A Dual Process Theory Explanation for Door-in-the-Face Effectiveness

Abstract

Two field experiments demonstrate that additive effects of heuristic and systematic processing can explain door-in-the-face technique (DITF) compliance. In Study 1, additive effects of heuristic and systematic processing resulted in an increase in charitable donations. Study 2 replicated the additive effects mean differences observed in the first. Attitude toward a charitable issue was also more strongly correlated with donations in a strong arguments than a no arguments DITF condition, compatible with an additive effects explanation. Both studies indicate that rational processing of argument information has a significant and independent effect on donations in addition to traditional heuristic processing typically inferred in explaining DITF compliance.

Notes

Notes

1 To our knowledge, no prior DITF study has explicitly examined the role of argumentation prior to the large request. Nevertheless, we reviewed 21 past DITF studies to see whether those that used some semblance of arguments advocating compliance prior to the request (Cann, Sherman, & Elkes, 1975; Cialdini & Ascani, 1976; Even-Chen, Yinon, & Bizman, 1978; Gueguen, 2014; Millar, 2002a, 2002b; O’Keefe & Figge 1997; Patch, Hoang, & Stahelski, 1997) had stronger effect sizes (obtained from Feeley et al., 2012) than those that did not (Bell et al. 1996; Cialdini et al 1975; Dillard & Hale, 1992; Gamian-Wilk & Lachowicz-Tabaczek, 2009; Goldman, 1986; Goldman & Creason, 1981; Miller, 1974; Miller 2010; Mowen & Cialdini, 1980; Patch, 1986; Rodafinos, Vucevic, & Sideridis, 2005; Tybout, Sternthal & Calder 1983). The average respective effect (correlation) sizes (.201 vs. .234) did not meaningfully differ. Those studies, however, were not designed to test for the effects of arguments prior to compliance requests. Further, studies with putative arguments typically gave only one, which was most often redundant with the request itself (e.g., arguing, “We need adults who are willing to work with children to help them improve their study skills”) when requesting help with tutoring (O’Keefe & Figge 1997); arguing, “This way we can be sure to have a continual supply of blood” when asking for blood donors (Gueguen, 2014). At other times, articles were unclear about what was even stated to respondents (Rodafinos et al., 2005). The research reported here focuses directly on the role of message argument processing on DITF compliance.

2 Using convention, a Cohen’s d statistic of .2 indicates a small effect size, .5 a medium effect size, and .8 or higher a large effect size.

3 Although we did not collect age and ethnicity data on individual respondents, we have no reason to believe our sample is not representative of the general population of undergraduate students used in our studies (see footnote 4). The university gender split was 50/50 when our studies began, and our samples splits on gender are reasonably close to the university statistics for both studies. The average age of undergraduates at the university is 20 years. The largest age segments are 20–21 years (24.9%), followed by 18–19 (22.7%) and 22–24 (19.4%). In terms of ethnicity, the largest group is Caucasian (57.3%), followed by nonresident alien (13.7%, primarily from China, India, and Mexico) and Hispanic/Latino (11.6%). We acknowledge that our decision not to collect this information at an individual level is a limitation of our study.

4 Respondents interceptions were made across the campus at locations most frequently trafficked by undergraduate students, such as areas close to 1st- and 2nd-year student dorms, and other common areas where undergraduates walk or sit. Because a large percentage of students walk or congregate in pairs or groups, individual respondents were selected using situational availability. Given random assignment to different treatment groups, this sample selection procedure does not represent internal validity problems. Whether those who walk/sit alone versus with others meaningfully differ in other ways, such a personality traits, is unknown.

5 Across both experiments, three respondents initially appeared agreeable to the large request. Those individuals were debriefed and not retained in the analytic sample. No donations were accepted from them.

6 Although two interviewers were used in both studies reported here, only one was used each interviewing day in order to capture all respondents for debriefing. All respondents were debriefed (by the author, wearing a university faculty identifying badge) after each experimental session was completed regardless of whether they donated money. They were asked verbally if they were agreeable to their responses being used in a study being conducted for publication in an academic journal. Seven respondents in total (across both studies) refused, and their data were not used. For respondents who donated money, they were informed that the contributions collected in the studies would be (and later were) donated either to the charity mentioned in the requests or to a closely related charity; the names of the organizations were revealed. All participants were also told they could have their money back if they wished. Five respondents asked for and received their money back, and their data were not used.

7 We followed Petty and Cacioppo’s (1986) definition of an argument, as well as a strong and weak argument in the development of our messages. They defined an argument as a parts of a communication “relevant to a person’s subjective determination of the true merit of an advocacy” (p. 16). They further stated that a strong argument is one that when respondents think about it, “the thoughts they generate are predominantly favorable,” whereas when respondents think about a weak argument, “the thoughts they generate are predominantly unfavorable” (p. 32)—with respect to the target request. Hence, by necessity and design, all strong and weak arguments contain different reasons advocating a position (see Petty & Cacioppo, 1986, pp. 54–59; also see Howard & Kerin, 2011; Burnkrant & Howard, 1984; Petty & Cacioppo, 1984; Petty et al., 1983).

8 Note that the zero median and mode, and the high skewness and kurtosis data in the tables are because the great majority of people in both experiments refused to donate money, especially in the weak arguments (Experiment 1) and control conditions (in both experiments). Nevertheless, the overall and condition comparable compliance rates we obtained are similar to many other DITF studies (see, e.g., the studies cited in footnote 1).

9 One question that can raised regarding the data in both studies is whether the means for “complying only” respondents are consistent with the means for the full samples. This can be tested only using the amount of money donated, as the compliance rate (for compliers) would be a constant across all conditions (100%). The results for money donated are presented next. Experiment 1: The ANOVA revealed meaningful differences between groups (η² = .14). The means are as follows: Strong arguments (n = 28) = 2.62 (SD = 1.67). Weak arguments (n = 5) = .91 (SD = .124). DITF (n = 16) = 1.90 (SD = 1.66). Control (n = 4) = 1.08 (SD = 1.63). Experiment 2: The ANOVA again revealed importance differences between groups (η² = .13). The means are as follows: Strong arguments (n = 35) = 2.87 (SD = 1.72). DITF (n = 18) = 1.98 (SD = 1.61). Control (n = 4) = .775 (SD = .263). In summary, the mean patterns for “compliance only” responders are consistent with the mean patterns observed for the entire sample (see Tables 1 and 2).

10 All data are available upon request from the author