Working Memory Training Across the Senses: A Systematic Review

Technical Summary: Working Memory Training Across the Senses

Problem Statement
Working memory (WM) training is a prevalent cognitive intervention aimed at enhancing cognitive abilities, often under the theoretical assumption that WM operates as a domain-general, modality-independent system. Current literature extensively investigates "near" and "far" transfer effects across different cognitive domains (e.g., from training tasks to fluid intelligence) within a single sensory modality. However, a critical gap remains regarding cross-modal transfer (CMT): the extent to which WM training in one sensory modality (e.g., vision) improves performance on analogous tasks in a different modality (e.g., audition). While often assumed that WM benefits generalize across sensory channels, this assumption is rarely empirically tested. Furthermore, the sensory modalities engaged in training and testing are frequently treated as incidental rather than as core design variables, potentially limiting the generalizability of findings to real-world, multisensory environments.

Methodology
The authors conducted a systematic review following PRISMA guidelines to address three objectives: (1) identify which sensory modalities are targeted in WM training paradigms; (2) examine whether studies have investigated CMT effects; and (3) evaluate the extent of support for CMT in the literature.

• Search and Selection: The review searched PubMed, Web of Science, and PsycINFO (as of November 13, 2024), supplemented by expert knowledge. Inclusion criteria required peer-reviewed empirical studies with adult human participants, a WM training intervention with pre- and post-training assessments, and explicit descriptions of the sensory modalities used.
• Data Extraction: From 1,772 initial records, 342 studies met the inclusion criteria. Studies were categorized based on the sensory modalities (visual, auditory, tactile, olfactory) engaged during both training and testing phases.
• CMT Analysis: Only studies with designs allowing for the isolation of modality-specific effects (i.e., training in one modality with testing in another, or crossed-modality training groups) were analyzed for CMT. A narrative synthesis and descriptive comparison of effect sizes were performed, as a formal meta-analysis was not feasible due to heterogeneity in outcome measures and reporting.
• Task Classification: WM tasks were classified into four functional families (maintenance, manipulation, updating, complex span) based on Oberauer et al.'s (2018) framework.

Key Results

• Sensory Modality Bias: The vast majority of studies (99.1%) relied exclusively on visual and/or auditory stimuli. Visual-only designs were the most common (31.0%), followed by mixed visual-auditory designs. Non-visual/auditory modalities (tactile, olfactory) were extremely rare.
• Scarcity of CMT Designs: Only seven studies (2.0%) out of the 342 included possessed designs suitable for evaluating cross-modal transfer.
• Inconsistent Transfer Patterns: The seven studies examining CMT yielded mixed and often asymmetrical results:
  • Evidence for CMT: Some studies reported transfer from visual to auditory (Kimura & Matsuura, 2020; Zhu et al., 2024), tactile to visual (Guo et al., 2021), and olfactory to visual (Olofsson et al., 2020).
  • Evidence for Modality-Specificity: Other studies found improvements confined strictly to the trained modality with no transfer to analogous tasks in other senses (Schneiders et al., 2012; Buschkuehl et al., 2014).
  • Asymmetry: Where transfer occurred, it was often asymmetric. For instance, Olofsson et al. (2020) found olfactory-to-visual transfer but not visual-to-olfactory; Zhu et al. (2024) found visual-to-auditory transfer but not auditory-to-visual.
• Task Types: Manipulation and updating paradigms (particularly n-back tasks) dominated the literature, while maintenance and complex span tasks were less common as training interventions.

Significance and Claims
The paper argues that the assumption of modality-independent WM plasticity is largely untested and potentially inaccurate. The findings suggest that:

1. CMT is not guaranteed: Unlike the assumption that strengthening central control mechanisms automatically generalizes across senses, the evidence indicates that training benefits may be constrained by the specific sensory channel.
2. Orthogonal Dimensions: The authors propose that WM transfer should be conceptualized along two orthogonal dimensions: cognitive domain (traditional near/far transfer) and sensory modality. A program might improve visual attention (near transfer within vision) but fail to improve the ability to remember spoken numbers (no transfer to audition).
3. Limitations of Current Theory: The scarcity of appropriate experimental designs and the inconsistency of available findings suggest that claims about the broad utility of WM training in multisensory environments rest on a narrow empirical basis. The literature's heavy reliance on visual and auditory stimuli limits the ability to generalize conclusions to WM as a general system.
4. Future Directions: The authors call for future research to treat sensory modality as a critical variable, requiring systematic manipulation, clear reporting of stimulus formats, and explicit testing of CMT to determine whether WM training produces genuinely broad benefits or effects tied to specific sensory systems.

The review concludes that while cross-modal transfer can occur, it is inconsistent and often asymmetric, challenging the view of WM as a purely modality-independent system and highlighting the need for more rigorous, modality-sensitive research designs.