[This is perhaps germane to yesterday's exchange with Gary re: Associationism vs Computationalism.]
It seems to me that "sensation" is used ambiguously in the EP literature. Sometimes (when discussing vision) "sensation" means something like retinal stimulation (as Andrew claims in his post here). Let's call this sensation-rs. Sometimes, however, "sensation" means something like an observation report of the sort discussed by the logical positivists earlier in the century.* Let's call this sensation-or. Very roughly, at least part of the difference is that the first does not seem to involve conscious mental experience, where the latter does.
So, this means that when one denies that perception is based on sensation, there are (at least) two distinct things one can mean:
1) Visual perception processes sensation-rs.
2) Visual perception processes sensation-or.
Gibson at least seems (to me) to want to reject both claims, but it seems to me that 1) is pretty plausible (it's what cognitivism typically asserts), where 2) is not.
Relatedly, I think that Merleau-Ponty spent some time arguing against 2), but I don't know about 1).
* Reed & Jones, (1978), refer to Agassi, (1966), as a target of their criticism. Agassi discusses observation reports.
Agassi, J. (1966). Sensationalism. Mind, 75(297), 1.
Reed, E., & Jones, R. (1978). Gibson's Theory of Perception: A Case of Hasty Epistemologizing? Philosophy of Science, 45(4), 519-530.