It seems to be that those who concentrate on remember their dreams are more likely to do so. Although this has not worked for me personally it is attested by many who attempt to concentrate on their dreams. To me the most interesting of the dreamwork practitioners is Robert Bosnak at CyberDreamwork.com. Even though I don’t like his books very much his workshops are really good. And what he says is that it really does not matter how much of the dream you remember. What really matters is to relive it without interpreting it. In his dreamwork what one does is connect the dreams aspects to their loci in the body. Once each image or part of the dream that is remembered and relived in imagination are anchored in a part of the body one attempts to feel these connections and feelings all together, and that can at time lead to transformational experiences which Bosnak interprets alchemically. This technique can be done in groups, and they have long distance groups that operate via their website and do this kind of dreamwork over internet. You can also take a series of courses from Bosnak and get certified to do this kind of dreamwork. Much of what goes on in this area of Dreamwork strikes me as totally bogus, but I have found that Bosnak’s way of doing it has some interesting connections to Somatic Experiencing of Peter Levine. Another completely different but similar psychotherapy is that of David Grove which is called Metaphor therapy. These three completely different kinds of therapy have many interesting aspects in common. The key in each case is not the interpretation of the dreams but relating to the sensations that they make one feel in different parts of ones body, and focusing on that and following that as it undergoes transformations in the light of awareness. Certain types of Buddhist meditation also focuses on sensations and their self-transformations called Vipassana (introspective observational) as opposed to Samantha or pacification types of meditation. (http://en.wikipedia.org/wiki/Vipassan%C4%81)
I believe that everyone remembers some dreams sometimes. But many times dreams are uninteresting and so it is less likely for people to even recognize them to try to remember them. But if people never remembered their dreams then they might not know what dreaming is, as only those who have dreamt sometime in their lives can know this. And you hardly ever run into anyone who denies dreams exist, even if they don’t have them very often, and do really remember what happened at least they know on waking that they have dreamed, and what dreaming is. So what I am taking issue with is the idea that some people don’t EVER remember their dreams. I believe that everyone dreams, and everyone knows what dreams are, but they might ignore them, not really remember then explicitly, and all they may really have is a residue of dreaming, the fading of the dream. Everyone has the experience of dreams fading from ones grasp as one awakes. Like anything in consciousness if you spend time concentrating on it then it is more likely that one will remember more of dreams, and for Bosnak’s approach to dreams to work all one needs is a snippet that is still active enough that one can enter back into the trance of the dream while waking.
It is an interesting fact that studies show that most peoples dreams are negative, and so this also militates against against people wanting to remember them.
But the most interesting thing about your question, which I really want to deal with is the relation between dreams and memories. This is because I think these two phenomenological experiences are of real interest because they both deal with the realm of the imagination in different ways. According to Kant Apriori Synthetic Manifolds that we intuit are all based on the imagination. One of those is the singular of SpaceTime. And the idea of General Schemas Theory is that we project not just a homogeneous plenum of space and time as Kant thought, but instead different nested scopes of templates of understanding for different resolutions of spacetime. And the key fact which I use in my dissertation is that schemas operate across various realms of consciousness. So they operate in Memory and Dream, as well as in our projective imagination. In other words schemas are more basic than the variations in our modes of consciousness. I think this, if true for everyone, is of great interest because what structures mundane consciousness, imagination, memory and dream is all the same and this is one way to think about its a priori nature.
In many ways remembering a dream, whether a REM dream or a hypnogogic dream is tantamount to schematizing it. Bosnak’s point is that dreams are real, they have their own separate reality, and it is that reality one wants to tap into as it animates feelings within the body through its images and the feelings that are produced within the dreams themselves. But the difference between normal REM dreams and Hypnogogic dreams are striking in that in normal REM dreams we are observing ourselves from the outside in many cases. While in Hypnogogic dreams the illusion is very real and it is in a space in relation to our bodies and we are inside our bodies usually trapped in a frozen bodily state. But even though our relation to our bodies and other bodies in the two kinds of dream are very different still we apprehend those relations on the basis of the same schemas. We do not have a schema for one and a different one for another mode of consciousness. Same is true with the mundane consciousness of being in the world and what we imagine. All Art and Architecture uses the same set of schemas as do everything else we produce. It is this continuity across spacetime of phenomenal experience that make the a prioiri schemas a priori, they come before the differentiation of experience into modalities. Even non-worldly experiences in dreams, imagination, and visions reinforce the same schematization that connects us to the mundane world.
Studies have been done which show that our waking consciousness is not much different from our dreamstates. And much of our time we spend in trance during the day. The part of the time that we are in objective reality, designated as intersubjectively real, is not very long each day. So in a sense it is not significant whether we remember dreams or not because we are always in trances that related to our finitude. Reading, Conversation, Eating, Dressing, taking showers, riding in conveyances, driving we are continually falling into trances. But those trances even the deepest ones induced by self-hypnosis where we enter into the unconscious directly are all schematized with the same schemas. This was shown by an experiment done by Erickson where he would hypnotize subjects prior to their actually being “officially hypnotized” and they they would be asked where the wall was in the room they were seeing, and if they pointed out the real wall it meant that they were not under hypnosis, but if they pointed to a wall that was not there, then they were in a hypnotic trance. But even in deep trances people would see walls, not something else. Imaginary objects were still objects following the same schematic templates as actual intersubjectively agreed to objects. And the same is true of hypnogogic and dream objects or environments.
So in a sense it does not matter if we remember our dreams or not because much of consciousness is in fact trance, and these trances can be very deep so as to be direct experiences of the unconscious which we do not remember on waking. But then self reports of glimpses of these states find that the people having those experiences are seeing visions of different parts of their lives, or dreamlike images which are schematized just like our mundane experiences.