Recognition memory: Adding a response deadline eliminates recollection but spares familiarity

A current controversy in memory research concerns whether recognition is supported by distinct processes of familiarity and recollection, or instead by a single process wherein familiarity and recollection reflect weak and strong memories, respectively. Recent studies using receiver operating characteristic (ROC) analyses in an animal model have shown that manipulations of the memory demands can eliminate the contribution of familiarity while sparing recollection. Here it is shown that a different manipulation, specifically the addition of a response deadline in recognition testing, results in the opposite performance pattern, eliminating the contribution of recollection while sparing that of familiarity. This dissociation, combined with the earlier findings, demonstrates that familiarity and recollection are differentially sensitive to specific memory demands, strongly supporting the dual process view.Receiver operating characteristic (ROC) analysis holds the promise of dissecting the contributions to recognition memory of episodic recollection and familiarity (Yonelinas 2001), and this method can be applied equally well to examine these memory processes in animals as well as humans (Fortin et al. 2004; Sauvage et al. 2008). According to the dual process model, recollection is indexed by the asymmetry of the ROC function whereas familiarity is measured by the degree of curvilinearity of that function, and correspondingly, these two parameters can vary independently (Yonelinas 2001). However, there is controversy about this interpretation of ROC components. Some have argued that the asymmetry and curvilinearity of the ROC function both reflect the strength of memories mediated by a single process (Wixted 2007), and correspondingly, these components of the ROC increase or decrease together in stronger or weaker memories, respectively (Squire et al. 2007).A resolution of this controversy can be advanced by examining whether the ROC asymmetry and curvilinearity are independently influenced by task manipulations that favor either recollection or familiarity, consistent with dual process theory, or instead are similarly influenced by conditions that affect memory strength. Recent data from an animal model of recognition have shown that adding a demand for remembering associations between independent stimuli eliminates the ROC curvilinearity without affecting the asymmetry, consistent with the dual process view (Sauvage et al. 2008; for discussion of associative recognition, see Mayes et al. 2007). However, in order to provide compelling evidence of independence of the two ROC components, it is also critical to show that other memory demands that favor familiarity produce the opposite pattern, elimination of the ROC asymmetry while sparing its curvilinearity. Together these findings would constitute a double dissociation between the two parameters of the ROC function that cannot be explained by a single process theory.As originally conceived in models proposed in the 1970s, familiarity is characterized as a perceptually driven, pattern matching process that is completed rapidly, whereas recollection is characterized as a conceptually driven, organizational process that requires more time (Mandler 1972; Atkinson and Juola 1973, 1974; for reviews, see Yonelinas 2002; Mandler 2008). Consistent with this view, the results of several studies that employ response deadlines in the test phase report that familiarity is more rapid than recollection. For example, forcing people to make speeded recognition responses has little effect on simple yes–no recognition but strikingly reduces performance when subjects must remember where or when an item was studied (Yonelinas and Jacoby 1994; Gronlund et al. 1997; Hintzman et al. 1998). Other studies that require subjects to oppose familiarity and recollection reveal a two-component temporal function that includes a rapidly available familiarity process and a slower recollective process (Dosher 1984; Gronlund and Ratcliff 1989; Hintzman and Curran 1994; McElree et al. 1999). In addition, studies that measure brain evoked response potentials (ERPs) have revealed two distinct ERP modulations commonly observed during recognition: a mid-frontal negativity onsetting about 400 msec after stimulus onset that is associated with familiarity, and a parietally distributed positivity beginning about 500 msec after stimulus onset that is associated with recollection (Smith 1993; Duzel et al. 1997; Curran 2004; Duarte et al. 2006; Woodruff et al. 2006; but see Voss and Paller 2009).Dual process theory predicts that applying an appropriate early response deadline should allow sufficient time for contribution of familiarity but not that of recollection, and so should reduce the ROC asymmetry while sparing its curvilinearity, opposite to the already observed effects of associative memory demands that favor recollection (Sauvage et al. 2008). Confirmation of this prediction combined with the previous findings of the opposite effects in associative recognition would constitute a double dissociation between the features of recollection and familiarity. This result would therefore strongly support the conclusion that the asymmetry and curvilinearity are independent parameters of the ROC function that are differentially linked to features of recollection and familiarity, respectively.