Revisiting Cognitive and Neuropsychological Novelty Effects

Transcription

1 Revisiting Cognitive and Neuropsychological Novelty Effects by Jordan Lindsay Poppenk A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Graduate Department of Psychology University of Toronto Copyright by Jordan L. Poppenk 2011

2 Revisiting Cognitive and Neuropsychological Novelty Effects Abstract Jordan Poppenk A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Graduate Department of Psychology University of Toronto 2011 Recent proposals have attributed a key role to novelty in the formation of new episodic memories. These proposals are based on evidence of enhanced memory and greater metabolic activity in the hippocampus in response to novel relative to familiar materials. However, such novelty effects are incongruous with long-standing observations that familiar items and lists are associated with better memory than novel ones. In four experiments, I explored possible reasons for this apparent discrepancy. In Experiment 1, I directly tested whether previously observed novelty effects were the result of novelty, discrimination demands, or both. I used linguistic materials (proverbs) to replicate the novelty effect but found it occurred only when familiar items were subject to source confusion. In Experiment 2, to examine better how novelty influences episodic memory, I used experimentally familiar, preexperimentally familiar, and novel proverbs in a paradigm designed to overcome discrimination demand confounds. Memory was better for both types of familiar proverbs. These cognitive results indicate that familiarity, not novelty, leads to better episodic memory for studied items, regardless of whether familiarity is experimentally induced or based on prior knowledge. I also conducted two fmri experiments to evaluate the neural correlates of the encoding of novel and familiar ii

3 forms of information. In Experiment 3, I compared the neural encoding correlates of source memory for novel and familiar visual scenes using fmri. Replicating previous neuroimaging studies, I observed an anterior novelty-sensitive region of the hippocampus specialized in novelty encoding. Unlike past studies, I also probed for familiarity-encoding regions and identified such regions in the posterior hippocampus. I replicated this pattern in Experiment 4 using proverbs as stimuli. As in Experiment 2, I found the effect held whether familiarity was based on prior knowledge or experimental induction. In both fmri experiments, anterior and posterior hippocampal regions were functionally connected with different large-scale networks, helping to explain local variation in hippocampal functional specialization in terms of different neural contexts. Together, these experiments show that stimulus familiarity enhances episodic memory for materials, and that novelty is processed differently, not preferentially, in the hippocampus. A new model of hippocampal novelty processing is proposed. iii

4 Acknowledgments I wish to thank the students and research assistants who contributed in various ways to the projects contained in this dissertation, including Kathy Li, Erol Ozcellik, Rajwant Sandhu, Christina Savarino, Signy Sheldon, Maria Tassopoulos, Hongye Wang and Jennifer Wang. I am also grateful to Marilyne Ziegler, who provided technical support, to Jennifer Ryan, who contributed to my supervisory committee, to Emrah Düzel, Cheryl Grady and Brian Levine, who sat on my defense committee, and to my peers, who showed me how to be a graduate student. Finally, I wish to thank my academic supervisors, Morris Moscovitch and Anthony McIntosh, who contributed theoretical and technical guidance, a supportive and unconstrained research environment, friendship, and advice and oversight concerning this dissertation and its component projects. iv

5 Table of Contents Acknowledgments...iv Table of Contents... v List of Tables... vii List of Figures... ix List of Appendices... xiii Chapter 1 General Introduction Cognitive evidence for benefits of familiarity on learning Cognitive evidence for benefits of novelty on learning Cognitive neuroscience support for novelty advantages in memory Evidence for hippocampal sensitivity to familiarity Interpreting the available evidence Overview of experiments...10 Chapter 2 Resolving conflicting novelty and familiarity findings in cognition Experiment Methods Results and discussion Experiment Methods Results and discussion General discussion of Chapter 2 experiments...26 Chapter 3 Neural correlates of novelty and familiarity encoding Methods Participants Stimulus materials Procedure MRI scanning and data analysis Results Behavioral results Functional neuroimaging results Discussion...56 v

6 Chapter 4 Modulation of encoding by experimental and pre-experimental forms of familiarity Methods Participants Stimulus materials Procedure Neuroimaging data acquisition and analysis Results Behavioral results Functional neuroimaging results Discussion...89 Chapter 5 General Discussion Summary of experimental effects Theoretical integration of cognitive results Theoretical integration of fmri results An alternative model of hippocampal memory function Further investigation Conclusions References Appendices vi

7 List of Tables Table 1. Schematic of experimental protocol and stimulus exposure in Experiment 1. Stimuli consisted of three lists of 25 Asian target proverbs (Familiar TARG, Novel- Repeated TARG, and Novel TARG ) and three lists of 25 Asian lure proverbs (Familiar LURE, Novel-Repeated LURE, and Novel LURE ). Table 2. Descriptive statistics for hits, false alarms (FA), and two measures of corrected accuracy (acc. and d') for pre-study repetition items, novel repeated items and novel items in the Experiment 1 memory test. Table 3. Schematic of experimental protocol and stimulus exposure in Experiment 2. Stimuli consisted of two lists of 20 English proverbs (English 1 and English 2 ) and four lists of 20 Asian proverbs (Asian-Familiar 1, Asian-Familiar 2, Asian-Novel 1, and Asian-Novel 2 ). Table 4. Descriptive statistics for Experiment 2 source memory accuracy for preexperimentally familiar proverbs, proverbs with experimentally-induced familiarity, and novel proverbs. Table 5. Descriptive statistics for hits, false alarms (FA), corrected recognition accuracy (acc) and source memory accuracy (src) for pre-study repetition items and novel items in the Experiment 3 memory test. Table 6. Activated voxel clusters in the block-level contrast of novel and familiar scenes in Experiment 3. Table 7. Voxel clusters that predicted subsequent memory success for both novel and familiar scenes in Experiment 3 (conjunction analysis). Table 8. Activated clusters in the Experiment 3 interaction contrast of subsequent memory with novel and familiar pictures. Table 9. Activated voxel clusters in the Experiment 3 event-related contrast of remembered and forgotten scenes (novel items only). Table 10. Activated voxel clusters in the Experiment 3 event-related contrast of remembered and forgotten scenes (familiar items only). Table 11. Functional connectivity of the anterior hippocampal novelty-encoding region versus posterior hippocampal familiarity-encoding region (Experiment 3). Table 12. Schematic of experimental protocol and stimulus exposure in Experiment 4. vii

8 Stimuli consisted of two lists of 40 English proverbs (English 1 and English 2 ) and four lists of 20 Asian proverbs (Asian-Familiar 1, Asian-Familiar 2, Asian-Novel 1, and Asian-Novel 2 ). Table 13. Descriptive statistics for Experiment 4 response frequency and source memory accuracy for pre-experimentally familiar proverbs, proverbs with experimentally-induced familiarity, and novel proverbs at high, moderate and low levels of confidence. Table 14. Regions that preferentially predicted memory success in Experiment 4 for novel or repeated proverbs in one of the two conditions. Table 15. Regions that preferentially predicted memory success in Experiment 4 for novel or prior knowledge proverbs in one of the two conditions. Table 16. Regions that preferentially predicted memory success in Experiment 4 for repeated or prior knowledge proverbs in one of the two conditions. Table 17. Regions that predicted memory success in Experiment 4 for more than one condition at a particular latency from proverb onset. Table 18. Voxel clusters distinguishing functionally-connected networks seeded from the anterior versus posterior hippocampus in Experiment 4. viii

9 List of Figures Figure 1. Experiment 1 memory performance. Hits, false alarms, and corrected accuracy scores for pre-study repetition items, novel repeated items and novel items in the Experiment 1 memory test. Error bars depict +1 standard error of the mean (SE). Figure 2. Experiment 2 memory performance. Experiment 2 source memory accuracy for pre-experimentally familiar proverbs, proverbs with experimentally-induced familiarity, and novel proverbs. A significant difference is depicted by * at P<0.05 and ** at P<0.01. Figure 3. Schematic of experimental design in Experiment 3. During the repetition phase (a), which took place immediately before fmri scanning in a mock fmri scanner, participants were exposed to one set of scenes three times. During the encoding phase (b), which took place in the fmri scanner, participants viewed the repeated scenes and a novel set while imagining themselves performing an action in the scene or imagining a possible intention associated with the scene. In a subsequent memory test outside of the scanner (c), participants were presented with all of the scenes from the encoding phase along with additional lures, some of which had been presented in phase 1 but not during scanning (repeated lures), and some of which had not previously been presented (novel lures). They were asked to determine whether each scene was presented during scanning and, if so, whether it was associated with action or intention instructions. Figure 4. Behavioral measures of Experiment 3 memory performance. A trend was observed towards the typical advantage for novel materials in recognition memory accuracy, the measure that has been used to compare memory for novel and familiar materials in all previous novelty studies (a). However, recognition memory is confounded here and in other studies investigating novelty and memory: whereas the goal is to compare memory strength for episodes involving novel and repeated materials, repeated lures are more difficult to reject than novel ones. Inset (b) is a plot illustrating how a higher false-alarm rate in the repeated relative to novel condition (*** P < 0.001) deflated recognition accuracy scores in the repeated condition. The novel approach described here was to measure source memory accuracy for novel and repeated items, a measure that ix

10 does not suffer from this confound. I observed the opposite pattern to that observed when using the recognition memory measure: memory was superior for repeated materials (** P < 0.005). Chance recognition memory accuracy was 0.00; chance source memory accuracy was Error bars depict +1 SE. Figure 5. Stable subsequent memory regions in Experiment 3. A transparent brain depicts those regions in which brain activity at encoding predicted subsequent source memory for both novel scenes viewed for the first time during the study phase and repeated scenes viewed three times before the study phase. The regions are shown in a sagittal view (a; anterior = right) and transverse view (b; left = up). Response plots for specific regions of interest are depicted separately on the right for novel and repeated scenes (c-e; +1 SE). To identify the common network, activation maps were combined from source memory hit > source memory miss contrasts for novel and repeated scenes. Depicted regions surpassed a threshold of P<0.05 in both contrasts, then survived a bootstrap resampling procedure designed to evaluate whether conjoint effects could be observed at the withinsubject level. Figure 6. MTL subsequent memory interactions in Experiment 3. A slice series is depicted moving medially in progression from top to bottom (both hemispheres; anterior = right), with functional activations overlaid on an MNI template brain. MTL regions predicting subsequent source memory for novel items seen for the first time during the study phase and repeated items seen three times before the study phase are plotted in yellow and blue, respectively. Activity during encoding in the right anterior hippocampus (a) predicted subsequent source memory (hits minus misses) for novel scenes only, whereas activity in the posterior hippocampi (b and c) predicted subsequent source memory for repeated scenes only (plots are shown +1 SE). Figure 7. Anterior vs. posterior hippocampal functional connectivity in Experiment 3. Left, right, superior and inferior views of a 3D rendering comparing the functional connectivity associated with the anterior and posterior hippocampus (a). Regions more functionally connected to the anterior hippocampus are shown in violet; those more functionally connected to the posterior hippocampus are shown in green. The contrast is displayed on an MNI template brain. In the top row, middle is posterior; in the bottom row, middle is right. All clusters of 12 or more x

11 contiguous voxels surpassing a threshold of BSR 2.0 (approximately corresponding to a threshold of P < 0.05) are displayed. Functional connectivity was greater between the right temporal pole and anterior versus posterior hippocampus, predicting source memory for novel but not repeated scenes (b). Functional connectivity was greater between the left precuneus and posterior versus anterior hippocampus, predicting source memory for repeated but not novel scenes (c). Figure 8. Subsequent source memory networks for novel and repeated scenes in Experiment 3. Renderings depict brain activity during encoding that predicted subsequent source memory for novel scenes seen for the first time at encoding (a), and repeated scenes seen three times before encoding (b). Activations predicting later memory success are in yellow and activations predicting later memory failure are in blue. Regions supporting memory for novel scenes were greater in number and extent than those supporting memory for repeated scenes. All activations are displayed on an MNI template brain (posterior = middle). All clusters of 12 or more contiguous voxels surpassing a threshold of BSR 2.0 (approximately corresponding to a threshold of P < 0.05) are displayed. Figure 9. Source memory performance in Experiment 4. The proportion of items for which participants recalled the correct encoding task was higher in the two familiarity conditions than in the novelty condition, replicating findings from Experiment 2. Memory was higher for pre-experimentally familiar items relative to memory for items with experimentally-induced familiarity. A statistically significant difference is designated by ** at P < 0.01 and *** at P < Figure 10. Novelty-based hippocampal encoding interactions in Experiment 4. Left hippocampal subsequent memory regions are overlaid on a sagittal slice (X=-20) taken from the average anatomical image of 200 healthy young adults scanned at the Rotman Research Institute. Regions contained peaks of at least BSR 3.5 and were thresholded at BSR 2.81, approximately corresponding to 99.95% and 99.5% confidence intervals, respectively. The sagittal image depicts three types of interaction regions: prior knowledge dm > novel dm (dark blue); repeated dm > novel dm (green); and novel dm > prior knowledge dm (orange). Overlapping interactions are shown using additive colors. No novel dm > repeated dm interaction regions were observed in the hippocampus. The left posterior xi

12 hippocampus (a) predicted successful familiarity encoding, regardless of whether familiarity was established through repetition or prior knowledge, and predicted unsuccessful novelty encoding. In contrast, the left anterior hippocampus (b) preferentially predicted successful novelty encoding, but also predicted encoding success in the prior knowledge familiarity condition. Figure 11. Novelty-based encoding interactions of the full brain in Experiment 4. Pair-wise interactions in the whole-brain subsequent memory predictors of the three novelty conditions (novel, repeated and prior knowledge items) are depicted using a glass brain view. Regions were thresholded as in Fig. 2. Figure 12. Resting-state functional connectivity differences between posterior and anterior hippocampus in Experiment 4. Rotated partial least squares analysis revealed a significant large-scale network dissociating posterior (phpc) and anterior (ahpc) covariance (a). No additional reliable patterns were observed. Reliable differences are depicted on a glass brain in a sagittal view (b) and transverse view (c). Differences are also overlaid on slices of a template brain based on 200 young adults scanned at the Rotman Research Institute (d-e). Differences corresponded to the cortical connections of two hippocampal pathways (f). Part f modified from Duvernoy (2005) with kind permission of Springer Science+Business Media. Figure 13. Schematic of the dual-comparator model of hippocampal novelty processing. Where appropriate, nodes are accompanied by region(s) hypothesized to support the processes described in them. Incoming information is evaluated in parallel by posterior and anterior hippocampal comparators that are specialized in semantic and episodic forms of novelty processing. Preliminary evidence in the model may be useful for guiding attention, whereas detailed evidence requires attention and elaborative processing to uncover in the first place. As illustrated here, it is hypothesized that novel materials must first undergo stimulus elaboration prior to contextual processing. xii

13 List of Appendices Appendix A. Proverb stimuli used in Experiments 1, 2 and 4. Appendix B. Scene stimuli used in Experiment 3. xiii

14 Chapter 1 General Introduction The current dissertation is concerned with the cognitive neuroscience of human memory. It presents a detailed investigation of the effects of novel materials on the formation of new memories, both in terms of the robustness of the memories created and the relationship between memory formation and memory-encoding structures in the brain. Anecdotal observation suggests that memory for one s first kiss and other such firsts can be far better than memory for more routine events. This observation is sometimes cited as evidence that novel experiences are remembered better than familiar ones, a notion that has been reinforced experimentally by evidence that memory can be superior for novel over previously repeated stimulus lists (Tulving & Kroll, 1995). Researchers have also reported increased activation in the medial temporal lobes (MTL), a brain area known to be important for long-term episodic memory (Squire, Stark, & Clark, 2004; Moscovitch et al., 2005; Eichenbaum, Yonelinas, & Ranganath, 2007), in response to novel relative to repeated stimuli (Tulving, Markowitsch, Craik, Habib, & Houle, 1996; Kirchhoff, Wagner, Maril, & Stern, 2000; Köhler, Crane, & Milner, 2002; Stark & Okado, 2003; Schott et al., 2004; Bunzeck & Düzel, 2006; Kumaran & Maguire, 2006; Danckert, Gati, Menon, & Kohler, 2007). Based on such evidence, it has been suggested that information is encoded only to the extent that it is novel (Tulving, Markowitsch, Craik, Habib, & Houle, 1996). This proposal, however, is not easily reconciled with classical findings revealing mnemonic benefits of advance familiarization with to-be-remembered items, such as advantages of item repetition prior to study in list-learning experiments (Ebbinghaus, 1913). The proposal is also complicated by evidence that the MTL is sensitive not only to novelty, but also to familiarity (Lepage, Habib, & Tulving, 1998; Rutishauser, Mamelak, & Schuman, 2006; Daselaar, Fleck, & Cabeza, 2006). 1

15 2 The purpose of this thesis is to try to resolve these apparent discrepancies, which I address directly in four cognitive and neuroimaging experiments. In a review of the literature below, I provide a general overview of the theoretical motivation for this topic and introduce several cognitive and neuroimaging experiments aimed at expanding current understanding of the role of novelty in memory encoding, both in terms of cognitive processes and neural correlates. 1.1 Cognitive evidence for benefits of familiarity on learning Items may become familiar through repetition in an experimental session or through pre-experimental exposures in the course of ongoing experience. Classical and more recent experiments have linked both forms of familiarity with mnemonic advantages. Concerning the effects of familiarity established through repetition, Hermann Ebbinghaus, when studying lists of nonsense syllables until he achieved errorless recitation, found that learning on any given trial was fastest when lists had been previously studied (Ebbinghaus, 1913). Repetition benefits in list-learning experiments have been replicated hundreds of times (Hintzman, 1976). I will refer to this form of familiarity established during a single experimental session, prior to a study phase, and through in-laboratory repetition of stimulus materials as repetition-based familiarity. Concerning the effects of item familiarity established through pre-experimental exposures in the form of semantic memory, it has been argued that semantic knowledge is a prerequisite for episodic memory (Tulving & Markowitsch, 1998). Indeed, children s episodic memory for pictures of objects appears to be limited by their semantic knowledge about the objects depicted (Robertson & Kohler, 2007). Other evidence from semantic dementia and Alzheimer s disease patients suggests that intact semantic representations of items are not strictly required to form episodic memories for items, but semantic item representations enhance perceptual flexibility of memories, facilitating recognition of perceptually different, but conceptually identical, items (Graham, Simons, Pratt, Patterson, & Hodges, 2000). Because semantic knowledge is defined in these experiments with respect to the learner s entire history, it could

16 3 alternately be described as familiarity based on prior knowledge, a more theoretically neutral description I will adopt to provide clear juxtaposition against repetition-based familiarity. Materials with neither form of familiarity (that are seen for the first time in the study phase) are described here as novel. 1.2 Cognitive evidence for benefits of novelty on learning The notion that there might be potential benefits of novelty over familiarity for new memory formation arises primarily from a line of cognitive research examining the detrimental effect of pre-study repetition of study materials on participants ability to study the repeated materials later and successfully identify them as studied in a memory test. This phenomenon was first observed by Kinsbourne & George (1974) and later revisited by Tulving and Kroll (1995), who conceptualized it in terms of novelty (i.e., as a novelty effect) and linked it to priming effects in the neuroscience literature. Some have recently suggested that novelty effects may be linked to distinctiveness (e.g., Tulving & Rosenbaum, 2006), a possibility I will consider in a later discussion. To obtain a novelty effect, Tulving and Kroll (1995) used a three-stage verbal memory procedure: in a familiarity induction stage, participants were exposed to two sets of 80 words with six repetitions. During one of the repetitions, recognition memory was evaluated using no lures, and a high rate of endorsement was observed. A subsequent encoding phase required participants to study words for a later memory test. Stimuli included one of the repeated sets of 80 words and a new set of 80 words. Finally, participants were asked to distinguish between the words that appeared in the study list (including one familiar and one novel set) and lures (including the unstudied familiar set and an unstudied novel set). Accuracy scores (hits minus false alarms) indicated substantially higher recognition performance for the novel words than for the repeated words, leading the researchers to conclude that novel stimuli are encoded better than familiar stimuli. This basic finding has been replicated many times (Aberg & Nilsson, 2001; Aberg & Nilsson, 2003; Kormi-Nouri, Nilsson, & Ohta, 2005). These novelty findings, when considered alongside those showing mnemonic benefits of familiarity (i.e., familiarity effects), appear paradoxical. As it is a contradiction that both familiarity and novelty enhance memory, it must be the case that one of the observations is incorrect, perhaps due to a faulty experimental design, or that an

17 4 unknown third variable exists that determines the times when it is novelty or familiarity that enhances memory. I will return to consider these possibilities after a review of the evidence available from the cognitive neuroscience literature and an evaluation of whether this literature supports one set of findings over the other. 1.3 Cognitive neuroscience support for novelty advantages in memory At about the same time that Tulving and Kroll were investigating novelty advantages in episodic memory (1995), Tulving and a group of collaborators (Tulving, Markowitsch, Kapur, Habib & Houle, 1994) employed a similar manipulation to search for the neural correlates of novelty and familiarity using neuroimaging. In their experiment, one day before scanning, participants were presented twice with a set of 80 photographs. One day later, they were scanned using positron emission tomography (PET) as they viewed the 80 repeated stimuli in blocks and 80 novel stimuli in separate blocks. The comparison of activity in these block types was noteworthy for revealing a significant difference in the hippocampus, a brain structure which, although the focus of significant theoretical interest to memory researchers, had not responded to any previous neuroimaging manipulation. It was nonetheless well established at the time, based on animal and neuropsychological evidence, that the hippocampus played a crucial role in long-term memory (Squire, 1992). Accordingly, it was reasonable to expect that any increase in hippocampal activity should be linked with conditions in which long-term memory was engaged, such as conditions involving previouslypresented stimulus materials likely to cue memory retrieval. Tulving and colleagues observed the opposite pattern, with greater activity in the novelty than familiarity condition. Thus, a second and lasting contribution of the study by Tulving and colleagues was to demonstrate that the hippocampus was sensitive to novelty. It was not long before such hippocampal novelty activations, defined as greater activity in response to novel relative to familiar stimuli, were again observed using other photographic materials (Stern et al., 1996; Tulving et al., 1996). In addition, converging evidence became available from the patient and animal literatures. An intracranial electrode recording study by Knight (1996) revealed that stroke patients with damage to

18 5 the posterior hippocampus did not show the normal event-related potential (ERP)- based differentiation of novel and repeated tone bursts that was present in controls. In another study, Grunwald and colleagues (Grunwald, Lehnertz, Heinze, Helmstaedter & Elger, 1998) observed reduced ERP-based differentiation of new and old verbal materials in temporal lobe epilepsy patients in the hemisphere of epileptogenic focus relative to the unaffected hemisphere. This pattern was apparent in epilepsy patients presenting with hippocampal sclerosis but not those with extrahippocampal damage only, and therefore further implicated the hippocampus in novelty processing. In animals, disruption of electrophysiological novelty responses was observed following hippocampal ablation (Vinogradova, 2001). Because of the known role of the hippocampus in long-term memory, the phenomenon of hippocampal novelty activations raised the tantalizing possibility of convergence with cognitive novelty effects. This idea was first formalized as the novelty-encoding hypothesis, which in its initial formulation posited that hippocampal encoding mechanisms only operate on information to the extent that the information is novel (Tulving et al., 1996). In this way, the hypothesis explained novelty advantages in cognitive tests in terms of activity in the hippocampus. According to the noveltyencoding hypothesis, this pattern arises because information passes through a novelty assessment stage in which brain structures filter out familiar information. Only information deemed to be novel is relayed to the hippocampus for encoding into longterm memory. It is reasonable to interpret this screening process as concerned with cognitive economy, because it reduces the load on long-term memory systems by eliminating the storage of potentially redundant information. Similar biological models have been developed by others that would support hippocampal selectivity of novel information for encoding into memory (e.g., Borisyuk, Denham, Hoppensteadt, Kazanovich, & Vinogradova, 2001; Lisman & Grace, 2005). These proposals received important support from one neuroimaging study that directly linked hippocampal novelty responses to memory encoding. Kirchhoff and colleagues (2000) presented participants with novel and repeated pictures and words and later tested recognition memory for the novel materials. As in earlier studies, hippocampal novelty activations were observed, although these were more apparent for scenes than for words. Consistent with earlier findings that MTL activity at encoding could be used

19 6 to predict subsequent memory (Brewer, Zhao, Desmond, Glover, & Gabrieli, 1998; Wagner et al., 1998), the researchers also identified regions of the hippocampus that correlated with subsequent memory success on the recognition memory test. Critically, hippocampal regions that predicted subsequent memory success (i.e., encoding regions), overlapped directly with regions where hippocampal novelty activations were observed. This evidence suggests that novelty leads to increased activity in regions that support long-term memory encoding. However, as memory was measured only for novel materials, it remains possible that familiar ones were similarly associated with greater activity in a different memory-encoding region. 1.4 Evidence for hippocampal sensitivity to familiarity Not all cognitive neuroscience evidence is as supportive of a novelty-encoding account of hippocampal function: reports of familiarity activations in the hippocampus appear to oppose this account. For example, in the context of several tasks during PET scanning, Kapur, Friston, Young, Frith and Frackowiak (1995) presented participants with novel and repeated non-famous faces, as well as famous faces that were familiar due to prior exposure. Contrary to reports linking hippocampal activity to novelty, hippocampal activity was greater in conditions in which stimuli were familiar than in conditions in which stimuli were novel. This was true whether stimuli were familiar due to repetition or due to prior knowledge. Similar observations of greater hippocampal activity in response to old relative to new items have been reported many times (Lepage et al., 1998). To help explain conflicting observations of novelty and familiarity effects in the hippocampus, LePage and colleagues (1998) attempted to distinguish the spatial coordinates of the two types of effects. In a meta-analysis of PET experiments, the researchers found that the majority of activation foci in the anterior hippocampus were encoding effects (such as greater activity for new vs. old materials), whereas the majority of activation foci in the posterior hippocampus were retrieval effects (such as greater activity for old vs. new materials). However, a follow-up meta-analysis by Schacter and Wagner (1999) that included additional PET and fmri evidence revealed this longitudinal organization to be less clear-cut than LePage and colleagues

20 7 description, with encoding foci distributed along the full longitudinal axis of the hippocampus. More recent fmri evidence has revealed hippocampal sensitivity to both new and old materials in the same dataset, reviving the possibility of differentiation along the longitudinal axis of the hippocampus. Prior to fmri scanning, Daselaar and colleagues (2006) presented participants with words and non-words in a lexical decision task. During scanning, they indicated whether the words were novel or old as well as their subjective confidence about their judgments. The highest-confidence old responses were characterized as recollection responses, corresponding to a form of ecphory with detailed information about the manner in which past events were experienced, interpreted and understood at the time of their occurrence (Tulving, 1983). Along with greater participant certainty that words were novel, the researchers observed increasing signal intensity in the anterior hippocampus. In contrast, high-confidence recollection responses were associated with greater signal intensity in the posterior hippocampus, and intermediate levels of confidence in oldness were correlated with other MTL regions. The observation of an anterior hippocampal novelty signal and a posterior hippocampal recollection signal in the same dataset provides direct evidence for regional differentiation in hippocampal sensitivity to novelty versus familiarity. An additional complication for efforts to spatially segregate novelty and familiarity sensitivity arises based on intra-cranial evidence of the type collected by Rutishauser and colleagues (2006). Using microwire recording from pre-surgical temporal lobe epilepsy patients, the researchers identified individual neurons in the anterior hippocampus that were sensitive to the novelty or familiarity of images. Approximately equal numbers of novelty and familiarity-sensitive neurons were observed in very close proximity in the anterior hippocampus, in some cases along the same microwire. While this evidence suggests that any absolute segregation of novelty- and familiaritysensitive neurons in the hippocampus is unlikely, the mass action of many hippocampal neurons covering a large area may nonetheless yield aggregate signals biased towards one or the other form of processing. This interpretation is supported by the fact that regional differences in novelty versus familiarity sensitivity do in fact routinely appear in neuroimaging contrasts (Lepage et al., 1998; Schacter & Wagner, 1999; Daselaar et al., 2006).

21 8 1.5 Interpreting the available evidence Summarizing the findings from the cognitive neuroscience literature, both novelty and familiarity activations may be found in the hippocampus, with some evidence for an anterior bias for novelty processing and a posterior bias for familiarity processing. These findings are not incompatible with a novelty-encoding account of hippocampal function, but neither do they favour it. The strongest support for a novelty-encoding account of hippocampal function remains the findings by Kirchhoff and colleagues (2000), who identified a hippocampal novelty activation that overlapped with an encoding region. However, their study did not evaluate how familiar stimuli are encoded. If the hippocampus predicted memory for such stimuli perhaps indicated by an encoding area overlapping with a familiarity activation it would argue against a privileged connection between novelty, the hippocampus, and memory. As the cognitive neuroscience literature does not provide clear support for either the noveltyencoding hypothesis or any other novelty-driven hypothesis, this literature cannot be used to resolve the original discrepancy observed between cognitive studies that show novelty effects and those that show familiarity effects. Possible factors leading to the seemingly paradoxical discrepancy in cognitive studies may be identified based on an appreciation of the differences in the paradigms used to obtain each type of result. In particular, two factors may explain why both novelty and familiarity effects can be observed, albeit under different conditions. I will refer to the information required of participants for successful memory discrimination as the discrimination demands. Rather than any special property of novel stimuli, the first factor concerns the possibility that novelty effects are explained by different discrimination demands for novel and familiar stimuli in paradigms where a novelty effect is obtained (Dobbins, Kroll, Yonelinas, & Liu, 1998; Greene, 1999). In such studies, better discrimination is typically required to reject familiar lures than novel ones. To reject familiar lures, participants must use source information to determine whether the item was seen during familiarity induction only, or was also seen in the encoding phase. In contrast, novel lures may be rejected by the absence of item familiarity alone, since these lures were not previously seen during the experiment. Because recognition accuracy

22 9 scores are calculated using both the rate of hits and the rate of false alarms (either through direct subtraction or calculation of the discriminability index d ), and only familiar lures are subject to confusion arising from the presence of multiple possible sources (i.e., source confusion), it may not be fair to compare recognition accuracy scores for novel and familiar items in these experiments. To illustrate the problem with this approach, it would not be appropriate to evaluate memory for friends and strangers based on one s ability to determine whether a particular friend and stranger attended a birthday party. Whereas it would be simple to recall that no stranger attended, it would be more difficult to rule out a friend who had attended past birthday parties. Just as this does not indicate that episodic memory for strangers is superior, so it is not fair to say episodic memory for novel items is superior based on comparable evidence. In support of this contention, Maddox and Estes (1997) found that separating the familiarity-induction and study phases by 24 hours led to fewer false alarms relative to no delay. This memory improvement, linked as it was with increased discriminability of sources, suggests that discrimination demands do play some role in memory for repeated materials. To circumvent this issue, some authors have focused on hits independently of lures, reporting novelty effects based on hits alone (e.g., Aberg & Nilsson, 2003); however, high discrimination demands could easily produce lower accuracy by way of fewer hits, more false alarms, or both, depending on participants specific response bias. A second possible explanation for why both novelty and familiarity effects are observed concerns the type of memory and familiarity under investigation. Studies of the kind reported by Robertson and Kohler (2007) involve manipulations of familiarity based on prior knowledge, whereas studies revealing novelty effects of the kind reported by Tulving and Kroll (1995) have manipulated repetition-based familiarity. Although Ebbinghaus reported mnemonic benefits of repetition-based familiarity (1913), his experiments assessed implicit memory measures (time in seconds or number of repetitions required for list mastery) rather than an explicit measure of episodic memory. Therefore, it is possible that prior knowledge-based familiarity has beneficial effects on episodic memory, whereas repetition-based familiarity has detrimental effects.

23 Overview of experiments In the above literature review, I identified several areas of conflicting findings in the cognitive and cognitive neuroscience literatures that bear upon the role played by novelty in the formation of new episodic memories. As current theory holds that novelty is an important factor indeed, in some formulations, a pre-requisite for human memory-encoding, the resolution of these problem areas is an important step towards developing a firmer understanding of the processes underpinning human memory function. In the current dissertation, I report a series of cognitive and cognitive neuroscience experiments aimed at helping to resolve conflicting findings in these areas. In Chapter 2, I describe cognitive evidence exploring the circumstances under which novelty rather than familiarity enhances memory encoding. The roles of discrimination demands and of the form of familiarity manipulated in the experiment are explored in two experiments. In Chapters 3 and 4, I describe fmri experiments in which I compare neural responses to novel and familiar materials as well as the neural predictors of how successfully each type of material is encoded. The experiment described in Chapter 3 involves the application of procedures developed in Chapter 1 in an fmri environment, allowing for evaluation of the neural correlates of processes isolated by those procedures. In Chapter 4, I attempt to replicate the findings from Chapter 3 using different materials and describe new comparisons involving alternate forms of familiarity. Because source information is considered an important attribute of episodic memory (Tulving, 1983) and can be used to measure mnemonic retention of details from a single, unique event, I use source memory as a measure in all of the experiments described in the current dissertation. In addition, while some recent work has explored effects of other forms of novelty, such as contextual and associative types, the current dissertation is constrained in scope to the role of stimulus novelty, since prevailing theory concerning the mnemonic and neurological effects of novelty has been established by manipulating novelty in this form.

24 11 Chapter 2 Resolving conflicting novelty and familiarity findings in cognition Because it seems contradictory that stimulus novelty could both facilitate and impair memory, there are two plausible explanations for apparently conflicting cognitive findings of this kind: one set of findings could be grounded in a faulty experimental design, or an unknown factor could be responsible for conditions in which the mnemonic effects of novelty are either positive or negative. As novelty and repetition advantages have been replicated by many different authors in many different experimental setups, it is more likely that an unknown factor is responsible for the conflict. The current chapter will examine the effects of two potentially important factors between cognitive studies that reveal novelty versus familiarity effects: differences in the role of discrimination demands (distinguishing targets and lures) and differences in the form of familiarity under study. The first factor, the role of discrimination demands, concerns criticism that has been levied at the experimental paradigm used by Tulving and Kroll (1995) to compare encoding of novel and repeated information. This factor describes the possibility that different types or amounts of information are required from participants to correctly respond to test items in novelty and familiarity conditions. As discussed in Chapter 1, the paradigm devised by Tulving and Kroll involves an uneven comparison of memory for repeated and novel items: distinguishing targets from recently-presented lures requires information about the source of items, whereas distinguishing targets from novel lures does not (Dobbins et al., 1998). The second factor concerns the different ways in which manipulations of novelty and familiarity have been manipulated. In some experiments, familiarity is operationalized as within-session experimental repetition of materials, whereas in others, it is characterized as prior, pre-experimental knowledge of materials. To evaluate the possible roles of these factors, a paradigm is needed that i) permits measurement of novelty and familiarity effects in a context where memory discrimination demands in novelty and familiarity conditions can be

25 12 equated or controlled; and ii) compares the impact of experimental versus preexperimental forms of familiarity on memory. In Experiment 1 of the current chapter, I designed a novelty paradigm that contained a familiarity condition and a novelty condition, as is typical, but also included a second novelty condition in which repetitions of targets and lures occurred after the study phase and before the memory test. If stimulus novelty at initial presentation at study underlies the novelty effect, then performance should be reduced when repetition occurs before the study phase but not when it occurs afterwards. On the other hand, if a difference in discrimination demands (i.e., the information needed to distinguish targets from lures) is the determining factor in novelty experiments, performance should be equivalent in the two conditions because the discrimination demands are the same. In Experiment 2, because familiarity effects in some classic studies are based on preexperimental familiarity, whereas novelty effects are based on experimentally-induced familiarity, I compared memory for stimuli that were novel, pre-experimentally familiar, or familiar due to experimental induction. In this experiment, I eliminated differences in discrimination demands by measuring memory for contextual features of encounters with familiar and novel information. The two experiments described in this chapter were conducted with native Englishspeaking participants, and used English and Asian proverbs as stimuli. These materials allowed us to introduce experimental familiarity through in-lab repetition of Asian proverbs and pre-experimental familiarity through real-world experience with English proverbs. 2.1 Experiment 1 In Experiment 1, I directly evaluated whether novelty effects are related to stimulus novelty, to different discrimination demands in the familiarity and novelty conditions, or to both factors. I undertook this task by implementing a typical novelty experiment paradigm and adding a second novelty condition, which I named the novel repeated condition because stimulus repetitions occurred after study but before the memory test (Table 1). This condition differed from the familiarity induction condition in terms of the timing of stimulus repetitions, which took place after, rather than before, the study

26 13 Table 1. Schematic of experimental protocol and stimulus exposure in Experiment 1. Stimuli consisted of three lists of 25 Asian target proverbs (Familiar TARG, Novel- Repeated TARG, and Novel TARG ) and three lists of 25 Asian lure proverbs (Familiar LURE, Novel-Repeated LURE, and Novel LURE ). Phase and purpose Lists presented and task instructions Phase I. Three repetitions for familiarity induction Familiar TARG Familiar LURE South American or Japanese? Phase II. Incidental encoding of proverbs in a valence task Familiar TARG Novel-Repeat TARG Novel TARG rate valence Phase III. Three repetitions for familiarity induction Novel-Repeat TARG Novel-Repeat LURE South American or Japanese? Phase IV. Test of memory for Phase II presentations (all items) rated valence?

27 14 phase. As a result, novel repeated items were novel at the time of the study phase, whereas repeated items were not. In contrast, both novel and novel repeated items were novel at study, but differed in terms of whether or not the stimuli were repeated following study along with new lures that were also repeated (Table 1). How accuracy fared in the novel repeated condition was central to the current experiment. If novelty effects arise from a stimulus novelty-based enhancement of memory encoding at study, then post-study repetition of targets and lures should have little effect. However, if novelty effects are a consequence of differences in discrimination demands between novelty and familiarity conditions, then no memory advantage should be observed for the novel repeated condition over the familiarity induction condition, and memory for both conditions should be at a disadvantage relative to proverbs in the novel condition Methods Participants Twenty-nine students of the University of Toronto, all English native-speakers with normal or corrected-to-normal vision and hearing, participated in the experiment (15 female; mean age 19.8). Two additional participants were excluded for failing to engage in the experimental tasks. Participants were screened for the absence of neurological and psychiatric conditions and received academic credit or financial compensation for their participation. The protocol for this experiment was approved by the Ethics Review Board at the University of Toronto Stimulus materials A list was created containing 150 Chinese and Japanese (Asian) proverbs translated into English (Appendix A). Each proverb consisted of a complete sentence at least five words in length that included no archaic or vernacular language. Some of the Asian proverbs were modified from their literal translations to ensure smooth, concise reading in English and to minimize the use of culturally specific terms or concepts. Proverbs from this list were randomly allocated to the familiarity induction, novel repeated and novelty conditions for each participant, thereby creating three sub-lists containing 50 items each. Each sub-list was then split evenly between targets and lures, resulting in 25 targets and 25 lures for each condition (Table 1).

28 Experimental tasks Participants were informed that the experiment consisted of several phases that together would take an hour to complete. Prior to each task, they were reminded that accuracy was the most important aspect of their response, but that the speed of their response was also important. The procedure consisted of four main phases: (I) familiarity induction through in-lab repetition of proverbs; (II) an encoding phase in which proverbs were familiar or novel; (III) repetition of a set of novel proverbs in the novel repeated condition; and (IV) a test phase, in which a recognition memory test was administered for the proverbs encountered in Phase II (Table 1). The pre- and post-study repetition phases allowed us to distinguish the effects on memory of familiarity induction (repetition before study) from the effect of repetition in general. In a familiarity-induction session (Phase I), participants were presented with 50 Asian proverbs that had been randomly allocated to the repetition condition (Table 1). Each proverb was visually presented for 3.5 s and was followed by presentation of a fixation cross for 0.8 s. During the time that each proverb was visible, participants were asked to indicate with a button press whether each proverb was South American or Asian in origin (while in fact all proverbs were Asian) to encourage deep encoding of the proverbs. Response key mappings were presented at the bottom of the screen. The full list was presented in random order three times; participants were told they had three tries and were asked to give each proverb fresh consideration each time. During an incidental encoding phase (Phase II), participants saw the 25 target familiar proverbs and 50 target novel ones (25 novel repeated and 25 novel proverbs; see Table 1). The materials were presented at the same rate and in the same format as in Phase 1. Participants rated the subjective valence of each proverb on a scale from 1 (very negative) to 5 (very positive), a task selected to help distinguish the study phase from other phases. Post-study repetition (Phase III) of the 25 novel repeated targets and 25 novel repeated lures took place following the encoding phase (Table 1). Aside from the use of different materials, Phase III was identical to Phase I. Participant responses from Phases I

29 16 through III were used only to ensure that participants engaged in the experimental tasks. At the end of the experiment, a surprise recognition memory test was administered for the items presented in Phase II. In this test, participants were asked to indicate with a button press whether or not each proverb was present in the Phase II study list and disregard whether they had seen the proverbs in the familiarity induction phase. In other words, participants decided whether or not they had rated proverbs in terms of valence (the Phase II incidental encoding task) while ignoring any memories related to rating the cultural origin of proverbs (the Phase I and III repetition tasks). Target items in the test included all 75 proverbs from Phase II, namely the 25 familiar, 25 novel repeated, and 25 novel proverbs (Table 1). Lures included 25 familiar proverbs that were seen in Phase I (familiarity induction) but not Phase II (study), 25 novel repeated proverbs that were seen in Phase III but not in Phase II, and 25 unseen (novel) proverbs. The targets and lures were presented in random order, again at the same rate as in earlier phases Results and discussion A high rate of responding in the Phase I and III repetition tasks indicated that participants engaged in these sections (Phase I: M = 0.98, SD = 0.05; Phase II: M = 0.97, SD = 0.04). On average, they did not show a particular preference for categorizing proverbs as belonging to a particular cultural origin: the mean proportion of proverbs identified as South American was not significantly different from an even 50/50 split in either the Phase I repetition task, M = 0.49, SD = 0.09, t(28) = 0.49, P = n.s., or the Phase III repetition task, M = 0.48, SD = 0.07, t(28) = 1.54, P = n.s. Accordingly, no distinctiveness advantage was expected for any of the repeated proverbs on the basis of responses in these sections. As a preliminary manipulation check, I first determined whether the novelty effect previously reported by Tulving and Kroll (1995) and others (Aberg & Nilsson, 2001; Aberg & Nilsson, 2003; Kormi-Nouri, Nilsson, & Ohta, 2005) was replicated. I computed a d measure of accuracy for the novelty condition using the rate of hits to novel targets and rate of false alarms to novel lures, and for the familiarity condition using the rate of hits to familiar targets and rate of false alarms to familiar lures. As is typical in novelty

30 17 studies, d was higher in the novel condition than the familiar condition, t(28) = 7.06, P < (Fig. 1; Table 2). To limit the number of t-tests performed in my analysis, I did not assess statistical differences in hits and false alarms. Numerically, there was a higher rate of hits for familiar items than for novel ones, but also a higher rate of false alarms to familiar lures than to novel ones. Having successfully replicated a novelty effect in my paradigm (with a characteristically large effect size, Cohen s d = 1.20), I next explored the relationship between the familiar and novel repeated conditions: to the extent that stimulus novelty enhances memory for materials, memory for novel repeated items should be superior. However, no novelty effect was observed in my d measure of accuracy, t(28) = 1.19, P = n.s. As I observed in the comparison between familiar and novel items, there was a numerically higher rate of hits for familiar target items than novel repeated targets, and a higher rate of false alarms to familiar lures than novel repeated lures (Fig. 1; Table 2). I also compared memory in the novel repeated and novel conditions. Comparisons between these conditions can be interpreted as a direct test of the influence of discrimination demands (i.e., whether the different information required for novelty and familiarity conditions had any impact on performance in the memory test). Accuracy scores were higher in the novel condition, t(28) = 5.36, P < 0.001, indicating that the repetition of stimuli outside of the study phase did introduce significant levels of source confusion. Numerically, there was little difference in the rate of hits between the conditions possibly reflecting the equivalence of the conditions during the study phase although there was a higher rate of false alarms to novel repeated lures than to novel lures (Fig. 1; Table 2). The current experiment strongly suggests that novelty effects, as typically measured, arise from different discrimination demands in novelty and familiarity conditions. The findings indicate clearly that source confusion, rather than stimulus novelty, produces the novelty effect. The introduction of post-study repetitions of targets and lures reduced the memory advantage for items that appeared in the study phase for the first time (novel), just as pre-study repetition (familiarity induction) did for items that were familiar at study.

31 18 Table 2. Descriptive statistics for hits, false alarms (FA), and two measures of corrected accuracy (acc. and d') for pre-study repetition items, novel repeated items and novel items in the Experiment 1 memory test. Mean Standard deviation Standard error Hits FA Acc d' Hits FA Acc d' Hits FA Acc d' Repeated before study Repeated after study Novel

32 (B!"#$%&'( '"!!"& #"! C<D0/ ;1<=<1*)<>!"%!"$ '"#!"&!"!"#!"$!"! ()*+,-./+0./ )*+9-3: !":!"!!"#$%&'()'*+,&%"-&./'('-&-0%1',&%20%-3.4&)')"*+,'-./+&'./.%0+,'.12'34%%&3*&2'.33$%.35'+34%&+'-4%'6%&7+*$25'%&6&*"*"41'"*&0+,'148&/'%&6&.*&2'"*&0+'.12' 148&/'"*&0+'"1'*9&':;6&%"0&1*'('0&04%5'*&+*<':%%4%'=.%+'2&6"3*'>('+*.12.%2'

33 20 What is still needed to understand the impact of novelty and familiarity on memory is a comparison of their impact in a context where discrimination demands are equivalent across all conditions. I addressed this need in a second experiment. In this same experiment, I evaluated the impact of the second potentially important factor discussed earlier: the form of familiarity (pre-experimental versus prior knowledge) that was employed to distinguish novel items from familiar ones. 2.2 Experiment 2 I used a three-phase design, with the phases corresponding to familiarity induction, study of novel and familiar materials, and memory testing (Table 3). To the extent that episodic memory incorporates critical information about the time and place of previous experiences, source information is a central attribute of episodic memory (Tulving, 1983; Johnson, 2005). Accordingly, to circumvent the discrimination demand confound observed in Experiment 1, I altered the memory test such that successful selection of targets required contextual information that was specific to individual encoding events. Using this approach, the same amount and type of information was required for successful item discrimination in all conditions of the memory test. To provide distinguishing contextual information, participants rated half of each type of material on one scale (vividness) and the other half on another (valence); in the test phase, they decided on which scale they had rated each proverb. Using this design, I compared the effects of pre-experimental familiarity, experimentally-induced familiarity, and novelty on memory Methods Participants Twenty-eight students of the University of Toronto, all English native-speakers with normal or corrected-to-normal vision and hearing, participated in the experiment (15 female; mean age 19.8). All participants had less than one year of experience with Asian languages or culture and all but one had at least one parent or guardian who was also an English native-speaker. One additional participant was excluded for failing to

34 21 Table 3. Schematic of experimental protocol and stimulus exposure in Experiment 2. Stimuli consisted of two lists of 20 English proverbs (English 1 and English 2 ) and four lists of 20 Asian proverbs (Asian-Familiar 1, Asian-Familiar 2, Asian-Novel 1, and Asian- Novel 2 ). Phase and purpose Lists presented and task instructions Phase I. Three repetitions for familiarity induction Asian-Familiar 1 Asian-Familiar 2 South American or Japanese? Phase II. Incidental encoding of proverbs in two tasks English 1 Asian-Familiar 1 Asian-Novel 1 rate vividness English 2 Asian-Familiar 2 Asian-Novel 2 rate valence Phase III. Test of memory for Phase II source information (all items) rated vividness or valence? Phase IV. Identification of proverbs known prior to the experiment (all items) learned today, or know from prior knowledge?

35 22 engage in experimental tasks. Participants were screened for the absence of neurological and psychiatric conditions and received academic credit as compensation for their participation. The protocol for this experiment was approved by the Ethics Review Board at the University of Toronto Stimulus materials Two base lists of proverbs were prepared, one containing 40 common English proverbs and the other 80 Asian proverbs (Appendix A). Each of the Asian proverbs was unfamiliar to at least eight of twelve undergraduate students polled in a preliminary norming investigation, whereas each of the English proverbs was familiar to at least eight students from the same group. The Asian proverb list was randomly divided evenly between the repetition condition and novelty condition to create two lists of 40 Asian proverbs Experimental tasks The procedure consisted of three main phases: (I) familiarity induction of Asian proverbs; (II) an encoding phase in which proverbs were novel, pre-experimentally familiar or familiar through experimental induction; and (III) a test phase, in which a source memory test was administered for the proverbs encountered in Phase II (Table 3). Familiarity induction (Phase I) was the same as in Experiment 1. All participants decided three times whether 40 Asian proverbs were South American or Japanese in origin (Table 3). During the study phase (Phase II), participants saw the 40 preexperimentally familiar English proverbs, 40 experimentally familiarized Asian proverbs and 40 unseen (novel) Asian proverbs (Table 3). One randomly-allocated half of each list (20 proverbs) was presented as part of a vividness rating task: participants were asked to rate the intensity of the mental imagery evoked by each proverb on a scale from 1 (not vivid) to 5 (highly vivid). The other half of each list was presented as part of a valence rating task: using a keyboard, participants rated how positive each proverb was on a scale from 1 (negative) to 5 (positive). Participants completed each rating task in 8 short blocks of 6 items (2 from each list). The rate and manner of presentation were the same as in Phase I, and responses from these first two phases were used only to ensure that participants engaged in the experimental tasks.

36 23 In a subsequent memory test (Phase III), participants were shown the full set of 40 preexperimentally familiar English proverbs, 40 experimentally familiarized Asian proverbs, and 40 novel Asian proverbs that were presented during Phase II (Table 3). Participants were asked to indicate using a button press whether they had rated each proverb for its vividness or its valence. The rate and manner of presentation were the same as in Phase I, although the inter-stimulus interval began as soon as a response was detected. At the end of the experiment, to confirm my assumption that participants would have better prior knowledge of English proverbs than Asian ones, I asked participants to complete a proverb identification task (Phase IV). In this task, they indicated which proverbs they knew prior to the experiment using a button press (Table 3). All materials used in the experiment were presented Results and discussion As in Experiment 1, a high rate of responding in the Phase I repetition task indicated that participants engaged in this task (M = 0.97, SD = 0.05). Again, the mean proportion of proverbs identified as South American was not significantly different from an even 50/50 split, M = 0.50, SD = 0.05, t(27) = 0.24, P = n.s. In the Phase IV proverb identification task, participants categorization of English and Asian proverbs was generally consistent with my expectation that they would know the English but not Asian proverbs prior to the experiment (known English proverbs: M=80.0%, SD=14.0%; known Asian proverbs: M=15.5%, SD=12.1%). Accordingly, I proceeded to explore the effect of experimentally-induced and pre-experimental familiarity on source memory measured in Phase III. Source memory accuracy for novel proverbs was lower than that for pre-experimentally familiar proverbs, t(27)=2.20, P<0.05, and experimentally familiarized proverbs, t(27)=3.02, P<0.01 (Fig. 2; Table 4). There was no difference in memory between the two types of familiar proverbs, t(27)=0.41, P=n.s. Based on these findings I conclude that familiarity, rather than novelty, provides an episodic memory advantage when discrimination demands are made equivalent, whether familiarity is pre-experimental or experimental in nature. This pattern is consistent with the established view that prior memory representations facilitate the

37 24 Table 4. Descriptive statistics for Experiment 2 source memory accuracy for preexperimentally familiar proverbs, proverbs with experimentally-induced familiarity, and novel proverbs. Source of familiarity Mean Standard deviation Standard error Prior knowledge Repetition None (novel)

38 (C!"#$%&'( 2354!"#$#"%&#'()#""*)% ???!"&#" 6'#70*89* :*$*%&%&#' +#,")*(#-(-./&0&."&%1 ;#'* <'#=*0>!"#$%&'()'*+,&%"-&./'('-&-0%1',&%20%-3.4&)')*+&%",&-.'('/0$%1&',&,0%2' 311$%312'40%'+%&5&*+&%",&-.3662'43,"6"3%'+%07&%8/9'+%07&%8/':".;' &*+&%",& "-<$1&<'43,"6"3%".29'3-<'-07&6'+%07&%8/='>'/"#-"4"13-.'<"44&%5

39 26 episodic encoding of information (e.g., Wickelgren, 1979; Hintzman, 1988; Tulving & Markowitsch, 1998; Moscovitch et al., 2005). 2.3 General discussion of Chapter 2 experiments In the current chapter, I attempted to resolve conflicting findings concerning beneficial or detrimental effects of novelty on memory by exploring two critical factors in experiments measuring novelty: i) type of familiarity (experimentally-induced versus pre-experimental) and ii) discrimination demands (equal versus unequal). In Experiment 1, I found that the novelty effect, as observed in the paradigm originally used to isolate it, is driven by differences in discrimination demands across conditions, rather than the influence of stimulus novelty. In Experiment 2, I found that type of familiarity did not account for the divergent findings of classical studies and novelty studies. I also found that context memory for familiar proverbs was consistently better than context memory for novel ones when discrimination demands were equated across conditions. Although my results cast doubt on the hypothesis that novel information enhances episodic memory encoding (Tulving & Kroll, 1995), they do support the idea of a link between novelty and distinctiveness. Numerous serial list-learning experiments have shown materials that are distinctive with respect to some intra-list contextual pattern, such as a common font or semantic category, are better remembered than other items (von Restorff, 1933; Hunt & Lamb, 2001). In novelty experiments of the kind designed by Tulving and Kroll, differences in discrimination demands may be considered a direct reflection of differences in distinctiveness. When events are based on entirely novel materials or even materials that are novel within a particular laboratory visit, they are distinctive as they may be identified on the basis of item information alone (recognition memory). In contrast, events based on materials repeated during a particular session must be distinguished from similar past events involving the same materials on the basis of specific contextual information (source memory). Viewed in this way, the novelty effect may be appreciated as a distinctiveness phenomenon, with different levels of discrimination demands for novel and familiar items acting as a mechanism

40 27 driving superior memory for novel items (rather than a confound clouding memory comparisons of novelty and familiarity conditions). The case for novel items exerting beneficial effects on memory through increases in distinctiveness has been made by a number of authors (e.g., Hunt & Lamb, 2001; Kishiyama & Yonelinas, 2006; Tulving & Rosenbaum, 2006). However, and returning to one of the goals of the current study, my objective was to test whether memory for source or other contextual information of single episodes (i.e., episodic memory) differed for novel and familiar stimuli. Working towards this goal, differing discrimination demands for novelty and familiarity conditions (i.e., item memory for novel items versus source memory for familiar ones) were indeed a confound. Once these were controlled, a familiarity episodic memory advantage was observed. A controlled comparison of this type has been needed as a critical test of the novelty-encoding hypothesis, which concerns episodic memory (Tulving & Kroll, 1995). Drawing on my results, I argue that previous claims of beneficial effects of novelty on memory, including the proposal that information is encoded only to the extent of its novelty (Tulving et al., 1996), appear to be based on over-generalizations of effects arising from the distinctiveness of novel materials. It is important to note that I did not aim to identify the mechanisms that underlie the observed familiarity effects in episodic memory in the current set of experiment. It is possible the effects arise because different types of information are encoded on initial and subsequent exposures to materials: representations of a novel item may be encoded during a comprehension process engaged during a first exposure and may contain little information other than a specification of the stimulus itself. In contrast, information regarding the item s context crucial for episodic memory may be the focus of encoding during subsequent exposures. Alternatively, item memory at encoding may be more likely to fail for novel items than familiar ones, leading to related source memory failures. As the current study was designed to explore the nature of conflicting novelty and familiarity effects in episodic memory, my results do not allow us to state whether either of these possible processes underlie the observed phenomena; however, an assessment of these and other possible processes will be a fine objective for future research. I believe that, irrespective of the mechanisms that may underlie the observed familiarity effects, the phenomenon is of intrinsic interest in light of its implications for the novelty-encoding hypothesis and notions of familiarity enhancing episodic memory formation.

41 28 In summary, my findings indicate that novelty effects arise specifically when source confusion resulting from increases in discrimination demands reduces accuracy for familiar items. Differences in discrimination demands, rather than differences between pre-experimental as compared to experimentally-induced forms of familiarity, appear to explain the long-standing inconsistency between the results from novelty-encoding experiments and those from classical experiments. Thus, my cognitive findings question the core notion of the novelty-encoding hypothesis that information is encoded into long-term memory networks only to the extent that it is novel (Tulving et al., 1996) and support the traditional perspective that memory formation is facilitated by familiarity.

42 Chapter 3 Neural correlates of novelty and familiarity encoding Reports of cognitive novelty advantages have had a considerable impact on the neuropsychological literature, motivating a number of neuropsychological hypotheses and experiments. These same reports have received reciprocal support from the neuropsychological literature: hippocampal activation is often observed at the time of encoding in response to stimulus novelty (e.g., Tulving et al., 1996; Poppenk et al., 2008), which, in light of the known importance of the hippocampus for memory, has been interpreted as support for the point of view that novelty and memory are linked (e.g., Lisman & Grace, 2005). However, in Chapter 2, I found that cognitive novelty effects appear to arise due to differences in discrimination demands between novelty and familiarity conditions rather than some special property of novel materials or the way they are processed. In supporting the perspective that memory formation is facilitated by familiarity rather than novelty, these findings challenge the theoretical basis of these interpretations and others that link novelty activations in the brain to enhanced memory encoding. In so doing, they demand alternative explanations concerning the nature of novelty activations. Formulating such an alternative explanation is difficult to do with the available evidence. Although contrasts of neural responses to novel and familiar items (e.g., (Tulving et al., 1994; Tulving et al., 1996; Poppenk et al., 2008) tell us that novel and familiar stimuli are processed differently, the meaning of these differences is not well understood, other than that they involve the hippocampus and other structures whose importance for memory is well-established. Further limiting possible inference is the fact that both novelty and familiarity effects are associated with activation of the hippocampus (Lepage et al., 1998; Schacter & Wagner, 1999). More direct evidence is available from event-related subsequent memory type studies, which involve postexperimental classification of items as remembered and forgotten based on memory test data (Brewer et al., 1998; Wagner et al., 1998). In evaluating the neural signature of 29

43 30 encoding success, these studies help to link brain activity directly to memory. This is useful in the context of the problem described above, i.e., interpreting hippocampal novelty activations as evidence of the extent to which encoding is taking place. Along these lines, one study revealed overlapping novelty and subsequent memory responses in the hippocampus (Kirchhoff et al., 2000; Stark & Okado, 2003), suggesting that novelty leads to more encoding activity. However, this interpretation requires the assumption that familiar stimuli are encoded using the same encoding network as novel items. Whether or not this is the case is unclear, because no previous study has evaluated how previously familiar items are encoded. This leaves open the possibility that familiar items are in fact encoded in other parts of the hippocampus. This pattern has likely arisen because subsequent memory studies, which are focused on encoding processes, have been designed to minimize contamination based on memory-retrieval events that may be triggered by familiar items. As a consequence, while much is known about how novel information is encoded into memory, little is known about the encoding correlates of materials that are not novel, and the correlates have never been compared. It is possible that familiar materials are encoded using the same network that supports encoding of novel materials, just as it is possible that they are encoded using an entirely different network. Here, I report a first look at how memories are formed when they involve previouslyrepeated materials and compare the functional magnetic resonance imaging (fmri) correlates of encoding for truly novel experiences. This experimental design of this experiment features a three-stage design involving familiarity induction, study and test, similar to the one used in Experiment 2. Behaviourally, I predicted that I would replicate the familiarity effect identified using source memory as a measure in Chapter 1. At the brain level, I hypothesized that the hippocampus would be more sensitive to novel than repeated materials and would also predict later memory for novel materials. As no past study has evaluated subsequent memory predictors for repeated materials, my assessment of memory predictors for these materials was exploratory.

44 Methods The study was conducted in three phases (Fig. 3): (1) a pre-scanning repetition phase, in which participants viewed various indoor and outdoor scenes three times; (2) an encoding phase, during which participants encoded repeated and novel scenes by viewing them under either action or intention instructions while being scanned using functional magnetic resonance imaging (fmri); and (3) a post-scanning phase, in which memory was tested outside the scanner. Phase 1 allowed us to familiarize participants with a number of items prior to their presentation during scanning in phase 2. Data from the phase 3 memory test were used to assess memory for novel and repeated items behaviorally, then to retrospectively categorize events from the phase 2 encoding task as subsequent memory hits or misses (Brewer et al., 1998; Wagner et al., 1998) Participants Sixteen right-handed volunteers from the Greater Toronto Area, all fluent in English with normal or corrected-to-normal vision and hearing, participated in the experiment (9 female; aged 22 to 35 years, mean age 25.8). Participants were screened for the absence of neurological and psychiatric conditions and received financial remuneration for their participation. Of these 16, two were excluded for chance-level performance on behavioral tasks, and a third due to excess imaging artifact. The protocol for this experiment was approved by the Research Ethics Board at Baycrest Hospital in Toronto Stimulus materials A collection of 384 scenes was prepared (320 x 240 pixels). Images were selected from a set of photographs depicting emotionally neutral object configurations, rooms and landscapes. No humans, animals or any well-known landmarks were depicted in the images. For each participant, this collection of scenes was randomly split into two sets of 192 images: a repetition set and a novelty set Procedure Participants were informed that the study consisted of three phases that together would take approximately two and a half hours to complete, and that their memory would be tested at the end of the experiment (see Fig. 3). In the phase 1 pre-scanning repetition

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

46 33 task, participants viewed the repetition set of scenes in a mock fmri scanner. This setting was used in order to maximize the contextual match between phase 1 and a later memory-encoding phase. Participants were informed that their memory for the scenes presented would not be tested, but that they were to view each of the scenes in order to familiarize themselves with the stimulus set. Each image was presented for 2.2 s and was followed by a 0.8 s presentation of a fixation cross. The full set of repetition scenes was presented in random sequence three times. To ensure that participants were alert for this task, they were prompted at several intervals to press a button to continue. During fmri scanning in phase 2, participants encoded scenes into memory while completing several tasks. Data were collected over four functional runs, each containing eight study blocks of eight images. Each block was preceded by 2 s of instruction and 2 s of fixation and was followed by 12 s of fixation. Each image was presented for 4 s and followed by 2 s of fixation. Half of the blocks contained repetition scenes (previously seen three times in phase 1) and the other half contained novel scenes (seen for the first time in phase 2). For each block, participants processed the scenes in accordance with instructions presented before each block and above each image. For blocks associated with action instructions, participants imagined themselves performing an action in the scene; for blocks associated with intention instructions, participants associated a future intention with the scene. Because I wished to relate any novelty and subsequent memory effects to findings reported by other investigators, and these effects are not typically assessed in the context of prospective memory, I did not include any data from the intention encoding condition in the current study (our findings on prospective memory are described elsewhere; (Poppenk, Moscovitch, Mcintosh, Ozcelik & Craik, 2010). Following the presentation of each stimulus, participants pressed a key on an MR-compatible keypad to indicate successful generation of an action or intention. Participants practiced both encoding tasks prior to scanning. In total, 256 pictures were presented, 64 in each of four block types: novel scenes with action instructions, repeated scenes with action instructions, novel scenes with intention instructions and repeated scenes with intention instructions. A different random allocation of scenes to these conditions was arranged for each participant. Each run contained two blocks of each type presented in a random sequence. In a memory test in phase 3, participants viewed on a computer the 128 novel and 128 repeated target scenes encountered during scanning, as well as 64 repeated lure scenes repeated three times in phase I but not

47 34 presented during scanning and 64 novel lure scenes not presented at all during the experiment. Each scene was presented for a maximum of 4 s and was followed by 1 s of fixation. Participants indicated whether each scene was studied as an intention, studied as an action, or not studied during scanning MRI scanning and data analysis All imaging was performed on a 3 Tesla whole-body MRI system (Siemens, Erlangen, Germany). 28 contiguous 5 mm-thick axial oblique slices were obtained, capturing the entire brain volume of each participant. The field of view was 200 by 200 mm (64 x 64 matrix) providing an in-plane resolution of 3 mm. T2-weighted EPI image acquisition was used for all functional scans (TE = 30 ms; TR = 2000 ms; flip angle = 70 ). Each run involved the acquisition of eight initial stabilization volumes that were discarded and 264 task volumes (33 volumes per block with eight blocks). An additional T1-weighted high-resolution MRI volume was obtained for the display of neuroanatomy during the same experimental session using a 3D MPRAGE pulse sequence in the same orientation as the functional scans (160 slices; 1 mm thick; FOV = 256 x 256 mm; 192 x 256 matrix; 1 mm in-plane resolution). Initial image preprocessing was performed using FSL (FMRIB Software Library version 4; (Smith et al., 2004). Following motion correction of the T2-weighted functional images, probabilistic independent component analysis was conducted on a run-by-run basis to identify and remove high-amplitude time course spikes as well as residual motion artifacts, high-frequency scanner noise and artifacts related to gradient timing errors. This step was performed using MELODIC (Beckmann & Smith, 2004) and detailed inspection of independent components by two raters. Timing differences between slices in the same volume were corrected using SPM software (Statistical Parametric Mapping version 5). Functional data were then transformed into MNI space (Cocosco, Kollokian, Kwan, & Evans, 1997), resampled into isotropic voxels (3 x 3 x 3 mm), and smoothed using a 3D Gaussian kernel with a full-width at half maximum value of 6 mm. In all analyses, intensity units were converted to a percent signal change score based upon the intensity of each image relative to a reference scan. In my blocked analysis, this scan was taken following a 16 s interblock interval and in the same TR in which

48 35 block instructions were presented. In my event-related analysis, I examined the percent signal change in each event from 2 to 10 s following each stimulus onset relative to onset of a reference scan, as taken 2 s after trial onset, to allow for a haemodynamic return to baseline from the preceding trial. I determined that this window was optimal for response detection on the basis of haemodynamic response function modeling as well as inspection of global intensity data. All functional neuroimaging analyses were conducted using non-parametric resampling statistics (non-rotated partial least squares in PLSGUI; (McIntosh & Lobaugh, 2004). Unlike rotated PLS, non-rotated PLS is a hypothesis-driven approach. It is suitable for testing network-level effects, in which case permutation testing is used to evaluate network stability, or for assessment of effects that are local in nature, in which case a bootstrap resampling procedure is used to evaluate the stability of signal differences in individual voxels. For each analysis, I created a singular profile containing a contrast matrix and a singular image describing the relationship of all voxels to the singular profile. In cases where a test of the stability of the singular profile was required, permutation testing was performed using 500 samples. In cases where local differences in signal were of focal interest, bootstrap resampling was conducted using 100 samples. Maps were created expressing the ratio of voxel salience over the estimated standard error (i.e., bootstrap ratio; BSR) for the purpose of identifying statistically reliable relationships between individual brain voxels and the singular profile. To characterize voxel responses in terms of a specific spatial distribution, I inspected BSR maps from the peak haemodynamic response at 4-6 s following stimulus onset for clusters of reliably differentiated voxels, defined as any set of at least 12 contiguous cortical or subcortical voxels above a BSR of 2.81 and a peak of 3.5 (approximately corresponding to a minimum spatial extent of 324 mm3, a % peak confidence interval and a 99% extent confidence interval) that was no closer than 12 mm to another cluster. This corresponds to a more conservative threshold than is often used in full-brain neuroimaging analyses, which are typically thresholded at an uncorrected P < and frequently include no restriction of minimum spatial extent. Labels for identified clusters were obtained by transforming peak MNI coordinates into Talairach coordinates using a best-fit icbm2tal transform (Lancaster et al., 2007) and localizing these coordinates in a Talairach brain atlas (Mai, Paxinos, & Assheuer, 2004).

49 36 All fmri analyses in my study pertained to fmri data collected during phase 2 (encoding), as no neuroimaging data was collected during any other phase. As a preliminary manipulation check, I attempted to replicate previous findings of novelty effects in the brain by contrasting neural activity associated with blocks of novel and repeated scenes. To evaluate whether novel and repeated scenes are encoded in similar or different ways, I assessed the interaction between novelty and subsequent memory success. Phase 2 encoding trials were classified as successful or unsuccessful for each participant based on source judgments in the phase 3 memory test. Source hits were action items identified as such, whereas miss items were action items identified as intention items or lures. For descriptive purposes, memory for the categorized scenes was also entered into two event-related contrasts exploring subsequent memory effects for novel and repeated stimuli. To identify regions that were associated with subsequent memory regardless of stimulus type, I ran a conjunction analysis. First, to create a mask limiting my analysis to voxels in which group subsequent memory effects were found for both novel and familiar items, each the two BSR maps associated with novel and repeated stimuli were thresholded at 1.96 (approximately corresponding to a 95% confidence interval) with supra-threshold voxels set to a value of one and sub-threshold voxels set to a value of zero. The product was taken of the two maps, yielding a binary mask occluding voxels failing to surpass threshold in both group contrasts (Friston, Penny, & Glaser, 2005). Next, using bootstrap resampling, I evaluated whether voxels located within the mask were reliably associated with a positive scaled singular-image product (i.e., were associated with positive subsequent memory effects in both novelty and familiarity conditions at the within-subjects level). First, the singular value-scaled singular image was taken from the contrast of hits and misses for both the novel and familiar conditions in each subject. The product of these scaled images was calculated and averaged across subjects to establish a group mean for each conjunction voxel. This procedure was repeated 100 times with bootstrap resampling to establish a 95% confidence interval about the mean. All voxels that included zero within this interval were exclusively masked. The surviving voxels were entered into a cluster analysis of the kind described for other group analyses above.

50 37 Finally, I compared the overall functional connectivity of anterior and posterior aspects of the hippocampus. Seeds for this analysis were selected from the peak anterior and posterior hippocampal voxels in the interaction analysis. To obtain within-subject voxelwise correlation values, a spatiotemporal stack of all events described in terms of percent signal change was assembled for each subject (as in above analyses) and used to generate a mean and standard deviation image for each time point of each event. Seed voxel standard deviation values were extracted and multiplied with their full standard deviation images to create a series of standard deviation product images for each event. Next, the difference was assessed between the mean event images and corresponding images for each event in the stack. Seed voxel difference values were extracted and multiplied with their full difference images to create a series of difference product images for each event. Finally, the difference product images were averaged at each timpoint and divided by the standard deviation product image to generate a withinsubject correlation image for each time point and subject. This procedure was conducted separately for the anterior and posterior hippocampal seeds. The difference between the resulting correlation images was compared using non-rotated contrasts (as above). The resulting overall connectivity map was inspected for clusters using the same method as in earlier analyses. In a follow-up analysis, I repeated my within-subject correlation computations, this time calculating functional connectivity separately for each of the novel hit, novel miss, repeated hit and repeated miss conditions. Separate maps for these conditions were created based on functional connectivity with each hippocampal seed. To avoid high Type I error rates associated with inspecting multiple whole-brain maps, I performed two tightly constrained analyses in place of further full-brain contrasts. First, I conducted a region-of-interest (ROI) analysis assessing whether hippocampal functional connectivity with the anterior temporal lobes or precuneus, regions of theoretical interest identified in the above overall connectivity analysis, predicted subsequent memory. These ROIs were delineated by the bounds of the original identified clusters. I searched for voxel sub-clusters falling within the ROIs of at least four voxels (64 mm3) with anterior or posterior hippocampal connectivity that reliably predicted subsequent memory for novel or familiar items (i.e., surpassed a threshold of BSR 1.96). Second, I performed permutation tests to evaluate whether large-scale

51 38 functional connectivity networks were present that distinguished between hits and misses, while also exploring for possibile subsequent memory interactions with novelty. 3.2 Results Behavioral results I first explored whether recognition memory performance was improved by repetition of scenes three times (phase 1) prior to the scanned memory encoding task in phase 2. In cognitive investigations on memory for novel and repeated materials, performance is typically assessed by asking participants to identify studied novel and repeated targets among novel and repeated lures. Similarly, in my memory test, I asked participants whether scenes were part of one of the scanning tasks or were not seen during scanning. Because intention data were not analyzed, recognition memory hits were specifically those novel or repeated phase 2 action items that participants correctly identified as being part of any scanning task (i.e., either as an action or intention item). Recognition misses were those phase 2 action items that participants incorrectly rejected as lures. Novel lures were items that were not presented at all prior to testing in phase 3, whereas repeated lures were items presented three times during the repetition phase (phase 1) but not at all during phase 2. False alarms were lures that participants identified as either action or intention items, whereas correct rejections were lures that were successfully identified as lures. Participants recognized more of the repeated target scenes as having been presented during scanning than target scenes that were novel as of the study phase, two-tailed t(12) = 4.44, P < (Fig. 4; Table 5). However, participants also made more false alarms to repeated lures than novel ones, two-tailed t(12) = 3.80, P < As a result, a trend towards higher overall recognition accuracy for novel items was observed, two-tailed t(12) = 1.84, P < 0.1. This pattern of high false alarms reducing accuracy in the repeated condition is typical of cognitive investigations of novelty (Tulving & Kroll, 1995). As discussed in earlier chapters, better discrimination is required to reject repeated lures than novel ones in this type of design (Dobbins et al., 1998). In the current experiment, participants would have needed to decide whether familiar lures were seen during fmri scanning or the

52 39 Table 5. Descriptive statistics for hits, false alarms (FA), corrected recognition accuracy (acc) and source memory accuracy (src) for pre-study repetition items and novel items in the Experiment 3 memory test. Mean Standard deviation Standard error Hits FA Acc Src Hits FA Acc Src Hits FA Acc Src Repeated before study Novel

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

54 41 pre-scanning repetition phase; in contrast, they would have been able to reject novel lures by the absence of recognition alone, since they were never previously encountered. As in Experiment 2, I circumvented this confound by directly evaluating memory for source (in addition to recognition memory). During the memory test, participants indicated whether the scene was presented in the scanner (recognition memory) and if so, which of two tasks encountered during scanning was associated with each scene (source memory). Source memory hits were action items specifically identified as such, whereas source misses were action items that were not identified correctly. Whereas the recognition memory decision required participants to discriminate using information acquired both before and during scanning, the source memory decision required them to discriminate exclusively on the basis of information acquired during scanning. When comparing source memory for the scenes that were novel or that had been repeated three times prior to the study phase, I observed an advantage of repetition, in contrast to the novelty advantage obtained using the confounded recognition memory measure. Participants correctly identified the source of repeated scenes more frequently than they correctly identified the source of novel ones, twotailed t(12) = 4.28, P < (Fig. 4, Table 5). When source memory accuracy was assessed together with recognition memory accuracy in a repeated-measures analysis of variance, an interaction was observed between the measures, with better recognition memory accuracy for novel scenes and better source memory accuracy for repeated ones, F(1,12)=16.97, P < (Fig. 4, Table 5). All behavioral effects were the same whether or not excluded participants were included in the analyses Functional neuroimaging results While my behavioral evidence favours the position that past experience in the form of repetitions, rather than novelty, benefits memory once confounding factors are removed, neuroimaging has consistently revealed greater medial temporal lobe (MTL) responses to novel relative to repeated materials (Tulving et al., 1996; Kirchhoff et al., 2000; Kumaran & Maguire, 2006; Poppenk et al., 2008). Because the MTL is known to be important for memory in general (Scoville & Milner, 1957), and because novelty and subsequent memory effects have been observed in the same hippocampal region

55 42 (Kirchhoff et al., 2000), many researchers have linked MTL novelty responses to memory encoding. Along these lines, various MTL memory-encoding mechanisms have been proposed that require novelty for efficacious encoding (Tulving et al., 1996; Borisyuk et al., 2001; Lisman & Grace, 2005). Accordingly, MTL responses are of particular interest here. To confirm that my novelty manipulation exerted comparable effects on brain activity to those observed in previous investigations, I first contrasted the fmri signal associated with novel and repeated scenes (see Table 6). I identified more activity in the right anterior hippocampus, fusiform gyrus, right visual cortex and left prefrontal cortex (PFC) in response to novel scenes, an effect often seen in neural priming studies comparing brain responses to novel and repeated stimuli (Schacter, Wig, & Stevens, 2007). In contrast, I found more activity in the right posterior hippocampus, motor cortex and bilateral PFC in response to repeated relative to novel scenes. Novelty/repetition dissociations in anterior/posterior hippocampus and left/right PFC have been observed in numerous neuroimaging investigations of novelty (Lepage et al., 1998; Habib, Nyberg, & Tulving, 2003) and were interpreted often as encoding/retrieval dissociations. A new departure in my fmri analysis was to assess how subsequent memory predictors in the brain interact with novelty. I first back-sorted novel and repeated scenes that appeared in the scanned encoding task using source memory data from the memory test (Brewer et al., 1998; Wagner et al., 1998). It is typical for recognition memory-based sorting to be employed in such analyses, but just as my behavioral evidence indicates that source confusion prevents a fair comparison of recognition memory for novel and repeated items, this same evidence illustrates that misses do not have a consistent meaning across novel and familiar items in an fmri study. Interactions between novelty and recognition memory-based subsequent memory observed using brain measures would be influenced by the same source confusion confound identified in my behavioral data. Whereas novel hits and misses would differ in terms of whether novel items were encoded, familiar hits and misses would differ in terms of whether source information was encoded. As a result of this issue, and because source information is an important attribute of episodic memory (Tulving, 1983), my source memory-based approach is fully consistent with the goal of my current

56 43 Table 6. Activated voxel clusters in the block-level contrast of novel and familiar scenes in Experiment 3. Region BA Hemi. Peak MNI coordinates X Y Z Peak BSR Spatial extent (mm³) Novel > Familiar Frontal lobe Rostral prefrontal ctx. 10 L Temporal lobe Anterior hippocampus - R Inferior temporal g. 19/37 R Occipital lobe Striate area 17 R Occipital g. 18 R Familiar > Novel Frontal lobe Lateral prefrontal ctx. 46 R Dorsolateral prefrontal ctx. 9 L Precentral g. 6 R Temporal lobe Posterior hippocampus - R Note: Regions were defined as any cluster with a minimum peak BSR of 3.5 (P < ) and an extent of at least 12 voxels (extent threshold was BSR 2.81 or P < 0.005). Peak coordinates are displayed in MNI space (Cocoso et al., 1997).

57 44 investigation: to measure the neural underpinnings of successful episodic memory formation associated with novel versus familiar information. I first conducted a conjunction analysis to identify regions that predicted memory independently of my novelty manipulation (Fig. 5 and Table 7). The regions matching this profile were primarily located in posterior neocortex and included bilateral occipital gyri, the right parietoccipital transition zone and bilateral parietal cortex. However, no MTL region predicted memory for both novel and repeated scenes. I next computed an interaction contrast and searched for MTL regions in which novelty interacted with subsequent memory. I identified a greater subsequent memory effect for novel scenes in the right anterior hippocampus and right amygdala, and a greater subsequent memory effect for familiar scenes in bilateral posterior hippocampal regions (Fig. 6, Table 8). An ROI analysis of the subsequent memory effects in these regions revealed that the anterior hippocampus peak predicted memory for novel scenes (BSR = 2.33, P < 0.05), but did not predict memory for repeated ones (BSR = 1.11, P > 0.25). In contrast, while memory for repeated scenes was predicted by both the left posterior hippocampus (BSR = 3.70, P < 0.001) and right posterior hippocampus (BSR = 4.02, P < 0.001), memory for novel scenes was not predicted by either region (left BSR = 0.76, P > 0.4; right BSR = 1.365, P > 0.15). A post-hoc test for a three-way interaction of region, novelty and subsequent memory revealed an effect between the right anterior and left posterior hippocampus, F(1,12) = 8.07, P < 0.05, and a weak effect between the right anterior and right posterior hippocampus, F(1,12) = 4.48, P = 0.056, but no effect between the left and right posterior hippocampus, F(1,12) = 0.37, P > No additional region predicted memory in the MTL. For descriptive purposes, full-brain subsequent memory effects for novel and repeated scenes are provided in Tables 9 and 10. Different connectivity of the anterior and posterior hippocampus could underlie this hippocampal double dissociation: the anterior hippocampus has been shown to have higher functional connectivity with the anterior and lateral temporal lobes, whereas the body and posterior aspects of the hippocampus has been shown to have higher functional connectivity with parietal cortex, posterior cingulate cortex and ventral PFC (Kahn, Andrews-Hanna, Vincent, Snyder, & Buckner, 2008). To confirm whether this same distinction in functional connectivity was observable in my data and was related

58 45 Table 7. Voxel clusters that predicted subsequent memory success for both novel and familiar scenes in Experiment 3 (conjunction analysis). Region BA Hemi. Peak MNI coordinates 95% CI of scaled SI product x 10 2 X Y Z LL M UL Spatial extent (mm³) Temporal lobe Inferior temporal g. 37 L Parietal lobe Superior parietal l. 7 R L Parietoccipital transition zone 19 R Occipital lobe Fusiform g. 19 R /37 R Occipital g. 18/19 L L R Sub-lobar Pons - L/R Note: Regions were defined as any cluster of 12 or more voxels surviving a threshold of BSR 1.96 (P < 0.05) in both of the two subsequent memory contrasts. Voxels were also required to survive bootstrap testing with 95% confidence prior to inclusion in a region. Coordinates are displayed in MNI space (Cocoso et al., 1997).

59 46 Table 8. Activated clusters in the Experiment 3 interaction contrast crossing subsequent memory with novelty. Peak MNI coordinates Region BA Side X Y Z Novel subsequent memory > Familiar subsequent memory Peak BSR Spatial extent (mm³) Frontal lobe Inferior frontal g. 11/47 R Dorsolateral prefrontal ctx. 9 L Precentral g. 6 L Temporal lobe Inferior temporal g. 20 R Parahippocampal g. 28/36 R Anterior hippocampus - R Inferior temporal g. 37 L Middle temporal g. 39 L Parietal lobe Parietoccipital transition zone 19 L Occipital lobe Occipital g. 19 L Familiar subsequent memory > Novel subsequent memory Parietal lobe Postcentral g. 1 R Temporal lobe Temporal pole 38 L Posterior hippocampus - L R Limbic lobe Cingulate g. 24/32 R Note: Regions were defined as any cluster with a minimum peak BSR of 3.5 (P < ) and an extent of at least 12 voxels (extent threshold was BSR 2.81 or P < 0.005). Peak coordinates are displayed in MNI space (Cocoso et al., 1997).

60 47 Table 9. Activated voxel clusters in the Experiment 3 event-related contrast of remembered and forgotten scenes (novel items only). Region BA Hem Peak MNI coordinates i. X Y Z Peak BSR Spatial extent (mm³) Novel remembered > Novel forgotten Frontal lobe Inferior frontal g. 47 L R Dorsolateral prefrontal ctx. 9/4 6 L Premotor ctx. 6/8 L L R Motor ctx. 6 L Temporal lobe Amygdala - R Anterior hippocampus - R Inferior temporal g. 37 L Medial temporal g. 19 L Parietal lobe Postcentral g. 3 L Precuneus 7 R L R Supramarginal g. 40 L Superior parietal l. 7 R Limbic lobe Posterior cingulate g. 23 L Table 9 cont d on next page

61 48 Table 9. (cont d from previous page) Region BA Hem Peak MNI coordinates i. X Y Z Peak BSR Spatial extent (mm³) Novel forgotten > Novel remembered Frontal lobe Rostral prefrontal ctx. 10 R Inferior frontal g. 47 L Precentral g. 6 L Temporal lobe Middle temporal g. 21 R Precuneus 7 R Limbic lobe Anterior cingulate g. 32 R Note: Regions defined as in Table 8.

62 49 Table 10. Activated voxel clusters in the Experiment 3 event-related contrast of remembered and forgotten scenes (familiar items only). Region BA Hemi. Peak MNI coordinates X Y Z Peak BSR Spatial extent (mm³) Familiar remembered > Familiar forgotten Frontal lobe Dorsolateral prefrontal ctx. 9 R Motor ctx. 6 L Temporal lobe Temporal pole 38 L Posterior hippocampus - L R Parietal lobe Supramarginal g. 40 R Limbic lobe Cingulate g. 32 R L Sub-lobar Lateral globus pallidus - L Thalamus - R Familiar forgotten > Familiar remembered Temporal lobe Entorhinal ctx. - R Note: Regions defined as in Table 8.

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

64 JG!"#$%&'( #$ #% E<3:9:<5 D48.9 $"#5 DQW KLSSRFDPSXV VLJQDO KLWV ± PLVVHV!"# *+,-./ :3+9: =2$ & *+,-./ ;<3:9:<52-1:3+9: AB%( AB%& ABA) ABAC ABA# ABAA 7ABA# 7ABAC VLJQDO KLWV ± PLVVHV SRVW KLSSRFDPSXV &' &( AB%A ABA' ABAC ABA$ ABA& ABAA 7ABA& 7ABA$ VLJQDO KLWV ±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

65 52 to the location of the hippocampal subsequent memory effects, I conducted a functional connectivity analysis seeded from anterior and posterior hippocampal subsequent memory regions. In this analysis, I calculated the within-subject correlation between activity in each seed and activity in the rest of the brain, then compared the correlations of the two seeds using a within-subject contrast. The pattern I observed was consistent with earlier findings: the anterior hippocampal novelty-encoding region preferentially correlated with anterior and lateral temporal lobe regions, as well as regions along the central sulcus, whereas the posterior hippocampal encoding regions correlated with inferior parietal cortex, visual cortex, posterior cingulate cortex as well as ventral and dorsolateral PFC (see Fig. 7 and Table 11). To determine how these differences in functional connectivity might contribute to memory encoding, I performed several ROI analyses, searching the precuneus and anterior temporal regions identified in the overall connectivity analysis described above for any small cluster of four or more voxels (64 mm 3 ) in which greater anterior or posterior hippocampal connectivity predicted subsequent memory success for novel or familiar items (at a threshold of BSR 1.96, approximately corresponding to a 95% confidence interval). Greater connectivity between the anterior hippocampal seed and a cluster in the right anterior temporal lobe predicted successful memory for novel scenes but not repeated ones, whereas greater connectivity between the posterior hippocampus and voxels in the left precuneus predicted successful memory for repeated scenes but not novel ones. A different approach to evaluating the significance of the observed subsequent memory differences is to test the ability of large-scale functional connectivity networks to predict subsequent memory as a whole. To this end, I performed a series of post-hoc nonrotated PLS contrasts, focusing on whether reliable differences in brain scores could be obtained for each contrast. I found no evidence of a shared anterior and posterior functional connectivity predicting subsequent memory, P > 0.5, nor did anterior hippocampal functional connectivity predict subsequent memory on its own, P > 0.5. Functional connectivity with the posterior hippocampus did predict memory, P < In addition, a functional connectivity interaction between subsequent memory and novelty was found in the posterior hippocampus, P < 0.05, although no corresponding interaction was observed with anterior hippocampus functional connectivity, P > 0.5.

66 53 Table 11. Functional connectivity of the anterior hippocampal novelty-encoding region versus posterior hippocampal familiarity-encoding region (Experiment 3). Region BA Hemi. Peak MNI coordinates X Y Z BSR product Spatial extent (mm³) Anterior > Posterior Frontal lobe Superior frontal g. 6 R Precentral g. 4 R /6 L Temporal lobe Temporal pole 38 R Medial temporal g. 21 R Superior temporal g. 22 L Anterior hippocampus - L Perirhinal ctx. 35/36 Medial temporal g. 21 R Medial temporal g. 21/22 L Superior temporal g. Parietal lobe Postcentral g. 3 R /2 L Sub-lobar Cingulate g. 32 R Table 11 cont'd on next page

67 54 Table 11. (cont'd from previous page) Region BA Hemi. Peak MNI coordinates X Y Z BSR product Spatial extent (mm³) Posterior > Anterior Frontal lobe Frontal pole 10 L Ventral PFC 11 R L Dorsolateral PFC 8/9 L R Temporal lobe Parahippocampal g. 36 R Posterior hippocampus - Thalamus - Parietal lobe Inferior parietal l. 39 L Superior parietal l. 7 R Precuneus Precuneus 7 L/R Parietoccipital transition zone 19 L Occipital lobe Occipital g. 17/18/19 L/R L Sub-lobar Posterior cingulate g. 23/31 L/R L Note: Seeds were the peak anterior and posterior hippocampal voxels from the interaction analysis. Regions defined as in Table 8, omitting autocorrelations.

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

69 56 Returning to my signal intensity data, as a final step, I searched the interaction contrast map for stand-alone regions predicting memory outside of the MTL. At the cortical level, a majority of regions were found to have greater subsequent memory effects for novel scenes than for repeated ones (see Fig. 8 and Table 7). Regions that preferentially predicted memory for novel scenes included bilateral prefrontal and temporal regions as well as left posterior parietal cortex and the left occipital lobe. Regions that preferentially predicted memory for repeated scenes included right postcentral gyrus, the left temporal pole and the right cingulate gyrus. 3.3 Discussion Like Experiment 2, Experiment 3 supports the view that recent past experiences in the form of repeated stimulus exposures can enhance the formation of new episodic memories. While a novelty effect was replicated in recognition memory, as discussed in earlier chapters, this measure was likely confounded by source confusion (Kinsbourne & George, 1974; Dobbins et al., 1998). Source memory, which was not affected by the confound, was superior for repeated over novel items, as it was in Experiment 2. In addition, analyses of fmri data revealed sub-regions of the hippocampus that were specialized for novelty and familiarity encoding. Both sets of findings are clearly incongruent with the hypothesis that incoming information is only encoded by the hippocampus to the extent it is novel (Tulving et al., 1996). With respect to my neuroimaging evidence, some investigators have proposed that neural mechanisms filter out redundant information before it reaches encoding structures, allowing only novel information to pass (Tulving et al., 1996; Borisyuk et al., 2001). Whereas such a process would entail a core encoding network that responds to information in a graded fashion to the extent the information is novel, I observed largescale interactions in the networks predicting encoding of novel and repeated scenes, involving constellations of MTL, temporal and frontal regions. Critically, in my exploration of encoding interactions, I noted a double dissociation of subsequent memory predictors for novel and repeated materials along the longitudinal axis of the hippocampus. While previous studies have implicated the hippocampus in both

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

71 58 novelty detection and subsequent memory for novel items, no studies have tested for an interaction of the type I observed between novelty, hippocampal region and subsequent memory effects. Memory was superior for repeated scenes once discrimination demands were equated. When I measured source memory, thus requiring the same information for responses in the novel and repeated conditions, memory was superior for repeated scenes relative to novel ones. This pattern replicates the findings of Experiment 2 using pictorial rather than verbal stimulus materials. As for recognition memory, participants were more likely to recognize repeated scenes, but were also poorer at rejecting repeated lures presented during the encoding phase than novel lures, and recognition accuracy scores were superior for novel scenes as a result. This same pattern has been observed numerous times and forms the basis for theoretical claims regarding superior memory for novel items (Tulving and Kroll, 1995; Aberg & Nilsson, 2001). This pattern was also observed in Experiment 1, and was found be the result of source confusion. As discussed earlier, recognition memory for novel and repeated items involves different memory constructs that are not appropriate to compare directly. This assertion was reinforced in the current experiment by the divergence between recognition and source forms of memory: a formal interaction was observed between novelty type and memory measure. At the brain level, an overall contrast of novel and repeated scenes revealed regions similar to those reported in other neuroimaging investigations of repetition and novelty (Schacter et al., 2007), serving as a manipulation check. Of particular note, I observed a greater response to novel scenes in the right anterior hippocampus and a greater response to repeated scenes in posterior hippocampi. This general pattern has been observed in a meta-analysis of similar studies (Lepage et al., 1998). The unique approach taken in the current experiment was to search for interactions in encoding mechanisms for novel and repeated materials. Because of the concerns raised above regarding the comparability of recognition memory measures in novelty and familiarity conditions, I conducted all analyses of this kind using source memory success. As discussed in Chapter 1, researchers have previously identified noveltysensitive hippocampal regions that also predict memory for novel materials (Kirchhoff

72 59 et al., 2000). Similarly, in the current study, I found that the right anterior hippocampus was both sensitive to novelty and predictive of subsequent source memory for novel scenes. However, activity in the region did not distinguish hits from misses in the repeated condition, suggesting that the right anterior hippocampus contributed to memory encoding only when materials were novel. I also observed posterior hippocampal regions that showed the opposite pattern of activity, responding more to repeated scenes than novel ones, and predicting subsequent memory for repeated scenes only. This new evidence suggests that while some hippocampal regions do appear to be specialized in processing novel materials, this sensitivity is not a general feature of the MTL encoding system, which also includes other hippocampal regions specialized in working with information that is familiar. Concerning the specific topography of the observed hippocampal responses, my evidence appears to complement an emerging understanding of functional organization along the long axis of the hippocampus. Numerous studies have linked the anterior hippocampus with acquisition and assimilation of new memories, and the posterior hippocampus with the retrieval of old ones (Lepage et al., 1998; Maguire, Woollett, & Spiers, 2006; Berlingeri et al., 2008), with remote memories engaging more posterior aspects than relatively recent ones (Gilboa, Winocur, Grady, Hevenor, & Moscovitch, 2004). In an apparent contradiction of this pattern, encoding-specific activation has been shown to occur along the entire hippocampal axis (Schacter & Wagner, 1999), much as activity in both anterior and posterior hippocampus predicted encoding success under different conditions in the current study. However, the encoding of repeated materials likely involves an element of retrieval resulting from previous exposures to the materials. Along these lines, the current results suggest that encoding operations that involve a measure of episodic retrieval fall more posteriorly in the hippocampus than those that do not. Some overlap in the networks predicting successful memory formation for novel and repeated scenes was observed. This overlap occurred primarily in posterior neocortex and included bilateral occipital gyri, the right parietoccipital transition zone and bilateral parietal cortex. These regions have been associated with the integration of visual and motor information in the dorsal visual pathway (Goodale & Westwood, 2004) and their contribution to memory could be related to my requirement that

73 60 participants imagine actions in visually presented scenes during encoding. Interestingly, many cortical subsequent memory responses were attenuated for repeated scenes relative to novel ones, in spite of my finding that source memory is superior for repeated items. However, greater brain responses do not always predict better memory (Daselaar, Prince, & Cabeza, 2004; Miller et al., 2008). One possible explanation for the link between signal attenuation and better memory for repeated scenes is that repetition increased cortical processing efficiency and subsequent memory by reducing the requirement for a basic parsing of the scenes. Functional connectivity differences between anterior and posterior hippocampus could help explain this difference in organization. I found that activity in the posterior hippocampal encoding regions was more strongly correlated with inferior parietal cortex, precuneus, ventral PFC and dorsolateral PFC than was activity in the anterior region. In contrast, anterior hippocampus activity primarily correlated with anterior and lateral temporal lobe. These connectivity findings confirm earlier observations regarding intrinsic functional connectivity differences between anterior and posterior hippocampus (Kahn et al., 2008) and directly link these differences to the hippocampal memory-encoding regions observed in the current study. My findings also hint at a possible functional role for these local variations in functional connectivity, since posterior hippocampal connectivity with the precuneus predicted memory for repeated, but not novel, scenes, whereas anterior hippocampal connectivity with the right anterior temporal lobe predicted memory for novel, but not repeated scenes. It is interesting to speculate about the possible cause for this dissociation. For instance, because the precuneus is part of a network that has been linked with autobiographical memory retrieval (Svoboda, McKinnon, & Levine, 2006; Buckner, Andrews-Hanna, & Schacter, 2008), its coactivation with posterior hippocampus may have predicted memory for familiar items because coactivation reflects a successful attempt to retrieve recently-encoded stimulus information from memory, which in turn facilitates the formation of a new memory. Because anterior and lateral temporal lobes are better known for their central role in semantic processing (Vandenberghe, Nobre, & Price, 2002; Rogers et al., 2006), their coactivation with the anterior hippocampus may have predicted memory for novel items because coactivation reflects successful initial interpretation of novel materials, drawing upon knowledge about the world.

74 61 It is interesting to consider what differences in novelty and familiarity processing led to the observed familiarity memory advantage and double dissociation of hippocampal subsequent memory responses. The current study was not ideally suited for answering such a question, as my main objective was to determine whether there exist any novelty-based differences in the brain basis of subsequent memory. Laboratory testing with a group of fourteen undergraduate students revealed no hint of novelty-based differences in the vividness of mental imagery generated during my encoding task, but other candidate mechanisms remain, such as possible differences in levels of item memory associated with novel and familiar items. That said, the current findings are not undermined by want of a deeper, underlying cause. Irrespective of the underlying cause of the novelty-based dissociations observed in the current study, the reported phenomena are of intrinsic interest and pose a challenge for existing notions of the role of novelty in memory formation. Efforts directed at exploring underlying causes will be a worthwhile goal for future research. To summarize, I replicated the behavioural finding in Experiments 1 and 2 of superior memory for repeated, relative to novel, scenes when controlling for confounds concerning discrimination demands. At the brain level, this memory benefit was associated with unique predictors of memory formation for repeated scenes. In the MTL, right anterior hippocampus predicted memory for novel scenes only, whereas posterior hippocampi predicted memory for repeated scenes only. Importantly, I found that the anterior region was functionally connected with the lateral and anterior temporal lobes, whereas the posterior regions were functionally connected with regions comprising the default network. As these differences in connectivity were linked with novelty and familiarity-specific encoding effects, they could help explain the location of novelty- and repetition-sensitive regions of the hippocampus. I conclude that memory representations established though past exposures to materials are not employed to filter out information, as has been suggested in recent proposals (e.g., Tulving et al., 1996), but are instead used to support new encoding. This conclusion, like the one drawn in Chapter 2 using cognitive rather than fmri-based evidence, is again in line with influential views of memory encoding emphasizing the role of prior representations in memory formation (Ebbinghaus, 1913; Nadel & Moscovitch, 1997).

75 Chapter 4 Modulation of encoding by experimental and pre-experimental forms of familiarity In comparing the neural encoding correlates of novel and familiar materials, the fmri investigation described in Chapter 3 helped to develop an understanding of the relationship between the cognitive familiarity effects observed in Chapter 2 and novelty/subsequent memory activations that are often observed in the hippocampus. Based on these results, it appears that at the level of the MTL, different regions support episodic memory encoding with respect to novel versus familiar materials. Specifically, there appear to be novelty-sensitive regions in the anterior hippocampus that are both more responsive to novel materials and more implicated in encoding episodes involving those materials. There also appear to be familiarity-sensitive regions in the posterior hippocampus that are both more responsive to familiar materials and more implicated in encoding episodes involving those materials. This pattern accommodates past observations of novelty and familiarity activations in the hippocampus, and also accommodates the current cognitive evidence showing that episodes involving familiar materials are better encoded into memory. An important question remains: does familiarity-based modulation of encoding activity in the brain result from phenomena specific to the recent repetition of stimulus materials, or is encoding modulated by interaction with long-term memory systems more generally? While memory advantages of familiarity over novelty were comparable in size in Experiments 1 and 2, this does not necessarily indicate that preexperimental and experimentally-induced forms of familiarity exert comparable effects at the brain level. Although I will not attempt equate these forms of familiarity with semantic and episodic memory systems directly, to the extent that they differentially recruit these systems, divergent influences could be expected on the basis of known differences in the neural correlates of episodic and semantic memory. For example, neuropsychological patients with damaged hippocampi perform poorly on tasks that 62

76 63 require episodic memory, even though their world knowledge is preserved (Moscovitch et al., 2005). In contrast, the degeneration of the anterior temporal lobes in certain dementias is associated with semantic memory loss and relatively preserved episodic memory function (Patterson, Nestor, & Rogers, 2007). Along these lines, hippocampallybased retrieval of recently encoded episodes (prompted by repeated materials) may influence local encoding within the structure (see, e.g., Hasselmo, Bodelon, & Wyble, 2002, for one such interaction model), whereas semantic memory retrieval, which does not require the hippocampus and is more closely associated with the neocortex, may not. On the other hand, because episodic memory was influenced by both prior knowledge and pre-experimental forms of familiarity in Experiments 1 and 2, and because episodic memory formation is thought to reflect processes in the hippocampus, a reasonable alternative hypothesis is that hippocampal encoding activity is influenced in both cases. Accordingly, the aim of the current experiment is to determine whether familiarity exerts similar or different encoding interactions depending on the type of familiarity in question. To explore this question, I designed an fmri protocol based on the design used in Experiment 2 of Chapter 2. Unlike in the original protocol, fmri data were collected during the familiarization and study phases in order to characterize the brain response to novel Asian proverbs, experimentally familiarized Asian proverbs and English proverbs known from prior knowledge. In a subsequent recognition memory test, participants viewed all of the stimuli from the study phase and attempt to determine which study task was associated with each item. Data from this memory test was used to categorize each proverb as remembered or forgotten for the purposes of subsequent memory analysis (Brewer et al., 1998; Wagner et al., 1998). I then compared the subsequent memory correlates of each type of proverb. Based on the subsequent memory interaction observed in the hippocampus in that same chapter, I predicted that the anterior hippocampus would predict memory for the novel Asian proverbs, whereas the posterior hippocampus would predict memory for the familiarized Asian proverbs. Because semantic memory is closely linked to the anterior temporal lobes, I predicted that these regions would predict memory for English proverbs along with hippocampal activity to support encoding of new information into episodic memory. However, I did not have strong predictions about whether this

77 64 activity would fall in the anterior or posterior hippocampus, since the prior knowledge items were in one sense novel (in the experimental context) and in another sense familiar (invoking memory retrieval). 4.1 Methods The procedure consisted of three main phases (Table 12): 1) repetition of Asian proverbs; 2) an incidental encoding phase in which participants were scanned using fmri while viewing proverbs that were novel, familiar due to prior knowledge, or familiar due to repetition in Phase 1; and 3) a post-scanning memory test phase, in which participants decided which of two tasks was associated with each proverb in Phase 2. Phase 1 allowed us to familiarize participants with items prior to the encoding phase. Data from the Phase 3 memory test were used to back-sort encoding events in the scanned Phase 2 encoding task as either successfully or unsuccessfully encoded items (Brewer et al., 1998; Wagner et al., 1998). During a separate follow-up session, participants indicated whether they had known each Asian and English proverb prior to the experiment Participants Nineteen right-handed volunteers from the Greater Toronto Area, all English nativespeakers with normal or corrected-to-normal vision and hearing, participated in the experiment (12 female; aged 21 to 34 years, mean age 26.2). All had less than one year of experience with East Asian languages or culture and had at least one parent or guardian who was also an English native-speaker. Participants were screened for the absence of neurological and psychiatric conditions and received financial remuneration for their participation. Of the 19 participants, one was excluded for excess head motion during scanning, one for chance-level performance on behavioral tasks, and a third due to incidental neurological findings. The protocol for this experiment was approved by the Research Ethics Board at Baycrest Hospital in Toronto.

78 65 Table 12. Schematic of experimental protocol and stimulus exposure. Stimuli consisted of two lists of 40 English proverbs (English 1 and English 2 ) and four lists of 20 Asian proverbs (Asian-Familiar 1, Asian-Familiar 2, Asian-Novel 1, and Asian-Novel 2 ). Phase and purpose Lists presented and task instructions Phase I. Three repetitions for familiarity induction Asian-Familiar 1 Asian-Familiar 2 multiple tasks Phase II. Incidental encoding of proverbs in two tasks English 1 Asian-Familiar 1 Asian-Novel 1 rate quality English 2 Asian-Familiar 2 Asian-Novel 2 rate target age Phase III. Test of memory for Phase II source information (all items) rated quality or target age? Follow-up. Identification of proverbs known prior to the experiment (all items) learned today, or know from prior knowledge?

79 Stimulus materials 160 of the Asian proverbs from Experiment 2 were employed. For each participant, this collection was randomly split into two sets of 80 proverbs: a repetition set and a novelty set. A separate list was compiled of 80 of the common English proverbs from Experiment 2. For use in a baseline task, a nonsense letter string was derived from each of the proverbs by substituting x s, o s, j s and g s in place of the original letters Procedure Participants were informed that the study consisted of two sessions: one involving two hours of fmri scanning and one hour of work at a computer, and a second, to be completed within one day of the first, involving an additional hour of computer tasks. Prior to the main experiment, demographic data were collected and experimental tasks practiced in a mock fmri scanner using a six-item set of proverbs (not included in analyses). In a familiarity-induction phase (Phase I), participants were familiarized with the repeated Asian proverb set in an fmri scanner (Table 12). Three exposures to the proverbs took place in the context of two tasks. Only the first task was scanned with fmri, and fmri data from this section were not evaluated as part of the current study. In the first task, participants were told they would see a series of proverbs, each with an underlying meaning. They were asked to indicate with an index-finger button press when they had imagined a plausible meaning for the presented proverb, and advised not to respond if no plausible meaning came to mind (i.e., a button press was not expected for every trial). The particular hand used for responding was counterbalanced across participants. Over the course of two scanned fmri runs, each of the 80 items from the repetition set of Asian proverbs was presented once for 7.5 s with an interstimulus interval of 4.0 to 10.5 s (average 5.9 s). Each of the two fmri runs took 9.4 minutes. Proverbs were displayed as centered black text in 18-point bold Courier New font over a white background. Throughout the task, instructions at the bottom of the screen in white text over a black banner (centered 12-point bold Courier New font) served as a reminder of which key to press if and when a plausible meaning was obtained for presented proverbs. Each proverb was then encountered a second and third time in a cultural origins task, which was not scanned with fmri; rather, a high-

80 67 resolution anatomical scan and diffusion tensor imaging scan were acquired during the task. The task was designed to encourage further elaborative processing of the repeated proverbs and maximize familiarity with them, while nonetheless remaining easily distinguishable from tasks presented later in the experiment. Participants were asked to decide whether each proverb was South American or Japanese (while in fact all proverbs were of East Asian origin, but not necessarily Japanese). Participants were told to make their best guess and to make a fresh judgment each time a proverb was presented, not allowing their memory of a past response to the same proverb to influence their judgment. During this task, proverbs were more rapidly presented at 3 s each with a 1 s inter-stimulus interval, and participants were allowed to make their response up until the time the next proverb was presented. Proverbs were presented in the same format as in the first task, and instructions linking each possible response to a particular key were presented at the bottom of the screen in the same format as in the first task. After all proverbs had been presented, a break was allowed before participants rated all of the proverbs a second time. The full task took 10.9 minutes. Following the task (at the end of Phase I) all 80 proverbs from the repeated Asian set had been presented three times. In Phase 2, participants were scanned using fmri over four functional runs while completing two incidental encoding tasks, labeled as target age and quality tasks (Table 9). In the target age task, participants decided whether proverbs were subjectively more appropriate to tell to an adult or a youth. In the quality task, participants decided whether proverbs were subjectively of high or low quality. The allocation of proverbs to the quality and target age tasks formed the basis of a later surprise source memory test (Table 12). A third low-level baseline task, not analyzed in the current study, required participants to decide whether a string of x s, o s and filler letters contained more x s or more o s. Each of the four runs contained two superordinate blocks for each of the three tasks. Each of the six blocks was preceded by a label designating the task to be performed (i.e., target age, quality or x s and o s ), which was presented for 4.5 s in 18-point bold Courier New font. Following presentation of the label, instructions at the bottom of the screen in white text over a solid banner (centered 12-point bold Courier New font) served as a reminder of task/response-key mappings. These instructions appeared on the screen 4.5 s prior to the first stimulus presentation in each block and remained until the end of the block. To

81 68 help distinguish the tasks and alert participants to shifts in task, all text in each task and the solid background color of the instruction bar was presented using a different color in each task (green, navy or maroon). To help detect task perseveration, participants were instructed to use a different hand for the quality and target age tasks. The specific color and hand assigned to each task was counterbalanced across participants. The index and middle finger of the assigned hand were randomly assigned to the two response options (high versus low quality; adult versus youth). The two response options in the baseline task were randomly assigned to contralateral index fingers. Within the target age and quality blocks, there was a set of five trials for each of the three proverb types (novel Asian, repeated Asian and English; for a total of fifteen trials per block). Within each baseline block, there were only five trials. In all blocks, a trial began with the 4.5 s presentation of a proverb randomly drawn from the appropriate list (or a nonsense letter string in the case of baseline blocks), in centered text (18-point bold Courier New font over a white background). This presentation was followed by an inter-stimulus interval of 2.0 to 7.5 s (average 4.8 s). In total, each of the four runs in Phase 2 consumed a total of 12.1 minutes in the scanner. Across all runs, 240 proverbs were presented 80 novel Asian, 80 repeated Asian and 80 English proverbs with each type split evenly among the quality and target age tasks. Following Phase 2 but before Phase 3, participants lay still in the scanner for ten minutes with eyes closed. During this period, fmri data measuring brain resting state activity were acquired. No visual or auditory stimulation was delivered. Prior to the scan, participants were asked not to fall asleep. Following the scan and upon leaving the fmri scanner, participants were asked whether they had in fact fallen asleep, whether they had been thinking about proverbs, and whether they had anticipated a memory test. In Phase 3, which took place in a quiet testing room, participants were seated at a computer and advised the next portion of the study involved a surprise memory test. All of the items originally presented in Phase 2 were tested, including the 80 novel Asian proverbs, 80 repeated Asian proverbs and 80 English proverbs (Table 12). For each proverb, participants were given as much time as they required to make two responses. First, an incomplete phrase containing the first few words from the proverb was centered on the screen in 18-point bold Courier New font, and participants were to

82 69 asked to either verbally recall the proverb based on the cue or indicate they did not know the proverb. An experimenter recorded the verbal response and pressed a button to reveal the full proverb. Participants then designated the proverb as either a target age or quality task item and rated their subjective confidence in each decision using a threepoint scale ranging from very sure to guess. An instruction bar of the type used in Phase 1 designated response key mappings. Because the verbal phrase-completion responses did not provide a fair comparison of episodic memory for the various stimulus classes (e.g., recall of English proverbs may be based on prior knowledge, rather than exposure during Phase 2), only source memory for proverbs and associated confidence measures were evaluated in the current study. A 1 s inter-stimulus interval followed each response. Proverbs were presented in a semi-random sequence, with proverbs of the same type (novel Asian, repeated Asian or English) grouped into proverb trial subsets of twenty. Each subset contained ten proverb trials from the target age task and ten proverb trials from the quality task. Following every three subsets of proverb trials, participants were given an opportunity for a break. All participants returned for testing within 24 hours of the first session to provide subjective reports concerning their knowledge of the 80 novel Asian, 80 repeated Asian and 80 English proverbs (Table 12). Items from the three lists were presented in random sequence. Participants were advised that all the items they would see had been presented in the study, and that the task, therefore, was not a memory test for the first session, but rather a subjective self-report of their general knowledge. They were asked to indicate with a button press whether they knew each proverb from prior knowledge or had seen it for the first time during the experiment. Participants were allowed as much time as they needed to make decisions. Proverbs were presented in the same font as in Phase 1 along with an instruction bar of the type used in Phase 1 designating key mappings. Each item was followed by a 1 s inter-stimulus interval Neuroimaging data acquisition and analysis All imaging was performed on a 3 Tesla whole-body MRI system (Siemens, Erlangen, Germany) using a 32-channel array head coil. T2-weighted echo planar image acquisition was used for all functional scans (TE = 30 ms; TR = 1500 ms; flip angle = 60 ). Each study run involved the acquisition of ten initial stabilization volumes that were discarded and 475 task volumes. 24 contiguous axial oblique slices were obtained

83 70 that captured the full neocortical brain volume of each participant, omitting the lower brain stem and inferior portion of the cerebellum (FOV = 200 x 200 mm; resolution 3 x 3 x 5 mm). Additional pre-experimental and resting state functional scans collected prior to and following the main session were acquired for unrelated investigations and are not reported here. A T1-weighted high-resolution MRI volume was obtained for the display of neuroanatomy during the same experimental session using a 3D MPRAGE pulse sequence in the same orientation as the functional scans (160 slices; FOV = 256 x 256 mm; resolution = 1 x 1 x 1 mm). Diffusion-weighted data were also collected in this orientation for analysis of anatomical connectivity (50 slices; FOV = 240 x 240 mm; resolution = 2.2 x 2.2 x 3 mm) using EPI acquisition (TE = 96 ms; TR = 9000 ms). Diffusion weighting was isotropically distributed along 30 diffusion directions (2 repetitions; b-value = 800 s/mm2). In addition, two images with no diffusion weighting (b-value = 0 s/mm2) were acquired prior to each series of 30 weighted images as an anatomical reference for motion correction. Initial preprocessing of the T2-weighted functional images was performed on a run-byrun basis using FSL (FMRIB Software Library version 4; Smith et al., 2004). Following correction for timing differences between slices in the same volume and motion correction within each image series, high-pass filtering was applied to exclude scanner drift and other low-frequency noise (sigma = 49.5 s). Each series was segmented to isolate the region containing the brain volume; voxels falling outside this region were exclusively masked. Each series was coregistered to the participant s high-resolution T1-weighted anatomical image after it underwent a similar masking procedure. Using the anatomical image as a reference image, both the anatomical and functional data were then transformed into standardized MNI space (Cocosco et al., 1997) and resampled into isotropic voxels (4 x 4 x 4 mm 3 ). Each normalized functional image series was entered into a probabilistic independent component analysis, which was conducted on a run-by-run basis to identify and remove residual motion artifacts, high-frequency scanner noise and artifacts related to gradient timing errors. This step was performed using MELODIC (Beckmann & Smith, 2004). Two raters conducted detailed inspection of components from one series of each participant. Evaluations from this inspection were used to train a component-sorting classification model, which was used to evaluate the remaining components in a semi-automated fashion. Manual spot checks of classifier output were made to verify sorting quality. Variance associated with noise

84 71 components was stripped from the data and high-amplitude spike volumes were dropped. Next, mean white matter intensity for each image volume was calculated using a mask composed of voxels with an 80 percent probability of being white matter (determined using an MNI standard space probability image) and this vector was residualized from the image series as a global intensity normalization step. Finally, all images were smoothed using a 3D Gaussian kernel with a full-width at half maximum value of 6 mm. Prior to statistical analysis, image series were segmented into single events with a temporal window size of 8 scans using stimulus onsets recorded during the experimental protocol. For each voxel and time point in each event, signal intensity was converted to a percent signal change value evaluated relative to the signal intensity of a reference scan collected immediately prior to stimulus onset. All events within each condition were merged for each participant to create a mean spatiotemporal image series for each condition and participant. All functional neuroimaging analyses were conducted using multivariate statistics (partial least squares in PLSGUI; McIntosh & Lobaugh, 2004), which is described in Chapter 2. Although contrast matrices were specified for most of these analyses (nonrotated PLS), it was not specified in analysis of resting state data (rotated PLS). For each analysis, a singular profile was generated containing a contrast matrix and a singular image describing the relationship of all voxels to the singular profile. In cases where a test of the stability of the singular profile was required, permutation testing was performed using 500 samples. In cases where local differences in signal were of focal interest, bootstrap resampling was conducted using 100 samples. BSR maps were created for the purpose of identifying statistically reliable relationships between individual brain voxels and the singular profile. To characterize voxel responses in terms of a specific spatiotemporal distribution, I inspected BSR maps from the time-course 0-12 s following stimulus onset for clusters of reliably differentiated voxels, defined as any set of at least 6 contiguous voxels at a particular time point with a BSR above 2.81 and a peak BSR of 3.5 (approximately corresponding to a minimum spatial extent of 384 mm3, a % peak confidence interval and a 99% extent confidence interval) that was no closer than 12 mm to another cluster. This corresponds to a more conservative threshold than is often used in full-

85 72 brain neuroimaging analyses, which are typically thresholded at an uncorrected P < and frequently include no restriction of minimum spatial extent. Labels for identified clusters were obtained by transforming peak MNI coordinates into Talairach coordinates using a best-fit icbm2tal transform (Lancaster et al., 2007) and localizing these coordinates in a Talairach brain atlas (Mai et al., 2004). I removed my extent requirement when searching for activations within the medial temporal lobes to allow for the possibility of focal effects in this a priori region of interest. First, I searched for regions that predicted encoding success in multiple conditions. To this end, I again compared the hits and misses for each condition and thresholded the resulting bootstrap time-course maps at BSR I then multiplied the maps for each pair of conditions, including novelty/repeated, novelty/prior knowledge and repeated/prior knowledge pairings. Next, I attempted to replicate my Chapter 3 findings of a novelty-based modulation of episodic memory encoding by comparing subsequent memory effects associated with novel and repeated proverbs. This was accomplished by 1) using source memory data from the Phase 3 memory test to classify each scanned encoding event in Phase 2 as a hit or miss ; 2) assessing hits minus misses to create difference time-course maps for both the novelty and repetition conditions; 3) performing a non-rotated contrast of the two difference time-course maps; and 4) limiting results to regions predicting memory success using a masking procedure. This latter step facilitated interpretation of the interaction contrast by allowing us to state that regions showing greater subsequent memory effects in one condition (i.e., an interaction) did in fact predict memory in that same condition. Without this step, positive subsequent memory interactions could have reflected forgetting effects in opposing condition, or weak effects in opposite conditions in the two conditions. To obtain memory success time-course masks, I compared hits and misses separately in each novelty condition and inclusively masked all memory success voxels surviving a liberal threshold (BSR 1.96, approximately corresponding to a 95% confidence interval). The novelty dm mask was applied to the novelty dm > repeated dm interaction contrast, whereas the familiarity dm mask was applied to the repeated dm > novelty dm interaction contrast. I repeated this overall procedure for detecting subsequent memory interactions between novel and prior knowledge proverbs, as well as between repeated and prior knowledge proverbs.

86 73 Because I wished to conduct functional connectivity analyses using the anterior and posterior hippocampus as seeds, it was necessary to obtain a time-series of the mean signal from each of the two regions in each hemisphere. The first step was to obtain a mask for the anterior and posterior hippocampus independently of my functional data. Subcortical segmentation of the high-resolution anatomical images and intracranial volume estimation were performed in a semi-automated fashion using the Freesurfer image analysis suite, which is documented and freely available for download online ( Briefly, this processing includes removal of nonbrain tissue using a hybrid watershed/surface deformation procedure (Segonne et al., 2004), automated Talairach transformation, segmentation of the subcortical white matter and deep gray matter volumetric structures (including hippocampus, amygdala, caudate, putamen and ventricles; (Fischl et al., 2002; Fischl et al., 2004) intensity normalization (Sled, Zijdenbos, & Evans, 1998), tessellation of the gray matter white matter boundary, automated topology correction (Fischl, Liu, & Dale, 2001; Segonne, Pacheco, & Fischl, 2007) and surface deformation following intensity gradients to optimally place the gray/white and gray/cerebrospinal fluid borders at the location where the greatest shift in intensity defines the transition to the other tissue class (Dale, Fischl, & Sereno, 1999). Manual quality control checks were performed during the procedure. While the automated structure segmentation was performed in standard Talairach space, generated vectors were projected into the subjects native anterior commisure - posterior commusure (AC-PC) space for volumetric assessment, thereby allowing for assessment of the hippocampal volume without distortions introduced through reorientation. Importantly, hippocampal volumes estimated using Freesurfer have been found to be generally consistent with those estimated using manual tracing, with a slight tendency to overestimate hippocampal size due to inclusion of boundary voxels and different assessments of the subicular / entorhinal / parahippocampal boundary (Tae, Kim, Lee, Nam, & Kim, 2008; Cherbuin, Anstey, Reglade-Meslin, & Sachdev, 2009). This overestimation does not obscure differences between groups (Tae et al., 2008) nor does it affect correlations between hippocampal volume and cognitive variables (Cherbuin et al., 2009). Hippocampal segmentations were divided manually into anterior and posterior portions with a coronal bisection in AC-PC space. The most posterior coronal slice of the anterior portion of the hippocampus was defined separately for each hemisphere by

87 74 one rater as the most posterior slice in which the uncal apex was visible. This commonly-employed landmark is easily recognized, unambiguous, and has been promoted as a standard reference for segmenting the anterior and posterior hippocampus (Weiss, Dewitt, Goff, Ditman, & Heckers, 2005). Anterior, posterior and overall hippocampal volumes were then assessed as the number of 1x1x1 mm3 voxels contained in each segmentation. To establish reliability, four raters performed segmentations for the dataset I collected and a high degree of agreement in the position of the uncal segmentation was observed, as determined by comparison of resulting anterior hippocampal volumes (two-way mixed average intraclass correlation coefficient for left hippocampus = 0.99; right hippocampus = 0.99). I projected hippocampal segmentations that were created for each participant using FreeSurfer validation analysis into the functional image space as separate masks. 4.2 Results Behavioral results Participants were alert and engaged in the experimental tasks, as indicated by a high rate of responding in the first Phase I task (proverb meaning task; M=0.85, SD=0.08), and second Phase 1 task (cultural origins task; M=0.90, SD=0.08). As in Experiments 1 and 2, the mean proportion of proverbs identified as South American was not significantly different from an even 50/50 split, M = 0.44, SD = 0.09, t(15) = 1.78, P = n.s. A high rate of responding was also observed in the three scanned Phase 2 tasks (target age task M=0.97, SD=0.02; quality task M=0.97, SD=0.02; control task M=0.98, SD=0.03). As the Phase 3 memory test and all second session tasks were self-paced, all items received responses in those tasks. Because the tasks performed during encoding in Phase 2 were the basis of correct source responses in Phase 3, it was important that participants performed the correct task for all items during Phase 2. For this reason, each task was assigned different response keys to facilitate detection of accidental task perseverations. Pooling all participant responses, perseveration errors were observed in only five trials, which were excluded from further analysis.

88 75 In the second-session proverb identification task, participants categorization of English and Asian proverbs was generally consistent with my expectation that they would know the English but not Asian proverbs prior to the experiment (known English proverbs: M=89.9%, SD=9.5%; known Asian proverbs: M=9.7%, SD=9.7%). I used participants reports to determine the novelty/familiarity status of proverbs: previously known English proverbs were familiar through prior knowledge ; known Asian proverbs were also familiar through prior knowledge unless they were repeated in Phase 1, in which case they were discarded; previously unknown Asian proverbs were familiar through repetition when repeated in Phase 1; and the remaining unknown Asian proverbs and the unknown English proverbs were novel. Applying these designations to the Phase 3 source memory data, I observed a main effect of novelty, F(2,30) = 34.99, P < 0.001, and confidence, F(2,30) = 6.84, P < (Table 13). As the variables did not interact, F(4,60) = 0.14, P = n.s., I collapsed across confidence levels for subsequent analyses. Post-hoc tests revealed better source memory for familiar than novel proverbs, whether proverbs were familiar through experimental repetition, t(15) = 3.16, P < 0.01, or prior knowledge, t(15) = 6.82, P < (Fig. 9). A trend was observed towards better memory in the prior knowledge condition than repetition condition, t(15) = 1.99, P = 0.07, and this trend was significant when the participant excluded from imaging analyses for head motion was included, t(16) = 2.23, P < None of the participants reported having suspected there would be a memory test in advance Functional neuroimaging results The goal of my analysis was not to map out the areas responsible for encoding different types of proverbs, but rather to appreciate the modulating effects of novelty on this particular class of materials. To this end, and to restrict the number of comparisons made in the current study, I limited my investigation to assessment of subsequent memory interactions and conjunctions among the three conditions. Because the medial temporal lobes (MTL) are known to be important for memory in general (Scoville & Milner, 1957), and are often implicated in theories of novelty encoding (e.g., (Tulving et al., 1996), I began my investigation with an assessment of effects within the hippocampus.

89 76 Table 13. Descriptive statistics for Experiment 4 source memory accuracy for and frequency of pre-experimentally familiar proverbs, proverbs with experimentally-induced familiarity, and novel proverbs at high, moderate and low levels of confidence. Number of responses Accuracy Confidence Source of familiarity Mean Standard deviation Standard error Mean Standard deviation Standard error Semantic High Repetition None (novel) Semantic Moderate Repetition None (novel) Semantic Low Repetition None (novel)

90 %% ;:'/':01'(85'::)508<'=:5)!"%#!"%!!"$#!"$!!"##!"#!!!!!! &'() *('+),-.)/)0101'( 2)34(015 67/)8'989431,14:107!"#$%&'(&("$)'*&$+"$(,-%&'.-'/0*&$.(&-1'2386A)8/:'/':01'(8'9810)3<89':8BA15A /4:0151/4(0<8:)54,,)C80A)85'::)508)(5'C1(?804<D8B4<8A1?A):81(80A)80B'89431,14:1078 5'(C101'(<80A4(81(80A)8('+),0785'(C101'(E8:)/,15401(?891(C1(?<89:'38FG/):13)(08H"8 I)3':78B4<8A1?A):89':8/:)J)G/):13)(04,,789431,14:810)3<8:),401+)80'83)3':789':8 10)3<8B10A8)G/):13)(04,,7J1(C=5)C89431,14:107"8K8<0401<0154,,78<1?(19154(08C199):)(5)81<8 C)<1?(40)C8L78MM8408;8N8!"!O84(C8MMM8408;8N8!"!!O"

91 78 In Experiment 3, I observed a novelty-dependent double dissociation of subsequent memory effects in the anterior and posterior hippocampus. In that experiment, whereas successful encoding of novel scenes was predicted by the right anterior hippocampus only, successful encoding of repeated scenes was uniquely associated with bilateral posterior hippocampi only. To determine whether this effect could be replicated using verbal rather than visual materials, I searched for hippocampally-based subsequent memory interactions between the novel and repeated conditions. Consistent with the past result, bilateral posterior hippocampi preferentially predicted encoding success for repeated over novel items (Table 14). The important contribution of the current study was to investigate whether familiarity established through years of prior experience would elicit similar subsequent memory interactions in the hippocampus relative to familiarity induced through Phase 1 stimulus repetition. When I compared subsequent memory correlates in the novel and prior knowledge conditions, I discovered that whereas novel proverbs were associated with greater subsequent memory predictors in the left anterior hippocampus, prior knowledge proverbs were associated with greater subsequent memory predictors in a left posterior hippocampus region that overlapped with the one found in the interaction analysis for novel and repeated proverbs (Tables 14 and 15). An ROI analysis of this overlapping area revealed that while activation in the left posterior hippocampus predicted memory for both repeated and prior knowledge proverbs, it predicted forgetting of novel proverbs (Fig. 10). An ROI analysis of the left anterior hippocampus revealed that the region preferentially predicted memory for novel proverbs, but nonetheless also reliably predicted memory for prior knowledge proverbs. It did not predict memory for repeated proverbs. No direct interaction between repeated and prior knowledge subsequent memory effects was reliably observed in any hippocampal region. To evaluate further the contributions of the left anterior and posterior hippocampal regions to novelty and familiarity encoding, I conducted post-hoc testing of longitudinal interactions using values extracted from the anterior and posterior hippocampal peaks. I focused on effects early in the time course, where the largest subsequent memory interactions were observed in the posterior hippocampus (TR=2), although similar effects were observed late in the time course, where the largest

92 79 Table 14. Regions that preferentially predicted memory success in Experiment 4 for novel or repeated proverbs in one of the two conditions. Time (TR) Region Hemi. BA Peak MNI coordinates X Y Z BSR Spatial extent (mm³) Novel dm > Repeated dm 1 Lateral temporal pole L 21/ Frontal pole R Lateral temporal pole L 20/ Repeated dm > Novel dm 1 Precuneus L Posterior hippocampus / thalamus L Medulla R Precuneus L Precuneus R Cuneus / Precuneus L 7/ Pons L/R Superior temporal g. L Middle temporal g. L Cuneus / Precuneus L/R 7/ Cerebellum R Middle temporal g. R Frontal pole L Cuneus L Middle temporal g. L Frontal pole L Anterior cingulate ctx. L Frontal pole L Note: regions are described in MNI space (Cocoso et al., 1997)

93 80 Table 15. Regions that preferentially predicted memory success in Experiment 4 for novel or prior knowledge proverbs in one of the two conditions. Time (TR) Region Hemi. BA Peak MNI coordinates X Y Z BSR Spatial extent (mm³) Novel dm > Prior knowledge dm 1 Inferior parietal lobule L Cerebellum R Inferior temporal g. R Temporal pole R Anterior hippocampus L Lingual g. R Lingual g. L Prior knowledge dm > Novel dm 1 Posterior hippocampus L Precentral / postcentral g. / precuneus L/R 3/4/6/ Precentral g. L Medial orbitofrontal g. R Precentral / postcentral g. L/R 3/4/ Medial orbitofrontal g. R Note: regions are described in MNI space (Cocoso et al., 1997)

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

95 82 subsequent memory interactions were observed in the anterior hippocampus (TR=7). There was a significant three-way memory x novelty x region interaction, F(2,15)=6.49, P<0.005, confirming a differentiation of subsequent memory correlates along the longitudinal axis of the hippocampus based on stimulus novelty. To facilitate interpretation, I repeated the analysis, including only one form of familiarity at a time. Limiting the analysis to the novel and repeated conditions, the three-way interaction was maintained, F(1,15)=13.17, P<0.005, replicating a three-way interaction observed in Experiment 3. Limiting the analysis to the novelty and prior knowledge conditions, the three-way interaction was again maintained, F(1,15)=7.36, P<0.05, allowing us to generalize the region x novelty interaction to conditions where familiarity is based on prior knowledge rather than stimulus repetition. To determine whether any particular condition contributed to this interaction, I assessed two-way interactions between memory and region separately for each novelty condition. There was an interaction of this type in the novelty condition, F(1,15)=21.45, P<0.001, but not in the repetition condition, F(1,15)=0.02, P=n.s., nor in the prior knowledge condition, F(1,15)=1.31, P=n.s. That is, a hippocampal encoding gradient was observed in the novelty condition but not in either of the familiarity conditions. Returning to inspection of whole-brain bootstrap maps, I also searched for subsequent memory interactions outside of the hippocampus. The left lateral temporal pole and right medial frontal pole preferentially predicted encoding success for novel proverbs relative to repeated ones (Table 14 and Fig. 11a). Repeated proverbs were preferentially encoded by a larger and predominantly left-lateralized set of regions, including bilateral cuneus/precuneus, the left anterior cingulate cortex, left lateral temporal frontal pole and middle and superior temporal gyri. In the subsequent memory interaction contrast of novel and prior knowledge proverbs, novel items were again associated with greater memory success effects in the temporal pole, whereas greater memory success effects for prior knowledge items were again found in bilateral precuneus regions, as well as in medial orbitofrontal gyrus (Table 15 and Fig. 11b). A direct comparison of the two familiarity conditions (repetition and prior knowledge) revealed a small set of regions that preferentially predicted encoding success for prior knowledge proverbs, including the left parahippocampal gyrus, left superior frontal gyrus and medial orbitofrontal gyrus (Table 16 and Fig. 11c). In contrast, preferential encoding of repeated proverbs

96 83 Table 16. Regions that preferentially predicted memory success in Experiment 4 for repeated or prior knowledge proverbs in one of the two conditions. Time (TR) Region Hemi. BA Peak MNI coordinates X Y Z BSR Spatial extent (mm³) Prior knowledge dm > Repeated dm 1 Medial orbitofrontal g. L/R Superior frontal g. L Posterior hippocampus / parahippocampal ctx. L Repeated dm > Prior knowledge dm 1 Inferior frontal g. R Inferior temporal g. R Inferior temporal g. R Inferior temporal g. R Frontal pole R 10/ Inferior frontal g. R Inferior temporal g. R Cuneus / Precuneus L/R 7/ Frontal pole R Entorhinal / perirhinal ctx. R Inferior / middle temporal g. R Cuneus / Precuneus L/R 7/ Frontal pole L Superior temporal g. R Cuneus / Precuneus L/R 7/ Frontal pole L Frontal pole L Entorhinal / perirhinal ctx. R Precuneus L Superior temporal g. / 8 insula R Precuneus R Note: regions are described in MNI space (Cocoso et al., 1997)

97 BC!"#$%&'((!"#$%&'(&&&&&&&)$*$+,+,"-&'( 34!"#$%&'(&&&&&&&./,"/&0-"1%$'2$&'( /,"/&0-"1%$'2$&'(&&&&&&&)$*$+,+,"-&'(!"#$%&'()*+$,-$./",0.1-0.&$2*/&0".+-"3-&4$-35%%-)2* $209$.&-:;-)*"%+,"-&' "./&%*0/"1.-'2&/,&&.'/3&'/3%&&'.14&5/6'01.7"/"1.-'8'.14&59'%&:&*/&7'*.7':%"1%';.1,5+ &7#&'8'".'/3&'2%*".'01%%&5*/&-'1<'-$2-&=$&./'>&>1%6'-$00&--'*%&'7&:"0/&7'$-".#'*'

98 85 was observed in a broad network that included right inferior, medial and superior temporal gyri, bilateral cuneus/precuneus regions, right frontal pole and right entorhinal/perirhinal cortex. To assess possible overlap in encoding correlates across conditions, I conducted several conjunction analyses. In the novel and repeated conditions, both of which involved materials not known prior to the study, memory success was predicted by the right anterior hippocampus/entorhinal cortex and insula (Table 17). This right anterior hippocampal/entorhinal region had a functional profile similar to that of the left anterior hippocampal region described above (there was no three-way interaction of memory x novelty x region between the left and right hemispheres, F(2,15)=0.85, P=n.s). In the novel and prior knowledge conditions, both of which involved materials presented for the first time in the experimental context, memory success was predicted by the left anterior hippocampus, left temporal pole and right amygdala. In the repeated and prior knowledge conditions, both of which involved materials that were familiar at the time of encoding, memory success was predicted by the left posterior hippocampus, left precuneus, right inferior parietal lobe and right posterior cingulate gyrus. Finally, I tested the hypothesis that functional connectivity in the anterior and posterior hippocampus would differ, thereby contributing to an explanation of the regional interactions described above in terms of differing neural contexts. In the current analysis, I assessed resting-state fmri collected following phase 2. I used as seeds the average signal from the entire anterior and entire posterior hippocampus, divided at the uncal apex, a standard landmark for longitudinally segmenting the hippocampus (Weiss et al., 2005). Thus, unlike in Experiment 3, not only were the data used for connectivity analysis drawn from an independent fmri image series during which no task was performed, but seeds were obtained independently of functional observations within the studied group. I then searched for any differences in the ambient functional networks associated with left and right phpc and ahpc. The largest (and only) stable pattern to emerge from this data-driven analysis corresponded to a contrast of phpc and ahpc bilaterally, an effect that was particularly strong in the right hemisphere (singular value = 8.9; P < 0.05; Fig. 12; Table 18). phpc and ahpc linked to different elements of an archetypical hippocampus-seeded default network (Buckner et al., 2008).

99 86 Table 17. Regions that predicted memory success in Experiment 4 for more than one condition at a particular latency from proverb onset. Time (TR) Region Hemi. BA Peak MNI coordinates X Y Z BSR product Spatial extent (mm³) Novel dm & Repeated dm 7 Anterior hippocampus R Anterior hippocampus / perirhinal ctx. R Insula R Novel dm & Prior knowledge dm 2 Temporal pole L 21/ Repeated dm & Prior knowledge dm 1 Posterior hippocampus L Inferior parietal l. R Posterior cingulate g. L Posterior cingulate g. R 23/ Note: regions are described in MNI space (Cocoso et al., 1997).

100 87 Table 18. Voxel clusters distinguishing functionally-connected networks seeded from the anterior versus posterior hippocampus in Experiment 4. Region BA Hemi. Peak MNI coordinates X Y Z BSR Spatial extent (mm³) Anterior > Posterior Temporal lobe Temporal pole 21/38 R Temporal pole 21/38 L Amygdala - L Pons L/R Amygdala - R Posterior > Anterior Frontal lobe Dorsolateral prefrontal ctx. 9/46 L Anterior cingulate g. 24/32 L Dorsolateral prefrontal ctx. 46 R Posterior cingulate g. 23/31 L/R Thalamus - L/R Parietal lobe Inferior parietal l. 7/40 R Inferior parietal l. 40 L Inferior parietal l. 39/40 L Precuneus 7 L Occipital lobe Cuneus / lingual g. 17/18 R Cuneus / lingual g. 17/18 L Note: Regions were defined as any cluster with a peak of at least BSR 3.29 (P < 0.001) and a spatial extent of 12 or more voxels surviving a threshold of BSR 2.81 (P < 0.005). Coordinates are displayed in MNI space (Cocoso et al., 1997).

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

102 89 ahpc correlated preferentially with lateral temporal cortex in both hemispheres, extending to the temporal poles bilaterally (Fig. 12b-e). Anterior-hippocampus linked motor regions observed in Experiment 3 did not reappear here, likely reflecting the fact that the current dataset was collected while participants rested, whereas the data evaluated in Experiment 3 was collected while participants responded to task demands. Activity in phpc correlated preferentially with bilateral dorsolateral prefrontal cortex (PFC), left anterior cingulate cortex (ACC), bilateral posterior cingulate cortex (PCC) / retrosplenial cortex, left precuneus, bilateral thalamus (including anterior and dorsomedial nuclei), bilateral inferior parietal lobe (IPL) and bilateral occipital gyrus regions (Fig. 12b-e; Table 18). This set of regions shows considerable overlap with the set of posterior-hippocampus linked regions observed in Experiment 3, which also included DLPFC, thalamus, ACC, PCC, IPL, and occipital regions. 4.3 Discussion In the current experiment, I sought to determine whether the modulation of hippocampal encoding activity in familiarity in Experiment 3 was attributable to shortterm factors caused by in-laboratory familiarization procedures or to effects related to the retrieval of information from long-term memory more generally. Behaviorally, I replicated the Experiment 1-3 source memory advantage for all types of familiar materials over novel materials. Also as in Experiment 3, encoding activity in the hippocampus was modulated by repetition of materials: left posterior hippocampal activity predicted memory success for repeated proverbs and memory failure for novel ones. Critically, this same encoding interaction was observed when comparing the subsequent memory correlates of prior knowledge proverbs. This result indicates that familiarity, whatever its form, modulates both episodic memory and its hippocampal substrates. As in Experiment 3, I found that familiarity induction through pre-experimental stimulus repetition modulated encoding activity in the hippocampus: the left posterior hippocampus predicted source memory for repeated but not novel materials. This was the case even though the current experiment involved a different class of stimulus

103 90 materials (proverbs instead of scenes). Interestingly, left posterior hippocampal activity did not just fail to predict memory success for novel items: it also predicted memory failure. Negative signals have sometimes been found to predict memory (Daselaar et al., 2004). To speculate about a possible explanation, novel proverbs may have sometimes been confused with similar repeated ones during study. Were this to occur and the novel item become processed as a familiar one, greater posterior hippocampal activity might result from the inappropriate retrieval. Meanwhile, memory would fail, since source information would not be linked to the correct proverb. In this way, the posterior hippocampal deactivation in the novelty condition can be interpreted as evidence of proactive interference based on the familiarization phase (or perhaps based on prior experience with proverbs). I also searched for modulation of hippocampal encoding correlates by familiarity arising from prior knowledge. Because pre-experimental exposures preceded neither the prior knowledge nor the novelty condition, this comparison can be said to reflect the modulatory effects of familiarity without the influence of priming, interference, or other short-term factors that could potentially cloud interpretation of possible differences in encoding correlates. Critically, an interaction of the same kind observed between the pre-experimental familiarity and novelty conditions was observed in the left posterior hippocampus, which also predicted memory success for prior knowledge proverbs. This result indicates that stimulus familiarity, whatever its form, modulates posterior hippocampal encoding activity. Specifically, the posterior hippocampus is associated with successful encoding of source information only in those trials in which materials are familiar. In Experiment 3, a different pattern was observed in the anterior hippocampus, which predicted memory for novel but not familiar items. However, in the current experiment, the anterior hippocampus predicted memory for novel items (both hemispheres) as well as repeated items (right hemisphere) and prior knowledge items (left hemisphere). To speculate, one possible explanation for the divergence is that the proverb stimuli used in the current experiment retained some novelty in terms of their deeper meaning even after multiple exposures. This may have been the case even for well-known proverbs, which participants previously may not have contemplated deeply. In contrast, the scenes that were repeated in Experiment 3 had only surface features and no deeper

104 91 meaning, and as such might have lost their novelty more completely after several exposures. It is interesting to interpret these effects in the framework of a hippocampal encoding gradient that describes the contribution of various aspects of the hippocampal long axis to memory encoding. One such framework is the HIPER model, which links the anterior hippocampus to memory encoding and the posterior hippocampus to memory retrieval (Lepage et al., 1998). Inconsistent with this framework, no gradient was observed for either familiarity condition; that is, both anterior and posterior aspects of the hippocampus predicted memory for familiar stimuli in the same way. A similar observation was made in a review by (Schacter & Wagner, 1999), who found that, in the fmri literature, encoding activations were reported all along the main axis of the hippocampus. Interestingly, however, the novelty condition did produce such a gradient, with the anterior hippocampus predicting memory success and the posterior predicting memory failure. This pattern, as well as the similar one observed in Experiment 3, can be accommodated by a small adjustment to the HIPER model. Because the posterior hippocampus was only recruited for encoding during trials that could have triggered retrieval events (i.e., that involved familiar materials), the posterior hippocampus appears to play a selective role in the integration of new episodic information with existing memory representations, regardless of the form of those representations. Such a role would most likely be based on long-lasting long-term memory effects rather than short-lived stimulus exposure effects (e.g., priming), since posterior hippocampal encoding of familiarity held in the current experiment regardless of whether materials were familiar due to prior knowledge or pre-experimentally repeated. Such a gradient could be explained in part by different neural contexts associated with anterior and posterior hippocampus. Interestingly, the regions associated with each area closely overlapped with the ones observed in Experiment 3. Both experiments replicate posterior and anterior hippocampal functional connectivity differences observed elsewhere (Kahn et al., 2008), although the current experiment more closely replicates the procedures of the experiment by Kahn and colleagues than Experiment 3, since the current experiment involved analysis of resting fmri data rather than data collected during an experimental task. The fact that participants were responding to

105 92 experimental tasks in Experiment 3 while data was being collected likely explains why a correlation with hippocampal regions and motor regions appeared in that experiment but not the current one. Interestingly, the sets of regions identified as posterior and anterior hippocampal correlates in these experiments appears to converge on the cortical connections of two known hippocampal pathways: the polysynaptic intrahippocampal pathway, which connects with frontal and parietal cortices via the fornix, and the direct intrahippocampal pathway, which projects to the anterior temporal lobe via the uncinate fasciculus (Duvernoy, 2005; Fig. 12). The posterior hippocampus correlated preferentially with the cortical connections of the polysynaptic pathway and the anterior hippocampus with those of the direct pathway. This pattern replicates although the current result extends prior evidence by formally demonstrating the stability of the overall pattern. Because there is little communication between posterior and anterior hippocampus (Moser & Moser, 1998; Fanselow & Dong, 2010), functional specialization could be expected in posterior and anterior hippocampus based on these different neural contexts. The cortical connections of the polysynaptic pathway are believed to support recollection by mediating perceptual (precuneus), attentional (inferior parietal) and strategic (lateral frontal) contributions to it (Spaniol et al., 2009). Along these lines, integrity of the fornix, which connects the polysynaptic pathway to cortex, is important for recollection memory (Aggleton & Brown, 1999; Gilboa et al., 2006; Tsivilis et al., 2008). In contrast, the anterior temporal connections of the direct pathway are more closely associated with processing of semantic information and social and emotional cues (Rogers et al., 2006; Olson, Plotzker, & Ezzyat, 2007). As the posterior hippocampus linked preferentially with polysynaptic pathway connections, which are important for recollection, a neural context interpretation could be consistent with the finding that activity in the posterior hippocampus predicted memory only for familiar materials, i.e., those involving memory retrieval. That is, because it is the posterior portion of the hippocampus has the most direct access to the network needed to recover old memories, it is reasonable that the posterior hippocampus should play a special role in memory operations that involve retrieval, such as building new associations with materials that already exist in memory.

106 93 Finally, although the current behavioural results do not directly bear upon the specific experimental question posed in the current chapter whether modulation of hippocampal encoding activity is attributable to long-term memory they do again replicate the Experiment 2 finding that source information is better retained for familiar items than novel ones, regardless of whether the familiarity is due to experimental induction or prior knowledge. This result reinforces the claim made in previous chapters that familiarity, rather than novelty, enhances encoding into episodic memory. In addition, performance was significantly higher for prior knowledge items than repetition items in the current experiment, whereas it was only numerically higher in Experiment 2. It is difficult to tell whether it is more robust familiarity or the difference between episodically- and semantically-based familiarity that drove this effect.

107 Chapter 5 General Discussion In the current dissertation, I conducted four experiments to investigate the effects of stimulus novelty on episodic memory and on the brain basis of episodic memory formation. In Experiment 1, I identified a confound between novelty and discrimination demands in conventional experiments used to evaluate the cognitive effects of novelty on episodic memory. I illustrated how the confound could have led to prior observations of novelty advantages in memory. In Experiment 2, I evaluated memory in the absence this confound and identified a mnemonic advantage of familiarity, whether familiarity was established through prior knowledge or pre-experimental repetition. In Experiment 3, I searched for interactions in the brain basis of episodic memory encoding when scene stimuli were novel or familiar. I identified a double dissociation in the hippocampal correlates of successful memory formation based on whether materials were novel or familiar, as well as distinctive functional connectivity associated with the novelty- and familiarity-specialized hippocampal regions. Finally, in Experiment 4, I replicated the hippocampal dissociation observed in Experiment 3 using verbal materials. I also extended findings of a special sensitivity of the posterior hippocampus to repeated materials to materials that were familiar due to prior knowledge. These experiments consistently showed that familiarity, not novelty, enhances episodic memory when a common confound is addressed, and that contrary to current hypotheses (Tulving et al., 1996), no privileged relationship exists among novelty, the hippocampus, and memory. Instead, they showed that hippocampal sub-regions that are separated into anterior and posterior portions along its long axis specialize in processing events involving novel and familiar materials, respectively. Other evidence from these experiments suggests these properties may have emerged on the basis of local differences in functional connectivity. 94

108 Summary of experimental effects An important goal of the current dissertation was to explore the basis of conflicting novelty and familiarity effects. Two factors that could potentially account for the discrepancy were considered: type of familiarity and conflation with distinctiveness effects. As prior knowledge and repetition-based familiarity types had similar impacts on memory performance and hippocampal encoding correlates, familiarity type cannot explain the discrepancy. Instead, a failure to distinguish novelty and distinctiveness was the important factor. In Experiment 1, I found that distinctiveness accounted for the obtained novelty effect (Tulving & Kroll, 1995; Aberg & Nilsson, 2001, 2003), in agreement with claims that the two factors are related (Hunt & Lamb, 2001; Kishiyama & Yonelinas, 2006; Tulving & Rosenbaum, 2006). Experiments 2, 3 and 4, which controlled for distinctiveness, consistently supported the traditional view that familiarity facilitates episodic memory, whether familiarity comes in the form of recent similar experiences (Ebbinghaus, 1913; Hintzman, 1976) or semantic memory (Wickelgren, 1979; Nadel & Moscovitch, 1997; Tulving & Markowitsch, 1998). The finding that the conditions eliciting better memory (familiarity based on prior knowledge and pre-experimental familiarity) were also uniquely associated with encoding correlates in the posterior hippocampus resonates with a growing body of evidence indicating that posterior hippocampus has a closer relationship with episodic memory than anterior hippocampus. Intracranial recordings from temporal lobe epilepsy patients show larger verbal encoding and retrieval successful effects the further posterior signal is recorded (Ludowig et al., 2008). Massive accumulation of spatial information by London taxi drivers is associated with increases in phpc volume (Maguire et al., 2000), and posterior, but not anterior, hippocampal volume positively predicts episodic memory ability in normal individuals (Poppenk & Moscovitch, submitted). Lesion patients with spared posterior hippocampi (phpc) retain spatial memories (Maguire, Burke, Phillips, & Staunton, 1996), just as medial temporal resections limited to the amygdala and uncus but including the uncal (anterior) part of the hippocampal head have a more limited impact on memory than resections extending more

109 96 posteriorly (although this finding is confounded with the amount of resected tissue; Scoville & Milner, 1957; Smith & Milner, 1981). In rats, damage to the dorsal but not ventral portion of the hippocampus (analogous to human posterior and anterior portions, respectively) impairs memory performance in the Morris water maze (Moser & Moser, 1998). Together, these studies indicate that phpc may play a particularly important role in episodic memory function. In contrast, anterior hippocampus (ahpc) has been more closely linked with habituation, novelty and motivational/emotional behaviour. Nadel (1968) found that rats with ventral hippocampal lesions were slower to habituate to a novel environment and slower to extinguish conditioned responses, and Strange and Dolan (2001) described a tendency of novelty activations in neuroimaging studies to appear in the anterior rather than posterior hippocampus. Concerning control of motivational/emotional behaviour, in a review of the literature, Fanselow and Dong (2010) found that ventral but not dorsal hippocampal lesions in rats are reported as enhancing stress ulcers but decreasing anxiety and corticosterone levels in confinement situations. The anterior/posterior dissociation in hippocampal encoding specialization observed in the current experiments may be explained in part by local variation in hippocampal connectivity. The notion that regions perform operations on the basis of the network to which they are functionally linked, i.e., their neural context, has been forwarded as a principle for understanding neural contributions to cognition (McIntosh, 1999). Drawing upon this principle, the functions a region can undertake may be constrained by the networks to which it has anatomical access. As discussed in Chapter 4, anterior and posterior aspects of the primate hippocampus are sparsely interconnected (Moser & Moser, 1998; Fanselow & Dong, 2010), leaving open the possibility that these areas could develop specialization as a consequence of potentially different connectivity with the neocortex. Experiments 3 and 4 both revealed such differences in large-scale connectivity in anterior and posterior hippocampus, a pattern that was established as stable in Experiment 4. Although the two experiments did not reveal precisely the same functional networks, differences could easily have arisen as a consequence of the different data sources used in the two

110 97 experiments: functional connectivity was evaluated using data from the study phase in Experiment 3, whereas it was evaluated using data from a post-study rest phase in Experiment 4. In spite of minor differences, the networks observed in the two experiments involved remarkable overlap: the lateral and anterior temporal lobes were preferentially associated with the anterior hippocampus, whereas the frontal, parietal and visual cortices as well as the thalamus were preferentially associated with the posterior hippocampus. A similar set of regions was linked to the anterior and posterior hippocampus in an experiment by Kahn et al., 2008, and these different sets of regions are believed to be part of different hippocampal anatomical pathways (Duvernoy, 2005). The posteriorlinked network has been shown to be important for recollection memory in humans (Aggleton & Brown, 1999; Gilboa et al., 2006; Tsivilis et al., 2008). Moreover, in rats, entorhinal cortex has been shown to transmits finely tuned spatial and possibly item information to the dorsal but not ventral hippocampus (Hargreaves, Rao, & Knierim, 2005; Kjelstrup et al., 2008), creating favorable conditions for associative memory. Combined with current observations, this evidence strongly suggests that the neural context of the posterior hippocampus is supportive of episodic encoding operations that involve retrieval, potentially explaining how specialization of function along the hippocampal axis may have arisen. In spite of general correspondence of hippocampal encoding and functional connectivity findings between Experiments 3 and 4, I observed puzzling differences in the extra-hippocampal subsequent memory correlates of novel and repeated items between the two experiments. In Experiment 3, memory for the repeated condition was predicted by a smaller network than the one that predicted memory for the novel condition; in fact, it was largely a subset of the one in the novel condition. This finding can be most easily understood in the context of priming, since either fewer neurons or less activation were required for successful memory formation (Grill-Spector, Henson, & Martin, 2006). In contrast, the network associated with memory success in the Experiment 4 repeated condition was larger than and largely non-overlapping with the network associated with memory success in the novelty condition. It is not clear

111 98 why this difference would appear across the two types of experiments. One possibility concerns the nature of the stimuli employed. Semantic information may be extracted very rapidly from novel scenes, with basic semantic decisionmaking response times as low as 150 ms (Fabre-Thorpe, Delorme, Marlot & Thorpe, 2001). In contrast, extraction of semantic information from linguistic materials is much slower, with basic decision-making responses consuming around 3500 ms (Carpenter, Just, Keller, Eddy, & Thulborn, 1999), and likely even more time required for a deeper comprehension of difficult materials such as proverbs. As a result, the potential savings in processing time from retrieval of proverbs is larger than the potential savings from retrieval of scenes, and the possibility of divergence in processing between repeated and novel scenes is also correspondingly larger. This interpretation, while speculative, is offered here to provide at least one possible explanation for the divergent findings between experiments. 5.2 Theoretical integration of cognitive results Although the current set of experiments consistently showed advantages of familiar materials in memory, there are conditions in which benefits of stimulus novelty for subsequent memory are obtained even when discrimination demands (and source monitoring requirements) are equated. For example, oddball materials in serial list-learning experiments, which are novel with respect to their list context, are typically better remembered than non-oddballs (von Restorff, 1933; Hunt & Lamb, 2001). However, novelty in oddball studies is conceptualized as deviance from some intra-list contextual pattern, such as a common font or semantic category, whereas I manipulated whether or not materials were associated with prior memory representations. At the brain level, oddball stimuli, like novel stimuli in the current experiment, have been associated with anterior hippocampal responses (Strange and Dolan, 2001); however, subsequent studies have found hippocampal responses to oddballs

112 99 arise from stimulus novelty of oddballs rather than oddballs status per se (Bunzeck & Düzel, 2006). Another approach has focused on word frequency. Words that are less frequent (hence more novel) are generally associated with more hits and fewer false alarms, a phenomenon known as the mirror effect (Glanzer & Adams, 1985; Hockley, 1994; Aberg & Nilsson, 2003). However, numerous features differentiate low- and high-frequency words, and the mirror effect may be inverted. Chalmers and Humphreys (1998) found familiarization worsened recognition memory for rare and high-frequency words on a later study list, but familiarization with rare words improved memory when accompanied by a definition. A similar interaction was observed by Stenberg, Hellman and Johansson (2007), who found famous names were much better recognized than non-famous ones, but high-frequency names were less well recognized than lowfrequency ones. Together with the current results, these outcomes suggest memory is improved by pre-study exposures when those exposures help to establish a semantic foundation; however, the exposures required to lay down this foundation bear a cost in terms of reduced clarity about the source of the memory. It is possible that this confusion reflects the buildup of many episodic memories one for each prior exposure in line with models that predict a new episodic memory is created each time information is attended (e.g., Nadel & Moscovitch, 1997). From this perspective, source confusion arises from the increased probability that the wrong episodic memory will be retrieved during testing. Another account is that the high familiarity of stimuli that have been presented many times encourages participants to endorse those materials as belonging to a particular list, even when recollected information suggests otherwise (Reder, Paynter, Diana, Ngiam & Dickison, 2008). While the novelty paradigms discussed so far in this dissertation have been based on intra-list effects, another approach has focused on effects of novelty that operate over longer time scales. In particular, five minutes of exposure to novel rather than familiar scenes shortly before a study phase has been shown to enhance participants later recollection memory and free recall of the studied

113 100 verbal materials (Fenker et al., 2008). Notably, these effects were not confounded by the distinctiveness effects discussed earlier, since the manipulation of novelty concerned stimulus events preceding the study phase that were not themselves the subject of a memory test. Rather, all of the words on the memory test were familiar, having been repeated in a previous experimental session (Fenker et al., 2008). These findings appear to have their basis in the pharmacological literature: hippocampal long-term potentiation (LTP), a cellular mechanism believed to be important for memory formation (Malenka and Bear, 2004), has been found to be encouraged in rats by exposure to spatial novelty, with effects persisting up to 30 minutes following exploration of a novel environment (Li et al., 2003). Facilitation of LTP has been shown to be dependent on activation of dopamine receptors in the CA1 subfield of the hippocampus, linking mnemonic facilitation by novelty to the function of the dopaminergic reward system (Li et al., 2003). Similar effects have been observed in humans, where neuroimaging studies have identified activation of dopaminergic midbrain structures in response to both stimulus novelty (Bunzeck & Düzel, 2006) and associative novelty (Schott et al., 2004). Novelty-seeking personality traits, reward motivation and attention to reward modulate these relationships (Bunzeck, Doeller, Fuentemilla, Dolan & Düzel, 2009; Krebs, Schott & Düzel, 2009; Krebs, Schott, Schutze, & Düzel, 2009). Taken together, this evidence suggests that novelty exerts an influence over memory via its reward characteristics, although this influence operates over a timescale of minutes to hours (for review, see Düzel, Bunzeck, Guitart-Masip, & Düzel, 2010). As a result, it is not surprising that such effects were not detected in the current experiments, since novel and familiar stimuli were intermixed in short (< 1 min.) blocks. When combined with the current findings, this line of results indicates that novelty can enhance memory in general by energizing participants; however, novel stimulus events are not themselves preferred by the episodic memory system, which more easily accommodates familiar information. It is also important to note that, in the current dissertation, novelty was operationalized in terms of stimulus novelty (i.e., the novelty of materials encoded into episodes), since this type of novelty is the focus of studies that have been influential on models of novelty encoding (e.g., Tulving et al., 1994; Tulving

114 101 & Kroll, 1995). However, other types of novelty are actively discussed in the cognitive neuroscience literature: for instance, the context of stimulus presentation may be either novel or familiar (i.e., contextual novelty; Nyberg, 2005). Notably, all study-phase presentations of novel and familiar stimuli in the current experiments involved contextual novelty. More concretely, the preexperimental repetition task always differed from the task(s) in the critical memory-encoding phase, which in turn were surely different from the life context in which prior knowledge materials were acquired. The presence of contextual novelty in all conditions was necessary for the behavioural measurement of memory for individual episodes; to determine whether a particular event has been encoded, it must have some unique contextual feature (i.e., novelty) that distinguishes it from earlier trials. Without such a feature, it is possible to determine whether a memory record for the event exists, but impossible to determine when the record was created (at least, based on explicit behavioural evidence). Because all conditions in the experiments in this dissertation contained contextual novelty, contextual novelty cannot explain the current findings. Indeed, some degree of contextual novelty will invariably separate any two events owing to the flow of time, the chaotic nature of the sensory world, and rapidly changing internal processes, making it a poor explanatory variable. In the words of the Greek philosopher Hetroclites, one can never step into the same river twice. 5.3 Theoretical integration of fmri results Turning to the neural substrates of novelty processing, the relation of the current results to neural priming deserves examination. As one researcher has pointed out, the paradigms most frequently used to evaluate novelty and repetition suppression effects are identical, since both types of paradigm require comparison of the neural signature of novel and repeated materials (Habib, 2001). The difference lies in interpretation: new > old activations may be interpreted as evidence of either novelty processing or repetition suppression,

115 102 whereas old > new activations may be interpreted as evidence of either familiarity processing or repetition enhancement. It is therefore not surprising that just as anterior hippocampal novelty responses are often observed in novelty investigations (Strange & Dolan, 2001), they are also reported in studies on repetition suppression (e.g., Jessen et al., 2002). In light of the tension between these theoretical concepts, it is important to be clear about where ambiguities do and do not exist in the current set of experiments. First, as a new > old design, the initial comparison of BOLD responses in the new and old conditions in Experiment 3 is subject to the same theoretical ambiguity described above. The main analyses of interest in Experiments 3 and 4 concern novelty-based interactions in subsequent episodic memory correlates, rather than a direct comparison of new and old conditions. These effects rely on interactions between novelty and another variable. While these analyses clearly reveal regions that specialize in encoding novel and repeated items, this distinction could still easily be described as either a specialization based on novelty status or based on the effects of recent neural priming. However, Experiment 4 hippocampal effects distinguishing signals in the novelty and prior knowledge conditions (i.e., new > known or vice-versa) are more difficult to interpret using a neural priming account, since these familiar materials may not have been encountered for many years. Neural priming effects are thought to operate on a much shorter time scale (although see Mitchell, 2006, for evidence that behavioural priming can persevere over a time scale of decades). Therefore, the perseverance of the effects of familiarity in the context of prior knowledge items argues against an interpretation based on repetition suppression. In the current dissertation, discussion of the neural basis of novelty processing has so far focused on whether or not the hippocampus selectively or preferentially encodes novel information. However, there is also detailed discussion about the mechanisms by which the hippocampus determines whether novelty is present in the first place. Current models draw upon the local circuitry of the hippocampus to support the hypothesis that the structure probes for differences between perception and predictions that are generated based on memory retrieval (Vinogradova, 2001; Lisman & Grace, 2005). The specific

116 103 implementation of this comparator operation varies among models: some involve differences in timing between cortical and subcortical inputs (e.g., Vinogradova, 2001), whereas others incorporate top-down pharmacological modulation from supervisory structures (e.g., Lisman & Grace, 2005). All models involve a loop with dual inputs that converge in either the CA1 or CA3 field of the hippocampus. The comparator s operation begins with perceptual or neocortical input to the hippocampus in one branch of the loop, which triggers pattern completion if there is sufficient overlap between the input and past experience. Any retrieved information from this branch of the loop may be conceptualized as a prediction that is relayed to the remaining CA1/CA3 field and met by the other branch of the loop, which carries a copy of the perceptual or neocortical input. A comparison of the predicted and veridical inputs is conducted (or the timing of their arrival; see Vinogradova, 2001), at which point various cognitive consequences are hypothesized in the case of a match or mismatch. Mismatch detection has been shown to evoke orienting responses in rabbits (Vinogradova, 2001). Drawing upon the novelty-encoding hypothesis (Tulving et al., 1996), it has also been suggested that mismatches lead to enhanced memory for novelty stimuli (Lisman & Grace, 2005). For reasons that have been discussed extensively already, the current results strongly argue against the latter proposal. While I have not explored the possibility that partial stimulus novelty (i.e., changed stimuli) enhances memory in the current set of experiments, pilot experiments based on the ones presented here have not revealed change-based enhancements of memory. Using fmri, Kumaran and Maguire have conducted several experiments testing the hypothesis that the hippocampus is specifically sensitive to mismatches (for review, see Kumaran & Maguire 2009). In their experiments, memory retrieval is initiated by a stimulus or stimulus sequence that was recently presented. This retrieval stage is required to generate predictions that could lead to a subsequent match or mismatch (Kumaran and Maguire, 2009). Contextual information that accompanied the original stimulus or stimulus sequence such as order, orientation, spatial location or an associated stimulus is replicated, altered, or replaced with new information. Kumaran and Maguire have consistently

117 104 observed activation in the hippocampus when contextual information is altered relative to when it is replicated or replaced (2009). Notably, all conditions in these experiments contain an equal amount of stimulus novelty; only the novelty of contextual information is varied. This setup helps to ensure that it is the degree to which predictions are met that is manipulated, rather than the degree to which predictions are generated in the first place. For this same reason, it is difficult to interpret stimulus novelty effects of the type examined here in the framework developed by Kumaran and Maguire. According to Kumaran and Maguire, designs that manipulate stimulus novelty introduce insufficient overlap between stimulus presentation in the study phase and previous experience to trigger pattern completion and generate predictions. Even in the case of repeated stimuli, contextual differences between earlier repetitions and the study phase prevent the generation of predictions. Similarly, no predictions would be generated for materials that were novel at study, since no pattern completion would be expected. Therefore, according to this interpretation, match-mismatch theory makes no predictions about the hippocampal correlates of stimulus novelty (Kumaran and Maguire, 2009). The claim that familiar stimuli do not trigger hippocampal pattern completion is unsubstantiated by Kumaran and Maguire (2009) and appears to be a weak point in this line of reasoning, since complete reinstantiation of an earlier event is not thought to be necessary for ecphory (Moscovitch et al., 2005). However, even were pattern completion to occur, it is still difficult to imagine what meaningful predictions participants would have derived from stimuli that were repeated in an earlier, irrelevant task, and whether those predictions would have been ignored or deemed a mismatch in the unrelated study task context. The awkwardness of a comparator interpretation under these circumstances is indicative of a broader problem for the mismatch framework: the conditions under which predictions, matches and mismatches are generated are not well specified. This is also the subject of disagreement among implementations of the comparator model: for instance, one formulation posits that stimulus novelty detection and mismatch detection are functionally equivalent (Vinogradova, 2001). This position was formed on the basis of electrophysiological recordings

118 105 from waking, unanaesthetized rabbits revealing systematic hippocampal mismatch responses to stimulus novelty (Vinogradova, 2001). For the purpose of exploration, let us consider the possibility that familiar stimuli did, in fact, trigger an initial match and elicit predictions that were subsequently violated by the presence of a different task accompanying the presented stimuli. The resulting cascade of hippocampal events could have led to enhanced hippocampal processing of familiar items and, as a consequence, accounted for the observed cognitive and hippocampal familiarity effects. However, in Experiment 3, unique areas of hippocampal activation were observed in both the novelty and familiarity condition. According to Kumaran and Maguire s (2009) model, only conditions in which a match is generated should lead to predictions that may be violated, leading to subsequent activation of the hippocampus. It is also not clear how comparator models ought to handle semantically-based memory representations. For present purposes, it can be assumed that no prediction is generated, since contextual information is not believed to be stored in semantic memory (Tulving, 1972). The neural architecture of the hippocampally-based mismatch detection loop also does not lend itself to predictions that are generated in the neocortex (Vinogradova, 2001; Lisman & Grace, 2005). In spite of this, hippocampal interactions were observed between the novelty and prior knowledge conditions, just as they were observed between the novelty and repetition-based familiarity conditions. Overall, this exploration reveals that hippocampal interactions based on stimulus novelty of the kind seen in the current experiments are difficult to accommodate using a comparator account. Because of this observation and the poor specification of key variables in comparator models, comparator models do not presently contribute a satisfactory model of novelty processing in the hippocampus.

119 An alternative model of hippocampal memory function Because no current neuropsychological model appears to offer a satisfactory account of novelty processing, memory and hippocampal function, a new one is proposed here. A complete model should accommodate the well-established effects obtained from the various lines of research discussed in this chapter. These include: Superior episodic encoding of familiar over novel stimulus events. Interactions between the novelty of information and locus of activation or encoding correlates along the longitudinal axis of the hippocampus. Greater importance of the posterior hippocampus for episodic memory than the anterior hippocampus. Differing neural contexts of the posterior and anterior hippocampus. Hippocampally-based pharmacological energization by novelty. Match-mismatch effects in the hippocampus. In addition, because novelty is not only the subject of discrimination in memory tests but serves as a signpost of important features of the environment, a complete model of novelty processing should include both an explicit decisionmaking mode that makes use of executive resources (e.g., for making explicit decisions about novelty) and a passive mode that uses few or no executive resources (such that novelty may usefully guide allocation of attention in the first place). I propose that hippocampal novelty processing operates according to the following principles, which are described here collectively as the dual-comparator model of hippocampal novelty processing. The innovative aspect of the proposal lies in principles 6 and 7, which restrict existing models to sub-regions of the hippocampus; the remaining principles are drawn from elsewhere in their

120 107 original form or follow from other principles. As a whole, the model describes parallel anterior and posterior hippocampal comparators that underlie automatic novelty processing and explicit memory judgment. The model is not intended to serve as a comprehensive model of hippocampal function, but does hypothesize elementary operations that may underlie some others. 1. The hippocampus binds relational information into physical episodic memory traces, each of which serves as a neocortical index of a previous event (Teyler and DiScenna, 1986; Moscovitch, 1992). 2. Because of sparse interconnectivity between posterior and anterior aspects of the hippocampus, the two sub-regions are functionally independent of one another (Moser & Moser, 1998; Fanselow & Dong, 2010). 3. Posterior and anterior aspects of the hippocampus have different neural contexts (Hargreaves et al., 2005; Kahn et al., 2008) that provide the sub-regions with access to different operations and information. Together with principle 2, this results in functional specialization in the posterior and anterior hippocampus. 4. The posterior hippocampus has access to detailed spatial/perceptual representations, item information, and stored episodic memories, whereas the anterior hippocampus has access to low-resolution spatial/perceptual representations and stored semantic information (Hargreaves et al., 2005; Kjelstrup et al., 2008). 5. Comparator operations in posterior and anterior hippocampus reflect the nature of the information available to each sub-region within its particular neural context. Thus, in the posterior hippocampus, a highresolution comparison is performed between incoming information and detailed information retrieved from episodic memory. In the anterior hippocampus, a low-resolution comparison is performed between incoming information and fuzzy spatial/perceptual information retrieved from semantic memory.

121 A novelty signal is transmitted to midbrain structures connected to the anterior hippocampus (including the substantia nigra and ventral tegmental area) when the anterior hippocampus comparator fails to match incoming perceptual information to existing semantic information. This operation is automatic and does not require executive resources. It may involve, for example, an evaluation of the speed with which information propagates from cortex relative to perceptual information transmitted from the brainstem (Vinogradova, 2001), a passive mechanism based on the idea that a cortical relay of perceptual input will proceed more slowly when no matching representation exists. A novelty signal, when generated in the anterior hippocampus, stimulates the release of dopamine from brainstem structures (Lisman & Grace, 2005). This pharmacological event has intrinsic reward properties, motivates exploratory behaviour (i.e., directing attention towards further evaluation of the novel information), and energizes the organism, leading to generalized memory enhancements over a period of minutes to hours (Düzel et al., 2010). An orienting response may also be triggered (Vinogradova, 2001). Elaborative processing of the novel stimulus (e.g., to derive an interpretation) is mediated by processing in PFC, ATL and anterior hippocampus. The presence or absence of a novelty signal may be used strategically during recognition (i.e., the discrimination heuristic; Gallo, Bell, Beier & Schacter, 2006). 7. Unlike the anterior hippocampus comparator, which operates on the basis of passive comparisons between perceptual information and semantic memory (principle 6), operations involving the posterior hippocampus are based on episodic pattern completion. Hippocampal pattern completion is thought to operate automatically on the basis of overlap between externally- or internally-generated patterns of activation in the neocortex and patterns previously indexed by the hippocampus (Eichenbaum, Yonelinas & Ranganath, 2007; Moscovitch et al., 1992, 2005; Norman and O'Reilly, 2003). In this way, the presence

122 109 or absence of ecphory can be interpreted as preliminary evidence of novelty. If pattern completion in the posterior hippocampus is not triggered, then the pattern is deemed novel in this initial, automatic stage of information processing. When time and executive resources are available, strategic memory search and monitoring processes mediated by interactions between PFC, parietal cortex and the posterior hippocampus (Moscovitch, 1992; Cabeza, Ciaramelli, Olson & Moscovitch, 2008) may be used to probe further the nature of overlap between current input and previous episodes, including direct comparison of particular stimulus features with memory (e.g., to detect subtle types of novelty; Ryan & Cohen, 2004) or a recall-to-reject strategy (Gallo et al., 2006). 8. Because novel materials are by definition dissimilar to previously exposed materials, they do not typically trigger hippocampal pattern completion. Accordingly, they may only be interpreted using stored knowledge about the nature of the world (i.e., from semantic memory). In contrast, familiar materials are likely to trigger hippocampal pattern completion because of high overlap with one or more earlier events (Moscovitch et al., 2005). When this happens, retrieval of stimulusspecific information gleaned from earlier exposures (i.e., from episodic or semantic memory) provides information that may facilitate performance on experimental tasks or integration of the stimulus with available contextual information and semantic knowledge. 9. The probability of successful encoding of information is a function of the extent to which information was processed in the first place (Craik & Lockhart, 1972). Together with the notion that novel materials trigger stimulus-focused processing (principle 6), whereas stimulus information is already available to facilitate task or contextual processing involving familiar materials (principle 8), this principle argues that contextual or task information will be better remembered for familiar relative to novel materials. In this way, the model

123 110 conforms to a hierarchical view of memory (e.g., Wickelgren, 1979), since it predicts that stable low-level (stimulus) representations are required before those representations may be bound together with higher-level (e.g., task or contextual) information. This model, as specified, accommodates the key empirical findings described earlier (see Fig. 13 for a schematic of the flow of information through the model as it has been described). Let us explore a subset of the possible inputs to the model to explore its predictions more fully. According to the model, all stimuli will be subjected to a comparator operation in the anterior hippocampus, the section of the hippocampus with access to semantic information. When no matching representation is found, the anterior hippocampus will trigger activation of the ventral tegmental area, generating an incremental but cumulative boost to the organism s energization and future probability of retaining long-term memories. Depending on available time and executive resources, the organism may choose to explore the stimulus further in hopes of accommodating it into the existing semantic memory framework, thereby initiating a processing loop between PFC, ATL and the anterior hippocampus. Alternatively, if incoming information were scored as a match by the anterior hippocampal comparator, no novelty signal would be generated and no further processing of the novel stimulus would be promoted. Instead, cognitive resources would typically be directed towards other goals, such as examining contextual information or responding to the demands of an experimental task. In the case of recently repeated stimuli that trigger episodic memory retrieval, ecphory serves as preliminary evidence of familiarity. Given sufficient time and executive resources, task or contextual processing can be undertaken. Alternatively, strategic memory search and monitoring processes may be initiated in order to detect potentially subtle differences between incoming information and information stored in memory. This processing would be expected to recruit PFC, parietal cortex and the posterior hippocampus (Moscovitch, 1992; Cabeza et al., 2008) and might be especially likely in the

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