This Title All WIREs
How to cite this WIREs title:
WIREs Cogn Sci
Impact Factor: 2.218

The Sapir‐Whorf hypothesis and inference under uncertainty

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

The Sapir‐Whorf hypothesis holds that human thought is shaped by language, leading speakers of different languages to think differently. This hypothesis has sparked both enthusiasm and controversy, but despite its prominence it has only occasionally been addressed in computational terms. Recent developments support a view of the Sapir‐Whorf hypothesis in terms of probabilistic inference. This view may resolve some of the controversy surrounding the Sapir‐Whorf hypothesis, and may help to normalize the hypothesis by linking it to established principles that also explain other phenomena. On this view, effects of language on nonlinguistic cognition or perception reflect standard principles of inference under uncertainty.

A stimulus S produces two cues c 1, c 2. From these cues, we obtain an estimate Ŝ of the original stimulus.
[ Normal View | Magnified View ]
Left: English (top panel) and Berinmo (bottom panel) color naming systems, both mapped against a standard color naming grid in which lightness varies by row and hue varies by column. Each false‐colored region represents the extension of a named color category. The white rectangles denote ranges of colors for which Roberson et al. collected color memory data; these are the same ranges for the two languages. (Adapted with permission from Ref . Copyright 1999 Nature Publishing Group). Right: Empirical color memory performance by Berinmo and English speakers for various color pairs, compared with the fit of category adjustment models based on native‐language color naming. Model fits are range‐matched to the empirical data. (Reprinted from Ref .)
[ Normal View | Magnified View ]
A category adjustment model accounts for bias patterns in color memory. In each panel, the horizontal axis denotes the hue of a target color that participants saw, and the vertical axis denotes bias in memory: positive values indicate that the color in question was remembered as being a color further to the left along the horizontal axis than it actually was, and negative values indicate that the color was remembered as being a color further to the right. Colored dots denote empirical data, and black circles denote estimates provided by a category adjustment model based on English color naming. Comparison of the upper and lower panels reveals that delay (and thus greater uncertainty) produces greater category‐congruent bias. (Reprinted from Ref .)
[ Normal View | Magnified View ]
A category adjustment model applied to the Sapir‐Whorf hypothesis in the domain of color. An observed stimulus is encoded in two ways: (1) a fine‐grained representation of the stimulus itself, shown as a (gray) distribution over stimulus space centered at the stimulus’ location in that space, and (2) the language‐specific category (e.g., English ‘green’) in which the stimulus falls, shown as a separate (green) distribution over the same space, centered at the category prototype. The stimulus is reconstructed by combining these two sources of information through probabilistic inference, resulting in a reconstruction of the stimulus (black distribution) that is biased toward the category prototype. The amount of bias is determined by the uncertainty of the fine‐grained representation. (Reprinted from Ref .)
[ Normal View | Magnified View ]
Probabilistic cue integration, weighted by cue certainty. Probability densities for haptic (H) and visual (V) cues are shown in dashed outline, and their combination (VH) is shown in solid outline. When the haptic and visual cues are equally certain (left panel), the combination is centered evenly between them. When the visual cue is more certain (right panel), the combination is centered nearer to the visual than the haptic cue. (Adapted with permission from Ref . Copyright 2002 Nature Publishing Group)
[ Normal View | Magnified View ]

Related Articles

Categorical perception
Bayesian models of cognition
Linguistic relativity
Bayesian approaches to sensory integration for motor control
How thought is mapped into words

Browse by Topic

Linguistics > Computational Models of Language
Linguistics > Language in Mind and Brain

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts