Earlier, the point was made that the reliable interpretation of complex words (whether derived or compounded) depends on an expectation that meaning should go hand in hand with structure. So far, this expectation has been fulfilled (provided we ignore words with totally idiosyncratic meanings). The meaning of a complex whole such as unhelpfulness or holiday car trip is built up out of the meanings of its two constituent parts, which in turn are built up out of the meanings of their parts, and so on until we reach individual morphemes, which by definition are semantically indivisible. However, we will discuss a few instances where this expectation is not fulfilled. Discussing these instances leads us to the question of whether a unit larger than a word (that is, a phrase) can ever be a constituent of a compound word. There is no agreed answer to these questions, but the kinds of English expression that give rise to them are sufficiently common that they cannot be ignored, even in an introductory textbook.

Consider the expression nuclear physicist. Its structure seems clear: it is a phrase consisting of two words, an adjective nuclear and a noun physicist. So, if the interpretation of linguistic expressions is always guided by their structure, it ought to mean a physicist who is nuclear. Yet that is wrong: a physicist is a person, and it makes no sense to describe a person as ‘nuclear’. Instead, this expression means someone who is an expert in nuclear physics. So we have a paradox: in terms of morphology and syntax, the structure of the expression can be represented by the bracketing [[nuclear] [physicist]], but from the semantic point of view a more appropriate structure seems to be [[nuclear physic-]-ist]. We thus have what has come to be called a bracketing paradox. In this instance, the meaning seems to direct us towards an analysis in which the suffix -ist is attached not to a word or root but to a phrase, nuclear physics. Is it possible, then, for a word to be formed by adding an affix not to another word but to a phrase?

A similar problem is presented by the expression French historian. This has two interpretations: ‘historian who is French’ and ‘expert in French history (not necessarily a French person)’. The first interpretation presents no difficulty: it is the interpretation that we expect if we analyze French historian as a phrase, just like green house (as opposed to greenhouse). This implies a structure [[French] [historian]]. However, the second interpretation seems to imply a structure [[French histori-]-an], in which a phrase is combined with an affix. We are faced with a dilemma. Should we acknowledge the second structure as the basis for the second interpretation? Or should we say that, with both interpretations, the structure of the expression is the same (namely [[French] [historian]]), but that for one of the interpretations this structure is a bad guide? Without putting forward a ‘right answer’, I will mention two further observations that must be taken into account – two observations that, it must be said, pull in opposite directions.

Examples of other adjective–noun combinations whose meanings diverge from their structure are plastic surgeon (denoting not a kind of doll, but an expert in cosmetic surgery) and chemical engineer (denoting an expert in chemical engineering, not a person who is ‘chemical’). These differ from nuclear physicist, however, in that there is no way of bracketing them so as to yield a structure that corresponds closely to the meaning. So, even if the meaning of nuclear physicist can be handled by the paradoxical bracketing [[nuclear physic-]-ist], no such device is available for plastic surgeon and chemical engineer. This means that some other way of reconciling their structure–meaning divergence must be found. It does not matter for present purposes how that reconciliation is achieved. What does matter is that, however it is achieved, the same method will presumably be available to handle nuclear physicist, and also French historian in the sense ‘expert in French history’. This weakens the argument for recognizing a ‘semantic’ bracketing distinct from the ‘grammatical’ one. Rather, we can simply say that, for example, [[French] [historian]], so structured, has two interpretations.

Those examples all involve derivation. What about any apparent bracketing paradoxes involving compounding? Consider the item French history teacher. In the sense ‘French teacher of history’, this is a phrase consisting of an adjective and a noun, just like French painter, the only difference being that the noun in French history teacher is the compound history teacher, just like the noun portrait painter in French portrait painter. But what about the interpretation ‘teacher of French history’? Is this a compound noun with the structure [[French history] teacher]? The trouble with that analysis is that French hístory, with its stress on history, seems clearly to be a phrase, not a word; yet, if a phrase such as French history is permitted to appear as a component of a compound word, we are faced with explaining why phrases cannot appear inside compounds generally – why, that is, we do not encounter compounds such as eventful history teacher, with the phrase eventful history as its first element, and with the meaninging ‘teacher of eventful history’, or history skilled teacher, with the phrase skilled teacher as its head. Perhaps, then, we should say of French history teacher essentially the same as what was suggested concerning French historian: it has only one structure, that of a phrase ([French [history teacher]]), even though it has two interpretations, one of which diverges from that structure.

Some implications of that analysis are unwelcome, however. Consider the expressions fresh áir fanatic and open dóor policy. Their main stress is on air and door, as indicated, and their meanings are ‘fanatic for fresh air’ and ‘policy of maintaining an open door (to immigration, for example)’. These are parallel to the meaning ‘teacher of French history’, which, we have suggested, diverges from its structure [French [history teacher]]. But, whereas French history teacher has a second meaning that corresponds exactly to that structure, fresh áir fanatic and open dóor policy have no such second meaning; one cannot interpret them as meaning ‘fresh fanatic for air’ or ‘open policy about doors’. So a bracketing such as [fresh [air fanatic]] would diverge not just from one of the meanings of fresh air fanatic, but from its only meaning!

A clue to a way out of this problem lies in comparing the actual expressions at (13) with the non-existent or ill-formed ones in (14):

(13) a. fresh air fanatic

       b. open door policy

       c. French historian ‘expert in French history’

       d. nuclear physicist

       e. sexually transmitted disease clinic

(14) a. cool air fanatic ‘fanatic for cool air’

       b. wooden door policy ‘policy on wooden doors’

       c. suburban historian ‘expert on the history of suburbs’

       d. recent physicist ‘expert on recent physics’ (not ‘recent expert on physics’)

       e. easily transmitted disease clinic

The phrases fresh air and cool air differ in that fresh air is a cliché, even if not precisely an idiom; that is, fresh air recurs in a number of stock expressions such as get/need some fresh air and get out into the fresh air, whereas there are no such stock expressions containing cool air. Similarly, French history is a cliché in that the history of France is a recognized specialism among historians; on the other hand, the history of suburbs is not recognized as a specialism to the same degree, so the phrase suburban history, though perfectly easy to interpret, is not a cliché. The same goes for open door versus wooden door, nuclear physics versus recent physics, and sexually transmitted disease versus easily transmitted disease; the first expression in each pair is an idiom or cliché, while the second is not. What we need to say, it seems, is that a phrase can form part of a compound or derived word provided that the phrase is lexicalized or in some degree institutionalized, so as to become a cliché.

From the point of view of the distinction carefully drawn between lexical items and words, this is a surprising conclusion. On the basis of the facts, it seemed that there was no firm link between lexical listing and grammatical structure. Now it appears that that view must be qualified: lexically listed phrases (i.e. idioms) or institutionalized ones (i.e. clichés) can appear in some contexts where unlisted phrases cannot. Whether we should analyze these contexts as being at the word level, so as to treat nuclear physicist and fresh air fanatic as words rather than phrases, is an issue that beginning students of word-structure should be aware of but need not have an opinion about.