The use of number words in natural language obeys Weber’s law.

Abstract

It has been suggested that the origins of number words can be traced back to an evolutionarily ancient approximate number system, which represents quantities on a compressed scale and complies with Weber's law. Here, we use a data-driven computational model, which learns to predict one event (a word in a text corpus) from associated events, to characterize verbal behavior relative to number words in natural language, without appeal to perception. We show that the way humans use number words in spontaneous language reliably depends on numerical ratio -a clear signature of Weber's law -thus perfectly mirroring the human and non-human psychophysical performance in comparative judgments of numbers. Most notably, the adherence to Weber's law is robustly replicated in a wide range of different languages. Together, these findings suggest that the everyday use of number words in language rests upon a pre-verbal approximate number system, which would thus affect the handling of numerical information not only at the input level but also at the level of verbal production.

Words with Consistent Diachronic Usage Patterns are Learned Earlier: A Computational Analysis Using Temporally Aligned Word Embeddings

“…Words with both high vector and neighborhood coherence, on the contrary, tend to refer to time‐related concepts, body parts, and numerals. The fact that we observe peculiar effects of language change (or better lack thereof) for number expressions in word embeddings is further evidence that language encodes important aspects of numbers and can inform the way we represent them cognitively (Rinaldi & Marelli, 2019). Our results show that number words tend to maintain coherent word representations over time, suggesting a central role in language, which should be further studied considering synchronic and diachronic patterns.…”

Language statistics as a window into mental representations

“…However, in the present article we show that there are serious limitations to this approach, demonstrating that a simple transposition between physical properties of the outside world and language statistics can break down completely (see also Rinaldi & Marelli, 2020b). We test as a prime example for such a breakdown the human body (and more specifically, the size of individual body parts), arguably one of the most salient visual stimuli we are exposed to throughout our lifetime.…”

“…Inspired by the work of Louwerse and colleagues (e.g. Louwerse, 2011), the researchers involved in this project have been working together for some time on the theoretical idea that natural language statistics -when properly examined through rigorous computational techniques as used in the field of natural language processing -offer a precious window into how the human brain represents experience with the external world (Günther et al, 2019;Rinaldi & Marelli, 2020a, 2020b. The present study examines as a prime example for this idea the human body, whose mental representation has long been the subject of Luca…”

Images of the unseen: Extrapolating visual representations for abstract and concrete words in a data-driven computational model

“…This is because language is not at all independent from the physical world we live in, but instead often used to communicate about this very world (Louwerse, 2011). This leads to statistical redundancies between the structure of the physical, directly-perceivable world on the one hand, and the structure of language on the other hand, so that relations between words tend to reflect the relations between their referents (Johns & Jones, 2012;Louwerse, 2011;Rinaldi & Marelli, 2020) -which in turn influences the training of distributional semantic models and the dimensional values of the resulting distributional vectors. The possibility of encoding grounded information through language usage makes it possible for our model to work, by capturing this information and systematically linking it to information from the visual domain.…”