How do statistical and narrative evidence affect persuasion?: the role of evidentiary features.

• Author: Han, Bing
• Position: Report

Persuasion has long been a focus for communication scholars, and message variables, such as order of arguments and message sidedness, have been investigated to determine the type of message that is most persuasive. More specifically, the persuasive effect of evidence has been extensively explored (e.g., Kellermann, 1980; McCroskey, 1969; Reinard, 1988; Stiff, 1986).

Persuasive messages typically offer evidence to produce attitude change (McCroskey, 1969). McCroskey (1969) identified two kinds of communication that function as evidence. Evidence consists of "factual statements originating from a source other than the speaker ... and opinions of persons other than the speaker that are offered in support of the speaker's claims" (McCroskey, 1969, p. 170). Evidence generally has been found to increase a message's persuasiveness as compared to no-evidence conditions (e.g., Hample, 1977, 1979; Levasseur & Dean, 1996; McCroskey, 1969; Nadler, 1983; Reinard, 1988; Stiff, 1986). However, most studies have ignored the message features that contribute to the persuasiveness of different evidence types. The current study investigates the specific evidentiary features that differentiate the persuasiveness of statistical and narrative messages.

Statistical evidence refers to evidence provided by quantitative or numerical information, whereas narrative evidence refers to anecdotal or personal evidence such as that provided by interviews, exemplars, stories, testimonials, and opinions (Kahneman & Tversky, 1973). Both statistical and narrative evidence have been found to be more persuasive than no-evidence control messages (e.g., Kazoleas, 1993; Limon & Kazoleas, 2004).

Many studies comparing the persuasiveness of different types of evidence, especially statistical and narrative evidence, have been inconclusive. Some studies have shown that the difference between the persuasiveness of these two types of evidence was not significant (e.g., Fisher as cited in Reinard, 1988; Kazoleas, 1993; Limon & Kazoleas, 2004; Nadler, 1983; Ryland, 1972). However, there have been studies that have shown the persuasive superiority of one type of evidence over the other. A majority of studies comparing the persuasiveness of statistical versus narrative evidence has reported that narrative evidence was more persuasive than statistical evidence (e.g., Borgida & Nisbett, 1977; Kahneman & Tversky, 1973; Taylor & Thompson, 1982). There is also research that has shown the opposite, that statistical evidence was more persuasive than narrative evidence (e.g., Baesler & Burgoon, 1994).

A meta-analysis comparing the persuasiveness of narrative and statistical evidence (Allen & Preiss, 1997) indicated that, in the 15 studies investigated, statistical evidence was more persuasive than narrative evidence, although the effect size for this difference was quite small. Another meta-analysis examining 23 studies (Reinhart, 2006), however, produced no significant differences in persuasiveness between the two message types when all dependent measures were compared together. Reinhart (2006) also found that when an attitude measure was compared independently of other outcome variables, narrative messages were slightly but significantly more persuasive than statistical messages. Allen et al. (2000) tested the persuasiveness of a combination of narrative and statistical evidence and found that a message combining narrative and statistical evidence was more persuasive than a message using statistical evidence alone, which was more persuasive than one with narrative evidence alone, which, in turn, was more persuasive than a message without any evidence. However, they noted that it was unclear why this ordering occurred and how message receivers evaluated these types of evidence.

The inconclusive results regarding the persuasive difference between these two evidence types may reflect a lack of control for some message features (such as message length and readability) or a failure to experimentally investigate evidentiary features that may differentiate one evidence type from another (such as amount of evidence and perceived vividness). The current study attempts to make the message features of statistical and narrative evidence that are not of theoretical interest as equal as possible so that the critical evidentiary features that may differentiate the persuasiveness of the two evidence types can be evaluated. A framework is proposed that identifies the specific evidentiary features of each of the two evidence types that contribute to its persuasiveness. The simple question as to which type of evidence, statistical or narrative, is more persuasive fails to elucidate the cognitive processes by which these messages persuade (e.g., Boster et al., 2000), whereas understanding how the persuasiveness of these message types are associated with particular evidentiary features leads to better theory as well as to an enhanced ability to create messages that are persuasive.

Narrative and statistical messages may bring about persuasion through different processes (e.g., Feeley, Marshall, & Reinhart, 2006; Kopfman, Smith, Ah Yun, & Hodges, 1998; Limon & Kazoleas, 2004; Lindsey & Ah Yun, 2003). This process-oriented research has mainly focused on the different responses to narrative and statistical messages. For example, Kopfman et al. (1998) studied organ donation and found that statistical messages scored higher on all cognitive measures, whereas narrative messages scored higher on all affective measures. However, few studies have directly focused on message features and how they may interact with message type. A message feature may enhance the persuasiveness of one type of evidence more than it increases that of the other, or it may enhance the persuasiveness of one type but undermine that of the other. The current study proposes that amount of evidence boosts the persuasiveness of statistical messages, whereas perceived vividness enhances the persuasiveness of narrative messages.

NARRATIVE AND STATISTICAL EVIDENCE

Narratives and Vividness of Evidence

The speculation regarding the effectiveness of narrative evidence has a long history. Schank and Abelson (1995) claimed that the human memory system stores knowledge as narratives. As Baesler and Burgoon (1994) indicated, story structures, such as schemata (Bartlett, 1932; Bower, 1976) and scripts (Mandler, 1984; Martin, 1982), help store and retrieve ideas (Myrsiades, 1987). Various theoretical frameworks have been proposed to explain how narratives become persuasive (e.g., Campbell & Babrow, 2004; Green & Brock, 2000, 2002; Oatley, 2002). Oatley (2002) postulated that the affective aspect of narrative made it an effective form of persuasion; more specifically, narratives were thought to persuade because they stimulate and evoke emotion-laden memories. Green and Brock (2000, 2002), in their Transportation-Imagery Model of narrative persuasion, focused on evoked imagery's ability to induce attitude and belief change because this imagery enables people to imagine themselves in a narrative world. However, this line of research focused on the responses to narratives but did not examine the features of these narratives. The hypothesized responses to narrative messages (i.e., emotion-laden memories and evoked imagery) may be induced more by vivid language than by a dry narrative.

The claim that vivid language is more effective than dry information has been widely discussed and investigated. Vividness may motivate message recipients to think about the causal relevance of message information (Taylor & Thompson, 1982), render information more available (Kahneman & Tversky, 1973), induce specific positive or negative emotions, elicit attention and interest, or produce memorable images (Fiske & Taylor, 1991; Nisbett & Ross, 1980). Nisbett and Ross (1980) suggested that it is the vividness of narrative information that renders it more persuasive as compared to pallid statistical data. Some researchers (e.g., Borgida & Nisbett, 1977; Nisbett, Borgida, Crandall, & Reed, 1976) have shown that a handful of vivid stories can be more effective than comprehensive statistical summaries.

There is no consistent support for the claim that vivid information is more effective than pallid information (Collins, Taylor, Wood, & Thompson, 1988; Taylor & Thompson, 1982). Vividness seems to boost the persuasiveness of only some types of information. Furthermore, vividness can become distracting and thus inhibit message persuasiveness (Frey & Eagly, 1993). Baesler and Burgoon (1994) explicitly examined the vividness effect on evidence type; their study was inconclusive regarding the ability of vividness to enhance the persuasive influence of narratives above that of statistics. In addition, they did not find a significant interaction between vividness and type of evidence. A possible explanation for their results is that vividness may have been distracting so that it undermined message persuasiveness (Frey & Eagly, 1993).

To reduce the possible distracting effect of vividness, in the present study vividness was experimentally manipulated at a relatively low level (see emotiveness values in Table 1) and was made relatively equal across all the experimental messages. Despite a constant level of vividness, perceived vividness, measured by how image-evoking a message was thought to be, may still differ across message types. Baesler and Burgoon (1994) manipulated emotiveness, concreteness, and imagery to create vivid and nonvivid messages. However, even if words in two different messages are equal in terms of emotiveness, concreteness, and imagery, one message may still be perceived as more vivid than the other. This difference in perceived vividness may be because (1) manipulated message variables (e.g., message type, topic involvement, or amount of evidence) may lead one message to be perceived as more vivid than another, or (2) individual differences...