Journal of Memory and Language 59 (2008) 183–199 Contents lists available at ScienceDirect Journal of Memory and Language journal homepage: www.elsevier.com/locate/jml Memory for items and associations: Distinct representations and processes in associative recognition Norbou G. Buchler a,*, Leah L. Light b, Lynne M. Reder a a Department of Psychology, Carnegie Mellon University, Pittsburgh, PA 15213, USA b Department of Psychology, Pitzer College, Claremont, CA 91711, USA article info abstract Article history: In two experiments, participants studied word pairs and later discriminated old (intact) Received 16 November 2006 word pairs from foils, including recombined word pairs and pairs including one or two pre- Revision received 28 March 2008 viously unstudied words. Rather than making old/new memory judgments, they chose one Available online 23 May 2008 of five responses: (1) Old–Old (original), (2) Old–Old (rearranged), (3) Old–New, (4) New– Old, (5) New–New. To tease apart the effects of item familiarity from those of associative Keywords: strength, we varied both how many times a specific word-pair was repeated (1 or 5) and Item how many different word pairs were associated with a given word (1 or 5). Participants Association could discriminate associative information from item information such that they recog- Recognition Interference nized which word of a foil was new, or whether both were new, as well as discriminating Recollection recombined studied words from original pairings. The error and latency data support the Familiarity view that item and associative information are stored as distinct memory representations Novelty and make separate contributions at retrieval. Fan Ó 2008 Elsevier Inc. All rights reserved. Introduction tive information are accessed and utilized in subsequent decision-making during recognition. The research reported Memory theorists distinguish between two qualita- in this article addresses these two issues. tively different sources of information underlying the pro- cess of recognition (Hockley, 1992; Humphreys, Bain, & The representation of item and associative information in Pike, 1989; Kelley & Wixted, 2001; Murdock, 1997; Reder associative recognition et al., 2000; Yonelinas, 1997). Memory for associative information, such as the co-occurrence of two words, can Global models of associative recognition, such as REM be distinguished from memory for the individual items (Shiffrin & Steyvers, 1997), TODAM (Murdock, 1997), (for reviews, see Clark & Gronlund, 1996; Raaijmakers & MINERVA II (Hintzman, 1984, 1988), and Matrix (Humph- Shiffrin, 1992; Yonelinas, 2002). However, memory theo- reys et al., 1989), assume that item and associative infor- rists differ with respect to just how item information and mation are inseparable and are stored as part of a associative information are represented in memory and common memory system, whereas local models such as how each contributes to decision-making in the recogni- SAC (Reder et al., 2000) assume that associative informa- tion of word pairs. One controversy is whether associative tion is represented separately from item information. In information is stored separately from item information or associative recognition, the REM, TODAM and Matrix mod- is stored in common with item information (Gillund & els use similar mathematical operations to represent asso- Shiffrin, 1984; Hintzman, 1988; Hockley, 1991, 1992; Mur- ciative information as a conjoined representation of two dock, 1993). A second core issue is how item and associa- sets of item features—either a concatenation (REM), convo- lution (TODAM), or tensor product (Matrix) of two item vectors. Local models, in contrast, store item and associa- * Corresponding author. Fax: +1 919 681 0815. E-mail address: [email protected] (N.G. Buchler). tive information in separate memory locations. For in- 0749-596X/$ - see front matter Ó 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.jml.2008.04.001 184 N.G. Buchler et al. / Journal of Memory and Language 59 (2008) 183–199 stance, the SAC model (Reder et al., 2000) representation of and associative information into a single composite mea- a studied word-pair is explicitly defined by two semantic sure of memory strength. Whether a stimulus is called nodes, one for each word in the pair, with each word node ‘‘old” or ‘‘new” depends on whether the composite mem- linked to the same episodic node encoding their associa- ory score falls either above or below a decision criterion. tion—the co-occurrence of both words in a particular study For instance, Hockley (1991, 1992) conceptualized word- context (see models 4 and 5, Buchler & Reder, 2007, pp. pair recognition as a signal detection process based on a 111–112; Reder, Paynter, Diana, Ngiam, & Dickison, 2007). single memory strength statistic, the sum of item informa- In the research reported in this article, we test these tion and associative information that must exceed a deci- assumptions underlying local and global models. If asso- sion threshold. ciative information is a product of item information, as as- In contrast, the dual-process model of Yonelinas (1997, sumed by global models, it follows that the strengthening 2002) argues that recognition is not based on memory of item information necessitates the strengthening of asso- strength aggregated across sources, but is instead based ciative information. However, if item and associative infor- on the outputs of two separable memory processes, famil- mation are stored separately, as assumed by local models, iarity and recollection (Mandler, 1980; Reder et al., 2000). it should be possible through appropriate manipulation of The process of recollection involves retrieving specific con- item and pair repetitions to strengthen item information textual associations (i.e. episodic traces) and the familiarity without strengthening associative information. process is based on the general strength of items in mem- In our study of word-pair recognition, we use two ory. Recollection is treated as a threshold process that can experimental manipulations—repetition (differential be distinguished by highly confident recognition and vivid strength of the word pairing) and fan or multiple word- mnemonic detail (for reviews, see Diana, Reder, Arndt, & pair associates (differential associative interference)—to Park, 2006; Yonelinas, 2002). Whether a stimulus is recog- further differentiate item and associative sources of infor- nized as ‘‘old” or ‘‘new” depends on whether either mem- mation. In two experiments, we manipulated how many ory process is successful. times a specific word-pair was repeated and how many Among dual-process models, however, there is no uni- different word pairs were associated with a given word versal agreement on how item and associative sources con- to tease apart the effects of item strength from associative tribute to an associative recognition judgment. That is, strength. If associative information is stored as a conjoined dual-process models all assume an attempt to recollect memory representation of item information as assumed by the association, but offer a number of different character- global models, then the memory representation for pairs izations of the familiarity process. Conceptually, a word- with multiple overlapping associates should be equivalent pair recognition task has three potentially discrete sources in strength to word pairs in which the same word-pair is of information—each word in the pair (word1, word2) and repeatedly presented. Examples are given in Table 1 (see the association. All three sources (word1, word2, associa- Fan 5–5 and Rep 5Â). In both cases the underlying item tion) contribute to familiarity in the Kelley and Wixted strength is identical, and by extension, global matching (2001; Wixted, 2007) model, both item sources contribute models predict that the associative strength should be to familiarity in the Yonelinas (1997; Yonelinas & Parks identical as well. In contrast, local models of memory pre- 2007) model, and each item can contribute separately in dict that it should be possible to additionally strengthen the SAC model (Buchler & Reder, 2007; Reder et al., 2000). associative information separately from item information. Theoretical treatments of associative recognition gener- In this case, repeated word pairs should be better recog- ally lump together the two words in the pair as contribut- nized than overlapping word pairs. Although these would ing to ‘item’ information. However, there is evidence that seem to be straightforward predictions, we were unable participants are able to accurately make frequency judg- to find any evidence of a test between these competing ments for words and word pairs and thus distinguish the views. We explicitly confirmed these predictions with a frequency of occurrence of specific item information (each Monte-Carlo simulation of both a global matching model word) and associative information (pairing). Hockley and of memory, REM, and a local memory model, SAC. A Cristi (1996) orthogonally varied the study presentation description of the REM and SAC model mechanics and frequency of word pairs and the frequency of the individ- the results of our Monte-Carlo simulations are provided ual items of the pairs across different study pairings. In in the General discussion. separate tests, participants were asked to judge the fre- quency of the word pairs, a word’s frequency as an individ- The use of item and associative information in recognition ual item, its frequency as a member