In early 1960's in French Canada, beginning of a change. Study of bilingualism, immersion schooling, led to an interest in attitude change i.e., to see whether changing schooling patterns (bilingual schooling etc.) led to a change in outlook among dominant sectors of society toward minority sector (i.e. French Canadians).
Lambert et al./footnote``Evaluational Reactions to Spoken Languages.", 1960 developed the matched guise technique:
Speakers are chosen who can pass as native in two different varieties (e.g. English and French) and are recorded speaking a short paragraph in both languages. Same content in both paragraphs (English is a translation of the French) so the only variable is the language. If there are 5 speakers, there are 10 `guises' and these are presented to subjects as if they are different speakers.
Subjects are asked to rank the different `speakers' on a number of different personality traits that the subjects think they can detect from the voices. Such traits as:
Outcome: Both English and French subjects rank the English `guises' higher on certrain traits; each group differs in some ways on other traits, but there is consistency from both groups on the higher ranking of English guises on the above traits; other traits (religiosity, kindness) were ranked higher for French guises by French S's.
After this initial study, this technique was applied to other sociolinguistic situations involving other varieties of language, e.g. AAVE (African-American Vernacular English) and SAE (Standard American English), Chicano English and SAE, Israeli Hebrew as spoken by Arabs vs. Israeli Hebrew as spoken by Jews, etc. etc.
Other traits or characteristics could also be asked about: what kind of job should this person (or would be this person be likely to) have:
Labov et al. tested the above for certain traits, i.e. the varieties differed only with regard to use of certain phonological (morphological syntactic, lexical) variables, and found significant rankings for (e.g.) the pronunciation of post-vocalic (r) in New York NSE, etc. Labov's work looked even more carefully at the socioeconomic `class' of the subjects, i.e. how they differed in their ranking of certain variables.