Figure 2. Contextual retrieval enables the extraction of global structure from isolated episodes.

a. Miniature example of a small-world network with connectivity chosen according to the structure of English as estimated by Steyvers & Tenenbaum (2005). b. The cue strength between pairs chosen from the small-world network are shown as a function of the shortest path between the items. Filled symbols show TCM with contextual retrieval. The open symbol shows the value for TCM without contextual retrieval. Only one point is shown for the model without contextual retrieval because the cue strength is zero for items not presented as part of the same pair.