The Effect of Preference Fluency on Consumer Decision Making

We propose that consumer preferences are often systematically influenced by preference fluency, i.e., the subjective feeling that forming a preference is easy or difficult. In five studies, we manipulated the fluency of preference formation by presenting descriptions in an easy or difficult to read font (Studies 1 and 2) or by asking participants to think of few vs. many reasons for their choice (Studies 3 to 5). As predicted, subjective experiences of difficulty increased choice deferral (Studies 1 to 3) and the selection of a compromise option (Studies 4 and 5), unless consumers were induced to attribute the experience to an unrelated cause. Unlike studies of decision conflict, these effects were obtained without changing the attributes of the alternatives or the composition of the choice sets. We discuss the theoretical and practical implications of the results. Understanding the factors that determine which options consumers choose and whether they make rather than defer purchase decisions is critical for the development of marketing strategies. A major contribution of behavioral decision research has been to establish the notion of uncertain preferences, the idea that consumer preferences are not well defined but rather constructed in the process of making a choice. This constructive viewpoint suggests that different tasks and contexts highlight different aspects of the options, focusing consumers on different considerations that lead to seemingly inconsistent decisions (Bettman, Luce, and Payne 1998). Past work has focused on the effect of decision conflict and preference uncertainty on consumer choice by manipulating the content of the choice or by manipulating which content is the focus of attention (e.g., Dhar and Simonson 1992; Nowlis and Simonson 1997; Shafir 1993). Recent judgment research indicates, however, that there is more to thinking than thought content. As reviewed below, the impact of thought content can be qualified by the person’s subjective experiences during the processing of information (for a review see Schwarz 2004). To date, research on constructed preferences has highlighted contexts that change the content of thoughts during a decision process and has largely ignored the effects of experiences arising from those thoughts. The present research begins to fill this gap. We examine the effect of changes in experiential states during the decision process. We term the subjective feeling of ease or difficulty experienced while constructing preference preference fluency. In the present research, we focus in on the effects of preference fluency on choosing the deferral and compromise option (Dhar and Simonson 2003). Prior research has Preference Fluency -3 demonstrated that purchase deferral and choice of the compromise option are both consequences of difficult decisions (Dhar 1997; Kivetz, Netzer, and Srinivasan 2004; Simonson 1989; Tversky and Shafir 1992). Building on recent research into the interplay of experiential and declarative information, we show that the perception of subjective preference formation experience bears on the decision to defer choice and to compromise, even when this experience is due to extraneous variables, unrelated to the choice alternatives. A major theoretical implication of our research is that in situations where preference uncertainty arises from the difficulty in trading off competing characteristics of the choice alternatives, its influence on decision behavior is potentially mediated by consumers’ subjective experience of indecision and conflict (e.g., Dhar 1997; Luce 1998). If so, it is likely that any other variable that elicits the experience of difficulty while contemplating a choice will have similar effects. The present studies hold objective tradeoff difficulty constant by presenting participants in all conditions with the same choice alternatives. To manipulate participants’ subjective experience of the difficulty of preference construction, we manipulated extraneous variables like the readability of the print font (Studies 1 and 2) and the number of reasons requested for a choice (Studies 3, 4, and 5). Throughout, we observed that variables that increase the difficulty of preference construction increase indecision, in ways that parallel the effects of tradeoff difficulty. In the remainder of the paper, we review key findings bearing on the role of experiential information in judgment. Next, we introduce the concept of preference fluency and report five studies in which preference fluency was manipulated through variables extraneous to the choice alternatives. We discuss the implications of this research with respect to the role of preference fluency in consumer decision making and managerial tactics that might promote preference fluency to influence purchase decisions. Experiential Information in Judgment and Decision Making The emerging consensus among decision researchers is that consumers often do not have well-defined preferences that can be retrieved, and that they construct their preferences when faced with the need to make decisions (for a review, see, e.g., Bettman, Luce, and Payne, 1998; Slovic 1995). For example, Levin and Gaeth (1988) showed that consumers who tasted ground beef that was described as being “75% lean” rated it more favorably than consumers who were Preference Fluency -4 told that the (same) ground beef had “25% fat.” Also, Simonson and Tversky (1992) demonstrated that consumers are more likely to exchange $6 for an elegant Cross pen when they also have the option of exchanging $6 for a less attractive pen. Past research has often focused on decision conflict and preference uncertainty resulting from the attribute values of the options and the need to choose among several about equally attractive alternatives. This focus on the impact of the content of information and attribute value tradeoffs has been complemented by increased attention to experiential information in the form of moods and emotions, first in judgment research (for a review see Schwarz and Clore 1996) and subsequently in decision making (e.g., Luce, Bettman, and Payne 1997). As research in social cognition indicates, however, experiential information is not limited to ambient affective states like moods and emotions, but includes cognitive experiences that accompany the reasoning process (Clore 1992). The notion of preference fluency is rooted in the dynamics of the reasoning process rather than in the content that is accessible. Broadly speaking, any variable that influences the cognitive experience of choosing is likely to influence the willingness to choose as well as which option is chosen. For example, if people find it easy or difficult to retrieve information from memory, or to generate relevant arguments, these accessibility experiences will be informative in their own right and qualify the implications of thought content ( for reviews see Schwarz 1998, 2004). Similarly, people may find it easy or difficult to process new information and this experience of processing fluency again affects the conclusions drawn from that new information (for a review see Winkielman, Schwarz, Fazendeiro, and Reber 2003). To date, these related cognitive experiences, pertaining to the fluency of recall and thought generation and the fluency of processing new information, have received considerable attention in the judgment literature, but their effects on preference construction have not been studied. We first review findings from the judgment literature and subsequently address implications for choice. Accessibility Experiences: Ease of Recall and Thought Generation Challenging the traditional notion that judgments are based solely on what comes to mind, Schwarz et al. (1991, Experiment 1) observed that the implications of accessible thought content are qualified by the ease or difficulty with which a given thought can be brought to mind. In their studies, participants rated themselves as less assertive after recalling 12 examples of their own assertive behavior (experienced as difficult) than after recalling merely 6 examples (experienced Preference Fluency -5 as easy). Apparently, they concluded from the difficulty of recalling 12 examples that they cannot be that assertive, or else recalling 12 examples would not be so difficult. Supporting this interpretation, the observed pattern reversed when the informational value of the subjective experience was undermined through a misattribution manipulation. That is, when participants could attribute the experienced difficulty to the distracting effect of background music, they reported higher assertiveness after recalling 12 rather than 6 examples (Schwarz et al 1991, Experiment 3). Subsequent research replicated this basic pattern across many content domains (for reviews see Schwarz 1998, 2004). Which inferences people draw from accessibility experiences depends on which naïve theory of the mind they bring to bear on them, which is context dependent. One set of naïve theories pertains to the relationship between external variables, like the frequency of events in the world, and the accessibility of thought content. Consistent with Tversky and Kahneman’s (1973) availability heuristic, individuals assume that frequent events are easier to recall than rare events. Hence, ease of recall feeds into judgments of frequency (e.g., Arts and Dijksterhuis 1999; Wänke, Schwarz, and Bless 1995) as well as judgments for which perceived frequency serves as input, ranging from assessments of risk (e.g., Raghubir and Menon 1998; Rothman and Schwarz 1998) to judgments of personality (e.g., Schwarz et al. 1991). Similarly, people assume that it is easier to generate arguments for a proposition when there are many rather than few good reasons that support it. Accordingly, they are more persuaded when they have to generate few arguments rather than many, despite the fact that more arguments were brought to mind in the latter case (e.g., Wänke, Bless, and Biller 1996). Moreover, they report higher attitude strength on measures like attitude importance and certainty after generating a few rather than many arguments (e.g., Haddock, Rothman, Reber, and Schwarz 1999). Finally, turning to consumer preferences, Wänke, Bohner and Jurkowitsch (1997) observed that consumers were more likely to prefer a BMW over a Mercedes after they generated one rather than many advantages of a BMW. In all cases, the obtained judgments were consistent with the implications of thought content when recall was easy, but opposite to the implications of thought content when recall was difficult. All these studies provide evidence that subjective experiences accompanying thoughts can affect judgments above and beyond the effects of the content of those thoughts. This research also includes studies to rule out the possibility that thought content is responsible for the effects. Preference Fluency -6 For example, in Wänke et al.’s (1997) persuasion experiment, yoked participants merely read the arguments generated by others, thus depriving them of the difficulty experience associated with argument generation. As expected, these participants were more persuaded by many rather than few arguments, testifying to the quality of the arguments themselves. Similarly, participants turn to the content of recall when their subjective accessibility experience is discredited through (mis)attribution manipulations. These manipulations change their judgments of personality (e.g., Schwarz et al. 1991) or attitude strength (e.g., Haddock et al. 1999) to be more extreme after generating many rather than few examples or arguments, again indicating that the content of their thoughts is persuasive unless qualified by the experience of difficulty. Finally, other studies held the amount of recall constant and varied the experienced difficulty by inducing participants to contract the corrugator muscle while listing their thoughts (e.g., Sanna, Schwarz, and Small 2002; Stepper and Strack 1993). As expected, the effects of this bodily feedback manipulation, which conveys a sense of effort, paralleled the effects of listing many thoughts while holding the number of thoughts constant. In sum, experienced ease or difficulty of recall or thought generation is informative in its own right and qualifies the implications of thought content. Processing Fluency: The Ease of Processing Information The premise behind processing fluency is that any stimulus may be processed with differing degrees of speed, effort, and accuracy. In some cases, these processing differences arise from visual variables like figure-ground contrast, stimulus clarity, presentation duration or the amount of previous exposure. These variables influence perceptual fluency (e.g., Jacoby, Kelley, and Dywan 1989). In other cases, processing differences arise from variables like the consistency between the stimulus and its context or the availability of appropriate mental concepts for stimulus classification. These variables influence conceptual fluency (e.g., Whittlesea 1993). While this distinction is relevant to some theoretical issues (Kelley and Rhodes 2002), perceptual and conceptual fluency have parallel effects for the issues of interest to the present research (for a discussion see Winkielman et al. 2003a,b). Hence we prefer the more generic term of processing fluency. As is the case for accessibility experiences, the inferences that people draw from processing fluency are highly malleable. Because we have only one window on our subjective experiences, people are vulnerable to misreading the fluency resulting from one of many sources Preference Fluency -7 (e.g., repeated exposure) as being due to a different source (e.g., visual clarity). Thus, people who have seen the stimulus before, and hence find it easier to process, may infer that the current presentation lasted longer, or had higher clarity, than people not previously exposed to the stimulus (e.g., Witherspoon and Allan, 1985; Whittlesea, Jacoby and Girard, 1990). Conversely, visual manipulations of fluency may result in erroneous inferences of previous exposure, often referred to as “illusions of familiarity” (Whittlesea 1993). One effect of processing fluency that is particularly relevant to decision making is its influence on judgments of truth. When the objective truth of a statement is difficult to evaluate, people often draw on social consensus information to arrive at a judgment – after all, if many believe it, there’s probably something to it (Festinger, 1954). Hence they are more likely to accept a statement as “true” when it seems familiar rather than novel. The statement’s perceived familiarity, however, is a function of how fluently it can be processed and variables like exposure frequency (e.g., Begg, Anas, and Farinacci 1992; Hawkins and Hoch 1992) or figureground contrast (Reber and Schwarz 1999) reliably increase the likelihood that a given statement is accepted as true. In fact, merely presenting the same statement in a color that makes it easy or difficult to read against the background influences truth judgments (Reber and Schwarz 1999). The fluency of processing can also have an impact on evaluative judgments: A given target is evaluated more positively, the more easily it can be processed. Thus, any variable that facilitates fluent perception is likely to increase liking, from figure-ground contrast (e.g., Reber, Winkielman, and Schwarz, 1998) and presentation time (Reber et al., 1998) to previous exposure (as known since Zajonc's, 1968, demonstration of the mere exposure effect). For example, Reber et al (1998) observed that participants liked a given picture more when it was preceded by a subliminally presented matching rather than mismatching, contour. Winkielman and Fazendeiro (reported in Winkielman, Schwarz, Reber, and Fazendeiro 2003b) obtained parallel findings with a conceptual fluency manipulation. In their studies, participants saw unambiguous pictures of common objects (e.g., a picture of a lock), preceded by a word. Participants reported liking the pictures more when they were preceded by conceptually related primes (e.g. “lock” or “key”) than unrelated primes (e.g. “snow”). Liking of the picture was a function of the processing fluency resulting from the primes. Lee and Labroo (in press) obtained similar findings in the consumer domain. They found, for example, that consumers reported more positive attitudes toward ketchup when they were previously exposed to a closely related product (mayonnaise) Preference Fluency -8 rather than an unrelated one (vitamins). Note that these effects of processing fluency cannot be traced to differential descriptive information about the unambiguous targets. Instead, the available evidence suggests that fluency itself is hedonically marked and experienced as positive. Consistent with this assumption, Winkielman and Cacioppo (2001) observed that high fluency is associated with increased activity over the region of the zygomaticus major ("smiling muscle"), which is indicative of a positive affective response. In sum, the reviewed findings illustrate that judgments are not necessarily based on descriptive information about the target. Instead, the fluency with which information about the target can be processed or the ease of thought generation or recall is informative in its own right and feeds into a variety of judgments, including judgments of truth and liking. Next, we turn to the implications of these phenomena for construction of uncertain preferences. Preference Fluency Past research has examined the consequences of difficulty of deciding arising from uncertain preferences (for a review, see Bettman et al. 1998). This research typically explained choices in terms of the characteristics of the options or in terms of reasons that are used to justify the choice (e.g., Kivetz 1999; Shafir, Simonson, and Tversky 1993; Simonson 1989; Simonson and Nowlis 2000). Difficulty was often manipulated by changing the relative attractiveness of the options or by changing the choice task. While the choice options and the task can influence the difficulty of the choice, the subjective experience during the decision process can also influence the perceived difficulty of the choice. In the judgment research reviewed above, individuals presumed that experiences that occurring while thinking about a judgment were related to that judgment and incorporated those experiences into their ratings. This is consistent with prior research on the “aboutness” principle (Higgins 1998). This research showed that when thoughts or feelings come to mind while a particular target is being considered, those thoughts or feelings are assumed to be relevant to the target or they would not come to mind at that moment. In preference construction, we believe that when the subjective experience of difficulty accompanies the decision making process, this experience will induce an inference that the choice itself is difficult. As a result, the experience Preference Fluency -9 will mimic the effects of difficulty found in previous research that manipulated the content of the choice. Research on the determinants of choice deferral shows that the tendency to not choose increases when the choice set offers several attractive alternatives but none that can easily be justified as the best (Dhar 1997, Tversky and Shafir 1992). Such decisions involve conflict and can be emotionally stressful (e.g., Luce 1998), because choosing one alternative implies that other alternatives and their unique attractive features should be foregone (e.g., Festinger 1962). Other sources of decision difficulty might also lead to choice deferral. If individuals experience low preference fluency when thinking about a choice, they may attribute this subjective feeling about preference formation to the difficulty of the decision. Building on prior work showing that decision difficulty promotes choice deferral, we propose that the likelihood of making a choice increases with preference fluency. Preference fluency can also influence the relative preference among options in a choice set. In particular, some options are selected not because they are more preferred but as a way to resolve a difficult decision (e.g., Dhar and Simonson 2003). As illustrated in Figure 1, the compromise effect occurs if the choice share of one option, b, relative to another alternative, c, is enhanced when a third option, a, is added to the choice set making b a “compromise” (middle) option. Consistent with earlier process data (Simonson 1989), many consumers who select a compromise option find the decision to be difficult and view the middle option as a way to resolve this difficulty. Thus, we propose that when the fluency of processing decreases, options whose status as conflict resolution mechanisms is transparent will gain share. In particular, we hypothesize that the compromise effect will be enhanced when preference fluency decreases. These predictions emphasize the role of experiential information in preference construction. Many variables are likely to influence preference fluency. At one extreme are variables that are fully extraneous to the features of the choice alternatives themselves. For example, we expect that consumers are more likely to make a choice when it is easy rather than difficult to read new information about the choice alternatives (e.g., due to different print fonts). At the other extreme are variables that are inherent to the choice alternatives themselves (e.g. equally attractive options). These latter variables were the focus of previous research that manipulated the difficulty of decisions by changing the content of the choice set or the attribute tradeoff difficulty. In between these extremes are experiences that may or may not depend on the nature Preference Fluency -10 of the choice alternatives. For example, listing reasons for one’s preference can be experienced as difficult because the choice alternatives are similar in atttractiveness or because one has been asked to list many of them. As this discussion indicates, we propose fluency of preference formation as an integrative concept that can account for cases of indecision resulting from features of the choice alternatives as well as cases of indecision resulting from extraneous variables. The present studies focus on these latter cases. We test our proposition that increased preference fluency decreases choice deferral in two ways. In studies 1 and 2, we manipulate fluency by presenting the same choice alternatives either in an easy or in a difficult to read font. In study 3, we ask participants to think either of a few (easy) or many (difficult) reasons for their choice. Extending the exploration of processing fluency from deferral of choice to the alternative chosen, studies 4 and 5 address the influence of preference fluency on the size of the compromise effect. In these studies, we manipulate fluency by asking for a few versus many reasons for the choice. Throughout, our results indicate that the experienced fluency of preference formation has a profound impact on choice behavior, unless the informational value of the experience is drawn into question. Importantly, these effects are obtained in the absence of any objective changes in the attributes of the individual choice alternatives, the composition of the choice sets, or the response options. Study 1: Print Fonts and Deferral of Choice Our first study tests the assumption that fluency experiences deriving from extraneous variables can affect the subjective difficulty of choice and in turn affect deferral. In this study, all participants received identical descriptions of two sets of choice objects (digital cameras and microwave ovens), except that the descriptions were printed in a font that was easy or difficult to read. We predicted that participants would misinterpret the experienced difficulty of processing information about the alternatives as being due to the difficulty of the decision, rather than the difficulty of reading the information. Accordingly, respondents should be more likely to defer choice when the information is presented in a difficult to read rather than standard font. Method As part of a large questionnaire completed for course credit, 99 undergraduate students at a large northeastern business school were presented with two hypothetical choice problems, one choice of digital cameras and one choice of microwave ovens. Each choice problem provided Preference Fluency -11 descriptions of two options. Like many real consumer purchase situations, subjects could either choose one of the two options presented or defer choice and continue looking for other options. Subjective difficulty was manipulated by presenting the choices to half of the participants with the descriptions presented in embossed italicized gray font (see example in Figure 1). This font has been shown to be fairly difficult to read, though it can be read accurately with some effort (Epley and Norwich 2004). In a pretest, 34 undergraduates rated the materials used in this study on a 9-point scale ranging from very easy to very difficult to read. As expected the materials shown in figure 1 were rated as significantly more difficult to read than the same materials presented in a standard font (Ms = 3.53 and 4.88, t(32) = 2.75, p < .01).