On the Possibility of Suspending Belief

We present two experiments that cast doubt on existing evidence suggesting that it is impossible to suspend belief in a comprehended proposition. In Experiment 1, we found that interrupting the encoding of a statement’s veracity decreased memory for the statement’s falsity when the false version of the statement was uninformative, but not when the false version was informative. This suggests that statements that are informative when false are not represented as if they were true. In Experiment 2, participants made faster lexical decisions to words implied by preceding statements when they were told that the statements were true than when the veracity of the statements was unknown or when the statements were false. The findings suggest that comprehending a statement may not require believing it, and that it may be possible to suspend belief in comprehended propositions. Is it possible to suspend belief in a comprehended proposition? In attempting to answer this question, Gilbert (1991) distinguished between the philosophies of Descartes and Spinoza. According to Descartes, people first comprehend a proposition and then later assess its veracity. Thus, in Descartes’ view, comprehension precedes belief, and belief suspension is therefore possible. In contrast, according to Spinoza, comprehending a proposition requires believing it to be true. Thus, in his view, comprehension and belief occur simultaneously, and belief suspension is therefore impossible. Although Descartes’ position on this issue is more intuitively appealing, existing evidence seems to favor the Spinozan view. This evidence comes in two forms. First, there is research suggesting that truth is represented more quickly and easily than falsity. People are quicker to assess the veracity of true than false statements (e.g., Just & Carpenter, 1976), and they more easily represent true than false propositions (Johnson-Laird & Savary, 1999). Second, Gilbert and his colleagues have demonstrated that when people are under cognitive load or time pressure, they tend to misremember and misrepresent false information as being true (Gilbert, Krull, & Malone, 1990; Gilbert, Tafarodi, & Malone, 1993). In a study ostensibly about language learning, Gilbert et al. (1990, Experiment 1) presented participants with statements about the meaning of words in the Hopi language (e.g., A tica is a fox). Each statement was presented on a computer screen for 8 s, and then immediately followed by the word true or false to indicate whether the proposition was an accurate or inaccurate statement about the Hopi language. Critically, on some learning trials, participants were asked to respond as quickly as possible to the sound of a tone, which was meant to interrupt the encoding of the proposition’s veracity by depleting participants’ cognitive resources. After this learning phase, participants were presented with some of these propositions and were required to recall whether they were originally presented as true or as false. Results were consistent with the Spinozan hypothesis; interruption decreased the recall accuracy of false but not true statements. Participants’ recall accuracy was 55% for uninterrupted true propositions, 58%for interrupted true propositions, and 55%for uninterrupted false propositions, but only 35% for interrupted false propositions. Gilbert et al. (1990, Experiment 2) replicated this effect with different materials (smiling faces that were either sincere or insincere) and a more difficult interruption task that required participants to judge whether the interrupting tone was high pitched or low pitched. Gilbert (1991) interpreted these demonstrations as evidence for a dual-process model of belief. At Stage 1, propositions are simultaneously comprehended and believed. Subsequently, at Stage 2, people effortfully ‘‘unbelieve’’ false propositions. Cognitive load interrupts the two-stage process before it runs to completion, causing the process to output after Stage 1. Consequently, cognitive load causes false propositions to remain believed, and therefore to be wrongly recalled as true. This dualprocess model has been used to explain a variety of effects, Address correspondence to Uri Hasson, Brain Research Imaging Center, Biological Sciences Division, The University of Chicago, 5841 S. Maryland Ave., MC-2030, Chicago, IL 60637; e-mail: uhasson@ uchicago.edu. PSYCHOLOGICAL SCIENCE 566 Volume 16—Number 7 Copyright r 2005 American Psychological Society including source-monitoring errors (Schacter, Norman, & Koutstaal, 1997), acquiescence effects (Knowles & Condon, 1999), and anchoring effects (Chapman & Johnson, 2002). Although extant evidence is certainly consistent with the Spinozan view, the verdict is still out on the possibility of belief suspension. First, people may encode assertions of truth more easily than assertions of falsity simply because true statements are encountered more often than false statements (Skurnik, 1998), or because false statements tend to be more grammatically complex than true statements (Clark & Chase, 1972). Second, although cognitive load may cause people to misrepresent some false propositions as true, this effect may not generalize to all propositions, because falsity may be represented differently depending on what is said to be false. In line with this argument, previous research has demonstrated that negated statements are more easily represented to the extent to which the negations promote meaningful, informative inferences (Fillenbaum, 1966; Wason, 1965). For example, Fillenbaum (1966) and Mayo, Schul, and Burnstein (2004) found that participants misrecalled negated statements as affirmations less frequently when the negation allowed for a meaningful and informative inference or could be accommodated by a preexisting schema. These findings suggest that the representation of negation may not always constitute a simple ‘‘tag’’ of the affirmation, but that whether or not a tagging system is employed may depend on the informativeness of the negation. Specifically, when the negation of a term does not offer a meaningful, informative inference, then negation might be represented as a tag (Clark & Chase, 1972). However, when negated content provides an available inference, people may not represent negation as a tag of the affirmative but, instead, make the allowable inference. In a study supporting this possibility, Fillenbaum (1966) found that the most frequent memory error for informative negations involved misrecalling a negated adjective as its antonym. Although Fillenbaum’s research focused on negation rather than falsity, it seems likely that negation and falsity are represented similarly (Gilbert, 1991; Just & Carpenter, 1976). Representing a statement as false may require applying a falsity tag to the statement if its false version is uninformative, because there isno other way to represent it. However, whenaproposition’s false version is informative, then the false statement may be represented in terms of what its falsity implies or suggests. Thus, a false statement may be much more easily represented, and less likely to be misrecalled as true, when knowing that it is false provides a perceiver with meaningful information. In their research, Gilbert et al. (1990) used primarily stimuli that were uninformative when false. Learning that the statementa tica is a fox is false does not allow for a meaningful inference, because knowing that a tica is not a fox implies nothing at all; tica could mean an infinite number of things. As we have suggested, when statements are not meaningful when false, participants may have no choice but to represent falsity in terms of the affirmative proposition, perhaps accompanied by an effortful application of a falsity tag. This in turn may lead people to incorrectly recall false statements as true when the encoding of falsity is interrupted. However, in general, many propositions that people encounter are highly informative when false. For example, consider the proposition George owns a television. This is an informative proposition when false, because knowing that it is false supplies you with knowledge about George—it suggests that George is unlike most people. Thus, when learning that such propositions are false, people may be able to represent them in terms of what their falsity suggests (e.g., George is atypical; he is the bookish type), rather than by applying a falsity tag to the affirmative proposition. Evidence from person perception is consistent with the possibility that this process may be completed independently of cognitive load. People are remarkably good at making spontaneous person inferences from minimal information (Todorov & Uleman, 2002). Such inferences persist despite lack of explicit memory for the information that triggered them (Carlston & Skowronski, 1994; Todorov & Uleman, 2002). More important, Todorov and Uleman (2003) have shown that spontaneous person inferences occur under conditions of cognitive load, rapid presentation of information, and shallow processing, suggesting that such inferences are automatic. Perhaps, then, when statements are informative when they are false, people will not misrecall them as true, even under time pressure or cognitive load. Such a finding would cast doubt on the claim that comprehending a statement requires believing it, and that belief suspension is impossible.