406 M.E. Hyland, B. Whalley / Journal of Psychosomatic Research 65 (2008) 405–413dispositions such as suggestibility and acquiescence are another. For any ritual, a person whose motivations arepredictors of placebo responding because they amount to a concordant with the ritual should be more engaged and, so,generic tendency to respond to suggestion . More have better outcomes (due to either of the two mechanisms ofrecently, the trait of optimism has also been found to predict behaviorally mediated therapeutic benefit). Intrinsic motiva-placebo outcome [16,17]; optimism correlates with expec- tion for the ritual depends on the fit between the personstancy, with optimists expressing more positive expectations. motives and the ritual. So, according to motivationalThe second and more popular view is that there is no such concordance theory, there should be no such thing as athing as a placebo responding personality [18,19]. Adherents generic placebo responder, but there should be context-of this second viewpoint to the considerable inconsistency in specific placebo responders, where the placebo responderthe placebo-responder personality literature and also make a characteristics depend on the therapy. The implication is thattheoretical point: expectancies are determined by an correlations between predictors and outcome should changeevaluation of the specific aspects of the situation—in the if the motivational context of the therapy is changed.context of the person×situation debate, they are the Fig. 1 provides a schematic representation of responseconsequence of the situation and the person and not just expectancy theory and the behaviorally mediated motiva-the person. For this reason, one would not predict a generic tional concordance theory. Classical motivation theoryplacebo responding personality—only context-specific cor- shows that motivation (i.e., the tendency to engage in goal-relations between expectancies and outcomes. oriented behavior) is the product of value and expectancy— i.e., expectancy×value theory . If response expectancy is the only mechanism (i.e., motivational concordance does notMotivational concordance and the placebo responder occur), then expectancy of positive outcome should correlate with outcome, and neither the intrinsic motivation for the The theory of motivational concordance is based on the therapy nor behavior should add additional variance. Ifassumption that the placebo response is a reaction to the motivational concordance is the only mechanism (i.e.,behavior of the therapeutic ritual. Two well-established response expectancy does not occur), then expectancy oftheories explain why the behavior of a therapeutic ritual positive outcome should not add additional variancemight affect outcome. First, several motivation theories (self compared to intrinsic motivation for the therapy. If responsedetermination theory, control theory, self-actualization and expectancy and motivational concordance are both true, thenpersonal growth theory) share a common assumption that expectancy of positive outcome on the one hand, andgoal fulfillment is a positive experience [20–25]. There is a intrinsic values and behavior on the other should bothgeneral consensus that the attainment of self-actualizing, contribute variance to outcome.self-defining, or self-relevant goals leads to positive affect.Additionally, there is a well-established link between affectand immune function [26,27], and so, positive goal Flower essences as placebosattainment can also create therapeutic physiological changes. Second, self-perception theory suggests that behavior is Flower essences are a form of complementary anda source of information about the self [29,30], and the alternative medicine that can be purchased over the counterbehavior of the ritual can therefore provide information that in pharmacies and health shops, or via the internet, as aaffects perception of symptoms. What is common to both remedy for psychological symptoms. They are widely usedthese theories is the idea that the therapy is effective to the in Western countries: a major pharmacy in the Unitedextent that a person engages in the ritual. Kingdom reports 650,000 bottles sold annually for a cost There are two reasons why a person may engage with a of £3.4 million in 2006 (personal communication). Each oftherapeutic ritual. One is the desire to get better and the the 38 Bach flower essences purports to treat a differentexpectation that the therapy will be effective—i.e., the psychological symptom (including anxiety, depression, andextrinsic value of the ritual which, when coupled with fatigue, as well as more unusual symptoms such as impatienceexpectancy, leads to motivated behavior. A second reason is or over concern with others). Users select the particular flowerthat the ritual satisfies important, self-actualising goals (i.e., essence using a chart that is placed near the essences in thethe intrinsic value of the ritual), which, when coupled with retail outlet or on the internet and which helps users decide onthe expectancy of achieving those self-actualising goals, the particular essence or essences they need. Like otherleads to motivation to engage in the ritual. Here, we focus on complementary medicines, flower essences are a spirituallythe second of these motivations to engage with a ritual, contextualized therapy , and the spiritual nature of flowernamely, the intrinsic motivation of the ritual. Note that essences was part of the rationale presented by their inventor,people may be more optimistic about self-actualising rituals Edward Bach . From a biochemical perspective, allbecause the extrinsic expectation of success can be 38 essences are identical (brandy 60% and water 40%), and noassociated with the intrinsic value of the ritual. difference has been detected between verum and placebo People have different self-actualising goals, so a ritual [34,35]. Flower essences can be considered a self-helpthat is self-actualising for one person may not be so for placebo, which is used regularly for clinical purposes.
M.E. Hyland, B. Whalley / Journal of Psychosomatic Research 65 (2008) 405–413 407 Fig. 1. Two different placebo mechanisms. When employed in placebo research, flower essences are would show that dispositional predictors depend on context,offered free of charge to volunteers in return for ques- as predicted by motivational concordance.tionnaire completion [10,11]. Like double-blind placebo The second aim is to test whether response expectancytrials, users in flower essence placebo studies believe they alone, motivational concordance alone, or both theoriesare engaging in real therapy and are not paid. Unlike clinical together contribute to outcome. To do this, we havetrials, participants believe that they are given verum (which measured expectancy and spirituality as before but haveis the case); uncertainty about group assignment in clinical added indicators of behavioral engagement, namely,trials may affect results . The use of flower essences in compliance and a retrospective measure of ease of taskplacebo research is not a laboratory analogue, and placebo completion. The latter measure could be biased bymechanisms for laboratory analogue studies may not be the perceived outcome (i.e., people perceive the task easiersame as those for real-life therapies. The methodology relies only because they have had a positive outcome); theon an existing set of beliefs and corresponds closely to the former measure is not subject to this reporting bias. Fornormal ritual use of the inert substance. this reason, these two behavioral measures will be Dispositional spirituality predicts response to flower analysed separately. However, they are, of course, onlyessences independently of expectancy , and this finding indicators of behavioral engagement, in the sensehas been replicated with a more conservative test where that they do not cover all variance attributable tomultiple measures of expectancy were taken . These behavioral engagement.findings can be explained by the motivational concordancemechanism. Spirituality is one of the high-level values orgoals that motivates behavior . If flower essences are Methodinterpreted as a spiritually oriented therapy, then peopleplacing high value on spirituality should be more motivated Overviewto perform the flower essence ritual, find it more satisfying,become more involved in the ritual, and gain greater benefit. Participants provided informed consent and completedThe idea of a behaviorally mediated form of placebo questionnaires at baseline and were then randomized to threeresponse is consistent with anthropological data showing groups (spiritual, affirmation, and neutral), each receivingthe beliefs are not essential for response to rituals that are different kinds of further information. They took floweracted out . essences for 3 weeks, during which time they provided There are two aims of this study. The first is to show that follow-up assessments. Excluding technical support, therethe correlation between spirituality and outcome occurs only was no human contact with participants, who enteredwhen flower essences are contextualized as a spiritual therapy baseline data online and provided follow-up data using anand not when contextualized as a nonspiritual therapy. This automated telephone system.
408 M.E. Hyland, B. Whalley / Journal of Psychosomatic Research 65 (2008) 405–413Procedure “Remember, flower essences work best if, while you are taking them, you imagine....” Finally, on Day 21, the The study was advertised through the media. Those spiritual and affirmation groups received the questiontaking part would be given a free bottle of flower essence in about the ease of the ritual.return for questionnaire completion and evaluation. Exclu-sion criteria were: use of flower essences in the previous 6 Assessmentsmonths, currently receiving psychiatric treatment, history ofalcohol abuse. Participants were instructed to log-on to a The SCQ-14 consists of seven positive and sevenWeb page where they (a) were provided with information negative items about the experience of spiritual connectionabout flower essences and the study, (b) gave consent, (c) with the universe and other people and the happiness suchconfirmed that they did not meet the exclusion criteria, (d) connection brings. Because it measures reported experience,completed baseline questionnaire assessments [Spiritual the scale can be considered to measure the motive to engageConnection Questionnaire 14 (SCQ-14) and Expectancy] in spiritual activity. A longer version of the scale predicts(e) selected any one of the 38 Bach flower essences with outcome for flower essence treatment . The scale isessence descriptions and picture of the flower taken from a secular in content and is consistent with the kind of new agecommercial website, and (f) gave a telephone number and beliefs associated with complementary medicine. Hightime of day for follow-up contact. The flower essences scores indicate more spirituality.(genuine commercially produced essences with a standard Expectancy was measured by a single seven-point scalelabel) were then posted to the participants. where participants were asked to rate “At this point in time Participants were randomized to one of three treatment do you expect the flower essence to help you?” The endgroups (spiritual, affirmation, and neutral) using a random points of the scale were marked Unlikely it will help (−3) andnumber list as they consented to the Web study. All Definitely think it will help (+3).participants received a brief introduction to flower essences To assess outcome, participants heard the followingon the Web site and were told that although they are message: “How much better do you feel from taking thebiologically inert, practitioners and users make controversial flower essence? Press a number from one to nine, where oneclaims that they work through a spiritual mechanism not yet means you feel much worse, five means you feel the same,understood by science. Participants were sent their flower and nine means you feel much better.” A final outcome scoreessence with an instruction to take three drops twice per day, was calculated from the mean of the last three outcomeand at this point in time, the ritual was extended for the assessments (positive scores indicate improvement). Anspiritual and affirmation groups. The spiritual group received initial outcome score was calculated from the mean of thethe written information: “Flower essences work best if, while first three outcome assessments.you are taking them, you imagine the essence connecting you Compliance was measured by the question: “Did you taketo a universal pool of healing and love.” In the affirmation the flower essence this morning? Press one for yes or zerogroup, participants received the written information: “Flo- for no.” Overall compliance was calculated as the mean ofwer essence works best if, while you are taking them, you responses made (high scores indicate greater compliance).imagine them helping you to solve your problem.” Neutral Ease of ritual was assessed by a single question in the finalgroup participants were not provided with additional automated telephone call, which corresponded with the ritualinformation. These extensions to the ritual are consistent extension in the spiritual and affirmation groups. In thewith instructions sometimes found in complementary spiritual group, this question was “How easy was it tomedicine (where autosuggestion is called “affirmation”). imagine the flower essence connecting you to a universal When participants received their essence they were pool of healing and love?” In the affirmation group, theinstructed to call an automated telephone line and register question was “How easy was it to imagine the flower essencetheir entry into the study with a unique identification code helping you solve your problem?” Participants responded onprovided. This telephone registration initiated a series of a nine-point scale, where 9 was defined as very easy and 1 ascalls by an automatic telephone system; calls were made on very difficult. Ease of ritual was not assessed for theDays 1, 2, 3, 4, 7, 14, and 21 after registration at the time of neutral group.day preferred by the participant (unanswered calls werefollowed up 30 min later, with up to five attempts madewithin the time period specified by the participant). During Resultseach call, participants were asked to provide an assessmentof outcome and compliance by entering numbers on the Three hundred fifty-six people registered on the Web site,telephone keypad. At the end of the call, the written of whom 118 were randomized to the spiritual group, 117 toinstructions for the spiritual and affirmation groups were the affirmation group, and 121 to the neutral group. Of these,repeated as part of the ritual extension. All participants 277 registered on the automated telephone system, andwere reminded to take the essence twice daily, but those 251 people responded on at least one of the three finalin the spiritual and affirmation groups heard additionally: days of data collection (42 male, 201 female; mean age=37,
M.E. Hyland, B. Whalley / Journal of Psychosomatic Research 65 (2008) 405–413 409 Fig. 2. Mean symptom change (with 95% confidence interval) on Days 1 through 21.S.D.=11.9, range=18–66 years), of whom 87 were in the these are generally much lower that the correlations withspiritual group, 75 in the affirmation group, and 89 in the final outcome.neutral group. There was no significant difference between An inspection of Table 2 suggests that correlations withgroups in the numbers registering on the telephone system outcome differ between groups. To test whether this dif-(χ2=3.8, P=.15; all P values reported are two-tailed) or ference was significant, we performed a multiple regressioncompleting one of the final assessments (χ2=3.4, P=.18). analysis. We entered final outcome (converted to z scores) as We also examined outcome scores for the three the dependent variable, with spirituality (also converted toinstructions groups; mean scores (standard deviations in z scores), group (coded 0/1 for affirmation/spiritual groups),parentheses) for final outcome were: spiritual: 5.7 (1.2), and the spirituality×group interaction term as predictors.affirmation: 5.7 (1.0), neutral: 5.8 (1.4). Note, the point of There was no main effect of spirituality (β=−.09, P=.47),no change is 5, and higher scores indicate improvement. but the interaction term was significant (β=.37, P=.01),There was no significant difference in final outcome showing that the correlation between spirituality and out-between the groups [F(2,248)=.04, P=.96]. An equivalent come was significantly different between the spiritual andcomparison between initial outcome scores was also not affirmation groups.significant [F(2,248)=1.77, P=.17]. Because the baseline Motivation theory predicts that spiritually motivatedscores between the three groups were not identical, we people would become more involved with spiritual instruc-examined the residualized change scores by carrying out an tions. We tested whether our measures of ease of ritual and ofanalysis of covariance with the final outcome as the compliance could be considered measures of involvement bydependent variable, Week 1 scores as covariate, and group examining the correlations between spirituality, ease ofas a fixed factor. There was still no significant difference ritual, and compliance. For the spiritual group, spiritualitybetween the groups. Analysis of final outcome showed that, and ease of ritual were correlated, r(69)=.37, Pb.01, but theyfor the total sample, 122 (50.7%) people had improved, 88 were not correlated for the affirmation group; r(65)=.15,(39.8%) remained the same, and 21 (9.5%) deteriorated. P=.23. Compliance did not correlate with ease of ritual forBecause group did not affect overall outcome, we combined the spiritual or affirmation groups (ease of ritual was notthe data for all three groups to provide a picture of measured in the neutral group). These results suggest thatimprovement over time. For the total sample, Fig. 2 shows ease of ritual can be considered a measure of involvementthe mean outcome score for each day of measurement.Improvement is gradual during the first week, but there islittle further improvement after Day 7. Table 1 We next investigated predictors of change. Table 1 Correlations between baseline variables (n=356)shows the correlations between the baseline measures for Predictor 1 2 3the sample as a whole, and Table 2 shows correlations with 1. Spirituality – .43 ⁎ .28 ⁎baseline variables and variables that may be affected by 2. Expectancy – .04group, namely, final outcome, ease of ritual, and compli- 3. Optimism –ance. The correlations with initial outcome are not shown: ⁎ P b.01.
410 M.E. Hyland, B. Whalley / Journal of Psychosomatic Research 65 (2008) 405–413Table 2Correlations between predictors and final outcome, ease of ritual, and compliance in three experimental groups Neutral (n=89) Affirmation (n=74) Spiritual (n=87) OUT EASE COM OUT EASE COM OUT EASE COMSpirituality .22 ⁎ – .18 .09 .15 −.04 .34 ⁎⁎ .37 ⁎⁎ .13Expectancy .10 – .21 ⁎ .24 ⁎ .35 ⁎⁎ .14 .28 ⁎ .37 ⁎⁎ .08Ease of ritual – – – .34 ⁎⁎ – .07 .41 ⁎⁎ – .15Compliance −.05 – – .25 ⁎ .07 – .23 ⁎ .15 –OUT indicates final outcome; EASE, ease of ritual; COM, compliance. ⁎ Pb.05. ⁎⁎ Pb.01.linked to motivation, whereas compliance is not. Compliance As a final examination of behavior during the study, weis an indicator of behavioral engagement with the task but tested whether completers differed from noncompleters onwhere the engagement is due to factors other than motivation- baseline variables. We computed two new variables: Regis-induced involvement. The psychological mechanisms lead- tered (0/1), indicating whether a participant who completeding to compliance are unclear. Compliance correlated with baseline data online went on to register with the automatedexpectancy for the neutral group, r(92)=.22, P=.04, but not telephone system, and Completed (0/1), indicating whetherthe spiritual group, r(91)=.14 [nonsignificant (NS)], nor the the participant answered at least one of the final threeaffirmation group [r(80)=.04 (NS)]. telephone calls and was thus included in the analysis above. For each of the three groups, we tested (a) whether We then ran two separate logistic regressions, first withspirituality and expectancy contributed independently to Registered and then with Completed as the dependentoutcome and (b) whether ease of task and compliance variable. Spirituality, optimism, expectancy, and groupexplained significant additional variance when added to the (dummy coded such that the neutral group was the referencetwo baseline measures. Because ease of task completion, but category) were entered as predictors. Only Spiritualitynot compliance, could be caused by outcome, we carried out predicted study completion: for Registered, b=.27, P=.02;the multiple regression in three steps. For each group, we Completed, b=.23, P=.03. A follow-up analysis indicatedcarried out a multiple regression where final outcome was there was no interaction between spirituality and instructionthe dependent variable, traits spirituality and expectancy group. As a further test of the contribution of expectancy towere added in the first step, compliance was added in the completion, we entered expectancy by itself for each of the twosecond step, and (for spiritual and affirmation groups only) measures of dropout. In neither case did expectancy predictease of ritual was added in the third step. The results are dropout; for Registered, b=−.04; for Completed, b=.02.shown in Table 3. The first step of the analysis shows that, for the Table 3affirmation and spiritual groups, spirituality and expec- Multiple regressions for final outcome shown separately for the threetancy independently predict outcome, but the negative β groups, showing β and P values (in parentheses) for the predictor variables entered in two stepsin the affirmation group suggests that participants who arerelatively high in expectancy and low in spirituality have Predictors: Neutral group Affirmation group Spiritual groupbetter outcomes, whereas those low in expectancy and Step 1high in spirituality have worse outcomes. Thus, in the Spirituality .23 (.06) −.27 (.03) .35 (.01) Expectancy −.01 (.92) .33 (.01) .25 (.03)spiritual groups, the correlations between expectancy and Model R 2 adj=.03 Model R 2 adj=.10 Model R 2 adj=.23spirituality and outcome are consistent with previous data, F(2,86)=2.2, F(2,64)=4.7, F(2,68)=11.9,but in the affirmation group, a different pattern emerges. P =.12 P =.01 P =b.01There were no independent predictors of outcome in Step 2the neutral group—spirituality just missed significance Spirituality −.23 (.05) −.24 (.05) .31 (.01) Expectancy .01 (.98) .30 (.02) .23 (.04)at P=.06. Compliance −.09 (.39) .18 (.13) .23 (.03) The second step shows that, for the affirmation and R 2 change=.01 R 2 change=.03 R 2 change=.05spiritual groups, expectancy remains significant after con- F(1,85)=.75, F(2,63)=2.39, F(1,67)=4.64,trolling for spirituality and compliance. In Step 3, Expectancy P =.39 P =.02 P =.04is no longer significant after controlling for Spirituality, Step 3 Spirituality −.26 (.03) .26 (.03)Compliance, and ease of task completion. These results show Expectancy .20 (.10) .17 (.12)that the question of whether expectancy is mediated via Compliance .17 (.13) .21 (.05)behavior depends on the behavioral measures taken. Ease of ritual .30 (.01) .22 (.05)Compliance and Ease of task completion are indicators of R 2 change=.08 R 2 change=.04engagement with a task and do not necessarily capture all the F(2,62)=6.42, F(2,66)=3.90, P =.01 P =.06variance associated with task engagement.
M.E. Hyland, B. Whalley / Journal of Psychosomatic Research 65 (2008) 405–413 411 We included a measure of optimism to see if we could compliance is a valid behavioral measure, then responsereplicate earlier findings [16,39] that optimism predicts expectancy appears to contribute to outcome and, based on βoutcome for therapies where a positive expectancy is values, to a similar degree as motivational concordance.generated. For the sample as a whole, the correlation Expectancy predicted outcome after controlling for spiri-between optimism and perceived change was −.06 (NS). tuality and compliance for both the spiritual and affirmation groups (but not the neutral group). Of course, compliance does not capture the full meaning of task engagement, so thisDiscussion test favors response expectancy. Second, if we assume that both compliance and ease of task completion are valid In this study, participants engaged in a therapeutic ritual measures of behavioral engagement (i.e., ease of taskthat resembles real-life purchase and use of flower essences completion causes outcome but not vice versa), then itover the internet. Our study ritual differed from the real life would appear that only motivational concordance predictsritual only because participants did not pay for the essence, outcome. Expectancy did not predict outcome in any of thecompleted assessments, and were aware they were taking part three groups after controlling for spirituality, compliance,in a study. There was a gradual improvement in outcome over and ease of task completion.the first 7 days of the study. We found that the previously On the basis of the model of motivational concordancereported correlation between outcome and spirituality is not shown in Fig. 1, why did Spirituality predict independentlydue to flower essences per se but due to the way they are after controlling for Compliance and Ease of Task comple-contextualized as a spiritual therapy. When flower essences tion? We believe the reason is that Compliance and Ease ofare contextualized as a less-spiritual therapy (i.e., affirmation Task completion are not ideal indicators of engagement withgroup), then the previously reported correlation disappears, a task. Because motivation is so closely linked with taskas predicted by motivational concordance and consistent with engagement, Spirituality (i.e., the measure of value) explainsresearch on gratitude therapy . Thus, we have achieved additional variance in behavior and, hence, outcome, notthe first aim of the study: we have shown that the predictors of covered by Compliance and Ease of Task completion.placebo outcome are context dependent, as predicted by In our study, the neutral group acted as a control conditionmotivational concordance. We do not know what motive is for the enhanced rituals that were provided in the spiritualcongruent with affirmation therapy; however, spirituality is and affirmation groups. Comparison between the neutralnegatively related to values such as power , and we found group and other groups leads to two conclusions. First, thethat people who were low in expectancy and high in provision of additional information stabilizes the meaning ofspirituality did badly with affirmation therapy. We received the ritual—the slightly lower correlation with spirituality ininformal feedback from a participant in the affirmation group the neutral, as compared with the spiritual, group suggeststhat he felt he was being manipulated by the instructions. It that not everyone interprets flower essences as a spiritualmay be that some people, particularly spiritual people, therapy. The brief initial reference to spirituality when therespond badly to autosuggestive instructions. essence was presented online was not necessarily remem- A second aim of this study was to compare the relative bered, and we know from contacts with participants in pastcontribution of response expectancy (i.e., directly mediated studies that some people associate Bach flower remedieseffect of expectancy) versus motivational concordance (i.e., with pharmacologically active herbal remedies such as St.values and expectancy are mediated via behavior). First, we Johns Wort or Echinacea. Second, in the neutral group, thereconfirmed previous research that motivational concordance was no correlation between expectancy and outcome—inis a mechanism for placebo response. We found that contrast to the other two groups, previous flower essencespirituality, as well as two behavioral measures, predicted research , and many other studies. The lack of anoutcome independently of expectancy. Thus, the effect of expectancy correlation is surprising but adds to dataexpectancy is not only mediated directly—there are also suggesting that response expectancy may not be as importanteffects that appear to derive from the behavior of engaging in a mechanism for real-life placebo responses as it is inthe ritual. The answer to the question of the relative laboratory analogue studies [40,41].contribution of response expectancy versus motivational A possible criticism of our previous research is thatconcordance is complex. We examined whether expectancy expectancy and spirituality are not equally reliable measures,predicted additional variance for outcome when spirituality and so, the independent effect of spirituality on outcome orand behavioral engagement were taken into account. We the weak effect of expectancy on outcome is an artefactused two measures of behavioral engagement, but these created by the properties of the scales. This study provided ameasures were taken at different points during the study. more robust test of the motivational concordance hypothesisCompliance was measured during the treatment period and by examining whether the placebo effect was mediatedmight thus cause perceived benefit but cannot be caused by through behavior—i.e., whether “doing the ritual” was moreperceived benefit. Ease of task completion was measured at important than “believing in the ritual.”the end of the study, and this could both cause or be caused The correlations with expectancy and spirituality suggestby perceived benefit. First, if we assume that only that our measure “ease of ritual” reflects involvement in the
412 M.E. Hyland, B. Whalley / Journal of Psychosomatic Research 65 (2008) 405–413task, i.e., the extent to which people get involved in doing the predicted outcome, as suggested in previous researchritual. By contrast, compliance failed to correlate with [16,39]. Although flower essences generated positiveexpectancy and spirituality, suggesting that compliance expectancies, we failed to find a correlation betweenmeasures the degree of doing the ritual but without tapping optimism and outcome, despite a large sample size. Theseinto involvement. For both enhanced information groups results are consistent with our overall conclusion that(i.e., spiritual and affirmation), we found that compliance dispositional predictors of outcome depend on the contextand ease of ritual predicted outcome; expectancy failed to in which the placebo is taken.explain additional variance when ease of ritual completion There are several limitations to this study. First,was included as a predictor, consistent with the prediction of participants did not have a serious illness, though they self-motivational concordance but not of response expectancy. selected to treat a problem for which they perceived it worthThus, these data provide further evidence to suggest that engaging in the study to treat. Second, the problems treateddoing the ritual rather than believing the ritual may be the were heterogeneous, and it is possible that the contribution ofimportant factor for long-term placebo effects. different mechanisms varies with the psychological problem Analysis of dropout rates shows that higher spirituality at being treated. Third, we have no objective monitoring ofbaseline improved a participants chances of both entering behavior during the study, though we do have a subjectivethe assessment phase and completing the study. During the measure of compliance, and internet studies are a validprocess of obtaining consent flower essences were briefly method for collecting this type of data . Fourth, we useddescribed as a spiritually oriented therapy for all three a single measure of perceived change—though on severalgroups, and participants prior knowledge of flower essences occasions—rather than a before/after measure of outcome.is likely to have been concordant with this spiritual Fifth, measurement deficiencies of a single-item expectancyorientation. Thus, for all groups spirituality predicted measure may be responsible for the failure of expectancy todropout. By contrast, expectancy at baseline failed to predict predict independently of other variables. However, previousdropouts, consistent with the assumption that dropout research  has shown that spirituality predicts indepen-behavior is determined by motivational concordance. Other dently of multi-item expectancy measures. Sixth, ourresearch has shown that participants who fail to improve in measure of ease of ritual could be biased by the participantspsychotherapy tend to drop out , so it would seem in our experience of outcome. As noted above, this leads tostudy that dropout and improvement are, in part, driven by uncertainty in our data as to whether response expectancythe same mechanism, namely, motivational concordance makes a significant but smaller contribution to outcome orwith the therapy. whether it makes no contribution. Finally, it should be noted Despite different instructions given to the three groups, that our results may have no bearing on short-term placebothere was no overall difference in mean outcome. Although effects, especially short-term placebo analgesia studies,the purpose of our study was to examine correlations rather where expectancy appears to have a direct effect.than mean differences, a failure to find any difference In conclusion, our data show that placebo responders canbetween groups was surprising—even though there is a be identified but also that, consistent with other researchhistory of research showing that different therapies are , placebo responders vary with the therapeutic context.equally effective [43,44]. Despite the absence of an overall This research also suggests that motivational concordance isdifference, the three instructions were not equally effective the primary mechanism for long-term change in self-helpfor individuals, suggesting that the advantage of one type of therapies involving an inert substance. Whether responseritual for one person in a group is counterbalanced by its expectancy provides an additional contribution cannot bedisadvantages for another person. Our enhanced rituals (i.e., determined from our data. We do not know why engaging inthe spiritual and affirmation groups) provided the opportu- motivationally concordant rituals is so important—whethernity for greater involvement, but the greater specification the behavior of the ritual alters affect or self-perceptions (ormeans that it is uncomfortable for those for whom the ritual both) was not investigated in this study. However, our datais nonconcordant. By contrast, the neutral condition allows do suggest that placebos in real life cannot be understoodgreater flexibility for people to interpret the ritual so as to be only as a cognitive appraisal of expectancy of outcome.concordant with their motives. In sum, there are two possibleexplanations for the overall equivalence between the three Acknowledgmentsgroups. The first is that instructions influence the way peopleinterpret the ritual, but without instructions, people construct We thank Ainsworths for providing us with flowertheir own interpretation of the meaning of the ritual, and essences free of charge and without precondition.there are equal numbers of those who find any particular typeof instruction congenial or noncongenial. The second Referencesexplanation is that another more important therapeutic  Kirsch I, Deacon BJ, Huedo-Medina TB, Scoboria A, Moore TJ,mechanism is yet to be discovered. Johnson BT. Initial severity and antidepressant benefits: a meta- Although it was peripheral to the main aim of the study, analysis of data submitted to the Food and Drug Administration.we measured optimism at baseline to see whether optimism Plosmedicine 2008;5:260–7.
M.E. Hyland, B. Whalley / Journal of Psychosomatic Research 65 (2008) 405–413 413  Kirsch I, Moore T, Scoboria A, Nicholls S. The emperors new drugs:  Deci EL, Ryan RM. The “what” and “why” of goal pursuits: human an analysis of antidepressant medication data submitted to the U.S. needs and the self-determination of behavior. Psychol Inquiry 2000;11: Food and Drug Administration. Prevention & treatment. 2002;5: 227–68. Article 23. Available from http://www.journals.apa.org/prevention/  Connor TJ, Leonard BE. Depression, stress and immunological volume5/pre0050023a.html (accessed 5 November 2007). activation: the role of cytokines in depressive disorders. Life Sci  Kaptchuk TJ. The placebo effect in alternative medicine: can the 1998;62:583–606. performance of a healing ritual have clinical significance? Ann Intern  Maes M. Major depression and activation of the inflammatory Med 2002;136:817–25. response system. In: Dantzer R, Wollman EE, Virmiya R, editors.  Hyland ME. A tale of two therapies: psychotherapy and complemen- Cytokines, stress, and depression. New York: Plenum, 1999. pp. 1–24. tary and alternative medicine (CAM) and the human effect. Clin Med  Hoppmann CA, Klumb PL. Daily goal pursuits predict cortisol 2005;5:361–7. secretion and mood states in employed parents with preschool children.  Walach H, Jonas WB. Placebo research: the evidence base for Psychosom Med 2006;68:887–94. harnessing self-healing capacities. J Altern Complement Med 2004;10  Bem DJ. An experimental analysis of self-persuasion. J Exp Soc (Suppl 1):S103–12. Psychol 1965;1:199–218.  Stewart-Williams S, Podd J. The placebo effect: dissolving the expec-  Bem DJ. Self-perception: an alternative interpretation of cognitive tancy versus conditioning debate. Psychol Bull 2004;130:324–40. dissonance phenomena. Psychol Rev 1967;74:183–200.  Geers AL, Weiland PE, Kosbab K, Landry SJ, Helfer SG. Goal  Atkinson J, Reitman W. Performance as a function of motive strength activation, expectations, and the placebo effect. J Pers Soc Psychol and expectancy of goal attainment. J Abnorm Social Psychol 1956;53: 2005;89:143–59. 361–6.  Marino J, Gwynn MI, Spanos NP. Cognitive mediators in the reduction  Petry JJ, Finkel R. Spirituality and choice of health care practitioner. of pain: the role of expectancy, strategy use, and self-presentation. J Altern Complement Med 2004;10:939–45. J Abnorm Psychol 1989;98:256–62.  Bach E. Collected writings of Edward Bach. Hereford, UK: Bach  Devine DP, Spanos NP. Effectiveness of maximally different cognitive Educational Programme, 1987. strategies and expectancy in attenuation of reported pain. J Pers Soc  Walach H, Rilling C, Engelke U. Efficacy of Bach-flower remedies in Psychol 1990;58:672–8. test anxiety: a double-blind, placebo-controlled, randomized trial with Hyland ME, Geraghty AWA, Joy OET, Turner SI. Spirituality predicts partial crossover. J Anxiety Disord 2001;15:359–66. outcome independently of expectancy following flower essence self-  Armstrong NC, Ernst E. A randomized, double-blind, placebo- treatment. J Psychosom Res 2006;60:53–8. controlled trial of a Bach Flower Remedy. Complement Ther Nurs Hyland ME, Geraghty AWA, Whalley B. Dispositional predictors of Midwifery 2001;7:215–21. placebo responding: A motivational interpretation of flower essence  Kirsch I, Rosadino MJ. Do double-blind studies with informed consent and gratitude therapy. J Psychosom Res 2007. yield externally valid results? An empirical test. Psychopharmacology Moerman DE, Jonas WB. Deconstructing the placebo effect and 1993;110:437–42. finding the meaning response. Ann Intern Med 2002;136:471–6.  Schwartz SH, Boehnke K. Evaluating the structure of human values Kirsch I. Response expectancy as a determinant of experience and with confirmatory factor analysis. J Res Pers 2004;38:230–55. behavior. Am Psychol 1985;40:1189–202.  Lévi-Strauss C. The sorcerer and his magic. Structural Anthropology. Kirsch I, Hyland ME. How thoughts affect the body—a metatheore- New York: Basic Books, 1963. pp. 167–85. tical framework. J Mind Behav 1987;8:417–34.  Geers AL, Handley IM, McLarney AR. Discerning the role of Fisher S, Greenberg RP. The curse of the placebo: fanciful pursuit of a optimism in persuasion: the valence-enhancement hypothesis. J Pers pure biological therapy. In: Fisher S, Greenberg RP, editors. From Soc Psychol 2003;85:554–65. placebo to panacea: putting psychiatric drugs to the test. New York:  Thomas K, MacPherson H, Thorpe L, Brazier J, Fitter M, Campbell M, John Wiley, 1997. pp. 3–56. et al. Longer term clinical and economic benefits of offering Geers AL, Helfer SG, Kosbab K, Weiland PE, Landry SJ. acupuncture to patients with chronic low back pain. Health Technol Reconsidering the role of personality in placebo effects: disposi- Assess 2005;9:1–109. tional optimism, situational expectations, and the placebo response.  Lewith GT, Hyland ME, Shaw S. Do attitudes toward and beliefs about J Psychosom Res 2005;58:121–7. complementary medicine affect treatment outcomes? Am J Public Geers AL, Kosbab K, Helfer SG, Weiland PE, Wellman JA. Further Health 2002;92:1604–6. evidence for individual differences in placebo responding: an  Bradley R, Greene J, Russ E, Dutra L, Westen D. A multidimensional interactionist perspective. J Psychosom Res 2007;62:563–70. meta-analysis of psychotherapy for PTSD. Am J Psychiatr 2005;162: Moerman DE. Meaning, medicine, and the ‘placebo effect’. Cam- 214–27. bridge: Cambridge University Press, 2002.  Wampold BE, Mondin GW, Moody M, Stich F, Benson K, Ahn H. Brody H. The placebo response. New York: HarperCollins, 2000. A meta-analysis of outcome studies comparing bona fide psycho- Sheldon KM, Elliot AJ. Goal striving, need satisfaction, and therapies: empirically, “all must have prizes”. Psychol Bull 1997;122: longitudinal well-being: the self-concordance model. J Pers Soc 203–15. Psychol 1999;76:482–97.  Benish SG, Imel ZE, Wampold BE. The relative efficacy of bona fide Gollwitzer P, Wicklund R. The pursuit of self-defining goals. In: psychotherapies for treating posttraumatic stress disorder: a meta- Kuhl J, Beckman J, editors. Action control: from cognition to behavior. analysis of direct comparisons. Clin Psychol Rev [in press]. New York: Springer-Verlag, 1985. pp. 61–85.  Gosling SD, Vazire S, Srivastava S, John OP. Should we trust web- Maslow A. Motivation and personality. New York: Harper & Row, based studies? A comparative analysis of six preconceptions about 1970. internet questionnaires. Am Psychol 2004;59:93–104. Hyland ME. Control theory interpretation of psychological mechan-  Hyland ME, Whalley B, Geraghty AWA. Dispositional predictors of isms of depression: Comparison and integration of several theories. placebo responding: a motivational interpretation of flower essence Psychol Bull 1987;102:109–21. and gratitude therapy. J Psychosom Res 2007;62:331–40. Carver CS, Scheier MF. Origins and functions of positive and negative  Whalley B, Hyland ME, Kirsch I. Consistency of the placebo effect. affect: a control-process view. Psychol Rev 1990;97:19–35. J Psychosom Res 2008;64:537–41.