COGMED: a database for Chinese olfactory and gustatory metaphor

The study examines how concrete sensory domains of smell and taste are used to express abstract meanings, within the frameworks of Conceptual Metaphor Theory and embodied cognition. Prior work indicates that sensory metaphors can shape judgments and activate sensory brain areas, yet behavioral and neuroimaging evidence for olfactory and gustatory metaphors in Mandarin Chinese remains limited and inconsistent, partly due to inadequate control of stimulus properties and parts of speech. The authors aim to build and validate a comprehensive Mandarin database of olfactory and gustatory expressions, categorized by metaphorical versus literal use and by adjective, noun, and verb, and normed on seven psycholinguistic dimensions (familiarity, meaningfulness, figurativeness, valence, difficulty, imageability, naturalness). The goal is to provide rigorously controlled materials to support cognitive, behavioral, and neuroimaging research on sensory metaphors.

Conceptual Metaphor Theory posits that metaphors map concrete source domains (e.g., senses) to abstract target domains (e.g., morality, social status). Embodied cognition suggests abstract concepts derive from bodily experiences. Behavioral studies have linked taste/smell experiences to abstract judgments (e.g., sweet taste influencing judgments about revenge; cleanliness and moral judgments) and distinguished conceptual vs. idiomatic metaphors. Neuroimaging shows sensory metaphors can recruit modality-specific cortices (e.g., somatosensory activation for texture metaphors; gustatory/olfactory cortex engagement), though findings vary with stimulus types and control, including part-of-speech differences and familiarity. The literature also debates hemispheric contributions to metaphor processing, with inconsistent evidence for right-hemisphere specificity and a recognized need for careful stimulus control and psycholinguistic norming (familiarity, valence, aptness, etc.). The lack of a dedicated Mandarin database for olfactory/gustatory metaphors, and cultural variability in odor/taste semantics, motivates the present resource, which focuses on quality terms (e.g., fragrant, sour) rather than concrete object labels (e.g., cinnamon).

Participants: 200 Mandarin-speaking students were recruited (41 males, 159 females; mean age 18.42±0.86). Inclusion criteria required good physical/mental health and no history of neurological injury or olfactory/gustatory dysfunction. Ethics approval: Southern Medical University (No. 2022-17); informed consent obtained.

Materials and item construction: Ten core single characters capturing olfaction and gustation in Mandarin were chosen as source domains: olfaction 香 (fragrant), 臭 (smelly), 腥 (fishy), 膻 (muttony), 臊 (urine odor); gustation 酸 (sour), 甜 (sweet), 苦 (bitter), 咸 (salty), 鲜 (umami), and 辣 (spicy). Vocabulary related to these senses was collected from authoritative dictionaries (Xinhua Dictionary, Modern Chinese Dictionary, Shuowen Jiezi) and the CCL corpus, supplemented by web searches. Ninety-seven words were identified (38 olfactory, 59 gustatory).

Stimulus normalization: Expressions (phrases/sentences/idioms) containing target terms were extracted from CCL and the Internet, then standardized to ensure consistent grammar and collocations across conditions and parts of speech. Formats: adjectives as X+Y (two words; e.g., 香艳的花朵), nouns as X是Y (three words; e.g., 鼻子是器官), verbs as X v Y (three words; e.g., 警犬嗅包裹). Minimal adjustments were made to ensure uniformity.

Screening: From initial pools (>400 items each for olfaction and gustation), 10 psychology students flagged incomprehensible/unfamiliar entries; such items were removed. Final item counts: olfaction 319 (adjectives 245: 145 metaphorical, 100 literal; verbs 52: 27 MI, 25 LI; nouns 22: 15 MI, 7 LI); gustation 352 (adjectives 270: 178 MI, 92 LI; verbs 30: 17 MI, 13 LI; nouns 52: 34 MI, 18 LI).

Procedure: Two alternating surveys (olfactory, gustatory) were administered in pen-and-paper format. Participants self-rated their smell and taste sensitivity (1–10) and then rated each item on seven 5-point dimensions: familiarity (1 very unfamiliar–5 very familiar), meaningfulness (1 difficult–5 easy to understand), figurativeness (1 very literal–5 highly figurative), valence (1 negative–5 positive), difficulty (1 very hard to interpret–5 very easy), imageability (1 no image–5 clear image), and naturalness (1 very unnatural–5 very natural). No time limit; typical completion time 45–90 minutes. Missing data were followed up with participants when possible.

Data screening and statistics: Of 199 completers (97 olfaction, 102 gustation), one olfactory and two gustatory participants with valence ratings negatively correlated with item means were excluded; one gustatory participant with poor inter-rater correlation (r ≤ 0.10) was also excluded, leaving N=195 (olfaction n=96; gustation n=99). Analyses used SPSS 20.0. For each dimension, inter-rater reliability (Cronbach’s α) was computed; Spearman correlations examined inter-dimensional relations by sense and part of speech; effects of gender, age, and self-perceived sensory ability were tested. Differences between metaphorical and literal items were assessed with paired t-tests or Wilcoxon signed-rank tests depending on normality/outliers (Shapiro–Wilk p<0.05).

Sample and descriptive norms: After exclusions, 96 participants rated 319 olfactory items and 99 participants rated 352 gustatory items. Overall means (SD) across items: Olfaction—Familiarity 3.42 (0.72), Meaningfulness 4.14 (0.53), Figurativeness 2.29 (0.85), Valence 2.84 (1.02), Difficulty 4.18 (0.45), Imageability 3.96 (0.52), Naturalness 3.93 (0.58). Gustation—Familiarity 3.63 (0.56), Meaningfulness 4.26 (0.39), Figurativeness 2.25 (0.73), Valence 2.87 (0.71), Difficulty 4.27 (0.34), Imageability 4.06 (0.37), Naturalness 4.08 (0.47).

Reliability: Inter-rater reliability was very high. Olfaction α: Familiarity 0.99, Meaningfulness 0.99, Figurativeness 0.99, Valence 0.92, Difficulty 0.99, Imageability 0.99, Naturalness 0.98 (mean 0.98, SD 0.03). Gustation α: 0.99, 0.99, 0.99, 0.96, 0.99, 0.99, 0.98 respectively (mean 0.99, SD 0.01).

Inter-dimensional correlations: Strong positive correlations emerged among Familiarity, Meaningfulness, Difficulty (ease), Imageability, and Naturalness in both senses (olfaction: r≥0.86; gustation: 0.76≤r≤0.93, all p<0.001). Figurativeness correlated negatively with these dimensions. Olfaction: Figurativeness vs. Familiarity r=-0.57, Meaningfulness r=-0.60, Difficulty r=-0.71, Imageability r=-0.66, Naturalness r=-0.62 (all p<0.001). Gustation: Figurativeness vs. Meaningfulness r=-0.19 (p<0.001), Valence r=-0.15 (p=0.006), Difficulty r=-0.28, Imageability r=-0.23, Naturalness r=-0.11 (p=0.036).

Metaphorical vs. literal comparisons: Generally, metaphorical items were rated as more figurative, while literal items were more familiar, meaningful, imageable, natural, and easier to comprehend. For adjectives, large differences were found (e.g., olfaction: figurativeness t=-8.572, p<0.001; difficulty t=-7.920, p<0.001; familiarity t=-6.743, p<0.001). Gustatory adjectives similarly differed (figurativess t=-8.066, p<0.001; difficulty t=-4.710, p<0.001; meaningfulness t=-3.049, p=0.002). Some valence differences were observed (e.g., gustatory adjectives t=-2.243, p=0.025). Differences for verbs and nouns were less consistent, likely reflecting smaller item counts.

Demographics: No significant effects of gender, age, or self-perceived sensory acuity on ratings (ps>0.05).

The project delivers COGMED, a rigorously controlled and highly reliable database of Mandarin olfactory and gustatory expressions. The strong internal consistency and expected correlation patterns among familiarity, meaningfulness, ease, imageability, and naturalness corroborate prior work, while the negative relations with figurativeness align with the notion that more literal items are easier and more natural to process. Clear differences between metaphorical and literal items—particularly in adjectives—validate the database’s capacity to dissociate figurativeness from other psycholinguistic properties, facilitating matched-stimulus designs.

By standardizing expressions across parts of speech and controlling keyword positions and length, COGMED addresses known confounds in metaphor research and supports diverse methodologies (ERP/EEG/MEG, eye-tracking, fMRI/TMS). The POS segmentation is especially relevant for neurocognitive studies given evidence that lexical categories may differentially engage sensory and language networks. The resource thus advances investigations into how olfactory/gustatory experiences ground abstract meaning and enables cross-study comparability through normed dimensions.

COGMED provides normative ratings on familiarity, meaningfulness, figurativeness, valence, difficulty, imageability, and naturalness for 671 Mandarin olfactory and gustatory items spanning adjectives, verbs, and nouns. The database demonstrates high inter-rater reliability and expected inter-dimensional relations, making it suitable for behavioral, corpus-based, and neurocognitive studies of sensory metaphors. The authors anticipate its use for controlled stimulus selection, manipulation of psycholinguistic variables (e.g., familiarity/novelty), and examination of POS-specific processing mechanisms, thereby promoting research on the cognitive and neural bases of olfactory and gustatory metaphors.

The database includes fewer verb and noun items than adjectives, producing imbalances across parts of speech and potentially reducing power and stability of correlations in those subsets. Some inconsistencies between ratings and means for verb and nominal expressions may stem from these smaller item pools. Valence showed comparatively lower (though still high) inter-rater reliability. While expressions were standardized to minimal forms for control, this may limit contextual richness unless researchers add context in downstream studies.