The article, authored by Manuela Macedonia of Johannes Kepler University Linz together with Lihui Hu and Brian Mathias of the University of Aberdeen, introduces what the authors call the Gesture Memory Support Model, or GMSM. Rather than offering yet another theory of embodied cognition, the framework shifts the level of analysis. Instead of asking whether gesture enrichment helps learning in general, it asks which properties of a specific gesture, in relation to a specific word and a specific group of learners, are likely to make that gesture memorable. The authors identify seven dimensions hypothesized to shape a gesture’s mnemonic contribution: iconicity, sensorimotor imagery potential, motional salience, emotional salience fit, gesture complexity fit, cultural familiarity, and enactment potential.

The theoretical foundation rests on embodied accounts of cognition, which hold that conceptual knowledge is grounded in sensory and motor experience rather than stored as purely abstract symbols. Neuroimaging studies have shown, for example, that reading action verbs such as “kick” or “lick” activates somatotopically organized regions of motor cortex corresponding to leg and mouth movements. When a new word is learned together with a meaningful gesture, the resulting lexical representation appears to be genuinely multimodal: words trained with gestures later engage sensory and motor brain regions during recognition even when the gestures themselves are no longer present. Causal evidence goes further, indicating that visual and motor cortices contribute to the later translation of vocabulary acquired through sensorimotor-enriched training. In other words, gesture enrichment may alter the functional representation of a newly learned word rather than merely serving as an attentional aid during encoding.

The first and most extensively studied dimension is iconicity, the degree to which a gesture visually resembles or symbolizes salient semantic properties of its referent. The evidence here is striking. In one study, learners of an artificial vocabulary corpus retained significantly more words paired with iconic, semantically congruent gestures than words paired with meaningless self-directed movements. Functional magnetic resonance imaging revealed that iconic gestures recruited broader motor networks, whereas meaningless gestures additionally activated a cognitive control network resembling that seen in Stroop-like interference tasks, suggesting that learners expect at least some semantic congruency between a word and its gesture. Electroencephalography studies using Stroop-like paradigms have found that incongruent gesture-speech pairings elicit larger N400 responses and slow reaction times, indicating automatic semantic integration and a processing cost when gesture and speech conflict. Importantly, the authors caution that iconicity is a matter of degree rather than a simple yes-or-no category, and that even arbitrary gestures can still enhance memorability relative to no gesture at all.

Two further dimensions capture different facets of a gesture’s expressive force. Motional salience refers to the perceptual prominence of the movement itself, how spatially extensive, dynamic, or vigorous it is, which plausibly modulates visual attention and the probability that gestural information is encoded. Emotional salience fit, by contrast, concerns whether the affective tone conveyed by the gesture and, where visible, the facial expression is appropriate to the meaning of the target word. The authors are careful to separate these constructs, because they can co-occur without logically entailing one another. A vivid example comes from item-level analyses of a classroom training study: the pseudoword “nunun,” meaning seal, was paired with a playful whole-body gesture of balancing an imaginary ball on the nose, and it became the most memorable word in the entire corpus, accounting for 9.59 percent of all retrieved items. Yet another animal word, “strattin” for bat, was paired with an equally conspicuous flight-mimicking gesture and was remembered far less well, illustrating that movement dynamics alone cannot explain a memory advantage and that affective and semantic associations of the referent may contribute.

The remaining dimensions extend the analysis to the learner and the context. Sensorimotor imagery potential captures the extent to which a gesture provides bodily, action-related, or tactile cues that may elicit internal simulation of the concept, a property that can be high even when visual resemblance is low, as in metaphorical gestures for abstract words. Gesture complexity fit asks whether a gesture’s observable motoric demands, its number of movement phases, limb involvement, and coordination requirements, are appropriate for the target meaning and learner population, with the authors explicitly rejecting the assumption that simpler is always better. Cultural familiarity acknowledges that gestures are embedded in sociocultural systems of meaning: the thumbs-up sign conveys approval in many Western contexts but may carry different or even offensive meanings elsewhere, and studies show that culturally unfamiliar emblems can reduce perceptual accuracy and social evaluations. Finally, enactment potential characterizes how readily a specified learner population could accurately reproduce a gesture, a relational property that is distinct from whether learners actually perform the gesture during instruction.

On the question of performance versus observation, the review delivers a nuanced verdict. The enactment effect, the well-established memory advantage for material that learners physically perform, is substantial, with a meta-analysis reporting an average effect of g = 1.23 in episodic memory. However, a systematic review and meta-analysis of gesture-supported foreign language learning, based on seven studies and 309 participants, found that observing gestures can be as effective as performing them for several vocabulary outcomes, including free recall and cued recognition. This convergence is consistent with embodied accounts proposing that action observation recruits perceptual and motor representations similar to those engaged during execution, while active production additionally supplies motor commands, proprioceptive feedback, and kinaesthetic experience. The authors also flag temporal synchrony between gesture and speech as a critical implementation condition rather than a gesture property: experiments with animated speakers showed that gesture strokes delayed by 500 milliseconds or omitted altogether reduced word recall, whereas strokes slightly preceding the stressed syllable were as effective as fully synchronized ones.

Operationally, the GMSM proposes that trained human raters score each gesture on all seven dimensions using five-point ordinal scales with dimension-specific anchors, ideally from standardized video recordings, and code a dimension as not applicable when reliable evaluation is impossible. Crucially, the ratings are retained as a multidimensional profile rather than collapsed into a single effectiveness score, because the dimensions are not assumed to be statistically independent or to contribute additively to learning. Worked examples illustrate the value of this approach: miming drinking from a cup earns high ratings across nearly every dimension, an arbitrary finger sequence scores low on most scales while still fitting the neutral emotional tone of the word, and outlining a roof shape for “house” yields a heterogeneous profile combining maximal iconicity with only moderate sensorimotor imagery potential and low motional salience.

The authors are explicit that the GMSM is at present a framework for characterization and hypothesis generation, not a validated diagnostic or predictive instrument. Its reliability, the empirical distinctiveness of its dimensions, and their individual and interactive relationships with learning outcomes all remain open questions. Future studies could use the dimension ratings as item-level predictors in regression or mixed-effects models, test whether sensorimotor imagery potential explains variance beyond iconicity, examine whether enactment potential moderates the advantage of production over observation, and explore whether multimodal artificial intelligence systems can assist large-scale gesture characterization, provided such computational ratings are validated against human judgments. For teachers and curriculum developers, the framework offers no simple recipe, but it does something arguably more valuable: it makes the design of instructional gestures a theoretically informed decision rather than an afterthought, treating the rich variability of human movement as a meaningful determinant of how well new words take root in memory.

Subject of Research: Gesture-supported second language vocabulary learning and the characterization of instructional gestures

Article Title: Why Some Gestures Work Better than Others: Toward a Model of Gesture-Supported Vocabulary Learning

Article References: Macedonia, M., Hu, L., & Mathias, B. (2026). Why Some Gestures Work Better than Others: Toward a Model of Gesture-Supported Vocabulary Learning. Educational Psychology Review, 38(1), Article 129. https://doi.org/10.1007/s10648-026-10222-8

Image Credits: AI Generated

DOI: 10.1007/s10648-026-10222-8

Keywords: gestures, vocabulary learning, second language acquisition, embodied cognition, memory, iconicity, enactment effect, cognitive load, educational neuroscience, multimodal learning, Gesture Memory Support Model, instructional design