Davidson uses ``Block'' and ``Slab'' to illustrate the blurring of the
distinction between words and sentences that Ogden and Richards and Morris
indulged in.
And one might say, at this point,
that all that I have said is beside the point, since the utterances in the
builders' language game should be thought of as one-word sentences.
I think when people say the utterances in question are ``one word sentences'' they have one conception of a sentence in mind, as that which conveys a complete thought. Since ``Slab'' conveys a complete thought, it must be a sentence, even though it does not meet some of the other conception of sentence-hood, like having both a noun phrase and a verb phrase.
Suppose someone asks you ``Who will win the National League, Western Division this year?'' and you reply, ``San Francisco.'' Your reply clearly conveys the thought that San Francisco will win the pennant this year in the National League, Western Division, which is a complete thought. You can express this complete thought merely by saying the name of a city, because the other parts of the thought were, as one might put it, already in place.
Let us introduce of subclass of propositions, simple
singular propositions. These may be identified with a major
constituent (an n-ary relation) and a sequence of n minor
constituents (individuals or non-individuals of the
appropriate sort to stand in relation R) and a polarity (1
or 0). So 
is the proposition that 
do stand in relation R, while 
is the proposition that they do not stand in that relation.
Let us say that an issue is a type of simple proposition that is missing one element. The missing element I indicate with a question mark. So here are some propositions with an element missing, and the questions that would naturally express them:
Who did A kill?
Who killed B?
Did A kill B or not?
What did A do to B?
Whenever there is an issue ``in the air,'' one can express a complete thought by supplying an answer, the missing propositional ingredient. One typically does this with a noun or a verb, rather than with a complete sentence.
I am not too concerned here with how we should treat answers to questions. My point is that there are two rather different ways of expressing complete thoughts. One can, out of the blue, utter a sentence that identifies all the necessary propositional constituents for a thought, as when I say, in the middle of a paper on the philosophy of language intended for an audience in Austria,
I'll call this ``constructing a complete thought.'' It is quite a different thing to do what I'll call ``completing a thought.'' This is what I do when the issue of who will win the Western Division of the National League this year is in the air, perhaps because someone has asked that question, and I merely say,
Here my words are responsible for completing the thought by introducing the last needed constituent to form a proposition. It is like Wittgenstein's connecting the rod and the lever to complete the brake assembly. The point goes both ways. Surely, in the absence of the rest of the assembly, just connecting the rod and the lever doesn't amount to much of anything. But, on the other hand, in the situation in which the rest of the assembly is in position, and connecting the rod and the lever is all that remains to finish the brake assembly, that connection doesn't amount to constructing the entire assembly, only completing it.
Suppose I claim to be able to express an important law of physics in one word. You challenge me, and we put some money on it. So I tell you, ``Ask me whether falling object near the surface of the earth accelerate and 32 feet per second per second.'' You do so, and I say, ``Yes.'' Have I won the bet? We may or may not say that I have uttered a one word sentence, but we should not fool ourselves. I have completed a thought with a word, but not constructed one.
Now in these cases, one word can complete a thought because of the other propositional constituents have been introduced linguistically, by the words spoken by the person who asked the question. In the builder's case, however, it is not language that has introduced the other constituents. They are present in the situation not because they have been introduced, but because a number of things about the builder's interaction with his assistant have been established by custom or tradition--it is what they do every day, and they never do anything else. The job of the assistant is to bring building stones to the builder. In a more complex language game the builder might need a verb, to tell the assistant whether to get or take or make or break a stone, but a verb isn't necessary, because it is established that the pattern of action is getting a building stone.
If by language we mean words that are spoken, written or signed, then language has a very small job to do in this case. An issue is in the air:
Bring-to-x-a-stone-of-type-y, the Builder,
? 
,1
No doubt the missing component is the type of stone. Supplying it is the only job that language has in this game. It is all that is missing, it is the link between the rod and the lever in an otherwise complete brake assembly. The choice of words, by indicating the choice of stones, indicates the choice of action that will comply with the builder's wishes.