Studs Terkel, who died recently, at the age of 96, had a special place in the hearts of some linguists — those who were studying the syntax (and accompanying pragmatics) of colloquial English, back in the old days, before very large corpora and automated search techniques were easily available.
There were essentially two says to investigate colloquial language non-anecdotally: collect your own corpora (by eliciting speech, recording it, transcribing the recordings, and coding the transcriptions); or use other people's corpora, collected for other purposes (in which case, other people do the elicitation, recording, and transcribing, leaving you to do the coding — and of course the counting and the analysis). That's still the case.
Collecting your own corpora is hard work, it takes time (transcription is particularly tedious), it takes skill (in elicitation, especially; you can't just stick a microphone in someone's face and tell them to talk to you). and it takes money. Piggy-backing on other people's corpora hugely simplifies the task. For many purposes, that just won't do. But sometimes, things sing.
That's where Studs Terkel comes in. He spent decades interviewing (mostly) ordinary people and publishing the conversations, and he was fantastic at establishing a rapport with the folks he interviewed. The result was a body of corpora that's a goldmine of data for (some) linguists, from Division Street (1966) on.
It has its limitations, of course. In particular, the material is edited, which means that variants that either the interviewees or the editors are conscious of as socially significant are probably out of bounds, and some kinds of statistical tests are probably not going to be doable. But if you choose the right question, you can gain some insight.
In the old days, you still had to search through the texts by hand, mark all the relevant examples, code them, put them on index cards (so that they could later be sorted and counted on your living-room floor, or its equivalent); it helps a lot if the stuff you're reading actually has some interest for you. Some of this — though not all — can now be automated and immeasurably speeded up.
Studs's books were a big help. In particular, I used them to provide data for my first work on "stranded to", Verb Phrase Ellipsis leaving infinitival to as a remainder (as in "They told me to leave, but I didn't want to ___"). (My first published discussion was here, in 1982.)
[I use his first name (well, the name he used; his birth name was Louis), partly because almost everyone did, and he invited it, but also because we actually talked, back in 2002, when I was trying to persuade him to give a plenary talk at the New Ways of Analyzing Variation conference at Stanford that year. He was charmed and honored by linguists' interests in the way be interviewed people and crafted his texts, but at around 90, the prospect of a plane trip from Chicago to San Francisco and a long crowded weekend of meetings, all of which would take him away from projects he wanted to finish before he died, was just too much. Who could argue with that?]
(Thanks to Ellen Prince, who first pointed me to Studs's books, very long ago.)