Conventional accounts of memory assume that associations among memorized items will cue recall (e.g., Anderson, 1983; Ebbinghaus, 1885/1913; Raaijmakers & Shiffrin, 1981) and the benefits of cues in aiding memory are well established (e.g., Eich & Metcalfe, 1989; Godden & Baddeley, 1975; Tulving & Pearlstone, 1966; Tulving & Thomson, 1971). At some level, remembering must depend upon cues because for memory to aid us, specific relevant information has to be retrieved and reconstructed from among the myriad possible memories. However, the extent to which cues from within a set of learned material aid recall was brought into question when part-set cuing impairment was observed, initially by Brown (1968) and Slamecka (1968), for recall from both episodic and semantic memory. The effect proved to be robust for the free recall of word lists and was then demonstrated across a range of materials and retention tests (for reviews, see Cole, Reysen, & Kelley, 2013; Muntean & Kimball, 2012; Nickerson, 1984). The purpose of the experiments reported here was to further extend the investigation of part-set cuing to two different types of experimental material: expository text and pictorial scenes.
Part-set cuing involves tests accompanied by a subset of the material from the set that is to be recalled, providing the opportunity for this subset to cue recall of the rest of the material. Since Aristotle (Morris, 1994), people have believed that our memory is, at least in part, guided by associations between items, with these associations depending upon their contiguity in time and their similarity in meaning. Because the cues are a subset of the tested material, they are at least weakly associated with the other items in that set. Through these associations, cues could be expected to facilitate the recall of the other items. A common outcome, however, is that recall in the absence of these cues is often better than recall in their presence. This pattern arises with both newly studied sets (e.g., Slamecka, 1968, 1969) and members of semantic categories such as the states of the United States of America (e.g., Brown, 1968).
Part-set cuing impairment seems inconsistent with very long-standing views about the nature of memory, associations, and cuing. If members of a set have inter-item associations and these associations can act as cues for recall, why does providing part of a set fail to facilitate recall of additional items from that set? Slamecka (1968, 1969) concluded from his demonstrations of the failure of part-set facilitation that this traditional view was incorrect and proposed that the words learned in free recall studies are not stored associatively, but are stored independently.
Early reports of part-set cuing impairment stimulated considerable research and theorizing, well reviewed by Cole et al. (2013), Marsh, Dolan, Balota, and Roediger (2004), Muntean and Kimball (2012), and Nickerson (1984). Brown's (1968) and Slamecka's (1968) findings have been extensively replicated and extended and the effect has become better described. One approach to understanding the phenomenon is to identify its limits – documenting when part-set cues facilitate rather than impair recall. For example, part-set cues can facilitate the recall of categorized lists, but only where they cue the recall of categories that would otherwise have been forgotten; recall within remembered categories is not aided by the cues (e.g., Hudson & Austin, 1970). Serial recall is another situation in which part-set cues can be beneficial. Serra and Nairne (2000) found that serial reconstruction of word lists was facilitated by presenting part-set cues in their original list positions but impaired when the cues were presented in different positions. Basden, Basden, and Stephens (2002) extended this research to serial and free recall: Serial recall was sometimes facilitated by consistent cues but impaired with inconsistent cues; for free recall there was little influence of cuing condition. This research suggests that in serial recall the associations between consecutive items are important, and when alternating items (i.e., odd or even positions) are provided as cues these inter-item associations can facilitate recall (as in Kelley & Bovee, 2007). When Cole et al. (2013) examined part-set cuing of spatial information using Snap Circuit objects, the cues facilitated reconstruction of the circuits.
Part-set cuing impairment (or lack of facilitation) is not limited to memory. Peynircioğlu (1987) demonstrated effects that resembled part-set cuing impairment in a range of non-memory tasks, including: constructing words from a lengthy word, identifying differences between pictures, recognizing blurred pictures, and making sense of nonsense figures. Similarly, Sloman (1991) reported part-set cuing impairment when generating category instances and when generating reasons.
Although part-set cues might be expected to benefit recall, but often impair recall, it is noteworthy that in some circumstances they appear to have little or no effect on the level of recall performance. Watkins, Schwartz, and Lane (1984) found that providing both tournament and casual chess players with the positions of half of the pieces from partly played chess games that they had studied neither facilitated nor inhibited the reconstruction of the remaining pieces (see also Huffman, Matthews, & Gagné, 2001). A similar failure to find either impairment or facilitation has sometimes been reported for free recall following the study of word lists (e.g., Experiments 2 and 3 of Basden et al., 2002; Slamecka, 1969) and when the measure of recall is fragment completion, an indirect word association task, or a cued recall task with extra-list associates as cues (Basden, Basden, Church, & Beaupre, 1991).
Several attempts have been made to account for the part-set cuing impairment often observed. The cue-overload principle (Watkins & Watkins, 1975) offered an explanation by considering the cues as additional items in the set – the larger set size for cued sets overloads the high level cue (to recall the set) thereby predicting poorer performance (Mueller & Watkins, 1977). The competition-at-retrieval hypothesis (Rundus, 1973) assumed that presenting the cue items strengthens the representation of these items making them more likely to be recalled and interfering with recall of the relatively weaker items outside the cue set. Rundus hypothesized that associations within word lists are hierarchical rather than involving associations between items. With the additional assumptions that recall of an item strengthens its representation and that cued and recalled items remain a part of the pool accessed for further retrieval, the model predicts poorer performance with part-set cuing. The associative sampling-bias hypothesis (Raaijmakers & Shiffrin, 1981) focused on inter-item associations and based its explanation on the clusters of associated items that are sampled in the recall attempt. It assumed that both part-set cued and control participants sample the same number of clusters but that when some items are used as cues fewer items remain for report. In cases where people are asked to recall the complete set, including the previously practised part of the set that might act as cues, Anderson and Neely (1996) and Bäuml and Aslan (2006) explain the part-set cuing impairment as functionally equivalent to retrieval-induced forgetting (Anderson, Bjork, & Bjork, 1994). Their explanation is similar to Roediger's (1973, 1974) suggestion that output interference could account for reduced recall of the items from the set that were not studied as cues. Efforts to retrieve members of the set lead first to the recently studied or practised items. Having retrieved these items, recall of the remaining items suffers.
These attempts to explain part-set cuing predict impairment but in some circumstances facilitation or a lack of effect has been observed. For example, serial recall with consistent cues is sometimes facilitated by part-set cues (Basden et al., 2002; Experiment 1) and free recall of uncategorized lists is sometimes not influenced by part-set cues (Basden et al., 2002; Slamecka, 1969). Similarly, recall of chess positions is also unaffected (Huffman et al., 2001; Watkins et al., 1984).
The strategy disruption hypothesis is an alternative approach predicting both impairment and facilitation. Proponents of this hypothesis (e.g., Basden & Basden, 1995; Basden et al., 2002; Brown & Hall, 1979; Sloman, Bower, & Rohrer, 1991) argue that the presented cues interact with the organization and/or the retrieval strategy of the participants. Although there is evidence against this explanation (e.g., Peynircioğlu, 1989), this hypothesis is quite successful at explaining the variations in impairment and facilitation: When the cues are consistent with the retrieval strategy they facilitate recall; when they disrupt the strategy they impair recall and when they are merely superfluous they have no effect. So, for example, serial recall with consistent cues will support the appropriate retrieval strategy, thereby facilitating recall, but inconsistent cues (i.e., words in the wrong positions) will disrupt the retrieval strategy and lead to impaired serial recall (Basden et al., 2002). Where neither impairment nor facilitation occurs, the hypothesis might explain that the cues neither disrupt nor support the retrieval strategies or that the cues both impair and benefit retrieval in equal measure.
Bäuml and Aslan (2006) argued for a two-mechanism account of part-set cuing, with retrieval strategy disruption and retrieval inhibition potentially playing parts, depending upon the specific experimental situation. Bäuml and Samenieh (2012) added context reactivation as a third mechanism, and Cole et al. (2013) interpreted their finding of facilitation when using Snap Circuit stimuli as potentially consistent with either the two- or the three-mechanism accounts.
Research on part-set cuing of recall has largely concentrated on word lists (with the exceptions described above). If the phenomenon is of general interest it should be examined in other situations, including those that may be closer to situations in which recall regularly occurs in everyday life. Bovee, Fitz, Yehl, Parrott, and Kelley (2009) extended the conditions under which part-set cuing was examined by studying cuing of shopping lists and campus buildings with real-world visits, demonstrating inhibition with free recall and facilitation with reconstruction.
We examined part-set cuing in two situations that extend the domains in which the phenomenon has been explored: recall from expository text and from pictures of scenes. For theoretical reasons it is useful to know whether, and in what form, part-set cues influence recall in a wider set of situations than those studied so far. How far do part-set cuing effects extend? This question has practical as well as theoretical implications. Expository text is often studied and tested in educational settings and eyewitnesses may be asked to recall the contents of a scene as in the Bovee et al. (2009) studies. It would, therefore, be useful to know whether providing part of a to-be-remembered text or scene will influence recall.
Positive effects of cuing might be expected both for expository text and pictorial scenes, as Serra and Nairne (2000) and Basden et al. (2002) have found for serial list recall. Expository text and pictures of scenes both provide coherent, integrated stimuli within which there are inter-associations among the elements of the material. Furthermore, reading a text leads to the creation of a schema (Bartlett, 1932) or situation model (e.g., Kintsch, 1998) that will guide retrieval. Although categorized word lists may be organized in the vertical fashion discussed by Rundus (1973) and Nickerson (1984), it is unlikely that either expository texts or pictures of scenes are so simply organized. These types of material would seem to involve interrelations between items if the text or the picture is to be understood and remembered.