Chemosensory sensations are often hard to describe and quantify. Language models may facilitate a systematic understanding of sensory descriptions. We accessed consumer and expert reviews of wine, perfume, and food products (English language; about 68 million words in total) and analyzed their sensory descriptions. Using a novel data-driven method based on natural language data, we compared the three chemosensory vocabularies (wine, perfume, food) with respect to their vocabulary overlap and semantic properties, and explored their semantic spaces. The three vocabularies primarily differ with respect to domain specificity, concreteness, descriptor type preference and degree of gustatory vs. olfactory association. Wine vocabulary primarily distinguishes between white wine and red wine flavors and qualities. Food vocabulary separates drinkable and edible food products and ingredients, on the one hand, and savory and non-savory products, on the other. A salient distinction in all three vocabularies is between concrete and abstract/evaluative terms. Valence also plays a role in the semantic spaces of all three vocabularies, but valence is less prominent here than in general olfactory vocabulary. Our method allows a systematic comparison of sensory descriptors in the three product domains and provides a data-driven approach to derive sensory lexicons that can be applied by sensory scientists.

A persistent belief holds that humans can imagine visual content but not odors. While visual imagery is regarded as recreating a perceptual representation, it is unknown whether olfactory mental imagery shares a perceptual format. Visual imagery studies have demonstrated this perceptual formatting using distance and shape similarity judgments, whereas olfactory studies often use single-odor vividness ratings, complicating the establishment of perceptual formatting for odors. Using odor pair similarity scores from two experiments (odor-based: 8,880 ratings from 37 participants, including 20 women; label-based: 129,472 ratings from 2,023 participants, including 1,164 women), we observed a strong correlation (r =.71) between odor-based and label-based odor pairs. The correlation was unaffected by gender and age and was present in a wide range of self-perceived olfactory functions. Pleasantness similarity was the main determinant of overall similarity for both odor-based (r=−.63) and label-based (r=−.45) odor pairs. We then used a large language model to derive semantic similarity scores for the labels of all odor pairs. Semantic similarity only mediated a small part of the observed correlation, further supporting our conclusions that odor imagery shares a perceptual formatting with vision, that odor percepts may be elicited from verbal labels alone, and that odor pair pleasantness may be a dominant and accessible feature in this regard.

Whereas human cognition develops through perceptually driven interactions with the environment, language models (LMs) are “disembodied learners” which might limit their usefulness as model systems. We evaluate the ability of LMs to recover sensory information from natural language, addressing a significant gap in cognitive science research literature. Our investigation is carried out through the sense of smell — olfaction — because it is severely underrepresented in natural language and thus poses a unique challenge for linguistic and cognitive modeling. By systematically evaluating the ability of three generations of LMs, including static word embedding models (Word2Vec, FastText), encoder-based models (BERT), and the decoder-based large LMs (LLMs; GPT-4o, Llama 3.1 among others), under nearly 200 training configurations, we investigate their proficiency in acquiring information to approximate human odor perception from textual data. As benchmarks for the performance of the LMs, we use three diverse experimental odor datasets including odor similarity ratings, imagined similarities of odor pairings from word labels, and odor-to-label ratings. The results reveal the possibility for LMs to accurately represent olfactory information, and describe the conditions under which this possibility is realized. Static, simpler models perform best in capturing odor-perceptual similarities under certain training configurations, while GPT-4o excels in simulating olfactory-semantic relationships, as suggested by its superior performance on datasets where the collected odor similarities are derived from word-based assessments. Our findings show that natural language encodes latent information regarding human olfactory information that is retrievable through text-based LMs to varying degrees. Our research shows promise for LMs to be useful tools in investigating the long debated relation between symbolic representations and perceptual experience in cognitive science.

Sensory expressions such as bright in bright light have traditionally been assumed to adhere to the Aristotetolian division of the five senses of sight, hearing, touch, taste and smell. Usages ofsensory terms outside of their primary sensory domain, such as in sweet melody where the taste term sweet describes a sound, have accordingly been seen as metaphorical and infrequent. However, recent research have questioned the notion of sensory metaphors, and findings from my own research indicate that the use of sensory terms in multiple sensory domains is pervasive. I suggest that sensory expressions often express multisensory meanings related to experiences of an emotional and bodily-internal nature (e.g., emotional evaluation, intensity or familiarity) that are shared between the senses. Investigating English sensory adjectives that are most strongly assiociated to the sensory domains in actual language use, I find that sensory adjectives of all sensory domains are highly multisensory, with one third of sight sensory adjectives and four fifths of smell adjectives being used in at least one more sensory domain. I also find that those sensory adjectives that are most multisensory, and thus describes sensory input in several sensory domains, express bodily-internal or interoceptive experiences to a higher degree. These findings indicate that sensory expressions reflect deeper underlying connections between the senses of an emotional nature. Further investigations of multisensory language may thus reveal how perceptual and cognitive-emotional processing is connected.

Parosmia is an olfactory disorder that involves distortions of specific odors that may co-occur with anosmia, loss of smell of other odors. Little is known about which odors frequently trigger parosmia, and measures of parosmia severity are lacking. Here, we present an approach to understand and diagnose parosmia that is based on semantic properties (e.g., valence) of words describing odor sources (“fish”, “coffee”, etc.). Using a data-driven method based on natural language data, we identified 38 odor descriptors. Descriptors were evenly dispersed across an olfactory-semantic space, which was based on key odor dimensions. Parosmia patients (n = 48) classified the corresponding odors in terms of whether they trigger parosmic or anosmic sensations. We investigated whether these classifications are related to semantic properties of the descriptors. Parosmic sensations were most often reported for words describing unpleasant odors of inedibles that are highly associated to olfaction (e.g., “excrement”). Based on PCA modeling, we derived the Parosmia Severity Index—a measure of parosmia severity that can be determined solely from our non-olfactory behavioral task. This index predicts olfactory-perceptual abilities, self-reported olfactory impairment, and depression. We thus provide a novel approach for investigating parosmia and establishing its severity that does not require odor exposure. Our work may enhance our understanding of how parosmia changes over time and how it is expressed differently across individuals.

Human olfaction can be extraordinarily sensitive, and its most common assessment method is odor identification (OID), where everyday odors are matched to word labels in a multiple-choice format. However, many older persons are unable to identify familiar odors, a deficit that is associated with the risk of future dementia and mortality. The underlying processes subserving OID in older adults are poorly understood. Here, we analyzed error patterns in OID to test whether errors could be explained by perceptual and/or semantic similarities among the response alternatives. We investigated the OID response patterns in a large, population-based sample of older adults in Sweden (n = 2479; age 60–100 years). Olfaction was assessed by a ‘Sniffin ́ TOM OID test with 16 odors; each trial involved matching a target odor to a correct label among three distractors. We analyzed the pattern of misidentifications, and the results showed that some distractors were more frequently selected than others, suggesting cognitive or perceptual factors may be present. Relatedly, we conducted a large online survey of older adults (n = 959, age 60–90 years) who were asked to imagine and rate the perceptual similarity of the target odors and the three corresponding distractors (e.g. “How similar are these smells: apple and mint?”). We then used data from the Swedish web corpus and the Word2Vec neural network algorithm to quantify the semantic association strength between the labels of each target odor and its three distractors. These data sources were used to predict odor identification errors. We found that the error patterns were partly explained by both the semantic similarity between target-distractor pairs, and the imagined perceptual similarity of the target-distractor pair. Both factors had, however, a diminished prediction in older ages, as responses became gradually less systematic. In sum, our results suggest that OID tests not only reflect olfactory perception, but also likely involve the mental processing of odor-semantic associations. This may be the reason why these tests are useful in predicting dementia onset. Our insights into olfactory-language interactions could be harnessed to develop new olfactory tests that are tailored for specific clinical purposes.

Grammatical encoding has been suggested to be driven by communicative efficiency - a balance between production ease and communicative success. Evidence for this view comes from studies indicating that speakers balance their use of morphosyntactic cues to grammatical functions with respect to animacy. However, these studies have not taken cues in the discourse context into account. In a picture-description task, we investigate the influence of animacy on the morphosyntactic encoding of grammatical functions in Swedish transitive sentences. These sentences are produced in discourse contexts with additional information about grammatical functions. We find various morphosyntactic cues to grammatical functions (e.g. SVO word order and case marking) to more frequently be used when the object referent is animate. Speakers thus balance their use of cues to grammatical functions, even when the discourse context is informative about those functions. These findings provide direct evidence for the view that grammatical encoding is influenced by communicative efficiency.

Odor identification is a common assessment of olfaction, and it is affected in a large number of diseases. Identification abilities decline with age, but little is known about whether there are perceptual odor features that can be used to predict identification. Here, we analyzed data from a large, population-based sample of 2,479 adults, aged 60 years or above, from the Swedish National study on Aging and Care in Kungsholmen. Participants performed both free and cued odor identification tests. In a separate experiment, we assessed perceived pleasantness, familiarity, intensity, and edibility of all odors in the first sample, and examined how odor identification performance is associated with these variables. The analysis showed that high-intensity odors are easier to identify than low-intensity odors overall, but also that they are more susceptible to the negative repercussions of old age. This result indicates that sensory decline is a major aspect of age-dependent odor identification impairment, and suggests a framework where identification likelihood is proportional to the perceived intensity of the odor. Additional analyses further showed that high-performing individuals can discriminate target odors from distractors along the pleasantness and edibility dimensions and that unpleasant and inedible odors show smaller age-related differences in identification. Altogether, these results may guide further development and optimization of brief and efficient odor identification tests as well as influence the design of odorous products targeted toward older consumers. 

The vocabulary for describing odors in English natural language is not well understood, as prior studies of odor descriptions have often relied on preselected descriptors and odor ratings. Here, we present a data-driven approach that automatically identifies English odor descriptors based on their degree of olfactory association, and derive their semantic organization from their distributions in natural texts, using a distributional-semantic language model. We identify 243 descriptors that are much more strongly associated with olfaction than English words in general. We then derive the semantic organization of these olfactory descriptors, and find that it is captured by four clusters that we name Offensive, Malodorous, Fragrant, and Edible. The semantic space derived from our model primarily differentiates descriptors in terms of pleasantness and edibility along which our four clusters are positioned, and is similar to a space derived from perceptual data. The semantic organization of odor vocabulary can thus be mapped using natural language data (e.g., online text), without the limitations of odor-perceptual data and preselected descriptors. Our method may thus facilitate research on olfaction, a sensory system known to often elude verbal description. 

A central component of sentence understanding is verb-argument interpretation, determining how the referents in the sentence are related to the events or states expressed by the verb. Previous work has found that comprehenders change their argument interpretations incrementally as the sentence unfolds, based on morphosyntactic (e.g., case, agreement), lexico-semantic (e.g., animacy, verb-argument fit), and discourse cues (e.g., givenness). However, it is still unknown whether these cues have a privileged role in language processing, or whether their effects on argument interpretation originate in implicit expectations based on the joint distribution of these cues with argument assignments experienced in previous language input. We compare the former, linguistic account against the latter, expectation-based account, using data from production and comprehension of transitive clauses in Swedish. Based on a large corpus of Swedish, we develop a rational (Bayesian) model of incremental argument interpretation. This model predicts the processing difficulty experienced at different points in the sentence as a function of the Bayesian surprise associated with changes in expectations over possible argument interpretations. We then test the model against reading times from a self-paced reading experiment on Swedish. We find Bayesian surprise to be a significant predictor of reading times, complementing effects of word surprisal. Bayesian surprise also captures the qualitative effects of morpho-syntactic and lexico-semantic cues. Additional model comparisons find that it—with a single degree of freedom—captures much, if not all, of the effects associated with these cues. This suggests that the effects of form- and meaning-based cues to argument interpretation are mediated through expectation-based processing.