I’ve been researching–or attempting to research–any academic work that’s been done on semantic representation (i.e. intensions/extensions) AND linguistic relativity. My search so far has proved unsuccessful.
Oppositely, I’ve found a wealth of information concerning the empirical justification for linguistic relativity. Undressmerobot.com offers an informative review of the issues concerning linguistic relativity, but mostly from the standpoint of social psychology.
That said, I find this evidence particularly favorable for linguistic relativity:
One of the most telling tests was one that dealt with the duplication of lines on a piece of paper. Gordon drew single and multiple lines on a piece of paper and asked the Pirahã members to copy those lines. For one, two, and three lines, the Pirahã had no difficulty completing the task. As the number of lines increased, the discrepancy between the number of lines and number of copied lines also increased. For instance, many only reproduced three lines when shown four (Holden).
Of course, in order to justify the strongest version of the hypothesis, the researchers (Holden, and I’m not sure who else) would have to show that the task itself was completely non-linguistic. Surely we think of simple copying tasks as non-linguistic, but the description above is obviously a summary and therefore incomplete.
Nevertheless, interesting evidence.