fMRI Evidence
In addition to the neuropsychological research reviewed in section 1.4, a number of fMRI investigations have also examined the neural correlates of recognition memory.
Importantly, the unitary memory strength model, dual-process models, the BIC model, and CRAFT make different predictions concerning differential fMRI BOLD responses in PrC, PhC, and HC in relation to familiarity-based item recognition and recollection. Pertinent fMRI studies have probed familiarity and recollection using either subjective reports of either familiarity- or recollection-based recognition, source memory judgments that assess recognition of a target item as well as the ability to recall contextual detail from the initial encounter, or recognition confidence ratings that index the certainty with which participants feel an item is old or new. In the context of experiments that have employed these paradigms, fMRI BOLD responses related to recollection have been isolated by contrasting recollected trials with those judged as being familiar [R > F], source memory correct versus incorrect [Source Correct > Incorrect], and highest confidence responses versus all lower confidence ratings [e.g., 5 > 1-4]. By contrast, familiarity related activity is assessed by contrasting correct familiar judgments with target items called ‘new’ [F < Miss], trials on which a target item is correctly identified
but not the source with misses [Target Correct + Source Incorrect < Miss], and by identifying regions in which changes in BOLD response correlate with changes in confidence levels after excluding the highest response option.
Evidence obtained across many studies using the paradigms noted above converges in a remarkably consistent manner with respect to the neural correlates of recollection and familiarity, although these findings have not gone unchallenged (Wixted and Squire, 2011). Specifically, most fMRI studies that have examined recognition memory report differential activity related to recollection at both the time of encoding and retrieval in the HC and PhC. Critically, however, these regions are generally insensitive to differences related to familiarity-based recognition (Eldridge et al., 2000; Davachi et al., 2003; Kahn et al., 2004; Ranganath et al., 2004; Weis et al., 2004; Dolcos et al., 2005; Uncapher and Rugg, 2005; Woodruff et al., 2005; Yonelinas et al., 2005; Montaldi et al., 2006;
Daselaar et al., 2006; Uncapher et al., 2006; Vilberg and Rugg, 2007; Wang et al., 2014; see Diana et al., 2007, for review). In contrast to results obtained in the HC and PhC, activity in PrC has been demonstrated to be sensitive to familiarity-based item
recognition, but not the recollection of contextual detail (Henson et al., 1999; Davachi et al., 2003; Ranganath et al., 2004; Weis et al., 2004; Uncapher and Rugg, 2005; Yonelinas et al., 2005; Daselaar et al., 2006; Montaldi et al., 2006; Kensinger and Schacter 2006; Uncapher et al., 2006; Yassa and Stark, 2008; Kafkas and Montaldi, 2012; Wang et al., 2014; see Diana et al., 2007, for review).
Rather than detailing the experimental design and specific results from each of these studies, I will elaborate on just one that has figured prominently in the literature. Ranganath et al. (2004) scanned participants during the encoding stage of a recognition memory task and in a subsequent un-scanned test stage asked that they discriminate between previously studied and novel words. During the test stage, participants indicated how confident they were that each item was studied and also performed a two alternative forced-choice source memory judgment that probed recollection of the encoding task associated with each trial. Scanned encoding trials were then scored according to subsequent memory judgments and whole-brain voxel-wise contrasts revealed that left PrC was the only MTL structure in which activity linearly indexed recognition
confidence ratings (i.e., a graded familiarity signal). By contrast, the right HC and right PhC were the only MTL structures that showed differential effects between correct and incorrect source memory judgments (i.e., recollection).
The results from Ranganath et al. (2004), together with those not reviewed in further detail (see above), are largely consistent with predictions derived from dual-process models of MTL organization regarding PrC and the HC. Specifically, evidence from the majority of fMRI studies link PrC to familiarity-based item recognition and the HC to the recovery of episodic contextual detail. However, in addition to the HC, these data also implicate PhC in the recovery of contextual detail. With respect to competing three- component models of MTL organization, the overall pattern of results obtained with fMRI studies favours the BIC model over CRAFT as differential activity in PhC is typically observed in relation to the encoding and retrieval of recollection or the recovery of source detail, not familiarity-based context recognition.
In addition to informing the development of the theoretical models of MTL organization, results from these fMRI investigations have also guided proposals concerning the
mechanism by which familiarity-based item recognition signals are coded in PrC.
Specifically, the observation that familiarity signals measured at the time of retrieval tend to manifest in fMRI BOLD responses as a relative decrement in activity has been taken as evidence favouring the notion that it is a decrease in PrC activity that codes for stimulus familiarity (c.f., Yassa and Stark, 2008; Kafkas and Montaldi, 2012). Interestingly, such findings are consistent with results from neurophysiological data obtained with single-cell recordings from PrC in both macaques and rats. Specifically, these studies most typically reveal a decrease in neural spiking for repetition of visually presented stimuli (Xiang and Brown, 1998; cf. Thome et al., 2012). However, as will be discussed in Chapter 4, there are a number of challenges associated with mapping neurophysiological recordings and fMRI BOLD responses onto one another.
By linking the functional contributions of MTL structures to differences in informational content, rather than mnemonic processes as such, the BIC model of MTL organization can account for much, though not all, extant fMRI and neuropsychological data that
speak to the issue of functional specialization pertaining to familiarity-based item recognition and recollection of contextual detail. Importantly, however, the neural correlates of familiarity-based item recognition have primarily been probed using words as memoranda. Accordingly, no systematic effort has been made to examine potential differences between contributions of MTL structures to recognition memory in relation to stimulus content. This point is particularly pertinent for discussions concerning the roles of PrC and PhC, as fMRI data obtained in non-mnemonic task contexts point to category- specific differences between these structures in relation to visually presented objects. How these categorical effects relate to the distinction between item and context representations remains unclear.