This may be one of the better stabs at understanding how late-night comedy programs affect learning that I've seen, and its discussion of political knowledge research is excellent. Now that I'm done gushing, let's get to the details. Using item response theory and some slick computational tricks to analyze 2004 election data, they looked at exposure to a general measure that includes most all late-night shows (Letterman, Stewart, et al., see below). Table 1 breaks down the item-by-item results, which for the statistically challenged will be something of, um, a challenge.
If you work carefully through the results, and read the conclusions, you get to a few main points:
- Fewer response alternatives (meaning fewer choices among possible answers) often equals an easier question, or at least less item difficulty.
- Watching late-night comedies does lead to slightly more knowledge.
- But -- on the point above -- this is mostly on easy questions. This is perhaps more of a TV effect than anything else, some incidental learning.
- And "this increase is especially pronounced among the viewers who do not pay close attention to public affairs." That's important too, because it fits some of what we are beginning to understand about television news, which seems to inform only the less informed.
Weaknesses? Every study has them. The biggest for me? It's easy to quibble with their catch-all independent variable. Letterman ain't Leno ain't Stewart ain't Colbert, so a single item probably fails to capture significant differences. This is a secondary analysis, so the authors worked with what they had available, but it's certainly a question that deserves further study.