The idea that language shapes thought has a name: the Sapir-Whorf hypothesis (also called linguistic relativity), associated with the linguists Edward Sapir and Benjamin Lee Whorf in the early 20th century. It exists in two versions of very different strength.
Strong linguistic determinism (Whorf's more radical formulation): language determines thought. You can only think what your language gives you the concepts to think. Speakers of different languages cannot have the same thoughts because their languages carve up conceptual space differently. The famous example Whorf used: the Hopi language has no words for time as a linear sequence of discrete units (past, present, future). Therefore, Whorf claimed, Hopi speakers cannot think about time as Westerners do.
This strong version has been largely discredited. The Hopi language does express temporal concepts, just differently. More broadly, the claim that you cannot think what you cannot linguistically encode is empirically false: pre-linguistic infants have sophisticated conceptual representations, deaf individuals who grow up without language still have complex thoughts, and mathematical concepts can be grasped before they are verbalized. The strong version turned out to be a very attractive and almost entirely wrong hypothesis.
Weak linguistic relativity (Sapir's more careful version): language influences thought. Having a word for something makes it easier to notice, remember, categorize, and reason about. It affects the default way you process experience, even if it does not determine what you can think in principle. This version has substantial empirical support.
The most famous experimental demonstration involves color. Russian has two separate basic terms for what English calls "light blue" (goluboy) and "dark blue" (siniy) where English has just "blue." Researchers found that Russian speakers are faster at discriminating colors at the goluboy/siniy boundary than at other blue-blue boundaries, and faster than English speakers at the same boundary. The language doesn't determine what colors you can see, but it primes your categorical perception. The lexical boundary becomes a perceptual shortcut.
The philosopher Anna Wierzbicka has developed a rigorous, empirically grounded version of linguistic relativity through decades of cross-linguistic research. Her finding: words like the English "fair" or "reasonable" encode specifically Anglo cultural values about individualism, mutual respect, and rule-following that don't map cleanly onto other languages. The Japanese concept of wa (harmony, smooth social functioning) encodes values about group cohesion that have no close English equivalent. These lexical differences are not superficial: they organize whole domains of social experience, moral evaluation, and emotional life differently.
What this means for translation is significant. Translating across languages is not merely finding equivalent words. It involves negotiating between conceptual systems that may be differently organized at a deep level. The translator who renders saudade as "nostalgia" is not just choosing the wrong word: they are importing a concept from a different system that distorts the original. The best translations acknowledge this explicitly, as when translators use footnotes, retain the original word, or coin neologisms to signal that something is happening here that resists smooth transfer.