“I am not guilty” vs “I am innocent”: Successful negation may depend on the schema used for its encoding

https://doi.org/10.1016/j.jesp.2003.07.008Get rights and content

Abstract

Negations (e.g., “Jim is not guilty”) are part of our daily language and communication. Linguistic and non-linguistic negations can occur when receivers counter-argue what communicators are saying, when hypotheses are disconfirmed, or through negative cognitive responses and many other social interactive processes. Our study explores how negations are encoded by considering the predictions of two theoretical models. According to the fusion model, the core of a negated message and the negation marker are integrated into one meaningful unit. Thus, Jim in the example might be encoded within the schema “innocence.” According to the schema-plus-tag model, a negated message is represented as a core supposition and a negation tag, allowing for dissociation of the two at a later point in time. We compare the two models by examining the nature of inferences that are facilitated by negations. Our results show that the existence of a schema that accommodates the meaning of the original negation is critical in determining how a negation will be encoded. When such a schema is not readily available, processing a negated message facilitates negation–incongruent associations, in line with predictions of the schema-plus-tag model. This model is also supported by analyses of respondents’ memory. We discuss implications of these findings for the communication of negated information, for discounting theories, and for the assessment of the truth of incoming information.

Introduction

“John is simply not a romantic person,” Mary muttered. Before you continue, please try to think of at least two examples of John’s behaviors that may have led Mary to this conclusion. Were your examples similar to “He forgets her birthday/wedding anniversary,” “He never brings her flowers,” “He does not express his love in poetry/love letters/surprises/gifts”? All of these behaviors are actually negated romantic behaviors. In other words, stating that John is not romantic can make one think of romantic behaviors and then negate them. Is this a general effect? What are the boundary conditions? What are its implications?

There is no doubt that negation is part of our daily language and communication. It also comes as no surprise that readers have no difficulty understanding the previous sentence, which contains a negation, as well as this one (which contains two negations). In spite of the greater complexity attributed to the comprehension of negations (Carpenter & Just, 1975; Clark & Chase, 1972; Wason, 1963), people usually succeed in understanding the intended meaning of negations. Yet, we believe that a message which is phrased as a negation can under certain conditions activate associations that are incongruent with the message meaning, and might thus introduce communication errors that may actually lead to inferences opposite to the message’s intended meaning. This paper explores whether and how this can happen.

The presence and function of negation in social interactions have puzzled philosophers, linguists, and psychologists throughout the centuries (Jordan, 1998). This paper examines whether there are processing and coding differences between messages phrased as negations (“Jim is not guilty”) and those phrased as affirmations (“Jim is innocent”). These might be central for understanding what happens in situations that induce receivers to use negation. For example, negations might be generated as one counter-argues against an untrustworthy source of communication while listening to it or reading it (Eagly, Wood, & Chaiken, 1978; Papageorgis, 1968; Petty & Cacioppo, 1977; Zuwerink & Devine, 2000). More generally, receivers are often informed that some knowledge they have is ill-founded, so that after having learned “X” in the past, they should now process “not-X” (Fiedler, Walther, Armbruster, Fay, & Naumann, 1996; Hornby, 1974; Johnson, 1988; Johnson, Hashtroudi, & Lindsay, 1993; Johnson & Raye, 1981; Koehler, 1991; Loftus, 1979; Loftus & Palmer, 1974; Ross, Lepper, & Hubbard, 1975; Schul, 1993; Schul & Burnstein, 1985; Schul & Manzury, 1990; Schul & Mayo, 1999; Sellars, 1954; Strack & Bless, 1994). Obviously, these situations involve diverse sets of causal factors that have an impact on how people process the given information. To study the role of negation, the current study simplifies the situations considerably by examining the kind of inferences people access at the time they process affirmations and negations that are given to them explicitly. In this endeavor we took advantage of our participants’ native language. In Hebrew, a language without any semantic prefixes or suffixes, a negated concept is simply a concept with the addition of the word “no” preceding it. Hence it provides a clear state of literal negation, with little leeway in interpretation. Therefore, we believe that any effect found for this simple form of negation may explain in part why some messages are easier to negate than others, in what cases discounting succeeds or fails, and why some beliefs are easier to change than others.

Let us start by considering what comes to mind spontaneously when one reads the statement, “Michael says that ever since he came to this country, he has not known peace of mind.” Does one think spontaneously about anxiety, or about serenity? Consider now our initial example “John is simply not a romantic person” and recall if you thought about romantic or unromantic gestures. Put more generally, does the processing of messages expressed as negations facilitate associations that are congruent with the intended meaning of the negation (e.g., unromantic gestures), or associations congruent with what is being negated (e.g., romantic gestures). In the latter case, the ensemble of spontaneously activated associations might leave residual meanings that can lead receivers astray when they are trying to think about the original message. The associative structures activated in processing negations can be explored in the light of two theoretical models, which will be discussed in the following sections.

Clark and Chase (1972) suggested that encoding negations requires more time because it entails processing an additional operator, the negation operator, which is separated or dissociated from the message’s core concepts. For example, they assume that the negation “B is not above A” is represented by the propositions (Not (B above A)). Accordingly, to process the negated message one must first process the core supposition (i.e., “B above A”) and then negate it. The critical point for our discussion is that the core supposition is processed as a cognitive unit, which is then marked with a negation tag. For this reason we term this the schema-plus-tag model. Under this model the core (“B above A”) could be dissociated from the negation tag at a later time. There are several lines of evidence consistent with this model:

Just and Carpenter (1976) tracked respondents’ eye movements during a verification task. They showed that respondents tend to focus on the core supposition more than on its negation. For example, when participants were presented with a probe saying “is not north,” referring to the location of a plus sign, their eye movements were focused on the north position rather than the south one. While this might be consistent with the logical way to test the truth value of negated information, it nevertheless implies that the focus of attention is on the information that is being negated.

Fiedler et al. (1996) asked their participants a series of questions about the interior of an apartment they had seen on a videotape. Some of the questions referred to objects that were not present in the apartment. Fiedler et al. reported that while participants correctly denied seeing the absent objects when they were questioned immediately afterwards, they mistakenly recognized these objects as having appeared in the video when asked to recognize them after a 20-min distraction task. Importantly, the extent of their mistaken memory was significantly higher than the rate of false alarms for other objects that had not appeared in the video and were not mentioned in the questions. Thus, it seems that thinking about and providing a negative response (e.g., “there was no hat rack in the apartment”) led to a false memory (e.g., there was a hat rack in the apartment). Fiedler et al. (1996) attributed the memory intrusions of absent objects to constructive memory, suggesting that people encode negated messages by adding a tag denoting negation to the core of the message. Finally, somewhat indirect support for the schema-plus-tag model emerges from studies showing that, when negation is used to deny a plausible misconception, comprehension is facilitated by prior consideration of the preconception (Johnson-Laird & Tridgell, 1972; Wason, 1963).

The schema-plus-tag model has two related consequences. First, the original negated message is assumed to be represented as a core supposition and a negation tag, allowing for dissociation between the two at a later point in time. Second, the consideration of the core supposition activates the associations that are congruent with the core, but incongruent with the intended meaning of the negation as a whole. For example, upon comprehending the message “Tom is not guilty,” one first thinks about the proposition “Tom is guilty” and activates associations that are congruent with the schema of “guilt.” Only then does the receiver attach the negation marker. This allows for the correct understanding of the intended meaning of the negation in spite of the activation of the negation–incongruent associations.

There is an alternative way to model the processing of negations. One may fuse the negation operator with the core, thus transforming the negation into an affirmation. For example, the message “Tom is not guilty” might be transformed into its affirmative counterpart and coded as “Tom is innocent.” In fusing, one spontaneously activates associations that are congruent with “innocence” (and “not guilty”), thus reinforcing the intended meaning of the message. The feasibility of changing negations into affirmations presupposes that there is an affirmation that captures the meaning of the negation (e.g., “innocent” means “not guilty”). This premise is offered by the literalism account of negation (Horn, 1989; Lyons, 1995).

The fusion model is consistent with a suggestion made by Gannon and Ostrom (1996), who argued that in processing uni-polar rating scales, the scale-point labeled “completely not X” is associated with a category which is the opposite of X. The study of Brewer and Lichtenstein (1975) also lends some support for this model, by showing that receivers recalled negations as affirmative assertions which preserve the original meaning (e.g., recalling “X is not warm” as “X is cold”).

The research of MacDonald and Just (1989) provides a mechanism for the fusion model by suggesting that a negation operator inhibits the activation level of concepts that appear in the core supposition (see also Lea & Mulligan, 2002). According to the findings of MacDonald and Just, for example, processing the negation “Tom is not guilty” should result in inhibiting the concept “guilty.” Functionally, this implies that receivers of such a negation are less likely to think of guilty-congruent concepts and, speculatively, more likely to think of guilty-incongruent concepts.

A basic distinction between the schema-plus-tag and the fusion model involves the nature of schema activated as one encodes a negation. The schema-plus-tag model assumes that negations such as “Tom is not guilty” are processed within the schema that refers to the core supposition (e.g., “guilt”). In contrast, the fusion model assumes that a negation–congruent schema (e.g., “innocence”) is activated. This distinction is highly significant because not only do the two schemas have opposite meanings, they are also embedded in markedly different associative networks. As Gannon and Ostrom (1996) asserted “The category dishonest is not merely the inverse or negative of the category honest, it is a different knowledge structure. That is, honesty and dishonesty do not differ only in terms of degree or positivity: The two categories are conceptually distinct.” (p. 338). Hence, the schema in which we process and code the negated message may be crucial, as in impression formation (Asch, 1946) or constructing measurement scales (Dholakia & Morwitz, 2002).

The differences between the two models might not be evident if we probe for the meaning of the message using direct questions. As we noted earlier, under normal conditions individuals have no trouble understanding the intended meaning of a negation. The models might be distinguishable if we probe the pattern of activation indirectly. To illustrate, imagine that shortly after encoding the negation, “Tom is not lazy,” one has to process one of the two probes, SLOW or QUICK. Assume that “quick” is message-congruent (it is associated with being not-lazy) while “slow” is message-incongruent (it is associated with being lazy). Our study examines the pattern of activation when processing negation by comparing the activation of the message-congruent (“quick”) and the message-incongruent (“slow”) probe. According to the fusion model, the description “Tom is not lazy” is interpreted within a negation-congruent schema (e.g., “industriousness”). Therefore, the “quick” probe should be activated more than the “slow” probe. In contrast, according to the schema-plus-tag model the description is interpreted within a negation-incongruent schema (e.g., “laziness”). Therefore, the “slow” probe should be activated more than the “quick” probe.

Of course the two models also differ in their implications for the long-term impact of negations. According to the schema-plus-tag model, the negation operator could be detached from the core supposition. As a result, individuals may remember the opposite of the intended meaning. Such state of affairs is unlikely under the fusion model, as the negation is interpreted within a congruent schema. Predictions about memory for negations are tested in Experiment 2. After describing the results of two experiments, we elaborate on other implications of the two models for the communication of negated information, for discounting theories and for the truth value assigned to incoming information.

Section snippets

Experiment 1

In each trial of Experiment 1 participants were presented with a description of a person, expressed as an affirmation or a negation (e.g., “Tom [is/is not] a tidy person”), followed by a probe sentence (e.g., “Tom’s clothes are folded neatly in his closet”). Participants were to determine whether the probe fits the description of the person. The question of interest is whether processing of congruent probes is facilitated relative to processing of incongruent probes. Our main interest concerns

Experiment 2

The design and procedure of Experiment 2 are similar to those of Experiment 1, with several important differences. First, the descriptions used in Experiment 1 were not chosen systematically on the basis of the uni-polar/ bi-polar distinction. Therefore, some of them were uni-polar, some bi-polar, and the others neither clearly uni-polar nor clearly bi-polar. In Experiment 2 we selected descriptions that were clearly either uni-polar or bi-polar (see below). Second, whereas participants in

Summary

Our study explored the consequences of processing negations. Experiment 1 showed that descriptions phrased as affirmations facilitated processing of behavioral probes that were congruent with their meaning as compared to probes that were incongruent. In contrast, descriptions phrased as negations facilitated processing of incongruent probes as compared to congruent ones. Experiment 2 demonstrated that this effect was limited to uni-polar descriptions. When negations had a well-defined and

References (53)

  • F. Strack et al.

    Memory for nonoccurrences: Metacognitive and presuppositional strategies

    Journal of Memory and Language

    (1994)
  • C.A. Anderson et al.

    Perseverance of social theories: The role of explanation in the persistence of discredited information

    Journal of Personality and Social Psychology

    (1980)
  • C.A. Anderson et al.

    Thinking about people: Contributions of typological alternative to associationistic and dimensional models of person perception

    Journal of Personality and Social Psychology

    (1991)
  • S.E. Asch

    Forming impressions of personality

    Journal of Abnormal and Social Psychology

    (1946)
  • I.M. Begg et al.

    Dissociation of processes in belief: Source recollection, statement familiarity, and the illusion of truth

    Journal of Experimental Psychology General

    (1992)
  • W.F. Brewer et al.

    Recall of logical and pragmatic implications in sentences with dichotomous and continues antonyms

    Memory & Cognition

    (1975)
  • P.A. Carpenter et al.

    Sentence comprehension: A psycholinguistic processing model of verification

    Psychological Review

    (1975)
  • H.L. Colston

    “Not good” is bad, but not bad is not good: An analysis of three accounts of negation asymmetry

    Discourse Processes

    (1999)
  • T.D. Cook et al.

    History of the sleeper effect: Some logical pitfalls in accepting the null hypothesis

    Psychological Bulletin

    (1979)
  • L. Cosmides et al.

    Cognitive adaptation for social exchange

  • U.M. Dholakia et al.

    The scope and persistence of mere-measurement effects: Evidence from a field study of customer satisfaction measurement

    Journal of Consumer Research

    (2002)
  • A.H. Eagly et al.

    Causal inferences about communicators and their effect on opinion change

    JPSP

    (1978)
  • S. Fein et al.

    Can the jury disregard that information. The use of suspicion to reduce the prejudicial effects of pretrial and inadmissible testimony

    Personality and Social Psychology Bulletin

    (1997)
  • I. Fleming et al.

    Information processing and the perseverance of discredited self-perceptions

    Personality and Social Psychology Bulletin

    (1979)
  • R.W. Gibbs

    On the psycholinguistics of sarcasm

    Journal of Experimental Psychology: General

    (1986)
  • D.T. Gilbert

    How mental systems think

    American Psychologist

    (1991)
  • Cited by (0)

    The research reported here was supported by grants from the US–Israel Binational Science Foundation (BSF) and the Israeli Foundation Trustees. We would like to thank Liran Rasinski, Yasmin Folder, and Orly Carmi for their help in conducting the experiments and Rachel Giora for comments on an earlier version of this paper.

    View full text