Words in Context

As we’ve repeatedly seen before, words don’t need to consist of single free morphemes, and if they don’t, they automatically have a specific context they occur in. At other times, they may require other types of ‘complements’, either on the syntactic or semantic side, in order to develop their complete meaning, as we, for instance, observed when we tried to determine the relevant features for verbs. And of course words don’t ‘pick’ these contexts randomly, but based on their semantics, i.e. what can be expressed through them in the first place, as well as certain kinds of conventions or topic areas, i.e. how and where they occur. Let’s explore these ideas further in the sections below.

Collocation

Although semantically compatible words can often combine freely to create ad-hoc – temporary – meaning, words that co-occur frequently within a certain span of each other often turn into relatively fixed and conventionally accepted combinations. In such cases, we say that these words collocate with one another, e.g.:

fair hair, fair play
nice try
black tea
fast car

Due to their frequency and commonness, we can then see them as prototypical constructions, where the accompanying word almost becomes another feature of the accompanied one. And, once we have such fixed, established connections, it then becomes odd – in other words (highly) marked – if we suddenly start using unfamiliar ones. If native speakers (or highly proficient non-native ones) do this, they often deliberately try to achieve a certain effect, i.e. to emphasise or exaggerate something, or to indicate that they’re being ironic. For less proficient non-native speakers, the situation is obviously somewhat different, and we can generally assume that they’re unaware of the collocation.

The list below shows pairs of marked an unmarked forms, where each time the unmarked one is a genuine collocation, and the one preceded by the *? symbol combination indicates that the construction may be doubtful as to its general acceptability.

strong coffee vs. *?powerful coffee
blond hair vs. *?blond coffee (c.f. French bière blonde)
good/bad omen vs. *?interesting/exciting omen

There are different ways of identifying collocations. One of them is that we simply think about/investigate which other words occur within a given span of a target word. This form of investigation generally works very well, but of course involves careful analysis and observation. Another way, which is frequently used to automate this task, is to compute collocations from a text by using statistical measures such as a mutual information (MI) score, t-Score, or log likelihood (LL) ratio, etc. While this may make it easier to find some collocations, one always needs to be cautious regarding the reliability of such measures, as factors such as the overall text size, quality of possible annotations, etc., may influence the computation.

To practise both analysis forms, first open the COCA page again.
Change the display option to KWIC, type the word omen into the box next to WORD(S), first click on the L in the sort options and then the number 1 three fields to the right.
Click SEARCH and look through the results on the right to see whether they confirm the collocations I indicated above. Tip: generally, the adjectives should appear in green, although some may also (mistakenly) be colour-coded in a different way.
Keep a list of the most relevant/interesting ones in the box below for your reference.
Next, first click on RESET, type in the word again, then select COLLOCATES, which will open up another box to the right, as well as two dropdown lists.
In the box, type [j*], select 1 from the first dropdown, and 0 from the second.
Change the option for SORTING to RELEVANCE. Initially, keep the MINIMUM set to FREQUENCY and the number 10, and click SEARCH.
Change the number to 5 to get some more results and observe the difference.
Finally, change the value of the dropdown list next to MINIMUM to MUTUAL INFO and click SEARCH again.
Look through the results list and think about whether all examples should really be considered collocates.
Keep the COCA interface open, as we’ll soon need it again.

Idioms

The meaning of collocations is essentially compositional and transparent. Thus, just like the endocentric compounds we investigated earlier, we can infer their meaning from the parts parts. Idioms, in contrast, tend to be opaque, similar to exocentric compounds. Their exact meaning has become fixed, and is usually almost completely unchangeable, unless one again changes some parts for a particular effect, and thus creates something marked again.

kick the bucket, buy the farm = ‘die’
pull s.o.’s leg, take the mickey = ‘make fun of s.o.’
have lost one’s marbles/have a screw loose = ‘have gone mad’

As you can see from the short list above, sometimes idioms may have some ‘universal’ background, e.g. that something is missing/lacking/not working properly, moving in the wrong direction or not at all, etc.

In COCA, first click RESET to clear all settings.
Type ‘[kick] * [nn*]’ (without the quotes) into the search box.
Investigate the individual results and try to determine whether they may be idioms or refer to literal acts of kicking.

Meaning Change

The meanings of words can also change over time. Thus, for instance, the original meaning of the word knight used to refer to a farm servant, as it still does in German. Some time towards the end of the Old English period, it then seems to have changed its meaning to refer to a ‘nobleman’ and ‘warrior’, while today obviously the latter part of the meaning is also absent. As in this example, often when such a change in meaning occurs, it doesn’t only affect the denotative meaning of the word, but also its connotation(s), as the change in the semantics of the word knight is clearly one from a noun indicating a low position, and of little esteem, to one of high respect and honour, where the latter can especially be seen in the modern title.

This change can essentially affect all (content) word classes. To verify this, let's take another look in the COCA.

First, type in ‘wicked [n*]’ to find all occurrences of the adjective followed by a potential noun collocate.
Take a look at the different adjective–noun combinations to see whether all of them are either wholly positive or negative, or can alternatively be one of the two.
As usual, keep some notes, together with appropriate examples.
Next, repeat the same thing for terrific.

As you’ll hopefully have observed, both adjectives may have positive or negative connotations. As a matter of fact, wicked was originally used to describe bad people, as can still be observed in the collocation wicked witch, which we often find in fairy tales. Some time in the latter half of the 20^th century, though, it started increasingly to be come used with a more positive connotation, meaning something like ‘clever’ and/or ‘amazing’. Something very similar happened with terrific, which was originally closer to the meaning of ‘terrifying’ or ‘terrible’, and often tended to collocate with negatively connotated nouns, such as accident, while, today, it has predominantly come to mean something positive, as in e.g. a terrific idea.

Sources & Further Reading:

Sinclair, J. (1991). Corpus, Concordance, Collocation. Oxford: OUP.

Stubbs, M. (1995). Collocations and Semantic Profiles: On the Cause of the Trouble With Quantitative Studies. Functions of Language 2, 1. pp. 23-55.