Derivation is a widely-used way of creating new words in human languages, but it is not the only one. As mentioned at the beginning of Section 5.2, we can also combine two (or more) bases, a process called compounding.

The bases involved can belong to every word-class so the first way in which we can classify basic types of compounds is in terms of the word-class of their constituent parts — for example, noun-noun compound (lawman), adjective-noun compound (blackbird), or verb-adjective compound (failsafe). Table 5.7.1 gives examples of compounds involving the three major word-classes, but keep in mind that there are also compounds involving adverbs (wellbeing, broadcast), prepositions (for-profit, downstream), and even pronouns (me time, she-wolf).

Note that not all of these compound-types are equally common — there are only a handful of cases each of established verb-adjective compounds, verb-verb compounds, adjective-verb compounds or adjective-adjective compounds, but hundreds of cases of noun-noun compounds and adjective-noun compounds, so we will focus on these types in the following (but that does not mean that the other types are not interesting — on the contrary, there is much less research on them, so in a sense they are more interesting)!

Before we continue however, here is an important reminder concerning spelling: remember that spelling is almost never a useful indication of linguistic structure. In particular, whether a sequence of words is spelled as a single orthographic word, with a hyphen or with whitespace between the roots has little to do with whether the sequence is a compound. English is different, in this respect, from other Germanic languages, where compounds tend to be spelled as single words, which sometimes leads to the perception that these languages have much longer words than English. Take the longest documented actual word from German, shown in (1):

It consists of six bases, one of which, Rindfleisch ‘cattle’ is itself a compound, and three of which are nouns derived from verbs using the suffix {-ung}; in addition, four of the bases contain a special affix referred to as Fugenmorphem ‘gap morpheme’, which sometimes marks words that are part of a compound. Totalling 67 letters (corresponding to 52 phonemes organized into 20 syllables), it seems much longer than any English word could ever be. But consider sentence in (2) from an academic paper, which contains the compound noun shown in (3):

The noun in (3) is very similar in length to the German noun in (1): it consists of six bases, three of which are derivatives with one affix (commun-ity, busi-ness, recruit-ment), and one with two affixes (advert-ise-ment), totalling 60 characters (counting the whitespaces), corresponding to 49 phonemes. The only difference is that, in English, compounds are often spelled with whitespace between the bases — but that is an orthographic convention that has nothing to do with morphology (or phonology).

Compounding is a recursive process — you can create one or more compounds and then use them in the creation of further compounds, and repeat this process until you get the word you want:

1. community + college
2. business + faculty
3. [community college] + [business faculty]
4. [ [community college] [business faculty] ] + recruitment
5. [ [ [community college] [business faculty] ] recruitment] + advertisement

So, while English, German and many other languages allow the creation of very long compounds, these compounds are created in a step-wise fashion, with every step combining just two bases.

The most important property of compounds is the relation between these two bases. Normally, this relation can be described in terms of a head, which determines what general type of entity, process or property the compound refers to and what word class it belongs to, and a modifier, which specifies the specific type of entity, process or property but does not have an influence on the compound’s word class. Such compounds are sometimes referred to as endocentric compounds, because their center — the element that determines their form and meaning — is contained within the compound (endo– means “in(side)”.

In English, the head in endocentric compounds is always the last element of the compound. For example,

• greenhouse refers to a type of house, not a type of green, and it is a noun (like house), not an adjective (like green);
• camera-shy refers to a particular type of being shy, not a particular type of camera, and it is an adjective (like shy), not a noun (like camera);
• to lipread refers to a particular type of reading, not to a particular type of lip, and it is a verb (like read), not a noun (like lip).

There are a few exceptions where the modifier follows the head. Most of these are French loanwords (in French, adjectives mostly follow the noun), such as attorney general (the government’s own attorney), court martial (a military court), or force majeure (a higher force) but some were formed on this model within English, such as president elect or sum total.

There are also compounds which do not seem to have a head-modifier structure at all:

Looking at the compounds in (4a), note that owner-occupier is both the owner and the occupier of their house or apartment, a singer-songwriter writes and sings their songs, spacetime is both space and time, and so on. It seems that both parts of the compound are head and modifier at the same time. Such compounds (which are rather rare) are sometimes referred to as dvandva compounds (dvandva means ‘pair’ in Sanskrit, where such compounds are more common).

Turning to the compounds in (4b), a redhead is not a type of head (or a shade of red), it is a word for a person (typically a woman) with red hair, a red-eye is not a type of eye, it is a word for a flight taken very early in the morning, a scatterbrain is not a type of brain but a person who behaves as though their brain was scattered all over the place, etc. In (4c), likewise, a cutthroat is not a type of cut or a type of throat, but a person who cuts other people’s throats, a killjoy is a person who kills other people’s joy, a pick-pocket is a person who picks other people’s pockets, etc.

Both of these types of compounds are sometimes called exocentric compounds, because they do not seem to contain a head that determines the type of entity they refer to and their part of speech. The idea is that their head is somewhere outside of the actual compound — that, for example, redhead has an invisible structure like [[red headed] person] or cutthroat has an invisible structure like [[throat-cutting] person].

With respect to the compounds in (b), this analysis is not really useful. First, notice that in all cases, the second part of the compound does determine the word class, and that in all cases, the first part does seem to modify the meaning of the second part to make it more specific. The difference to the more typical compounds shown in Table 5.7.1 is simply, that the word as a whole does not literally refer to whatever the second part would refer to, but to a related entity. A red-head is a ‘person with a red head’, a red-eye is a ‘flight that causes red eyes (because you don’t get enough sleep’ etc. But this shift in meaning is not specific to compounds, it also occurs with simple words — we can talk about wanting to hire a few good heads or brains for a team, meaning, of course, people with good heads/brains, we can talk about a vintage car being the pride and joy of its owner, meaning, of course, that it is an ‘entity causing pride and joy’, and so on. This process is referred to as metonymy, and the compounds in (b) are plainly and simply endocentric compounds that are used metonymically.

The compounds in (c) could be called exocentric, if we really want to. In these cases, the second part of the compound really does not seem to be a head: it does not specify what type of entity the compound refers to, the first part of the compound does not seem to modify the second part, and the second part does not determine the word-class of the compound. However, the idea that these compounds have a head somewhere outside of the word itself is a bit weird, and we would have to be able to say what that head is. It could be a phonologically empty noun with the meaning ‘person’, so that the structure of cut-throat would be [cut-throat Ø_NOUN], but we already talked about the dangers of positing phonologically empty elements, and this structure would still not explain how the element cut-throat itself is formed — normally, in English, compounds consisting of a verb and a noun functioning semantically as its object have the form [N V], as in baby-sit, trouble-shoot, back-stab, etc.

Instead, we could simply treat the different types of compounds as resulting from different word-formation rules. Unlike the WFRs for derivational and inflectional affixes, such rules would not include any morphemes, but this is not a problem: we have already seen that it is possible to have WFRs without any morphemes in the case of conversion.

For “endocentric” noun-noun compounds, this rule could look like this (similar rules can be posited for the other types of “endocentric” compounds):

This WFR only assigns a very general meaning to such compounds: the modifier somehow characterizes the head. The specific relationship is left open — if speakers create a novel compound, the relationship is determined by context, in conventionalized compounds, it is determined by the past usage of the word. For example, lawman refers to a sheriff or marshal, but theoretically, it could refer to any man with a relationship to the law, such as a judge or lawyer or a politician who passes laws; sunflower refers to a flower that looks a bit like a child’s drawing of a sun and that turns its head to the sun at all times of the day, but it could also refer to a flower growing on the surface of the sun (in a world where that were possible), or a flower that needs sunlight (i.e., any flower). Thus, like the WFRs for derivational affixes we saw in the preceding section, the WFRs for compounds only determine the range of possible meanings, but once a word is established, its meaning will not cover the whole range but be much more specific.

For dvandva compounds, the WFR would look like this:

And for the cutthroat type of “exocentric” compound, the WFR would look like this:

Unlike the first two WFRs, however, this last rule does not seem to exist in current English: all compounds of this type are fairly old and it does not seem possible for form new words based on it. For example, a bus driver is not a *drive-bus, a math teacher is not a *teach-math, a lawn mower is not a mow-lawn, a drug dealer is not a *deal-drug, etc.

← Previous section Next section →

CC-BY-NC-SA 4.0, Written by Anatol Stefanowitsch