Memory for associations and order

点击翻译按钮获取中文摘要

From learning our way to work, to memorizing a script or notes of a melody, memory is essential for complex knowledge and behaviour. Understanding the cognitive and brain-activity basis of memory is therefore central to understanding naturally occurring behaviour and to designing effective artificial memories. This means remembering not just individual pieces of information, but also their relationships, the simplest of which are pairings (e.g., BRAD-ANGELINA), called "association-memory," or sequences (e.g., phone numbers, sequence of events in a story), called "order-memory." One intriguing idea is that the brain may carry out association- and order-memory in very similar ways. If this "common-processes" view were supported, that would allow us to understand a broader range of memory functions more concisely. There are two ways to build a common-processes model: 1) "Chaining" models learn an ordered list (ABC...) by linking together associations (A-B, B-C, ...). 2) "Positional coding" models learn a position-code for each item (A-1, B-2, C-3, ...), and can learn a set of pairs by linking paired items to a common position. Behavioural findings have both supported and challenged both models. Brain-activity measures (event-related potentials, ERPs, and rhythms) could help us compare models, but EEG studies have only begun to address order-memory. We suggest one model's weaknesses are the other model's strengths. Thus, people may use both mechanisms, depending on the situation. Combining behavioural, modeling and brain-activity (ERPs and rhythms) methods, we will build a common-processes model that uses chaining and positional coding, by testing how they trade off, in two ways: 1) We test memory of item-order within a pair (was it STEPHEN-HARPER or HARPER-STEPHEN?), which can test model mechanisms. 2) We will instruct participants to use popular strategies that resemble chaining ("Link Method") or positional coding ("Method of Loci") to identify which factors make chaining superior to positional coding, and vice-versa. This will reveal how brain activity may support relational memory across species and could enhance the design of artificial memories.