Individual differences in working memory performance as a function of the local integrity and regional connectivity of the hippocampus

Very proud of my research assistants(first and second authors) putting this one together!


Kommers, C.1,+, Raccah, O.1,+, Reggente, N.,1 Rissman, J1,2

Although the hippocampus is well known to contribute to the storage and retrieval of long-term memories, emerging data suggests that the hippocampus may also contribute to the online maintenance of task-relevant representations in some tests of working memory. To the degree that hippocampal mechanisms serve to facilitate performance on short delay memory tasks, individual differences in hippocampal microstructure could contribute to across-subject variance in working memory performance. To examine the relationship between hippocampal structure and function, we obtained the diffusion-weighted images (DWI) of a large cohort of subjects from the Human Connectome Project MRI dataset. We used the DWI to compute diffusion tensor images (DTI), which in turn were used to generate whole-brain mean-diffusivity (MD) maps. MD in deep gray matter has been construed as an indirect measurement of local microstructural deficits (Kim et al., 2013). Thereby, we aimed to assess the underlying integrity of each subject’s hippocampal gray matter and use examine whether these measures can account for variance in memory performance across subjects. Hippocampal regions of interest (ROIs) were identified using Freesurfer’s automated segmentation algorithm. Average MD within the left hippocampus was found to be significantly correlated with performance on a Working Memory List Sorting Task. This result is consistent with prior work showing that hippocampal MD serves a predictor for verbal and visuospatial memory (Carlesimo et al., 2010). This current study extends these previous findings and contributes to the debate surrounding the role of the hippocampus in working memory. We plan to conduct further analyses aimed at characterizing the potentially important role of fronto-hippocampal connectivity in working memory performance.

+ These authors contributed equally to this effort.

1 – Dept. of Psychology, University of California, Los Angeles

2 – Dept. of Psychiatry & Biobehavioral Sciences, University of California, Los Angeles




Kim, Hengjun J., et al. “Alterations of mean diffusivity in brain white matter and deep gray matter in Parkinson’s disease.” Neuroscience letters 550 (2013): 64-68.


Carlesimo, Giovanni A., et al. “Hippocampal mean diffusivity and memory in healthy elderly individuals: A cross-sectional study.” Neurology 74.3 (2010): 194-200.


Characterizing common and dissociable involvement of medial temporal lobe regions during episodic source memory retrieval and analogical reasoning

Westphal, A.J., Reggente, N., Ito, K., Fortuna, W.H., Nawabi, Y., Milstein, M., & Rissman, J.


 SfN 2013 Abstract

Episodic memory and analogical reasoning tasks tend to engage many common frontoparietal structures, perhaps owing to their common demands for declarative memory retrieval and relational integration. Regions of the medial temporal lobe (MTL), well known to play a critical role in the encoding and retrieval of episodic memories, have also been shown to contribute to relational reasoning. We aimed to expand upon these findings by performing a direct comparison of memory- and reasoning-related MTL activity profiles and assessing how these regions communicate with distinct cortical networks to support different task demands. We examined fMRI activity and functional connectivity (FC) of the hippocampus (HIP), parahippocampal cortex (PHC), and perirhinal cortex (PRC) in a novel experimental paradigm featuring closely matched memory and reasoning tasks, both requiring judgments on 4-word stimulus arrays. One day prior to fMRI scanning, subjects (N = 20) encoded 80 words under two different mental imagery conditions. During the scanned memory task, subjects were to identify the word they previously studied and specify the encoding context, if possible. During the analogical reasoning task, subjects were to assess if the top and bottom word pairs shared the same semantic relationship or else indicate the number of non-analogous semantic relationships. Univariate parameter estimates extracted from HIP, PHC, and PRC all showed greater activity for source retrieval versus item familiarity. Activity in the PRC was significantly greater for correct versus incorrect source judgments; this effect also trended in HIP and PHC. During the reasoning task, HIP and PHC showed significantly greater activation on trials with valid analogies than on trials with no semantic relationships, whereas PRC activated strongly during all reasoning task conditions where semantic relationships were present. Task-dependent FC contrasting reasoning and memory was analyzed using psychophysiological interactions analysis. Left HIP demonstrated preferential coupling with both default mode and cognitive control network (CCN) structures for memory and bilateral MTL and lateral temporal regions for reasoning. Left PHC showed preferential coupling with CCN structures for memory and the supramarginal gyrus for reasoning. Left PRC demonstrated stronger coupling with precuneus for memory and occipital structures for reasoning. Taken together, these results confirm prior findings of MTL involvement in episodic source retrieval, while also documenting putative MTL contributions to analogical reasoning and distinct profiles of cortical network coupling across task sets.

Decoding cognitive task-sets from rostral prefrontal cortex functional connectivity patterns

Westphal, A.J., Reggente, N., Ito, K., Fortuna, W., & Rissman, J.


 HBM 2013 Abstract

Resting state fMRI connectivity analyses have identified a number of distinct functional brain networks, including the fronto-parietal task control network (FPTCN), the dorsal attention network (DAN), and the default mode network (DMN) (e.g., Vincent et al., 2008, Power et al., 2011). While these networks are typically defined based on intrinsically correlated BOLD fluctuations during periods of undirected thought, engagement of these networks is also observed during goal-oriented cognition. For instance, the FPTCN has been shown to co-activate with the DMN to facilitate internally-focused mentation and with the DAN to promote externally-focused attention (Spreng et al., 2010). In the present investigation, we sought to evaluate the degree to which task-set representations, particularly those requiring relational integration such as analogical reasoning and episodic memory retrieval, could be decoded from functional connectivity patterns within and between these networks. We were most interested in examining the representational content of connections originating in the rostral prefrontal cortex (RPFC), since RPFC may play a key role in relational integration, in addition to supporting the maintenance of superordinate goal-states (e.g., Badre & D’Esposito, 2009).

20 subjects healthy adult subjects underwent fMRI scanning (3T Siemens Trim Trio scanner, TR = 2 s, voxel size = 3 x 3 x 3.7 mm), performing alternating blocks of analogical reasoning, episodic source memory retrieval, and visuospatial attention tasks. These tasks were closely matched for reaction times, response demands, and bottom-up visual stimulus processing (all trials involved 4-word arrays, with the tasks only differing in what subjects had to decide about these words). Our data analysis procedure involved calculating the pairwise correlations between the concatenated BOLD time-courses for each task for each of 264 functional areas (10 mm spheres, identified by Power et al., 2011). We then supplied a regularized logistic regression classification algorithm with the full connectivity matrix from a given network (within-network connectivity) or from the set of connections that linked a pair of networks (between-network connectivity). All classification analyses used a leave-one-subject-out procedure, such that the classifier was trained on the connectivity data from 19 of 20 subjects and then applied to predict the task-sets associated with the remaining connectivity matrices from the held-out subject.

Using correlations between all 264 nodes, our classifier was 100% accurate at differentiating between the three cognitive task-sets. When trained solely on the correlations between the 16 RPFC nodes, the classifier was unable to differentiate between the reasoning and memory task-sets, indicating that within-RPFC connectivity patterns are not necessarily diagnostic of task-set. However, when trained on the correlations between RPFC nodes and nodes outside of RPFC, classification accuracy was quite robust (Fig. 1), reaching accuracy levels of up to 85% depending on which network was paired with RPFC. This result provides novel evidence that RPFC flexibly adjusts its interactivity with all three of the core networks to facilitate both internally and externally-oriented cognition.

By measuring the pattern of correlations between distinct nodes in a subject’s brain, one can reliably decode information about that subject’s cognitive task-set, even when a classifier has not been trained on data from that subject. The connection strengths between RPFC nodes and nodes in other core brain networks can be used to predict whether a subject is engaged in analogical reasoning or episodic source memory retrieval, despite the common demands of these tasks for relational integration. Given its position at the apex of a rostral-caudal hierarchy (Badre & D’Esposito, 2009), these data suggest that RPFC may differentially collaborate with posterior networks depending on task goals.



Vincent, J. L., Kahn, I., Snyder, A. Z., Raichle, M. E., & Buckner, R. L. (2008). Evidence for a frontoparietal control system revealed by intrinsic functional connectivity. Journal of neurophysiology100(6), 3328-3342.

Spreng, R. N., Stevens, W. D., Chamberlain, J. P., Gilmore, A. W., & Schacter, D. L. (2010). Default network activity, coupled with the frontoparietal control network, supports goal-directed cognition. Neuroimage53(1), 303-317.

Badre, D., & D’Esposito, M. (2009). Is the rostro-caudal axis of the frontal lobe hierarchical?. Nature Reviews Neuroscience10(9), 659-669.

Power, J. D., Cohen, A. L., Nelson, S. M., Wig, G. S., Barnes, K. A., Church, J. A., … & Petersen, S. E. (2011). Functional network organization of the human brain. Neuron72(4), 665-678.

Cho, S., Moody, T.D., Fernandino, L., Mumford, J.A., Poldrack, R.A., Cannon, T.D., Knowlton, B.J., & Holyoak, K.J. (2010). Common and Dissociable Prefrontal Loci Associated with Component Mechanisms of Analogical Reasoning. Cerebral Cortex, 20(3),524-533.

Shared and distinct contributions of rostral prefrontal cortex to analogical reasoning and episodic memory retrieval: Insights from fMRI functional connectivity and multivariate pattern analyses

Westphal, A.J., Reggente, N., Nawabi, Y., & Rissman, J.


CNS 2013 Abstract


The rostral prefrontal cortex (RPFC), positioned at the apex of the prefrontal processing hierarchy, has been implicated in a diverse array of high-level cognitive processes including analogical reasoning and episodic memory retrieval—tasks that may share demands for relational integration. However, because reasoning and memory tasks have not been compared in the same studies, the degree of neuroanatomical overlap is unclear. To address this gap, we developed an fMRI paradigm that required subjects to periodically shift between Reasoning, Memory, and Perception tasks, closely matched for response demands, reaction times, and bottom-up stimulus processing. On all trials, participants were presented with an array of four words, with the cognitive operations to be performed on this array specified by a task set cue provided at the beginning of each block. Although RPFC regions showed highly overlapping recruitment during successfully solved analogy and source memory retrieval trials, without significant univariate differences, multi-voxel pattern analysis identified areas of RPFC wherein local activity patterns could facilitate robust decoding of these trial types. One such prominent cluster in left lateral RPFC was then seeded in a psychophysiological interaction analysis. Strikingly, this region showed divergent profiles of functional connectivity across task blocks, coupling more strongly with frontoparietal control network structures during Reasoning and with default mode network structures during Memory. These findings suggest that common areas of RPFC may differentially contribute to analogical reasoning and episodic retrieval via their coordinated interactions with distinct brain networks that respectively facilitate the integration of complex semantic or episodic relationships.