In Alciphron VII, Berkeley has Euphranor give three distinct arguments based on verbal phenomena that we can have meaningful language without cashing it out into clear and distinct ideas.
(1) Words are signs, but in the case of other signs, we find signs that are intelligible, but do not require us always to connect them with ideas, even though there are ideas to which they can be connected. For instance, counters in card games represent money, but to use them, it's completely unnecessary to keep track of this. You can just play for chips in poker, not keeping track of the money until later you cash them out. Thus we have signs that can meaningfully be used as signs without determining their ideas, because they can be cashed out in ideas later on. This is not unique to game counters; we do the same with figures, where we can get the right results just by manipulating the symbols using the right rules. By extension, it seems that this can be true of words, as well.
(2) Words themselves seem to have uses, connected to "influencing our conduct and actions", that are not connected to "marking and suggesting distinct ideas". This can happen because they are used to formulate rules or "by raising certain passions, dispositions, and emotions in our minds." Thus it seems that a speech intended to direct us to actions or to excite us in some way can be meaningful as such even if we are not attaching distinct ideas to all the words.
(3) Ideas as such are wholly inactive; agents and actions therefore cannot be ideas or like ideas. Words for agents and actions, therefore, do not stand for ideas but for the agents and actions themselves. But we understand agent words (like 'I' or 'myself') and action words (like 'memory', 'will', 'love', or 'hate').
It's easy to miss that these are three distinct arguments and concern three different phenomena, and thus describe three different ways in which words can be used meaningfully without being connected to clear and distinct ideas. The counter use of words, (1), involves use of words in their own right that can eventually be cashed out into clear and distinct ideas. The influencing use of words, (2), involves use of words as parts of a discourse that holistically can be effective even if not all the words are connected with ideas. And the agential use of words, (3), involves use of words that Berkeley claims never have any clear and distinct ideas attached at all. The three together form a stepping structure, as we get purer cases of idealess meaningful words -- (1) points to words that have a capacity of being cashed out in terms of ideas, although we are not presently cashing them out that way; (2), at least as Euphranor develops the argument, indicates words that get their meaningfulness from being in a larger discourse which is performing an intelligible function (this is probably not intended to be a necessary condition, though, as when we shout 'Stop!' at someone); and (3) is concerned with words that Euphranor claims cannot have any consistently associated ideas at all.
Euphranor goes on to give a fourth, more notorious argument, namely the argument based on the premise that there can be no abstract general ideas. This argument doesn't deal with a distinct class of words, though; Euphranor explicitly links general terms to (2) -- we use general terms to "direct us in the disposition, and management of our affairs".
