or did people who already had more unfavorable views of Clinton prefer Sanders?
There are polls over a body of time. It would be reasonable to attribute some of the changes to different people being Sanders supports over time different time frames but there are some time frames where things would be pretty steady. Between January and June the race was pretty static.
See the research I cited (edit) in the last post. It's mentioned in this TED talk, the whole talk is interesting, but the stuff on multiple-choice stuff is after 11 minutes
https://www.ted.com/talks/dan_ariely_asks_are_we_in_control_of_our_own_decisionse.g. one of the research examples was two guys faces and they asked girls which one was the best looking (50/50) then they added a third face which was an "uglified" version of one guy or the other. This had the paradoxical effect of making women strongly favor whichever guy most resembled "uglified dude" - because he looked better in comparison to the ugly version, and paradoxically this also affected women's perception of him relative to the other guy. And this has been repeated in a number of unrelated contexts, often ones that involve working out how much money something is worth to pay for. Adding a "choice C" that's an inferior or over-priced version of one of the choices skews decisions towards the choice it most closely resembles, even though the choice C is selected 0% of the time. And skews things
a lot not just a little: in one example, 2/3rds of people picked the cheap option rather than an expensive option, but after adding a meaningless third choice (which 0% of people select), 84% of people selected the expensive option. So the mere presence of a "useless" third choice can in fact turn a decision completely on it's head.
Extra choices, even if they're
never picked can changes the outcome, so as people drop out of the primaries on both sides, everyone's perception of the remaining candidates can change in non-linear ways. It's not just "people who favored candidate X will switch to either candidate Y or candidate Z", people who already favored Y or favored Z can shift after X is no longer a choice (even though they never liked X in the first place). The problem is that humans aren't as rational as we like to think - things that clearly
should not factor into certain decisions, such as choosing between X number of choices, clearly have a big impact.