Intrusions and the Decision to Terminate Memory Search

Intrusions and the Decision to Terminate Memory Search J. Isaiah Harbison (isaiah.harbison@gmail.com) Department of Psychology, University of Maryland at College Park College Park, MD 20742 Eddy J. Davelaar (eddy.davelaar@gmail.com) Departiment of Psychological Sciences, Birkbeck University of London Malet Street, WC13 7HX, London, UK Erica C. Yu (erica.c.yu@gmail.com) Erika K. Hussey (erikahussey@gmail.com) Michael R. Dougherty (mdougher@umd.edu) Department of Psychology, University of Maryland at College Park College Park, MD 20742 Abstract termination (exit latency or E L ; Dougherty & Harbison, 2007). These variables allow for the testing of different memory search stopping rules previously proposed in the literature, with much of the available data uniquely sup- porting the cumulative-failures stopping rule (Harbison, Dougherty, Davelaar, & Fayyad, 2009) proposed within the search of associative memory model (SAM; Raaij- makers & Shiffrin, 1981). According to this rule, every retrieval attempt that does not produce a new retrieval is counted as a retrieval failure and search is terminated when the number of these failures reaches a threshold. Recent research, however, has suggests that search ter- mination might also be influenced by the presence of memory intrusions (Miller, Weidmann, & Kahana, 2012; Unsworth, Brewer, & Spiller, 2011). Miller et al. (2012) showed that memory search was more likely to be ter- minated after an intrusion from a previous list (prior list intrusion or PLI), an extra list intrusion (ELI), or after outputting a list word that had previously been re- trieved (repetition). They suggested that the increase in the probability of stopping may be due to the effect such retrieval errors have on subsequent recall. Each retrieved word is thought to influence subsequent retrievals either by the use of the retrieved word as a cue for subsequent retrieval attempts (Kimball, Smith, & Kahana, 2007; Sirotin, Kimball, & Kahana, 2005) or by the retrieved item shifting the contextual retrieval cues closer to the retrieved items own context (Howard & Kahana, 2002). Intrusions and repetitions then would decrease the prob- ability of retrieving a new target list word. A PLI would increase the relative probability of another word from the prior list being retrieved; an ELI would increase the probability of sampling related extra-list words, and rep- etitions would increase the probability of retrieving other words that have already been retrieved. One potential problem with the Miller et al. anal- ysis is that they had to infer when participants termi- nated search since the experiments they used in their analysis used a closed-interval design. To determine Little is known about the how the decision is made to terminate memory search, though there have been sev- eral recent attempts to uncover this process. In one recent study, Miller et al. (2012), re-analyzed data from a large number of free-recall experiments and identified intrusions as a factor that influenced search termination decisions. One potential problem with this re-analysis is that all the data were drawn from experiments in which it was impossible to determine if or when search was ter- minated. Using data from experiments in which search termination decisions were directly measured, we con- firmed Miller at al.’s (2012) original findings but also demonstrated that intrusions influence the time taken to generate the final retrieval and the time between the final retrieval and search termination. The pat- tern of data is consistent with a simple, sample-with- replacement model in which intrusions are less active than items from the target list. Keywords: recall; memory search termination; stop- ping rules Every search of memory is eventually terminated. When an individual decides to terminate their own memory search (e.g., when they are not interrupted or given a fixed time limit for search), what factors influence this decision? The long history of memory research is rel- atively silent regarding this question as most memory recall experiments give participants a pre-determined amount of time to search memory (a closed-interval) and have no method of determining when or, even if, partic- ipants terminate their search before the retrieval inter- val expires. When participants are allowed to terminate their own search, as is the case in the open-interval de- sign discussed below, a number of dependent variables emerge that allow the measurement of memory search termination decisions. These variables include the total time spent in search (total time or T T ), which is con- trolled by the participant in this design. The total time can be divided into the time from the beginning of search to the time of the final retrieval (time-to-last retrieval or T L ) and the time between the final retrieval and search