Denominal -ed Adjectives and Their Adjectival Status in English Morphology

Abstract

In the study of English denominal adjectives, scholarly attention has predominantly centred on those with Latinate suffixes (e.g., -al, -ary, and -ic/-ical), which are well-known as relational adjectives (RAdjs) and are extensively scrutinised in the existing literature. Conversely, those with English native suffixes (e.g., -en, -ern, -y, and -ed) have not undergone thorough examination to date. In the present study, I delve specifically into denominal adjectives with the suffix -ed (-ed Adjs), such as bearded, long-tailed, and shirt-sleeved. I present a novel basic picture of these adjectives, setting forth the following two central propositions: (i) -ed Adjs are a type of RAdj and (ii) undergo conversion to qualitative adjectives (QAdjs) (e.g., bearded man vs. bearded rock) akin to the better-known Latinate RAdjs (e.g., grammatical error vs. grammatical sentence). The analysis is conducted by examining suffixal etymology (i.e., Latinate or Germanic), suffixal properties (i.e., all-purpose or dedicated), and the driving factor for QAdj-forming conversion (i.e., the modal attribute true). These propositions and analyses collectively enrich our comprehensive understanding of the semantic and morphosyntactic properties of -ed Adjs within the realm of English morphology.

1. Introduction

In English derivational morphology, as clearly articulated by Bauer et al. (2013, at the very beginning of Ch. 14), Fábregas and Marín (2017), and Sleeman (2021, p. 551), studies on adjectivalisation remain comparatively scarce compared to those on verbalisation (e.g., Clark and Clark 1979; Hale and Keyser 1993; Levin and Rappaport Hovav 1994; Goldberg 1995; Plag 1999; Borer 2005) and nominalisation (e.g., Chomsky 1970; Grimshaw 1990; Alexiadou 1998; Harley 2009; Lieber 2016).

On the other hand, there has been a gradual increase in scholarly efforts focussing on adjectivalisation, particularly within the domain of denominal adjectives, commencing with seminal works by Ljung (1970), Levi (1978), and Beard (1995). Notable contributions have emerged from scholars such as Giegerich (2005, 2009), Fábregas (2007, 2020), Fradin (2008, 2017), Bisetto (2010), Cetnarowska (2013), Rainer (2013), Spencer (2013), Spencer and Nikolaeva (2017), and Nikolaeva and Spencer (2013, 2020), as well as, significantly, a comprehensive series of works by Nagano (2013, 2015, 2016, 2018a, 2018b), Nagano and Shimada (2015), Shimada and Nagano (2018), and Nagano (2023) as the latest contribution. Thanks to these remarkable previous studies, the properties of denominal adjectives have been systematically and deliberately advanced, with cross-linguistic implications.

However, the majority of the existing literature on adjectivalisation, with the exception of Nagano (2023), has generally specialised in denominal adjectives with typical non-native (i.e., Latinate) suffixes such as -al, -ary, or -ic/-ical (e.g., Levi 1978; Kaunisto 2007; Bauer et al. 2013, p. 318; Plag 2018, pp. 95–96). In this paper, I specifically address adjectives suffixed with the native -ed. While the suffix -ed remains considerably productive in English, especially when compared to other native suffixes that are no longer productive, such as -en and -ern (Bauer et al. 2013, pp. 304–6), there are limited systematic studies on this suffix and its derivatives due to the inherently complex and multifunctional nature of -ed as a homophonous affix (e.g., -ed appears in various grammatical and morphological contexts, including passive constructions, verbal past tense, verbal/adjectival participles, and the derivation of adjectives). The present study explores denominal adjectives suffixed with -ed, such as bearded, long-tailed, and shirt-sleeved, hereinafter referred to as ‘-ed Adjs’.

The immediate concerns surrounding -ed Adjs have primarily revolved around their semantico-pragmatic aspects. For example, Hudson (1975) observes that constraints exist concerning the acceptability of phrases containing these adjectives, such as a bearded/*wifed man, an eyed *boy/probe, a verandahed/*doored bungalow, and a *large-gardened/red-roofed house.1 Researchers have rigorously explored these constraints, employing key notions such as inalienable possession (i.e., a close, inherent relationship between two entities where the possessee cannot be conceptually or physically separated from its possessor) and informativity (i.e., the degree to which a given piece of information or linguistic expression provides new, relevant, or useful knowledge to the recipient) (e.g., Hirtle 1969; Hudson 1975; Beard 1976; Ljung 1976; Quirk et al. 1985; Goldberg and Ackerman 2001; Nevins and Myler 2014; Ishida 2023; Morita 2024). While semantico-pragmatic constraints have been discussed from many angles, the morphosyntactic properties and adjectival status of the adjectives in question have not yet undergone adequate investigation and clarification. Against this backdrop, the questions that remain to be solved are threefold:

• What specific properties do -ed Adjs intrinsically possess?
• Into which adjectival categories should -ed Adjs be classified?
• If -ed Adjs are primarily relational adjectives, how can their behaviours as qualitative adjectives be analysed and understood?

The aim of this study is to provide plausible answers to these inquiries and to offer a holistic view of the derived adjectives, exploring and unveiling their adjectival status. Specifically, the contention posits that -ed Adjs can be legitimately categorised as a type of relational adjective. This theory can be reasonably deduced through a discussion of conversion.

The examination unfolds in the following manner. Section 2 first provides an overview of the canonical properties of the following two distinct categories of adjectives: relational adjectives (hereafter RAdjs) and qualitative adjectives (hereafter QAdjs). This enables us to form a general framework of the domain of denominal adjectives and allows us to understand the subject of analysis, i.e., -ed Adjs. Section 3 provides an overview of the basic (albeit irregular and idiosyncratic) properties of -ed Adjs, drawing particular attention to their phonology, semantics, and morphosyntax. Section 4 examines the adjectival status of -ed Adjs, revealing their intrinsic nature as RAdjs, and further discusses why we should assume the adjectives to be a type of RAdj, taking into account the suffix’s etymology and properties (e.g., Rainer 2013). Section 5, predicated on Nagano’s (2018a) conversion analysis, elucidates the enigma and rationale behind the idiosyncratic behaviour exhibited by -ed Adjs from the perspective of proprietive–similative polysemy. Finally, Section 6 concludes the paper with some summarising remarks.

2. Two Types of Denominal Adjectives and Their Canonical Properties

Relational adjectives (RAdjs), also known as associative adjectives, constitute a subset of denominal adjectival formations wherein the role of the adjectivalising morpheme is solely transpositional (e.g., senatorial, polar, algebraic). Morphosyntactically, they function as adjectives by directly modifying nouns and agreeing with the modified noun in languages featuring agreement morphology. Semantically, RAdjs classify the noun that they modify by establishing various semantic relationships between their base nouns and the modified nouns (Giegerich 2005, 2009; see also Shimamura 2014; Nagano 2016). For example, algebraic mind means ‘a mind having to do with algebra, referring to algebra, characterised by algebra’ (Plag 2003, p. 94; Nagano 2016, p. 43). It is also widely recognised that, due to their exclusively transpositional nature, RAdjs exhibit “exactly the same range of denotations” (Nikolaeva and Spencer 2020, p. 92) as their base nouns. Therefore, RAdj-N(oun) and N-N pairs are “fully synonymous and, more interestingly, are apparently used in free variation” (Levi 1978, p. 224), as demonstrated in (1) (examples are adopted from Levi 1978, p. 38; see also Beard 1995, p. 188; Levi 1978, p. 224).2

(1)	a.	atomic bomb	≈	atom bomb
	b.	maternal role	≈	mother role
	c.	industrial output	≈	industry output
	d.	marine life	≈	ocean life
	e.	linguistic skills	≈	language skills
	f.	urban parks	≈	city parks

The relational meaning that combines the base noun of the adjective and the head noun is regarded as extra-linguistic, closely tied to the conceptual knowledge associated with the two nominal concepts (Rainer 2013, p. 15; see also Nikolaeva and Spencer 2020, pp. 91–95).

In contrast, qualitative adjectives (QAdjs), also referred to as ascriptive adjectives, serve to qualify the noun from which they are predicated (e.g., beautiful, picturesque, nervous). To succinctly illustrate these two distinct types of adjectives, compare grammatical in (2), where (2a) is cited from Plag (2003, p. 94, italics mine) while (2b) is from Nagano (2018b, p. 276).

(2)	a.	a grammatical sentence
	b.	a grammatical error

The adjectives in (2) have the same form, yet they exhibit entirely different meanings and functions in modifying each head noun. Grammatical in (2a) is considered a QAdj, as it means ‘conforming to the rules of grammar’ and is predicated on the attribute of sentence (i.e., the sentence is grammatical). In contrast, grammatical in (2b) is an RAdj, as it has a relational sense such as ‘having to do with grammar’ and does not predicate the characteristic of error (i.e., *the error is (very) grammatical) (Nagano 2018b, p. 276) but rather subcategorises the type of error (i.e., an error of grammar). Thus, whereas QAdjs have a qualifying/predicating function, RAdjs have a classifying function.

In addition to their relational semantics and classificatory function above, the other widely assumed prototypical characteristics of RAdjs can be summarised in (3), as detailed by scholars such as Levi (1975, 1978), Beard (1995), Giegerich (2005, 2009), Bisetto (2010), Cetnarowska (2013), Shimamura (2014), and Nagano (2015, 2016, 2018a).

(3)	a.	Predicability:	*this output is industrial, *this decision is senatorial
	b.	Gradability:	*a very industrial output
	c.	Comparability:	*more industrial
	d.	Coordinability:	*the big and wooden table
	e.	Quantifiability:	monochromatic, binational, triconsonantal, multiracial
	f.	Nominalisability:	??presidentialness, ??racialness

Regarding (3a), RAdjs cannot be used in predicative constructions. For (3b) and (3c), RAdjs do not exhibit gradability or comparativeness. Concerning (3d), RAdjs cannot be coordinated with QAdjs. As for (3e), RAdjs can be affixed by numerical prefixes. Lastly, as in (3f), RAdjs do not allow for further nominal affixation. Incidentally, as observed by Farsi (1968), the choice of negative prefix can also differentiate one word from another, such as nongrammatical (RAdj) vs. ungrammatical (QAdj) (see Farsi 1968, p. 53 for other contrastive pairs).

The behaviour of RAdjs in (3) should be derived from their base nouns, as nouns generally have these properties. For example, when the head noun is a deverbal noun, the base noun of an RAdj is interpreted as its internal/external argument with specific prepositions, such as of and by, as shown in (4), cited from Nagano (2015, p. 5).

(4)	a.	Japanese democratisation after World War II
		= the democratisation of Japan after World War II
	b.	the Japanese attack on Pearl Harbour
		= the attack on Pearl Harbour by Japan

Building on this, Nagano (2013, 2015) extends Beard’s (1995) argument and analyses RAdjs as alternative forms of prepositional phrases (hereafter PPs), as indicated in (5) and (6), excerpted from Nagano (2013, p. 123; 2015, pp. 6–7; see also Fradin 2008, §4).

(5)	a.	presidential plane	a′.	plane of the president
	b.	cellular structure	b′.	structure of cells
	c.	dental disease	c′.	disease of teeth
	d.	bearded man3	d′.	man with a beard
	e.	southern exposure	e′.	exposure to the south

(6)	a.	preadverbial expression	a′.	expression in front of an adverb
	b.	postnominal adjective	b′.	adjective after a noun
	c.	suprasegmental phonemes	c′.	phonemes above segments
	d.	sub-Saharan Africa	d′.	Africa below the Sahara
	e.	a trans-global expedition	e′.	expedition across the globe

Interestingly, even if the PPs are complex, such as those in (6), the corresponding RAdjs similarly exhibit the same nominal properties, as aforementioned in (3). As further emphasised by Nagano (2015, p. 7), the observations so far lead us to understand that there is a morphological parallelism between RAdjs and PPs; namely, while nonprefixed RAdjs correspond to simple PPs, as in (5), prefixed RAdjs conform to complex PPs, as in (6).

3. Basic Properties of -ed Adjectives

This section provides an overview of the basic properties of -ed Adjs, focussing on their (i) input structure, (ii) phonology, (iii) semantics, and (iv) morphosyntax. Although this study cannot examine and discuss each property of -ed Adjs in depth, it is crucial to identify and shed light on the issues involved. The semantic and morphosyntactic properties and problems particularly relevant to the present study are summarised at the end of the section (Section 3.5).

3.1. Input Structure

Let us first survey the conceivable structures that can serve as the input for -ed Adjs. Before proceeding, it is important to note that the examples are primarily cited from Marchand (1969, pp. 264–67), Meys (1975, pp. 163–64), Allen (1978), Gram-Andersen (1992), Bauer et al. (2013, p. 304), and COBUILD (2017, p. 94) (see their respective works for other various examples), in addition to the Oxford English Dictionary Online (henceforth OED 2023). To the best of my knowledge, the cited data from OED (2023) remain synchronically in use.

Regarding the typical morphological characteristics of -ed Adjs, Bauer and Huddleston (2002, p. 1709) assert that “[t]he construction is extremely productive” and OED (2023, s.v. -ed suffix₂) goes as far as to say that the adjectivalising suffix -ed can be attached “without restriction to any noun” (see also Bauer et al. 2013, p. 306; Plag 2018, pp. 150–51). This statement, however, may appear too strong, particularly when considering the unacceptable cases mentioned in the Introduction (e.g., *an eyed boy; *a carred lady; *a wifed man). Nevertheless, it is true that this suffix can be affixed to various kinds of bare noun in (7) and even more complex elements, as enumerated in (8) to (10).

(7)	Bare noun:
	a.	bearded, toothed, freckled, tongued, voiced, skinned, throated, toed, heeled
	b.	booted, braceleted, jacketed, armoured, barbed, gloved, hooded, turbaned
	c.	cultured, fashioned, mannered, principled, skilled, diseased, jaundiced
	d.	patterned, pointed, spotted, dotted, striped, beaded, sized, styled, curved
	e.	wooded, pebbled, treed, tufted, flowered, leaved, mossed, hued, branched

(8)	Noun phrase (NP):4
	a.	blue-eyed, red-faced, long-legged, four-footed white-haired, heavy-handed
	b.	heavy-booted, white-gloved, double-buttoned, black-coated, high-crowned
	c.	broad-minded, low-spirited, soft-mannered, rough-featured, crazy-minded
	d.	one-sided, three-cornered, double-edged, multi-coloured, middle-sized
	e.	muddy-rivered, sandy-beached, deep-rooted, thick-stemmed, three-awned

(9)	Compound noun:
	a.	swallow-tailed, honey-tongued, pigeon-toed, pig-headed, lion-hearted
	b.	shirt-sleeved, bowler-hatted, rubber-soled, leather-aproned, sun-hatted
	c.	crime-minded, party-spirited, soul-diseased, grit-tempered, world-minded
	d.	pencil-pointed, hill-sided, cone-shaped, heart-shaped, razor-edged
	e.	beach-pebbled, oak-treed, garden-flowered, pine-forested, rain-weathered

(10)	Bound stem:
	a.	naked (legs)
	b.	wicked (uncle)
	c.	blackavised (chap)
	d.	jagged (edge)
	e.	fructed (tree)

The five divisions of input in (7) to (9), namely (a) to (e), can be roughly categorised as follows: (a) body parts; (b) wearable items such as clothes and accessories; (c) human properties or characteristics; (d) dimensions; and (e) natural objects. The remarkable productivity of the -ed suffix is evident.

The examples provided in (7) suggest that the suffix -ed can be attached to a wide range of bare nouns without strict limitations.5 The input structure of -ed Adjs in (8) and (9), composed of sequences combining a two-element nominal base and -ed, is sometimes referred to as an extended bahuvrihi compound, as exocentric Adj-N and N-N compounds are occasionally equivalent to extended forms in -ed, such as bareback(ed), duckbill(ed), long-nose(d), bighead(ed), and faint-heart(ed) (Marchand 1969, pp. 265–66; Adams 2001, p. 94).6

Regarding the bound stems in (10), some of the stems are not found in OED (2023) and other dictionaries, which may suggest that they are borrowed or have obscure origins. Harley (2006, pp. 157–58) argues that the base of naked in (10a), for instance, has become obsolete or uncommon, asserting that “[t]he infrequency of the root in words like naked, jagged, and wicked means that it might not have occurred to you before that the -ed in these words is a suffix at all”. This prompts speculation that the suffix may have undergone a process of reanalysis, amalgamating with a cranberry morpheme like nake-.7 However, this assumption warrants careful examination.8 Notably, the editors of this issue perspicaciously remark that these -ed Adjs can also be classified, albeit approximately, into the same semantic types. While naked in (10a) may appear challenging to classify, it has either or both semantic properties associated with body parts and wearable items.9 The next two -ed Adjs, wicked in (10b) and blackavised in (10c), connote human properties. Lastly, jagged in (10d) denotes dimensions, while fructed in (10e) denotes natural objects (or potentially dimensional attributes as well).

3.2. Phonological Properties

In the examination of -ed Adjs, attention is consistently drawn to their phonological properties due to their intriguing peculiarities (e.g., Marchand 1969, p. 264; Allen 1978, p. 39; Fudge 1984, pp. 65–66).

(11)	a.	/t/:	after voiceless consonants	e.g., shaped, forked
	b.	/d/:	after voiced consonants or vowels	e.g., diseased, cultured
	c.	/ɪd/:	after dental sounds such as /t/ or /d/	e.g., booted, bearded

In comparison to the general phonological rules outlined in (11), it has been acknowledged that the suffix -ed, in certain words originating from Old and Middle English, is consistently pronounced, as illustrated in (11c). Examples include crabbed, cragged, crooked, cussed, dogged, forked, jagged, knobbed, peaked, piked, ragged (cf. rugged is the ablaut variant), wicked, and wretched (Marchand 1969, p. 265). Allen (1978, pp. 39–40) explains that these words are “clearly historical relics, having a literary or picturesque flavour in most cases”, such as a peaked /pikɪd/ mountain (contrasted with high-peaked /pikd/) and a forked /fɔrkɪd/ tail (contrasted with three-forked /fɔrkd/), and are no longer synchronically productive formations with -ed (Allen 1978, p. 40).

In addition to Marchand’s denominal -ed examples, Fudge (1984, p. 66) observes that certain examples, including deverbal adjectives such as accursed, beloved, blessed, cursed, naked, and sacred, are also pronounced /ɪd/ even when not preceded by /t/ or /d/. Notably, some of them are variants of verbal past participle forms, as demonstrated in (12).

(12)		Xed	/ɪd/	/d/ (past form of ‘to V’)
	a.	aged	‘old’	‘to age’
	b.	dogged	‘persevering’	‘to dog’
	c.	learned	‘erudite’	‘to learn’
	d.	ragged	‘in rags’	‘to rag’

For instance, dogged in (12b) means ‘persevering’ when pronounced with /ɪd/ (i.e., dogged as a denominal -ed Adj), whereas it means ‘to dog, to follow someone closely and continuously’ with /d/ (i.e., dogged as a past form of the verb dog).10

Similar phonological distinctions are observed between adjectival and verbal passives, i.e., in the deverbal domain. Dubinsky and Simango (1996, §6) scrutinise the syntactic distribution of two identical passive forms (adjectival and verbal) based on their phonologically distinct behaviour. They suggest that verbs such as seem, remain, sound, and look exclusively select adjectival complements with -èd (i.e., /әd/, /ɪd/), such as abhorrèd, stripèd, avowèd, markèd, cursèd, believèd, hallowèd, redeemèd, despisèd, and unstoppèd. Dubinsky and Simango (1996, p. 777–78) conclude, based on their English historical survey of metrical verse, that -ed represents two distinct affixes and merged (sometime between the seventeenth and nineteenth centuries) into the identical spelling -ed in modern English. Importantly, although the adjectival -ed /ɪd/ has disappeared in its full syllabic form, Dubinsky and Simango (1996, p. 779) note that there are speakers who are “still quite able to use it correctly and productively (in some dialects)”. This possibility implies that, from a phonological perspective, there is still room to investigate a means of differentiating not only between the adjectival -ed and the verbal -ed in the deverbal domain, but also amongst the examples in (11) and (12), i.e., in the denominal domain.

3.3. Semantic Properties

While considering the semantic properties of -ed Adjs, the terminological ambiguity surrounding them should also be addressed and disambiguated. There exist several terms for -ed Adjs, including possessive, possessional, and proprietive. In accordance with the arguments presented by Nikolaeva and Spencer (2020, p. 101) and Nagano (2023, pp. 500–2), I adopt the term proprietive in this paper as a more precise semantic descriptor for -ed Adjs rather than the alternative terms.

Bauer et al. (2013, p. 313) analyse the suffix -ed as expressing an ornative sense, generating adjectives from nouns with the overarching meaning ‘having X, being provided with X’, wherein X corresponds to the base noun of the derivative.11 Illustrative instances include a bearded man ‘a man with a beard’; a wooded area ‘an area covered with trees’; and a blue-eyed boy ‘a boy having blue-eyes’. Here, the semantic relationship between the base noun of the derivative (e.g., beard) and the head noun (e.g., man) is the property−proprietor relationship. This semantic relationship can be illustrated by Nikolaeva and Spencer’s (2020, p. 102) examples in (13), where the nouns attached by prop.adj stand for potential derived adjectival forms of those nouns, accompanied by certain possible suffixes or case markers.

(13)	a.	campsite-prop.adj riverbank
		‘a riverbank which has (many) campsites along it’
	b.	riverbank-prop.adj campsite
		‘a campsite which is by a riverbank (as opposed to in the forest)’

In both examples in (13), albeit the constituents, campsite and riverbank are indeed identical, and their proprietive relationship diverges by virtue of the proprietive marker. Thus, whereas campsite in (13a) serves as a property for riverbank as the proprietor, this semantic relationship is patently reversed in (13b).

Despite previous studies traditionally categorising -ed Adjs as possessive adjectives (e.g., Marchand 1969, p. 264), this terminology appears to be misleading and, at times, obfuscating, as possessive adjectives basically refer to pronominal possessors (or possessive determiners), like my/your/their (book), and ethnic adjectives, such as Galilean (revolution) and Jovian (moon) (Nikolaeva and Spencer 2020, p. 96). Nikolaeva and Spencer (2020, p. 101) regard proprietive adjectives as being akin to possessive and relational adjectives, as they share the general concept of ‘being possession’ between the semantic content of the base noun and the head noun. However, the difference between proprietive and possessive functions should be attributed to the distinct nature of semantic headedness in noun phrases (Nikolaeva and Spencer 2020, p. 101; see also Beard 1995, pp. 219–27). With possessive adjectives (Poss.Adj), the relationship with the modified noun takes the form in (14a); conversely, with proprietive adjectives (Prop.Adj), the relationship is inverted, as in (14b).

(14)	a.	Poss.Adj N:	Poss.Adj<possessor)> N<possession>	e.g., my book
	b.	Prop.Adj N:	Prop.Adj<possession> N<possessor>	e.g., bearded man

Additionally, as observed by Akiko Nagano (personal communication), a similar reversal of the possession relationship can be found between the clitic and the -ed Adj, as compared in (15).

(15)	a.	the boy’s blue-eye:	the boy’s<possessor> blue eye<possession>
	b.	a blue-eyed boy:	a blue-eyed<possession> boy<possessor>

In this respect, even though clitics are not adjectives, they can be grouped together with possessive adjectives in terms of the modification relationship between the modifiers and the head nouns, although not with proprietive adjectives like -ed Adjs.

On the other hand, another term that can be employed is possessional adjectives (e.g., Jespersen 1954, II, p. 375; Beard 1976, 1981a, 1981b, 1995), which generally indicates that a noun is in possession of (or exhibits a quality of) some other noun. This term could be considered a candidate to be applied to the class of -ed Adjs; however, in reality, it encompasses a broad range of adjectives with other rival suffixes, such as warty, knowledgeable, intellectual, sorrowful, stylish, nodose, nodulous, and modular (Beard 1976, p. 155; 1981a, p. 34; 1981b, p. 36; Szymanek 1996, p. 258). Hence, I regard possessional as a superordinate term for -ed Adjs and proprietive as a more specific term for them.

On the native suffixes deriving proprietive adjectives, Nagano (2023, p. 509) analyses the affixal rivalry between -ed and -y (e.g., boned vs. bony; brained vs. brainy; edged vs. edgy; mooded vs. moody; see Nagano 2023, pp. 506–7 for additional examples). Building on Kennedy and McNally’s (2005) analysis of gradable adjectives and their four scale structures (i.e., (i) closed scale; (ii) lower closed scale; (iii) upper closed scale; and (iv) open scale (cf. Murphy 2010, §11.5.2)), she posits that, while proprietive -y adjectives (hereafter -y Adjs) represent the totally open-scale type, -ed Adjs manifest the closed-scale type. To exemplify this distinction, -ed and -y Adjs exhibit dissimilarities in their manner of modification with very, a marker associated with totally open-scale attributes, and can exclusively modify -y Adjs, as compared in (16) (examples taken from Nagano 2023, p. 507; cf. ibid., p. 510).

(16)	a.	??very dressed	very dressy
	b.	??very fished12	very fishy
	c.	??very (left-)handed	very handy

Additionally, Nagano (2023, p. 509) highlights that -ed Adjs have a morphological antonym, such as -less (e.g., clouded vs. cloudless, legged vs. legless) (see also Nikolaeva and Spencer (2020, p. 105), for discussions of caritive (or privative) adjectives, meaning ‘lacking-N’ or ‘without N’ as negative counterparts of proprietive adjectives. They further consistently accept maximality modification via completely, as evidenced in phrases such as completely cloudless.13

In addition, Allen (1978, p. 245) explains that while the typical semantics articulated by the -ed suffix is ‘characterised by the presence of X’ and extends beyond a proprietive meaning (see also OED 2023, s.v. -ed suffix₂ for the sense ‘possessing, provided with, characterised by’ (something), italics mine), certain derivatives express a simile-based meaning, as exemplified in (17), cited from Allen (1978, pp. 269–70; cf. ibid., p. 249).14

(17)	a.	child-fingered	‘having fingers like a child’
	b.	eagle-eyed	‘having eyes like an eagle’
	c.	gorilla-armed	‘having arms like a gorilla’
	d.	lion-clawed	‘having claws like a lion’
	e.	grass-seeded	‘having seeds like grass’
	f.	butcher-boned	‘having bones like a butcher’

This reading gains support from the observation that such -ed Adjs can be applied to entities beyond the denotation of the modifier. For instance, eagle-eyed in (17b) can describe things other than eagles, as illustrated in an eagle-eyed third baseman. This flexibility is similarly evident in phrases like swallow-tailed {hawk/kite}, which is ‘a {hawk/kite} which has a tail like a swallow’s’ (Allen 1978, pp. 290–1; see also Adams 2001, p. 95 for analogous instances and Norrick 2010 for discussions on this type’s scalarity and metaphoricity). This proprietive–similative (polysemy) relationship is a significant semantic factor and will be discussed in detail in Section 5.3.

3.4. Morphosyntactic Properties

In contrast to the considerable attention devoted to the phonological and semantic properties discussed earlier, the morphosyntactic properties of -ed Adjs have generally received scant exploration in previous studies. Nevertheless, scholars such as Allen (1978), Gram-Andersen (1992), Koenig and Launer (1997), Adams (2001), and Nevins and Myler (2014) have meticulously examined these properties and delineated their idiosyncratic morphosyntactic characteristics.

Initially, it is recognised that the base of the second element in -ed compound Adjs can be either a noun or verb, as illustrated in (18) (examples from Adams 2001, pp. 95–96; see Allen 1978, §4.3.3.4 for more data and discussions).

(18)	a.	Are the British colour-prejudiced?
		‘imbued with colour prejudice’ / ‘prejudiced with regard to colour’
	b.	his oak-panelled study
		‘having oak panels’ / ‘panelled in oak’
	c.	the pedal-powered drive chain
		‘having pedal power’ / ‘powered by pedals’

This issue has engendered multiple analyses and extensive discussions in Levin and Rappaport Hovav (1986), Borer (1990), Dubinsky and Simango (1996), and Bruening (2014). Addressing this categorial ambiguity, Hirtle (1969, p. 25) suggests that “the two morphemes remain distinct, one (that of the past participle) belonging to the grammatical morphology of English, the other (that of the adjective) belonging to its lexical morphology”. Gram-Andersen (1992, p. 77) briefly notes that the perfective aspect of the participles of verbs can denote a condition or state (i.e., resultative) resulting from a completed action and that, in the latter, the participle is identical to the -ed Adj not only in form but also in meaning. However, this matter is not further pursued here as it lies beyond the scope of this study.

The remainder of this section proceeds in light of the criteria for RAdjs, summarised in (3) in Section 2. The criteria are reiterated in (19) for facile reference.

(19)	a.	Predicability:	*this output is industrial, *this decision is senatorial
	b.	Gradability:	*a very industrial output
	c.	Comparability:	*more industrial
	d.	Coordinability:	*the big and wooden table
	e.	Quantifiability:	monochromatic, binational, triconsonantal, multiracial
	f.	Nominalisability:	??presidentialness, ??racialness

First, as confirmed by (19a), RAdjs do not exhibit the predication possibility. However, this characteristic cannot be simply confirmed in -ed Adjs, as demonstrated in (20) from Nevins and Myler (2014, p. 248) and (21) from Tsujioka (2002, p. 163).

(20)	a.	John is bearded.
	b.	John is moneyed.

(21)	a.	John is blue-eyed.
	b.	Mary is dark-haired.

Clearly, both -ed Adjs derived from bare nouns in (20) and from complex nouns in (21) can appear in a predicate position (see also Beard 1981b, p. 120). Nevins and Myler (2014) contend that -ed Adjs can pass the seem test (e.g., John seems blue-eyed (in this light)), establishing their truly adjectival nature. In this regard, Gram-Andersen (1992, p. 164; cf. ibid., pp. 31–32) provides several examples from diverse literary texts, as illustrated in (22).

(22)	a.	In philosophy, where truth seems double-faced, there is no man more paradoxical than myself.
	b.	The College became definitely ill-tempered.
	c.	For her sake I strove to appear light-hearted.
	d.	You shall have a present and not remain empty-handed.
	e.	It seemed simple enough. In fact, it seemed downright simple-minded.

However, compared to these observations, there are instances where the predicability cannot be confirmed. Compare the examples in (23), cited from Nevins and Myler (2014, p. 248).

(23)	a.	John is moneyed.
	b.	*John is carred.

Evidently, (23a) is acceptable, whereas (23b) is not, despite the fact that both base nouns do not have an inalienable nature. Furthermore, the complex -ed Adjs in (24) also cannot appear in a predicate position (Tsujioka 2002, p. 163).

(24)	a.	*John is white-housed.	(intended: John has a white house.)
	b.	*John is big-carred.	(intended: John has a big car.)

As an illustration, from the perspective of informativity, although carred in (23b) is rendered sufficiently informative through adding another adjective (big, as in (24b)), it remains unacceptable. In relation to predicability, the concepts of inalienability and informativity, which have been extensively debated in prior research, do not seem to provide a satisfactory resolution to this issue.15

Second, whereas RAdjs do not show gradability and comparability, as observed in (19b) and (19c), this does not seem to be the case for -ed Adjs. As evidenced in (25), complex -ed Adjs are amenable to modification by the degree adverb very, indicating their qualitative (or open-scale) status.

(25)	a.	John is very foul-mouthed.16	(Nevins and Myler 2014, p. 243)
	b.	She is very kind-hearted.	(Gram-Andersen 1992, p. 37)
	c.	You … will end up very narrow-minded.	(Gram-Andersen 1992, p. 164)

Moreover, -ed Adjs can manifest comparative forms in (26) and superlative forms in (27), respectively (examples cited from Gram-Andersen 1992, pp. 22–29).

(26)	Comparative:
	(inflectional) larger-hearted, smaller-boned, harder-headed, shorter-winded, sweeter-tempered, greyer-skinned, truer-hearted, brighter-eyed; (periphrastic) more spirited than ever, more vulgar-minded, more old-fashioned, more narrow-minded, more good-humoured, less ready-witted

(27)	Superlative:
	(inflectional) largest-hearted, truest-mannered, sharpest-sighted, swiftest-footed, longest-haired, mightiest-brained, coolest-headed, dullest-witted; (periphrastic) the most spirited person, most vulgar-minded, most old-fashioned, most narrow-minded, most renowned, the best-natured

However, as noted in (16) from the prior section, -ed Adjs can be considered a close-scale type, thus typically resisting degree modification. Similarly, concerning the comparative degree, bearded, for instance, lacks gradability and thus resists modification by the comparative modifier more, as shown in (28), taken from Koenig and Launer (1997, p. 66).

(28)	??Santa Claus is more bearded than Lenin.

Third, as regarding the coordination possibility, whereas RAdjs cannot be coordinated with QAdjs, as in (19d), the -ed Adjs in (29) are conjoined with the RAdjs anarchic and aquiline (the suppletive adjectival form of eagle), and they do so in (30) with the QAdjs tall and large.

(29)	a.	the culture is anarchic and free-spirited	(Adams 2001, p. 94, italics mine)
	b.	Ramrod-tall, blue-eyed and aquiline, with a high forehead …

(30)	a.	When the tall and bearded careers advisor…
	b.	… was at the centre of his own large and warm-spirited family
		((29b), (30): Wordbanks Online n.d., italics mine)

Generally, different types of adjectives cannot be coordinated, such as *the fierce and yellow dog and *the red and sturdy house (Koenig and Launer 1997, p. 68). Still, the examples in (29) and (30) lead us to surmise that -ed Adjs have flexible coordinability.

Fourth, in terms of quantifiability in (19e), -ed Adjs exhibit the same behaviour as RAdjs. Koenig and Launer (1997) explain that the proprietive meaning of bearded, ‘having a beard’, is “similar to dead and pregnant, one either is bearded or one is not” (p. 66). This remark can align with the three-way distinction: (i) binary oppositions (e.g., dead/alive); (ii) multiple oppositions (e.g., primary colour terms); and (iii) polar oppositions (e.g., rich/poor) (Leech 1974, p. 106ff; cf. Levi 1978, p. 20). In this regard, -ed Adjs can be classified under the second category (i.e., multiple oppositions), thus permitting their combination with quantificational (or numeral) prefixes such as mono-, bi-, multi-, and so forth, as enumerated in (31). The data are cited from OED (2023) and Webster’s (1913).

(31)	a.	mono-:	monoeyed, monolegged, monocoloured, monoped, monocled
	b.	uni-:	unilobed, unisexed, unicelled, unifaced
	c.	bi-:	biforked, biparted, bisexed, bi-lingued, bilobed, bivaulted
	d.	tri-:	tri-pointed, tri-shaped, tri-membered, trifasciated, tri-sceptred
	e.	quad-/ quarter-:	quadrisulcated, quadrigabled, quatrefoiled, quarter-pierced
	f.	penta-:	pentahydrated
	g.	hexa-:	hexaped
	h.	septi-:	septi-fronted, septi-coloured
	i.	multi-:	multieyed, multiroomed, multicored, multi-engined
	j.	poly-:	polychromed, polyspored, polyped, polytyped, polybuttoned
	k.	demi-/semi-:	demi-piqued, demi-natured, semi-calcined, semi-fasciated
	l.	omni-:	omnitooled

This property markedly distinguishes -ed Adjs from QAdjs, as most of the latter fall into binary or polar oppositions, precluding occurrence with numerical prefixes such as *mono-high, *bi-red, *tri-strong, *quadralow, *multidense, *polynear, and *omnistupid (Levi 1978, p. 24).17

Finally, serving as genuine derivatives, -ed Adjs can undergo further nominalisation with the -ness suffix, as illustrated in (32), which appears to be a diametrically opposite characteristic to RAdjs in (19f).

(32)	a.	beardedness, nakedness, wickedness, crookedness, craggedness, forkedness, ruggedness, honeyedness, multicolouredness (OED 2023, s.v. -ness suffix)
	b.	tight-fistedness, level-headedness, short-sightedness, wrong-headedness (Allen 1978, p. 240); cold-bloodedness, kind-heartedness, left-handedness (Gram-Andersen 1992, p. 38)

Consequently, -ed Adjs potentiate further nominal affixation, providing decisive evidence for the derivative function of the -ed suffix.

3.5. Summary and Problems of -ed Adjectives

The characteristics of -ed Adjs that have been examined thus far suggest that they appear to have an intermediate or dual-standard adjectival status situated between RAdjs and QAdjs. This is evident when comparing their properties to the canonical features of RAdjs presented in (19). This hybrid status can be encapsulated, as in (33).

(33)	a.	Predicability:	the man is bearded vs. *the man is white-housed
	b.	Gradability:	very foul-mouthed vs. *very handed
	c.	Comparability:	more spirited    vs. ??more bearded
	d.	Coordinability:	tall and bearded  vs. anarchic and free-spirited
	e.	Quantifiability:	monolegged, bicorned, tri-pointed, multiroomed
	f.	Nominalisability:	beardedness, nakedness, kind-heartedness, left-handedness

With respect to the first three properties outlined in (33a–c), some -ed Adjs exhibit these traits while others do not. In (33d), -ed Adjs can be coordinated with QAdjs and RAdjs. Additionally, while -ed Adjs demonstrate parallel behaviour with respect to quantifiability, as indicated in (19e) and (33e), they differ distinctly in terms of nominalisability when comparing (19f) and (33f).

Consequently, at this point, it appears plausible to suggest that -ed Adjs occupy an intermediate status between RAdjs and QAdjs, displaying a subtle departure from the quintessential properties of generic RAdjs. In other words, the classification of -ed Adjs under specific adjectival categories remains inconclusive due to their intermediate status. Moreover, previous studies have failed to reach a consensus on whether -ed Adjs align more closely with RAdjs or QAdjs, thereby providing ample opportunity to deepen our understanding of adjectivalisation with the -ed suffix. Although the ambiguous adjectival status of -ed Adjs initially obfuscates our comprehension, a thorough examination of their distinctive characteristics as QAdjs can shed light on this issue.

4. The Adjectival Status of -ed Adjectives and Their Suffixal Properties

Expanding upon the findings presented in the preceding section, in this section, I further investigate the adjectival status of -ed Adjs and provide a clear classification for these adjectives. I contend that -ed Adjs should be genuinely classified as RAdjs, adding that their QAdj-like behaviours become apparent through an analysis of conversion, which is discussed in detail in Section 5.

4.1. -ed Adjectives Are Intrinsically Relational Adjectives

Upon examining the observations in Section 3, it can be provisionally concluded that -ed Adjs bear a closer resemblance to RAdjs than to QAdjs. This is because some -ed Adjs have not exhibited QAdjs-like properties, such as predicability, gradability, and comparability, as summarised in (33a–c). These three properties are, in fact, considered the most distinct and prototypical characteristics of RAdjs, a view that is widely shared amongst the scholars mentioned in the Introduction. Furthermore, in terms of quantifiability, as indicated in (33e), -ed Adjs permit a range of numerical prefixes in a manner similar to nouns, as illustrated by examples such as monoplane, biped, triangle, quadrangle, multicylinder, and polysyllable (Levi 1978, p. 24). My contention is simply based on the fact that, because all the properties of RAdjs (as outlined in (19)) are analogous to those of nouns, if the adjectives in question display nominal properties in certain respects, it is appropriate to classify them as a type of RAdj. However, a rationale for this is necessary.

Nagano (2015, §6), from the perspective of the Weak Lexicalist Hypothesis (which deals with inflection in syntax), explains the interesting asymmetry between RAdjs and QAdjs. Nagano argues that, while derivatives with RAdj-forming suffixes (e.g., -al, -ary, -ic/-ical) can additionally possess QAdj properties, such as grammatical of a grammatical sentence in (2a) (cf. Nikolaeva and Spencer 2013, p. 224; Shimamura 2014, §4.3.5), those with QAdj-forming suffixes (e.g., -ful, -ish) never concurrently possess RAdj properties. This asymmetry arises because properties that are inherited from the base through derivation can have the potential to be lost over time, whereas properties that were not inherited through derivation cannot subsequently emerge (Nagano 2015, p. 17). Nagano then underscores that her analysis can be extended to deverbal nominalisations in a similar manner.

The degree of property inheritance from the base noun between RAdjs and QAdjs draws a parallel between the distinction observed within two categories of deverbal nouns, specifically event nominals (hereafter ENs) and result nominals (hereafter RNs). ENs exhibit significantly more verbal characteristics of their base verbs compared to RNs, and, while RNs can refer to a concrete entity, ENs refer to an event (Grimshaw 1990, p. 49). We can compare the pairs of deverbal nominals in (34) and (35), as elucidated by Grimshaw (1990, pp. 49–50, italics mine).

(34)	a.	The examination of the patients took a long time.
	b.	The examination was on the table.

(35)	a.	The constant assignment of unsolvable problems is to be avoided.
	b.	The assignment is to be avoided.

For example, examination in (34a) serves as an EN, signifying an event and taking the argument patients as its underlying transitive verb examine. In contrast, examination in (34b) functions as an RN, possessing a referential meaning while lacking an argument structure. This distinction is similarly exemplified in the usage of assignment in (35a) and (35b), with the former serving as an EN and the latter as an RN. Hence, the two types of deverbal nominals differ in the extent to which they inherit verbal properties from their bases.

Moreover, verbal gerunds (e.g., destroying in (36) cited from Spencer and Nikolaeva 2017, p. 80), compared to ENs (e.g., assignment in (35a)), express more verbal properties. For example, they take obligatory arguments (see Grimshaw 1990, pp. 50, 56 for other properties).

(36)	The Government’s/Their systematically destroying all the evidence appalled us.

Similar to a verb, destroying in (36) undergoes modification by an adverb (i.e., systematically) and directly takes a direct object (i.e., all the evidence). These verbal properties are only observed in gerunds, not in complex event nominals. In contrast, assignment in (35a) is fully nominalised, has a referential reading, is modifiable by adjectives (i.e., constant) instead of adverbs, and is incapable of directly taking a grammatical object of its base verb.

Interestingly, however, even -ing nominals have referential readings. Whereas Grimshaw (1990, p. 56) regards handwriting as a lexicalised case and does not discuss the referential reading of -ing nominals in detail, Andreou and Lieber (2020, §5) and Lieber and Plag (2022, pp. 327–28) argue that referential readings of -ing nominals are viable, though there exists an evident proclivity for numerous -ing nominals to denote eventive readings. Lieber and Plag (2022) propose that the referential readings of -ing nominals, such as rock-climbing, medical assisting, and bungee jumping, are expressed when they denote “[t]he name for a kind of hobby, pastime, occupation, or job” (p. 315). In addition, the -ing nominals covering(s) in (37) and cutting(s) in (38), taken from Andreou and Lieber (2020, p. 347, bold mine), clearly exhibit referential readings.

(37)	covering/coverings
	a.	Military History 2005: Finally he would add a waterproof covering of thin leather, bark or snakeskin.
	b.	Saturday Evening Post 1993: You want bulbs without deep scars, but small nicks and loose paperlike coverings (tunics) are acceptable.

(38)	cutting/cuttings
	a.	Horticulture 1992: Firm the moss around the cutting to protect it until it roots.
	b.	Harpers Magazine 1993: His wife walked in from outside, carrying some cuttings.
	c.	Ebony 2000: ‘Ye shall not make any cuttings in your flesh for the dead, nor print any marks upon you’.

For instance, the singular form covering in (37a) and the plural form coverings in (37b) denote the same meaning, ‘a layer that covers something’. Both cutting in (37a) and cuttings in (38b) similarly express ‘something that has been cut off (e.g., a part of a plant)’. However, cuttings in (38c) means ‘an injury made when the skin is cut with something sharp’. Andreou and Lieber (2020, p. 347) explicate that this meaning is similar to the converted form, cuts, and not derived from the singular form, cutting. Furthermore, they observe that some plural -ing nominals such as winnings, takings, meltings, and drippings have quite distinct meanings (e.g., a collective reading) from their singular counterparts (ibid., pp. 347–48). The polysemy/polyfunctionality can therefore be observed not only in Latinate nominalising suffixes such as -tion and -ment, but also in the Germanic -ing suffix.18

Accordingly, it appears that nominalisations derived as ENs can extensionally acquire RN usage, thus losing the properties of their base verbs. Conversely, it seems rare for nominalisations derived as RNs to extend to the usage of ENs; namely, it is uncommon for them to acquire verbal characteristics at a later stage.19 Turning back to -ed Adjs, it is reasonable to presume that -ed adjectivalisations as RAdjs can subsequently acquire the usage of QAdjs, but the contrary is not feasible. Hence, it is appropriate to regard -ed Adjs as an intrinsic type of RAdj.

4.2. -ed as a ‘Dedicated’ Relational Adjective Suffix

Given that -ed Adjs are unequivocally RAdjs, the following question arises: Is there any divergence between general RAdjs and -ed Adjs? I argue that considering the etymology and semantic properties of the suffixes can be a critical factor and offers a plausible answer to this question. Specifically, along with Rainer’s (2013) two distinct types of suffixes, while Latinate suffixes are considered all-purpose relational suffixes, native suffixes are regarded as dedicated relational suffixes. This analysis leads to the corollary that -ed is a dedicated RAdj suffix.20

The different properties between the two types of deverbal nominals (see Section 4.1) can be attributed to the features of their suffixal etymology. For instance, assignment in (35b) exhibits true nominal behaviour as an RN, whereas destroying in (36) shows more verbal behaviour as a verbal gerund. Paying careful attention to the suffixal etymology of these examples, it is important to acknowledge that -ment possesses the [+Latinate] feature and -ing has the [–Latinate] feature. Based on this and Nagano’s (2015) analysis, Ishida et al. (2021) (see also Ishida 2021) hypothesise that the etymological difference between suffixes (whether Latinate or Germanic) can be identified in the lexicon with the [±Latinate] feature (cf. Giegerich 1999). For example, RAdjs with [+Latinate] suffixes and [–Latinate] suffixes can be depicted as in (39).

(39)	a.	RAdjs with [+Latinate] suffixes:	senatorial, polar, algebraic
	b.	RAdjs with [–Latinate] suffixes:	bearded, wooden, southern

Thus far, the [±Latinate] feature of suffixes seems to distinguish between the two types of RAdjs (see Section 4.3 for discussion of native suffixes other than -ed), but it remains unclear how this is related to the divergence between general RAdjs and -ed Adjs. To clarify this, another perspective on the suffixal property should be examined.

Rainer (2013, §1.4) divides denominal adjectivalising suffixes into the two categories of all-purpose and dedicated. All-purpose suffixes are essentially relational and encompass various kinds of semantic relations. In contrast, dedicated suffixes express “only one specific relation or a small set of such relations” (Rainer 2013, p. 16), and they are therefore, in theory, subsumed under the all-purpose suffix(es). Based on this suffixal dichotomy, Latinate suffixes that form RAdjs can be categorised as all-purpose suffixes, as RAdjs with these suffixes exhibit the same range of denotations as their base nouns. Recall that linguists do not typically assign any meaning to the suffixes of RAdjs or, at most, attribute a highly abstract meaning to them; namely, the absence of semantic predicates in Latinate suffixes results in a contextually determined meaning of RAdj-N combination, being akin to compounds (see (1)). Conversely, the suffix -ed can be classified as a dedicated one, as it primarily expresses a proprietive meaning. Thus far, in deriving RAdjs, two types of suffixal properties and their functions can be identified, as shown in (40).

(40)	a.	If the suffix of an RAdj has the [+Latinate] feature, it serves as an all-purpose suffix and expresses genuinely relational semantics.
	b.	If the suffix of an RAdj has the [−Latinate] feature, it serves as a dedicated suffix and expresses a certain specific meaning (e.g., proprietive).

Accordingly, the [±Latinate] feature specification underscores the disparity between general RAdjs and -ed Adjs in ascertaining the type of RAdj that can be derived from a given suffix.

One may wonder how both proprietive and relational properties coexist in -ed Adjs, or which property of the two is their primary characteristic. The concurrent properties should be categorised into two different aspects, i.e., a purely semantic property and a modifying function. The former directly corresponds to the proprietive meaning of -ed Adjs and the latter occurs collaterally through their classificatory function in noun modification. The rationale for the dual semantic relationship can be associated with meronymic (or part-whole) relations between the base nouns of -ed Adjs and the head nouns (cf. Nikolaeva and Spencer 2020, p. 102; see also Nagano 2023, §3.3). It is important to note that meronymy is a distinct concept from hyponymy, as hyponymy expresses a kind relation, whereas meronymy expresses a part relation; thus, “a dog is a kind of animal, but not a part of an animal; a finger is a part of a hand, but not a kind of hand” (Cruse 2006, pp. 105–6). A meronymic relation is closely related to the input categories of -ed Adjs, given that they are typically derived from relational nouns such as kinship (e.g., brother, husband), social non-kinship (e.g., friend, governor), operational (e.g., picture, rumour), relative part (e.g., edge, corner), body part (e.g., leg, eye), and properties (e.g., height, colour) (cf. Newell and Cheung 2018). Relational nouns are known to pertain to a particular entity based on its association with another entity (Barker 2011).

Scholars, such as Takehisa (2017), Ishida (2023), and Morita (2024), argue that the input for -ed Adjs is restricted to relational nouns.21 This strongly suggests that scholars should connect relational nouns with the notion of the (in)alienable possession of -ed Adjs (see Introduction), as indicated by Spencer (2018), and the input property of -ed Adjs is generally regarded as inalienable rather than alienable. On the basis of these considerations, the particularly prominent semantic property of -ed Adjs (i.e., proprietive) prevails in their adjectival status as a dedicated RAdj suffix over the relational meaning. This outcome is consistent with the etymological feature of the suffix (i.e., Germanic) and aligns with Rainer’s (2013) suffixal dichotomy.

4.3. Relational Adjectives with Other Germanic Suffixes

The above conclusions pique our interest in similar native RAdjs in English, such as those formed by the suffixes -en (e.g., wooden) and -ern (e.g., southern). I argue that these Germanic suffixes should also be regarded as dedicated RAdj suffixes, similar to -ed (see (39b)).

Regarding the suffix -en, Adjs derived from this suffix (hereafter -en Adjs) express specific semantics, such as ‘made of X, consisting of X’, where X represents material-denoting nouns (e.g., wooden, woollen, earthen, wheaten) (Marchand 1969, pp. 270–71; see Ishida et al. 2021 for more detailed discussions about -en Adjs). Similar to -ed Adjs, the material semantics of -en Adjs clearly suggests that the suffix is a dedicated one, in contrast to Latinate suffixes (see (39a)).

As for another Germanic suffix, -ern, Marchand (1969, p. 271) explains that this only forms adjectives from nouns denoting points of the sky, such as northern, eastern, southern, and western (hereafter -ern Adjs), and it appears to be a historic extension of the suffix -en. Although a definitive conclusion regarding -ern as another type of dedicated suffix requires further exploration, Marchand’s account suggests that it possibly represents a specific meaning, such as ‘situated in X, towards X’ (see OED’s (2023) account of those -ern Adjs).

Therefore, including -ed, the discussion about dedicated suffixes thus far can be summarised in (41).

(41)	a.	-ed:	proprietive (‘having X, being provided with X’)
	b.	-en:	material (‘made of X, consisting of X’)
	c.	-ern:	locational (‘situated in X’) / directional (‘towards X’)

In comparison to these, comparison suffixes such as -al, -ary, or -ic/-ical are purely relational in nature since they do not exhibit specific semantics. This proposition bolsters my contention in (40) and offers a new perspective on RAdjs in the English language.

One might question whether -less can be considered the opposite counterpart to -ed; however, this is not the case. The true counterpart of -less is the suffix -ful, as denominal adjectives derived from -less and -ful can be inherently QAdjs based on their QAdj-properties. In this regard, the suffixes -less and -ful are dedicated QAdj suffixes. As discussed by Rainer (2013, p. 17), RAdjs seemingly never convey the meaning ‘without N’, which seems likely to be why numerous languages possess dedicated privative suffixes, such as English -less, German -los, and so forth (see Footnotes 8 and 9 for relevant discussions).

4.4. Summary and Additional Remarks

Hitherto, an examination of the adjectival status of -ed Adjs and their suffixal properties has yielded the following succinct synthesis:

• -ed Adjs are intrinsically RAdjs, as evidenced by their property inheritance from base nominals.
• -ed Adjs distinguish themselves from other RAdjs by virtue of their native (i.e., Germanic) suffixal etymology, which contrasts with the non-native (i.e., Latinate) origins of their RAdj counterparts.
• -ed is a dedicated RAdj suffix (i.e., proprietive) together with other Germanic suffixes, such as -en (i.e., material) and -ern (i.e., locational/directional).

It is pertinent to address the question of why only RAdjs formed by the native suffixes are considered to be meaning-bearing RAdjs while those formed by non-native (i.e., Latinate) suffixes are not. The answer may reside in the inherent multifunctional nature of native suffixes, a characteristic that non-native suffixes generally lack due to their limited ability to penetrate the inflectional domain. For example, the suffix -ed exhibits versatility as a regular past-tense marker, a past-participle marker, and a denominal adjectivaliser. In addition, -en/en-, which also serves multiple functions, appears as a verbal prefix (e.g., enlighten), diminutive marker (e.g., chicken), feminine marker (e.g., vixen), plural marker (e.g., oxen), deadjectival verbaliser (e.g., darken), participle marker (e.g., broken), and denominal adjectivaliser (e.g., wooden). Originally an inflectional element in Old English, -en has, probably through ‘being copied’, established its multifunctional status, much like -ed. As for -ern, it is no longer productive and does not exhibit any multifunctional nature at all (cf. Marchand 1969, p. 226; Bauer et al. 2013, p. 304). Conversely, it is challenging for Latinate suffixes forming RAdjs to acquire such multifunctionality, owing to their borrowed nature (see Ishida 2021, §5.4.2.2 for an in-depth discussion on this matter). Therefore, an examination of suffixal multifunctionality requires meticulous scrutiny and should be systematically investigated in detail.

As an additional remark, although it will not be discussed in detail, another Germanic suffix, -ly, should be mentioned. This suffix derives adjectives not only from base nouns of persons that denote ‘in the manner of X, like an X’, as in brotherly, daughterly, fatherly, and womanly, but also from temporal nouns (e.g., half-hourly, daily, monthly) and directions (e.g., easterly, southwesterly) (Plag 2018, p. 97; Marchand 1969, pp. 329–31; cf. Bauer et al. 2013, pp. 304–6). Crucially, Marchand (1969, §4.64.3) observes that, in Old and Middle English, the suffix denotes a neutral (i.e., relational) sense ‘characteristic of X, belonging to X’, as in deathly, fleshly, heavenly, lively, lovely, summerly, homely, and shapely; namely, the semantics are still the same as those of RAdjs. This fact seems to support my contention regarding native suffixes (i.e., -ly is another type of dedicated RAdj suffix). However, a more detailed analysis is necessary, such as an examination of the differences and competition between the forms of -ly Adjs and suppletive RAdjs (e.g., fatherly vs. paternal; summerly vs. aestival; nightly vs. nocturnal).

5. The Qualitative Adjective(-like) Behaviour of -ed Relational Adjectives

The puzzle of QAdj(-like) behaviour of -ed Adjs still persists. As previously mentioned in (2) of Section 2, we have observed two types of grammatical. Recall that grammatical in (2b) serves as an RAdj, while, in (2a), it functions as a QAdj. The question arises as to how these two distinct adjectival categories can be reconciled. This section employs a conversion analysis to provide an answer to this question and explores the very relationship between -ed Adjs as RAdjs and those as QAdjs.

5.1. Nagano’s (2018a) Conversion Analysis

Bauer and Valera (2005) discuss the definition of conversion as a word-class change without a form change (e.g., a bridge and to bridge), emphasising the importance of considering “how narrowly a word-class is to be defined” (p. 10) (cf. Valera 2014, 2021; Martsa 2021). They question whether homophonous pairs within a word-class (often described as secondary word-class conversion) can be considered conversion cases, such as those in (42) (cf. Bauer et al. 2013, pp. 557–59).

(42)	a.	Noun:	proper vs. common	e.g., Mary vs. the four Marys
			countable vs. uncountable	e.g., a cake vs. some cake
			concrete vs. abstract	e.g., a beauty vs. beauty
	b.	Verb:	intransitive vs. transitive	e.g., to walk vs. to walk a dog
	c.	Adjective:	relational vs. qualitative	e.g., English vs. very English

Although investigating the definition of conversion is not the primary purpose here, Bauer et al. (2013, pp. 557–59), for instance, regard these examples as cases of type-shift (or coercion; cf. Pustejovsky 1995) rather than conversion. For example, Bauer et al. (2013, p. 318) state that most RAdjs can be semantically coerced into a qualitative interpretation; however, their definition of qualitative reading remains unclear (Nagano 2018a, pp. 185–86). For example, it is uncertain whether they are implying that RAdjs can be coerced into similative adjectives. Nagano (2018a, p. 187) examines RAdj-to-QAdj alternation as a conversion rather than a coercion. Based on her analysis in the following, I similarly contend that the RAdj-to-QAdj alternation in (42c) constitutes a type of conversion.

Note that the QAdj(-like) behaviour of -ed Adjs has been clearly ascertained from Section 3. I then propose a more refined perspective, having argued that -ed Adjs are intrinsically RAdjs, with QAdj properties being subsequently acquired via conversion. This assertion gains efficacy when examining -ed Adjs from the vantage point of QAdjs rather than exploring them in the realm of RAdjs. The differentiation between homophonous pairs of RAdjs and QAdjs has previously been underscored by Bauer et al. (2013, p. 318), as illustrated in (43).

(43)	a.	Newsweek 1997: France is second—75 percent of French electricity is nuclear, which has reduced French air pollution fivefold—followed by Russia and Japan.
	b.	USA Today 2005: The outspoken Texas conservative, who displays the Ten Commandments in his office but admits he has a hard time loving his enemies, declined to run for House speaker in 1998 because he considered himself “too nuclear”.

Interpreting these examples, it becomes apparent that nuclear in (43b) undeniably carries a qualitative connotation (e.g., ‘emitting dangerous radiation’) and conveys a property amenable to gradability. In contrast, nuclear in (43a) lacks this characteristic. Instead, the latter instance preserves the basic classificatory function inherent in RAdjs. Thus, the sentence in (43a) signifies that 75 percent of electricity generated in France pertains to the nuclear kind. Here, the expression 75 percent in (43a) is a quantifier (cf. primarily, mostly, mainly, all), not a degree head. It specifies the proportion of the subject referent that belongs to the class denoted by the predicative complement NP. This interpretation is supported by the class-membership sentence without an adjective, as in This object is partially a bed (Fábregas 2014, p. 284). Thus, nuclear in (43a) is an RAdj and indicates the class membership of French electricity, while that in (43b) is a QAdj and expresses that the subject has a particular property represented by the adjective.

Addressing the issue in (43), Nagano (2018a, p. 196) expounds that the two manifestations of nuclear in (43) can be demonstrated through intransitive structures in (44a) and (44b), respectively.

(44)	a.	nuclear in (43a):	Subjecti + be + [RAdj + onei/Øi]NP
	b.	nuclear in (43b):	Subject + be + [Deg [QAdj]AP]DegP

Building upon Levi’s (1978, §7.2) intricate exposition, Nagano (2018a) argues that RAdjs occupying predicate positions results from the ellipsis originating in their initial prenominal position (cf. Nagano 2016, 2018a; Shimada and Nagano 2018; Ishida 2020, 2021; Ishida and Naya 2022). As delineated in (44a), nuclear in (43a) does not operate as a true adjectival predicate akin to QAdjs in (43b); rather, it stands as a stranded prenominal modifier, left by coreferential one-substitution (i.e., 75 percent of French electricity is the nuclear one) or bereft of its corresponding modifying target (i.e., the head noun) due to deletion.

Nagano (2018a) advances a convincing mechanism for the RAdj-to-QAdj conversion process. Her central proposition is that an RAdj within an NP “is raised into a QAdj slot through the force of the modal attributive true” (Nagano 2018a, p. 206), as depicted in (45) from Nagano (2018a, p. 202).

(45)	Input structure		Output structure
	[true [RAdj + N]NP]NP	>	[(truly) QAdj […]NP]NP

The impact of true-modification of RAdjs serves to accentuate the assertion of classificational truthfulness. As is widely acknowledged, modal attributes provide a specific value for the truthfulness operator associated with their host NP (cf. Pullum and Huddleston 2002). This assessment determines the accuracy of the classification of an entity or concept. For example, the modal attributive apparent in an apparent electrical fire “does not modify any stated ascription (it was not apparent fire or apparent electricity), but the implicitly ascribed meaning ‘caused by’” (Feist 2012, p. 85). This reveals the fact that the expression denotes “a fire that is apparently caused by electricity” (Nagano 2018a, p. 201), where apparent also takes an adverbial form, apparently, in the related clausal structure (cf. the actual cause ‘that which is actually the cause’; a certain winner ‘one who will certainly be a winner’, Pullum and Huddleston 2002, p. 557). On this basis, Nagano (2018a) applies true as a modal attributive to the constructed RAdj+N combination. The examples in (46) are taken from Nagano (2018a, p. 202).

(46)	a.	true electrical fire
		= fire that can truly be called an electrical fire
	b.	a true musical clock
		= clock that can truly be called musical clock
	c.	true Dalmatian wine
		= wine that can truly be called Dalmatian wine
	d.	true nuclear weapon
		= object that can truly be called nuclear weapon
	e.	a true English {woman/Øperson}
		= person who can truly be called English {woman/Øperson}

Nagano explicitly contends that the introduction of true and its function of changing semantics precisely represent the process of RAdj-to-QAdj conversion.

5.2. A Conversion Analysis of -ed Adjectives

We are now in a position to apply Nagano’s (2018a) insightful analysis to the case of -ed Adjs as QAdjs. Let us first observe how bare N-ed Adjs occur in gradable structures in (47). All the sentences are sourced from Wordbanks Online (n.d.).)

(47)	a.	They demonstrated that icecaps were smaller and tundra zones more wooded than had previously been imagined.
	b.	Rio Ferdinand is more cultured and will relish playing at his Manchester United home.
	c.	The ship’s insurers argue that they threw the slaves overboard because they were too diseased to be sold.
	d.	The resulting emptinesses are too mannered and elegant to be real.
	e.	St. Clair is certainly very much freckled, although I try to prevent the others from commenting on it … for I was freckled once and well do I remember it.

The italicised bare N-ed Adjs are all used as QAdjs, as they straightforwardly show gradability. Employing true-modification, we can derive the following interpretations for each case, as indicated in (48).

(48)	a.	true wooded tundra zones
		= tundra zones that can truly be called wooded tundra zones
	b.	true cultured Rio Fernand
		= Rio Fernand that can truly be called cultured Rio Fernand
	c.	true diseased slaves
		= slaves that can truly be called diseased slaves
	d.	true mannered emptinesses
		= emptinesses that can truly be called mannered emptinesses
	e.	true freckled St. Clair
		= St. Clair that can truly be called freckled St. Clair

As observed in each pair of interpretations, the internal adjective acquires distinct scalar semantics as a result of the accentuation of truthfulness. Furthermore, the analysis can be applied to complex N-ed Adjs, as illustrated in (49).

(49)	a.	a true foul-mouthed John	(cf. (25a))
		= a man named John who can truly be called foul-mouthed
	b.	a true kind-hearted lady	(cf. (25b))
		= a lady that can truly be called a kind-hearted lady
	c.	a true narrow-minded person	(cf. (25c))
		= a person that can truly be called a narrow-minded person

In the previous section, there were only a few points to corroborate our view of regarding -ed Adjs as a type of RAdjs; however, upon reflection, the analysis of -ed Adjs as QAdjs sufficiently convinces us that -ed Adjs are originally RAdjs, as well as that their QAdj behaviour, such as predicability and gradability, can be acquired through RAdj-to-QAdj conversion, which is implemented by true-modification (Nagano 2018a). In this way, the idiosyncratic properties of -ed Adjs are no longer perplexing.

5.3. The Semantics of Converted -ed Adjectives

Thus far, -ed Adjs exhibiting QAdj behaviour can plausibly be considered the outcome of the morphological process, i.e., RAdj-to-QAdj conversion. This section discusses how -ed Adjs can undergo the same conversion process to QAdjs as general Latinate RAdjs (e.g., grammatical), exploring the underlying motivation behind RAdj-to-QAdj conversion. Specifically, I argue that there are two semantic factors driving the RAdj-to-QAdj conversion of -ed Adjs: (i) class-membership relation and (ii) proprietive–similative polysemy.

In terms of class-membership relations, as reviewed in Section 5.1, RAdjs as classifying modifiers establish a class-membership relation with their head nouns. Nagano (2018a) explains that converted QAdjs deviate from similative (e.g., -esque, -ish, -like) and -ful adjectives, as the latter QAdjs are “highly compositional” (p. 197) in nature, being predicated on perceptual similarity (e.g., similative adjectives: [(subject) be close to base noun]; -ful adjectives [(subject) be full of base noun]). In contrast, the former converted QAdjs do not exhibit such compositional characteristics (e.g., nuclear in (43b) ≠ [(subject) be close to nucleus], ≠ [(subject) be close to nuclear (state)], or ≠ [(subject) be full of nucleus]). Instead, they are grounded in the notion that an entity or individual constitutes a true member of the base class, either as a representative member or a metaphorical class membership. The former case can be a highly professional professional writer (Farsi 1968, p. 56; cf. Nagano 2018a, pp. 186, 204), where professional, positioned far from the head noun, is a converted QAdj. This usage emphasises that the subject writer is a quintessential representative of the class of professional writers, showcasing their exceptional skills and expertise in the field. As for the latter, too nuclear in (43b) suggests that the individual in question truly emits hazardous radiation, implying that he possesses a dangerous level of nuclear properties or characteristics. This interpretation is based on the notion that the person is a true representative member of the nuclear class, metaphorically speaking. Another typical example is the sentence He’s more English than the English (Quirk et al. 1985, p. 1565; cf. Nagano 2018a, p. 204), which expresses the idea that the person’s characteristics encapsulate the genuine essence of Englishness, extending beyond mere resemblance to English people in the appearance or superficial traits. This interpretation indicates that the individual embodies the true nature of Englishness, making them a metaphorical representative member of the English class. These inferences are reasonably obtained from the semantic interpretations of true modification.

The case of converted -ed Adjs can be compared to RAdjs with Latinate suffixes, in that they establish a class-membership relation with their modified head nouns. For example, wooded in (47a) expresses that the tundra zones are a representative member of a class characterised by having wooded vegetation. The use of the operator true in (48a) emphasises this class membership by suggesting that the subject is a genuine example of this class. Metaphorical class membership can also be conveyed through the use of converted -ed Adjs, as demonstrated in (50a) and (50b). In these examples, wooded and bearded are used metaphorically to describe non-literal class membership.

(50)	a.	The room was wooded with dark, heavy furniture that seemed to crowd in on them, making it difficult to breathe.
		(From The Thirteenth Tale by Diane Setterfield)
	b.	The rocky cliff was bearded with a thick growth of moss and lichen, giving it a wild and untamed appearance.
		(From The Light Between Oceans by M. L. Stedman)

In the example in (50a), wooded serves as a metaphorical descriptor for the room, which is replete with dark, heavy furniture that resembles trees in a forest. In the case of (50b), bearded functions as a metaphorical characterisation of the rocky cliff, which is covered with moss and lichen that resemble a beard. Furthermore, the case of complex -ed Adjs can be found, as shown in (51), where foul-mouthed and kind-hearted are used to describe inanimate objects and abstract concepts, respectively.

(51)	a.	The car’s engine was foul-mouthed, sputtering and cursing as it struggled to turn over.
		(From The Girl Who Kicked the Hornet’s Nest by Stieg Larsson)
	b.	The small town was kind-hearted, welcoming strangers with open arms and offering help to those in need.
		(From The Help by Kathryn Stockett)

In (51a), foul-mouthed metaphorically describes the car’s engine, serving to depict the unpleasant noises emitted by the machine as being akin to cursing and swearing. Conversely, kind-hearted in (51b) figuratively characterises the subject small town by its friendly and generous inhabitants. These examples effectively illustrate metaphorical class membership through inferences derived from the semantics of the true operator.22

Overall, the use of converted -ed Adjs, whether simple or complex, can serve to establish class membership or convey metaphorical class membership. Such a class-membership relation thus not only explains the mechanism of RAdj-to-QAdj conversion of Latinate RAdjs but also the very motivation for that of -ed Adjs. This analysis further strengthens the analysis of defining -ed Adjs as a type of RAdj in Section 4.

With respect to proprietive–similative polysemy, the assertion that the -ed suffix should be considered a dedicated suffix is grounded not only in its proprietive semantics per se but also in this semantic property. First, the claim that -ed Adjs inherently convey a proprietive meaning is founded on their asymmetrical possessive–proprietive relation (see Section 3.3). Unlike other possessive or ethnic adjectives (e.g., my/your/her/their (house); Napoleonic (era), Darwinian (theory)), -ed Adjs are uniquely specialised in proprietive semantics. Next, like -y Adjs, and as discussed by Nagano (2023, §3.3), proprietive–similative polysemy is discernible in -ed Adjs. Nagano (2023, p. 519) applies Fradin’s (2007) analysis of French -eux Adjs to -y Adjs (e.g., rusty) and explains that rusty is polysemous with two readings: proprietive (as in rusty knife) and similative (as in rusty hair). In alignment with the observations in (17), -ed Adjs in (52) to (54) plainly exhibit similar behaviour; thus, the (a)-examples convey a proprietive meaning and the (b)-examples convey a similative meaning.

(52)	a.	bearded lady	‘lady having a beard’
	b.	bearded rock	‘rock covered with tufts or appendages like a beard’

(53)	a.	shirt-sleeved man	‘man wearing a shirt without a coat or jacket’
	b.	shirt-sleeved weather	‘weather pleasantly warm enough to live or work in clothing like short-sleeves’

(54)	a.	long-tailed tit	‘titmouse having a very long tail’
	b.	long-tailed pair	‘electronic circuit configuration in which the output voltage is proportional to the difference between two input voltages, characterised by having one transistor with a significantly larger current than the other, similar to how data points extend out like a long tail in a graphical representation’

For example, bearded in (52a) can be naturally interpreted as proprietive, as the head noun, lady, can have a beard as part of her body. In contrast, regarding bearded in (52b), although it still retains the general meaning of -ed, ‘covered with’, the semantic relationship between its base and the head noun (i.e., rock) cannot be inferred in a proprietive manner, as a rock as a natural object does not possess a human beard; therefore, this triggers an alternative similative meaning (i.e., ‘beard-like’). Similar contrastive observations can be applied to the other examples in (53) and (54).

In contrast to the analysis proposed by Nagano (2018a), which posits that converted QAdjs exhibit semantic distinctions from similative and -ful adjectives, the previous observations challenge this assertion. I contend that the crux of this issue resides in the suffixal attributes of RAdjs, specifically in regard to whether these suffixes are all-purpose or dedicated (see Section 4.2). RAdjs with Latinate suffixes undergo conversion into QAdjs, thereby establishing a (metaphorical) class-membership relation through their suffixal all-purpose properties. Conversely, -ed Adjs undergo a similar conversion into QAdjs; however, due to their suffixal dedicated properties and the intrinsic proprietive semantics, as well as the nature of proprietive–similative polysemy, they convey both similative semantics and a (metaphorical) class-membership relation.

Thus, analogous to the case of general RAdjs, -ed Adjs also undergo conversion into QAdjs. Nevertheless, -ed Adjs differ from typical RAdjs in that their conversions are propelled by the semantics of the dedicated suffix as well as the inference indicated by true-modification.

5.4. More on the Complex Qualitative Adjective Semantics of -ed Adjectives

Ishida (2023) provides further insights into the semantic properties of -ed Adjs as QAdjs. Certain QAdjs that undergo conversion from RAdjs can indeed exhibit a more complex figurative extension of their semantics. For instance, monochromatic, as an Radj, expresses ‘having one colour’, but its converted QAdj imbues it with the sense of ‘drab, unvarying’ (see also Ishida 2020, pp. 39–40, 42 for cases involving non-numeral prefixed RAdjs converted into QAdjs). Such further complex semantic extension can also be evinced in the case of -ed Adjs.

Ishida (2023) argues that a specific class of -ed Adj, conventionally treated as lexicalised (or idiomatic) cases (cf. Bauer and Huddleston 2002, p. 1709) due to their unpredictable meanings (e.g., forked road does not mean ‘road with or having a fork’ but rather ‘a bifurcated, branching road’), retains its intrinsic proprietive semantics. Typical examples are provided in (55), cited from Ishida (2023, p. 64).

(55)	a.	forked road	=	fork [-shaped/-formed] road
	b.	dogged persistence	=	dog [-natured/-tempered] persistence
	c.	cupped flower	=	cup [-shaped/-formed] flower
	d.	crooked business	=	crook [-natured/-tempered/-dispositioned] business
	e.	ragged coat	=	rag [-natured/-typed/-patterned] coat

For example, if the phrase forked road in (55a) is paraphrased with a certain relational noun, such as shape or form, the semantically corresponding -ed Adjs, fork-shaped/formed road, can be obtained, as provided on the right side of (55a). This analysis can be substantiated by its corresponding with-phrase (namely, road with a fork shape/form) and further enable the avoidance of unwarranted exclusions of such lexicalised cases, considering even these -ed Adjs as preserving their inherent proprietive meanings as original RAdjs. Significantly, the derived meaning in this context can be considered reminiscent of the class of similative -ed Adjs in (17) (e.g., forked road as a ‘road with a shape like a fork’).

Additional examples can be comprehended in a manner akin to the case of (55). Gram-Andersen (1992, pp. 70–71) highlights that, in certain complex -ed Adjs, even if the -ed Adjs as their second components are elliptical, the meaning of the first components is essentially preserved and is the same as that of the whole expression, as illustrated in (56).

(56)	a.	a deep-toned voice	=	a deep voice
	b.	an oval-shaped face	=	an oval face
	c.	a kind-hearted old lady	=	a kind old lady
	d.	a simple-minded fellow	=	a simple fellow

This observation underscores the true qualitative status of -ed Adjs as QAdjs, although there are instances that deviate from this pattern (e.g., a short-armed person ≠ a short person). It seems safe to say that Ishida’s (2023) analysis and Gram-Andersen’s (1992) findings are interrelated, evincing a reverse of the complex semantic pattern exhibited by -ed Adjs, as compared in (55) and (56).

Provided that Ishida’s (2023) analysis is well-founded, more intricate QAdj semantics of -ed Adjs can also be captured in a similar manner. The examples are illustrated in (57), which are cited from Tsujioka (2002, p. 164; cf. Nevins and Myler 2014, pp. 253–54).

(57)	a.	blue-blooded	=	noble
	b.	white-livered	=	timid
	c.	close-fisted	=	stingy
	d.	hard-headed	=	stubborn

For example, blue-blooded (man) in (57a) metaphorically denotes a nobility. The expression distinguishes the upper class from the working class, in that a nobility’s superficial veins through untanned skin appear blue due to not engaging in outdoor labour. This can also instantiate the similative case of -ed Adjs, namely, blue-blooded man being a ‘man with blood which looks like blue in colour’. In this manner, Ishida (2023) reveals the very mechanism of semantic extension of -ed Adjs, and the discussion of proprietive–similative polysemy further lends plausible support to this analysis. Additional instances are presented below (see also Gram-Andersen 1992, pp. 91–93 for similar examples). The bare N-ed in (58) and complex N-ed in (59) are taken from OED (2023).

(58)	a.	cankered	=	malignant, spiteful, envious (cf. bad-tempered)
	b.	floored	=	overpowered, done for
	c.	haltered	=	fettered, hampered
	d.	laureled	=	honoured, illustrious
	e.	misted	=	dulled, blurred

(59)	a.	blue-eyed	=	innocent, ingenuous
	b.	quick-eyed	=	perceptive (cf. sharp-eyed, keen-sighted)
	c.	ill-tempered	=	morose, cross, peevish
	d.	half-assed	=	contemptible, ridiculous
	e.	free-spirited	=	lively and independent; unconstrained by convention

Although each example is not examined here, as expected, these examples now serve to qualify the referents (predominantly human properties) as true QAdjs. Moreover, for more lexicalised -ed Adjs derived from missing inputs, such as wicked, naked, and jagged, this semantic extension is considered fixed (or fossilised) over time, alongside the demise of the base nouns (cf. Harley 2006).

5.5. Summary and Further Remarks

It is now evident that -ed Adjs are essentially RAdjs converted into QAdjs, providing a comprehensive understanding of their inconsistent behaviour. The preceding analysis can be summarised as follows: (i) The conversion process of -ed Adjs, in conjunction with Latinate RAdjs, is elucidated by postulating the true modal attribute. (ii) Through conversion, both -ed Adjs and Latinate RAdjs establish a class-membership relation (either as a representative member or a metaphorical class membership) with their base classes, and (iii) the resultant semantic distinction between Latinate RAdjs and -ed Adjs arises from their suffixal semantics (i.e., all-purpose ‘relational’ or dedicated ‘proprietive’). The QAdj-like behaviour of -ed Adjs is therefore not a likeness; rather, they are truly converted QAdjs, akin to Latinate RAdjs. This recognition enables the resolution of the idiosyncratic grammatical characteristics of -ed Adjs.23

As a further remark, as mentioned at the beginning of Section 5, given that RAdjs exhibit undeniable nominal properties in comparison to prototypical QAdjs (see Section 2), RAdj-to-QAdj alternation is, indeed, a conversion rather than a coercion. In this respect, although the relationship between English and very English, as in (42c), is deemed an intra-categorial conversion, it can also be viewed as a cross-categorial (or primary word-class) conversion (i.e., N-to-Adj). Conversely, the nominal domain in (42a) and verbal domain in (42b) can be considered genuine instances of secondary word-class conversion, as the former alters referents and the latter modifies argument structures within the respective domains. The adjectival domain has thus been studied in connection with the concept of mixed categories (cf. Spencer and Nikolaeva 2017; Nikolaeva and Spencer 2020). More importantly, the same imbalance in scholarly attention is evident not only in cross-categorial derivations, as mentioned in the Introduction, but also here in such intra-categorial conversions between nouns in (42a), verbs in (42b), and adjectives in (42c) (cf. Nagano 2018a, pp. 185–86). This situation necessitates a more thorough investigation into the case of the secondary word-class conversion of adjectives.

6. Conclusions

In this study, my primary achievement lies in unveiling the adjectival properties inherent to denominal -ed Adjs. The three main questions in the Introduction, the answers to which were revealed through this analysis, are repeated as follows:

• What specific properties do -ed Adjs intrinsically possess?
  ―The adjectives in question exhibit nominal properties that mirror those of their base nouns, along with distinct proprietive semantics.
• Into which adjectival categories (RAdjs or QAdjs) should -ed Adjs be classified?
  ―The relevant adjectives are categorised as a type of RAdj rather than QAdjs due to their nominal properties, which resemble those of general RAdjs. This categorisation is supported by the property inheritance observed in deverbal nominalisations and the semantic nature of proprietiveness. Additionally, it primarily hinges on the etymological features of the suffix, particularly the [±Latinate] feature. General RAdjs bear the [+Latinate] feature, while -ed Adjs exhibit the [–Latinate] feature. It is noteworthy that -ed Adjs distinctly embody proprietive semantics due to the dedicated suffix -ed, creating meaning-bearing RAdjs, with their relational meaning intricately interwoven with the classificatory function―a characteristic that stands as prototypical for all RAdjs.
• If -ed Adjs are primarily RAdjs, how can their behaviours as QAdjs be analysed and understood?
  ―The adjectives, originally classified as RAdjs, undergo conversion into QAdjs. Similar to Latinate RAdjs, this conversion process can be analysed by introducing the true modal operator. However, these adjectives still set themselves apart from Latinate RAdjs in that they represent not only a class-membership relation but also similative semantics due to the effect of proprietive–similative polysemy induced by the dedicated suffix.

Despite the limited attention afforded to this particular class of adjectives, -ed Adjs, as a subset of RAdjs, present notably intriguing characteristics. In my summarising remarks, however, -ed Adjs emerge as a distinctive category within English RAdjs, seamlessly transitioning into genuine QAdjs when imbued with additional qualitative properties.

Finally, a matter that seems highly relevant to this study, and that warrants consideration, is the issue of adjectival passives prefixed with be-. As elucidated by Bruening (2014, §7.2), these adjectival passives pose challenges involving specific adjectives, such as bedecked, bedraggled, bejewelled, belated, beloved, bemused, benighted, benumbed, bereaved, beribboned, beringed, and besotted. OED (2023, s.v. be-₇) also notes that “[s]ome of these words have no form without be-, […]”. The inherent ambiguity surrounding the bases of these adjectives, some of which seemingly derive from verbs, introduces a noteworthy complexity. However, the be- prefix, conventionally construed as aspectual, prompts intriguing inquiries regarding its attachment to potential bases. How does it harmonise with bases that may not conform to its typical aspectual function? This intricate issue, which is relevant to our exploration of -ed Adjs, demands additional scrutiny and remains earmarked for future research endeavours.