The hippocampus operates in a threshold manner during spatial source memory Scott D. Slotnick a and Preston P. Thakral b

Transcription

1 Cognitive neuroscience and neuropsychology 265 The hippocampus operates in a threshold manner during spatial source memory Scott D. Slotnick a and Preston P. Thakral b Long-term memory can be based on general familiarity or detailed recollection. Although familiarity is thought to be a continuous/graded process, the nature of recollection is currently under debate. In the present functional MRI spatial source memory study, we evaluated the pattern of activity in the hippocampus to assess whether this region operates in a threshold/all-or-none or a continuous manner during recollection. During the study phase, abstract shapes were presented to the left or right of fixation. During the test phase, old and new shapes were presented at fixation, and participants classified each shape as old-left, old-right, or new, followed by a sure unsure confidence rating. Accurate spatial memory for old-left shapes produced a single activation in the left hippocampus. The corresponding event-related activation profile revealed a threshold above which old-left-sure responses produced positive activity for old-left but not old-right shapes. This hippocampal activation profile was used to generate a source memory receiver operating characteristic that was adequately fit by a threshold model of recollection but was not adequately fit by a continuous model of recollection. By contrast, there was no evidence of a threshold in the behavioral response profile, which is consistent with previous behavioral source memory receiver operating characteristic results indicating that recollection is a continuous process. The present results indicate that the hippocampus can operate in a threshold manner during spatial source memory and further suggest that this discrete signal is transformed into a continuous process through the operation of other brain regions that also contribute to behavioral performance. NeuroReport 24: c 213 Wolters Kluwer Health Lippincott Williams & Wilkins. NeuroReport 213, 24: Keywords: continuous model, functional magnetic resonance imaging, receiver operating characteristic, recollection, threshold model a Department of Psychology, Boston College, Chestnut Hill, Massachusetts and b Center for Vital Longevity, School of Behavioral and Brain Sciences, University of Texas at Dallas, Dallas, Texas, USA Correspondence to Scott D. Slotnick, PhD, Department of Psychology, Boston College, Chestnut Hill, MA 2467, USA Tel: ; fax: ; sd.slotnick@bc.edu Received 24 December 212 accepted 18 January 213 Introduction Long-term memory can be based on general familiarity or detailed recollection. Although familiarity is widely thought to be a continuous/graded process, the nature of recollection has recently been a source of contentious debate. Some believe recollection, like familiarity, is a continuous process [1], whereas others believe recollection is a threshold/all-or-none process [2]. The threshold and continuous models of recollection have been distinguished by evaluating the shape of the source/ context memory receiver operating characteristic (ROC): a plot of hit rates versus false alarm rates. Of importance, source memory for items with similar levels of familiarity is assumed to be mediated by recollection alone. In such paradigms, the threshold recollection model predicts a linear source memory ROC, whereas the continuous recollection model predicts a curved source memory ROC [1 5]. Although an initial behavioral study suggested that source memory ROCs were linear [3], a large body of subsequent behavioral evidence has accumulated indicating that they are curved [4 1] but can appear linear if nonmemorial Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal s Website ( information is included in the analysis [5]. These behavioral results provide strong evidence in support of the continuous model of recollection. In the present functional MRI study, we evaluated the shape of the source memory ROC generated from the pattern of activity in the hippocampus, given that this region has been linked to recollection on the basis of functional MRI evidence [11 14] and lesion evidence [15 18]. In this way, we aimed to assess whether the hippocampus operates in a continuous manner during recollection, as indicated by behavioral evidence, operates in a threshold manner. This is the first time, to our knowledge, that the hippocampal activation profile has been used to distinguish between the threshold and continuous models of recollection. Materials and methods Twelve right-handed participants with normal or corrected-to-normal visual acuity (seven women, age years) provided informed and written consent. The Massachusetts General Hospital Institutional Review Board approved the protocol. Each participant completed six study test phases (Fig. 1). Each shape was constructed by connecting four c 213 Wolters Kluwer Health Lippincott Williams & Wilkins DOI: 1.197/WNR.b13e32835f282d

2 266 NeuroReport 213, Vol 24 No 5 Bézier curves along each side of a 5.51 visual angle square and filling it with colored oriented lines [19]. During the study phase, 32 shapes were presented 31 of visual angle to the left or right of fixation (Fig. 1a). Each shape was displayed for 2.5 s followed by.5 s of fixation. Participants were instructed to remember each shape and its location. During the test phase, the old shapes and 16 new shapes were presented in a random order at fixation (Fig. 1b). Each shape was displayed for 2.5 s followed by s of fixation. Participants classified each shape as old-left, oldright, or new followed by a sure unsure confidence rating. During the study and test phases, no more than three shapes of a given event type were sequentially presented and participants were instructed to maintain fixation. Oldleft, old-right, and new shapes were counterbalanced across participants using a Latin square design. Imaging data were acquired using a 3 T Siemens Allegra scanner (Siemens AG, Erlangen, Germany). Anatomic images were acquired with a multiplanar, rapidly-acquired gradient echo sequence [3 ms time of repeat (TR), 3.3 ms time of echo (TE), 128 slices, mm resolution] andfunctionalimageswereacquiredwithanechoplanar Fig. 1 Study phase + + Test phase Old-left-sure Old-right-unsure During the study phase, abstract shapes were presented to the left or right of fixation. During the test phase, old and new shapes were presented at fixation, and participants classified each item as old-left, old-right, or new, followed by a sure unsure confidence rating (possible responses are shown to the right). imaging sequence (2 s TR, 3 ms TE, acquisition matrix, 3 slices, 4.5 mm isotropic resolution). Analysis was carried out using BrainVoyager QX (Brain Innovation B.V., Maastricht, the Netherlands). Functional images were slice-time corrected, motion corrected (runs > 3 mm excluded), temporally filtered (linear trends and frequencies < 3 cycles/run removed), and Talairach transformed. A random-effect general linear model analysis was carried out. Models of activation were constructed by convolving a canonical hemodynamic response with each event protocol (active from stimulus onset until the behavioral response or, if no response was given, stimulus offset). Event types consisted of encoding shape and location, accurate item and source memory (old-left-hit-hits, old-right-hit-hits), accurate item memory alone (old-left-hit-misses, old-righthit-misses), complete misses, new false alarms, new correct rejections, no responses, and a constant (collapsed over sure unsure responses). The conjunction of old-left-hithit > old-left-hit-miss [11 14,2] and old-left-hit-hit > old-right-hit-hit [21 22] was used to isolate accurate old-left shape activity (i.e. old-left-correct > baseline; baseline refers to both old-left-hit-miss and old-right-hit-hit events). The analogous conjunction was used to isolate accurate old-right shape activity. Although such comparisons have been described as measuring memory for object place associations [17 18], the term source memory is used to maintain consistency with the ROC literature [1 1]. On the basis of established anatomic criteria [12 14], each activation was confined to the hippocampus. Note that we did not expect hippocampal activity to be lateralized in any systematic way. An individual voxel P-value of less than.1 (t = 4.28) was enforced to ensure false discovery rate correction for multiple comparisons to less than.5. After identifying each hippocampal activation using the conjunction, a P-value of less than.1 (t = 3.9) was enforced for display and all contiguous significant voxels served as a region of interest for the subsequent time course analysis. Event-related activation time courses from 2 to 12 s after stimulus onset were extracted and baseline corrected from 2 to s. The eight event types included in the analysis reflected increasing memory strength for old-left shapes and decreasing memory strength for oldright shapes (Fig. 2b). To ensure that activation magnitudes were greater than or equal to corresponding to a lower boundary on neural firing at spikes/s the minimum activation magnitude across all event types was subtracted from each activation time course. For each event, the mean activity 6 s after stimulus onset (the time point of maximum old-left-hit-hit and old-right-hithit amplitudes) was used to carry out the analysis. To compute activation probabilities as a function of memory strength, the magnitude of each stimulus/ response was divided by the sum of all magnitudes for that stimulus type (Fig. 2b; each old-left shape activation

3 Threshold recollection in the hippocampus Slotnick and Thakral 267 Fig. 2 Signal change (%) Signal change (%) Source memory activity (old-left-correct > baseline) t = 4.28 (c) 1.9 Hit rate Hippocampal activation profile Old-left shapes Hippocampal source memory ROC t = 3.9 Response (increasing memory strength) Old-right shapes Response (decreasing memory strength) Threshold recollection.1 Continuous recollection False alarm rate Source memory activity (old-left-correct > baseline) in the left hippocampus is circled (coronal view, key to the right). Event-related activity (mean±1 SE) extracted from the hippocampal activation above. For old-left shapes and old-right shapes, respectively, memory strength increases and decreases to the right. The dotted lines demarcate a threshold above which old-left-sure activity is greater than for old-left but not old-right shapes. (c) Source memory ROC (circles) generated from the hippocampal activation profile. Best-fit threshold and continuous model ROCs are shown by solid and dotted lines, respectively (key at the bottom). The chance line is illustrated along the diagonal. ROC, receiver operating characteristic. magnitude was divided by the sum of all old-left shape activation magnitudes). Hit rates were then computed by cumulating the probabilities from the highest to lowest source memory strength for that stimulus type (e.g. old-left), and false alarm rates were computed by cumulating the probabilities from lowest to highest source memory strength for the other stimulus type (e.g. old-right) [5]. The hippocampal source memory ROC was generated by plotting these hit rates versus false alarm rates. The corresponding numbers of behavioral responses were used to compute the behavioral source memory ROC. The threshold recollection model (parameter R) and the continuous recollection model (parameter d ) were each fit to the hippocampal source memory ROC and the behavioral source memory ROC by adjusting model parameters using maximum-likelihood estimation. These single-parameter models were used as the sources in such paradigms have similar levels of familiarity [4,5], which predicts a symmetrical ROC about the diagonal. A threshold continuous model (parameters R and d [1]) was also fit to both ROCs to evaluate whether an additional parameter would improve the fit over that of a single-parameter model (which was possible as the other models were nested within this model). Log-likelihood w 2 was used to assess the adequacy of each model with lower w 2 -values reflecting a better fit (P >.5 indicates an adequate fit). Results The source memory comparison of old-left-correct versus baseline produced a single activation in the tail of the left hippocampus (Fig. 2a; most significant Talairach coordinate x = 13, y = 38, z = 3; cluster size 162 mm 3 ). The source memory comparison of old-right-correct versus baseline produced a single activation in the body of the left hippocampus (Supplemental digital content; Fig. 1; but this activation did not survive correction for multiple comparisons and thus was not analyzed further. These distinct spatial memory activations are reminiscent of place cells in the rat hippocampus [23]. The old-left-correct versus baseline activation profile revealed a threshold between old-left-unsure and oldleft-sure responses (Fig. 2b, dotted lines) above which old-left shapes (t = 16.6, P <.1) but not old-right shapes (t = 1.5, P =.16) produced positive activity. All other event types produced positive activity (minimum t = 6.37, P <.1). This pattern of activity indicates that the hippocampus can operate in a threshold manner. The hippocampal ROC (Fig. 2c) was linear in shape and adequately fit by the threshold model (w 2 =3.73,P =.15; R =.18), whereas the continuous model did not provide an adequate fit (w 2 =6.91, P <.5; d =.64). The threshold continuous model provided an adequate fit

4 268 NeuroReport 213, Vol 24 No 5 Fig. 3 Number of responses Number of responses Hit rate Behavioral response profile Old-left shapes Response (increasing memory strength) Old-right shapes Old-rightsurunsurunsursure Old-right- Old-left- Old-left- Response (decreasing memory strength) Behavioral source memory ROC Threshold recollection Continuous recollection False alarm rate Source memory behavioral response profile (mean±1 SE). Source memory ROC (circles) generated from the behavioral response profile with best-fit threshold and continuous model ROCs (key at the bottom). The chance line is illustrated along the diagonal. ROC, receiver operating characteristic. (w 2 = 1.36, P >.2; R =.25, d =.38) that did not differ from that of the threshold model (w 2 =2.37, P =.12) but was better than that of the continuous model (w 2 =5.54, P <.5). The latter comparisons discount the possibility of the threshold and continuous w 2 -values being just below and above the statistical threshold value and indicate that a threshold process is necessary to account for the linear shape of the hippocampal ROC. Unlike the activation profile of the hippocampus, there was no threshold in the behavioral response profile (Fig. 3a), as old-left-sure responses were associated with positive activity for both old-left shapes (t = 5.57, P <.1) and old-right shapes (t = 3.5, P <.1). All other event types were also associated with positive values (old-right-sure responses to old-left shapes, t = 3.22, P <.1, other responses, minimum t = 5.32, P <.1). The corresponding behavioral source memory ROC (Fig. 3b) appeared to be more curved than the hippocampal ROC, which suggests that behavioral source memory is not a threshold process. Although the behavioral response profile deviated from that predicted by the threshold model, the behavioral ROC was adequately fit by both the threshold model (w 2 = 2.4, P >.2; R =.47) and the continuous model (w 2 = 2.76, P >.2; d = 1.53). The threshold continuous model also adequately fit the behavioral ROC (w 2 = 1.54, P >.2; parameters R =.43, d =.26), and this fit did not differ from that of the threshold model (w 2 <1)or the continuous model (w 2 = 1.22, P >.2). Discussion In the present study, both the hippocampal activation profile and the hippocampal ROC indicated that this region can operate in a threshold (all-or-none) manner. Specifically, the hippocampal activation profile revealed a threshold above which old-left-hit-hit-sure but not oldright-hit-miss-sure event types produced positive activity, and the linear hippocampal ROC was only fit by the threshold model. The pattern of behavioral results was distinct from the pattern of hippocampal results. Unlike the activation profile of the hippocampus, there was no evidence for a threshold in the behavioral response profile. Further, the continuous model did not adequately fit the hippocampal ROC (P <.5) but adequately fit the behavioral ROC (P >.2). The threshold continuous model provided a better fit than the continuous model for the hippocampal ROC (P <.5), but not the behavioral ROC (P >.2). The d -value of the threshold continuous model was negative for the hippocampal ROC (d =.38), which is opposite to that predicted by the continuous model of recollection [1 5], and was positive for the behavioral ROC (d =.26). Thus, the current behavioral findings are consistent with a large body of behavioral ROC evidence indicating that source memory is a continuous process [4 1]. The present results suggest that the hippocampus can operate in a threshold manner during source memory, whereas the present and previous results indicate that the behavioral responses reflect a continuous process. Although seemingly at odds, it is important to keep in mind that many neural regions beyond the hippocampus mediate visual spatial source memory including the striate and extrastriate cortex (BA17/18/19/37), the parahippocampal cortex, the parietal cortex, and the prefrontal cortex [2,24,25]. Thus, even if the hippocampus operates in a threshold manner, many other brain

5 Threshold recollection in the hippocampus Slotnick and Thakral 269 regions, which may operate in a threshold or a continuous manner, also contribute to any given behavioral response. Indeed, a comparison of the hippocampal ROC memory strength parameters (R =.18, d =.64) and the behavioral ROC memory strength parameters (R =.47, d = 1.53) reveals that less than half of the source memory signal can be attributed to the hippocampus. Future work will be needed to evaluate the nature of processing in different neural regions during source memory. Conclusion The present results indicate that the hippocampus operates in a threshold manner during spatial source memory. By contrast, we observed no evidence of a threshold process contributing to behavioral performance. These findings suggest that the discrete signal in the hippocampus is transformed into a continuous signal through the operation of other brain regions that also contribute to behavior. Acknowledgements Conflicts of interest There are no conflicts of interest. References 1 Wixted JT. Dual-process theory and signal-detection theory of recognition memory. Psychol Rev 27; 114: Yonelinas AP, Parks CM. Receiver operating characteristics (ROCs) in recognition memory: a review. Psychol Bull 27; 133: Yonelinas AP. The contribution of recollection and familiarity to recognition and source-memory judgments: a formal dual-process model and an analysis of receiver operating characteristics. J Exp Psychol Learn Mem Cogn 1999; 25: Slotnick SD, Klein SA, Dodson CS, Shimamura AP. An analysis of signal detection and threshold models of source memory. J Exp Psychol Learn Mem Cogn 2; 26: Slotnick SD, Dodson CS. Support for a continuous (single-process) model of recognition memory and source memory. Mem Cognit 25; 33: Qin J, Raye CL, Johnson MK, Mitchell KJ. Source ROCs are (typically) curvilinear: comment on Yonelinas (1999). J Exp Psychol Learn Mem Cogn 21; 27: Hilford A, Glanzer M, Kim K, DeCarlo LT. Regularities of source recognition: ROC analysis. J Exp Psychol Gen 22; 131: Glanzer M, Hilford A, Kim K. Six regularities of source recognition. JExp Psychol Learn Mem Cogn 24; 3: Dodson CS, Bawa S, Slotnick SD. Aging, source memory, and misrecollections. J Exp Psychol Learn Mem Cogn 27; 33: Slotnick SD. Remember source memory ROCs indicate recollection is a continuous process. Memory 21; 18: Diana RA, Yonelinas AP, Ranganath C. Imaging recollection and familiarity in the medial temporal lobe: a three-component model. Trends Cogn Sci 27; 11: Ross RS, Slotnick SD. The hippocampus is preferentially associated with memory for spatial context. J Cogn Neurosci 28; 2: Slotnick SD. Does the hippocampus mediate objective binding or subjective remembering? Neuroimage 21; 49: Slotnick SD. Controversies in cognitive neuroscience. Basingstoke, UK: Palgrave Macmillan; Thaiss L, Petrides M. Source versus content memory in patients with a unilateral frontal cortex or a temporal lobe excision. Brain 23; 126: Thaiss L, Petrides M. Autobiographical memory of the recent past following frontal cortex or temporal lobe excisions. Eur J Neurosci 28; 28: Parkinson JK, Murray EA, Mishkin M. A selective mnemonic role for the hippocampus in monkeys: memory for the location of objects. J Neurosci 1988; 8: Bachevalier J, Nemanic S. Memory for spatial location and object-place associations are differently processed by the hippocampal formation, parahippocampal areas TH/TF and perirhinal cortex. Hippocampus 28; 18: Slotnick SD, Schacter DL. A sensory signature that distinguishes true from false memories. Nat Neurosci 24; 7: Thakral PP, Slotnick SD, Schacter DL. Conscious processing during retrieval can occur in early and late visual regions. Neuropsychologia 213; 51: Slotnick SD. Memory for color reactivates color processing region. Neuroreport 29; 2: Slotnick SD. Rapid retinotopic reactivation during spatial memory. Brain Res 29; 1268: O Keefe J, Dostrovsky J. The hippocampus as a spatial map. Preliminary evidence from unit activity in the freely-moving rat. Brain Res 1971; 34: Slotnick SD, Moo LR, Segal JB, Hart J Jr. Distinct prefrontal cortex activity associated with item memory and source memory for visual shapes. Brain Res Cogn Brain Res 23; 17: Mitchell KJ, Johnson MK. Source monitoring 15 years later: what have we learned from fmri about the neural mechanisms of source memory? Psychol Bull 29; 135: