I propose that we do not perceive anything directly, that when we believe that we perceive something, that it is an inference.
This is from the nature of perception: some external entity or effect inducing ideas in our minds. The ideas are not the external entities, and our consciousnesses only have direct access to those ideas and not to the external entities. We unconsciously conclude that those ideas are due to external entities, and we continually maintain a model of the external world that is fitted to our perceptions.
This model is usually successful, but it sometimes has rather revealing failures. Like when we believe that a rainbow is a solid object. It looks like one, and many premodern people have believed that rainbows are solid objects, if mythology and folklore is any guide.
The whole idea that we believe is a human construct used to make sense of what is beyond our experience. I'm not sure belief is useful as a tool for explaining given what we know of how we operate today.
Obviously we don't perceive anything directly unless one considers our sensoria as part of perception. Our sensoria are, by the way, part of our perceptions since they have evolved to provide us with capabilities in that task set and they lay at the root of what
steve_bnk and
metachristi discuss. What they discuss is an entirely different problem from the role of perception in knowing based on the construct of belief. It is so because noting glitches in what we perceive only gets at whether we can discriminate what it actually there with what we have. It has little do do with belief other than it is noise for that construct.
Sure one who hasn't experienced or one who doesn't have the equipment to get stuff into that area we still call consciousness (another pet peeve of mine) so however we act in the realm of bad information or inexperience we are going to point to as evidence we need the construct of belief to attack the problem of knowledge. What happens is the construct of belief interferes with what one is actually capable of doing by presuming what one believes is actually germain to the function of perception in the process of knowing. That,to me, is a very bad way to study knowing.
We aren't talking about either perception or whether knowing can be based on belief when we do so. We are talking about equipment not suited for parlor tricks. That's an entirely different discussion from the one where belief is at the base of how we consider the problem of knowing. I find it much simpler to analyse what we have and how we got to what we have and how we use it to our advantage when I'm about trying to understand perception with respect to the problem of knowing. When I do I actually look at what is going on and how we come to understand that as part of the knowing problem.
If one considers perception a process and how it is employed by us when we consider perception's products (sensing, perceiving, recognizing, deciding, etc) when we come to know more than what others in the past have known the whole idea of confounding perceiving and believing with parlor tricks becomes play time. Looked at in this way, the notion of realism as the explanation for how we come to knowing (I prefer understanding since we'll never be a completed product and what we 'know' changes over time) is probably the only sane way to treat the problem of knowing.
We are geared to operate successfully in the world in which we exist and our equipment is adapting to the end all the time. How much more realistic need one get?
I am not a a formally trained philosopher. So if I tweak the sensibilities of those who do philosophy for a living I apologize. However it makes no sense to me to try to force constructs on ourselves as one would require eyelets for every use of thread.
As far as I'm concerned belief will be one of the areas where humans will likely end up bringing themselves to an end because of it. Humans are capable of learning to engage in empirical process as a means for living together.