I have just finished reading this book. It is argued that we can say anything we need to say in any language but that each language forces us to make some distinctions that others might not. For example, in Russian, there is a need to decide whether something is siniy (dark blue) or goloboy (light blue). This, he argues makes the distinction more significant in speakers of Russian than it is for those who do not speak Russian although most people can recognise different shades of blue and label them as we do in English by, for example, using light and dark.
There is another key example, which is that of Matses, a language spoken in Peru. Here, speakers have to state how they know facts they state (Deutscher refers to them being like "the finickiest of lawyers" (page 153)). There are separate verbal forms for whether you know something from direct viewing, inferred from evidence (eg a footprint), conjecture or from rumour. This would presumably give listeners a good opportunity to be critical of other people's claims to knowledge.
Overall, a very interesting book although there are some aspects of the style that are slightly irritating such as a tendency to refer to language as weird or outlandish.
Deutscher G (2010) Through the Language Glass London: Heinemann.