I've got data from an unpublished prospective memory (PM) study (got an abstract at a conference, but never wrote it up). We used caps and exclaimation symbols, both of which work in the lab to make cues more salient, and put them in email subject lines in a study where we were attempting to test different email header cues to get participants to respond. Whereas they work in lexical decision tasks in the lab, in email headers, the plain text was non-significantly more effective at cueing response.
According the PM theory, despite getting participants to form the intention to reply (with a reward) the association between those cues and replying is reduced becuase of their assocaition with a different response, namely deleting of spam. However, we'd need to do another study to test that ('cos it's not what we naively hypothesised), and me and the mate I did the study with (in the final few months of our PhDs 'cos we were sick of writing up) have moved on to more interesting things (using the half-life2 engine to model multitasking and PM).
