Bound context features are integrated at encoding.

In the present research, we tested the predictions of different accounts of context binding in "remember" judgements. Context binding, defined as the stochastically dependent retrieval of two different context features, has previously been suggested to be due to mechanisms operating at retrieval either by cueing among context features (Meiser & Bröder, 2002) or by cueing between item and context features (Starns & Hicks, 2008). These accounts, however, do not make specific assumptions regarding the underlying memory representation supporting context binding. By contrast, here we propose that a binding process at encoding integrates item and context information into a coherent memory representation. Varying the presentation of the context features during encoding either with both context features presented simultaneously or with features spread over two encoding episodes, data from two experiments corroborate the notion that binding is produced at encoding. This result suggests that a binding process integrating the context features at encoding is necessary for stochastically dependent retrieval of context features.