1 Introduction
In 1968, Miloš Dokulil (Reference Dokulil1968: 215) placed derivation by conversion at the crossroads of morphology, syntax and lexical semantics. Barely a decade later, Dieter Kastovsky (Reference Kastovsky1977) located word-formation ‘at the crossroads of morphology, syntax, semantics, and the lexicon’, again for the intricate connections that these levels of language description manifest in word-formation.Footnote 2 This article finds inspiration in Dokulil and Kastovsky for the description of a rather neglected side of the process known as conversion,Footnote 3 namely the relationship between conversion-related pairs.
This is a relevant question because, even if conversion shares morphological, syntactic and lexical properties with other word-formation processes, it raises questions of its own as a result of the formal identity involved in conversion, and these questions do not apply in other word-formation processes. The relevance of the issue is shown by the number of interpretations available in the literature for the relationship between conversion-related words and also by the lack of a precise term or by the uncertainty of the terms of the description. This is witnessed, among others, by such major texts as Bauer, Lieber & Plag (Reference Bauer, Lieber and Plag2013: 546, our emphasis), which cites this as one of the issues that are still to be answered in conversion:
[s]omewhere in the range between homonymy and polysemy is a set of homophonous words which are generally taken to be derivationally related. The base forms or citation forms of two lexemes are homophonous, but the two lexemes belong to different word classes …, despite being semantically related.
Similar references can be found in the literature at least since as far back as Quirk, Greenbaum, Leech & Svartvik (Reference Quirk, Greenbaum, Leech and Svartvik1985: 70), who pointed out that ‘[t]here is no standard term for words which also share the same morphological form’. The same point can be found in Persson (Reference Persson, Hyldgaard-Jensen and Zettersten1988: 271):
[i]n many cases one ‘parent lexeme’ (cf. Clark & Clark Reference Clark and Clark1979) is seen to yield two or more phonologically identical forms which are not analyzable in terms of either homonymy or polysemy [and] some other solution will have to be sought for.
These and other texts show repeated attempts in the literature at the identification of the relationship between conversion-related pairs as homonymy, as polysemy, or in other terms (see section 2). As pointed out by an anonymous reviewer, it should be stated first why and whether such a relationship needs description. The formal identity inherent in conversion refers converted words to relations of formal identity, more evidently in word-based than in stem-based or root-based conversion, i.e. more evidently e.g. in English (work V vs work N) than in German (arbeit⋅en V vs Arbeit N) or Spanish (trabaj⋅ar V vs trabaj⋅o N). Otherwise, and despite the literature on the subject, it can be argued whether the relation between base and derivative in conversion is in special need of description compared with the relation between base and derivative in other word-formation processes.
The article conducts a systematic review of material that deals with conversion in English as a dynamic word-formation process whereby a base is used for formation of a derivative under the conditions of formal identity (with the usual exceptions of stress shift, final voicing or vowel reduction) and word-class contrast (Leech Reference Leech1974: 214; Lieber Reference Lieber1981: 187 et passim; Bauer Reference Bauer1983: 32; Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1520, 1558; Don Reference Don1993; Štekauer Reference Štekauer1996: 15 et passim; Vogel Reference Vogel1996: 258 et passim; Bauer & Huddleston Reference Bauer, Huddleston, Huddleston and Pullum2002: 1640; Manova Reference Manova2011: 55 et passim; Bauer et al. Reference Bauer, Lieber and Plag2013: 562–3). Comparatively few of these titles discuss the topic at issue here, so any material within the recent literature on word-formation and/or lexicology involving conversion and/or lexical relations was reviewed for potentially relevant reference to this topic.
This article does not consider views that do not involve a relationship between words, e.g. the interpretation of conversion as ‘multifunctionality’ or ‘word-class underspecification’ (originally in Whorf Reference Whorf1945; later recalled e.g. in Lipka Reference Lipka1971: 213–15; Nida Reference Nida1975: 99; Robins Reference Robins and Burchfield1987: 58 under the term ‘grammatical neutrals’ and, more recently, in Farrell Reference Farrell2001: 109, 116, 128 and Bauer et al. Reference Bauer, Lieber and Plag2013: 565–6, among others). Under the latter interpretation, these words are not the result of a process and behave as roots unspecified for word class, such that the same lexical element can surface as various arbitrary linguistic categories, i.e. as various word classes. A similar case can be made for the interpretation of these examples as transflexion (e.g. in Ivanová & Ološtiak Reference Ivanová, Ološtiak and Müller2015), described as ‘a word-formation procedure common in Slavic languages in which a new word is coined by a change of grammatical morpheme’ (Ivanová Reference Ivanová2016: 22; cf. similarly ‘dérivation immédiate’, ‘transfert grammatical’ and ‘unmittelbare Ableitung’ in Pennanen Reference Pennanen, Ringbom and Rissanen1984 and Kastovsky Reference Kastovsky and Britton1994, ‘paradigmatic derivation’ in Nagórko Reference Nagórko and Müller2015 or mere ‘inflection’ in Riese Reference Riese and Müller2015).
The contents of the article are arranged as a review in order to consider the most common interpretations of the relationship between conversion-related pairs attested in the literature. For easier reading, section 2 is arranged according to which side of conversion-related pairs can be used for the description of their relationship: form or meaning. This arrangement does not assume any formal or semantic priority to the material reviewed. Section 3 reviews types of conversion with respect to these relationships. Section 4 argues for a description in terms of none of the relations reviewed, and section 5 closes with a conclusion and a recapitulation of the relationship between conversion-related words available in the literature and of their descriptive accuracy. As a result, the article identifies an area that has been dealt with very dissimilarly, and one on which no agreement has been reached and is often unnoticed. The article brings together titles from three fields (morphology, syntax, lexicology), many of which are unknown to specialists outside their own field, and argues for a redefinition of this relation away from the homonymy/polysemy debate in which recent and authoritative titles (e.g. Bauer, Lieber & Plag Reference Bauer, Lieber and Plag2013) describe the relation.
2 The relationship in conversion-related pairs
This section reviews the relationships that are proposed in the literature for conversion-related pairs. The review considers general material, and texts in lexical semantics and in morphology, and separates the description of conversion into relationships according to formal identity (section 2.1.1) and according to morphological identity (section 2.1.2).
2.1 Relationships according to identity
2.1.1 Mere formal identity: homonymy
Conversion-related pairs are most often described in terms of homonymy. In some cases, the use of this term can be considered to stand for mere formal identity rather than for the relationship in the strict sense:
The notion of zero-morpheme … accounts for the fact that two homonymous lexemes, which are superficially identical, are in a synchronically directed relationship. One lexeme is the base, while the other one is derived from it by means of a zero-suffix. (Lipka Reference Lipka1990: 86, our emphasis; cf. similarly Adams Reference Adams1973: 38; Reference Adams2001: 100; Robins Reference Robins and Burchfield1987: 56, among many others)
Even if the use of the term is not intended as the lexical relation but as plain formal identity, the literature offers texts that view conversion-related pairs as homonymy, or as a type of homonymy (Jespersen Reference Jespersen1909–49, vol. VI: 84; Reference Jespersen1933: 43; Lyons Reference Lyons1977, vol. II: 560; Paul Reference Paul1982: 305; Lipka Reference Lipka1986: 134, 137; Reference Lipka1990: 138, 140; Bauer & Huddleston Reference Bauer, Huddleston, Huddleston and Pullum2002: 1641, 1644). Thus, Lipka (Reference Lipka1990) continues by arguing that:
it is nevertheless advisable to include zero-derivation with homonymy, since there are clear grammatical differences (word class) and there is also the close parallelism with overt suffixal derivatives (Lipka Reference Lipka1990: 140)
Some of these texts qualify the term in one way or the other, e.g. as ‘grammatical homonyms’ or ‘grammatical homophones’ (Jespersen Reference Jespersen1933: 43; Marchand Reference Marchand1963: 176; cf. also Kerleroux Reference Kerleroux1999: 99).
Various interpretations of homonymy are available in the literature. The traditional and most widely accepted view treats homonymy as the paradigmatic relation between two units, words or morphemes, which accidentally have the same phonological and orthographic form (Ullmann Reference Ullmann1962: 159; Leech Reference Leech1974: 228–30; Lyons Reference Lyons1977, vol. II: 550; Reference Lyons1981: 146; Cruse Reference Cruse1986: 80; Lipka Reference Lipka1986: 128–9; Reference Lipka1990: 75; Magnusson & Persson Reference Magnusson and Persson1986: 27–34; Persson Reference Persson, Hyldgaard-Jensen and Zettersten1988: 269–70; Reference Persson1990: 141–3, 160; van Sterkenburg Reference Sterkenburg1996: 401; Saeed Reference Saeed1997: 63;Footnote 4 Tournier Reference Tournier2007: 337; Goddard Reference Goddard2011: 22; Martsa Reference Martsa2013: 201). By ‘accidentally’ (as, e.g., in Goddard Reference Goddard2011), ‘coincidental’ or ‘causally unrelated’ (as, e.g., in Magnusson & Persson Reference Magnusson and Persson1986: 31–4; cf. also Persson Reference Persson, Hyldgaard-Jensen and Zettersten1988: 269–70), it is meant that the units are historically or morphologically unrelated (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 70), their meanings being unrelated too (Plag, Braun, Lappe & Schramm Reference Plag, Braun, Lappe and Schramm2007: 164), as emphasized in Stockwell & Minkova (Reference Stockwell and Minkova2001: 147–8). By implication, no relation can be established between homonymous words beyond the form, as lexemes are semantically, etymologically and morphologically unrelated, and the formal identity is unmotivated.
Of the conditions set in the literature for so-called absolute homonymy (semantic unrelatedness, formal identity and syntactic equivalence), only the former seems to be a necessary condition (Lyons Reference Lyons1977, vol. II: 560). This is because special types of homonymy (‘partial homonymy’) are considered when the second condition does not apply (i.e. they are homographs or homophones), or when the third condition does not apply (i.e. where two formally identical and semantically unrelated lexemes belong to different syntactic categories or word classes). Of these two less central cases of homonymy, only the latter is of relevance for this article.
Most texts appear to relate homonymy only to lexemes of the same word class, perhaps because it is in pairs of the same category where the lexical ambiguity usually associated with homonymy typically emerges. Word-class contrast is not one of the conditions for homonymy, even if some variation can be noticed in the literature in this regard. Thus, Lyons (Reference Lyons1977, vol. II: 560) argues that ‘no two lexemes can be absolutely homonymous if they are members of different parts-of-speech’.Footnote 5 Ullmann (Reference Ullmann1962: 159), Lipka (Reference Lipka1990: 138, 140) and Saeed (Reference Saeed1997: 64) accept homonymy across different word classes, and Bergenholtz & Agerbo (Reference Bergenholtz and Agerbo2014) argue that a separate type of homonymy may obtain when there is word-class contrast.
The description of conversion-related pairs as homonymy accounts successfully for the condition of formal identity mentioned in section 1, but not for the fact that ‘the two functions are nonaccidental variants of the same fundamental linguistic unit’ or for the ‘basic invariance of core meaning across these [two different syntactic] functions’ displayed in conversion (both as described by Sanders Reference Sanders, Hammond and Noonan1988: 157). The different etymological origins and the semantic contrast that are usual tests for identification of formally identical terms as separate words do not apply in conversion-related pairs:
(i) In the first test, there seems to be little discussion as to whether a new word is formed in conversion or not:Footnote 6 conversion-related pairs are morphological base and derivative with respect to each other and their formal identity is, unlike in homonymy, not unmotivated.
(ii) In the second test, because the semantic contrast in conversion usually limits itself to contrast in categorial meaning, and this contrast entails a relatedness that has even been described as senses (e.g. in Murphy Reference Murphy2010: 96 and Koskela & Murphy Reference Koskela, Lynne Murphy and Brown2006: 743; the contrast in terms of categorial meaning can be seen in the semantic patterns found, e.g. in noun-to-verb conversion as summarized, among others, in Plag Reference Plag1999: 219–20; cf. also Štekauer Reference Štekauer1996, and Bauer & Valera Reference Bauer, Valera, Bauer and Valera2005, on the interpretation of conversion in terms of what belongs in specific lexical classes or word classes). The description of homonymy across word classes depends partly on the semantic proximity perceived in the words in question, and this proximity has been valued differently in the literature: from change of meaning across the word classes involved (e.g. as in Leech Reference Leech1974: 214–16 or Akmajian, Demers, Farmer & Harnish Reference Akmajian, Demers, Farmer and Harnish2001: 28) to little difference in ‘relevant information content’ (as in Deane Reference Deane1988: 350; cf. similarly Beard Reference Beard, Spencer and Zwicky1998: 60 on transposition in general).
The description of conversion-related pairs as homonymy in the light of these tests leads to contradictions in terms. These manifest themselves, e.g. in the report of homonymous lexemes that share a ‘common meaning’ (Lipka Reference Lipka1990: 138, 140; cf. also Lipka Reference Lipka1986: 134, 137), ‘etymologically related’ homonyms (Koskela & Murphy Reference Koskela, Lynne Murphy and Brown2006: 743), or homonymy, even if ‘this entails that some of the homonymous meanings are semantically closer connected than others’ (Bergenholtz & Agerbo Reference Bergenholtz and Agerbo2014: 32). Homonymy has therefore been explicitly disqualified as a description of conversion-related pairs on these grounds in a number of texts (cf. Magnusson & Persson Reference Magnusson and Persson1986: 45–6 and Persson Reference Persson, Hyldgaard-Jensen and Zettersten1988: 277–8).
2.1.2 Morphological identity: heterosemy, homomorphy
In the second possibility under review, identity as a morphological property, the relation between conversion-related pairs is described by terms that lay emphasis on the semantic diversity associated with one form (heterosemy) or on the occurrence of one morpheme associated with various meanings (homomorphy), i.e. focusing in each case on one of the conditions for conversion to obtain.
‘Heterosemy’ has been proposed as a middle term between homonymy and polysemy (Persson Reference Persson1990: 156, 158), as Persson (Reference Persson, Hyldgaard-Jensen and Zettersten1988: 276) argues, ‘because of the “split but united” relation which is thus one of its [conversion's] chief characteristics’ and that disqualifies both homonymy and polysemy. ‘Heterosemy’ has been interpreted variously, also because Persson's (Reference Persson, Hyldgaard-Jensen and Zettersten1988) words may be interpreted as applying to the terms related by conversionFootnote 7 and also to the various senses within one derivative by conversion. Of these, the latter sense has been adopted in the literature, e.g. in Kreidler (Reference Kreidler, Cannon, Barton and Kreidler1988: 75) to define heterosemy as ‘the relation between co-hyponyms of a homophonous superordinate’, in line with Magnusson & Persson (Reference Magnusson and Persson1986: 42–5), and Persson (Reference Persson, Hyldgaard-Jensen and Zettersten1988: 276–7).Footnote 8 This view of heterosemy requires a manifest causal link between the members of the relation, such that the loss of this link results in plain homonymy even if the terms are originally related by a causal link as derivatives of the same parent lexeme (Persson Reference Persson, Hyldgaard-Jensen and Zettersten1988: 278–9; cf. similarly Magnusson & Persson Reference Magnusson and Persson1986: 36–7, 41–2). For this reason, Persson (Reference Persson, Hyldgaard-Jensen and Zettersten1988: 278–9) limits heterosemy to synchronic derivation, meaning that heterosemy refers only to where the link between the terms of the relation is clear, i.e. not obscured or lost over time.
Magnusson & Persson (Reference Magnusson and Persson1986: 43–4) and Persson (Reference Persson, Hyldgaard-Jensen and Zettersten1988: 274–6) do not use ‘heterosemy’ for the relationship between words across word classes but, e.g., for two nouns derived from the same verb (e.g. the various nouns drive derived from the verb drive). Still, the term ‘heterosemy’ has been interpreted as applicable to the relation between conversion-related pairs (Persson Reference Persson, Hyldgaard-Jensen and Zettersten1988, cited in Lichtenberk Reference Lichtenberk1991: 476; Enfield Reference Enfield, Ameka, Dench and Evans2006: 297), and has also been adapted to apply to cases ‘where two or more meanings or functions that are historically related, in the sense of deriving from the same ultimate source, are borne by reflexes of the common source element that belong in different morphosyntactic categories’ (Lichtenberk Reference Lichtenberk1991: 476). In the latter case, Lichtenberk's (Reference Lichtenberk1991: 476) interpretation does not necessarily assume that conversion qualifies as heterosemy, nor does it assume a common core sense across terms, as usually happens in conversion. At least in Lichtenberk (Reference Lichtenberk1991), ‘heterosemy’ is therefore applied only to instances of grammaticalization.Footnote 9 Thus, Brugman (Reference Brugman1984, cited in Heine Reference Heine1997: 9) notes that over ‘is an instance of polysemy even if it has prepositional, adverbial, and derivational uses and hence is associated with different morphosyntactic categories’. In this view, heterosemy is considered a special type of polysemy.
A partly similar case to Lichtenberk (Reference Lichtenberk1991) is Enfield (Reference Enfield, Ameka, Dench and Evans2006). Unlike Magnusson & Persson (Reference Magnusson and Persson1986) and Persson (Reference Persson, Hyldgaard-Jensen and Zettersten1988), Enfield (Reference Enfield, Ameka, Dench and Evans2006: 297–8, 310) uses ‘heterosemy’ for conversion, if a word is both polysemous and a member of two word classes. Enfield (Reference Enfield, Ameka, Dench and Evans2006: 297 et passim) distinguishes three types of heterosemy:
(i) open-class heterosemy, in which central conversion belongs,
(ii) closed-class heterosemy (e.g. to as the infinitive mark and as a preposition), and
(iii) cross-class heterosemy (e.g. have as a main verb and as an auxiliary verb).
Leaving aside the potential conceptual ambiguity of the occurrence of ‘meanings or functions’ as in Lichtenberk's (Reference Lichtenberk1991: 476) quotation above, and the uneven use of the term ‘heterosemy’, this definition in principle covers central cases of conversion and avoids some of the contradictions raised by the use of the concept ‘homonymy’ listed at the end of section 2.1.1. Despite the disparate use of the term, it offers advantages for the higher descriptive accuracy and for the avoidance of the said contradictions. It also allows higher internal homogeneity within homonymy by excluding conversion-related pairs and the additional possibility of occurrence of homonymy across word classes.
Whereas heterosemy lays emphasis on semantic contrast, homomorphy lays emphasis on formal identity. ‘Homomorphy’ was specifically proposed for conversion-related pairs, or cases where a relation by conversion may be posited and, unlike zero-derivation (as used in Sanders Reference Sanders, Hammond and Noonan1988), it is specific to the relationship, i.e. it is not used for both the process and the relationship. Originally coined in this sense by Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985: 70–1), homomorphy is presented alongside other equivalence relations between words, namely homonymy and synonymy. Of these, homonyms and their subtypes, homographs and homophones, are defined as morphologically unrelated words that have the same phonological or orthographic shape. For words with the same morphological form (i.e. not just formally identical) and different syntactic function, Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985: 70) propose the term ‘homomorph’. Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985: 71) thus account for instances of conversion, i.e. where words share the same stem morpheme and are related through a process of word-formation in a semantically systematic way. Remarkably, the term is ambiguous in several respects (cf. Valera Hernández Reference Valera Hernández1996): e.g. regarding its use with reference to grammatical or lexical words (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 70–1), and regarding other processes where words share the same stem or morphological form and belong to different syntactic categories, including cases where grammatical words are formed from apparently open word classes (i.e. potentially involving grammaticalization, as in down Adv and down Prep). In the latter respect, Quirk et al.'s (Reference Quirk, Greenbaum, Leech and Svartvik1985) use of ‘homomorphy’ and Lichtenberk's (Reference Lichtenberk1991) use of ‘heterosemy’ coincide, despite the substantial differences between conversion and grammaticalization, both from an output-oriented and from a process-oriented point of view.
2.2 Relationships according to semantic contrast: polysemy, synsemy, hyponymy
The relationship between conversion-related pairs has also been described with a focus on semantic contrast. The most frequent interpretation under this approach is in terms of polysemy (cf. e.g. Miller Reference Miller, Halle, Bresnan and Miller1978: 102–5 on a rule for sense extension from a core sense as a semantic link between verbs and nouns, cited in Martsa Reference Martsa2013: 208; cf. also Lehrer Reference Lehrer1990: 208–9, cited in Martsa Reference Martsa2013: 207–10; Tătaru Reference Tătaru2002, cited in Frăţilă Reference Frăţilă2011: 56). Under this approach it is assumed that conversion is a non-standard, non-prototypical type of polysemy (Martsa Reference Martsa2013: 209–10; cf. also Martsa Reference Martsa, Gottlieb, Mogensen and Zettersten2002, cited in Martsa Reference Martsa2013: 202), ‘intercategorial polysemy’ (cf. Zawada Reference Zawada2006: 247, cited in Martsa Reference Martsa2013: 210), or ‘functional polysemy’ (Tătaru Reference Tătaru2002), applicable to formally identical, semantically and derivationally related, and syntactically different word pairs.
Polysemy is typically described as the paradigmatic lexical relation between different, related senses (for some authors, meanings, e.g. Lipka Reference Lipka1990: 75, 136; Heine Reference Heine1997: 8; Saeed Reference Saeed1997: 64; Stockwell & Minkova Reference Stockwell and Minkova2001: 147–8) of one lexeme, lexical entry or word (Ullmann Reference Ullmann1957: 117; Reference Ullmann1962: 159; Leech Reference Leech1974: 228–30; Lyons Reference Lyons1977, vol. II: 550 et passim; Reference Lyons1981: 146; Cruse Reference Cruse1986: 80; Magnusson & Persson Reference Magnusson and Persson1986: 27–34; Persson Reference Persson1990: 141–3; van Sterkenburg Reference Sterkenburg1996: 409; Plag et al. Reference Plag, Braun, Lappe and Schramm2007: 163; Goddard Reference Goddard2011: 23; Rainer Reference Rainer, Lieber and Štekauer2014: 338 et passim).
Unlike homonymy, polysemy attests a semantic connection between senses of the general meaning of one lexeme. For this reason, it has sometimes been described under the condition that the senses be ‘causally related’ (Persson Reference Persson, Hyldgaard-Jensen and Zettersten1988: 270; Reference Persson1990: 143–6, 152–3, 160). The connection is often one of a literal versus a figurative or extended sense, and this is one of the respects where polysemy does not cover all the properties of the relationship between conversion-related pairs: while, unlike homonymy, an interpretation in terms of polysemy accounts for the formal identity, for the morphological relation resulting from the use of one and the same morpheme, and for the common etymological origin, it does not account for the semantic contrast, and this in two senses:
(i) the semantic contrast in polysemous pairs is usually understood as a contrast between literal and figurative senses, and the contrast in categorial meaning is not typically accounted for in most cases of polysemy or in regular polysemy, and
(ii) the semantic contrast between conversion-related pairs is usually understood as a contrast between categorial meanings,Footnote 10 even if figures of speech like metonymy or metaphor may be involved in conversion too.
Lexical extension mechanisms, mainly metaphor and metonymy, have been proposed to account for or to occur in conversion (Paul Reference Paul1982: 305; Crocco-Galeas Reference Crocco-Galeas, Dosuna and Pensado1990: 27–8; Twardzisz Reference Twardzisz1997: 35 et passim; Kövecses & Radden Reference Kövecses and Radden1998: 54–5; Dirven Reference Dirven1999: 277–9; Radden & Kövecses Reference Radden and Zoltán1999: 37; Schönefeld Reference Schönefeld, Bauer and Valera2005: 149–50; Martsa Reference Martsa2013: 202, 205, 211; cf. also Manova Reference Manova2011: 62–4 and Bauer Reference Bauer2018, the latter especially on noun/verb conversion). Under this view, semantic change and word-class change are taken as the result of figurative extension in conversion.
Metaphor and metonymy are extremely productive semantic processes in the lexicon (Lipka Reference Lipka1990: 163–5), and major processes involving figurative extension of meaning and word-class change. These processes are sometimes explained as ‘dual categorization’ (Lipka Reference Lipka, Hüllen and Schulze1988, cited as ‘Lipka 1988b’ in Lipka Reference Lipka1990: 124), as ‘secondary’ or ‘multiple categorization’ (Lipka Reference Lipka1990: 186, for metaphor), or as ‘semantic’ or ‘inner derivation’ (Lipka Reference Lipka1986: 134), on the grounds that metaphor and metonymy occur within one and the same lexeme. Of these, metaphorical extension of meaning creates new relations (Ullmann Reference Ullmann1962: 212–18), e.g. to fish for information/for a rich man (Dirven Reference Dirven1999: 281), and is typically described as:
a linguistic instrument whereby a linguistic form designating an element from one conceptual sphere is used (figuratively) to talk about an element from another conceptual sphere. (Lichtenberk Reference Lichtenberk1991: 478, emphasis as in the original)
By contrast, metonymy is simply described as ‘a word used for another, with which there is some special association, as in space or time’ (Stockwell & Minkova Reference Stockwell and Minkova2001: 149) or as belonging:
to the wider set of strategies of finding a point in the common reference space between a speaker and a listener that can serve as a bridge or link to the intended referent. (Langacker Reference Langacker1991: 170 et passim, cited in Dirven Reference Dirven1999: 275)
This is often within the same word class (cf. e.g. Ullmann Reference Ullmann1962: 218–20). Dirven (Reference Dirven1999: 277–9) cites a special type of metonymy, ‘event-schema metonymy’ for cases of conversion, on the argument that a salient feature or participant becomes the main designation for the event itself, as in e.g. The waitress cleaned the table. Dirven (Reference Dirven1999: 277 et passim) also distinguishes mainly between the action schema e.g. to fish, the location or motion schema e.g. to bottle, and the essive schema e.g. to author. In line with Dirven, Colman & Anderson (Reference Colman and Anderson2004: 547) argue that lexicalized metonymies are conversions, and that they do not result in polysemy.
Conversion is therefore related via metaphor and metonymy to polysemy, which is often claimed to be the result of metaphorical extension (e.g. in Ullmann Reference Ullmann1962: 185; Lyons Reference Lyons1977, vol. II: 567; Tournier Reference Tournier2007: 49; Saeed Reference Saeed1997: 307). Specifically, it has been argued that:
[t]he polysemous nature of the semantic link between conversion pairs … is based on the assumption that the senses of converted words are largely predictable through metonymic and/or metaphoric mappings from the relevant senses of parent words, or, perhaps more frequently, from the language users’ encyclopaedic knowledge encapsulated in the lexical meanings of parent words. (Martsa Reference Martsa2013: 224)
Thus, even though the view of polysemy as the lexical relation between conversion-related pairs may not be frequent in the literature, it lies behind the descriptions that argue that figurative or lexical extension of meaning may give rise to conversion.
The description of conversion-related pairs as cases of polysemy has been explicitly excluded on the grounds of the existence of a word-class contrast (cf. e.g. Béjoint Reference Béjoint2010: 299; Manova Reference Manova2011: 62–4, also against metaphor in conversion; cf. also Goddard Reference Goddard2011: 25), and on the grounds of the existence of separate words instead of senses of one word (cf. Magnusson & Persson Reference Magnusson and Persson1986: 45–6 and Persson Reference Persson, Hyldgaard-Jensen and Zettersten1988: 277–8). In general, descriptions of conversion-related pairs as polysemy are fewer than as homonymy, and usually use some specification of the term: besides the abovementioned ‘intercategorial polysemy’ and ‘functional polysemy’, the literature also features reference to ‘closed referential polysemy’ in Deane (Reference Deane1988: 349) and in Dirven (Reference Dirven1999, cited in Rainer Reference Rainer, Lieber and Štekauer2014: 339), or to ‘regular polysemy’ in Koskela & Murphy (Reference Koskela, Lynne Murphy and Brown2006: 743). Another example is Nida's (Reference Nida1975: 97–9) description of homophones that are semantically related, versus homophones that are not, and where the former are ‘a single morpheme’ and the latter are not.Footnote 11
A second relationship can be listed in this section: synsemy. Various interpretations are available in the literature. This is partly because Magnusson & Persson (Reference Magnusson and Persson1986: 47–51) and Persson (Reference Persson, Hyldgaard-Jensen and Zettersten1988: 277–8) use ‘synsemy’ for cases where there is causal relation, co-sense, same category and phonological identity (e.g. transitive and intransitive stand V), and which are graphically represented and described as a ‘“[v]ertical” sense relation between phonologically identical superordinate and subcategorial hyponyms’ (our emphasis). In a later publication, Persson (Reference Persson1990: 156, 159) defined synsemy as ‘the sense relation that obtains between a parent lexeme and a phonologically identical subcategorial hyponym’, exemplified again with transitive and intransitive stand V, but where it seems clear that synsemy does not apply across word classes (despite the use of the term ‘zero-derivation’ in this regard: ‘[t]he most typical cases occur among verbs with intransitive and zero-derived transitive versions’, Persson Reference Persson1990: 159; cf. also Magnusson & Persson Reference Magnusson and Persson1986: 50–1).Footnote 12
This interpretation of ‘synsemy’ differs from Magnusson & Persson's (Reference Magnusson and Persson1986, cited in Steinvall Reference Steinvall2002: 146, emphasis as in the original), where Steinvall notes:
there is a qualitative distinction between homophonous transcategorial derivatives and polysemous items belonging to the same category. The transcategorial relation (i.e. involving shift of word class) is called synsemy.
In a review of Magnusson & Persson (Reference Magnusson and Persson1986), Kreidler (Reference Kreidler, Cannon, Barton and Kreidler1988: 75, our emphasis) argues that the term ‘synsemy’ is used for:
the vertical sense relation between superordinate and hyponym of the same category which are phonologically identical, as the transitive causative verb stand is a hyponym of the intransitive verb stand.
Kreidler explains that Magnusson & Persson (Reference Magnusson and Persson1986) use ‘categorial hyponymy’ for conversion instead of ‘synsemy’. Specifically, Persson (Reference Persson, Hyldgaard-Jensen and Zettersten1988: 273–4; Reference Persson1990: 157–8) and Magnusson & Persson (Reference Magnusson and Persson1986: 8–10, 35–8, 40–1, 44) refer to it as ‘categorial hyponymy’ of a morphologically unmarked transcategorial type, i.e. by contrast with morphologically marked cases (discover V vs discovery N or discoverer N) and with subcategorial cases (e.g. as transitive stand V vs intransitive stand V). The term ‘categorial hyponymy’ is used analogously to ‘lexical hyponymy’, e.g. flower–rose–tulip, where rose and tulip would be hyponyms of the superordinate flower but co-hyponyms to each other. For Magnusson & Persson, the difference between categorial hyponymy and lexical hyponymy lies in the fact that categorial hyponymy is concerned with derivation across word classes and also across subclasses, e.g. transitive and intransitive verbs, gradable and non-gradable adjectives, etc. An example of categorial hyponymy is wide Adj/width N/widen V. However, it is noted that in categorial hyponymy:
there is sometimes a problem in determining which lexeme is the superordinate and in what order that various hyponyms are derived from the superordinate and/or from one another. (Kreidler Reference Kreidler, Cannon, Barton and Kreidler1988: 75; cf. also Magnusson & Persson Reference Magnusson and Persson1986: 38–40)
This issue about directionality is similar to what often happens in conversion. From the examples cited above, it can be seen that not only zero-derived terms are considered, but word-class-changing derivation in general falls under the term ‘categorial hyponymy’.
The following section reviews the major cases of conversion and the adequacy of the relationships reviewed above for each of the cases.
3 Conversions and relationships
This section reviews the adequacy of the above relationships with respect to specific types of conversion. Like the preceding sections, it is based on the literature. For a comprehensive summary and, in view of the dissimilar profile that conversion-related pairs may display, this section relies on Bauer et al.'s (Reference Bauer, Lieber and Plag2013: 546–62) list of cases that can be or have been described in terms of conversion, namely the central type described in Bauer et al. (Reference Bauer, Lieber and Plag2013: 545–9; see section 3.1), which also covers compounds and phrases in this article, ‘formations related to prepositions’, ‘minor phonological modification’, ‘participles’, ‘type coercion’, ‘adverb formation’ and ‘compounds and phrases’. All these are briefly reviewed in this subsection for convenience, except participles (discussed in section 3.2), except the type on adverb formation, for whose description this article follows Bauer et al. (Reference Bauer, Lieber and Plag2013: 559–60) and refers to Valera Hernández (Reference Valera Hernández1996), and except the type of compounds and phrases, which falls within the type described in section 3.1 for the reasons argued at the beginning of that section.
As the identification of central conversion with respect to less central cases is not the objective of this article, the types described by the above source as syntactic and therefore not central cases of conversion are not considered below. This exclusion refers to instances of mentioned words (i.e. ‘mention versus use’, e.g. but me no buts), the so-called ‘adjective to noun cases’ (Bauer et al. Reference Bauer, Lieber and Plag2013: 549–51), and instances of type coercion.
Of these, the former diverge from the latter two in several respects, like the type of bases and the productivity, and in that it is not as common as the other two cases or as central conversion. Adjective-to-noun cases cover three types of structures which are described as syntactic developments rather than as conversion:
(i) the types the rich with plural reference and the type the new with singular reference,
(ii) the type an intellectual, and
(iii) the type news, burnables.
The latter is described as less clearly syntactic. As far as this article is concerned, and as a number of structures may hide under this formal profile, only when the latter type becomes lexicalized as a word with different lexical content (e.g. as in hopeful Adj ‘full of hope’ vs hopeful N ‘candidate’) is it considered to qualify as central conversion. In that case, it can be argued that it shares the essential properties of formal identity, semantic contrast and grammatical contrast and can be subsumed under the description of central conversion in section 3.1 as far as the relationship is concerned. Finally, and although type coercion has been marginally associated with conversion in the literature (e.g. in Magnusson & Persson Reference Magnusson and Persson1986: 6–12, 47–9, 54–9), it is usually acknowledged that conversion occurs between primary (not secondary) word-class categories.
Otherwise, it is unclear whether less central cases of conversion can be described in the same terms as central conversion not only as regards the identity of the process, but, as pointed out by an anonymous reviewer, also as regards the relation between base and derivative.
3.1 Central conversion
Central conversion refers to examples like leaflet N vs leaflet V, dirty Adj vs dirty V and spy N vs spy V, which have been identified with ‘a definable type of homophony between homonymy and polysemy’ (Bauer et al. Reference Bauer, Lieber and Plag2013: 546–7). This section also covers conversion of compounds and phrases, insofar as they:
fairly straightforwardly meet the criteria for conversion set out in Section 25.2: the … forms are homophonous, they clearly belong to distinct word classes, they are semantically related, and their semantic relationships are similar to those found between bases and overtly affixed forms derived from them. (Bauer et al. Reference Bauer, Lieber and Plag2013: 561)
The intermediacy of this type of homophony between homonymy and polysemy may find justification in the interpretation of these relationships as ends of a gradient instead of as an opposition (cf., among others, Lipka Reference Lipka1986: 138; Reference Lipka1990: 138–9, 186; Robins Reference Robins and Burchfield1987: 66; Murphy Reference Murphy2010: 91–3; or Rainer Reference Rainer, Lieber and Štekauer2014: 339). The relationship that holds in conversion-related pairs would display an intermediate degree of formal/semantic unity that does not fit either end, i.e. it does not fit standard homonymy or standard polysemy. It does not fit standard homonymy, because there is semantic relatedness and because there is a common source. It does not fit standard polysemy, because the semantic contrast is between two words instead of between two senses of one word, and because it is not always in terms of literal versus figurative meaning, but in terms of contrast in categorial meaning, regardless of whether figurative meaning is involved or not. Finally, it does not fit the two of them, homonymy and polysemy, because word-class contrast is a necessary condition in conversion, but is not a condition either in standard homonymy or in standard polysemy. Still, what conversion-related pairs show, and what these authors prove in terms of semantic properties (dependency, scope) and of productivity, is that the relationship between the members of these pairs is derivational, that it is across word classes and entails grammatical contrast, and that it is between at least two words. The point here is that this derivational link, word-class contrast and the existence of two words are not in the definition of homonymy, of polysemy, or in any of the intermediate degrees that may be defined between the two.
In terms of the cases of central conversion, pairs that are perceived in present-day English as conversion, even if their formal identity results from diachronic leveling of inflections, are in a better position to qualify as homonymy, at least as far as their identity is, as in standard homonymy, accidental. Nonetheless, neither these nor present-day formations for which leveling cannot be attested display the semantic or the historical unrelatedness of homonymy, or the semantic contrast of polysemy.
The terms of this review may also apply to less central types or cases which are described in Bauer et al. (Reference Bauer, Lieber and Plag2013: 552–5) as formations related to prepositions, and as formations with minor phonological modification, e.g. with devoicing or stress shift. Extension of these arguments to prepositions is in line with other descriptions where open and closed word classes are treated alike (e.g. Persson Reference Persson, Hyldgaard-Jensen and Zettersten1988, cited in Lichtenberk Reference Lichtenberk1991: 476 with respect to ‘particles’). Although some of these cases have been considered cases of polysemy, the literature also includes texts where the analysis in terms of polysemy alongside similar word-class transfer between nouns and verbs has also been discarded on the grounds presented above in this section for the rejection of polysemy in conversion (e.g. Goddard Reference Goddard2011: 25 with respect to examples like behind).
3.2 Participles
The description of participles as instances of conversion refers to present and past participle forms. The relationship between verbs and other categories under one form in both cases has been described as categorial hyponymy alongside standard conversion, e.g. in Magnusson & Persson (Reference Magnusson and Persson1986: 10–12).
The present participle ending -ing qualifies in part as a case of homonymy and as a case of polysemy. The former functions as the contemporary counterpart to the Old English participial suffix in verbs and the Old English derivational suffix for the formation of deverbal nouns -ung (cf., among others, titles in historical linguistic work such as Quirk & Wrenn Reference Quirk and Wrenn1957: 112; Wik Reference Wik1973 and titles in morphological work such as Marchand Reference Marchand1969: 302–5; Dalton-Puffer Reference Dalton-Puffer1996: 37–9 or Tabor & Traugott Reference Tabor, Traugott, Ramat and Hopper1998: 240–4). The profile of the original inflectional -ing form and the derivational form -ing has been interpreted in terms of homonymy but also in terms of polysemy, such that contemporary English -ing is polysemous in that it may form deverbal adjectives or deverbal nouns (cf. O'Grady & De Guzman Reference O'Grady, de Guzman, O'Grady, Dobrovolsky and Katamba1996: 171). This type of participle thus appears to illustrate, according to several authors, the ends of the gradient between homonymy and polysemy often referred to in the literature.
A similar point can be raised with reference to the past participle, regular and irregular. It can be interpreted as involving both an inflectional suffix and as a derivational suffix for the formation of deverbal adjectives and, therefore, as a case of two separate coexisting affixes (thus, homonymy, e.g. in Adams Reference Adams1973: 22–3; cf., however, the description in Adams Reference Adams2001: 6–7 without explicit reference to homonymy or to a different relation). A gradient from pure inflectional to derivational meaning can, however, be traced back from the scale between purely participial instances and purely adjectival instances (cf. Granger Reference Granger1983), the latter supplying an analogical basis for derivational instances without attested verbs (e.g. unknown).
4 Conversion outside paradigmatic relationships
Section 2 reviewed a number of relationships described in the literature for conversion-related pairs. This section reconsiders the main arguments why the relationships may be adequate or not, in addition to the points raised in the above review.
For reasons of space, this section focuses on the relationships most often cited in the literature, specifically homonymy and polysemy. In terms of the relationships that deal with identity (section 2.1), heterosemy and homomorphy are not discussed beyond what was presented in section 2. Both are comparatively less frequently accepted, are relatively ambiguous according to how they are applied, and also involve cases that are not conversion, thereby giving rise to conceptual heterogeneity too. In particular, homomorphy is also listed as an equivalence relation between words and, whereas it may be in wide use in certain forums, it does not feature in the research literature.Footnote 13 In terms of the relationships that deal with semantic contrast (section 2.2), no further arguments are given below against the adequacy of synsemy and categorial hyponymy, largely for similar reasons: the two concepts have been interpreted in rather different ways in the literature (in the case of synsemy with substantial conceptual conflict), and both are assumed to involve cases within the same word class, i.e. they are assumed to apply in cases that are described as secondary word-class conversion or type coercion and which are not frequently considered to be cases of conversion.
This section also shows that the use of the concepts of homonymy and polysemy with respect to conversion-related pairs often occurs in a vague sense (presumably intended as mere formal identity). Otherwise, the use of these terms needs some qualification to set them apart from the standard cases that do not involve conversion, or as interpretations of these relationships that encompass a range of cases, one of which (precisely, conversion) is less prototypical in that it involves a word-class contrast or grammatical contrast associated with various word classes. Still (as shown in previous sections), both homonymy and polysemy have been disqualified for the relation between conversion-related pairs.
This section proposes a view that lays emphasis on the separation between a process and a relationship (unlike, e.g. in Sanders Reference Sanders, Hammond and Noonan1988, where zero-derivation is seen as both a relationship and a process, or in Bauer Reference Bauer2018, where metonymy is seen as both a relationship and a process). For higher descriptive accuracy in terms of the relationship between conversion-related pairs and in terms of the type of relationship used for the description, this view is also intended to avoid the heterogeneity that the inclusion of conversion adds to the relationships discussed in section 2, e.g. the heterogeneity that conversion as homonymy (or within polysemy) entails with respect to canonical cases of homonymy (or of polysemy).
This view refers specifically to Cruse's separation of semantic relations of a paradigmatic type from paronymic relations. Paronymic relations are described in Cruse (Reference Cruse1986: 55) as ‘involving identity of root, but difference of syntactic category’ between words, e.g. act > actor. The term ‘paronym’, reportedly used as early as in Aristotle's Categories, is rare in the discussion of conversion. Aristotle apparently used it for substantives named for their qualities, e.g. grammarians get their name from grammar (cf. Jones Reference Jones1972: 117–18; Malcolm Reference Malcolm1981: 291–2), and the brave get theirs from bravery (Jones Reference Jones1972: 118; Reference Jones1975: 148). The term also appears, e.g., in Hintikka (Reference Hintikka1959: 141, cited in Jones Reference Jones1975: 147): ‘two things are paronyms when they are called by different “names” (terms) of which one is nevertheless derived (grammatically) from the other’. Other authors translate or describe Aristotle's notion of paronymy with a focus on the change of word class by modification of the word ending (e.g. Owen Reference Owen, Düring and Owen1960: 175, cited in Jones Reference Jones1975: 147; Ackrill Reference Ackrill1963, cited in Jones Reference Jones1972: 118). The same type of relationship seems to have been given a different name by other authors: Malcolm (Reference Malcolm1981: 286–7) describes the term ‘eponymy’ as Plato's equivalent to Aristotle's ‘paronymy’. Still, Malcolm (Reference Malcolm1981: 291–2) notes that Allen (Reference Allen1970: 126) also treats cases such as the just and just, in which there is no formal difference between the terms, as deviant cases of eponymy or paronymy. Ross (Reference Ross1981: 137) also uses the term in a broader sense, such that ‘[p]aronyms are morphologically variant (and, for the most part meaning related, but not univocal) n-tuples derived (synchronically) from a common root’, with ‘morphologically variant’ referring to both inflection and derivation. Overall, Cruse (Reference Cruse1986) uses the term ‘paronymy’ not only for the relation in cases involving class-changing derivation where the word ending indicates the change of word class, but also for zero-derived paronyms, i.e. where there is no overt sign of category change (Cruse Reference Cruse1986: 132), e.g. hammer N vs hammer V. Under this view, then, paronymy is the relation between any two derivatives that can be identified as members of different word classes:
The relationship between one word and another belonging to a different syntactic category and produced from the first by some process of derivation will be called paronymy; the derivationally primitive item will be called the base, and the derived form the paronym. (Cruse Reference Cruse1986: 130, emphasis as in the original)
Paronymic relations involve another relationship, endonymy, defined as a lexical relation ‘based on the notion of semantic encapsulation, and [that] involves the incorporation of the meaning of one lexical item in the meaning of another’ (Cruse Reference Cruse1986: 123). Although endonymy is initially used by Cruse for ‘proportional series’ of related words (of different origins), and of the same word class, e.g. horse vs stable or cow vs byre, it is also applied to series where the contrast involves lexical items of different syntactic categories, e.g. the relation between dig and spade. The relevance of this contrast may appear to be far-fetched in that these examples do not involve conversion, but it is not if we consider that the semantic relation between spade N and dig V is claimed to be virtually identical to that between comb N and comb V (Cruse Reference Cruse1986: 133). The following parallelism can therefore be inferred in these pairs:
[Exonym: dig V (endonym: spade N)]
[Exonym: comb V (endonym: comb N)]
The two relationships referred to above as ‘paronymic relations’ are connected to each other: ‘in a normal case of base and paronym, the base is an endonym of the paronym’ (Cruse Reference Cruse1986: 130).
The profile of conversion-related words is different from that of words which, as in paradigmatic relations, display formal identity or contrast, alongside semantic contrast or identity, and vice versa, without additional grammatical considerations. In this sense, it is not just a lexical relationship, because it involves grammatical contrast. In turn, the profile of conversion-related words is different from that of words which are morphologically related by any other class-changing word-formation processes, because formal identity in conversion is a necessary condition. This brings the question of the form to the fore, so much so that it may have prompted their classification as paradigmatic relations at the cost of the grammatical contrast.
If all the facets of the profile of conversion-related words are to be covered by one relationship, then it must be one that is not only a lexical relationship. Like conversion itself, it must comprehend properties of morphology, syntax (or, in broad terms, grammar) and lexical semantics (as regards both contrast in categorial meaning and lexical formation) because, as in the opening lines of the article, conversion displays properties of morphology, syntax, semantics and the lexicon. This condition seems to be satisfied not by standard paradigmatic relations, but by paronymic relations of the type described by Cruse (Reference Cruse1986). Even so, and considering the range of cases that the term ‘conversion’ covers in the literature, not all instances of so-called conversion may be adequately described as conversion. Aside from the issue of the identification of conversion, this prompts the question whether different types of conversion may involve different types of relations between their bases and their derivatives.
The type discussed as central in section 3.1 qualifies as a paronymic relation in Cruse's (Reference Cruse1986) sense: paronymy describes the relation between base and derivative in central conversion more accurately than any of the relations discussed in section 2, because it responds to conversion's profile of formal identity, grammatical contrast and semantic relatedness. Very relevantly for this article (as pointed out in section 1, to cite an anonymous reviewer on the relevance of the description of the relationship between conversion-related pairs), paronymy occurs between derivatives regardless of whether there is formal identity, as in conversion, or not, as in other word-formation processes.
Also, the relation between base and derivative in these examples is not adequately described by any of the relations reviewed in section 2, among others, because the standard definitions of those relations do not cover grammatical (word-class) contrast. The qualifications of those relations that incorporate the grammatical component are, as far as the cases under consideration here are concerned, an unnecessary addition of peripheral cases that is redundant with other relations (in this case, paronymy), and this introduces unnecessary heterogeneity.
This applies to central conversion regardless of the specific semantic pattern involved, e.g. noun-to-verb conversion; i.e. it holds equally well for the converted verb to widow when it expresses a resultative meaning (‘make into [noun]’) as when it expresses an inchoative meaning (‘become [noun]’), because the relation between the base noun widow and the derived verb to widow does not change according to whether the semantic pattern is one or the other.
In principle, the same applies regardless of other variables, like the word classes involved, the directionality, or whether literal or figurative meaning is involved, because the above examples with widow are described in terms of paronymy as accurately as when other word classes are involved (e.g. adjective-to-verb conversion to clean ‘make [adjective]’), as when a different direction is involved (i.e. verb-to-noun conversion say ‘an instance of [verb]’ compared with noun-to-verb conversion in widow), or when literal or figurative meaning is involved (e.g. to watermark ‘provide with [noun]’, whether it is a physical or figurative hallmark). In all these cases, the relation between base and derivative involves ‘identity of root, but difference of syntactic category’, as in Cruse's (Reference Cruse1986: 55) quotation above in this section: the identity of root entails semantic relatedness (whether literal or figurative), and the difference of syntactic category responds to the word-class contrast inherent in conversion. As pointed out by two anonymous reviewers, opposite directions and the occurrence of figures of speech respectively may lead to different types of paronymic relations, because e.g. noun-to-verb conversion has rather different semantic properties and uses different semantic patterns than verb-to-noun conversion, and conversion involving metaphor and metonymy involves different semantic properties than conversion involving literal meaning. Even so, all these cases are expected to fall within Cruse's description above and, ultimately, paronymy qualifies for this type of conversion better than the other relations reviewed here. Also, Cruse's (Reference Cruse1986: 132) qualification ‘zero-derived paronyms’ does not introduce heterogeneity to the concept of paronymy, it only specifies the type of units in which paronymy may be found.
By contrast, the type involving participles discussed in section 3.2 lends itself to description as homonymy or polysemy better than as paronymy. Unlike the central type, this type of conversion results from the polyfunctionality and/or polysemanticity of affixes, some of which are known to produce word-class contrast partly as a result of homonymy (cf. the case of -ing in section 3.2), some others possibly as a result of polysemy (cf. both the case of -ing and of -ed in section 3.2). Which of the two possibilities (homonymy or polysemy) fits each best is a more complex question than this article can solve, and one that probably needs different answers for different words, even if broad patterns can be identified. At any rate, what matters here is that in participial conversion homonymy and polysemy fit better than any of the other relations reviewed in this article, including paronymy, because the word-class contrast results from relations of identity: the leveling of several forms as one (homonymy) or the polysemanticity of participial units in -ed as states (polysemy).
At a more general level, the point of this article is that two types of conversion, central and participial, involve two types of relations between their respective bases and derivatives, that only one relation cannot adequately describe the relation between several types of conversion-related words, and that relations between conversion-related pairs do not necessarily have to be in terms of paradigmatic lexical relations: they must also be in terms of paronymic relations, of which even several subtypes may be found according to the participation of figures of speech or according to directionality in conversion.
5 Conclusion
The relationship between conversion-related words has been considered and reconsidered for decades, probably for the reasons listed in the first lines of the article: conversion lies at the crossroads of morphology, syntax and lexical semantics, and each field views the relation differently, from something that is unimportant (and sometimes not even mentioned) to one among various types of relations. Conversion-related pairs have been described in terms of a process (zero-derivation, metonymy), in terms of an ad hoc relationship (homomorphy) or, more often, without special consideration, i.e. very much like bases and their derivatives formed by affixation or compounds. When they have been described in terms of a paradigmatic relation, this is usually homonymy or polysemy, but even in those cases they have been presented as special or non-prototypical cases of each relationship.
There may not be a need for a relation to describe these words, as in affixation, but if one is considered necessary, then it seems that the usual description in terms of paradigmatic relations like homonymy or polysemy is not entirely accurate, at least for all that is described as conversion, or it may be more accurate for some cases (participial forms) than for others (central cases of conversion). This is because the best framework for the description of the relation between conversion-related pairs should be able to respond to Persson's (Reference Persson1990: 161, our emphasis) claim that:
phonological identity may disguise a number of different underlying syntactico-semantic relations and that the interplay between meaning and form may be much more than the traditional division into homonymy and polysemy seems to imply.
The literature on derivational morphology seldom brings lexical relations in line with morphological properties, and limits itself to the occurrence of formal identity or not, and of semantic relatedness or not. The literature on lexical semantics seldom relates the description of lexical relations based on semantic structures with the morphological connection that may exist between semantically related terms. This approach, which is made explicit e.g. in Magnusson & Persson (Reference Magnusson and Persson1986: 2–4), allows us to describe the relation between conversion-related pairs as lexical relations with a grammatical qualification, as in categorial hyponymy (cf. Magnusson & Persson Reference Magnusson and Persson1986; Persson Reference Persson, Hyldgaard-Jensen and Zettersten1988), or as a paronymic rather than paradigmatic relation (cf. Cruse Reference Cruse1986). This approach to the relationship between conversion-related terms has received comparatively less attention than any other reviewed above. This is paradoxical considering that the words under discussion display a unique profile that brings together formal, semantic and grammatical conditions, as in the quotations from Dokulil (Reference Dokulil1968) and Kastovsky (Reference Kastovsky1977) that open the article, and that these three are better covered in the definition of paronymic relations than in terms of the paradigmatic relations that are usually cited and are recalled in the opening paragraph of this section.
Not all the pairs that meet the profile of conversion respond equally well to the conditions or the definitions of each of the relationships discussed here: paronymic relations respond well to central cases of conversion, but paronymic relations do not cover all the possible cases that meet the profile of conversion. Homonymy and polysemy may respond well to the type involving participial forms.
Use of a relationship-oriented rather than an output-oriented approach to conversion, where processes and relationships are clearly separated, may contribute to the identification and separation of the range of formations that meet the profile of the definition of conversion in English.
