hmm... also on the topic of goal-states... i guess if the representation of the outcome is a strong driver, it makes sense that you would get "pavlovian" behaviors that are uncoupled from instrumental contingencies. the dominant representation isn't "i'm here, what should i do?" (which you could imagine would favor instrumental behavior), but it's "here's what i want". this could favor actions that are associated with that outcome (like salivating for food).

maybe you could also view lots of the little internal actions that we do during goal-states as pavlovian. orienting/preparatory actions that are consistent with consuming (or avoiding) the outcome. i wonder if we're doing lots of these online during listening to music.

http://cercor.oxfordjournals.org/content/early/2014/09/20/cercor.bhu210.long