Foraging memory: Retrieving words from one and from two semantic categories

Traditionally applied to an animal's search for food, the concept of foraging has been extended to include the search for information in such places as the Internet and libraries (Pirotti & Card, 1999). The premise behind the research reported here is that memory searching can also be construed as foraging. The goals of this investigation are to uncover mental factors that may affect memory production during memory search and to use this knowledge to guide a prediction of foraging production. Prior to testing, four such mental factors were identified: a time cost when producing an initial item from a different category (switch cost); a production benefit driven by a release of proactive interference (time-out benefit); a production cost caused by the additional mental load of executing an autonomous switching strategy (executive-decision cost); and sub-optimal allocation of time between categories. Experiment 1 tested whether switching between categories leads to a switch cost and/or time-out benefit by having subjects produce items from a category in a continuous three-minute block or multiple blocks that add to three minutes. Experiment 2 addressed the possibility of an "Executive-decision" cost by either allowing subjects to autonomously switch between categories or yoking them to another subject's switch schedule. Experiment 3 tested whether memory foragers divide their time optimally between categories.;Data from the first experiment demonstrated that like external foraging, moving from category to category (patch to patch) results in a production downtime. These data also demonstrated that switching production between domains may lead to a time-out benefit. Experiment 2 showed that the execution of an autonomous switching strategy leads to less production then when switching is forced. The third experiment demonstrated that, unlike animals, humans do not have an innate sense of how to divide their time between patches (categories) to maximize gain.;Our prediction was derived by having subjects produce category exemplars from a single category alone or from two categories at once. Data from single-category production trials as well as adjustments inspired by Experiment I through 3 were used to predict production from two categories. Though accurate, the flexibility of our prediction is limited. Research needed to allow for greater flexibility is discussed.