Translation Emerges from Mind, Body, and Tools

For centuries, the act of translation has been viewed through a representational lens: the mind processes symbols internally to map one language onto another. A new study proposes a radical alternative, framing translation as an enacted activity where meaning emerges dynamically from the interplay of brain, body, tools, and context. This shift, based on the ABC framework—integrating affective, behavioral, and cognitive processes—suggests that translators don't merely manipulate static correspondences but participate skillfully in sociocultural practice, with their minds extending into the environment through real-time interactions.

The researchers found that translation unfolds through three interdependent layers: affective (feeling and emotion-related), behavioral (action and sensation-related), and cognitive (reflection and thought-related). These layers operate on different timelines but are hierarchically organized, as shown in Figure 1, forming a system where belief states guide the translator's coupling with their environment. Rather than storing symbolic representations, translators enact translation through sensorimotor engagement, where actions like typing and gazing shape perception and meaning. This model builds on predictive processing and enactive inference, emphasizing that the mind emerges from loops of brain-body-environment interactions, not as a fixed entity but as a dynamic pattern.

Ologically, the study draws on the Extended Mind theory and radical enactivism, rejecting classical representationalist views. It uses the ABC framework to categorize translation processes, with data illustrated through progression graphs from eye-tracking and keystroke logs. For example, Figures 3 and 4 show two translators rendering an English sentence into Japanese, revealing different behavioral policies: one as a 'head starter' with linear translation and short orientation phases, and another as a 'large-context planner' with reverse-order translation and longer initial orientation. The researchers model this using an enactive-inference architecture, depicted in Figure 2, where internal states interact with external states via sensory and active states, forming closed perception-action loops. This approach allows simulation of translation strategies, such as entropy reduction in word-order choices, as detailed in Table 2.

Analysis indicates that affective states modulate precision weighting, influencing how translators handle uncertainty. In the example, affective beliefs like confidence or anxiety determine whether prediction errors are treated as noise or prompt re-evaluation, affecting behavioral patterns such as pauses or resource consultation. The data shows that translators with different styles exhibit distinct policy cycles: the head starter engages in immediate typing with more revisions, while the large-context planner delays action for higher certainty, as evidenced by hesitation states in Figure 4. The study also highlights how orientation, hesitation, and revision states reflect dynamic reconfigurations of the ABC layers, with affective cues guiding transitions between these states to maintain workflow coherence.

This research has significant for understanding translation as a holistic practice, moving beyond technical skill to include emotional and embodied dimensions. It suggests that tools like CAT software and cultural norms are not mere aids but constitutive parts of the cognitive system, affecting performance in real-world settings such as legal or literary translation. For regular readers, this means translation is more akin to a dance between person and environment, where factors like stress or confidence directly impact output quality, potentially improving training s by emphasizing affective regulation and environmental design.

Limitations of the study include its focus on basic modeling, as it does not fully capture the integration of affective, behavioral, and cognitive dynamics in complex real-world translation. The framework's falsifiability criteria, such as precision-modulation claims and entropy-reduction dynamics, require further empirical validation. Additionally, the example simplifies lexical selection, focusing only on word-order choices, which may not represent the full ambiguity translators face. The researchers acknowledge that more work is needed to test layer interactions and apply the ABC model to diverse translation contexts.