Expectations are studied to help explain experience quality but are often measured after the conclusion of an activity. Psychological researchers suggest that such recall may be inaccurate. To assess accuracy of recall, 120 Owyhee River boaters indicated their expectations concerning experience conditions (e.g., amount of people, litter) and internal states (e.g., happiness, boredom) on a pre-trip questionnaire. At the conclusion of the trip, they were asked to recall their original expectations. Results showed that individuals recalled most of their original expectations accurately (9 of 13 conditions; 12 of 12 internal states). Recall accuracy was equally high among commercial and private boaters, novices and repeat visitors, and boaters on different length trips. These findings suggest that measuring expectations after a trip is as valid as measurements taken prior to the trip.
This research was supported by the Bureau of Land Management, Vale District, Oregon. It is based on Shannon Dickson's Masters Thesis at the University of Idaho. The authors would like to acknowledge the assistance of Erin Seekamp for data collection and Craig Parks with statistical analysis. The assistance of the Vale District BLM is also appreciated.
Notes
1Analysis of change scores is debated in the psychological and statistical literatures (CitationMaxwell & Howard, 1981; CitationWilliams, Zimmerman, & Cummings, 1996). One concern commonly expressed is that a high correlation between pre- and post-tests may reflect memory for the pre-test rather than a treatment effect. In our case, it is precisely this memory that we hoped to assess. Several have also pointed out that change scores have lower reliability than either the pre- or post-treatment measures, particularly if all individuals exhibit approximately the same amount of change (CitationMiller & Kane, 2001). Additionally, use of a change score masks the typical correlation between the pre-test score and change (CitationGardner & Neufeld, 1987). To address these issues, some statisticians recommend analysis of covariance (ANCOVA), comparing post-test scores while including the pre-test as a covariate. ANCOVA thus corrects for the problem of regression to the mean (CitationTwisk & Proper, 2004). ANCOVA of post-test scores is appropriate where researchers are interested in the actual values on the post-test. In our case, we were interested in the change per se, and not the post-trip values. (In other words, we cared whether people could accurately recall their prior expectations, which would generate small change scores, and not about what precisely people expected.) Miller and Kane point out that in such cases, “the main goal is to assess how much each individual has changed in absolute terms rather than in comparison to some population” (p. 308). Thus, we chose to use change scores as the dependent variable rather than the post-test values, and we included pre-test values as covariates.
**p < .005.
***p < .0005.
**p < .005.
***p < .0005.