Just by way of explanation, B L Whorf's hypothesis (a considerable expansion of Sapir's original opinion) was that vocabulary, and more importantly, grammar, guided (and was guided by) modes of thought. A particularly famous example was that Hopi, having no past or future tense, which led its speakers to conceive of time as cyclical in nature, as opposed to the almost-neurotical need for Europeans to keep historical records and plan everything out. Quite apart from the fact that his analysis of the Hopi language was hideously inadequate (it is as tense-filled as most European languages, lacking only the explicit present progressive that is a peculiar feature of English and Celtic languages), he seems to have managed to miss the fact that the culture had developed a sophisticated set of timekeeping technologies as well. Much of the follow-on research ("language as culture" is a tasty morsel for anyone bent on believing that language is nothing more than learned behaviour) can be shown to use specific meanings gathered from informants for words that can be used to express much more general concepts. The problem is that the researchers failed to ask if the words could be used in another way.
The provable aspects revolve mostly around concepts like the colour "grue". That is, people whose languages lack separate words for blue and green, using instead a single word for both that could be rendered in English as "grue", take longer to distinguish between blue and green. It's not that they can't see the difference, but they initially file blue things and green things under the same heading, and only go on to distinguish between them when it's necessary. We have the same problem in English with yellow -- only people who've spent enough time in the world of colour would be likely to immediately distinguish warmer/redder yellows (like cadmium yellow) from colder/greener ones (like lemon yellow). And they learn that (usually) by figuring out that the two are not interchangeable when mixing colours -- both yellow and blue done wrongly and yellow and red done wrongly result in a muddy brown rather than the expected orange and green. It doesn't affect our perception, just our initial means of filing things, and it means that some circumlocution is needed when describing a particular yellow to somebody who doesn't have the technical vocabulary.
We also don't have a single word for "someone who always feels cold" as the French do, but we all know a few frileux even if we don't have a word for it, and are aware of that knowledge. We have dropped hither, thither and whither, but are still aware of the difference between direction and location when we use here, there and where. A large number of the world's languages don't have sex-linked gender, but their speakers are perfectly capable of telling the difference between men and women.
The only place where the Sapir-Whorf hypothesis really holds any water is in the realm of computer languages (and similar ordering-machines-around environments). Sometimes it is impossible to express a thought/instruction even when the machine is perfectly capable of carrying it out. Sometimes it's just so far away from idiomatic use of the language that it's better from any perspective to find a suitable paraphrase that is idiomatic. But these are toy languages designed to accomplish specific goals, not general-purpose languages that anybody has to live in. (Try having a relationship in Lisp and see how far you get.) In the human world, the ability or inability to express a thought has a lot more to do with experience and culture than with any limitations that the language you speak may impose on you.