Has odor conditioning been demonstrated? A critique of “Unconscious odour conditioning in human subjects”

Has odor conditioning been demonstrated? A critique of “Unconscious odour conditioning in human subjects”

BIOLOGICAL PSYCHOLOGY ELSEVIER Biological Psychology Technical 37 (1994) 265-267 Note Has odor conditioning been demonstrated? A critique of “Un...

201KB Sizes 2 Downloads 45 Views

BIOLOGICAL PSYCHOLOGY ELSEVIER

Biological

Psychology

Technical

37 (1994) 265-267

Note

Has odor conditioning been demonstrated? A critique of “Unconscious odour conditioning in human subjects” Stephen L. Black I, David G. Smith 2 Department of Psychology, University College of Swansea, Swansea SA2 8PP, UK

Key words:

Odour

conditioning;

anxiety

A study was reported in this journal (Kirk-Smith, Van Toller, and Dodd, 1983) which reached the novel and theoretically interesting conclusion that anxiety could be evoked in female subjects by an odor which the subject was unaware had been previously paired with the emotion. This finding of “unconscious odour conditioning” has been described as “important” (King, 19881, a characterization supported by citations to it in at least 21 articles, including six in 1992. As interest in the relation of odor to emotion grows (see Ehrlichman and Bastone, 19921, it is reasonable to expect that this study will continue to be highly cited. Unfortunately, an examination of the experiment indicates that it is flawed in design and analysis, and its conclusion therefore unsound. The purpose of this note is to bring these unrecognized problems to the attention of researchers. The design of the study was to require two groups of subjects to work at a task which was presumed to be anxiety inducing, one group in the presence of the conditioning odor. Several days later, the mood of the subjects was assessed before and after exposure to the odor. However, all of the subjects were initially required to work on the same stressful task, and all were later exposed to the same odor in the test phase of the experiment. The experiment therefore lacked appropriate controls to determine whether the reported increase in anxiety could have oc-

’ Correspondence should be addressed to Stephen L. Black, Department of Psychology, Bishop’s University, Lennoxville, Quebec, Canada JlM 127. ’ Present address: Department of Psychology, Queen’s University, Kingston, Ontario, Canada K7L 3N6. 0301-0511/94/$07.00 0 1994 Elsevier SSDI 0301-0511(94)00937-S

Science

B.V. All rights reserved

266

XL. Black, D.G. Smith /Bdogical

Rychology

37 (1994) X5-267

curred without prior stress or without exposure to the odor in the test phase. In either case, conditioning would not then have been demonstrated. Moreover, no evidence was provided that the task intended to induce anxiety was effective in doing so. As the task was to assemble block patterns and was administered for obvious experimental purposes to sophisticated subjects (postgraduates and staff), it seems unlikely that it was. The statistical analysis of mood changes before and after odor presentation is also problematic. Despite an interest in the two variables of sex and odor, for which a two-way ANOVA would have been appropriate, a one-way ANOVA was carried out. There were four groups: female/odor present, female/no odor present, male/odor present, and male/no odor present (the labels used by Kirk-Smith et al. refer to odor conditioning during the initial phase of the experiment). A significant F ratio was reported. It was then asserted as a post hoc analysis without statistics that “the female group with odor present showed increased “anxiety’ ratings through the session”, presumably in relation to the female group with odor absent. As the results were nevertheless reported in sufficient detail to ahow the calculations to be made, we did so using the Tukey HSD test (Kirk, 1982) and a significance level of p = .0.5. This indicated that the mean changes in mood of the female/odor present group did not differ significantly from that of the female/no odor present group. Thus the analysis provides no support for the conclusion that odor conditioning had been achieved for the female subjects. An alternative and preferable means of analysis, however, would be a planned test of the original hypothesis that all subjects (male and female) conditioned to the odor would show increased anxiety scores in the test phase compared with subjects not conditioned. unfortunately, Kirk-Smith et al. did not make this comparison. The analysis is also flawed in another way. Although it is clearly stated that the subjects were 12 male and 12 female subjects, thus giving a group size of n = 6 for each of the four groups, Table 2 states in contradiction that “FZ= 48/group”. A group size this large would result in 188 degrees of freedom for the error term, and this value was in fact used to calcuiate the mean square error (the tabled value of 186 df appears to be a misprint). The only possible explanation for this striking contradiction is that each of the eight responses every subject made on the anxiety questionnaire was treated as though it was a separate independent observation from a different subject, thus improperly inflating the number of subjects in each group. That this must be the case is suggested by the otherwise cryptic comment “the eight scales comprising each factor were equally weighted and included separately in the appropriate ANOVA.” Similar considerations may explain why the error degrees of freedom for the “hostility” factor rise to 764, implying the use of 192 subjects per group, a rather unlikely number. These considerations Iead to the following conclusions regarding the study of irk-Smith et al. (1983): (a) the design of the study is inadequate to determine whether conditioning has taken place; (b) their data as presented fail to demonstrate that “female subjects who had experienced.. . odour in the stress condition showed an increase in anxiety ratings”; (c) the data have been employed in an

S.L. Black, D.G. Smith /Biological Psychology 37 (1993) 265-267

analysis of variance in an incorrect manner. Thus their conclusion conditioning can take place in human subjects is unfounded.

267

that odor

References Ehrlichman, H., & Bastone, L. (1992). Olfaction and emotion. In M. Serby & K. Chobor (Eds.), The science of olfaction (pp. 410-438). New York: Springer. King, J. (1988). Anxiety reduction using fragrances. In S. Van Toiler and G. Dodd (Eds.1, Perfumery: fhe psychology and biology of fragrance. London: Chapman and Hall. Kirk, R. (1982). ~peyj~e~~al design: procedures for the be~a~.ioral sciences (2nd. edn.1. Monterey, CA: Brooks/Cole. Kirk-Smith, M., Van Toiler, C., & Dodd, G. (1983). Unconscious odour conditioning in human subjects. Biological Psychology, 17, 221-231.