pith. sign in

arxiv: 2605.22616 · v2 · pith:KZRN4NN2new · submitted 2026-05-21 · 💻 cs.CL

Chinese sensorimotor and embodiment norms for 3,000 lexicalized concepts

Jing Chen , G\'abor Parti , Yin Zhong , Chu-Ren Huang , Marco Marelli This is my paper

Pith reviewed 2026-05-22 05:57 UTC · model grok-4.3

classification 💻 cs.CL
keywords sensorimotor normsembodiment ratingsMandarin Chineselexical decisionrepresentational similarity analysisdistributional semanticsembodied cognitionlexical processing
0
0 comments X

The pith

Sensorimotor ratings for 3,000 Mandarin concepts predict lexical decision speed and recover from text alone.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper builds a database of how 3,000 concepts in Mandarin Chinese register across eleven senses plus an overall embodiment score, gathered from 378 native speakers. The ratings hold up well in consistency checks and match earlier smaller Chinese collections. When tested against real-time word recognition, two ways of combining the ratings stand out as the best predictors of faster decisions for more embodied words. The same ratings can also be estimated with decent accuracy from statistical patterns in language use, though visual and sound dimensions come through more clearly than taste or smell.

Core claim

A novel normative database supplies 11-dimensional sensorimotor ratings and unidimensional embodiment ratings for 3,000 lexicalized Mandarin concepts, obtained from 378 native speakers. These ratings exhibit high reliability and cross-norm validity with prior Chinese resources. In lexical decision validation, the PSE-Sensorimotor composite and Minkowski-3 metric emerge as the strongest predictors of processing speed. Sensorimotor ratings prove substantially recoverable from purely linguistic representations through regression, yielding a mean Spearman correlation of .62 across dimensions, with visual and auditory dimensions showing higher recoverability than chemosensory ones; the relational

What carries the argument

Eleven sensorimotor dimensions (visual, auditory, haptic, olfactory, gustatory, interoceptive, and others) plus a unidimensional embodiment rating, aggregated into composites such as Perceptual Strength of Embodiment (PSE) to quantify grounding effects on lexical access.

If this is right

  • PSE-Sensorimotor and Minkowski-3 composites best capture how sensorimotor information speeds lexical decisions.
  • Simple regression recovers sensorimotor ratings from linguistic data at mean Spearman r = .62.
  • Visual and auditory dimensions recover more faithfully from language than chemosensory dimensions.
  • The geometry of the sensorimotor space is partially preserved in distributional patterns (r = .540).

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Text statistics appear to encode enough embodied structure for partial simulation of sensorimotor knowledge in language models.
  • Weaker recovery of taste and smell suggests inherent limits in what distributional data can capture about chemical senses.
  • The resource opens direct tests of embodied cognition theories in a major non-Indo-European language.
  • These norms could serve as training targets for models that aim to ground Chinese lexical representations in simulated perceptual experience.

Load-bearing premise

The ratings collected from 378 speakers accurately reflect embodied grounding for the broader Mandarin population and causally shape lexical processing rather than merely correlating with other linguistic properties.

What would settle it

A replication in which new Mandarin speakers produce ratings that correlate below .60 with the reported set on multiple dimensions, or in which the PSE-Sensorimotor and Minkowski-3 composites lose all predictive power for lexical decision times after full statistical control for word frequency and length.

Figures

Figures reproduced from arXiv: 2605.22616 by Chu-Ren Huang, G\'abor Parti, Jing Chen, Marco Marelli, Yin Zhong.

Figure 1
Figure 1. Figure 1: An example of the survey questions for a random target word: [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Sensorimotor and embodiment profiles of six example words. Each radar plot shows ratings [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Modality dominance and exclusivity categorized by part-of-speech. (a) the proportion of [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Correlation matrix of the 11 sensorimotor dimensions and embodiment. Color intensity [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Distribution of human and predicted sensorimotor ratings across 11 dimensions. Each [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Representational dissimilarity matrices (RDMs) for human sensorimotor ratings (left) and [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Relationship between cross-validated R2 and Spearman ρ across 11 sensorimotor dimen￾sions. Each point represents one dimension: circles denote sensory dimensions and triangles for motor. Error bars show 95% bootstrap confidence intervals. non-dominant dimensions (e.g., Minkowski-3 and PSE-Sensorimotor), compared to undifferentiated composites (e.g., Summed strength) or single-dimension metrics (e.g., Maxim… view at source ↗
read the original abstract

Understanding how conceptual knowledge is grounded in bodily experience, and to what extent machine systems can acquire such knowledge without direct sensorimotor experience, are central questions in both cognitive science and embodied artificial intelligence research. Large-scale normative resources are essential for investigating these questions empirically, yet such resources remain sparse for non-Indo-European languages. We present a novel normative database for 3,000 lexicalized concepts in Mandarin Chinese, comprising 11-dimensional sensorimotor ratings and unidimensional embodiment ratings collected from 378 native Mandarin speakers. The ratings demonstrate high reliability and strong cross-norm validity with existing Chinese resources, each of which covers fewer words and a subset of the 11 sensorimotor dimensions. In a validation study, we tested new variables derived from a theoretically motivated metric, Perceptual Strength of Embodiment (PSE) (Huang et al., 2025), together with seven common composite variables, on lexical decision tasks. The results suggest that PSE-Sensorimotor and Minkowski-3 are the strongest composite predictors of lexical decision performance, capturing the facilitatory effects of sensorimotor information on lexical processing. A further exploratory study showed that sensorimotor ratings are substantially recoverable from purely linguistic representations using simple regression models (mean Spearman r = .62 across dimensions), though recovery varied markedly: visual and auditory dimensions yielded higher correspondence than chemosensory ones. Representational similarity analysis further showed that the relational geometry of the sensorimotor space is also partially recoverable (r = .540), consistent with the view that distributional language use encodes aspects of embodied conceptual structure.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The paper presents a novel normative database for 3,000 lexicalized concepts in Mandarin Chinese, comprising 11-dimensional sensorimotor ratings and unidimensional embodiment ratings collected from 378 native Mandarin speakers. It reports high reliability (split-half and Cronbach's alpha), strong cross-norm validity with existing Chinese resources, predictive utility of derived metrics including PSE-Sensorimotor and Minkowski-3 in lexical decision tasks, and substantial recoverability of the ratings from linguistic representations via regression (mean Spearman r = .62), with representational similarity analysis showing partial recovery of relational geometry (r = .540).

Significance. If the results hold, this work supplies a large-scale, publicly useful resource that addresses the scarcity of sensorimotor norms for non-Indo-European languages. The lexical-decision validation and linguistic-recovery analyses provide concrete evidence that sensorimotor information facilitates lexical processing and that distributional language statistics encode aspects of embodied structure, with dimension-specific variation (stronger for visual/auditory than chemosensory). These contributions are likely to support downstream modeling in cognitive science and embodied AI.

minor comments (3)
  1. [§3.2] §3.2: The seven common composite variables are referenced but not enumerated in a single location; a short table or explicit list would improve readability.
  2. [Figure 3] Figure 3 (recovery correlations): axis labels and dimension abbreviations are not fully expanded in the caption, making it difficult to map visual/auditory vs. chemosensory results without cross-referencing the methods.
  3. [Table 4] Table 4 (lexical decision regressions): the exact set of linguistic covariates included alongside the sensorimotor composites is not stated in the table note, although the text indicates standard controls were used.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary of our work and for recommending minor revision. We are pleased that the contributions of the normative database, its reliability, cross-norm validity, lexical-decision validation, and linguistic recoverability analyses are recognized as addressing an important gap for non-Indo-European languages.

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper's core contributions consist of newly collected sensorimotor and embodiment ratings from 378 native speakers for 3,000 concepts, along with reliability analyses, cross-norm validations, lexical decision validation studies, and exploratory regression-based recovery analyses from linguistic representations. These elements rely on primary data collection and standard statistical procedures rather than reducing to self-citations or fitted inputs by construction. The reference to the PSE metric from Huang et al. (2025) introduces a composite variable for validation but does not underpin the primary empirical findings or recovery results, which remain independent of that prior definition. No steps in the reported derivation chain exhibit self-definitional, fitted-prediction, or load-bearing self-citation patterns that would render the claims equivalent to their inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work rests on standard psycholinguistic assumptions about the validity of speaker ratings rather than on fitted parameters or new postulated entities.

axioms (1)
  • domain assumption Ratings collected from native Mandarin speakers provide valid measures of sensorimotor and embodiment properties of concepts
    Invoked when treating the collected norms as ground truth for validation and recovery analyses.

pith-pipeline@v0.9.0 · 5814 in / 1287 out tokens · 46046 ms · 2026-05-22T05:57:21.547146+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.