فریب در واکنش پذیری استرس و پژوهش بازیابی

Abstract Objective Testing stress reactivity in the laboratory often requires deception or at least concealment of the hypothesis in order to mimic real-life provocations. Researchers routinely conduct a post-experimental validity check about the success of deception in order to rule out competing hypotheses. The research literature on the impact of failed deception offers contradictory results about the ‘cost’ of failed deception. To date, no evaluation of this threat to internal validity has used objective physiological indices to assess the extent of damage to the results when deception or concealment fails. In this study we evaluated whether or not post-experimental assessment of participants' ability to see through a protocol affected physiological and subjective responses to an anger-provoking laboratory task. Method One hundred and thirty-seven participants were subjected to an anger provocation task disguised as a ‘cognitively challenging arithmetic task’. Results Forty-six participants declared during debriefing that they had seen through or suspected that the underlying hypotheses were related to anger provocation but neither blood pressure, heart rate, or self-reported affect responses to the tasks differentiated the ‘aware’ from the ‘unaware’ group. Discussion We posit that concealment of the hypothesis in anger provocation experiments is usually effective and may not be a threat to the study's internal validity.

. Introduction 1.1. Objective of this manuscript Psychologists conducting experimental research often use elements of deception in order to test their hypotheses (Hertwig and Ortmann, 2008). Such deception is implemented to assure experimental control and improve the internal validity of experimental protocols (Stang, 1976). The current manuscript provides an empirical test of what happens when research participants claim to have seen through deception or concealment. We had the opportunity to test the impact of attempted concealment in a study designed to evaluate emotional and physiological reactivities to an anger provocation stressor. A literature search revealed that no previous publication had tested this question using physiological outcomes. 1.2. Deception and concealment in psychological research The current research needs to be placed within the context of existing literature on the use of deception in experiments. The word ‘deception’ is likely to trigger associations in the reader with the intriguing world of spies, marital infidelities, unsavory sales practices and, for this readership at least, certain types of studies in psychology. There is widespread consensus that deception is generally undesirable and should be avoided (Hertwig and Ortmann, 2008, Ortmann and Hertwig, 1997, Ortmann and Hertwig, 2002 and Fisher and Fyrberg, 1994), yet many psychological phenomena (for example, bystander effects, consumer behaviors, mood inductions, or social skills assessments) are best studied using outright deception or at least some form of concealment of the true experimental hypothesis (Bonetti, 1998 and Kimmel, 1998). Use of deception is by no means rare; in Social Psychology journals, between 30% and 50% of experiments (somewhat variable between different journals) contain elements of deception (Hertwig and Ortmann, 2008). Institutional review boards for research ethics habitually require that participants are fully debriefed after completion of the experiment (and before leaving the laboratory) so that they depart without emotional turmoil and without having developed antipathies against psychological research (Epley and Huff, 1998). To increase the probability of successful concealment, researchers prefer to work with relatively naïve subjects and, in the case of university students as participants, favour 1st and 2nd year students because they have yet to learn about the intricacies of experimental design. This particular article deals with one aspect of deception in research, namely what happens to a study's internal validity (and ultimately the study results) when concealment of the hypothesis is reported post-experimentally as having failed. Use of deception in experimental designs is particularly undesirable when it fails to achieve its intention, namely to assure that responses are not confounded by expectations and knowledge about the experimenters' intentions (Taylor and Shepperd, 1996). This raises the question of how often and under what circumstances participants see through psychologists' concealments. Researchers have reported contradictory results on this account and results may also vary between areas of research and types of protocols (Hertwig and Ortmann, 2008): [a] Research on conformity is greatly affected by participants' discovery of the study hypotheses, [b] participant responses are also changed if participants repeat a similarly deceptive study later, but [c] vague or non-specific announcements of potential deception appear to have little, if any impact. Interestingly, all available studies involving manipulation checks for suspicion of deception used self-report or overt behavior to compare the response of naïve versus suspicious participants. None have used objective physiological indices (Hertwig and Ortmann, 2008). When psychologists began research involving deception and concealment, they reported that few subjects were suspicious (Stricker et al., 1967) whereas our own observations and unpublished anecdotal reports suggest that students have become more knowledgeable and suspicious over time. 1.3. The rationale for this study The prevailing pathway model of how mood and affect relate to the development of high blood pressure and coronary artery disease is that some individuals react more strongly to affect provocation, and that sustained activation of the physiological stress response is the presumed mechanism for later disease development (Schwartz et al., 2003). A number of researchers have shown that a more comprehensive method of studying this pathway model is to carefully evaluate reactivity to and recovery from stressors (Andersen et al., 2005, Hocking-Schuler and O'Brien, 1997, Linden et al., 1997 and Brosschot and Thayer, 1998). Furthermore, there is evidence that emotion provoking, interpersonal stressors are more likely to slow down recovery and maintain activation than are mere physical or cognitive challenges (Dickerson and Kemeny, 2004 and Linden et al., 1998). To tap physiological activation we measured heart rate (which is mostly an index of motor activity), systolic blood pressure (which is mostly based on alpha-adrenergic activity) and diastolic blood pressure that largely reflects vascular changes and is sensitive to negative affect, in particular anger (Brosschot and Thayer, 1998 and Schwartz et al., 2000). Intentional provocation of anger in laboratory reactivity paradigms has been a key tool in the psychophysiological study of affect inducement and health outcomes. Ideally, such linkages would be studied in the natural environment, however, there are pervasive and difficult to overcome methodological problems with any attempt to maintain control in field studies; lack of control, in turn, makes interpretation of field studies results very difficult. An additional concern is that of ethics, because researchers don't want to provoke anger in situations where they cannot assure the safety of research participants (and possibly the experimenters themselves). This has led to the development of laboratory tasks where anger is intentionally provoked but where the protocols are carefully crafted to ensure the hypothesis is concealed and the resulting anger level is not excessive. As well, researchers ensure that there is a comprehensive debriefing at the conclusion of the experiment (Andersen et al., 2005 and Schwartz et al., 2000). In order for the anger provocation to be effective and interpretable, it is a methodological requirement that the participant does not fully understand the study's hypothesis and cannot prepare for his or her own emotional response. Therefore, researchers using anger provocation want to ensure that the intended deception or concealment actually worked and that the participant responded with genuine anger. If, for example, hostile participants were consistently more likely to see through concealment of an intended anger provocation than non-hostile participants, they might modulate their responses correspondingly and become even angrier or pull back in their efforts to perform well on the task. In such a case, researchers have a serious threat to their study's internal validity and results may be difficult to interpret. It has become a standard protocol feature to ask participants after completion of the lab task whether or not they have seen through the deception or concealment. In this manuscript we are reporting a study where healthy undergraduates were exposed to anger provocation and the researchers had the opportunity to use information garnered post-experimentally to objectively assess the impact of failure to fully conceal the hypothesis. “Impact” was operationally defined here as the affective and cardiovascular responses of participants.

. Results One hundred and thirty-seven participants with complete data were available for analysis. 46 were coded as aware and 91 were coded as unaware. An independent sample t-test indicated that there were also no mean differences in age for aware and unaware groups (M = 22.0, SD = 4.9 for unaware, M = 22.8, SD = 6.3 for aware, t(132) = –.87, p > .50). Nor did they differ in terms of gender (23.9% male in exclude group, 28.6% male in the include group, Fisher's Exact p > .50) or ethnicity, (27.6% Caucasian in unaware group, 33.3% Caucasian in aware group, Fisher's Exact p > .50; note that our sample consisted of very few individuals that did not endorse Asian or Caucasian decent, therefore for this analysis ethnicity was coded as Caucasian or other). Independent t-tests indicated that aware and unaware individuals did differ in terms of baseline affective or physiological values (see Table 1). Table 1. Baseline means and standard deviations for aware vs. unaware participants. Variable Aware of intention Unaware of intention M SD M SD Baseline SBP 99.18 6.65 102.53 9.82 Baseline DBP 65.76 7.68 66.53 8.29 Baseline HR 64.41 11.64 68.56 9.57 Baseline happiness 3.33 1.69 3.37 1.93 Baseline anger .46 .75 .95 1.30 Baseline anxiety 1.93 1.52 2.21 1.73 Baseline sadness .45 .74 1.06 1.38 Baseline disgust .12 .18 .59 .99 Baseline surprise .88 1.15 1.55 1.73 Table options The analytical approach used for testing the main question was MANOVA. Initially we conducted one MANOVA for all reactivity change scores from baseline to task (physiological and affective), one for all recovery change scores from baseline to recovery (physiological and affective) and another one for all areas under the curve (physiological and affective). We first computed these analyses with simple change scores and repeated them with residualized change scores. Because the results were essentially the same, we are only reporting the raw change score results because they are easier to interpret. The three initial MANOVAs violated assumption of normality given a significant Box's M result (p < .001). This was assumed to result from large differences in SD for affect scores versus physiological scores (see Table 1 and Table 2). Therefore, we conducted 6 separate MANOVAs separating out state affect scores and physiological measures for both reactivity, recovery and AUC calculations. These 6 separate MANOVAs still indicated a violation of assumptions of normality but because no SD differed more twice that of other cells and each cell's n is greater than 30, MANOVA should still be robust to this violation. Further, because we have no significant results we are unconcerned with this violation as there is no threat of Type I error. Table 2. Means, standard deviations, and intercorrelations for physiological dependent variables. Measure Pearson's correlations M SD (2) (3) SBP AUC 91.1 76.6 .56⁎⁎ .34⁎⁎ DBP AUC (2) 84.3 73.5 − .30⁎⁎ HR AUC (3) 59.8 54.2 − (5) (6) SBP reactivity (4) 15.7 9.2 .66⁎⁎ .40⁎⁎ DBP reactivity (5) 14.9 7.8 − .36⁎⁎ HR reactivity (6) 15.9 8.5 − (8) (9) SBP recovery (7) 4.14 6.30 .56⁎⁎ .19⁎ DBP recovery (8) 3.52 6.50 − .17⁎ HR recovery (9) −.18 4.21 − ⁎p < .05, ⁎⁎p < .001, n = 137. Table options As Table 2 and Table 3 reveal, there was a large amount of overall change in affect ratings as well as physiological responsivity. Nevertheless, the 6 separate MANOVAs on participants' responses to the stressor all showed non-significant results (see Table 4). The multivariate F indicated no significant effect (not even near-significant) for self-reported awareness vs. lack of awareness and its consequences for the magnitude of responses (see Fig. 1, Fig. 2 and Fig. 3 for physiological activation over experiment time for both groups). Therefore, the results indicate that awareness of intention to provoke anger did not alter the physiological or affective reaction to the experimental harassment. Table 3. Means, standard deviations, and intercorrelations for affective dependent variables. Measure Pearson's correlations M SD (2) (3) (4) (5) (6) Anxiety AUC .08 2.59 −.13 .20⁎ .22⁎ .11 .15 Happiness AUC (2) − 1.40 2.14 − −.10 −.13 −.09 −.04 Anger AUC (3) .98 2.06 − .28⁎⁎ .43⁎⁎ .09 Sadness AUC (4) .11 1.78 − .15 −.08 Disgust AUC (5) .77 1.76 − .25⁎⁎ Surprise AUC (6) .81 2.08 − (8) (9) (10) (11) (12) Anxiety reactivity (7) .88 1.68 −.15 .27⁎⁎ .29⁎⁎ .14 .16 Happiness reactivity (8) −.97 1.17 − −.23⁎⁎ −.16 −.16 −.04 Anger reactivity (9) .93 1.54 − .29⁎⁎ .54⁎⁎ .07 Sadness reactivity (10) .30 1.29 − .20⁎ −.15 Disgust reactivity (11) .78 1.43 − .21⁎ Surprise reactivity (12) .50 1.63 − (14) (15) (16) (17) (18) Anxiety recovery (13) −.80 1.37 −.16 .21⁎ .25⁎⁎ .24⁎⁎ .01 Happiness recovery (14) −.43 1.23 − −.06 −.12 −.13 −.11 Anger recovery (15) .05 .82 − .27⁎⁎ .23⁎⁎ .11 Sadness recovery (16) −.19 .70 − .12 .10 Disgust recovery (17) .00 .62 − .30⁎⁎ Surprise recovery (18) .31 .96 − ⁎p < .05, ⁎⁎p < .001, n = 137. Table options Table 4. Multivariate tests for aware vs. unaware group differences in physiological and affect change. Effect Λ F df1 df2 p Physiological reactivity Awareness .98 .93 3 131 .43 Gender .95 2.10 3 131 .10 Awareness × Gender .99 .57 3 131 .63 Physiological recovery Awareness .98 1.13 3 131 .34 Gender .98 1.08 3 131 .36 Awareness × gender .98 .82 3 131 .49 Physiological AUC Awareness .99 .41 3 131 .75 Gender .96 2.00 3 131 .12 Awareness × gender .98 .98 3 131 .40 Affect reactivity Awareness .97 .69 6 128 .67 Gender .91 2.03 6 128 .07 Awareness × gender .97 .61 6 128 .73 Affect recovery Awareness .99 .20 5 129 .96 Gender .99 .31 5 129 .91 Awareness × gender .97 .69 5 129 .63 Affect AUC Awareness .98 .36 6 128 .91 Gender .94 1.32 6 128 .25 Awareness × gender .98 .52 6 128 .79 n = 137. Table options Systolic blood pressure activation over experiment time. Fig. 1. Systolic blood pressure activation over experiment time. Figure options Diastolic blood pressure activation over experiment time. Fig. 2. Diastolic blood pressure activation over experiment time. Figure options Heart rate over experiment time. Fig. 3. Heart rate over experiment time. Figure options Also shown in Table 2 and Table 3 are the intercorrelations of various change parameters. This was included to show that the different dependent variables within each class of measures (biological versus self-report) were only partially overlapping thus justifying that they could be treated as distinct variables. As expected, systolic and diastolic blood pressure changes are significantly inter-correlated whereas both are much less closely tied to heart rate changes. With respect to affect ratings, there was remarkable little overlap in correlation coefficients for change scores; an exception is the overlap between disgust and anger suggesting that these feelings share common variance.