False memories and the DRM paradigm: effects of imagery, list, and test type

ABSTRACT

Several researchers have reported that instructing participants to imagine items using the Deese-Roediger-McDermott (DRM) paradigm lowers false memory rates (Foley, Wozniak, & Gillum, 2006). However, other researchers have found that imagery does not always lower false memory rates (Robin, 2010), and investigators have examined the effects of imagery manipulation on semantic but not phonological lists. In the present study, we presented 102 participants with semantic and phonological DRM lists, followed by a free recall test and final recognition test. Some participants received instructions to imagine list items during the study phase to facilitate memory, and others were simply told to remember list items. Imagery instructions enhanced correct memories and further suggested a trend for decreased false memories. A test type by list type interaction also emerged, with phonological lists eliciting higher false memories at recall, and semantic lists eliciting higher false memories at recognition. Directions for future research are discussed.

Keywords:

• DRM
• false memories
• false recall
• false recognition
• imagery
• phonological word lists
• semantic word lists

Author notes

Merrin Oliver is earning her PhD in Educational Psychology at Georgia State University. She is an active doctoral fellow and teaches undergraduate classes centered around memory and human development. Her research focuses on false memories, testing effects, and the interaction of different learning strategies and student-level variables, like prior knowledge. Rebecca B. Bays is currently an assistant professor of cognitive psychology at Skidmore College. Her research investigates the effects of imagination on the veracity of memories for past personal experiences. Dr. Karen M. Zabrucky is a professor of Educational Psychology at Georgia State University, where she has been since 1986. In her research, she focuses on selected aspects of applied cognitive psychology, (e.g., memory, metacognition and discourse comprehension in the real world). She has received a university-wide award for excellence in the three areas of teaching, research and service and has received funding from several external grant agencies, including a five-year grant from the National Institute on Aging.

Notes

1. A Multivariate Repeated Measures ANOVA on correct memories replicated results reported in the main text: significant multivariate main effects of test type, λ = .41, F(1, 100) = 143.97, p < .001, ŋρ² = .59 and list type, λ = .63, F(1, 100) = 143.97, p < .001, ŋρ² = .59, qualified by a significant interaction between test and list, λ = .83, F(1, 100) = 143.97, p < .001, ŋρ² = .59. Tests of simple effects revealed that both phonological and semantic lists elicited higher correct recognition (Ms = .89 and .92, respectively) than correct recall (Ms = .74 and .83 respectively). No other interaction effects were significant, Fs < 1. The univariate main effect of imagery was also significant, F(1, 100) = 5.90, MSE = .02, p = .02, ŋρ² = .06. Imagined lists (M = .86) elicited more correct memories than lists not imagined (M = 83).

2. A Multivariate Repeated Measures ANOVA on false memories replicated results reported in the main text: significant multivariate main effects of test type, λ = .33, F(1, 100) = 143.97, p < .001, ŋρ² = .59 and list type, λ = .92, F(1, 100) = 143.97, p < .001, ŋρ² = .59, qualified by a significant interaction between test and list, λ = .64, F(1, 100) = 55.36, p < .001, ŋρ² = .36. Tests of simple effects revealed higher false memory rates for phonological than semantic lists during recall (Ms = .14 and .06, respectively), and higher false memory rates for semantic than phonological lists during recognition (Ms = .48 and .28, respectively). A univariate main effect of imagery approached significance, F(1, 100) = 3.46, MSE = .10, p = .066, ŋρ² = .03. Imagined lists elicited fewer false memories (M = .21) than lists not imagined (M = .27). No other interaction effects were significant, all Fs < 1.