Distribution

It is one of the merits of Bloomfield to have shown the importance of distribution as a criterion for classifying word-groups. Before discussing distribution as a feature of types of groups, I should like to formulate two objections to Bloomfield’s procedure.

From Bloomfield’s own point of view, it would, in my opinion, have been more logical and fruitful to start with a distinction between different distribution of the groups themselves, instead of starting with different distribution of members. The result of the latter procedure is that his category of ‘exocentric constructions’ is a catch-all, comprising, for instance, the predicative group John ran and the connective group with John. If the distribution of the groups themselves had been taken into account first, especially by noticing that John ran is a ‘favourite sentence-form’, the predicative group would have received its unique position in English syntax. [...]

A minor objection concerns the terms ‘endocentric’ and ‘exocentric’, which I consider unnecessary neologisms, but thisquestion depends, perhaps, on the other question whether a special name is desirable for what I have called the catch-all, or not.

Other objections, concerning the value of his procedure fromthe point of view of structural linguistics, will be discussed below.

In dealing with distribution as a feature of groups, I distinguish between the distribution of the group, and the distribution of its members.

1) Distribution of the group

For this purpose we have to distinguish the various types of distribution of syntactic unitsin general. A syntactic unit is either a word or a word-group.

There are only two main types of syntactic distribution. The unit is used either аs an independent, or as a member of a word-group. An independentis used either as the whole word-content of a sentence, Alas Mary! John died, or as a part of it, Alas, Mary, John died. The last sentence has three independents. They are not members of a word-group. They are not, for instance, members of a coordinative group. The criterion, in this case, is that the use of coordinating conjunctions would be impossible: Alas and Mary and John died; this distinguishes the combination from the coordinative group Eat, drink, be merry, into which co­ordinative conjunctions may be inserted: Eat, drink, and be merry. Eat, or drink, or be merry.

A member of a word-groupis either a head, or an adjunct, or a conjunct, or a coordinate. The difference is defined in terms of syntactic omissibility of members of the group. As syntactic omissibility is a matter of degree, no rigid lines of demarcation can be drawn between the three categories, but the distinction itself is of fundamen­tal importance. The test to be applied is the omission test.

A headis defined as a member that cannot be omitted without affecting the structure of the rest of the sentence, whereas the other member can be omitted in the same sense. In the sentence I have fresh milk, the word milk is head of the group fresh milk. I have milk is possible, I have fresh is impossible.

An adjunctis defined as the omissible member of a group of which the other member is not omissible, e. g. fresh in fresh milk.

A conjunctis defined as a non-omissible member of a group of which the other member is equally non-omis­sible, e. g. both John and ran in John ran, or both with and John in with John.

A coordinateis defined as a member of a word-group of which each member is omissible. Examples are men, women, and children in the group men, women, children (cried).

It is obvious that each of these five types of distri­bution, and several combinations of them, is a typical feature of some type of group in some language. [...]

3) Distribution of members of the group

This is Bloomfield’s basic criterion in classifying word-groups.

It is obvious that it is a typical feature of many types of groups. For examples I may refer to the discussion of Bloomfield’s classification in the Introduction.

In-the-following I shall use the terms ‘coordinative group’ (consisting of coordinates, as defined above), ‘adjunc­tive group’ (consisting of a head and an adjunct), and ‘conjunctive group’ (consisting of two conjuncts). I believe that this terminology is preferable to the use of endocentric and exocentric constructions. Its disadvantage is that ‘coordinative’ is used both in a structural and in a purely distributional sense. As, however, the two concepts in prac­tice apply to exactly the same groups, I do not believe that there will be any ambiguity. I may add that in analyzing the structure of word-groups, and defining types in terms of distribution, excellent work has been done especially by Ch. F. Hockett, Rulon S. Wells, and Zellig S. Harris. See e. g., R. S. Wells, Immediate constituents, Language 23, 1947, 81 — 117; Ch. F. Hockett, Problems of Morpheme Analysis, Language 23, 1947, 321 — 343; Zellig S. Harris, Methods in Structur­al Linguistics, Chicago, 1947.