Early in his life, Berkeley developed the belief that sense perceptions form a language by which the originating mind (God) communicates information to us. This is one of the primary contentions of his Essay Towards a New Theory of Vision, the first major philosophical work he published. His claim is that the whole of the physical world is a series of signs which always have the same meaning. We shall see later that this consistency is a critical feature of reality.
I think it's a great time to be studying Berkeley, because there are some long overdue changes in the air, in which the old clichés are being swept away and a closer consideration of Berkeley's work is becoming more common. The features of the 'New Berkeley' are, roughly this:
(1) One of the most important concepts in Berkeley's philosophy is signification; and the real center of his philosophical work is his view that the world is constituted by signs (the divine language Kenny mentions).
(2) While he's a nominalist 'epistemologically', he's a Platonist speculatively. That is, he's definitely a Platonist, although a nondogmatic one; he manages to combine this with typical empiricism (which tends to be very un-Platonist) by the means of his focus on the divine language. (There is some room to think that his Platonism is developed over time, i.e., that he becomes more Platonistic as time goes on. This question is an important one for Berkeley scholarship in the years ahead.)
(3) Berkeley's immaterialism, and particularly the negative part (the attack on materialism), has to be put into historical context to be properly understood, and should never, ever be construed as suggesting that the external world does not exist. Rather, it's a rigorous criticism of a very common (especially in the eighteenth century) way of understanding what that external world is.
(4) Far from just being thrown in as an ad hoc device in the Principles to save knowledge of minds, the term 'notion' is important, because on Berkeley's view of the mind, the primary cognitive activity is not perception of ideas but understanding of signs (which perception of ideas subserves).
(5) Far from being an inconsistency, Berkeley's appeal (in a number of his works) to particles (e.g., of light) is entirely consistent with his more general views.
These shifts in interpretation, although in places in need of refinement, have all the advantage of actually taking the evidence more seriously than anyone has for a long time. As I said, an exciting time for studying Berkeley. At present I'm not sure I entirely like the gloss in terms of domains of quantification that Kenny suggests in his post, although it does have the advantage of sticking reasonably close to Berkeley's interest in signs and language, and although I have definitely used similar language myself. It seems to me to be much closer than standard interpretations, but still to be missing something (although I can't quite put my finger on what). If I find out what, I'll let you know.