Dario Ringach wrote in a letter titled Neuroscience: States of Mind:
(inline references removed, see link for references)
If the ability to perceive oriented stimuli is found to depend on the
cortical state, we can then ask a number of important questions
about cortical function. Is the time the cortex spends visiting
particular states a function of attention? In other words, if an
animal is trained to perceive horizontal stimuli, will there be a
corresponding increase in the time the cortex spends sampling the
horizontal orientation map relative to the other maps? Perhaps
perceptual learning has more to do with 'top-down' mechanisms (brain
processes learning how to 'control' the ongoing state of the early
visual cortex) rather than 'bottom-up' mechanisms (external stimuli
producing dramatic and long-lasting changes in the way the visual
cortex processes incoming information). In more general terms,
one might ask if such intrinsic cortical states represent the brain's
'current hypothesis' about the state of the external world. In
this case, the purpose of the cortical machinery would be to
continuously update the current hypothesis by considering new visual
information.
Rabinovich is well-known in dynamical systems neuroscience for highlighting the importance of transient states (as opposed to attractor states). That is, it is the active dynamics that correlate with perceptive state, moreso than the steady-state dynamics. This view has experimental evidence from several different organisms (See the Rabinovich link) but the most notable example is Locust olfactory states.
The Rabinovich paper above is really the goldmine for this question, including fMRI evidence, but another source of fMRI theory in relation to brain states is Karl Friston's Free Energy Principle for the Brain