While the idea of spatial violation has a certain appeal, this research article is about people who tend to vote for candidates who, logically, they shouldn't vote for.
Lemme explain. You ask folks in a survey to rank candidates on a bunch of issues, then you ask people where they stand on the same issues. Logically, people should vote for the candidate who looks more or less like them on those issues. And most people do exactly that. But between 1972 and 2004, the range of "spatial violators" fell between 9% and 15% in the presidential election years.
Education is a key factor. Less-educated respondents tended to be violators as compared to those with greater education. Falls into the ideological and partisan consistency, or you can easily explain this through various persuasion paradigms, such as the elaboration likelihood model. Then again, ELM explains everything!, or it seems that way sometimes.
No real media aspect to this study, though I can think up lots of different and interesting ways to integrate media consumption patterns into the likelihood of being a violator, or not. News media consumption might cut both ways. High consumption could arguably lead to less consistency as people pick up on other traits or factors that sway their vote, or high consumption could reinforce ideological or partisan cues that lead to greater consistency. It's a neat question. Time to grab some NES data and crank up the SPSS...