Knowing that you don’t know, and consistency in false memory James A. Hampton and Priya Gorasia Department of Psychology, City University London Abstract Student’s memories for details of a TV show were tested after 3 days and again after 5 days. A group who judged statements Definitely True, Uncertain, or Definitely False were more consistent across tests than a group who judged the statements simply True or False, implying the involvement of metamemory. Improved accuracy was however limited to True statements. The Rumsfeld Effect General Knowledge Using three options led to greater consistency across time There are statements which are “Known Unknowns” These attract stable Uncertain judgments Categorization, autobiographical memories, likes and aspirations Using three options is no more consistent than using two There are only “Unknown Unknowns” Results: Consistency of memory A B 3 2 1 1 4 3-option group more consistent than the 2-option group for both True and False statements. “Uncertain” responses for the 3-option group were as consistent as Definitely True responses (around 80%) – the “known unknowns” False statements (red) judged as “Definitely” True remained firmly in memory with near 80% consistency But True statements (blue) judged as Definitely False were unstable on retest, compared with False statements correctly judged Definitely False Calculating consistency For the 2 response case, consistency is the likelihood of giving the same response on each occasion For the 3 response case, consistency is the mean of the consistency of a Definite Yes the consistency of a Definite No Second response First response Definitely YES MAYBE Definitely NO The yellow cells are counted as Consistent for a Definite Yes Response. A Definite No uses the equivalent cells based on the lower right corner Metamemory and the Rumsfeld Effect Hampton et al. (2011): a new method for testing the reliability of judgments of truth/falsity of uncertain or vague statements Sociology is a Science – True or False? GROUP A: judged statements with a binary choice: TRUE or FALSE GROUP B: judged statements with three options: DEFINITELY TRUE UNCERTAIN DEFINITELY FALSE Group B were able to choose only the “easy” statements for judging true or false, and any statements that were doubtful could be left as uncertain. Group A had to guess if they were not confident of the answer. How reliable would people be when tested in the same way a week later? Does the option of responding only to the obviously clear statements lead to a greater consistency across time? Or does the metacognitive difficulty of deciding when a statement’s truth is uncertain or not lead to no improvement in consistency? A number of different domains were studied: General knowledge: the Uruguayan flag has red in it Categorization in semantic domains: a tomato is a fruit Autobiographical memories: I once was on a bus when it broke down Likes and dislikes: I like peanut butter Aspirations: I would like to meet the Queen Accuracy Testing certainty for memories Will judgments of event memories show greater consistency in the 3 response condition that includes an Uncertain response? Will the 3 response condition give more accurate responses for those statements judged Definitely True or False? How stable are highly confident but false memories? Participants: 59 students aged 17-18 at a London college Procedure: Watched a clip from a TV show. Memory tested after 3 days and retested after 5 days, using a list of 40 statements. Monk washed his toothbrush with soap Design: A) Number of Response Options – between subjects Two response options: True/False Three response options: Definitely True, Uncertain, Definitely False B) Statement Truth 20 true and 20 false statements Test/Retest Each statement was tested after 3 days, and again after 5 days the 3-option group were more accurate than the 2-option group for True statements, t(57) = 7.4, p < .001, but not for False statements t(57) = 1.2, NS Conclusions Allowing an “Uncertain” response in this task leads to improved consistency and improved accuracy for True statements judged Definitely True Even when definitely certain of a response, stability was only 80% after a 2 day delay False Positives were very stable, but False Negatives were not True statements benefited in accuracy from the 3-options, but False statements did not. Reference: Hampton, J.A., Aina, B., Andersson, J.M., Mirza, H., & Parmar, S. (2012). The Rumsfeld Effect: the Unknown Unknown. Journal of Experimental Psychology: Learning Memory & Cognition, 38,340-355. Contact: hampton@city.ac.uk Web: www.staff.city.ac.uk/hampton