When remembering a past event, we typically cannot pluck a whole memory neatly from our brains, as we would a folder from a filing cabinet. The individual features comprising a memory are distributed throughout our brains, so remembering a past experience is more similar to assembling a jigsaw puzzle; all the pieces must be located and assembled correctly to construct a coherent memory.
This constructive memory system provides a flexibility that is generally adaptive (Schacter, Guerin, and St Jacques 2011), in that we can update memories with new information (Lee 2009), and assemble memory fragments in novel ways to imagine events that have not yet taken place (Schacter and Addis 2007). Yet the downside to this system is that it is prone to mistakes. Pieces can go missing, resulting in blank spaces in the memory (which may be filled in with schematic knowledge). Another mistake occurs when a piece of one memory trace is erroneously incorporated into another (Burt, Kemp, and Conway 2004), forming a memory conjunction error (Odegard and Lampinen 2004).
In our recent study we were interested in illuminating the factors that contribute to the formation of these conjunction errors in autobiographical memory (that is, memory for personal life experiences). Inducing enough errors in autobiographical memory to allow empirical study is challenging, due to the complex and deeply personal nature of the memories. We used a novel recombination paradigm to generate autobiographical memory conjunction errors in a laboratory environment, so that we might shed some light on the conditions under which these errors come to be accepted as authentic memories.
In our three session paradigm, participants first recall a number of autobiographical memories, specifying a person, place and object involved in each. For example, participant X might remember 'going to the cinema to see Mission Impossible with Susan, and eating popcorn', as well as 'watching Cats at the Civic theatre with Jill, and drinking a frozen coke'. We then take these memory details and recombine them across events to form recombined detail sets (such as: Susan, Civic theatre, frozen coke). In the second session, participants either imagine a novel past event involving these recombined details, or, as a control for exposure to the lures, rank the details by how pleasant they feel. In the third and final session, participants determine the source of original and recombined detail sets, then describe the associated memories and rate them for phenomenological detail (such as vividness and emotion). A conjunction error occurs when a recombined set is falsely recognised as an authentic memory: for instance, participant X might now claim she had gone to see Cats with her friend Susan.
We found support for the source monitoring account of memory errors, which suggests that false memories occur when they have phenomenological characteristics similar to that of authentic memories, and are thus confused as such. Conjunction errors were more likely to occur when the detail substitution was plausible, and when only a single detail was altered. An imagination inflation effect (Garry and Polaschek 2000) was prevalent, where imagining a novel past event in response to a recombined set resulted in more conjunction errors than simply being exposed to the recombination of details. Imagining a conjunction event increases the phenomenological richness of the false event, making it more memory-like in quality, and therefore more difficult to distinguish from authentic memories of real events. Indeed, generating a highly vivid and plausible event during imagination further increased the probability of a conjunction error later forming. Furthermore, when retrieving memories, conjunction errors were rated as similar in vividness and emotion as authentic memories, and descriptions of both types of memories contained a similar amount of perceptual detail.
While the majority of conjunction errors that occur in day-to-day life will pass by unnoticed, conjunction errors in autobiographical memory can cause unique difficulties in establishing the authenticity of eyewitness accounts. Indeed, it has previously been shown that spectators can confuse the perpetrator of a crime with innocent bystanders (Kersten, Earler, Curtain, and Lane 2008). Our results indicate that witnesses may be particularly vulnerable to these conjunction errors if they are endowed with recollective quality, as might happen during discussions with other witnesses (Gabbert, Memon, and Allan 2003), or even during interrogation itself (Kassin, Gisli, and Gudjonsson 2004). We have shown that because the elements of a conjunction event have been truly experienced, albeit in a different combination, conjunction errors may be highly compelling for the rememberer.