2.1.1 Formal Criteria

2.1.2 Syntactic Criteria

2.1.3 Structural Criteria

2.1.4 Semantic Criteria

2.2.1 Compounds versus Simplex Lexemes

In the majority of cases, the distinction between compounds and simplex lexemes should not pose any practical problems, as a simplex like tree will rarely provoke uncertainty regarding potential categorisation as a compound. However, there are two types of exception: on the one hand, so-called fossilised compounds (Dressler Reference Dressler, Libben and Jarema2005: 40) or obscured compounds (Götz Reference Götz1971) are no longer recognisable as compounds, so that e.g. lord and lady cannot be analysed into constituents in present-day English anymore. As a consequence, one may argue in favour of their classification as simplex lexemes from a synchronic perspective. Conversely, unmotivatable but transparent lexemes (cf. Sanchez Reference Sanchez2008: 87) could theoretically be analysed into free pseudo-constituents, e.g. forget into for+get. This is for example the case of mush+room, a popular etymological interpretation of the French loanword mousseron (cf. Oxford English Dictionary [OED] s.v. mushroom). For the purposes of the present study (cf. also 4.1), the concept of the compound is therefore restricted to compounds which can be analysed into motivating – and thus semantically relevant – constituents in present-day English. While excluding pseudo-analyses such as forget and mushroom, the present approach includes all compounds whose parts appear morpho-semantically relevant, without consideration of their actual etymological origin.

2.2.2 Compounds versus Derivatives

In the majority of cases, the distinction between compounds and affixations is relatively unproblematic: compounds (e.g. pen friend) are formed from freely occurring constituents, and derivatives (e.g. befriend or friendship) contain at least one bound lexical affix. However, the definition of compounds cannot rely solely on the absence of lexical affixes, since some compounds (e.g. ozone-friendly) contain prefixes or suffixes within the compound’s constituents. Yet this is not the only problem with regard to the dividing line between compounding and derivation:

1. 1. The status of a particular morpheme as a prefix is not always easy to determine, because some prefixes (e.g. after-) are formally and semantically identical with free morphemes (in this case, the preposition after). For the sake of consistency, constructions containing the affixes listed in Table A.8 on the highest level of analysis are therefore considered prefixations rather than compounds in the present study (unless the meaning of the morpheme in question is different from the meaning of the affix, e.g. in the case of postcard or adland, which clearly refer to mail and advertising rather than to the meanings of the listed affixes).

2. 2. Compounds containing one or more combining forms of Greek or Latin origin (Plag Reference Plag2003: 74), such as the neoclassical compound biology, are special in that their constituents cannot occur as free lexemes in their own right. Since this contradicts the requirements of the compound definition at the beginning of this chapter, such items are not considered compounds in the present study.

3. 3. The same is true of lexemes of the type cranberry, which contain morphs that are unique in the language (cf. Aronoff Reference Aronoff1976: 15). Although berry is a noun in its own right, cran never occurs on its own with the meaning it has within the complex word (i.e. not the homonymous Scotch form of crane recorded in the OED). Once again, this contradicts the vital requirement for compound status, “the declaration of independence” of a construction’s constituents (Bauer Reference Bauer, Dressler, Kastovsky, Pfeiffer and Rainer2005: 97).

4. 4. Compound-final man forms a large number of compounds with solid spelling (e.g. fireman, policeman) and has relatively general semantics, with its meaning corresponding to little more than the suffix –er. It can therefore be considered an affixoid (cf. Bußmann Reference Bußmann2002 s.v. Affixoid, Suffixoid), but since the important difference to suffixation is that man also occurs as a free lexeme in its own right, such constructions are regarded as regular compounds here.

5. 5. The absence of lexical affixes on the highest level of analysis is a particularly important criterion for the distinction between compounds and affixations. It plays a role in the categorisation of a number of borderline cases, particularly constructions ending in -ed or -er (e.g. bite-sized and do-gooder). Since bite and sized occur as lexemes in their own right in the English language and are also listed in the OED, in contrast to the verb ^?to bitesize (from which bite-sized could have been derived), the complex adjective bite-sized is conferred compound status. The potential constituent ^?gooder, by contrast, does not seem to exist as an English word and is not listed in the OED, which makes the derivation of do-gooder from the existent phrase to do good by means of suffixation the more plausible choice. However, even if all constituents of a potential compound exist as individual words (e.g. black and marketeer), the semantics of their combination need to be taken into account: since black marketeer does not refer to a marketeer with black skin but to someone selling objects on the black market, it is considered a suffixation of a phrase, i.e. [black market] + -eer, and not a compound. In the literature, the term synthetic compound is sometimes used in order to accept as compounds constructions which combine an initial noun with a final deverbal noun whose status as a word in its own right is doubtful: thus goer in church+goer and swallower in sword+swallower are “possible, but not established English words”, which “function as building blocks in word-formation” through the simultaneous application of derivation and compounding (Booij Reference Booij2007: 92).

Usually, derivatives are spelled solid, which should simplify the discrimination between compounds and derivatives, but there are some exceptions: we find hyphenation in some prefixations (e.g. co-operate or re-cover), particularly if there are orthographic or semantic reasons, such as the avoidance of sequences of identical letters or an unusual meaning (cf. 5.1.1), and also in some suffixations (e.g. bell-less, shell-like; cf. Ritter Reference Ritter2005a: 56) – but no spacing.

2.2.3 Compounds versus Other Word Formations

In most cases, the difference between compounds and other types of word formation beside derivatives is relatively clear, as demonstrated by these classical examples:

• Compounds differ from acronyms in that the latter (e.g. BBC) are usually capitalised throughout and consist of constituents which have been shortened so extremely (British Broadcasting Corporation) that they cannot be considered recurring free lexemes anymore.

• Blends such as brunch also contain constituents that have been shortened in such a way that they do not meet the requirement of representing recurring free lexemes (breakfast lunch).

• Clippings (e.g. lab from laboratory) as shortenings are unlikely to be confused with the more complex compounds. When clippings occur as parts of potential compounds, however (e.g. language+lab; cf. 2.5), one may disagree whether such constructions are better classified as compounds consisting of two free constituents (language and lab) or as clippings of a longer compound (language+laboratory). Whenever the clippings in such constructions recur as free words in English, the present study assigns compound status to them, because they fulfil the requirements of the compound definition (cf. 2.6).

• Back-formations (e.g. to edit from Latin editor; cf. OED) are not necessarily complex – but when they are, e.g. in the case of to baby-sit (from baby-sitter), the line is difficult to draw. Since they consist of free lexemes and have no affixation on the highest level of analysis, back-formations are not singled out by the compound definition used here (cf. 2.6). However, that is not necessary either, because back-formations are not distinct from compounds in their structure synchronically; only historically (Huddleston and Pullum Reference Huddleston and Pullum2012: 286). A synchronic account of compounding therefore permits a certain extent of overlap between the two categories. As long as the result is compatible with the compound definition used, a compound may have undergone various word formation processes (cf. also Huddleston and Pullum Reference Huddleston and Pullum2002: 1660).

• Conversions are frequently simple lexemes, e.g. the verb to bottle, which goes back to the noun bottle (cf. e.g. Sanchez Reference Sanchez2008: 93–94). When phrases or sentences consisting of more than two components are transformed into a construction with a single part of speech (e.g. in an I-don’t-care-what-you-do attitude), this could be regarded as an instance of conversion. The present study will, however, follow the relatively common view that hyphens or concatenation in such constructions are an indication of a unified idea (cf. also Schmid Reference Schmid2011: 122) and categorise them as phrase compounds (cf. 2.5.2, 5.2 and Huddleston and Pullum Reference Huddleston and Pullum2002: 1660).

2.5.1 Length

Many English compounds – in any case the most typical ones – contain only two constituents (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1567; Schmid Reference Schmid2011: 121), but each constituent of a determinative compound (cf. 2.5.2) may itself be a compound, e.g. [[motor + cycle] + [outlet + store]], and the existence of longer compounds seems to be generally accepted in the literature. Coordinative compounds (e.g. secretary-treasurer-editor or Metro-Goldwyn-Mayer; cf. Huddleston and Pullum Reference Huddleston and Pullum2002: 1648 and 2.5.2) can also consist of more than two constituents, and phrase compounds such as love-lies-bleeding (cf. 2.5.2) even have more than two constituents by definition. According to Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985: 1567), English compounds can involve any number of constituents, but while there is no clear cut-off point, readers seem less prepared to accept a construction as a compound with growing complexity (Donalies Reference Donalies2003: 78). As a consequence, some scholars may disagree with Plag’s (Reference Plag2003: 133) example university teaching award committee member or Adams’ (Reference Adams2001: 79) claim that UK film industry task force appointment controversy represents a compound.

2.5.2 Structure

Table 2.2 summarises the various compound categories recognised in the present study from the perspective of compound structure. It draws on Adams (Reference Adams2001: 82), Bauer (Reference Bauer1983: 30–31, 212–213, 233), Marchand (Reference Marchand1969: 11–127, 380–389) and Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985: 1570–1578).

The status of phrase compounds is frequently disputed in the literature (e.g. by Meibauer Reference Meibauer2003: 185; Adams Reference Adams2001: 3; Plag Reference Plag2003: 136), because they are supposedly not formed according to the usual rules of word formation and rather similar to conversion (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1563, 1569). However, other accounts (e.g. Bauer Reference Bauer1983: 206–207) do recognise them as a type of compound. Among the small number of established phrase compounds, there are many family terms ending in -in-law, plant names (e.g. forget-me-not) and coordinated constructions (e.g. bread-and-butter; cf. Schmid Reference Schmid2011: 133–134). Since phrase compounds in general do not contradict the compound definition adopted here (cf. 2.6), and since some phrase compounds cannot be syntactic due to their ill-formedness (cf. the missing article in a pain-in-(the-)stomach gesture; Bauer Reference Bauer1983: 207), phrase compounds are also recognised as a subcategory of compounds in the present study.

The status of genitive compounds such as bull’s-eye (cf. also 5.1.3.1 and 5.5.1.2) is by no means undisputed, either (cf. Sauer Reference Sauer and Bammesberger1985: 309): they are frequently denied compound status (cf. Bauer Reference Bauer1983: 240–241) due to the fact that their constituents are linked by inflection, which makes them relatively phrase-like. However, interruption and reordering may result in semantic changes compared to phrases, e.g. if the compound bull’s-eye ‘target’ is contrasted with the modified phrases a bull’s blue eye and the eye is a bull’s. As a consequence, the approach adopted here follows Adams (Reference Adams2001: 80), who recognises a compound category “[n]oun-genitive s + noun”, whose members need not necessarily be as idiomatic as the example given earlier, e.g. potter’s wheel.

By contrast, a number of categories considered compounds by other accounts of English compounding (cf. the sources of Table 2.4) were not generally included, as they might contradict the preliminary definition:

• Neoclassical compounds or combining form compounds such as biology consist of bound roots (bio-, -logy) by definition, which contradicts the requirement of freely occurring constituents for the whole category.

• Rhyme-motivated compounds consist of two rhyming elements (e.g. hoity-toity). According to Bauer (Reference Bauer1983: 213), one of these may not exist independently, and judging from his examples, this may extend to both parts. While this contradicts the present approach’s requirement of free occurrence, rhyme-motivated compounds are accepted when their constituents occur on their own, e.g. in brain-drain.

• Ablaut-motivated compounds consist of two elements differing only in their stressed vowel, e.g. shilly-shally. If at least one element does not occur freely (e.g. the first part of wishy-washy; cf. OED), they cannot be accepted as compounds here – but if both constituents do (e.g. riff-raff; cf. OED), such items are recognised as compounds.

• Clipped compounds contain one or more clippings, e.g. optical art or situation comedy (Bauer Reference Bauer1983: 233–237). Since the clipping must have taken place after the compounding process in the forgoing examples (cf. 2.2.3), they are not compounds according to the present approach. However, if the constituents of such constructions occur on their own, usually as informal clippings (e.g. in language laboratory or parachute jump), they are considered compounds here.

Last but not least, the existing compound categories from the literature could be complemented and/or refined by the following potential structural compound types:

• By analogy to Aronoff’s (Reference Aronoff1976: 15) cranberry morph, the term cranberry compound could be coined to designate constructions such as cranberry, which contain constituents that are unique in the language but at the same time contrast with compounds containing two (or more) free constituents, e.g. black+berry or blue+berry. Since cran does not occur on its own in the language (except in linguistics texts pointing out this fact, as in the present sentence, which would skew any corpus search), this category cannot be part of the compound definition used here but might complement other types of approach.

• The categories of copulative compounds and dvandva compounds could be further refined by considering some recurring but descriptively neglected patterns:

  1. – In dictionary titles, the order of the constituents expresses directionality: thus English–Irish is interpreted to refer to a dictionary providing translation equivalents of English words into Irish, whereas the opposite is the case of Irish–English. The relation expressed is thus ‘from… into… ’.

  2. – In addition to the meaning component ‘against’, the order of the constituents in football matches expresses that the match takes place in the first of the two locations (e.g. Scotland in the Scotland–France match)

  3. – In the term “yes-no interrogative clauses” (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 724), yes and no are connected by an ‘or’ relation instead of the usual ‘and’ relation found in copulative and dvandva compounds.

2.5.3 Part of Speech

The classification of compounds according to the parts of speech involved in their formation is of particular interest with regard to spelling conventions. The overview in Table 2.4, which draws heavily on Bauer (Reference Bauer1983: 201–216) and occasionally on Adams (Reference Adams1973, Reference Adams2001), Aarts (Reference Aarts2011: 34–35) and Tournier (Reference Tournier1985: 113–119), seeks to provide examples for all three spelling variants when these are attested as examples in the reference works on word formation or in the dictionaries used in the present study (particularly MED and LDOCE; cf. 4.1). Table 2.4 shows the large variety of compound types in English. It lists compounds with a maximum of three constituents and thus more types than most other accounts of English compound formation but is still far from being exhaustive. Usually, a single example is given, except in order to draw attention to large differences between the examples. Additional compound types were added whenever the compounds encountered in the dictionaries could not be classified into any of the existing categories. The names of some categories (e.g. gerund + n or n + deverbal noun) were modified (in this particular case into n-ing + n and n + n-ing), and numbers were classified as numerals rather than adjectives. Since the part of speech of compound constituents is often difficult to determine, the empirical study described here uses a very limited set of parts of speech, which subsumes all the grammatical parts of speech under a hyperonymous concept (cf. 5.6.1). In Table 2.4, however, an intermediate approach between the more traditional part-of-speech classification used for the whole compounds and that followed for the constituents in the empirical study is applied in order to do justice to the more detailed constituent-based compound type classifications found in the literature. Table 2.3 lists the parts of speech, phrases and clauses that are distinguished in Table 2.4. In addition, the genitive and the word and are used for recurring patterns.

Note that in Table 2.4, the simultaneous presence of open and solid spelling (e.g. in the triconstituent compound Wellington airport) is placed in the middle of the scale for want of a better location, whereas combinations of hyphenation with one of the other two types are situated in more iconic positions.

The listing of part-of-speech-based compound types in Table 2.4 follows Marchand’s (Reference Marchand1960a: 20) view that “[c]ompounding occurs in all word classes” by including compound verbs (which are e.g. not considered by Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985), compounds formed from grammatical morphemes (which are denied compound status e.g. by Schmid Reference Schmid2011: 128) as well as compound adverbs, determiners, numerals and interjections. However, the comparison between the compounds found in the literature with the parts of speech in Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985: 67–77) suggests that some parts of speech are never compounded, namely modal verbs, primary verbs and the members of the category ‘unclassified’ (i.e. the negative particle not and the infinitive marker to). Nonetheless, patterns which are missing from Table 2.4 are not automatically impossible (for a discussion of potential word formations, cf. e.g. Burgschmidt Reference Burgschmidt, Brekle and Kastovsky1977; Bauer Reference Bauer2001): as a considerable number of patterns in Table 2.4 are based on the empirical study’s compounds with identical spelling in five to six dictionaries, future research is likely to yield even more compound types to add to the list.

2.5.4 Spelling

When separate lexemes or grammatical words are combined into compounds, a distinction has to be made in writing which is not required in oral speech (cf. Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1614): while varying vowel and consonant length, different degrees of loudness or pauses of different length may blur word or constituent boundaries in spoken language (e.g. between a nice drink and an ice(d) drink; Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1614), the “visual indicators of word limits” permit no gradience, and writers of English are usually forced to make an absolute decision between “total separation, hyphenation, and total juxtaposition” (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1614), the three main types of English compound spelling.

The constituents of so-called open compounds (Merriam-Webster 2001: 99) are separated by one or more spaces, depending on the number of constituents (e.g. cable television and central nervous system). Alternative terms are spaced compound (Juhasz, Inhoff and Rayner Reference Juhasz, Inhoff and Rayner2005) or separated compound (Huddleston and Pullum Reference Huddleston and Pullum2002: 1759–1760), the first of which seems to be more common. Since – as we have seen – the “status of spaced two-word segments is uncertain” (Sundby Reference Sundby, Ramisch and Wynne1997: 225), some accounts of English compound spelling (e.g. Morton Ball Reference Morton Ball1951: 3) and some dictionaries from the nineteenth and early twentieth centuries denied compound status to open sequences (cf. Morton Ball Reference Morton Ball1939), but most present-day treatments of compounding accept and include open compounds.

In hyphenated compounds (Merriam-Webster 2001: 99), the constituents are separated by one or more hyphens, depending on the number of constituents (e.g. hunter-gatherer and forget-me-not). The alternative terms hyphened compound, hypheme (Morton Ball Reference Morton Ball1951: 3), half-compound and occasional compound (Sundby Reference Sundby, Ramisch and Wynne1997: 225) seem to be uncommon. While some psycholinguistcs studies (e.g. Juhasz et al. Reference Juhasz, Inhoff and Rayner2005) restrict compound spelling to the distinction between open and solid spelling and do not consider hyphenated compounds, this is relatively unusual.

Solid compounds (Merriam-Webster 2001: 99), such as software and twentysomething, constitute orthographic words and thus uninterrupted sequences of letters (cf. Plag Reference Plag2003: 4). They are the only type of compound that is not excluded by any compound definition. Strumpf and Douglas (Reference Biber1988: 52) use the alternative term closed compound, which is common compared to less established alternatives such as juxtaposed compound (Huddleston and Pullum Reference Huddleston and Pullum2002: 1759–1760), solideme (Morton Ball Reference Morton Ball1951: 3) or absolute compound (Sundby Reference Sundby, Ramisch and Wynne1997: 225).

Of the three spelling variants, solid spelling is the only one that can be considered immaterial. The hyphen obviously has a form, and although the gap in open compounds is not filled, it occupies a space which could have been used otherwise. The three types of compound spelling in English make use of the application or non-application of two principles:

1. a) concatenation vs. non-catenation

2. b) use of a hyphen vs. non-use of a hyphen.

The combination of these two options with their two possible values (i.e. the presence or absence of each feature) yields the three most common types of compound spelling in English, namely

A fourth variant can be derived from the logical combination of these parameters, namely non-concatenation + use of a hyphen, e.g. boy-friend. While this variant does not seem to occur on its own in the English language, there are contexts in which it may be used, namely in combinations of identically structured hyphenated compounds, the first of which is incompletely realised on the formal level because of the ellipsis of the shared second constituent, e.g. in car-owners and ship-owners (Morton Ball Reference Morton Ball1939: 96) or in boy- and girl-friends. Longer sequences are conceivable, with the proviso that the shared second constituent is always retained in the last of the compounds (cf. Morton Ball Reference Morton Ball1939: 96), e.g. in These are paragraph-, sentence-, or clause-boundaries (Nunberg Reference Nunberg1990: 69). However, this phenomenon, which is termed “floating hyphens” by Butcher (Reference Butcher1992: 154) and “elliptical compounds” by Morton Ball (Reference Morton Ball1939: 96), is frequently regarded as undesirable (cf. Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1347), and Butcher (Reference Butcher1992: 154) suggests its avoidance by rewording.Footnote ³

While the present study focuses on the three most common ways of spelling English compounds, alternatives using punctuation marks other than the hyphen exist or existed in the past:

• Between the fourteenth and eighteenth centuries, the equals sign < = > was sometimes used instead of a hyphen, at least in end-of-line hyphenation (McDermott Reference McDermott1990: 13). Especially in the United States, it is still used by proofreaders to “signify the instruction to insert a hyphen” and therefore now corresponds to an alternative variant with a different pragmatic focus (Clark Reference Clark1990: 196).

• A conventional (but not extremely frequent) alternative to the hyphen is the oblique stroke or slash < / >. Since slashes are not normally flanked by spaces (although usage may vary), Huddleston and Pullum (Reference Huddleston and Pullum2002: 1731) define them as part of word-level punctuation, i.e. “the marking of word boundaries and the use of punctuation marks … within a word”. Slashes are used in very specific types of compound, e.g. some noun+noun compounds describing a double title or function, such as bar/restaurant (Merriam-Webster 2001: 100–101). Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985: 1570) extend their use to coordinate compounds without restricting part of speech and give adjectival examples, e.g. aural/oral (approach). Interestingly, the slash has an additive function rather than its usual alternative function in these compounds. In addition, the oblique stroke may be preferred over the hyphen “where one or more elements consist of more than one word, e.g. Bedford/Milton Keynes boundary” (Butcher Reference Butcher1992: 152).

• Single or double quotation marks may be used in the spelling of long English compounds (Morton Ball Reference Morton Ball1951: 6), e.g. those including a film title such as “Gone with the Wind” remake (Google search, June 2017). The pairwise occurrence of quotation marks (Meyer Reference Meyer1987: 4–6) opens a unified slot, whose closing is indicated by the second part of the pair. While quotation marks are clear delimitators of long components, they combine with spaces or hyphens and thus cannot be considered a completely new spelling variant.

• In programming languages, underscores are often used in order to connect the parts of a lexical entity, e.g. window_id_format (cf. Venezky Reference Venezky1999: 41).

From a systemic and logical point of view, any punctuation mark could be used to link the constituents of compounds – except those which usually indicate syntactic boundaries and need to be followed by a space. However, even though commas, full stops, colons, semi-colons, question marks and exclamation marks break up the unity of the compound, which would seem to prevent their use, there are some exceptions even here: thus the original compound list from the Longman Dictionary of Contemporary English (cf. 4.1) contained the two comma-separated items all-singing, all-dancing and two up, two down, and in the American spelling of Mr. Right, a compound-internal full stop occurs at the end of the abbreviated first constituent. Furthermore, in metalinguistic language use, compounds with undetermined spelling may be concatenated with a question mark between the constituents (Rot?wein ‘red?wine’; Jacobs Reference Jacobs2007: 54).

Yet another possible means of marking constituent boundaries is the modification of the standard font: thus it is customary to italicise foreign phrases within compounds (cf. Fowler’s Reference Fowler1921: 10 example “an ex officio member”), and superscript is occasionally combined with solid spelling when chemical elements and figures are conjoined (e.g. Sr⁹⁰; GPO Style Manual 2008: 82).