Order information and free recall: Evaluating the item-order hypothesis

Abstract

The item-order hypothesis proposes that order information plays an important role in recall from long-term memory, and it is commonly used to account for the moderating effects of experimental design in memory research. Recent research (Engelkamp, Jahn, & Seiler, 2003; McDaniel, DeLosh, & Merritt, 2000) raises questions about the assumptions underlying the item-order hypothesis. Four experiments tested these assumptions by examining the relationship between free recall and order memory for lists of varying length (8, 16, or 24 unrelated words or pictures). Some groups were given standard free-recall instructions, other groups were explicitly instructed to use order information in free recall, and other groups were given free-recall tests intermixed with tests of order memory (order reconstruction). The results for short lists were consistent with the assumptions of the item-order account. For intermediate-length lists, explicit order instructions and intermixed order tests made recall more reliant on order information, but under standard conditions, order information played little role in recall. For long lists, there was little evidence that order information contributed to recall. In sum, the assumptions of the item-order account held for short lists, received mixed support with intermediate lists, and received no support for longer lists.

Notes

¹Although the correlations between recall and IO scores were not significant, it should be noted that, in the case of EPTs, the correlation between recall and other order measures approached significance (which was not the case for SPTs).

²As noted, the item-order account assumes a role for serial-order information in free recall. More specifically, it is assumed that participants spontaneously use order information in typical experiments on recall (i.e., in single-trial recall from long-term memory). With this in mind, it should be noted that some of the studies cited as supporting the order-based recall assumption had one or more of the following features: (a) multiple study–test trials in which the same study materials were presented repeatedly; (b) method of incrementing presentation of the study list (in which a new item is added to the list on each presentation; Mandler & Dean, 1969); and (c) immediate recall tests (e.g., Kintsch, 1970; Mandler, 1969; Mandler & Dean, 1969; Postman, 1972; Tulving & Patkau, 1962). Multiple recall tests (with constant order of study presentation) and the method of incrementing presentation both induce a reliance on order information in recall (e.g., Mandler, 1969; Mandler & Dean, 1969; Tulving & Patkau, 1962). Immediate recall (especially of short lists, as was typical in these studies) permits short-term memory influences, known to rely on order information (e.g., Healy, 1974; Jahnke, 1965, 1968; Postman & Phillips, 1965). Consequently, although these studies indicate that order information plays a role in recall under certain circumstances, these studies do not provide unambiguous support for the more specific, central assumption of the item-order account—namely, that order information plays a role in recall from long-term memory within a single study–test trial. Secondly, other studies cited as supporting the order assumption (Toglia & Kimble, 1976; Tzeng, Lee, & Wetzel, 1979) demonstrate that order information may be encoded for common, unrelated items, indicating that this information is potentially available to aid recall. However, these studies do not demonstrate that the order information is actually used in free recall.

³Although the IO score is commonly used to assess order information in this experimental paradigm, Engelkamp et al. (2003) argued that these scores may be insufficiently sensitive. To assess the generality of our results, we evaluated order information using an alternative measure used by Engelkamp et al. (2003): bidirectional pair frequency (PF; Sternberg & Tulving, 1977). All analyses using IO score in this and subsequent experiments were duplicated using the PF measure. These analyses produced the same pattern of significant effects and correlations, with one exception noted later. In general, the results and conclusions are not materially affected by the choice of IO or PF as the measure of order information in free recall.

⁴We place more emphasis on the correlations between order usage (i.e., IO scores) and free recall than on the absolute values of the IO scores themselves. Some studies have focused on IO scores (e.g., McDaniel et al., 2000; Nairne et al., 1991) whereas others have focused on the FR–IO (or comparable) correlation (Engelkamp et al., 2003; McDaniel et al., 1995). We take the latter tack because the item-order account assumes not just that order information is reflected in retrieval order but that using order information contributes to recall. Consequently, although above-chance IO scores are implied by the item-order account, this by itself does not compel us to conclude that order information contributes to recall. Positive IO–free-recall correlations are more completely consistent with the item-order account and provide a stronger basis for the acceptance of its assumptions.

⁵As described earlier (Footnote 3), order information in recall was assessed with both the IO score and the PF measure. The analyses were consistent with one exception. In the FR-order condition, the correlation between PF scores and free recall was significant, r(30) = .56, in contrast to the nonsignificant FR–IO correlation.