Ambivalence and Attitudes - OSF

Full Terms & Conditions of access and use can be found athttps://www.tandfonline.com/action/journalInformation?journalCode=pers20

European Review of Social Psychology

ISSN: 1046-3283 (Print) 1479-277X (Online) Journal homepage: https://www.tandfonline.com/loi/pers20

Ambivalence and Attitudes

Mark Conner & Paul Sparks

To cite this article: Mark Conner & Paul Sparks (2002) Ambivalence and Attitudes, EuropeanReview of Social Psychology, 12:1, 37-70, DOI: 10.1080/14792772143000012

To link to this article: https://doi.org/10.1080/14792772143000012

Published online: 15 Apr 2011.

Submit your article to this journal

Article views: 1168

View related articles

Citing articles: 117 View citing articles

https://www.tandfonline.com/action/journalInformation?journalCode=pers20

https://www.tandfonline.com/loi/pers20

https://www.tandfonline.com/action/showCitFormats?doi=10.1080/14792772143000012

https://doi.org/10.1080/14792772143000012

https://www.tandfonline.com/action/authorSubmission?journalCode=pers20&show=instructions

https://www.tandfonline.com/action/authorSubmission?journalCode=pers20&show=instructions

https://www.tandfonline.com/doi/mlt/10.1080/14792772143000012

https://www.tandfonline.com/doi/mlt/10.1080/14792772143000012

https://www.tandfonline.com/doi/citedby/10.1080/14792772143000012#tabModule

https://www.tandfonline.com/doi/citedby/10.1080/14792772143000012#tabModule

Chapter 2

Ambivalence and Attitudes Mark Comer

Paul Sparks University of Leeds, UK

University of Sussex, UK

ABSTRACT

This chapter explores the concept of ambivalence and its relationship to attitudes. Definitions and different measures of ambivalence are reviewed. We present three dimensions on which measures of ambivalence can be classified. We then go on to examine some of the elicitors of ambivalence and present some preliminary findings on attitude objects that generate high levels of ambivalence. The major section of the chapter then reviews findings related to treating the ambivalence construct as an aspect of attitude strength. Research is presented in relation to four major predictions: (a) temporal stability of attitudes; (b) impact of attitudes on intentions and behavior, (c) pliability of attitudes; and (d) impacts on information processing. The final section offers suggestions for further research in this area.

Attitude has retained its position in social psychology as the most distinctive and indispensable concept for well over 50 years (Allport, 1935). Interest remains high, with significant advances in our understanding of the antecedents and consequences of attitudes (see Eagly & Chaiken, 1993, for an extensive review). A particular focus of research interest in recent years has been on aspects of the structure (e.g., Pratkanis, Breckler, & Greenwald, 1989) and strength (e.g., Petty & Krosnick, 1995) of attitudes. The recent

Correspondence regarding this chapter should be addressed to: Mark Comer, School of Psychol- ogy, University of Leeds, b e d s LS2 9JT, UK. E-mail: [email protected]

European Review of Social Psychology, Volume 12. Edited by Wolfgang Stroebe and Miles Hewstone. Q 2002 John Wiley & Sons Ltd.

38 MARK CONNER AND PAUL SPARKS

interest in attitudinal ambivalence can be partly attributed to the resurgence of interest in the dimensions of attitude structure and strength within which ambivalence fits. Attitudinal ambivalence may be defined as a psychological state in which ". . . a person holds mixed feelings (positive and negative) towards some psychological object" (Gardner, 1987, p. 241). Ambivalence thus is at odds with traditional views of the attitude concept as a unidimensio- nal evaluation (Eagly & Chaiken, 1993). This chapter explores the ambivalence concept through a review of research on its definition, measurement, antecedents, and consequences.

The ambivalence concept was introduced into social psychology by Scott (1966,1968,1969), followed by other important contributions by Kaplan (1972), and Katz and Hass (1988). Nevertheless, the resurgence of interest in the concept has occurred principally in the 1990s (see Jonas, Bromer, & Diehl, 2000; Priester & Petty, 1996; Thompson, Zanna, & Griffin, 1995, for reviews of earlier research with the ambivalence concept). Research with the ambivalence concept since 1990 has principally focused on predictions for the general concept of ambivalence, although there has also been interest in ambivalence in specific domains, such as racist attitudes (e.g., Katz & Hass, 1988), sexist attitudes (e.g., Glick & Fiske, 1996), and consumer attitudes (Otnes, Lowrey, & Shrum, 1997) (see Jonas er al., 2000, for a useful review of this research). The present chapter will focus on the general concept of ambivalence.

Jonas et al. (2000) provided an earlier review of the literature on ambivalence. That chapter provided an excellent review of the origins and history of the ambivalence concept and of the relative utility of different formulae employed to tap measures of ambivalence. Neither of these issues are the focus of attention of the present chapter. Both chapters consider the implications of treating ambivalence as a measure of attitude strength. Jonas er al. focus on distinguishing ambivalence from attitude strength and give relatively brief attention to empirical data supporting the presumed implications of treating ambivalence as a measure of attitude strength. In contrast, the focus in the present chapter is reversed. The first section of the present chapter explores various definitions of the ambivalence concept that have been offered. The differing definitions suggest a number of different forms of ambivalence. In considering the variously used measures of ambivalence, we present three dimensions along which different measures of ambivalence can be classified. We then present some preliminary data on the interrelationship of these different measures of ambivalence and show that the measures possess discriminant validity, but suggest that further work on this issue is required. We then go on to explore some of the elicitors of ambivalence. After presenting some preliminary findings on the types of attitude objects that generate high levels of ambivalence, we examine the relationship between various individual difference variables and ambivalence.

The largest section of the present chapter then explores the consequences of holding ambivalent attitudes. Based on treating ambivalence as a dimen-

AMBIVALENCE AND AlTITUDES 39

sion of attitude strength, we examine the evidence to support four predicted consequences of ambivalence-its impacts on: (a) the temporal stability of attitudes; (b) the relationship of attitudes to intentions and behavior; (c) the pliability of attitudes; and (d) the processing of information (Krosnick & Petty, 1995). The final section of our chapter offers a number of suggestions for further research with the concept of attitudinal ambivalence.

DEFINITIONS OF AMBIVALENCE

Ambivalence has been defined in various ways. Given our belief in the importance of ambivalence to attitude research, we should spend some time on an assessment of different approaches to the concept. Most definitions in the psychological literature make reference to the simultaneous existence of positive and negative evaluations of an attitude object (see Table 2.1 for a selec- tion). However, there are important differences between the various definitions of the concept that merit some attention.

Table 2.1 Internretations and definitions of ambivalence

“The existence of simultaneous or rapidly interchangeable positive and negative feelings toward the same object or activity, with the added proviso that both the positive and negative feelings be strong” (Meehl, 1964, quoted in Emmons, 19%) “A psychological state in which a person holds mixed feelings (positive and negative) towards some psychological object” (Gardner, 1987, p. 241) “The extent of beliefs’ evaluative dissimilarity (or inconsistency)” (Eagly & Chaiken, 1993, p. 123) “A conflict aroused by competing evaluative predispositions” (Breckler, 1994, p. 359) “. . . the extent to which one’s reactions to an attitude object are evaluatively mixed in that both positive (favorable) and negative (unfavorable) elements are included” (Wegener, Downing, Krosnick, & Petty, 1995, p. 460)

An inclination “. . . to give it [an attitude object] equivalently strong positive or negative evaluations” (Thompson er al., 1995, p. 367). “An approach-avoidance conflict-wanting but at the same time not wanting the same goal object” (Emmons, 19%, p. 326)

First, the idea that positive and negative evaluations need to be simultaneous is not a universal feature of orientations to the ambivalence concept. Ainslie (1986), for example, with his focus on temporal dimensions of choice, suggests that long-term interests (e.g., health) and short-term interests (e.g., hedonic pleasure) are not invariably evaluated in conjunction with one another to form a consistent choice because each may be “dominant” at

40 MARK CONNER A N D PAUL SPARKS

different times. From Ainslie’s decision-making perspective on competing choice options, ambivalence is defined as an “inability to make a stable choice between two alternatives despite familiarity with their consequences” (Ainslie, 1992, p. 370). Moreover, a distinction has been made between diu- chronic ambivalence and synchronic ambivalence (see Stocker, 1990) to reflect the distinction between “inconstancy” and “inconsistency” in evaluations. Like diachronic ambivalence, the notion of “attitude variability’’ (e.g., Sparks, Hedderley, & Shepherd, 1992) focuses on vacillating attitudes over time, whereas the simultaneous experience of positive and negative evaluations is central to the (synchronic) ambivalence, which is of predominant current interest to social psychologists researching the relationship between attitudes and ambivalence. However, lay people’s responses to, for example, direct measures of ambivalence in research work (see below) will of course reflect their own interpretations of the questions put to them.

Second, attitude objects may be, for example, behaviors, targets, events, or states of affairs. People may be ambivalent towards behaviors (e.g., reducing fat consumption), towards goals (e.g., losing weight), towards events (e.g., Millennium celebrations) or states of affairs [e.g., the existence of 70 mph (112 kph) motorway speed limits]. They may feel ambivalent about a particular behavior because the behavior has both negative and positive attributes, but there may also be a fundamental ambivalence about any one attribute. In this chapter, we are principally concerned with ambivalence towards behaviors, since most contemporary attitude research also displays this orientation and because we believe that closely specified personal behaviors are an appropriate focus for the study of motivational conflicts.

Third, some approaches to ambivalence focus more on people’s beliefs, while others more directly on their attitudes. Given that the relationship between beliefs and attitudes may be complex, we have a general preference for the more direct focus on people’s attitudes (although exploratory work on both is of great interest), since these are central to the evaluative issues at stake in the notion of ambivalence. While outcome evaluations may be assessed in conjunction with behavioral beliefs within the Theory of Reasoned Action (TRA; Fishbein & Ajzen, 1975) and the Theory of Planned Behavior (TPB; Ajzen, 1991), it is the specific evaluation of behaviors on which we are concentrating here for our assessment of ambivalence.

Fourth, it is likely to be fruitful to examine other approaches to ambivalence, given that this concept attracts attention across different disci- plines. For example, ambivalence is implicated in political science perspectives on people’s conflicting values (Zaller & Feldman, 1992), sociological work on conflicting role demands (Merton, 1976), and psychological approaches to both the mixed feelings that people may have about the objects of their attitudes (Tourangeau, Rasinski, Bradburn, & D’Andrade, 1989) and to conflicting parts of their personality (Freud, 1962). It is expected that

AMBIVALENCE AND ATTITUDES 41

insights from these approaches may well serve to develop and expand social psychologists’ interests in attitudes and ambivalence.

Most attempts to amve at a method for quantifying ambivalence in recent social psychological research have acknowledged the need to examine both (a) the “intensity” of people’s feelings (i.e. both how positive and negative those feelings may be) about an attitude object and (b) the “similarity” of the intensity of positive and negative feelings. This way of thinking about attitudes involves a departure from some traditional conceptualizations, which tend to be characterized as referring to attitudes as reflecting either positive or negative evaluations of an attitude object (see Allport, 1935, for an early discussion of these issues). As well as having implications for the theoretical status of the attitude concept, there are also a number of possible empirical consequences that follow from this way of interpreting attitudes: these form the focus of the current chapter.

MEASUREMENT OF AMBIVALENCE

There is currently a lack of consensus about the best way to measure ambivalence (Breckler, 1994; Conner, Sparks, Povey, James, & Shepherd, 1999~; Eagly & Chaiken, 1998; Jonas el al., 2000; Priester & Petty, 1996; Thompson et al., 1995). Previous research into ambivalence has operationalized ambivalence in different ways. For example, whilst Lavine, Thomsen, Zanna, and Borgida (1998) examined affective-cognitive consistency, Bassili (1996) measured global ambivalence, and Sparks et al. (1992) investigated attitude variability. The extent to which these are interchangeable is unclear. However, there is good reason to expect differences between affectivexognitive, evaluative-cognitive, and evaluative-affective inconsistency (e.g., Chaiken, Pomerantz, & Giner-Sorolla, 1995). In order to ensure that the present discussion of ambivalence measurement is internally coherent, the present chapter focuses on global measures of ambivalence (see Thompson er al., 1995). Even amongst global measures of ambivalence there may be a number of forms that possess discriminant validity. The different measures that have been employed to tap ambivalence can be classified along three dimensions (see Table

The first dimension is in terms of direct versus indirect measurement. Bassili (1996) usefully divides measures of attitude strength into meta-attitudinal and operative indexes (see also Greenwald & Banaji, 1995). Meta-attitudinal measures of attitude strength are based upon the attitude holder’s subjective impression of hidher own attitude. In relation to attitudinal ambivalence, such measures might tap the subjective impression of the extent to which the individual’s attitude was “mixed” or consisted of both positive and negative evaluations, beliefs or feelings. We refer to such measures as direct measures

2.2).


Table 2.2 Classification of ambivalence measures

Overall evaluation measure]

Direct measures Indirect measures

Affective conflicts

Cognitive conflicts I have conflicting thoughts My thoughts about X are . . . about X (extremely positive-not at all (strongly agree-strongly positive; extremely negative- disagree) not at all negative)

I have mixed feelings about X (strongly agree-strongly disagree) pleasant; extremely unpleasant-

My feelings about X are . . . (extremely pleasant-not at all

not at all unpleasant)

Cognitivelaffective conflicts

My thoughts and feelings about X are conflicting (strongly agree-strongly positive) disagree)

My thoughts about X are . . .* (extremely positive-not at all

My feelings about X are . . . (extremely unpleasant-not at all unpleasant)

I Note that belief-based measures all fall into the indirect category in this scheme but can be used to tap cognitive, affective, or cognitive-affective conflicts, based on a classification of the beliefs. Belief-based measures can also be either closedended (e.g., Van Harreveld et al., 1999) or open- ended (e.g.. Conner ef al., 1999c), Closed-ended measures present sets of beliefs from which the respondent selects. Open-ended measures require the respondent to self-generate beliefs.

This represents a measure of pro-cognitivelcon-affective cognitivelaffective ambivalence (see Thompson et al., 1995). An alternative con-cognitivelpro-affective cognitivelaffective ambivalence might use items such as, “My thoughts about X are . . . extremely positive-not at all positive” and “My feelings about X are . . . extremely unpleasant-not at all unpleasant”.

(Jonas et al., 2000, use the term “experienced ambivalence”) because they require respondents to report directly the extent to which they experience mixed thoughts or feelings (see Tourangeau et al., 1989, for one simple example of the use of such a measure). For example, Lipkus, Green, Feaganes, and Sedikides (in press) report a smoking-specific measure of “felt” ambivalence. This was a six item self-report measure (e.g., “You have strong feelings both for and against smoking”, strongly agree-strongly disagree). Ambivalence is reported as the mean or summed score across a number of related statements. One advantage of such measures is the possibility of com- puting internal reliability measures (e.g., Cronbach’s alpha). Such measures are often regarded as the “gold standards” against which to compare other measures of ambivalence (see Jonas et al., 2000; Thompson et al., 1995). However, it has been suggested that, in general, meta-attitudinal measures seem to be more open to extraneous influences that can undermine their validity (Bassili, 1996). For example, it is not clear whether the information required for an accurate judgment is readily available to consciousness and so easily retrievable from memory (cf. Wilson, Lindsey, & Schooler, 2000, on explicit and implicit attitudes).


Indirect or “formula-based’’ (Jonas et al., 2000) measures of ambivalence consist of separate measures of the positive and negative thoughts or feelings that an attitude object produces. Such measures possess some similarity to Bassili’s (1996) operative indexes of attitude strength. Operative indexes of attitude strength are derived from the attitude judgment process or its outcomes and have an advantage over meta-attitudinal measures because they stem from the same cognitive response and information processing respons- ible for judgmental and behavioral manifestations of the attitude (Bassili, 1996, p. 638). Operative indices appear to provide a less reactive means of assessing the properties of attitude strength and might be preferred to direct (or meta-attitudinal) measures of ambivalence because they are less reactive to extraneous influences (Conner et al., 1999~). Kaplan (1972) was the first to suggest such a measure of ambivalence. He modified a widely used measure of attitudes (the semantic-differential technique) by separately assessing the positive and negative attributes ascribed to an attitude object; one measure assessed the positive attributes and another assessed the negative attributes (usually with the instruction to ignore the oppositely valenced attributes when making the judgment). Generally, the two measures have been found to show low negative correlations with one another. For example, Sparks, Conner, James, Shepherd, and Povey (2001) employed such split semantic differential measures to tap ambivalence towards eating chocolate (e.g., “For a moment consider only the positive things about eating chocolate and ignore any negative things about it. Please rate how positive those positive things are . . .”, not at all positive-extremely positive; “For a moment consider only the negative things about eating chocolate and ignore any positive things about it. Please rate how negative those negative things are . . .”, not at all negative- extremely negative). These separate judgments are then combined within a formula into an overall measure of ambivalence (see below). If several positive and negative judgment items are used, it is possible to compute reliability coefficients for the sets of positive and negative judgments (Conner et al., 1999~).

The second dimension along which ambivalence measures can be classified is whether the measures tap overall evaluations or whether they tap beliefs. Overall evaluation measures are exemplified by the split semantic differential measure of Kaplan (1972): the attitude object is simply rated on one or more evaluations designed to tap either positive or negative reactions. Belief-based measures of ambivalence tend to focus on the nature of object-relevant information and take two main forms: either closed-ended measures or open- ended measures. Closed-ended belief-based measures of ambivalence require respondents to select (or evaluate) a series of positive and negative beliefs about an attitude object. For example, Van Harreveld, Van der Pligt, and De Vries (1999) required participants to select the important attributes they associated with an attitude object from a list. Ambivalence is then computed using


a formula based on the number of positive and negative attributes selected as being important (see Van der Pligt, De Vries, Manstead & Van Harreveld, 2000, for a review of this research). Open-ended belief-based measures employ a similar approach, but require respondents to self-generate their beliefs in response to a prompt (Bell, Esses, & Maio, 1996). For example, Conner er al. (1999c), in a study of ambivalence towards eating a low-fat diet, used the prompt, “The thoughts and feelings I associate with eating a low fat diet are . . .” to elicit such beliefs. Participants can also be requested to rate each belief in terms of how positive or negative it is (alternatively, independent judges can rate each belief as positive or negative). The numbers of positive and negative beliefs or the sum of the ratings of the positive and negative beliefs can then be entered into an equation to compute a measure of ambivalence (see Erber, Hodges, & Wilson, 1995 for a similar “belief homoge- neity” measure).

The third dimension along which measures of ambivalence can be classified is in terms of the content of ambivalence judgments (focusing on cognitive conflicts, affective conflicts, cognitive-affective conflicts). So, for example, a direct ambivalence measure can refer to mixed cognitions or thoughts (e.g., “I have conflicting thoughts about X”, strongly agree-strongly disagree), mixed affect or feelings [e.g., “I have mixed (both positive and negative) feelings about X”, strongly agree-strongly disagree) or a conflict between thoughts and feelings (e.g., “I have positive thoughts but negative feelings about X”, strongly agree-strongly disagree). As Thompson et al. (1995) make clear, this latter form of ambivalence measure (i.e., the heart vs. the mind) can take two distinct forms: pro-cognition and con-affect versus con-cognition and pro- affect. For example, an individual may like the taste of one particular food, but dislike the high number of calories that the food contains (pro-affect, con- cognition) or dislike the taste of another food, but like the low number of calories the food contains (con-affect, pro-cognition) (Grogan, Bell, & Con- ner, 1997). These different contents of ambivalence measures may be dif- ferentially important in different areas (see also MacDonald & Zanna, 1998). For example, while thinking about an event such as the introduction of a new social policy may arouse principally cognitive ambivalence, thinking about a particular politician (an object) may arouse principally affective ambivalence. Similarly, thinking about an action or behavior (e.g., using a condom with a new sexual partner) may arouse conflicting thoughts and feelings. Ortony, Clore, and Collins (1988) similarly propose three aspects of the world that we can appraise emotionally: events, actions, and objects. However, they further suggest that, within these categories, it is only appraisals of events and actions that are likely to produce conflicting emotions (and ambivalence) because these appraisals can focus on consequences for self and other or on the immediate future and the more distant future (pp. 51-52). In this scheme, objects only produce positive or negative reactions for the self in the immedi-


ate future. However, appraisal of objects and eventdactions could produce between-category ambivalence. For example, a disliked person (object) performing a kindness (action) might produce such conflicting emotions.

The best way to combine separately obtained positive and negative judgments of an attitude object into an overall measure of ambivalence has received attention in a number of studies (Breckler, 1994; Jonas et al., 2000; Priester & Petty, 1996; Thompson et al., 1995). Kaplan (1972) suggested that in a measure of ambivalence, scores should increase as the positive and negative judgments become more polarized (both larger in value) and the more equal they are in absolute value. A variety of measures of ambivalence have been proposed to capture these conditions in a measure of ambivalence (see Jonas et al., 2000, for a recent review). Perhaps the most widely employed such equation is that devised by Griffin and presented in Thompson et al. (1995): ambivalence is calculated as half the polarization of the positive (P) and negative (N) judgments, minus the absolute difference between the two. That is, ambivalence is denoted by the equation:

Ambivalence = (P +N)/2 - IP - IVl

where P and N are measured on unipolar scales (in absolute values) in two separate questions. So, for example, an individual who rates an attitude object as both moderately positive (2) and moderately negative (2) would be assigned an ambivalence score of [(2 + 2)/2 - 12 - 21 = 21, while an individual who rates the same attitude object as extremely positive (3) and only mildly negative (1) would be assigned an ambivalence score of [(3 + 1)/2 - 13 - 1 1 = 01. This equation is the one employed in all the analyses reported in the present chapter where separate positive and negative ratings are combined into a single index of overall ambivalence.

Table 2.2 presents the different dimensions of ambivalence and provides examples of items used to measure ambivalence. In general, very few studies have reported comparison amongst these different measures of ambivalence. Thompson et al. (1995) report a significant correlation of 0.40 between a direct measure of ambivalence and a multi-item split semantic differential measure (with ambivalence calculated using the Griffin formula; n = 146). Lipkus, et al. (in press) reported a significant correlation of just 0.19 between the same measures in a study of smoking (n = 157). In our own research we have compared levels of ambivalence based on open-ended belief-based measures of ambivalence and split semantic differential-based measures of ambivalence (Conner et al., 1999c) and employed the Griffin equation to compute a measure of overall ambivalence. For both behaviors examined (eating a low fat diet, n = 140, eating five portions of fruit and vegetables per day, n = 142), the degree of correspondence was only modest ( r = 0.18 and r = 0.27 respectively, both significant). Hence, across the studies that have reported the relevant


data, there appears to be evidence that the different measures of ambivalence do not show strong convergent validity.

We believe that an important area for further research with the ambivalence construct is a systematic examination of the convergent and discriminant validity of the various ambivalence measures employed across studies (Table 2.2). Without such data it is difficult to offer good advice about the most appropriate means to measure ambivalence. Each of the different measures has potential problems. For example, we have reservations about direct measures of ambivalence because it is not clear to what extent individuals have access to the relevant information on which to base such judgments. In addition, as noted earlier, there may be individual differences in the way ambivalence is interpreted, posing further problems for the potential useful- ness of direct measures of ambivalence. For example, Priester and Petty (2001) have recently reported data to indicate that direct measures of ambivalence tap both intrapersonal ambivalence and discrepancy between one’s own attitude and that of valued others (i.e., interpersonal ambivalence).

We also have some concerns about split semantic differential measures of ambivalence because it is not clear whether respondents are able to comply with instructions to ignore judgments of the opposite valence (e.g., ignore negative evaluations when making a positive evaluation) when making a judgment. In addition, there is as yet only limited evidence to support the view that the positive and negative evaluations supposedly assessed by such measures are both available for retrieval from memory (Cacioppo, Gardner, & Bertson, 1997). However, we should note that much of our own research employing measures based on obtaining separate positive and negative judgments has produced findings consistent with theoretical predictions. Hence, such measures may have considerable validity.

Nevertheless, we also see the value of belief-based measures of ambivalence (see Conner et al., 1999c) as have other authors (e.g., Haddock & Zanna, 1998; van der Pligt et al., 2000). In fact, we have found both open- ended and closed-ended belief measures of ambivalence to be useful (see Haddock & Zanna, 1998; Esses & Maio, this volume, for useful discussions of open-ended measures of attitudes). Belief-based measures of attitude are consistent with Erber et al.’s (1995) attitudes-as-constructions model and al- low one to examine cognitive, affective, and cognitive/affective ambivalence by classifying the different beliefs. However, such measures are also not without problems. For example, it is not clear that individuals can easily report all the information upon which their overall attitude judgments are based (Fish- bein & Ajzen, 1975; Sparks, 1994). Respondents may also find it easier to generate cognitive correlates of their attitudes compared to other influences such as affective factors (Conner et al., 1999c; Sparks, 1994). This appears to be in stark contrast with evidence that affect-based evaluations are more accessible in memory than cognitive-based evaluations (Verplanken, Hofstee,


& Janssen, 1998). Hence, belief-based measures of ambivalence and split semantic differential measures may be based upon different information, thus accounting for their lack of convergent validity. Nevertheless, our own research (Conner ef d., 1999c) demonstrates that despite this lack of convergent validity, belief-based and split semantic differential measures of ambivalence are equally well able to moderate the attitude-behavior relationship (a key prediction for an ambivalence construct).

ELICITORS OF AMBIVALENCE

Comparatively little research has examined the possible elicitors of ambivalence. This section reports findings from two approaches to understanding the bases of ambivalence. The first approach examines what things commonly produce higher and lower levels of ambivalence across individuals. The second approach examines individual difference variables assumed to be important in the genesis of ambivalent attitudes.

Few studies have investigated what it is that tends to produce considerable degrees of ambivalence. Theoretically we might expect anything that tends to generate both strong positive and negative evaluations to be more likely to produce higher levels of ambivalence. One consequence of the work of Ortony et af. (1988), referred to earlier, is that actions/events likely to produce different emotions when focusing on the immediate future compared to the more distant future should produce greater ambivalence. Our previous research had indicated that this conflict was particularly common for health- related behaviors, where the conflict was between the immediate pleasantness/unpleasantness of the behavior and the longer-term positive/ negative health outcome of the behavior. Two types of health-related behavior can be distinguished: health protective behaviors, which can be unpleasant or onerous (e.g., exercising compared to non-exercising), and health- risking behaviors, which are pleasant (e.g., drinking alcohol compared to not drinking alcohol). We did not expect clear differences in ambivalence between health-protective and health-risking behaviors. Conner, Povey, Sparks, James, and Shepherd (1998a) identified 12 such health behaviors relevant to students. These behaviors varied in overall evaluation and also in whether engaging in the behavior was health-protective or health-risking. Split semantic differential measures were obtained in relation to these 12 different behaviors from a group of 143 students. The positive and negative evaluations were combined into measures of ambivalence using the equation suggested by Griffin (Thompson et al., 1995) and overall evaluation (by summing the two evaluations). The results are shown in Table 2.3. There were clearly considerable differences in the extent to which these different behaviors produced feelings of ambivalence. Drinking alcohol produced the highest degree of


ambivalence, while eating a low-fat diet produced the second highest level of ambivalence. Taking the drug Ecstasy and regularly eating breakfast produced the lowest levels of ambivalence. The reasonably large standard devia- tions for several behaviors indicate that people do differ in their reported ambivalence for a number of behaviors. While health-protective behaviors did tend to be evaluated positively and health-risking behaviors negatively, there did not appear to be a consistent relationship to levels of ambivalence (Table 2.3). In addition, there was no apparent relationship between overall evaluation and ambivalence.

Table 2.3 Ambivalence and evaluation measures for each of 12 behaviors (Conner el al., 1998a)

Behaviork Ambivalence' Evaluation"

Mean SD Mean SD ~~~ ~

Drinking alcohol (risk) 0.49a 0.16 0.15e 1.88 Reducing fat intake (protect) 0.34b 0.25 2.53* 2.25

Avoiding stressful situations (risk) 0.30bC 0.23 2.65' 2.36 Exercising (protect) 0.27Cde 0.21 3.51ab 1.83 Sleeping 7-8 hours per night (protect) 0 . 2 9 0.24 3.32ab 2.14

Using illicit drugs (risk) 0.31h 0.24 -2.33f 2.82

Healthy eating (protect) 0.24e 0.23 3.66' 1.96

Increasing fruit andor vegetable consumption (protect) 0.23e 0.21 3 . 4 w 1.96

Smoking (risk) 0.24e 0.25 -3.599 2.27 Using condoms (protect) 0.24e 0.21 3.71a 2.17

Regularly eating breakfast (protect) 0.23e 0.21 3.07bC 2.42 Taking Ecstasy (risk) 0.22e 0.23 -3.5% 2.58

k Means in the same column not sharing the same letter are significantly different from each other: risk = health-risking behaviors; protect = health-protecting behaviors. I Ambivalance is calculated from the positive (scored 0 to +6) and negative (0 to -6) evaluations. Higher scores indicate greater ambivalence (measures all scaled on the range 0-1). m Evaluation is the sum of the positive and negative evaluations (possible range -6 to +6). Higher scores indicate more positive attitudes.

In a second study, we asked a sample of 102 students in a more open-ended manner to name the things that either made them feel neither positive nor negative, just positive and not at all negative, just negative and not at all positive, or both positive and negative (Waterman & Conner, 1999). This latter category should contain those things that produced high degrees of ambivalence. The responses were independently coded into a number of categories. Among this sample the things that most commonly produced both positive and negative reactions were eating, drinking alcohol, and personal relationships. Eating chocolate and foods rich in calories and binge drinking were particularly frequently mentioned. In both these cases the conflict appears to be between a more immediate pleasant outcome and a more distant negative outcome. The fact that eating (and eating these foods in

AMBIVALENCE AND AlTITUDES 49

particular) arouses high levels of ambivalence was an important factor in our choice to examine these behaviors when examining the consequences of ambivalence (see Sparks et al., 2001; Armitage & Conner, 2000) (see below). Amongst the relationship category, the relationship with sexual partner and family were particularly frequently mentioned. Common here was a conflict between a liked person performing an action that was not approved of. Exam- ination of the other (non-ambivalent) categories revealed considerably more variation in the types of attitude objects reported.

Across these two studies it is interesting to note that ingestive behaviors (drinking alcohol, eating) are the most frequently reported sources of ambivalence. These are behaviors commonly engaged in by the vast majority of respondents and ones for which both positive and negative outcomes are common. Interestingly, they are also both behaviors where the ambivalence is often between appraising immediate outcomes and those some time in the future or between appraising an object (e.g., family member) and appraising an event (e.g., performance of a disapproved behavior). The mapping on to the predictions of Ortony et al. (1988) is noteworthy.

A second approach to understanding the elicitors of ambivalence examines individual difference variables (Thompson et al., 1995). It is suggested that the experience of ambivalence is not merely associated with a particular attitude but may characterize an attitude domain. Thompson and Zanna (1995) demonstrated some convergence in individual levels of ambivalence across five different social policy issues. Such individual differences in ambivalence across attitude topics might be attributable to the operation of other individual difference variables. For example, individuals with particular personality traits might be more or less likely to resolve conflicts between positive and negative evaluations towards various attitude topics, and so exhibit lower or higher levels of ambivalence across topics. Several different personality dimensions have been examined in this regard. These include more specific personality dimensions, such as need for cognition, personal fear of invalidity (Thompson & Zanna, 1995), need for closure, and self-monitoring (Sparks, Conner, James, Shepherd, & Povey, 1995). As well as these personality dimensions, research has also investigated issue involvement as a potential moderator of this relationship (Sparks et al., 1995; Thompson & Zanna, 1995).

Thompson and Zanna (1995) report the relationship of a split semantic differential measure of ambivalence to the personality dimensions of need for cognition (Cacioppo & Petty, 1982), personal fear of invalidity (Thompson, Naccarato, & Parker, 1989), and level of involvement with the attitude object. Across five social policy issues, degree of ambivalence was significantly negatively related to need for cognition and positively related to personal fear of invalidity. Whilst involvement did not show consistent direct relationships with ambivalence, there was evidence of moderation. In particular, the above relationships of need for cognition being negatively related to ambivalence


and personal fear of invalidity being positively related were only apparent when respondents were highly involved with an issue. Hence, the mechanism by which these personality variables influence ambivalence (i.e., through influencing propensity to engaging in resolving evaluatively inconsistent reactions) may only operate when individuals are highly involved with an issue.

However, this effect may be relatively weak and only be detectable when considering the average ambivalence across several attitude topics. For example, Sparks et al. (1995) found no evidence that a split semantic differential measure of ambivalence about eating chocolate or meat was significantly related to need for cognition, need for closure (Kruglanski, Webster, & Klem, 1993), or self-monitoring (Snyder & Gangestad, 1986). In addition, there was no evidence for a moderating effect of involvement with the behavior. Hence, we believe the current evidence that individual difference variables elicit ambivalence is weak. We suspect that more specific research hypotheses are required in this area. One interesting hypothesis is that the consequences of ambivalence may be different for different personality types (e.g., ambivalence may be more unpleasant or anxiety-provoking for those intolerant of ambiguity).

As the above review makes clear, we still know comparatively little about the elicitors of ambivalence or about how ambivalence might best be manipulated. These are important areas for further research (see below).

CONSEQUENCES OF AMBIVALENCE

If we consider attitudinal ambivalence as a dimension of attitude strength, then we might anticipate four consequences of holding an ambivalent attitude (Converse, 1995, p. xi; Krosnick & Petty, 1995): (a) impacts on the temporal stability of attitudes; (b) impacts on the relationship of attitudes to intentions and behavior; (c) impacts on the pliability of attitudes; and (d) impacts on information processing. A number of authors have investigated the effects of attitudinal ambivalence in relation to each of these consequences, although the evidence in support of each effect varies considerably. In the following sections we review the research in relation to each of these four consequences of having an ambivalent attitude, focusing on the findings from our own research.

Temporal Stability

A variety of studies have demonstrated that strong attitudes are more stable over time. Indeed, a number of authors have argued that attitude stability is the factor that all attitude strength measures hold in common (Ajzen, 1996; Doll & Ajzen, 1992; Erber et al., 1995). Other research with intentions (e.g.,


Conner, Sheeran, Norman, & Armitage, 2000, Sheeran, Orbell, & Trafimow, 1999) and personality constructs (e.g., Biesanz, West, & Graziano, 1998) has noted the importance of temporal stability as a measure of the strength of a psychological construct. On this view, attitudes high in ambivalence would represent “weak” attitudes and be predicted to show lower temporal stability compared to attitudes low in ambivalence.

Wilson and Hodge’s (1992) am’tudes-us-constructions model provides one framework for understanding this proposed relationship between ambivalence and attitude stability (see also Erber et af., 1995; Zaller & Feldman, 1992). Central to the model is the idea that when attitudes are required (e.g., for decision-making), they may be constructed from available information (e.g., via contextually-cued beliefs about the attitude object), rather than being stored in long-term memory. Ambivalence is therefore a function of the extent to which the underlying information (e.g., beliefs, feelings) on which the attitude is based is inconsistent. That is, when a person constructs an attitude at different time points, the degree of attitude @)stability between time points should vary as a function of the level of evaluative incongruity between positive and negative evaluations of the object or behavior at each time point. From this perspective, inconsistency in the determinants of attitudes provides a structural basis for ambivalence, which in turn regulates stability. It also implies that levels of ambivalence and attitude stability are inextricably linked. Hence, less ambivalent attitudes should be more stable over time.

Two types of research have tackled this predicted consequence of ambivalence. In one approach, ambivalence measures have been correlated with self-report measures of attitude variability and stability (i.e., meta-attitudinal measures). In our own data using this approach, ambivalence is found to be positively related to self-reports of attitude variability and negatively to attitude stability. For example, in a study of the use of the drug Ecstasy by clubbers (Conner, Sherlock, & Orbell, 1998b), we obtained multi-item self- report measures of ambivalence (using a split semantic differential measure). Although not in the journal report of this research, we also measured (see Conner & Sherlock, 1997) attitude variability (e.g., “Sometimes my attitude towards Ecstasy varies, sometimes it is positive and sometimes it is negative”, strongly disagree-strongly agree). Ambivalence and attitude variability were found to be significantly positively related ( r = 0.43). We found a similar level of correlation between a two-item measure of felt ambivalence (e.g., “I have conflicting feelings about having casual sex”, strongly disagree-strongly agree) and attitude variability (“My attitudes about casual sex”, do not vary at all-vary greatly) in a study of intentions to engage in casual sex amongst students (i.e., a “one-night stand”), r = 0.61, significant (Conner & Flesch, 1998; see Conner & Flesch, 2001, for further details of this research). Such data is not wholly convincing because it is not clear that respondents are clearly able to differentiate the two constructs when making their responses.


In a second and more defensible approach, ambivalence has been related to more indirect measures of attitude stability (i.e., operative measures). For example, Bargh, Chaiken, Govender, and Pratto (1992) showed that respondents who gave the same evaluation to an attitude object at two time points tended to have less ambivalent attitudes than those who did not give the same evaluation, ( r = 0.27, significant). Bassili (1996) reported contrasting results; ambivalence was unrelated to stability for attitudes towards introducing quotas, r = 0.00, or towards pornography, r = 0.03.

In our own research we have examined the stability predictions in one study (Armitage & Conner, 2000). These authors (study 1) examined the impact of ambivalence on the stability of attitudes towards eating a low-fat diet over three time points in a sample of 346 hospital workers. Am- bivalence was assessed using a split semantic differential measure, while attitudes were assessed using a reliable multi-item bipolar semantic differential measure. As attitudes were measured at three time points, a total of three comparisons were possible, representing time intervals of 3 months, 5 months and 8 months between measurement of attitudes. The extent to which ambivalence moderated the degree of attitude stability was investigated in two ways. First, we computed a within-subjects correlation measure of attitude stability. Across respondents, this measure of the stability of attitudes was unrelated to our measure of ambivalence (over 3 months, r = -0.02; over 5 months r = -0.01; over 8 months, r = -0.01). Second, we attempted to discount any problems attributable to error variance by com- puting measures of attitude stability based on between-subject correlations disattenuated for measurement error (using LISREL 8; Joreskog & Sorbom, 1993). In this analysis correlations were computed separately for the higher and lower ambivalence groups (based on a median split). In general, the degree of attitude stability was high (mean disattenuated r = 0.71), although, as might have been expected, stability was inversely related to the time interval between measurements. In no case did the differences in disattenuated correlations reach significance (over 3 months, Zr = 0.00, over 5 months, Zr = 1.59; over 8 months, Zr = 1.24), although the difference was in the predicted direction (Le., less stability in the high ambivalence group) for comparisons over the longer time intervals.

Thus, research provides only limited support for the predicted impact of ambivalence on attitude stability. Whilst self-reported stability of attitudes appears to be relatively consistently related to ambivalence, designs employing prospective measures of stability do not tend to observe this relationship of ambivalence to attitude stability. The studies based on self-report measures of stability are particularly open to consistency biases. The one study reporting a significant effect using a more defensible measure of attitude stability (Bargh et al., 1992) employed a relatively small sample (but see Jonas et al., 2000, for report of a further study producing supportive results).


Attituddntention and Attitude-Behavior Relationships

A key prediction for a measure of attitude strength is that strong attitudes should be more likely to guide behavior than weak attitudes (Converse, 1995, p. xi; Krosnick & Petty, 1995, p. 3). As Schwartz (1978) noted, an attitude assessed at one time is unlikely to predict behavior at a later time if the attitude does not persist over the intervening time interval. Fishbein and Ajzen (1975) make a similar observation in relation to the impact of intentions on behavior (see Conner et d., 2000). Some researchers (e.g., Erber. et al., 1995) have argued that the greater impact of strong attitudes on behavior is solely attributable to strong attitudes being more likely to persist over time. However, other mechanisms may also influence the ability of strong attitudes to influence behavior. For example, Fazio (1986, 1995) has argued that attitudes influence our behavior in part by shaping our perceptions of the world (the automatic activation effect). That is, the capacity of an attitude to predict behavior is partly dependent on the attitudes’ ability to bias perceptions of the attitude object and the context in which the behavior is performed. Strong attitudes are assumed to be more readily accessible and so more likely to produce these biasing effects. Thus, one might therefore expect low- ambivalent attitudes to be more capable of automatic activation and so more predictive of behavior than high-ambivalent attitudes.

For many attitude objects that elicit high levels of ambivalence, it is useful to consider consistency of expressed attitudes over context as well as consistency over time. Ambivalence towards a behavior may be associated with a temporal and/or contextual instability in dominant motives (or evaluations) that arises as a consequence of fluctuations in the salience and/or strength of different motives. Ambivalence may not only reflect “mixed feelings” towards an attitude object at a particular moment in time but also changes in the dominance of competing motives at different points in time as a result of variations in internal and external contextual cues (cf. Ainslie, 1992). While measured attitudes may show a consistency over time if elicited in the same context, this consistency may not be achieved if variations in context are introduced into the measurement procedure: for example, many people’s expressed attitude towards eating fruit as dessert may be stable over time if measured in similar laboratory conditions and may only show variation when measured after dinner in full sight of the chocolate cake alternative! The strength of the relationship between attitudes, intentions, and behavior may be expected to be influenced by whether or not the same evaluations or motives are dominant when attitudes, intentions, and behavior are measured (cf. Ajzen, 1996 Tesser, 1978). Since ambivalence is associated with mixed or conflicting motives, the strength of measured attitude-intention-behavior linkages may be consequently attenuated at higher levels of ambivalence (compared to lower levels of ambivalence).


In addition, much research in the attitude domain has considered behavioral intentions as mediators of attitude-behavior relations (e.g., Ajzen, 1991; Fishbein & Ajzen, 1975) and a considerable body of research supports this relationship (see Armitage & Conner, in press a; Conner & Armitage, 1998; Conner & Sparks, 1996; Eagly & Chaiken, 1993; Godin & Kok, 1996; Sheppard, Hartwick, & Warshaw, 1988; Van den Putte, 1991). Indeed, in researching the attitude-behavior relationship, many studies have employed measures of behavioral intention as a proxy for actual behavior. Previous work on attitudinal ambivalence has examined the moderation of either attitude-intention relations (e.g., Jonas, Diehl, & Bromer, 1997) or attitude- behavior relationships (e.g., Lavine et af., 1998).

The earliest examination of the moderation effect was by Moore (1973, 1980). His research demonstrated a moderating effect of attitudinal ambivalence on the relationship between attitudes and behavior, such that there was lower consistency between attitudes and behavior for those high in attitudinal ambivalence. However, measures of attitude and behavior were weak and taken in a cross-sectional manner. Thus, this moderation effect is more accurately described as influencing attitude-past behavior relationships. Sparks et al. (1992) examined the moderating role of a mixed measure of attitude ambivalence and attitude variability on attitude-intention relationships in relation to food choices. Higher levels of ambivalence were associated with an attenuation of the attitude-intention correlation. Norman and Smith (1995) also used a mixed measure of attitude ambivalence and attitude variability, but found no moderating effects on belief-attitude or attitude- intention relationships for exercise behavior. However, in comparing differences in zero-order correlations they failed to take account of potential differences in variance in the higher and lower ambivalence groups (Aiken & West, 1991; Baron & Kenny, 1986). This problem can be addressed through the use of moderated regression analysis or structural equation modeling. Jonas et af. (1997), in a laboratory study, found evidence of stronger attitude-behavioral intention relations for more ambivalent attitudes. Finally, research by Lavine et af. (1998) reported that attitude-behavior relations were attenuated by high levels of affective-cognitive inconsistency.

Hence, previous studies have produced mixed findings on the moderating impact of ambivalence on attitude-intention and attitude-behavior relationships. However, each of the above studies can be criticized in relation to the design (Moore, 1973, 1980), the mixed measure of ambivalence employed (Lavine et af., 1998; Norman & Smith, 1995; Sparks er af., 1992), the analysis method employed (Norman & Smith, 1995), or the focus on laboratory- generated attitudes (Jonas ef af., 1997). Our own research has attempted to address this issue through five cross-sectional studies on attitude-intention relationships (Conner & Flesch, 1998; Conner, Povey, & Sparks, 1995; Conner et af., 1998b; Conner, Povey, Sparks, James, & Shepherd, in press; Sparks et


al., 2001) containing seven independent tests, two prospective studies of the attitude-behavior relationship (Conner et al., 1999c, in press) containing four independent tests, and one study that examined these relationships simultaneously in a prospective design (Annitage & Conner, 2000, study 1). In each case we employed a measure of ambivalence that was consistent with our definition of ambivalence (although the precise measure varied between studies) and used appropriate statistical tests (see Conner, Sparks, & Armi- tage, 1999b; Conner et al., 1998b; Sparks, Conner, James, Povey, & Shepherd, 1997a; Sparks, James, Conner, Shepherd, & Povey, 1997b, for earlier reports of this work).

A ttitude-Zntention Relationships

One of our studies examined the moderating role of ambivalence on attitude- intention relationships in relation to students engaging in casual sex (Conner & Flesch, 1998). The study employed a hypothetical scenario about meeting a new sexual partner, and involved the respondents producing ratings of attitude, intentions, and ambivalence in relation to having sex with this new partner. Attitude was assessed by a series of bipolar semantic differentials, which factor analysis revealed to fall into components assessing cognitive (three items) and affective (two items) factors. Intentions were assessed by six items. Ambivalence was assessed by two items intended to tap direct measures of affective ambivalence (e.g., “I have conflicting feelings about casual sex”, strongly agree-strongly disagree). Moderated regression analysis (Baron & Kenny, 1986) demonstrated that both attitude measures significantly predicted intentions. Ambivalence also explained significant additional variance in intentions (higher levels of ambivalence were associated with stronger intentions to engage in sex). However, most importantly, ambivalence significantly moderated the impact of affective attitudes, but not cognitive attitudes on intentions. This effect was also in the hypothesized direction, with lower ambivalence being associated with stronger affective attitude-intention relationships.

In a series of six related studies, we have examined the impact of ambivalence in moderating the relationship between attitudes and intentions in relation to a series of food choices (Conner, Sparks, Povey, James, & Shepherd, 1996). Food choices were selected for study because this is a domain where ambivalence is commonly reported (e.g., conflict between pleasure and health; see Beardsworth, 1995; Conner er al., 1998a). The design of the studies was similar. In each case, reliable multi-item self-report measures of attitude (semantic differential measure) and intention (e.g., “I intend to eat a low-fat diet”, strongly agree-strongly disagree) were taken. In addition, a split semantic differential measure of ambivalence was taken in each case. The studies examined healthy eating (n = 239 Conner et al., in press), eating a low-fat diet (n = 158;


Conner et al., 1995; n = 145; Conner et al., 1999c), eating chocolate (n = 155; Sparks et al., 2001), eating meat (n = 165; Sparks et al., 2001), or eating five portions of fruit and vegetables per day (n = 149; Conner et al., 1999~). The results from these studies are summarized in Table 2.4.

Table 2.4 Summary of moderating effects of attitudinal ambivalence on attitude- intention and attitude-behavior relationships across studies

Numbers of supporting results Frequency-weighted correlation

Significant Non- Non- Lower Higher moderating significant significant ambivalence ambivalence effect effect effect

(high > low) (low > high) ~~~ ~~ ~

Attitude-intention 2 3 1 0.67.a 0.60' Attitude-behavior 3 1 0 0.53'b 0.40'

Differences of r from zero: p < 0.001; differences between r: p < 0.09; p < O.OOlb

Moderating effects of ambivalence were assessed by moderated regression analysis (Baron & Kenny, 1986). However, for presentational purposes, we split the sample at the median on the ambivalence measure in order to create higher ambivalence and lower ambivalence groups. Correlations between attitude and intention measures were then computed separately for these two groups and then compared. As can be seen in Table 2.4, the results were mixed, with some studies demonstrating the predicted direction of moderation effect and others failing to show such an effect. However, in summarizing across studies (Table 2.4) and comparing the frequency-weighted correlations, it becomes clear that across studies there is evidence for the moderating impact of ambivalence on the attitude-intention relationship (lower ambivalence group mean, r = 0.67; higher ambivalence group mean, r= 0.60), although the difference is modest. We suspect that the failure to find a moderation effect in some studies is attributable to one of two factors. The first factor is degree of ambivalence. Where the behavior that is the focus of the intention does not produce a considerable degree of ambivalence in at least some respondents, the moderation effect is unlikely to be observed (i.e., the lower and higher ambivalence groups are too similar). To test this effect we examined the mean degree of reported ambivalence in each of the six studies reported in Table 2.4. Consistent with these predictions, the highest mean levels of ambivalence were found in the two studies reporting significant effects in the prediction direction (Conner et al., 1999c; Sparks et al., 2001). However, this does not represent a sufficient explanation because one study where we failed to find an effect also showed high mean levels of ambivalence (Conner er al., in press).


A second factor which might explain the failure to find consistent effects is the reliability of the ambivalence measure. A common psychometric principle to increase reliability is to obtain multiple measures of a construct and then average or sum across measures where they show high intercorrelations. In relation to the studies reported in Table 2.4, which reported high mean levels of ambivalence, the one study not showing a significant effect (Conner et al., in press) only employed a single-item measure of ambivalence. We suspect this may have compromised our comparison of higher and lower ambivalence individuals and masked any moderation of the attitude-intention relationship. Hence, across studies of food choices that produce considerable levels of ambivalence (when reliably assessed), we find consistent evidence for ambivalence attenuating the attitude-intention relationship, as assessed in between-subject correlations. Perhaps because the measures tended to be taken at the same time point, and so perhaps be masked by consistency biases, the differences were modest.

One problem with between-subject analyses such as these is that they as- sume that, for a given relationship, the person with the strongest attitude will form the strongest intention. However, a more psychologically realistic view would suggest that the behavioral intention selected by an individual reflects the behavior towards which they have the most positive attitude. Within- person correlations more accurately reflect this process. A final study examined the moderating role of ambivalence on attitude-intention relationships using a within-subjects design (Conner et al., in press); 239 members of the public completed a questionnaire measuring attitudes and intentions in relation to 20 healthy eating behaviors (e.g., avoiding high-fat foods, eating fruit and vegetables, etc.). Ambivalence was assessed by a split semantic differential measure concerning eating a healthy diet. As before, we created two groups based on a median split of the ambivalence measure. The analysis then compared the mean within-subjects correlation between attitudes and intentions in the lower ambivalence group (M = 0.86; SD = 0.54) and the higher ambivalence group (M = 0.79; SD = 0.57). This difference was found to be significant. Such analyses avoid some of the problems with the other datasets and provide good evidence that ambivalence moderates the attitude- intention relationship, such that higher levels of ambivalence are associated with weaker attitude-intention relationships (see also MacDonald & Zanna, 1998, for consonant results).

Attitude-Behavior Relationships

In a series of four independent prospective datasets, we examined the impact of ambivalence on moderating the relationship between attitudes and self- reported behavior in relation to a series of food choices (Conner et al., 1996). Again, the design of these studies were similar. In each case, self-report


measures of attitude (semantic differential measure) and ambivalence (split semantic differential measures) were taken in an initial questionnaire; while self-report measures of behavior (e.g., “I have eaten a low-fat diet in the past month”, strongly agree-strongly disagree) were taken in a later questionnaire. The studies examined healthy eating (Conner er al., in press), eating a low-fat diet (Conner et al., 1999c; Conner et al., 1995), or eating five portions of fruit and vegetables per day (Conner el al., 1999~). The results from these studies are summarized in Table 2.4. For presentational purposes we again split the sample at the median on the ambivalence measure in order to create a higher and a lower ambivalence group. Correlations between attitude and behavior measures were computed separately for these two groups and then compared. As can be seen in Table 2.4 the results were fairly consistent in showing the predicted moderation effect. The one study failing to show a significant effect (Conner et al., in press) employed a single-item measure of ambivalence, which may have compromised its reliability. Across studies the frequency-weighted correlation shows this effect to be substantial, with more ambivalent attitudes being less strongly correlated with behavior assessed prospectively, r = 0.40 vs. r = 0.53 (Table 2.4). In addition, this effect was also apparent in within-subjects correlations for healthy eating in a prospective design (Conner et al., in press). The lower ambivalence group (M = 0.57; SD = 0.37) and the higher ambivalence group (M = 0.47; SD = 0.35) differed significantly on attitude-behavior correlations in the predicted direction (i.e., weaker correlations for the higher ambivalence group).

One problem with the above studies is the failure to take account of the effect of ambivalence on attitude-intention and attitude-behavior relationships simultaneously. Armitage and Conner (2000, Study 1) attempted to rectify this problem by taking measures of attitude, intention, and behavior at three different time points. In addition, this study used structural equation modeling in order to remove any effects due to measurement error. Data were collected from a sample of 350 hospital workers in relation to eating a low-fat diet. The measure of ambivalence was a split semantic differential measure that was used to split the group into a higher and lower ambivalence group (based on a median split). We then fitted a model to the two groups with structural paths from attitude to intentions, from attitude to behavior, and from intentions to behavior. As the interest was in the relationship between the latent constructs (attitude, intentions, and behavior), we con- strained the model such that the factor loadings for each construct were equal in the two groups (the factor loading constraint model). This allows us to make direct comparisons between the regression paths for the higher and lower ambivalence groups. As predicted, the path from attitude to intention was stronger in the lower compared to the higher ambivalence group. Also as predicted, the path from attitude to behavior was stronger in the lower compared to the higher ambivalence group. Formal tests of these differences


revealed them both to be significant. These effects were present despite any evidence that ambivalence influenced attitude stability in the same data.

Thus, there appears to be growing evidence that ambivalence moderates the attitude-intention relationship and, particularly, the attitude-behavior relationship, such that higher levels of ambivalence are related to weaker relationships. It would also appear that such effects are not purely attributable to attitude stability effects. It has been suggested that this moderation effect may be due to ambivalence being confounded with attitude extremity. For example, Thompson et al. (1995) noted that attitudinal ambivalence is necessarily negatively correlated with attitude extremity (p. 382), that is, univalent attitudes should be more extreme than ambivalent attitudes and therefore more predictive of behavior. This hypothesis is directly supported by Lavine et al. (1998; but see Jonas et al., 1997) and indirectly supported by Kallgren and Wood (1986). The latter compared experimental participants, whose attitude extremity had been manipulated, with a no-treatment control group, and reported attitude- behavior correlations of 0.85 and 0.19 for the experimental and control conditions, respectively. One might therefore expect less ambivalent attitudes to be more predictive of behavior than ambivalent attitudes, just because of greater attitude extremity. Nevertheless, there is evidence that the moderation effects observed for ambivalence are not merely attributable to differences in attitude extremity. For example, Conner et af. (in press) while reporting the predicted attitude-intention and attitude-behavior moderation effects for ambivalence, found these relationships remained even when ambivalence was unconfounded from extremity. Most convincing, however, is the data reported by Wester and Petty (2000). In a study of safer sex behaviors, they compared a lower and a higher ambivalence group. Ambivalence was found to moderate the attitude- behavior relationship. They then examined only those respondents checking the most positive (extreme) attitude category. In this case, lower-ambivalent respondents were significantly more likely to report performing the behavior than higher-ambivalence respondents, despite being identical on attitude extremity. In examining the moderation effects of ambivalence on attitude- intention and attitude-behavior relationships, it is clearly important to ensure that these effects are not simply attributable to attitude extremity effects. In addition, these reported moderation effects need to be confirmed with objective measures of behavior and in experimental studies.

Pliability of Attitudes

More ambivalent attitudes are expected to be more susceptible to the influence of a persuasive communication (i.e., more pliable). For example, Eagly and Chaiken (1995) argue that, “attitudes are strong to the extent that they are well embedded in an existing attitudinal structure’’ (p. 414). Thus strong attitudes are held to be more securely “anchored” in knowledge structures.

60 MARK CONNER A N D PAUL SPARKS

Given that ambivalent attitudes are based on conflicting evaluations and inconsistent information, they should be more weakly anchored and hence more pliable than univalent attitudes.

Only a few studies have investigated this prediction for ambivalence. Bassili (1996) tested the pliability of attitudes towards three attitude objects in the face of persuasive messages. Pliability was found to be significantly related to ambivalence for two attitude objects ( r = 0.15 in both cases), but not for a third (two tests: r = 0.06 and r = 0.03) (see also Pomerantz, Chaiken, & Tordesillas, 1995). MacDonald and Zanna (1998) showed individuals high in (cross-dimensional) ambivalence about a particular social group to be more susceptible to a priming manipulation in terms of their judgment about that group. Finally, Maio, Esses, and Bell (2000) found an interaction between strength of message and ambivalence on attitudes towards residents of Hong Kong, such that only in the high ambivalence group did message strength produce differential effects on attitude change.

In our own research we have investigated the impact of ambivalence on the pliability of attitudes in one study (Armitage & Conner, 2000, Study 2). The study was a simple pre-post quasi-experimental design. Attitudes and levels of ambivalence towards eating a low-fat diet were assessed at baseline in a sample of 344 hospital workers. Attitude measures were based on a multi- item semantic differential measure, while ambivalence was assessed using a split semantic differential measure. Five months later, participants were ran- domly assigned to either an attitude change or control condition and received experimental materials designed either to change attitudes or provide information only. Both sets of intervention materials included UK Government recommendations on dietary fat intake, epidemiological data on population levels of fat intake, and a description of sources of fat in the diet. The experimental intervention also included a section designed to change individual attitudes (see Armitage & Conner, in press b). Based on Fishbein and Ajzen’s (1975) model of belief-attitude relations, we provided persuasive information that targeted five beliefs about eating a low-fat diet, which had previously shown to distinguish who did and did not intend to eat a low-fat diet (Armi- tage & Conner, 1999).

Analysis of variance revealed that the experimental and control groups did not differ at baseline on the attitude measures. In order to test the prediction that less ambivalent attitudes are more resistant to change, we split the group into higher and lower levels of ambivalence, based on a median split of this measure. Across both our intervention conditions, attitudes became more positive, although the change was greater in the experimental group. Of particular interest was the three-way interaction between condition, ambivalence, and time. This only approached significance. Nevertheless examination of the effects of the intervention demonstrated that it had differential effects in the two ambivalence groups. For the lower ambivalence


group, there was no difference between the control and experimental conditions; while for the higher ambivalence group, attitudes became significantly more positive following the experimental intervention. Hence, the attitude intervention (experimental) group had significantly more impact on more ambivalent attitudes supporting the pliability prediction.

Hence, the evidence to date generally supports the idea that ambivalent attitudes tend to be more pliable in the face of persuasive messages. Nev- ertheless, further studies across a more diverse range of attitude objects are required.

Information Processing Impacts of Ambivalence

A further predicted consequence of attitudinal ambivalence is an impact on the processing of attitude-relevant information. Attitudes characterized by higher levels of ambivalence are assumed to be less likely to guide processing of information about the attitude object, because of their lower accessibility. Several studies have found evidence to support the relationship of higher levels of ambivalence to lower attitude accessibility (Bargh et al., 1992; Bassili, 1996; Bromer, 1998; see Jonas et al., 2000, for a review). For example, Van der Pligt et al. (2000) hypothesized that those individuals with ambivalent attitudes would take longer to integrate information into an overall attitude judgment compared to non-ambivalent individuals, and hence ambivalent attitudes would show longer response latencies when compared to non-ambivalent attitudes. Across three different attitude objects support was found for this effect. In each case the response latency for an overall attitude judgment was longer for ambivalent compared to non- ambivalent individuals.

In our own research we have investigated this effect in one study (Conner, Bainbridge, & Halton, 1999a). This study used a response latency paradigm to examine the time taken to respond to different attitude objects. Consistent with previous research, we expected bipolar attitude judgments to show longer latencies when the respondents reported higher degrees of ambivalence. Ambivalence was assessed using split semantic differential measures to separately assess positive and negative reactions to the attitude object. However, we further hypothesized that ambivalence would only influence the response latency to the bipolar attitude measure when the two unipolar judgments had similar short latencies. Consistent with Erber et d ’ s (1995) attitudes-as-constructions model, this prediction was based on the view that, in arriving at an overall judgment about an attitude object (i.e., as tapped by the bipolar rating), respondents can take account of both negative and positive evaluations (i.e., as tapped by the unipolar ratings), but will only do so if they are salient (i.e., have a shorter latency). Thus, only when both unipolar judgments are salient and in conflict (i.e., ambivalent) will response


latencies for the overall attitude judgment be longer. Within-subject analyses supported this view, with latencies for bipolar attitude judgments being significantly longer when the respondent was ambivalent about the same attitude object and both unipolar evaluations had similar and short response latencies (see McGregor, Newby-Clark, & Zanna, 1999, for a similar finding).

Thus, there appears to be reasonably good evidence that ambivalence leads to lower accessibility. However, the evidence that this reduced accessibility has significant impacts on the processing of information about the attitude object has been addressed in fewer studies. Several studies appear to support the idea that higher levels of ambivalence lead to a more systematic processing of attitude-relevant information. The assumption is that an ambivalent attitude is weaker than a non-ambivalent attitude and so is less likely to produce a biased processing of attitude-relevant information (e.g., information consistent with one’s initial attitude receives more favorable processing than information that is inconsistent). For example, Maio, Bell, and Esses (1996) showed that higher levels of ambivalence about a minority group resulted in more systematic processing of information about that group. Similarly, Bromer (1998) showed that only individuals lower in ambivalence showed a preference for attitude-congruent information in processing messages about European Monetary Union. Jonas et al. (1997) suggest that the mediating mechanism is confidence in one’s attitude: high ambivalence produces a reduction in confidence in one’s attitude and reduced confidence leads to more systematic processing of information.

Summary of Research Findings on the Consequences of Ambivalence

As the above review demonstrates, there is a growing body of research on the differing consequences of ambivalence. The findings support the view that ambivalence is an important issue for those involved in attitude research (Breckler, 1994; Thompson et al., 1995; Priester & Petty, 1996), and that ambivalence produces many of the effects one might expect of a measure of attitude strength. Perhaps the strongest evidence concerns the moderating effects of ambivalence on attitude-intention and attitude-behavior relationships. Those with ambivalent attitudes show significantly weaker relationships between their attitudes and intentions or attitudes and behavior. We would anticipate that a range of behaviors that show conflict between cognitive and affective influences (e.g., safer sex practices, exercise, food choices) might be characterized by high levels of ambivalence and so show attenuated attitude- behavior relationships. However, the generality of the moderation effect across different behaviors still needs to be established and demonstrated experimentally. For example, Jonas et al. (1997) and Maio et al. (1996) suggest that ambivalence can, under certain circumstances, lead to stronger attitude- behavior relationships.


The predicted effects of ambivalence on the stability and pliability of attitudes were less consistent. The prediction that less ambivalent attitudes will be more stable over time has only been confirmed with meta-attitudinal measures of stability and has been refuted in two studies using more appropriate measures of attitude stability (Bassili, 19%; Annitage & Conner, 2000, Study 1; but see Bargh et al., 1992). Similarly, the prediction that ambivalent attitudes will be more pliable has only received limited support to date (Armitage & Conner, 2000, Study 2; Bassili, 1996). In relation to information-processing effects, several studies demonstrate that ambivalent attitudes are less accessible. However, only two studies (Bromer, 1998; Maio er al., 1996) have shown impacts of ambivalence on information processing; greater ambivalence appears to be associated with more systematic processing of attitude-relevant information.

Given that ambivalence measures show some evidence of producing all the effects one might expect of a measure of attitude strength, one might ask what is the relationship of ambivalence to other measures of attitude strength. Although there have been several empirical studies of the relationships amongst attitude strength measures (e.g., Bassili, 1996; Eagly & Chaiken, 1995; Jonas et al., 2000; Krosnick, Boninger, Chuang, Berent, & Carnot, 1993; Pomerantz et al., 1995) the relationship of ambivalence to the full range of attitude strength measures remains under-researched. Ambivalence does tend to be associated with low accessibility of the attitude (Bargh et al., 1992; Bromer, 1998), lower attitude extremity (Maio et al., 1996), and lower attitude certainty (Bassili, 1996; Jonas et al., 1997), and unrelated to evaluative- cognitive or evaluative-affective consistency (Maio et al., 1996, 2000). However, the relationship of ambivalence to other attitude strength dimensions, such as attitude importance, knowledge, interest, and intensity remains an issue for future research.

FUTURE DIRECTIONS FOR RESEARCH WITH THE AMBIVALENCE CONSTRUCT

Whilst our review has demonstrated that we now have a greater understanding of the ambivalence construct, particularly in relation to some of the consequences of ambivalence, we believe that a number of areas are worthy of further research attention. We would highlight three such key areas for research. The first is the need to further explore the interrelationship of ambivalence measures. As the above review makes clear, there is evidence for the divergent validity of various different ambivalence measures. However, there is also a need for researchers working with the ambivalence concept to be clear about their interpretation of the concept, in order that the area is not dogged by differences of interpretation that are not made explicit. Further research is necessary to identify the relationships amongst ambivalence


measures across different attitude objects. A multi-trait multi-method matrix study (see Campbell & Fiske, 1959) might be particularly informative in this regard in isolating the degree to which variations were due to differing types of ambivalence measurement method (e.g., open-ended vs. closed-ended measures; direct ambivalence vs. split semantic differential measures; split semantic differential vs. belief-based measures) or attributable to different attitude objects (e.g., events, objects or actions).

A second key issue for future research is the factors that elicit ambivalence. We still know relatively little about which attitude objects produce considerable levels of ambivalence or the factors promoting the formation of ambivalence. Future studies might usefully employ experimental methods to study such effects (e.g., see Jonas et al., 1997), although it is not clear what might represent the best means for manipulating ambivalence (Wegener et al., 1995). A third key area is the consequence of holding an ambivalent attitude, and in particular the implications for information processing. Whilst a number of studies have examined other consequences of holding an ambivalent attitude, only a few studies have considered the implications for information processing (but see Bromer, 1998). An investigation of these effects may throw light on a number of the other consequences of ambivalence. Other information-processing implications of holding an ambivalent attitude also deserve further attention.

CONCLUSIONS

In the present chapter, we have reviewed the various definitions of ambivalence that have been suggested in the literature and indicated a number of different forms of ambivalence. We have presented methods for measuring different forms of ambivalence and some preliminary findings about their degree of interrelationship. We have also described findings relating to the elicitors of ambivalence. In the major section of the chapter, we then reviewed findings for the ambivalence construct in relation to four major predictions based on considering ambivalence a measure of attitude strength: temporal stability of attitudes, impact of attitudes on intentions and behavior, pliability of attitudes, and impacts on information processing. There is evidence for each consequence, although the strongest evidence supports the moderating role of ambivalence on attitude-intention and attitude-behavior relationships. Finally, we outlined a number of suggestions for further research: in particular, the need to further explore the interrelationship of ambivalence measures, the key determinants of ambivalence, and the information- processing implications of holding ambivalent attitudes. Further exploration of the concept of ambivalence promises to provide us with further insights into broader questions about the structure of attitudes.


REFERENCES

Aiken, L. S., & West, S. G. (1991). Multiple Regression: Testing and Interpreting Interactions. Newbury Park, CA: Sage.

Ainslie, G. (1986). Beyond microeconomics: Conflict among interests in a multiple self as a determinant of value. In J. Elster (Ed.), The Multiple Self (pp. 133-75). Cambridge: Cambridge University Press.

Ainslie, G. (1992). Picoeconomics: The Strategic Interaction of Successive Motivational States within the Person. Cambridge: Cambridge University Press.

Ajzen, 1. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179-21 1.

Ajzen, I. (1996). The directive influence of attitudes on behavior. In P. M. Gollwitzer & J. A. Bargh (Eds), Psychology of Action (pp. 3 8 5 4 3 ) . New York Guilford.

Allport, G. W. (1935). Attitudes. In C. Murchison (Ed.), Handbook of Social Psychol- ogy (pp. 798-844). Worcester, MA: Clark University Press.

Armitage, C. J., & Comer, M. (1999). Distinguishing perceptions of control from self- efficacy: Predicting consumption of a low fat diet using the theory of planned behavior. Journal of Applied Social Psychology, 29,72-90.

Armitage, C. J., & Comer, M. (2000). The effects of ambivalence on attitude stability and pliability, prediction of behavior, and information processing. Personality and Social Psychology Bulletin, 26, 143243.

Armitage, C. J., & Comer, M. (in press, a). Efficacy of the theory of planned behaviour: A meta-analytic review. British Journal of Social Psychology.

Armitage, C. J., & Conner, M. (in press, b). Reducing fat intake: Lessons from the theory of planned behaviour. In D.R. Rutter (Ed.), Changing Health Behaviour: Intervention and Research with Social Cognition Models. Buckingham: Open Uni- versity Press.

Bargh, J. A., Chaiken, S., Govender, R., & Pratto, F. (1992). The generality of the automatic attitude activation effect. Journal of Personality and Social Psychology,

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51,1173-82.

Bassili, J. N. (19%). Meta-judgmental versus operative indexes of psychological attributes: The case of measures of attitude strength. Journal of Personality and Social

Beardsworth, A. (1995). The management of food ambivalence: Erosion or reconstruc- tion? In D. Maurer & J. Sobal (Eds), Eating Agendas: Food and Nutrition as Social Problems (pp. 11743). New York: De Gruyter.

Bell, D. W., Esses, V. M., & Maio, G. R. (1996). The utility of open-ended measures to assess intergroup ambivalence. Canadian Journal of Behavioral Science, 28,12-18.

Biesanz, J. C., West, S. G., & Graziano, W. G. (1998). Moderators of self-other agree- ment: Reconsidering temporal stability in personality. Journal of Personality and

Breckler, S. J. (1994). A comparison of numerical indices for measuring attitudinal

Bromer, P. (1998). Ambivalent attitudes and information processing. Swbs Journal of

Cacioppo, J. T., Gardner, W. L., & Berntson, G. G. (1997). Beyond bipolar conceptualizations and measures: The case of attitudes and evaluative space. Personality and Social Psychology Review, 1,3-25.

Cacioppo, J. T., & Petty, R. E. (1982). The need for cognition. Journal of Personality and Social Psychology, 42,116-31.

62,893-912.

Psychology, 71,637-53.

Social Psychology, 75,467-77.

ambivalence. Educational and Psychological Measurement, 54,35065.



Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validity by the multitrait-multimethod matrix. Psychological Bulletin, 56,81-105.

Chaiken, S., Pomerantz, E. M., & Giner-Sorolla, R. (1995). Structural consistency and attitude strength. In R. E. Petty & J. A. Krosnick (Eds), Attitude Strength: Antece- dents and Consequences (pp. 387-412). Mahwah, NJ: Erlbaum.

Conner, M., & Armitage, C. J. (1998). Extending the theory of planned behavior: A review and avenues for further research. Journal of Applied Social Psychology, 28, 1430-64.

Comer, M., Bainbridge, J., & Halton, B. (1999a). Within-subjects correlations between bipolar and unipolar measures of attitude and intentions: Moderating effects of response latency (unpublished raw data). School of Psychology, University of Leeds, UK.

Comer, M., & Flesch, D. (1998). Attitudes, intentions and ambivalence about having casual sex. Unpublished raw data. School of Psychology, University of Leeds, UK.

Conner, M., & Flesch, D. (2001). Having casual sex: Additive and interactive effects of alcohol and condom availability on the determinants of intentions. Journal of Ap- plied Social Psychology, 31,89-112.

Comer, M., Povey, R., & Sparks, P. (1995). The role of attitudinal ambivalence in the theory of planned behaviour: A study of low fat diets (unpublished raw data). School of Psychology, University of Leeds, UK.

Comer, M., Povey, R., Sparks, P., James, R., & Shepherd, R. (1998a). Understanding dietary choice and dietary change: Contributions from social psychology. In A. Murcott (Ed.), The Nation’s Diet: The Social Science of Food Choice (pp. 43-56). London: Longman.

Comer, M., Povey, R., Sparks, P., James, R., & Shepherd, R. (in press). Moderating role of attitudinal ambivalence within the theory of planned behaviour. British Journal of Social Psychology.

Conner, M., Sheeran, P., Norman, P., & Armitage, C. J. (2000). Temporal stability as a moderator of relationships in the theory of planned behaviour. British Journal of Social Psychology, 39,469-93.

Comer, M., & Sherlock, K. (1997). Attitudinal ambivalence and the use of ecstasy (unpublished raw data). School of Psychology, University of Leeds, UK.

Conner, M., Sherlock, K., & Orbell, S. (1998b). Psychosocial determinants of ecstasy use in young people in the UK. British Journal of Health Psychology, 3,295-318.

Conner, M. T., & Sparks, P. (1996). The theory of planned behaviour and health behaviours. In M. Conner and P. Norman (Eds), Predicting Health Behaviour: Re- search and Practice with Social Cognition Models (pp. 121-62). Milton Keynes: Open University Press.

Comer, M., Sparks, P., & Armitage, C. (1999b). The moderating role of attitudinal ambivalence on attitude-behaviour relationships. Proceedings of the British Psycho- logical Society, 7( l), 58.

Comer, M., Sparks, P., Povey, R., James, R., & Shepherd, R. (1996). Attitude- intention-behaviour links: moderating role of attitudinal ambivalence. Paper presented at 11th General Meeting of the European Association of Experimental Social Psychology, Gmunden, Austria, 13-18 July 19%.

Conner, M., Sparks, P., Povey, R., James, R., & Shepherd, R. (1999~). Moderator effects of attitudinal ambivalence on attitude-behaviour relationships (manuscript submitted for publication).

Converse, P. E. (1995). Foreword. In R. E. Petty & J. A. Krosnick (Eds), Attitude Strength: Antecedenls and Consequences (pp. xi-xvii). Hillsdale, NJ: Erlbaum.

Doll, J., & Ajzen, I. (1992). Accessibility and stability of predictors in the theory of planned behavior. Journal of Personality and Social Psychology, 63,754-65.


Eagly, A. H., & Chaiken, S. (1993). The Psychology of Attitudes. Fort Worth, Philadelphia, PA: Harcourt Brace Jovanovich.

Eagly, A. H., & Chaiken, S. (1995). Attitude strength, attitude structure, and resistance to change. In R. E. Petty & J. A. Krosnick (Eds), Attitude Strength: Antecedents and Consequences (pp. 413-32). Mahwah, NJ: Erlbaum.

Eagly, A. H., & Chaiken, S. (1998). Attitude structure and function. In D.T. Gilbert, S. T. Fiske, & G. Lindzey (Eds), The Handbook of Social Psychology, 4th Edn (Vol. 1, pp. 269-322). Boston, MA: McGraw-Hill.

Emmons, R. A. (1996). Strivings and feeling: Personal goals and subjective well-being. In P. M. Gollwitzer & J. A. Bargh (Eds), The Psychology of Action: Linking Cogni- tion and Motivation to Behavior. (pp. 313-37). New York: Guilford.

Erber, M. W., Hodges, S. D., & Wilson, T. D. (1995). Attitude strength, attitude stability, and the effects of analyzing reasons. In R. E. Petty & J. A. Krosnick (Eds), Attitude Strength: Antecedents and Consequences (pp. 433-54). Mahwah, NJ: Erlbaum.

Fazio, R. H. (1986). How do attitudes guide behavior? In R. M. Sorrentino & E. T. Higgins (Eds), Handbook of Motivation and Cognition: Foundations of Social Be- havior (Vol. 1, pp. 204-43). New York: Guilford.

Fazio, R. H. (1995). Attitudes as object-evaluation associations: Determinants, consequences, and correlates of attitude accessibility. In R. E. Petty, & J. A. Krosnick (Eds), Attitude Strength: Antecedents and Consequences (pp. 247-82). Mahwah, NJ: Erlbaum.

Fishbein, M., & Ajzen, I. (1975). Belief, Attitude, Intention and Behavior: An Introduc- tion to Theory and Research. Reading, MA: Addison-Wesley.

Freud, S. (1962). The Ego and the Id. London: Hogarth. Gardner, P. L. (1987). Measuring ambivalence to science. Journal of Research in

Science Teaching, 24,241-7. Glick, P., & Fiske, S. T. (1996). The ambivalent sexism inventory: Differentiating

hostile and benevolent sexism. Journal of Personality and Social Psychology, 70,

Godin, G., & Kok, G. (1996). The theory of planned behavior: A review of its applica- tions to health-related behaviors. American Journal of Health Promotion, 11,87-98.

Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition: Attitudes, self- esteem, and stereotypes. Psychological Review, 102,627.

Grogan, S. , Bell, R., & Comer, M. (1997). Eating sweet snack foods: Gender differences in attitudes and behaviour. Appetite, 28,19-31.

Haddock, G., & Zanna, M. P. (1998). On the use of open-ended measures to assess attitudinal components. British J o u m l of Social Psychology, 37,12949.

Jonas, K., Bromer, P., & Diehl, M. (2000). Attitudinal ambivalence. European Review of Social Psychology, 11,3574,

Jonas, K., Diehl, M., & Bromer, P. (1997). Effects of attitudinal ambivalence on information processing and attitude-intention consistency. Journal of Experimental Social Psychology, 33,190-210.

Joreskog, K. G., & Sorbom, D. (1993). LISREL 8 User’s Reference Guide. Chicago, IL: Scientific Software.

Kallgren, C. A., & Wood, W. (1986). Access to attitude-relevant information in memory as a determinant of attitude-behavior consistency. Journal of Experimental Social Psychology, 22,328-38.

Kaplan, K. J. (1972). On the ambivalence-indifference problem in attitude theory and measurement: A suggested modification of the semantic differential technique. Psy- chological Bulletin, 77,361-72.

Katz, I., & Hass, R. G. (1988). Racial ambivalence and American value conflict: Correlational and priming studies of dual cognitive structures. Journal of Personality and Social Psychology, 55,893-905.

491-512.


Krosnick, J. A., Boninger, D. S., Chuang, Y. C., Berent, M. K., & Carnot, C. G. (1993). Attitude strength: One construct or many related constructs? Journal of Personality and Social Psychology, 65,1132-51.

Krosnick, J. A., & Petty, R. E. (1995). Attitude strength: An overview. In R. E. Petty & J. A. Krosnick (Eds), Attitude Strength: Antecedents and Consequences (pp. 1-24). Mahwah, NJ: Erlbaum.

Kruglanski, A. W., Webster, D. M., & Klem, A. (1993). Motivated resistance and openness to persuasion in the presence or absence of prior information. Journal of Personality and Social Psychology, 65, 861-76.

Lavine, H., Thomsen, C. J., Zanna, M. P., & Borgida, E. (1998). On the primacy of affect in the determination of attitudes and behavior: The moderating role of affectivexognitive ambivalence. Journal of Experimental Social Psychology, 34, 398-421.

Lipkus, I. M., Green, J. D., Feaganes, J. R., & Sedikides, C. (in press). The relationship between attitudinal ambivalence and desire to quit smoking among college smokers. Journal of Applied Social Psychology.

MacDonald, T. K., & Zanna, M. P. (1998). Cross-dimension ambivalence toward social groups: Can ambivalence affect intentions to hire feminists? Personality and Social Psychology Bulletin, 24,42741.

Maio, G. R., Bell, D. W., & Esses, V. M. (19%). Ambivalence and persuasion: The processing of messages about immigrant groups. Journal of Experimental Social

Maio, G. R., Esses, V. M., & Bell, D. W. (2000). Examining the conflict between components of attitudes: Ambivalence and inconsistency are distinct concepts. Ca- nadian Journal of Behavioral Science, 32,7143.

McGregor, I., Newby-Clark, I. R., & Zanna, M. P. (1999). “Remembering” dissonance: Simultaneous accessibility of inconsistent cognitive elements moderates epistemic discomfort. In E. Harmon-Jones & J. Mills (Eds), Cognitive Dissonance: Progress on a Pivotal Theory in Social Psychology (pp. 325-53). Washington, DC: American Psychological Association.

Merton, R. (1976). Sociological Ambivalence and Other Essays. New York: Free Press. Moore, M. (1973). Ambivalence in attitude measurement. Educational and Psycholog-

ical Measurement, 33,481-3. Moore, M. (1980). Validation of attitude toward any practice scale through use of am-

bivalence as a moderator. Educational and Psychological Measurement, 40,205-8. Norman, P., & Smith, L. (1995). The theory of planned behaviour and exercise: An

investigation into the role of prior behaviour, behavioural intentions and attitude variability. European Journal of Social Psychology, 25,403-15.

Ortony, A., Clore, G. L., & Collins, A. (1988). The Cognitive Structure of Emotions. Cambridge: Cambridge University Press.

Otnes, C., Lowrey, T. M., & Shrum, L. J. (1997). Toward an understanding of attitudinal ambivalence. Journal of Consumer Research, 24,80-93.

Petty, R. E., & Krosnick, J. A. (Eds) (1995). Attitude Strength: Antecedents and Con- sequences. Mahwah, NJ: Erlbaum.

Pomerantz, E. V., Chaiken, S., & Tordesillas, R. S. (1995). Attitude strength and resistance processes. Journal of Personality and Social Psychology, 69,408-19.

Pratkanis, A. R., Breckler, S. J., & Greenwald, A. G. (1989). Attitude Structure and Function. Hillsdale, NJ: Erlbaum.

Priester, J. R., & Petty, R. E. (1996). The gradual threshold model of ambivalence: Relating the positive and negative bases of attitudes to subjective ambivalence. Journal of Personality and Social Psychology, 71,43149.

Priester, J. R., & Petty, R. E. (2000). The relationship of ambivalence to behavior: A reconceptualization of the antecedents and consequences of evaluative tension.



Paper presented at the European Association of Experimental Social Psychology Small Group Meeting on the Psychology of Attitudes, Newtown, Wales, 26-28 July 2000.

Priester, J. R., & Petty, R. E. (2001). Extending the bases of subjective attitudinal ambivalence: Interpersonal and intrapersonal antecedents of evaluative tension. Journal of Personality and Social Psychology, 80,19-34.

Schwartz, S. H. (1978). Temporal instability as a moderator of the attitude-behavior relationship. Journal of Personality and Social Psychology, 36,715-24.

Scott, W. A. (1966). Measures of cognitive structure. Multivariate Behavioral Research,

Scott, W. A. (1968). Attitude measurement. In G. Lindzey & E. Aronson (Eds), The Handbook of Social Psychology (Vol. 2, pp. 204-73). Reading, MA: Addison- Wesley.

Scott, W. A. (1%9). Structure of natural cognitions. Journal of Personality and Social

Sheeran, P., Orbell, S., & Trafimow, D. (1999). Does the temporal stability of behavioral intentions moderate intention-behavior and past behavior-future behavior relations? Personality and Social Psychology Bulletin, 25,721-30.

Sheppard, B. H., Hartwick, J., & Warshaw, P. R. (1988). The theory of reasoned action: A meta-analysis of past research with recommendations for modifications and future research. Journal of Consumer Research, 15,32543.

Snyder, M., & Gangestad, S. (1986). On the nature of self-monitoring: Matters of assessment, matters of validity. Journal of Personality and Social Psychology, 51,

Sparks, P. (1994). Attitudes towards food: Applying, assessing and extending the “theory of planned behaviour”. In D. R. Rutter & L. Quine (Eds) Social Psychology and Health: European Perspectives (pp. 25-46). Aldershot: Avebury.

Sparks, P., Conner, M., James, R., Povey, R., & Shepherd, R. (1997a). Taking ambivalence seriously in attitude research. Proceedings of the British Psychological Society, 5(1), 47.

Sparks, P., Conner, M., James, R., Shepherd, R., & Povey, R. (1995). Ambivalence about health-related behaviours: Relationships to personality measures (unpublished raw data). Institute of Food Research, Reading, UK.

Sparks, P., Conner, M., James, R., Shepherd, R., & Povey, R. (2001). Ambivalence about health-related behaviours: An exploration in the domain of food choice. British Journal of Health Psychology, 6,53-68.

Sparks, P., Hedderley, D., & Shepherd, R. (1992). An investigation into the relationship between perceived control, attitude variability and the consumption of two common foods. European Journal of Social Psychology, 22,55-71.

Sparks, P., James, R., Conner, M., Shepherd, R., & Povey, R. (1997b). Ambivalence about food: Implications for attitude theory and research. Appetite, 29,259.

Stocker, M. (1990). Plural and Conflicting Values. Oxford: Clarendon Press. Tesser, A. (1978). Self-generated attitude change. In L. Berkowitz (Ed.), Advances in

Experimental Social Psychology, 11,289-338. Thompson, M. M., Naccarato, M. E., & Parker, K. (1989). Measuring cognitive needs:

The development and validation of the personal need for structure (PNS) and personal fear of invalidity (PFI) measures. Paper presented at the annual meeting of the Canadian Psychological Association, Halifax, NS.

Thompson, M. M., & Zanna, M. P. (1995). The conflicted individual: Personality-based and domain specific antecedents of ambivalent social attitudes. Journal of Person- ality, 63,259-88.

Thompson, M. M., Zanna, M. P., & Griffin, D. W. (1995). Let’s not be indifferent about (attitudinal) ambivalence. In R. E. Petty & J. A. Krosnick (Eds), Attitude Strength: Antecedents and Consequences (pp. 361-86). Mahwah, NJ: Erlbaum.

1,391-5.


125-39.


Tourangeau, R., Rasinski, K. A., Bradburn, N., & D’Andrade, R. (1989). Belief awess- ibility and context effects in attitude measurement. Journal of Experimental Social

Van den Putte, B. (1991). 20 years of the theory of reasoned action of Fishbein and Ajzen: A meta-analysis (unpublished manuscript). University of Amsterdam.

Van Harreveld, F., Van der Pligt, J., & De Vries, N. K. (1999). Belief importance and ambivalence in expectancy value models (manuscript submitted for publication).

Van der Pligt, J., De Vries, N. K., Manstead, A. S. R., & Van Harreveld, F. (2000). The importance of being selective: Weighing the role of attribute importance in attitudinal judgment. In M.P. Zanna (Ed.), Advances in Experimental Social Psychology, 32,

Verplanken, B., Hofstee, G., & Janssen, H. J. W. (1998). Accessibility of affective versus cognitive components of attitudes. European Journal of Social Psychology,

Waterman, M., & Conner, M. (1999). An open-ended exploration of attitude objects which generate neutral, positive, negative, or both positive and negative evaluations (unpublished raw data). School of Psychology, University of Leeds, UK.

Wegener, D. T., Downing, J., Krosnick, J. A., & Petty, R. E. (1995). Measures and manipulations of strength-related properties of attitudes: Current practice and future directions. In R. E. Petty, & J. A. Krosnick (Eds), Attitude Strength: Antecedents and Consequences (pp. 456-87). Hillsdale, NJ: Erlbaum.

Wilson, T. D., Lindsey, S., & Schooler, T. Y. (2000). A model of dual attitudes. Psychological Review, lCn, 101-26.

Wilson, T. D., & Hodges, S. D. (1992). Attitudes as temporary constructions. In A. Tesser & L. Martin (Eds), The Construction of Social Judgment (pp. 37-65). Hills- dale, NJ: Erlbaum.

Zaller, J., & Feldman, S. (1992). A simple theory of the survey response: Answering questions versus revealing preferences. American Journal of Political Science, 36, 579-616.


135-200.

28,22-35.

Ambivalence and Attitudes - OSF

Documents