Running head : SERIAL COMBINATION 1 Ditta , A . S . , & Steyvers , M .

This article describes a new approach for studying collaborative memory that examines people’s editing processes for naturally occurring memory errors. In this approach, memories of individuals are combined via a chaining method in which each participant indirectly receives information from the previous participant. Participants were asked to individually study word lists and recall as many words as possible in an online setting. Once a participant completed the recall task, his/her answers were provided for the next participant as suggested answers for their own recall. However, that participant was allowed to add or subtract words from the provided list of suggested answers. The final answer of the group was an aggregate of recalled words based on the answer given by the last participant in the chain. Results showed that participants displayed a very high accuracy of recall throughout the chain, though they were not able to replicate the entire study list or eliminate all errors by the end of the chain. This procedure has the advantage that it allows examination of the memory-editing processes individuals utilize when they communicate information indirectly, independent from social factors that arise in face-to-face group memory settings. Serial Combination Procedure 3 In collaborative memory settings, people work together to retrieve information from memory. Many recent studies have investigated the cognitive and social effects of collaboration on memory (Roediger & McDermott, 2011; see Rajaram & Pereira-Pasarin, 2010 for an overview). In some cases, memory improves when information is recalled in groups. For example, Edwards and Middleton (1986) found that individuals in groups use information provided by others as cues for their own memories, thus adding to the overall number of items recalled. However, groups remember fewer items than when the same people recall separately and their recall is subsequently combined (Thorley & Dewhurst, 2007), a phenomenon called collaborative inhibition (Weldon & Bellinger, 1997). This could be because when people remember in groups, they are subject to social variables that may influence their performance ( e.g. social loafing; Latané, Williams, & Harkins, 1979), or because of cognitive factors such as retrieval interference (Weldon, Blair, & Huebsch, 2000). Another problematic effect of collaboration is the creation and persistence of false memories. A study by Gabbert, Memon, Allan, and Wright (2004) found that when people in groups encounter misinformation about an event, they are more likely to falsely recall this information later (when they would not otherwise have made these mistakes). In some situations, false memories are easily passed from one group member to another, leading to the “social contagion of memory” (Roediger, Meade, & Bergman 2001). For example, Roediger et. al. (2001) showed false memories implanted by a confederate are more likely to be accepted as truth (when compared to a control group with no false memory plant) when Serial Combination Procedure 4 they are more consistent with the scene being studied (e.g. a toaster in a kitchen, as opposed to oven mitts) and when participants had a limited amount of time to study the scene. It is important to note that typical memory studies of social contagion involve only two people: a confederate and a participant, or two subjects (see Gabbert, Memon, & Wright, 2006 and Meade & Roediger, 2002). Such studies that involve only two participants have traditionally shown that there are far fewer errors in collaborative groups when compared to nominal groups (Ross, Spencer, Blatz, & Restorick, 2008). However, larger groups might provide an advantage in recreating more accurate memories, as more points of view can be assimilated into the group memory. It is entirely possible that one participant catches a detail that several others do not, which would be missed in a group of only two people. Practical problems arise when studying group memory involving multiple individuals, especially when individuals can communicate freely among each other about their retrieved memory. This is because the large number of social and cognitive factors influencing group memory can make it difficult to isolate the factors of interest. We propose a simple group-memory procedure in which multiple individuals communicate serially in a chain. This allows us to examine the dynamics of memory errors over multiple individuals and situations where errors occur naturally, without being planted by confederates. Our approach is inspired by the seminal research by Bartlett (1932) on the serial reproduction procedure. His study was structured much the way that the game telephone is played: one person sees the target stimulus (the “target message”) and reconstructs it from memory, and then the next person receives this reconstruction as the target stimulus, studies it in order to reconstruct it again, and so on. He found that the final Serial Combination Procedure 5 reconstruction could be very different from the target message presented at the beginning of the chain, because each person introduced minor errors in each reconstruction, which then accumulated over each participant in the chain. The purpose of Bartlett’s study was only to examine how memories change over time, as opposed to how participants correct these errors. Ross et al. (2008) examined error correction in spouses, and found that they do engage in this sort of correction, but our procedure seeks to investigate these correction mechanisms in indirect collaboration and with larger group sizes. Our proposed method to study group memory does this by adopting some of the essential features of Bartlett’s serial reproduction procedure, but with one key difference. In the original procedure, each individual only receives information from the previous individual in the chain (or the set of target stimuli in case of the first individual). In our variant of this procedure, each individual in the chain receives the target stimuli that need to be remembered but also receives the reconstruction from the previous individual in the chain. Therefore, each individual in the chain (excluding the first individual) has the ability to rely on their own memory of the original information, as well as the retrieved memories from one other individual. Because the “truth” individuals are trying to recall is shown to every individual in the chain, individuals are theoretically capable of remembering the entire truth, and correct any mistakes that arise. This is very unlikely to occur in Bartlett’s serial reproduction procedure because only the first individual in the chain has access to the true answer. From here on, we will refer to our procedure as the serial combination procedure, to distinguish it from Bartlett’s procedure. Serial Combination Procedure 6 In a way, the social contagion studies also examine editing memory in a chain, but they employ “single shot” chains of two in which there is only one chance to accept or reject an answer. In these studies, one person deliberately feeds another person incorrect material, a memory test follows, and the “chain” is terminated. However, our approach will allow us to examine the editing process over time with multiple people, as errors introduced by one person in a chain will be either copied or corrected by subsequent individuals in the chain. Our design also has the additional benefit in that the errors will arise naturally – there is no confederate slipping false answers to see how people will react. Finally, our procedure has the advantage that it can be implemented in non-social situations. Each individual can observe the reconstruction from another individual without being in the presence of that person by utilizing an online transmission of information. This allows us to study the dynamics of the editing process in absence of social factors. Admittedly, our procedure stretches the definition of “collaboration,” as the first individual does not have access to anyone else’s answers, and there is no direct communication among the collaborators. Therefore, we are examining a more indirect type of collaboration as opposed to real-time face-to-face or online procedures (e.g. Ekeocha, & Brennan, 2008). Overall, the editing process in our serial combination procedure is related to the collaborative editing process in Wikipedia: any individual has the ability to delete and insert information in articles, and if something is blatantly wrong, someone else will delete the incorrect piece of information (e.g. Viégas, Wattenberg, & Dave, 2004). From time to time, people will add new bits of information, and if they are correct, they will remain in the article, while the additional incorrect information continues to be culled. Serial Combination Procedure 7 In this paper, we first describe an experiment where subjects study lists fashioned after the Deese-Roediger-McDermott paradigm and retrieve memories in the serial combination procedure. We then examine the types of editing processes that people utilize when correcting others’ responses, and measure overall performance with regards to correct recall and precision of words recalled as a function of the position in the serial chain