> Rejection
:      Comprehension                            --->Acceptance --->Rejection

An argument for the cognitive ability to ‘reserve judgment’ is predicted by Cartesian and Cartozan models, which propose the possibility of a non-evaluated initial representation state.
(Wegner, Coulton, Wenzlaff, 1985) However, people given information (appraisals of their performance of a task) which they are previously told is invalid (completely arbitrary and not related to their actual performance) show evidence of believing the appraisal, regardless.[114]

(Gilbert, Krull & Malone, 1990): Participants presented with images of smiling faces were told either before or after whether each image portrayed true or false happiness. When later recalling the truth of the happiness of images while under cognitive load, false faces were misidentified as true (true faces were not misidentified as false) whether they had been advised of the truth value before or after exposure. It is assumed that giving the truth value before exposure should cause the subject to exercise the option to delay truth evaluation, if such an option existed, but there was no significant difference between before and after condition data. The Spinozan model predicts this result. [115]

(Gilbert et al., 1990): Participants were asked to read statements about a fictional ‘glark’ as quickly as possible, either evaluating or withholding evaluation of their truth. Though RT times of the non-evaluation task verified the absence of evaluation, when asked later, participants tended to remember non-evaluated statements as true. There is a lack of evidence for the possibility of evaluation-neutral comprehension. [115]

Dealing with Mere Possibility

When there is no way to evaluate a proposition as true or false, Cartesian and Cartozan systems should reserve judgment, while Spinozan systems should fail to label a proposition as false at the assessment, and default to belief.
(Trope & Bassock, 1982): People find a search for information which confirms a hypothesis which they believe to be true more subjectively useful (informative) than a search for neutral (non-confirming) information.
(Swann & Giuliano, 1987): The subjective ‘informativeness’ of a search which confirms a 'merely entertained' (not deliberately evaluated) statement is rated as more highly than a confirmation-neutral search.
(Lord, Lepper & Preston, 1984): If a subject believes an unverified hypothesis to be true, confirmational searches tend to follow. When asked to comprehend both a hypothesis and its opposite, the two conflicting representations yield no single believed hypothesis and no confirmational search follows. [115]


Strong, converging evidence supports the asymmetry and unity hypotheses entailed by the Spinozan belief model. [116]

Though a Spinozan system appears procedurally redundant and prone to error, early output of such a system allows quick active responses, similar to the allowances made by the perceptual system; “As perception construes objects, so cognition construes ideas”. (pg.116)

Automated belief may be an ‘evolutionary outgrowth’ of the older perceptual system. The perceptual system gainfully assumes that most percepts are accurate, the procedural system of belief may assume the same of a social culture where lying is considered immoral. The innovative addition of the subsequent evaluation stage catches costly exceptions. [116]

Generally, people believe easily, but doubt with comparatively great effort and difficulty. Believing appears to be so automatic as to be involuntary, rather than a deliberate application of rationality. [117]

Summary author's notes:

Back to the Cognitive Science Summaries homepage
Cognitive Science Summaries Webmaster:
JimDavies (jim@jimdavies.org)