Levine, McCornack / THE PROBING EFFECT 471 Behavioral Adaptation, Confidence, and Heuristic-Based Explanations of the Probing Effect TIMOTHY R. LEVINE STEVEN A. MCCORNACK Michigan State University Researchers have found that asking probing questions of message sources does not enhance deception detection accuracy. Probing does, however, increase recipient and observer per- ceptions of source honesty, a finding we label the probing effect. This project examined 3 potential explanations for the probing effect: behavioral adaptation, confidence bias, and a probing heuristic. In Study 1, respondents (N = 337) viewed videotaped interviews in which probes were present or not present, and in which message source behaviors were controlled. Inconsistent with the behavioral adaptation explanation, respondents perceived probed sources as more honest than nonprobed sources, despite the fact that source behav- iors were constant across conditions. The data also were inconsistent with the confidence bias explanation. Studies 2 and 3 investigated the probing heuristic explanation. The data from Study 2 (N = 136) were ambiguous, but the results of third study (N = 143) were consistent with the heuristic processing explanation of the probing effect. Common sense suggests that interrogative probing should be an effective strategy for detecting deception. By asking message sources probing questions, message recipients gain additional in- formation, and consequently should render more accurate judgments regarding the veracity of sources’ messages. Police and military interro- gations, cross examinations in courtrooms, and interviews by investiga- tive reporters are just a few examples of how probing is used as a strat- egy for uncovering truths and lies. Timothy R. Levine (Ph.D., Michigan State University, 1992), is an associate professor in the Department of Communication at Michigan State University. Steven A. McCornack (Ph.D., University of Illinois, 1990), is an associate professor in the Department of Communication at Michigan State University. The authors would like to thank Dr. Carlos G. Aleman and Dr. John Oetzel for their assistance with the first two studies. The authors also would like to thank Dr. Bella DePaulo for her helpful comments regarding earlier drafts. Correspondence concerning this article should be addressed to Tim Levine, Department of Communication, Michigan State University, East Lansing, MI 48824-1212; email: [email protected] Human Communication Research, Vol. 27 No. 4, October 2001 471–502 © 2001 International Communication Association 472 HUMAN COMMUNICATION RESEARCH / October 2001 Contrary to intuition, interrogative probing has not been found to sig- nificantly enhance deception detection accuracy (Buller, Comstock, Aune, & Strzyzewski, 1989; Buller, Strzyzewski, & Comstock, 1991; Stiff & Miller, 1986). Even more surprising, researchers repeatedly have found that prob- ing causes judgments of message source honesty to increase (Buller et al., 1989; Buller, Strzyzewski, & Comstock, 1991). Regardless of the ac- tual truth or falsity of the messages presented (Buller et al., 1989; Buller, Strzyzewski, & Comstock, 1991), the interactants’ relationship history (Buller, Strzyzewski, & Comstock, 1991), message recipients’ suspicion (Buller, Strzyzewski, & Comstock, 1991) or whether the individuals mak- ing truth/deception judgments are interaction participants or observers (Buller, Strzyzewski, & Hunsaker, 1991), the simple act of probing, or witnessing a source being probed, enhances source believability. We la- bel this consistent, robust, and counterintuitive finding the probing effect. The behavioral adaptation explanation (BAE) is regarded by many as the most plausible account for why the probing effect occurs (e.g., Buller et al., 1989; Buller, Strzyzewski, & Comstock, 1991; Buller, Stiff, & Burgoon, 1996; Burgoon & Buller, 1996; Burgoon, Buller, Ebesu, & Rockwell, 1994; Burgoon & Floyd, 2000; Miller & Stiff, 1993; Stiff & Miller, 1986). The BAE proposes that the reason probed sources are judged as more truth- ful than nonprobed sources is that probed sources adapt their postprobe behaviors to appear “honest.” Specifically, the BAE posits a three-link causal chain: (a) the probing of a source by a message recipient causes the source to believe that the recipient may be suspicious, (b) the source’s recognition of recipient suspicion causes the source to alter his or her behaviors to appear more truthful, consequently (c) the source’s truth- ful-appearing, adapted postprobe behaviors cause the message recipi- ent to judge the source as honest (relative to nonprobed sources). Hence, probing (albeit indirectly) leads message recipients to attribute truthful- ness to message sources. Although several studies have examined probing (e.g., Buller, Comstock, Aune, & Strzyzewski, 1989; Buller, Strzyzewski, & Comstock, 1991; Stiff & Miller, 1986), the BAE has garnered little direct empirical support. The source behavioral changes that occur subsequent to prob- ing typically are small in effect size, infrequent in occurrence, inconsis- tent across studies, and most often not in the directions predicted by the BAE (Levine & McCornack, 1996a, 1996b). Consequently, the probing effect remains a robust finding that must be judged to be without a cred- ible explanation. The goal of the current project is to test three different explanations for the probing effect. Our first study tests both the BAE and a confi- dence explanation derived from McCornack and Parks’s (1986) research on relational deception. Our second and third studies test a third possi- bility, the existence of a probing heuristic. Levine, McCornack / THE PROBING EFFECT 473 Probing Research Probing has been defined and operationalized as the direct question- ing of a message source regarding the veracity of information presented, or requesting that additional information be presented. Weiler and Weinstein (1972) examined the effects of supportive and suspicious probes on the type and frequency of credibility-enhancing statements used dur- ing a simulated job interview. The use of credibility-enhancing statements did not differ significantly between probing conditions. Stiff and Miller (1986) probed subjects either in a fashion that sup- ported belief in their veracity, or a fashion that questioned their veracity. Subjects’ responses were videotaped and coded for 10 nonverbal behav- iors. Positive and negative probes were edited from the audio-track of the videotapes and replaced with neutral probes. A different sample of subjects subsequently viewed the videotaped interrogations and were asked to rate the honesty of the videotaped, interrogated subjects’ re- sponses. Although probing did not enhance detection accuracy, sources who originally had been subjected to negative probes were judged as slightly more truthful than those who had been interrogated with posi- tive probes (r = .14). Stiff and Miller (1986) then correlated each of 10 coded source nonver- bal behaviors with the type of probe that had been used, and with verac- ity judgments. Four of the 10 source behaviors correlated significantly with honesty judgments. None of the probing-behavior correlations were statistically significant. For 8 of the 10 coded behaviors, however, the sign of the probing-behavior effect was in the direction opposite from the behavior-veracity judgment correlation. Thus, the same behaviors apparently caused by probing were apparently associated with truthful attributions. Based upon these findings, the results of Kraut and Poe (1980), and research on impression management (e.g., Silverman, Rivera, & Tedeschi, 1979), Stiff and Miller speculated that communicators faced with negative probes may realize that message receivers are suspicious, and alter their behaviors accordingly in order to appear truthful. Buller et al. (1989) extended upon the Stiff and Miller (1986) study by having respondents interact face-to-face, use probing during these inter- actions, and then make veracity judgments regarding their conversational partners. Probing caused sources to encode more speech errors, talk longer, and pause more. Also, individuals who had probed sources were more likely to rate the sources as honest than individuals who had not probed (r = .22). Buller et al. (1989) interpreted their results as consistent with the Stiff-Miller (1986) study and the BAE. They argued that probing causes both deceivers and truthtellers to present cues indicative of truth- fulness. These cues, in turn, purportedly caused receivers to attribute honesty to sources whom they probed, because these sources encoded 474 HUMAN COMMUNICATION RESEARCH / October 2001 behaviors linked to stereotypes of truthtellers (see Buller et al., 1989, p. 167). Similar results and interpretations were provided by Buller, Strzyzewski, and Comstock (1991). They coded 25 nonverbal behaviors, and found 6 significant main effects for probing on source behaviors. Probing resulted in longer turns and response latencies, and more paus- ing, speech errors, gestures, and laughing. Similar to Buller et al. (1989), truth judgments increased following probing (r = .57). Buller, Strzyzewski, and Comstock (1991) also reported the only test to date of the first two links in the causal model specified by the BAE (i.e., probing leads to increases in source perceived suspicion, and perceived suspicion pro- duces source behavioral adaptation). The effect for the first link was moderated by probe type. Coded probe skepticism was positively asso- ciated with perceived suspicion, but the effect for the probe manipula- tion was found to be nonsignificant. In examining the source-perceived- suspicion-leads-to-source-behavioral-adaptation link, 6 behaviors were found to be associated with perceived suspicion: illustrators (r = -.17), body activity (r = -.13), turn length (r = -.14), laughing (r = -.19), head shaking (r = .18), and response latencies (r = .17). Based upon these re- sults, Buller et al. concluded that “as anticipated, perceptions of suspi- cion caused sources to manage their behavior” (p. 20), and “deceivers monitored receivers’ reactions to determine whether their deceit was succeeding and, when suspicion was detected, they altered their behav- ior to appear truthful” (p. 18). Most recently, Vrij (1995) investigated the effects of probing on non- verbal behaviors in a simulated police interview. Vrij reported that rela- tive to baseline behaviors, probed subjects engaged in fewer head, trunk, and leg movements; smiled less; gazed more; and had more changes in voice pitch. Probing and Behavioral Adaptation Close examination of previous findings indicates that data fail to pro- vide compelling support for the BAE. If, when probed, both deceivers and truthtellers strategically adapt their behaviors to appear more truth- ful, then the behaviors that change should be those that research-naive individuals stereotypically associate with deceptiveness. In order for this behavioral adaptation to account for the increase in honesty judgments that follows probing, however, the behavioral cues that are altered must include those that influence receiver truth-deception judgments. Hence, for the BAE to account for the probing effect, sources must change some of those behaviors that are both stereotypically deceptive and decoded- as-deceptive (Levine & McCornack, 1996a). Zuckerman, DePaulo, and Rosenthal’s (1981) meta-analysis found that although the behaviors Levine, McCornack / THE PROBING EFFECT 475 stereotypically linked with deception correlate moderately (r = .44) with those linked to judgments of deception, six of the stereotypical decep- tion cues also cause deceptiveness judgments to increase (i.e., gaze, pos- ture shifts, response latencies, speech errors, hesitations, pitch). Conse- quently if the BAE is a viable explanation for the probing effect, sources who are strategically adapting their behavior to appear honest and who also will be perceived as honest should use longer gazes, fewer posture shifts, shorter response latencies, fewer speech errors and hesitations, and a lower pitch. The six behaviors specified by Zuckerman et al. (1981) related to both stereotypes about honesty and judgments of honesty are listed in Table 1, along with the behaviors found to change as a result of probing. As Table 1 displays, six of the nine behaviors found to vary as a function of probing vary in a fashion indicative of source deceptiveness rather than source honesty. Thus, two thirds of the cues purportedly indicative of sources’ strategic behavioral adaptation are in the direction opposite of that predicted by the BAE. To summarize, the BAE suggests a causal model in which probing leads message sources to perceive recipient suspicion, this perceived suspicion causes sources to adapt their behaviors to appear truthful, and this behavioral adaptation dupes message recipients. The evidence for this model is, at best, mixed. Research has not documented a significant relationship between probing and perceived suspicion (Buller, Strzyzewski, & Comstock, 1991), and the relationships between probing and behavioral changes have been nonsignificant (Stiff & Miller, 1986; Weiler & Weinstein, 1972), in the wrong direction (Buller et al., 1989; Buller, Strzyzewski, & Comstock, 1991), or generally consistent with the BAE (Vrij, 1995). The majority of behaviors that do change should lead to judg- ments of source deceptiveness, rather than the judgments of source hon- esty repeatedly observed by past researchers. Nevertheless, authors con- tinue to cite the BAE as established fact (e.g., Burgoon & Floyd, 2000, pp. 245–246). Because only one in five studies have actually produced data consistent with the BAE, a search for an alternative explanation for the probing effect appears warranted. Confidence One alternative explanation for the probing effect can be derived from the McCornack and Parks (1986) model of relational deception. McCornack and Parks found that as relationship involvement increases, people become increasingly confident in their ability to detect their part- ners’ duplicities. These increases in confidence generate a “truth-bias,” or tendency to judge all messages as truthful. Given this, confidence trig- 476 HUMAN COMMUNICATION RESEARCH / October 2001 n ot o n perceived suspici der et al., 1991 nrnr+nrnr ut main effects bing ects of Bull ated b y Pro Eff estig b v Produced eVrij, 1995 +- ns - nrcts, = in nd c1 effe n, a 199 ain ABLE 1Decoded as Truth-Deceptio Effects of probing ber et al., 1989Buller et al., nsnrnsnrns+++++ nsnship; = nonsignificant m To Truth-Deception, a& Miller, 1986Bull nsnsnsns es a negative relatio ues Stereotypically Related t Stereotypes andtruth judgments uckerman et al., 1991Stiff +----- ositive relationship, - indicat C Z a p s y n e Cue GazePosture shiftResponse latencSpeech errorSpeech hesitatioPitch NOTE: + indicatreported.aN = 40.bN = 148.cN = 210.dN = 210.eN = 64. Levine, McCornack / THE PROBING EFFECT 477 gered by probing can be advanced as a potential explanatory mecha- nism underlying the probing effect. Although researchers have not yet tested the relationship between probing and confidence, it seems plausible that such a link exists. Just as relational development produces confidence, which in turn produces a truth-bias, probing too may lead to increased confidence, which in turn produces a truth-bias. As receivers probe potential liars, they should gain more information upon which to base truth-lie judgments. As a function of having more information, and a belief in the efficacy of probing as a detection strategy, receivers should become more confident in their abil- ity to make accurate judgments; confidence that ultimately results in truth-bias (DePaulo, Charlton, Coooper, Lindsay, & Muhlenbruck, 1997; McCornack & Parks, 1986; Levine & McCornack, 1992). A model based upon confidence also provides an explanation for why probing appears to have little effect upon judgmental accuracy. Early probing researchers (e.g., Stiff & Miller, 1986) predicted that probing a source would provide a receiver with more information resulting in in- creased detection accuracy. This reasoning renders the repeated finding that probing does not increase accuracy counterintuitive. If, however, the accuracy boost generated by the information gained through prob- ing is offset by a processing bias stemming from confidence, one would expect not to find an effect for probing upon accuracy. Consistent with this reasoning, research suggests that confidence and accuracy are unre- lated (DePaulo et al.,1997; Vrij, 2000). STUDY 1 Previous researchers who have examined interrogative probing have argued that the increase in honesty ratings that occurs subsequent to probing stems from behavioral changes on the part of message sources. If behavioral adaptation causes the probing effect, one should find that when source behaviors are held constant, there will be no significant difference between sources who are probed and sources who are not probed in ratings of truthfulness. If, however, the line of reasoning re- garding the confidence explanation is valid, behavioral adaptation on the part of the source should not be required to observe the probing ef- fect. Thus, we expect to find that when source behaviors are held con- stant, sources who are probed will be rated as significantly more truthful than sources who are not probed. Although data consistent with this hypothesis would be consistent with the confidence explanation, this would provide only indirect sup- port for the confidence explanation. In order to provide a direct test of 478 HUMAN COMMUNICATION RESEARCH / October 2001 the role of confidence, one would need to assess the effect of probing upon confidence, and the effect of confidence upon truth judgments. To the extent that the confidence explanation provides a viable account for the probing effect, we should find that individuals will be more confi- dent in their judgments of sources who are probed than sources who are not probed, and confidence will be positively associated with judgments of honesty. The type of probe that is used also may have an effect upon confi- dence. Stiff and Miller (1986) found a small but significant effect for the valence of probes, but did not have a no-probe control group. Later studies using only probe and no-probe groups found substantially larger effect sizes (Buller et al., 1989; Buller, Strzyzewski, & Comstock, 1991). Nega- tive probes often imply suspicion and disbelief (Buller et al., 1989; Stiff & Miller, 1986). Given that suspicion may result in less confidence (Fein, Hilton, & Miller, 1990; Levine & McCornack, 1992; Stiff, Kim, & Ramesh, 1992), the use of negative probes may be significantly related to decreases in confidence. On the other hand, face-threatening, pointed probing is commonly believed to be more likely to trap deceivers (Buller et al., 1989). Thus, the use of negative probes may actually function to enhance confi- dence, by enhancing individuals’ feelings that they indeed have “hit sources with their best shot.” To explore this issue, our study tested if the type of probe used (i.e., positive-negative) influences the strength of the probing effect. Method Study 1 occurred in several stages. First, respondents were videotaped discussing truthful and deceptive answers to an attitude questionnaire that they had completed. The probes were altered or removed from the tape to form the stimuli for four probing conditions in which source be- haviors remained constant (i.e., positive, negative, neutral, and no probe). Next, the four tapes were pretested for realism. In the final stage, re- spondents each viewed one of the four tapes, and made judgments about the truthfulness of the sources on these tapes. All participants received extra credit in exchange for their participation. Phase 1: Interviews. Seventeen undergraduate students (11 females, 6 males) from a large Midwestern (U.S.) university participated on a vol- untary basis. Upon arriving at the experimental site, each respondent was told that the purpose of the experiment was to “examine the percep- tions that people have of the communication behaviors of others.” Re- spondents were then asked to complete a four-item questionnaire. Simi- lar to other deception studies (e.g., McCornack & Parks, 1986), the ques- tionnaire contained randomly selected items from the Mach IV scale Levine, McCornack / THE PROBING EFFECT 479 (Christie & Geis, 1970). Responses to these items were recorded on 10-point Likert-type scales.1 After completing the four items, an experimental assistant collected the questionnaire and changed two of the respondent’s answers to an- swers that were five points different (i.e., half of the scale length) from the respondent’s original true answers. Six different orderings of the items to be changed were used to minimize order effects. The respondent was then informed that he or she would be interviewed regarding his or her answers on the altered questionnaire. The respon- dent was told to report his or her original answer for the two unchanged items during the interview. For the remaining two items, the respondent was told to report the answer that was changed (half the scale length different from his or her true answer). Each respondent was given 10 minutes to prepare his or her responses. At the end of this time, an experimenter entered the room and sat facing the respondent. Each respondent was asked to state his or her attitude on each of the items and to briefly explain his or her answer. The experimenter was not informed as to which items were truthful (i.e., un- changed) and which items were lies (i.e., changed). This interview was videotaped from behind a one-way mirror. The camera was placed so that the respondent’s entire body would appear, while the experimenter could not be seen. During the interview, each respondent reported his or her answer and then briefly discussed reasons for his or her choice of that answer. Fol- lowing each answer, the experimenter probed the respondent with a neu- tral probe, saying “tell me a little bit more about why you answered it that way.” Following this probe, each respondent elaborated briefly upon his or her answer. At the conclusion of the interview, each respondent was debriefed and thanked for his or her participation. Phase 2: Creating experimental conditions. From the original question- naire protocol, one item was randomly selected as the stimulus item for the final test tapes. A single item was selected as the basis for all responses on the tapes in order to minimize potential effects due to variation in item content. Eight videotaped respondents were then randomly selected from the original sample of 17 interviewees with the constraint that there were 2 males who had lied, 2 males who had told the truth, 2 female liars, and 2 female truthtellers. A master tape was then created, with each of these eight interviews (regarding respondents’ answers on item one). The order of the eight interviews on the tape was randomly determined. Following the creation of the “master” tape, three additional “stimu- lus” tapes were created. In the negative-probe stimulus tape, the audio- track of the tape was altered, so that the experimenter’s original “neu- 480 HUMAN COMMUNICATION RESEARCH / October 2001 tral” probe was replaced with a negative, face-threatening probe. The experimenter said, “I don’t think you really circled that. Tell me a little more about why you answered it that way.” In the positive-probe stimu- lus tape, the experimenter’s original probe was replaced with a positive, face-supportive probe. The positive probe was “that sounds reasonable, but tell me a little more about why you answered it that way.” In the no- probe control stimulus tape, the experimenter’s probe was deleted and the tape spliced, so that each respondent’s answer to the item and his or her explanation of the answer were joined together. This resulted in four tapes each corresponding to a probing condition: no probe, neutral probe (i.e., the master tape), positive probe, and negative probe. Phase 3: Pretesting of stimulus tapes. The final step (prior to conducting the primary study) was to test each of the tapes for realism. This was to ensure that, despite the editing, the three altered tapes would not be per- ceived by naive observers as significantly different from the master tape (i.e., the original, unedited videotape of the eight respondents discuss- ing item one, with the experimenter using neutral probes). One hundred and thirty-four undergraduates (76 women and 58 men) from a large Midwestern university participated on a voluntary basis outside of class time. Respondents were informed that the purpose of the experiment was to “examine their perceptions of the communication behaviors of others.” Respondents were randomly assigned to view one of the four videotapes (i.e., one of the three edited tapes or the master tape). After viewing the tape, respondents completed a six-item ques- tionnaire designed by the experimenters to measure the realism of the test tapes. The questionnaire used Likert-type items with seven-point response formats. A sample item read “the interviews on this videotape seemed natural.” The internal consistency of the realism scale was tested using confir- matory factor analysis. The scale was found to be reliable (α = .91) and consistent with a unidimensional measurement model. The six items were then summed to create an overall index of realism for each respondent. A one-way ANOVA was then conducted, with the four different test tapes as the independent variable and perceived realism as the depen- dent measure. The ANOVA found a marginally significant, and small difference between the four test tapes in terms of realism F(3, 130) = 2.64, p = .052, η2 = .02, r =.14). Upon examination of the means, it became apparent that the negative probe tape (M = 32.56) seemed to be the cause of the difference (means in the other three cells ranged from 35.45 to 36.57). Given that the means for all of the tapes were over 32 (i.e., potential range for realism being 6 to 42), it is obvious that each of the test tapes was seen as fairly realistic