The Strength of the Probability of Co-occurence Relationship

The strength of the probability of co-occurrence relationship between a pair of items can be measured by comparing the number of times that the two items occur together, with the total number of times that each of the items occurs, cat and scratch will occur together quite a large number of times in proportion to the total number of times that cat occurs and in proportion to the total number of times that scratch occurs. Cat and music will occur together only a small number of times in proportion to the total number of times that each of the items occurs. The and cat might at first sight seem to have a strong probability of co-occurrence relationship, since they do often occur together. But when the number of times they occur together is compared with the total number of times that they occur they are found to have a relatively weak probability of co-occurrence relationship. The more times a particular item occurs, the less significant is its co-occurrence with any other particular item, unless it can be shown to co-occur with that item a very high proportion of the total number of times on which it occurs.

Some pairs of items have such a weak probability of co-occurence relationship that it is unprofitable to pursue the relationship between them. They can be considered irrelevant to each other from a lexical point of view.

Another problem which arises from the lack of generality of collocations and their items is the question of knowing where to begin. Elements of structure can be compared with each other on the basis of shared or not shared properties and focal points of interest arise from the comparison. For instance, if one is considering the structure of the nominal group, the obvious focal point of interest is the headword. Modifiers and qualifiers can be considered in relation to the headword. On the other hand, if one is considering a collocation, there is no natural focal point, no natural headword. There is no reason why any one of the items of the collocation The waitress spilt the soup on the tablecloth should be regarded as a headword any more than any of the others. The items are unique and cannot be compared.

This problem is usually solved by the artificial creation of a headword. An item is selected on which interest is to be focused. This item temporarily becomes the headword and the other items in the collocation are considered in relation to the selected item. The grounds in which the selection is made are usually not lexical grounds at all. One might, for instance, have a grammatical reason for the selection; one might be interested in the items which realize the element headword of nominal groups which act as the subjects of clauses. In this case one would choose waitress as the “headword” of the above collocation. Alternatively one might have a contextual reason; one might be interested in items which realise food concepts. In this case one would choose soup. These artificial headwords are called nodes. The items which occur with a node in a collocation are called the collocates of the node. If waitress was chosen as the node of the above collocation, soup would be one of its collocates. If soup were chosen as the node, waitress would be one of its collocates. (In an ideal lexical study every item in turn would be treated as the node for all the vast number of items which occur in a worthwhile amount of data to be treated in this way. The assistance of a computer is essential. Often one has to make do with a less than ideal lexical study).

The strength of the probability of co-occurrence relationship between two items varies according to which of the items is being treated as the node and which the collocate. Cat occurs with the more times in proportion to the total number of occurrences of cat than the occurs with cat in proportion to the total number of occurrences of the. If cat is the node, the relationship is stronger than if the is the node.

A collocation, then, can be said to consist of a node and its collocates. It is important however to recognize that ‘consists of’has a different implication here from the implication it has in the statement, that the structure of a nominal group consists of the elements modifier, headword and qualifier. This latter statement represents a truth about the English language. The statement that a collocation consists of a node and its collocates simply refers to a convenient way in which a linguist organises his data. [...]