Dissociation of Maintenance and Manipulation in Working Memory: A Functional MRI Study

Abstract

Working memory comprises functionally distinct processes: maintenance of information and manipulation of maintained representations. We used 3T fMRI with 32 participants performing a spatial n-back task with parametric increases in demand. Left dorsolateral prefrontal cortex and parietal regions showed load-dependent activation (p < .001, FWE). Medial prefrontal regions engaged preferentially during manipulation blocks. Effective connectivity analysis revealed task-dependent reorganization of prefrontal-parietal networks.

Introduction

Working memory comprises the capacity to transiently maintain and manipulate information needed for ongoing cognition. Neuroimaging investigations over the past decade have begun to delineate the neural substrates of these processes (Jonides et al., 2005; D'Esposito & Postle, 2009), yet the precise functional differentiation between maintenance and manipulation remains unclear. Current models suggest that dorsolateral prefrontal cortex (dlPFC) supports active maintenance via sustained elevation of neural firing rates, while posterior parietal regions may serve as a buffer for information storage. However, neuroimaging evidence for dynamic switching between maintenance and manipulation states within a single network architecture remains limited.

To address this gap, we designed a modified spatial n-back paradigm requiring participants to maintain sequences of spatial locations and, on separate blocks, to mentally rotate or manipulate those representations. This dissociation permits identification of neural regions selectively recruited by manipulation beyond domain-general load effects. We hypothesized that left dlPFC would show load-dependent activation regardless of manipulation demand, while regions including medial prefrontal cortex (mPFC) and posterior parietal cortex (PPC) would show preferential engagement during manipulation-heavy blocks.

Method

Participants

Thirty-two right-handed adults (14 female, age M = 26.3, SD = 4.1) with normal or corrected-to-normal vision and no history of neurological or psychiatric illness participated. All provided written informed consent. The study was approved by the university research ethics board. Participants received CAD $25 compensation.

Procedure

The experiment employed a 2 (maintenance vs. manipulation block) × 3 (load: n = 1, 2, 3) within-subjects design. Participants performed a spatial n-back task inside a 3T Siemens Magnetom scanner (8-channel head coil). In maintenance blocks, a sequence of spatial locations appeared (250 ms presentation, 1500 ms inter-stimulus interval), and participants indicated via button press whether the current location matched the location n steps back. In manipulation blocks, locations appeared identically but participants mentally rotated each location by 90° before performing the match judgment, increasing working memory demand beyond n-back load alone. Each condition comprised 12 trials per block, with 6 blocks per load level, yielding 108 trials total per condition. Block order was counterbalanced across participants.

fMRI data were acquired using a gradient-echo T2*-weighted echo-planar imaging sequence (TR = 2000 ms, TE = 30 ms, flip angle = 90°, 3 × 3 × 3 mm voxels, 42 slices). High-resolution T1-weighted structural images were acquired for anatomical reference (MPRAGE, 1 × 1 × 1 mm). FSL preprocessing pipeline was applied: motion correction (MCFLIRT), high-pass temporal filtering (100 s), and spatial smoothing (5 mm FWHM). Images were registered to standard MNI152 space using affine and nonlinear registration.

Results

Behavioral data revealed expected load effects: reaction time increased with n-back level (F(2,30) = 47.2, p < .001), and error rates were higher in manipulation conditions (F(1,31) = 36.8, p < .001). Accuracy remained above 75% across all conditions. Univariate fMRI analysis employed FSL's FEAT with task regressors convolved with the canonical hemodynamic response function. Left dlPFC (MNI coordinates -42, 30, 24; peak Z = 4.8) showed significant load × manipulation interaction (p < .001, family-wise error corrected at p < .05 using cluster thresholding at Z > 2.3). Left intraparietal sulcus (IPS, -30, -54, 42; Z = 4.3) displayed similar load effects. Critically, medial prefrontal cortex (mPFC, 6, 48, 18; Z = 3.9) showed selective engagement during manipulation blocks, independent of load (p = .002). Posterior cingulate cortex (PCC) activation was reduced during maintenance conditions (Z = -3.2), consistent with task-induced suppression of default-mode activity. Effective connectivity analysis using dynamic causal modeling revealed increased dlPFC → IPS coupling during maintenance-heavy blocks (F = 6.4, p = .016) and strengthened mPFC → dlPFC connectivity during manipulation blocks (F = 5.2, p = .029).

Discussion

This neuroimaging investigation provides functional dissociation between working memory maintenance and manipulation at both univariate and network levels. The selective engagement of mPFC during manipulation, independent of n-back load, suggests a specialized role for medial prefrontal regions in the executive transformation of working memory contents. The differential connectivity patterns—with dorsolateral-parietal coupling supporting maintenance and medial-dorsolateral coupling supporting manipulation—align with distributed-network models of executive function (Seeley et al., 2007) and extend previous findings from single-cell recording in nonhuman primates (Asaad et al., 2000).

The current findings have implications for understanding how frontal lobe pathology and age-related cognitive decline selectively impair manipulation while sometimes leaving maintenance relatively preserved (Daigneault & Braun, 1995). Future investigations should employ higher-field imaging (7T) and multimodal approaches combining fMRI with transcranial magnetic stimulation to establish causal relationships between regional activation and behavioral performance. Longitudinal designs tracking working memory network architecture across development may illuminate the neural basis of improvement in executive function during adolescence.

References

Asaad, W. F., Rainer, G., & Miller, E. K. (2000). Task-specific neural activity in the primate prefrontal cortex. Journal of Neurophysiology, 84(1), 451–459.
D'Esposito, M., & Postle, B. R. (2009). The cognitive neuroscience of working memory. Annual Review of Neuroscience, 18, 193–222.
Daigneault, S., & Braun, C. M. (1995). Working memory and the Post-Rolandic right hemisphere: Specific sequential resource demands for verbal and visuospatial working memory? Neuropsychology, 9(4), 566–579.
Jonides, J., Lewis, R. L., Nee, D. E., Lustig, C. A., Berman, M. G., & Moore, K. S. (2008). The mind and brain of short-term memory. Annual Review of Psychology, 59, 193–224.
Seeley, W. W., Menon, V., Schatzberg, A. F., Keller, J., Glover, G. H., Kenna, H., & Greicius, M. D. (2007). Dissociable intrinsic connectivity networks for salience processing and executive control. Journal of Neuroscience, 27(9), 2349–2356.
Rainer, G., Asaad, W. F., & Miller, E. K. (1998). Memory fields of neurons in the primate prefrontal cortex. Proceedings of the National Academy of Sciences, 95(25), 15008–15013.
Buckner, R. L., Andrews-Hanna, J. R., & Schacter, D. L. (2008). The brain's default network: Anatomy, function, and relevance to disease. Annals of the New York Academy of Sciences, 1124(1), 1–38.
Smith, E. E., & Jonides, J. (1999). Storage and executive processes in the frontal lobes. Science, 283(5408), 1657–1661.