Contrasting visual working memory for verbal and non-verbal material with multivariate analysis of fMRI

Habeck, Christian; Rakitin, Brian; Steffener, Jason; Stern, Yaakov · 2012 · Crossref

DOI: 10.1016/j.brainres.2012.05.045

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates the neural substrates of visual working memory, specifically contrasting the encoding and maintenance of verbal versus non-verbal material. Motivated by competing theoretical models—Baddeley’s model, which posits dedicated, modality-specific storage buffers, and Cowan’s embedded-process model, which suggests a generic attentional focus reactivating long-term memory—the authors sought to determine whether neural networks for storing information are material-specific or shared. The research aimed to clarify whether encoding and maintenance processes rely on distinct or overlapping brain regions depending on whether the stimuli are verbal (letters) or non-verbal (shapes). The researchers employed event-related functional magnetic resonance imaging (fMRI) in 25 young adults performing a Delayed-Item-Recognition task with unnamable nonsense line drawings ("Shape task"). Participants encoded 1, 2, or 3 shapes, retained them during a 7-second delay, and then identified a probe. To analyze the data, the authors used Ordinal Trend Canonical Variates Analysis (OrT CVA) to identify spatial covariance patterns showing monotonic increases with memory load. Crucially, they prospectively applied load-related patterns derived from a previously published verbal "Letter task" (using 1, 3, or 6 letters) to the current Shape task data. This allowed for a direct comparison of topographic composition and subject utilization between verbal and non-verbal neural networks across encoding (STIM), retention (RET), and probe (PROBE) phases. The results revealed a dissociation between encoding and maintenance processes. During the encoding phase, no significant relationship was found between the neural patterns for letters and shapes; their topographic compositions and subject utilizations were distinct, indicating that encoding is critically dependent on the specific material type. In contrast, during the retention phase, the neural patterns for letters and shapes were highly correlated in both topography and subject expression. Both patterns involved similar frontoparietal regions increasing in activity with load, alongside mediofrontal and temporal regions decreasing in activity. Furthermore, the expression of these shared retention patterns positively correlated with recognition accuracy in the Shape task. No significant load-related patterns were identified during the probe phase for the Shape task. These findings suggest that while encoding processes require material-specific neural substrates, the maintenance of information in working memory relies on a shared, generic neural network regardless of whether the material is verbal or non-verbal. This supports the existence of a modality-independent episodic buffer for rehearsal, challenging strict interpretations of Baddeley’s model that predict separate buffers for all phases. Instead, the data align with aspects of Cowan’s model regarding a unified attentional focus during maintenance, while confirming that initial encoding remains distinct based on stimulus type. The study demonstrates that multivariate analysis can effectively disentangle these phase-specific and material-specific neural contributions.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• working memory