The picture I want to suggest is that:
In the builders' language, there is a basic pattern of action and circumstance. The builder is speaking to the assistant during working hours; he is building something of stones, and the assistant is to help him by passing him the stones he needs. The four words create four action-types for the builder. Each type of action has its compliance conditions, bringing a slab, bringing a pillar, and so forth.
Only within this context, does the builder's saying, e.g., ``Slab'' amount to his expressing the command that the assistant is to pass a slab. Such an utterance is like connecting the rod and the lever--it amounts to what it does only in conjunction with certain circumstances. But now, given this, what is the contribution the individuals words make to the propositional content?
First, consider a case in which the builder's mumbles ``Slab'' and the assistant doesn't understand what command has been expressed. He knows that he is to fetch a building-stone of a certain kind, but he doesn't know which kind. When he realizes which word was spoken, he realizes which kind of stone he is to fetch. That is exactly the contribution that ``Slab'' makes. It is just this that we capture by saying that ``Slab'' stands for slab.
Next, compare this situation with the one in which there is no communication and no language involved at all. The builder works alone; when he wants a building stone he walks over to the piles of stones, picks one up, and brings it back to the building site. He goes down one of four paths to one of the four piles. Depending on which of the four piles he goes to, he gets a different kind of stone. So we have a basic pattern of action, and four different actions of this type, with four different results, that would be satisfactory for four different desires.
Now consider the sub-action of turning down the pillar path--call it ``P'' instead of the block, slab or beam path (call them ``B1,'' ``S'' and ``B2''). That sub-action determines that the overall action will be of a type that satisfies the desire to get a pillar, rather than the desire to get a slab, beam or block. That particular sub-action has a certain determinate contribution to make to the result of the whole act of which it is a part. It would be quite misguided to say that it was the turning alone which had the result that the builder got a pillar. However, given all of the other things that were set up, this is the effect of taking this sub-action rather than another.
That sub-action has that effect, because of a pre-existing fact, that this path leads to the pillars and not the slabs, beams, or blocks.
So what we have here is a pre-existing relation between a path and a kind: this path leads to pillars. And we have the path involved in one of four action-types. The result of the builder's action is due to the particular sub-action he chose, and the relation the path involved in that sub-action had to a certain kind of object.
Similarly, when our original builder says ``Pillar'', he performs a certain act that involves a certain object, the word ``Pillar''. That word has a pre-existing relationship to a certain kind of object: pillars. Its the relationship we call standing for or referring. The builder's utterance is a case of commanding that a pillar be passed, because of the type of action it was, and the pre-existing relation, reference, between the word involved in that type of action and a kind of stone.
There are of course a great many differences between the two cases. The fact that the path leads to the pillars does not depend on the habits of the builder and his assistant. The fact that the words ``Pillar'' means pillar does depend on the habits of the builder and his assistant. That is of course a tremendously important difference.
But the fact that the relationship depends on the habits of the builder and his assistant (and various other mental facts about them), does not mean that the relationship depends on a relationship between sentences involving the word ``Pillar'' and the truth or compliance conditions of those sentences. The two dependencies are just different issues.
One indication that this is so, is that the same fact could play such a role in determining the result of a number of different types of action. For there are many things that one could do, once one got to the end of the path, besides fetching the pillars. One could throw pillars or write graffiti on pillars or trip over pillars or learn what pillars were. In understanding how going down path P, rather than S or B1 or B2 determines that a given subtype of a pattern of action is writing graffiti on a pillar or learning what a pillar is or tripping over pillars, one appeals to the very same fact, that path P leads to pillars.
When we think of the builders' language, we think that there is a connection in the builder's and the assistant's minds between the words and the different kinds of blocks. (I am just using ``minds'' in the normal way, not contesting the somewhat behavior-oriented position Wittgenstein is advocating.) These connections may be the result of what the Builder did to establish and teach his system to the assistant, a system he originated for the purposes of getting building stones brought to him. But one can easily imagine that the same connection could be used in new ways to create families of sub-actions for a given pattern of action. Suppose that as the builder grows frail and the assistant learns the trade, the builder has the assistant not only fetch blocks but, later in the day, place the blocks into a wall. The builder has the plans in his mind, though, and he indicates to the Assistant at each step of the way what type of block should go in each place, using the four words. Now this is a different language game, in the sense that it is a different family of actions that the commands pick out. But I think it would be rather natural to suppose that once the Builder has shown the Assistant what the pattern of action was, he could use the very same words he had been using in the original language game and things would work fine. He would say ``Block" and the Assistant would put a block in the place designated, and so forth. Our analysis of this language game might look like this:
(B) Analysis of Second Language Game
Here clause (B3) is unchanged from (A3) of the analysis of the original language game. The difference between the two games is not in what the words stand for, but in the overall pattern of action to which each of them standing for what it does makes a differential contribution.