A Study of Grammatical Gradience in Relation to the Distributional Properties of Verbal Nouns in Scottish Gaelic

Abstract

Verbal nouns in Insular Celtic languages have long been a subject of interest because they are capable of exhibiting both nominal and verbal properties, posing a persistent challenge when it comes to determining their precise categorization. This study therefore seeks to examine the intersective gradience of verbal nouns in Scottish Gaelic from a functional-typological and multidimensional perspective, providing an insight into the interaction between their morphosyntactic, semantic, and pragmatic properties and their lexical categorization, and, consequently, encouraging a broader discussion on linguistic gradience. This hybrid category plays a central role in the clause structure of Scottish Gaelic, as it appears in a wide range of distinct grammatical constructions. Drawing on a range of diagnostic tests revealing the morphosyntactic and semantic properties of verbal nouns across various contexts (e.g., etymology, morphological structure, inflection, case marking, TAM features, syntactic function, types of modification, form and position of objects, distributional patterns, cleft constructions, argument structure, subcategorization, etc.), this line of research identifies two key environments, depending on whether the construction features a verbal noun functioning either as a verb or a noun. This distinction aims to illustrate the way in which these contexts condition the gradience of verbal nouns. By doing so, it provides strong evidence for their function along a continuum ranging from fully verbal to fully nominal depending on their syntactic context and semantic and pragmatic interpretation. In conclusion, the findings of this study suggest that the use of verbal nouns blurs the line between two lexical categories, often displaying mixed properties that challenge a rigid categorization.

1. Introduction: Categories

At least until the 19th century, linguists showed little interest in or—perhaps more accurately—lacked the opportunity to become thoroughly acquainted with the structural differences of languages. As a result, the prevailing approach among grammarians has been to assume that notions such as ‘word’ as well as grammatical categories like ‘noun’, ‘verb’, ‘adjective’, ‘adverb’, and ‘preposition’—whose study goes back to antiquity—were cognitively grounded and thus universally applicable (Haspelmath, 2014). However, with the increasing availability of data from non-Indo-European (IE) languages beginning in the 19th century, linguists began to observe that these languages often display markedly different organizational patterns and properties, including significant variation in the number and type of categories, word classes, or parts of speech. This led to growing scepticism about the classical teaching of nine parts of speech (pronoun, noun, verb, participle, adverb, article, preposition, conjunction, and interjection) and even raised doubts about the universality of the notion of ‘word’ itself, particularly with the rise of the concept of the ‘morpheme’. According to Haspelmath (2014), it, then, became increasingly common to argue that the analysis of non-IE languages should not be forced into the categories developed for IE languages, since descriptive categories “depend entirely on the inner form of each language” and, accordingly, each language should be described “in its own terms” (Boas, 1911).

While early comparative linguists were primarily concerned with uncovering historical and genetic relationships among languages, structuralist linguistics—which emphasized the synchronic study of individual languages—gained prominence from the mid-20th century onwards. This approach led to the proposal of new language-specific categories and the recognition that certain categories may be unique to particular languages (Sapir, 1921; Bloomfield, 1933). However, this perspective clashed head-on with the new rising Chomskyan paradigm, which posited that grammatical categories—or at least their constituent features—are universal (Chomsky, 1970; Baker, 2003). As a result, linguists largely abandoned the earlier practice of identifying and justifying the most appropriate categories for describing individual languages. Instead, the focus shifted to revealing the fundamental principles governing language across all humanity, and lexical categories such as noun, verb, adjective, and preposition came to be assumed as universal. In the following decades, alternative classificatory frameworks received limited attention and, particularly with regard to categories, it was common to essentially reaffirm Chomsky’s position.

From the late 20th century on, however, a number of scholars (e.g., Croft, 1991, 2000, 2023; Hengeveld, 1998; Van Valin & LaPolla, 1997; Dik, 1997; Evans, 2000; Baker, 2003; Dixon, 2010; Chung, 2012; among others) began to question the universality of these four major lexical categories, but a lack of consensus regarding the number of cross-linguistic categories became evident and proved difficult to resolve. This may, to some extent, be related to the difficulty in finding the right set of criteria to distinguish between word classes (Crystal, 1967) or, more importantly, in establishing a definition enabling the universal identification of distinct word classes.

In an effort to address the limitations of strictly morphosyntactic criteria for defining lexical categories, various scholars have proposed semantically grounded characterizations of major word classes. Givón (1984) offers a semantic perspective, for example, by characterizing nouns and verbs in terms of time-stability. Nouns are thus said to represent experiences that remain stable over time, whereas verbs refer to rapidly changing states of affairs. He also acknowledges the prototypical nature of some members of a category, which are considered to be more central than others. Similarly, Langacker’s Cognitive Grammar (Langacker, 1987a, 1987b, 2002, 2008, 2009) concurs that prototypical nouns and verbs denote physical objects and overt physical actions, respectively, but introduces the idea of higher-level schemas, which represent universal, schematic realizations that underlie all category members and are grounded in basic cognitive processes. Word classes are thus defined on two levels: a more general level showing universal semantic schemas where nouns and verbs are defined as linguistic expressions that designate a thing and a process, respectively, and a more specific level offering local instantiations that range from prototypical to non-prototypical members. For his part, Croft (1990, 2001, 2007) synthesizes semantic and pragmatic perspectives by proposing three semantic classes—objects, actions, and properties, which correspond to three pragmatic functions—reference, predication, and modification. These classes and functions are in turn aligned with the traditional categories of nouns, verbs, and adjectives, respectively, which serve as typological prototypes rather than fixed grammatical classes. Importantly, these prototypes describe only the core of each category, while their boundaries are defined by the grammar of individual languages. Finally, Haspelmath (2010) suggests that grammatical comparison must not begin with language-specific categories but rather with comparative concepts that are clearly distinct from descriptive categories. More specifically, an accurate analysis of word classes must take into account the semantic properties associated with the comparative concepts considered fundamental to communication, namely the capacity to refer and to predicate, and even to describe properties of entities. These functions appear to correspond to the traditionally used categories of noun, verb, and adjective, which, as Haspelmath (2012) argues, can be readily identified across languages and, as Croft (2000) points out, can be considered typological prototypes. In sum, these semantically oriented models offer a more flexible and cognitively grounded understanding of lexical categories, while acknowledging cross-linguistic variation in their grammatical instantiation.

In an attempt to sum up the situation, the evidence provided by the analysis of a wide range of languages from very diverse families reveals that there are hardly any languages that lack a definitive distinction between nouns and verbs (e.g., Schachter, 1985; Sasse, 1993; Davis et al., 2014; Vogel & Comrie, 2000; Haspelmath, 20011, 2023), although this distinction is not always as clear-cut as it is in IE languages2. Adjectives also appear to be widely attested cross-linguistically, although they are less universally represented than verbs and nouns. Finally, as far as the other categories are concerned, for example prepositions, adverbs, pronouns, determiners, quantifiers, numerals, coordinators and subordinators, and particles, these tend to vary much more significantly from one language to another.

One of the language families in which the distinction between the categories of verb and noun is not as well-defined as might be expected is the Celtic family. In these languages, a hybrid category exists that is commonly referred to as the verbal noun, which, from the earliest attested stages of its historical development, appears to demonstrate that the boundary between verbal and nominal categories is far from clear-cut, despite the bias suggested by the name generally attributed to this category. As a result, the study and analysis of verbal nouns have consistently attracted considerable scholarly interest, so that now this work will focus on an investigation of this category in Scottish Gaelic from the perspective of intersective gradience.

The organization of this paper is as follows: after a brief account of the most relevant typological properties in Scottish Gaelic in Section 2, Section 3 introduces the notion of the verbal noun in Celtic linguistics and provides a summary of the way in which this category has traditionally been analysed with respect to its origin and nature as either a nominal or a verbal element. Section 4 explains the concept of gradience in relation to word classes, parts of speech, or categories, and more specifically of intersective gradience, where an item exhibits characteristics of multiple word classes, thereby illustrating the intersection between categories. Section 5 offers a comprehensive investigation of the distribution and functions of the verbal noun in Scottish Gaelic by examining a range of morphosyntactic and semantic properties across various contexts with the aim of determining whether the category of verbal nouns in Scottish Gaelic has more verbal or more nominal properties. A detailed discussion of the results derived from this analysis is provided in Section 6, and, finally, a summary of the most relevant findings obtained in this research concludes this paper.

2. Typological Description of Scottish Gaelic

Scottish Gaelic is a language belonging to the Goidelic branch of Insular Celtic languages3 and is mostly spoken in Scotland by about 69,000 speakers (NRS, 2022), who exhibit a considerable degree of diatopical variation in their speech. In order to clarify the discussion concerning verbal nouns in this language, an introductory typological overview of some of its most relevant phonological and morphosyntactic features is included here. The most distinctive phonological feature of Celtic languages may be the existence of an elaborate system of initial mutations, which entail the alteration of the initial phoneme of words. More specifically for Scottish Gaelic, the most common type of mutation is lenition, which involves the softening of the initial consonant of a word—often marked orthographically by the insertion of an “h” in spelling (e.g., bòrd ‘table’ → bhòrd)—in a variety of morphosyntactic contexts, such as after the vocative particle, the article in some cases, certain possessive pronouns, and specific prepositions, particles, adverbs, numerals, and adjectives, among others.

According to conventional morphological typology, Scottish Gaelic is considered a mildly synthetic, largely fusional, language, since, while it now has a relatively low morpheme-to-word ratio and a fairly regular morphology, in the past its verb forms have included a number of inflectional affixes expressing different grammatical meanings. Also, there are now instances of inflected, fused, or conjugated prepositions, which have historically developed from the contraction of a preposition with a personal pronoun (i.e., prepositional pronoun) or with a possessive (i.e., prepositional possessive):

(1)	Cha	chaidilinn	gu	math-san	taigh-tughaidh
	neg	sleep.cond.dep.1sg4	adv	good in.the.dat	house-thatched.dat
	aca	rè	mo	shaor-làithean	samhraidh.5
	at.3pl	during	1sg.poss	free-day.gen.pl	summer.gen
	‘I would not sleep well in their thatched house during my summer holidays.’6

Some of the words contained in this sentence reflect the previously synthetic nature of Scottish Gaelic. One example is the verbal dependent form chaidilinn, which is composed of the verbal root caidil7—this verb appeared in Old Irish8 as co-tlud, which is formed by a prepositional preverb meaning ‘with’ and a verbal root meaning ‘sleep’ (Calder, 1923, p. 197)—and the suffix-inn, which expresses the first person singular in conditional. The formation of verbal forms by adding pronominal suffixes to a verbal stem was common in Old Irish (Thurneysen, 1946/1980, pp. 455–460), but this method gradually gave way to a more analytical procedure. Consequently, in Modern Scottish Gaelic, personal pronouns are expressed as independent forms, rather than by means of verbal endings.9 Other morphological properties of Scottish Gaelic are: the distinction made between the masculine and feminine genders—a neuter gender once existed but has now disappeared, the existence of a four-way case distinction (nominative-accusative, dative, genitive, and vocative)10, the presence of inflected, fused, or conjugated prepositions, the distinction between independent and dependent verb forms, the existence of an impersonal verb form, and the use of ablaut as a morphological device to mark grammatical distinctions in terms of number in nouns, tense in verbs, and degree in adjectives.

With respect to word order, Scottish Gaelic tends to have the tensed verb in clause-initial position—after preverbal particles indicating subordination, illocutionary force, and negation in dependent, interrogative, and negative sentences11, although it is also possible to arrange a non-verbal constituent in a preverbal position for pragmatic purposes, as an epiphenomenon of information structure. As regards the parameter referred to as argument-type, Scottish Gaelic may be considered a lexical-argument language now, as the overt lexical nominal phrases are the true arguments of the predicate. It is, likewise, a dependent-marking language because all grammatical relations are coded in nominal phrases—which can be morphologically unmarked and which occur in the common case, or marked as dative or genitive, rather than within the verb. Scottish Gaelic is an instance of a right-branching language owing to the fact that it tends to place dependents after heads (i.e., it has prepositions, and the nouns are commonly followed by adjectives, although nominal objects may precede or follow their verbs depending on the type of construction). In terms of alignment, this language can be said to exhibit a nominative-accusative alignment, as, in the early stages of the language, the subjects of intransitive verbs were treated like the subjects of transitive verbs and were therefore distinguished from the objects of transitive verbs in basic clause constructions.12 Finally, like the other Celtic languages, Scottish Gaelic makes use of two non-finite verbal forms, namely the verbal adjective and the verbal noun (or verb-noun). Regarding the first non-finite form, the verbal adjective, it suffices to note that it descends from the old past participle found in Old Irish. It was declined as an adjective and had verbal features, such as voice and, to a certain extent, tense (Thurneysen, 1946/1980, pp. 441–443; Russell, 1995, pp. 258–259), and is used today to express a passive meaning and perfect aspect in a specific passive-like construction:

(2)	Tha	an	leabhar	sgrìobhte	mu thràth.
	be.pres	the	book	write.va	already
	‘The book is already written.’

The second form, the verbal noun, will be discussed in more detail in the following sections, as it constitutes the main focus of this study due to its ambiguous nature.

3. The Verbal Noun

Verbal nouns in Insular Celtic languages have long been a subject of scholarly interest due to fact that they demonstrate both nominal and verbal properties. Consequently, numerous studies have been devoted to their investigation, both from a comparative perspective encompassing all Celtic languages (Wagner, 1959; Gagnepain, 1963; Russell, 1995, 2015) as well as through focused analyses of individual Celtic languages (Jeffers, 1978; Disterheft, 1980; for Old Irish; McCloskey, 1983; Guilfoyle, 1997; for Irish; Sproat, 1985; Willis, 1988; Fowkes, 1991; Borsley, 1993, 1997; Asmus, 2025; for Welsh; S. R. Anderson, 1981; Stephens, 1990; Timm, 1990; for Breton; Cram, 1984; Ramchand, 1993; Adger, 2010; for Scottish Gaelic; Lewin, 2016; for Manx; among others) in order to ascertain whether they should be considered either as verbs or nouns.

Despite extensive research, the synchronic status of verbal nouns in Celtic languages remains a subject of scholarly debate. Moreover, the limited historical evidence from stages prior to Old Irish precludes a definitive conclusion as to whether their origin was fundamentally nominal or verbal. Consequently, given the scarcity of early textual documentation prior to Old Irish, and although any claims regarding their diachronic development should be considered tentative, various scholars have nevertheless ventured to take a position on the issue by examining the origin of verbal nouns in other languages, tracing them back to Proto-Indo-European (PIE).

Although any hypothesis about their precise origin is speculative—especially considering that the earliest evidence available stems from a protolanguage, this line of inquiry led early scholars to argue for the nominal origin of verbal nouns based on their presumed historical derivation from abstract action nouns (Zeuss, 1871; Brugmann, 1886–1900/1897–1916).13 For instance, Thurneysen (1946/1980) and Gagnepain (1963) classified the verbal noun unequivocally as a substantive. Similarly, Jeffers (1978) asserted that it was clearly nominal in nature, a characteristic that he associated with the change to a VSO structure, as Lehmann and Lehmann (1975) suggests. Disterheft (1980) also noted that its nominal function was recognized very early, but began to acknowledge that some of its uses would indicate properties more typically associated with verbs. Subsequently, in a similar vein, in their examination of Modern Irish, Breton, and Welsh, McCloskey (1980, 1983), S. R. Anderson (1981), Willis (1988), Timm (1990) and Fife (1990) also treated the verbal noun as a nominal element, but one that also showed verbal properties in some contexts, a position later echoed by Cram (1981), Ramchand (1993), and Adger (1996, 2010) in their analyses of Scottish Gaelic.14

However, this dual interpretation has been increasingly refined by scholars who tend to attribute the mixed categorial properties of verbal nouns to their complex diachronic development. Russell (1995, 2015) supports this view, arguing that verbal nouns in Celtic started out as nouns and have gradually become verbs across their documented history. This process of reanalysis is also emphasized by Ronan (2006), who states that during earlier stages of the Celtic languages, the verbal noun was unambiguously nominal in inflection and usage, whereas the incorporation of verbal features over time reflects an increasing tendency toward verbalization15. Lewin (2016) likewise points out that, while the verbal noun in Manx was originally a noun, it has been gradually reanalysed as a non-finite verb, though it continues to exhibit formal traces of its erstwhile nominal character.

Finally, as an illustration of an alternative, yet not necessarily exclusive perspective, Borsley (1993) states that, in Welsh, verbal nouns are basically verbs, although they can sometimes appear as nouns. Rouveret (1994) goes even further by arguing that, in Welsh, verbal nouns should be analysed as verbs embedded within a Determiner Phrase (DP). Similarly, Adger (2022a, 2022b) describes the verbal noun in Scottish Gaelic as a nominalization of the verb that operates across contexts typically occupied by infinitives or participles in other IE languages.

In conclusion, although the evidence provided by PIE appears to support a nominal origin of the verbal noun, and subsequent research on Celtic languages has increasingly highlighted both its diachronic dynamism and its current categorial ambiguity, which may lead us to recognize the verbal noun synchronically as a mixed or transitional category, there appears to be no scholarly consensus as to whether the verbal noun is more nominal or verbal in nature.

4. Gradience

For centuries, the dominant framework for understanding categorization was grounded in the rigid classical or Platonic-Aristotelian model. Under this model, category membership was determined by a set of necessary and sufficient conditions with rigid boundaries that allowed no room for variation, so an entity either fully belonged to a category or it did not; there was no in-between. However, despite its longevity and influence, this classical view began to be seriously questioned in the 20th century, as a growing body of empirical evidence highlighted its limitations. Studies from multiple fields began to challenge the rigidity of the classical model. For instance, key research in colour terminology (Berlin & Kay, 1969), speech variation (Labov, 1973), and psychological categorization (Rosch, 1978) illustrated the fact that natural categories often do not conform to strict binary divisions. Instead, these findings pointed to the idea that some members of a category are perceived as more central or typical than others, leading to the understanding that categories are structured around prototypes—ideal or the most representative examples—rather than through a fixed checklist of defining features. Over the following decades, semantically oriented scholars such as Jackendoff (1983, 1990), Langacker (1987a, 1990, 2002), and Lakoff (1987) were especially instrumental in articulating and refining this new view of categorization, which viewed category membership as graded rather than absolute, allowing for degrees of membership and recognizing the notion of gradience (Bolinger, 1961), which posits that linguistic categories and grammatical constructions often demonstrate internal variation and overlap with other categories. This conception of gradience has been fundamental to cognitive and functional approaches to language, and grammatical phenomena that previously appeared anomalous or irregular under the classical model have been reinterpreted as natural variations within a graded structure, illustrating the fact that categories are better understood as continua rather than discrete units.

In relation to word classes, parts of speech, or categories, the distinction between subsective gradience and intersective gradience pertains to two different types of categorical overlap or boundary blurring. Thus, on the one hand, subsective gradience refers to a type of gradience within a single category, whereby certain members of that category are more prototypical than others. In this case, all items are still classified under the same grammatical category but tend to vary in the degree to which they exemplify the core features of that category. On the other hand, intersective gradience involves an overlap between two distinct grammatical categories, since it refers to those linguistic elements with properties of two (or more) categories simultaneously, thereby occupying an intermediate position between them. Intersective gradience, therefore, highlights the permeability and continuity between categories, challenging the notion of strictly discrete grammatical boundaries. An example of intersective gradience can be seen in the category of verbal nouns in Celtic languages, as they are words that blur the boundaries between verbs and nouns.

5. Analysis of the Verbal and Nominal Characteristics of Verbal Nouns

Despite the theoretical challenges involved in defining word classes more generally, in practice it is often not difficult to reach an agreement regarding the identification and use of verbs and nouns in a given language. This is largely due to the fact that these categories tend to exhibit significant cross-linguistic similarities in their behaviour. Since the aim of the present study is not to contribute to the ongoing debate concerning the number of universal categories or to define the criteria that should be applied to distinguishing between them cross-linguistically, but rather to determine whether the category known as the verbal noun displays more verbal or nominal properties in Scottish Gaelic, my analysis will focus on examining its behaviour in a number of grammatical contexts with features traditionally employed to define verbs and nouns, which are the two categories between which verbal nouns exhibit intersective gradience16; hence the topic of this paper is to investigate the properties of this mixed or hybrid category, which is at times considered as nominal, and at other times regarded as an uninflected, verbal form, in Scottish Gaelic. To this end, the grammatical contexts are classified in two distinct subsections, depending on whether the features of the category under examination are more nominal (Section 5.1) or verbal (Section 5.2) in nature. While the evidence for the existence of gradience in verbal nouns in Scottish Gaelic is purely synchronic, as the examples are instances of Modern Scottish Gaelic, references to diachronic developments in specific constructions involving this syntactic category are provided whenever they are available.

5.1. Nominal Contexts

Verbal nouns act as nouns in sentences in several ways, as they serve as arguments of predicates, can show case, gender, and number distinctions, appear with determiners and numerals, and may accept adjectival, rather than adverbial, modification. Furthermore, their complement may take a genitive case and they themselves have genitive case marking both when they complement other nouns and when they appear in compounds. Additionally, they may have possessives instead of personal pronouns to realize their objects, may be modified by a relative clause, do not appear to show verbal morphology, are negated differently from finite verbs, can occur after prepositions, do not show a difference between independent and dependent forms, and lack voice distinctions. Finally, besides these properties, which are widely regarded as hallmarks of nominal elements, verbal nouns also display other features that could further underscore their nominal nature (e.g., the ambiguous role and the position of the possessive, the NP and PP objects, and the reflexive pronoun, the formation of compounds and of periphrases with the light verb dèan ‘do’, and the presence of lenition effects in aspectual constructions). The following grammatical contexts serve to identify the verbal noun as a category of a nominal nature in Scottish Gaelic.17

One of the most distinctive features of nominals is that they can act as the subject18 or object of an inflected verb and as the object of a preposition. Like nominals, verbal nouns can behave as subjects (3) and objects (4) of verbs as well as objects of prepositions (5):

(3)	Tha	cuideachadh	dhaoine	eile	gam	fhàgail	toilichte.
	be.pres	help.vn	person.pl	other	at.1sg	leave.vn	happy
	‘Other people’s help makes me happy.’

(4)	Chunnaic	iad	mo	bhualadh.
	see.past	3pl	1sg.poss	beat.vn
	‘They saw my beating.’

(5)	Tha	mi	ann	an	Gaol	le	còcaireachd.
	be.pres	1sg	in.3sg.m	in	love.dat	with	cook.vn.dat
	‘I am in love with cooking.’

The nominal status of verbal nouns seems undeniable when we observe that they decline like nouns. Although nominal inflection has been retained in Modern Scottish Gaelic only in a gradually reduced form, morphologically, a verbal noun can be inflected for the common case (6)—which is unmarked, the dative case (7), and the genitive case (8), just like a noun, such as in the examples given below:

(6)	Tha	seinn	furasta	do	Phòl.
	be.pres	sing.vn	easy	for	Paul.dat
	‘Singing is easy for Paul.’

(7)	Cha	tuirt	i	dad	mun	t-seinn.
	neg	say.past.dep	3sg	anything	about.the.dat	sing.vn.dat
	‘She did not say anything about the singing.’

(8)	Cha	do	bhruidhinn	duine	rè	na	seinn(e).
	neg	pfv	talk.past.dep	nobody	during	the.gen	singing.gen
	‘Nobody talked during the singing.’

Additionally, when used as a noun, the verbal noun is normally masculine (MacLaren, 1923, p. 144; MacFarlane, 1912/1948, p. 165)—like leughadh ‘reading’ in (9), but there are also feminine verbal nouns—like togail ‘building’ in (11). In both genders, the verbal noun may also appear in the plural (10 & 12):

(9)	Chòrd	an	leughadh	aice	rium.
	agree.past	the	read.vn	at.3sg.f	to.1sg
	‘I liked her reading.’

(10)	Bidh	mòran	leughaidhean	anns	an	tachartas	sin.
	be.fut	many	read.vn.pl	in	the.dat	event.dat	dem.dist
	‘There will be many readings at that event.’

(11)	Bha	a	theagasg	Gearmailtis	airidh	air	moladh.
	be.past	3sg.m.poss	teach.vn	German.gen	merit	on	praise
	‘His teaching of German was praiseworthy.’

(12)	Bha	teagasgan		Ìosa		glè	fheumail	dha
	be.past	teach.vn.pl		Jesus.gen		very	useful	to
	Na		deisciobail		aige.
	the.pl.dat		disciple.pl.dat		at.3sg.m
	‘Jesus’ teachings were very useful to their disciples.’

Verbal nouns also behave like nouns because they appear with determiners and numerals. Both articles (13), which are almost always definite, and numerals (14) precede the verbal noun:

(13)	Bha	an	tadhal	aca	glè	thaitneach
	be.past	the	visit.vn	at.3pl	very	pleasant
	‘Their visit was very pleasant.’

(14)	Cha	do	rinn	mi	ach	aon	tadhal	air	bràthair
	neg	pfv	do.past.dep	1sg	but	one	visit.vn	on	brother.dat
	mo		mhàthar.
	1sg.poss		mother.gen
	‘I made only one visit to my uncle.’

Another nominal feature of verbal nouns is that they can also be modified by adjectives:

(15)	Ceart-sgrìobhadh19
	correct-write.vn
	‘Orthography’ (lit. ‘Correct writing’)

(16)	Bha	coiseachd	bhrèagha	ann	an-dè.
	be.past	walk.vn	beautiful	in.3sg.m	yesterday
	‘There was a beautiful hiking yesterday.’

The fact that the phrase containing the verbal noun is essentially nominal in character can be observed in that, unlike the direct object of a finite verb (17) and like the object of a noun (18), the direct object of a verbal noun may appear in the genitive case (18), although in contemporary usage it most commonly does so if the object is definite. This situation, which can be observed in constructions expressing the progressive (19) and purposive (20) aspect20, means that what is assumed to be the direct object of an untensed verb—marked with the genitive—is in fact the possessive complement of a noun.

(17)	Dh’ionnsaich	mo	bhràthair	na	cànanan	sin.
	learn.past	1sg.poss	brother	the	language.pl	deic.dist
	‘My brother learnt those languages.’

(18)	Ionnsachadh	nan	cànanan	sin.
	learn.vn	the.pl.gen	language.pl.gen	deic.dist
	‘The learning of those languages’

(19)	Tha	mo	bhràthair	ag	ionnsachadh	nan
	be.pres	1sg.poss	brother	prog	learn.vn	the.pl.gen
	cànanan	sin.
	language.pl.gen	deic.dist
	‘My brother is learning those languages.’

(20)	Tha	mo	bhràthair	a’	dol	a21	dh’ionnsachadh
	be.pres	1sg.poss	brother	prog	go.vn	purp	learn.vn
	nan	cànanan		sin.
	the.pl.gen	language.pl.gen		deic.dist
	‘My brother is going to learn those languages.’

Verbal nouns can be modifiers of other nouns—so the verbal noun may be said to perform an adjectival function in this context—and, as Scottish Gaelic still retains nominal distinctions like case, the verbal noun is in the genitive case in this situation almost categorically:

(21)	Cunnart	dràibhidh
	danger	drive.vn.gen
	‘The danger of driving’

(22)	Ealain	Sgrìobhaidh
	art	write.vn.gen
	‘The art of writing’

Another use of the verbal noun that could be considered nominal—or even adjectival, as it expresses a property of another noun, which serves as the head of the noun phrase—is observed in compounds, where verbal nouns can bear the genitive case. While this does not normally occur in Scottish Gaelic, some examples may be found:

(23)	draibheadh-cadail
	drive.vn-sleep.vn.gen
	‘Sleepdriving’ (lit. ‘Driving of sleep’)

(24)	Teisteanas-breithe
	certificate-bear.vn.gen
	‘Birth certificate’ (lit. ‘Certificate of birth’)

A clearer nominal property exhibited by verbal nouns is that, unlike finite verbs, they have possessives22 or prepositional objects, rather than object personal pronouns. Thus, the fact that, unlike a finite verb (25), a verbal noun cannot be followed by a direct object pronoun, and that the so-called direct object of this untensed verb is in fact a prepositional possessive (26) or (a preposition plus) a possessive (27), which generally accompanies nouns, seems to suggest that as gam chuideachadh (26) or (air) mo chuideachadh (27) could be rendered in English as ‘at my helping’ or ‘(on) my helping’, respectively.23

(25)	Chuidich	e	mi.
	help.past	3sg.m	1sg
	‘He helped me.’

(26)	Tha	e	gam	chuideachadh.
	be.pres	3sg.m	at.1sg	help.vn
	‘He is helping me.’

(27)	Tha	e	air	mo	chuideachadh.
	be.pres	3sg.m	perf	1sg	help.vn
	‘He has helped me.’

It is also commonly said that, although arguments are obligatory, they can sometimes be ambiguous, especially in nominal contexts. This happens, for instance, in the two following examples where the argument cross-referenced by the prepositional possessive gam ‘at my’ is ambiguous, as it can be understood as coreferential with either an agent subject (28) or a theme subject (29):

(28)	Tha	an	saighdear	sin	ga	leigheas.
	be.pres	the	soldier	dem.dist	at.3sg.m	cure.vn
	‘That soldier is curing him.’

(29)	Tha	an	saighdear	sin	ga	leigheas.
	be.pres	the	soldier	dem.dist	at.3sg.m	cure.vn
	‘That soldier is being cured.’

Additionally, like a regular noun, a verbal noun may also serve as the head of a relative clause, although this use appears to be rather unusual:

(30)	Cha	robh	an	t-ionnsachadh	a	rinn	iad	math.
	neg	be.past.dep	the	learn.vn	rel	do.past	3pl	good
	‘The learning that they did was not good.’

Also, it is widely acknowledged (Pedersen, 1909–1913; Baudiš, 1913; Thurneysen, 1946/1980; Gagnepain, 1963; Disterheft, 1980; among others) that verbal nouns in Celtic languages display nominal characteristics as a result of their nominal functional morphology24. While they are typically derived morphologically from the same root as the base verb, in Scottish Gaelic, like other Celtic languages, there is no single dedicated verbal noun marker; instead, verbal nouns are typically, though not exclusively, formed by suffixation to the root.25 The most common—and the most productive suffixes today—are: (e)adh (e.g., pòs ‘marry’—pòsadh (vn)), -inn (e.g., faic ‘see’—faicinn (vn)), -sinn (e.g., ruig ‘arrive’—ruigsinn (vn)), -tinn (e.g., cluinn ‘hear’—cluinntinn (vn)), -tainn (e.g., can ‘say’—cantainn (vn)), -(e)amh (e.g., feith ‘wait’—feitheamh (vn)), -ad (e.g., greas ‘hurry’—greasad (vn)), -ail (e.g., gabh ‘take’—gabhail (vn)), -eil (e.g., tilg ‘throw’—tilgeil (vn)), -e (e.g., ith ‘eat’—ithe (vn)), -(e)achadh (e.g., èist ‘listen’—èisteachadh (vn)), -aidh (e.g., iarr ‘want’—iarraidh (vn)), -(a)ich (e.g., ràn ‘roar’—rànaich (vn)), -t (e.g., freagair ‘answer’—freagairt (vn)), and -eam (e.g., tuit ‘fall’—tuiteam (vn)).

Most verbal nouns have the same stem as the corresponding finite verb forms (e.g., òl ‘drink’—òl (vn), falbh ‘leave’—falbh (vn), ruigh ‘run’—ruigh (vn), reic ‘sell’—reic (vn), snàmh ‘swim’—snàmh (vn), cluich ‘play’—cluich (vn), coimhead ‘watch’—coimhead (vn), seinn ‘sing’—seinn (vn), etc.).

Some verbal nouns are slightly different from the corresponding verb stem (e.g., beir ‘catch’—breith (vn), thoir ‘give’—toirt (vn), teagaisg ‘teach’—teagasg (vn), cuir ‘put, send’—cur (vn), ceannaich ‘buy’—ceannach (vn), caidil ‘sleep’—cadal (vn), etc.).

Others change the final syllable of the root and add a suffix like -eachd, -adh, -eachadh, or –eadh (e.g., coisich ‘walk’—coiseachd (vn), faighnich ‘ask’—faighneachd (vn), fuirich ‘stay, wait’—fuireach (vn), ionnsaich ‘learn’—ionnsachadh (vn), cuidich ‘help’—cuideachadh (vn), dannsa ‘dance’—dannsadh (vn), peant ‘paint’—peantadh (vn), cuimhnich ‘remember’—cuimhneachadh (vn), dràibhig ‘drive’—dràibheadh (vn), etc.).

Finally, there are verbal nouns that are not actually morphologically related to their corresponding verb, so they use suppletive verbal noun forms (e.g., abair ‘say’—ràdh (vn), rach ‘go’—dol (vn), etc.).

While this context is not so clearly nominal, the verbal noun cannot be negated like the finite verb. Conversely, the verbal noun is regularly negated by the negative preposition gun ‘without’, rather than by the negative particle cha ‘not’:

(31)	Is	ann	ainneamh	a	dhèanadh	iad	e
	be.pres	foc	seldom	rel	do.cond	3pl	3sg.m
	gun	innse.
	without	tell.vn
	‘Seldom would they do it without mentioning (it).’

(32)	Thuirt	mi	ris	gun	a	dhol	a-mach.
	say.past	1sg	to.3sg.m	without	inv	go.vn	out
	‘I told him not to go out.’

Another context that appears to reflect nominal properties—although it may be influenced by the distinction between finite and non-finite verb forms—is related to the contrast between independent and dependent verbal forms. Unlike the finite verb (33 & 34), the verbal noun does not show the independent/dependent distinction (35 & 36):

(33)	Chunnaic	mi	Màiri.
	see.past	1sg	Mary
	‘I saw Mary.’

(34)	Chan	fhaca	mi	Màiri.
	neg	see.past.dep	1sg	Mary
	‘I didn’t see Mary.’

(35)	Cho-dhùin	mi	Màiri	fhaicinn.
	decide.past	1sg	Mary	see.vn
	‘I decided to see Mary.’

(36)	Cho-dhùin	mi	gun	Màiri	fhaicinn.
	decide.past	1sg	without	Mary	see.vn
	‘I decided not to see Mary.’

Also, in addition to lacking tense distinctions, the verbal noun also does not inherently express voice, as can be seen in the following example where the form of the verbal noun is neutral to the active/passive distinction, although contextual factors typically determine whether an active or passive interpretation is intended:

(37)	Bha	mi	gan	teagasg.
	be.past	1sg	at.3pl	teach.vn
	‘I was teaching them.’

(38)	Bha	iad	gan	teagasg	agam.
	be.past	3pl	at.3pl	teach.vn	at.1sg
	‘They were being taught by me.’ (lit. ‘They were at their teaching at me.’)

Another piece of evidence for the nominal status of the verbal noun could be that the reflexive particle fèin ‘self’ may be added to the verbal noun (41) to derive reflexive pronouns, which means that the verbal noun shows the same property as the pronoun and noun in examples (39 & 40):26

(39)	mi-fhìn, thu fhèin, e fhèin, i fhèin, sinn fhìn, thu fhèin, iad fhèin.
	1sg-self, 2sg self, 3sg.m self, 3sg.f self, 1pl self, 2pl self, 3pl self

(40)	Mo	theaghlach	fhìn
	1sg.poss	family	refl
	‘My own family’

(41)	Tha	iad	air	an	cuideachadh	fhèin.
	be.pres	3pl	perf	3pl.poss	help.vn	refl
	‘They have helped themselves.’

The widespread use of verbal nouns in prepositional phrases is commonly considered to be another important example of nominal syntax. As we can see in the following examples, the verbal noun can function as the object of a preposition, which is very common in Scottish Gaelic:

(42)	Tha	seo	co-cheangailte	ri	co-obrachadh.
	be.pres	dem.prox	related	to	collaborate.vn
	‘This has to do with collaborating.’

(43)	Tha	seo	math	airson	an	sgrùdaidh.
	be.pres	deic.prox	good	for	the.gen	study.vn.gen
	‘This is good for my study.’

(44)	Tha	brosnachadh	cudromach	ann	an	ionnsachadh.
	be.pres	motivation	important	in.3sg.m	in	learn.vn.dat
	‘Motivation is important in learning.’

Another context that highlights a similarity between nouns and verbal nouns is the position of the pronominal objects. Unlike finite verbs, which generally precede their personal pronouns, an object of a verbal noun must appear preverbally as a possessive (46)27, with the resulting structure being similar to the ordinary nominal structures of the language, where a possessive adjective or determiner precedes a noun (45):

(45)	Do	theaghlach.
	2sg.poss	family
	‘Your family.’

(46)	Tha	iad	air	do	chuideachadh.
	be.pres	3pl	perf	2sg.poss	help.vn
	‘They have helped you.’ (lit. ‘They are on your helping.’)

Also, unlike nominal and pronominal complements, prepositional complements generally follow the verbal noun in constructions involving modal verbs—such as feum ‘must’ and faod ‘can, may’, adjectival or nominal predicates, and the expression of a perfect aspect (48), mirroring its behaviour when functioning as a noun (47).

(47)	Ar	sabaid	an	aghaidh	eucoir.
	1pl.poss	fight.vn	in	face	crime.gen
	‘Our fighting against crime.’

(48)	Tha	sinn	air	sabaid	an	aghaidh	eucoir.
	be.pres	1pl	perf	fight.vn	in	face	crime.gen
	‘We have fought against crime.’

Additionally, the appearance of the verbal noun in a periphrastic construction involving the verb dèan ‘do’ appears to suggest that, in these instances, the verbal noun is serving as a noun, and specifically as the direct object of the verb dèan ‘do’, as is the case with expressions in which this verb is combined with a noun to denote an event, such as dèan sgeig ‘mock’, dèan casad ‘cough’, dèan ùrnaigh ‘pray’, dèan prothaid ‘make a profit’, or dèan iomrall ‘make a mistake’, for instance.

(49)	Dèan	cadal	a-nis!
	do.imp	sleep.vn	now
	‘Go to sleep right now!’

(50)	Chan	eil	e	math	sodal	a	dhèanamh
	neg	be.pres.dep	3sg.m	good	flatter.vn	inv	do.vn
	ris.
	to.3sg.m
	‘It is not good to flatter him.’

Finally, another argument for considering verbal nouns as nouns is that the typical Celtic phenomenon of lenition affects verbal nouns in the same way as regular nouns, as is illustrated by the initial mutation effected by first, second, and third person masculine singular possessives in verbal nouns in aspectual constructions28:

(51)	Tha	a	thidsear	air	a	mholadh.
	be.pres	3sg.m.poss	teacher	perf	3sg.m	praise.vn
	‘His teacher has praised him.’

(52)	Tha	a	tidsear	air	a	moladh.
	be.pres	3sg.f.poss	teacher	perf	3sg.f	praise.vn
	‘Her teacher has praised her.’

While this body of data might lead us to the conclusion that verbal nouns are simply nominals, several other contexts challenge such an interpretation and instead appear to support their analysis as verbal forms.

5.2. Verbal Contexts

Verbal nouns may also be considered verbal in nature, insofar as they typically denote events—such as actions, processes, or states, and most of them do not share the same form as deverbal nouns29. Furthermore, like tensed verbs, verbal nouns take arguments that are both unambiguous and obligatory, may assign a structural case to their complements, which may often be realized before the lexical verb, and can be modified by adverbials. They also function as non-finite complements to modals and nominal and adjectival predicates, and function as the only predicate in periphrastic aspectual and passive constructions. Additionally, verbal nouns occur in raising and control predicates, exceptional case marking (ECM) constructions, and causative constructions. Their behaviour in cleft constructions further aligns them more closely with verbs than with nouns, and the clauses in which they appear can often be substituted by finite clauses, which is a feature traditionally associated with the verbal category. Beyond these, verbal nouns have other properties (e.g., the distinction between aspectual particles and prepositions, the presence of verbal nouns as the only predicate in non-finite clauses as well as their occurrence as complements of interrogative phrases) that, while not uniquely verbal, further support their verbal status. The following grammatical contexts are proposed as a diagnostic of the verbal nature of verbal nouns in Scottish Gaelic.30

Apart from the fact that the verbal noun is primarily derived from the same root as the verb itself, traditionally lexical categories have been distinguished based on the type of meaning they denote. Seen from this perspective, verbal nouns appear to resemble verbs more, as they express states, actions, or processes, rather than to nouns, which typically denote people, objects, places, animals, abstract concepts, ideas, and similar entities. This is illustrated by the contrast between the meaning conveyed by a noun (53) and that conveyed by a verbal noun (54), as well as by the fact that a verbal noun can independently express an action in response to a question (55):

(53)	Bha	an	t-òran	aige	glè	bhòidheach.
	be.past	the	song	3sg.m.poss	very	beautiful
	‘His song was very beautiful.’

(54)	Bha	an	t-seinn	aige	glè	bhòidheach.
	be.past	the	sing.vn	3sg.m.poss	very	beautiful
	‘His singing was very beautiful.’

(55)	A.	Tha	Pàdruig	a’	cluich	a’	ghiotàir.
		be.pres	Peter	prog	play.vn	the.gen	guitar.gen
		‘Peter is playing the guitar.’
	B.	Dè	tha	e	a’	dèanamh?
		what	be.pres	3sg.m	prog	do.vn
		‘What is he doing?’
	A.	A’	cluich	a’	ghiotàir.
		prog	play.vn	the.gen	guitar.gen
		‘Playing the guitar.’

Another feature that appears to show the verbal nature of verbal nouns is that, although they may be etymologically related, most verbal nouns differ from deverbal nouns in form31, such as dannsadh (vn) vs. dannsa (n) (<danns (v) ‘dance’), marcachadh (vn) vs. marcach (n) (<marcaich (v) ‘ride’), foilleachadh (vn) vs. foill (n) (<foillich ‘cheat’), gàireachdainn (vn) vs. gàire (n) (<gàir (v) ‘laugh’), feadaireach (vn) vs. fead (n) (<feadairich (v) ‘whistle’), osnachadh (vn) vs. osann (n) (<osnaich (vn) ‘sigh’), faireadh (vn) vs. faire (n) (<fair (v) ‘watch’), or rèiteach (vn) vs. rèite (n) (<rèitich (v) ‘reconcile’), among others. This means that in cases where distinct verbal and nominal forms have developed the characterization of verbal nouns as verbal and deverbal nouns as nominal would seem logical, in a fashion parallel to the distinction made between examples like ‘reconciling’ and ‘reconciliation’ in English32.

An important verbal feature of verbal nouns is that they display an equivalent argument structure to that of finite verbs, such that finite verbs like tuit ‘fall’, ionnsaich ‘learn’, and ceannaich ‘buy’ require one, two, and three arguments, respectively, just like their corresponding verbal noun forms tuiteam, ionnsachadh, and ceannach. As with finite verbs and unlike nouns and adjectives, the arguments of verbal nouns are generally obligatory although they may be omitted when they can be easily retrieved from the context, and there is no ambiguity as to the role of these arguments relative to the verbal noun. Unquestionably, e ‘he’ is the agent and Gàidhlig ‘Gaelic’ is the theme in (56 & 57):

(56)	Dh’ionnsaich	e	Gàidhlig.
	learn.past	3sg.m	Gaelic
	‘He learnt Gaelic.’

(57)	Tha	e	ag	ionnsachadh	Gàidhlig.
	be.pres	3sg.m	prog	learn.vn	Gaelic
	‘He is learning Gaelic.’

The verbal noun should also be treated as a verb because it has selectional restrictions and strict subcategorization features (58), like inflected verbs (59):

(58)	Bha	mi	ag	ithe	feòil	is	glasraich	/!bòrd	/*de
	be.pst	1sg	prog	eat.vn	meat	and	vegetables	/!a table	/*of
	chaise.
	cheese
	‘I was eating meat and vegetables/!a table/*of cheese.’

(59)	Dh’ith	mi	feòil	is	glasraich	/	!bòrd	/* de	chàise.
	eat.pst	1sg	meat	and	vegetables	/	table	/ of	cheese.gen
	‘I ate meat and vegetables/!a table/*of cheese.’

In these two examples, regardless of whether the verb ith ‘eat’ appears in the past tense or as a verbal noun ithe, it requires an NP expressing an inanimate, edible object.

Another verbal characteristic of verbal nouns is reflected in the case assigned to their direct object in certain constructions, particularly those expressing a perfect aspect, involving a modal verb, or containing a non-verbal predicate (61). In such contexts, the nominal object of the verbal noun appears in the common case—paralleling the syntactic behaviour of the direct object of a finite verb (60), rather than in the genitive case, which would typically mark the complement of a noun33:

(60)	Cheannaich	an	duine	sin	càr	ùr.
	buy.past	the	man	dem.dist	car	new
	‘That man bought a new car.’

(61)	Bu	chòir	don	duine	sin	càr	ùr	a
	be.past	obligation	to+the.dat	man.dat	dem.dist	car	new	inv
	cheannach.
	buy.vn
	‘That man should buy a new car.’

Verbal nouns may also appear in structures where the object precedes the verbal noun (63)34, which seems to have no counterpart in nominal structures, as it is impossible for a complement of a noun to assume a prenominal position (62):

(62)	Dealbh	an	eich.
	painting	the.gen	horse.gen
	‘The painting of the horse.’

(63)	Tha	mi	air	an	t-each	a	pheantadh.
	be.pres	1sg	perf	the	horse	inv	paint.vn
	‘I have painted the horse.’

Another typically verbal feature shown by verbal nouns appears in some constructions where adverbial, rather than adjectival, modification is available. In this context, the verbal noun is modified by an adverb35, instead of an adjective, as would be expected if it had a nominal status, and the adverb tends to appear rightmost in the clause:

(64)	Tha	i	a’	dràibheadh	gu	faiceallach.
	be.pres	3sg.f	prog	drive.vn	adv	careful
	‘She is driving carefully.’

(65)	Tha	an	dotair	a’	ciùradh	a’	bhalaich	gu
	be.pres	the	doctor	prog	cure.vn	the.gen	boy.gen	adv
	ceart.
	correct
	‘The doctor is curing the boy properly.’

Another feature that renders verbal nouns more similar to verbs than to nouns is their occurrence in constructions including modal verbs (66) or non-verbal predicates (67 & 68)36, where many languages employ a non-finite form, which is comparable in function to the infinitive in other European languages:

(66)	Feumaidh	clann	gèilleadh	do	am	pàrantan.
	must.fut	children	obey.vn	to	3pl.poss	parent.pl.dat
	‘Children must obey their parents.’

(67)	Tha	mi	toilichte	sin	a	dhèanamh.
	be.pres	1sg	happy	dem.dist	inv	do.vn
	‘I am happy to do that.’

(68)	Bha	dùil	aca	a	dhol	a	Lunnainn
	be.past	expectation	at.3pl	inv	go.vn	to	London.dat
	as	t-samhradh.
	in	summer.dat
	‘They expected to go to London in summer.’

Celtic languages are rich in periphrastic verbal constructions, most of which involve the use of verbal nouns. They are conventionally regarded as taking a form consisting of the auxiliary verb bi ‘be’, which is marked for person, number, and tense, an aspect marker or a preposition37, and a verbal noun. Assuming that aspect is a verbal rather than a nominal property, the appearance of the verbal noun in a number of aspectual constructions containing the substantive verb bi ‘be’—progressive (69), perfect (70), predictive (71), prospective (72), and stative (73)—lends further support to its categorial status as a verb.

(69)	Tha	iad	a’	cluich	ball-coise.
	be.pres	3pl	prog	play.vn	football
	‘They are playing football.’

(70)	Tha	iad	air	ball-coise	a	chluich.
	be.pres	3pl	perf	football	inv	play.vn
	‘They have played football.’

(71)	Tha	iad	a’	dol	a	chluich	ball-coise.
	be.pres	3pl	prog	go.vn	purp	play.vn	football
	‘They are going to play football.’

(72)	Tha	iad	gu	bhith	a’	cluich	ball-coise.
	be.pres	3pl	to	be.vn	prog	play.vn	football
	‘They are about to play football.’

(73)	Tha	sinn	nar	suidhe.
	be.pres	1pl	in.1pl.poss	sit.vn
	‘We are sitting.’

Another characteristic of verbal nouns suggestive of a verbal more than a nominal nature is their frequent occurrence in constructions involving an auxiliary verb. Thus, in addition to the aspectual constructions illustrated above, a verbal noun may also occur in certain passive-like constructions, where it is preceded by the auxiliary verbs rach ‘go’ (75) or bi ‘be’ (76). The active counterpart is provided in (74) for purposes of comparison:

(74)	Thug	sinn	tòrr	shiùcairean	dhan	a	chloinn.
	give.past	1pl	a.lot.of	sweet.pl.gen	to.3sg.f	the.dat	children.dat
	‘We gave a lot of sweets to the children.’

(75)	Chaidh	tòrr	siùcairean	a	thoirt	dha	na
	go.past	a.lot.of	sweet.pl	inv	give.vn	to.3sg.f	the.dat
	cloinn.
	children.dat
	‘A lot of sweets were given to the children.’

(76)	a.	Bha	tòrr	shiùcairean	air	an		toirt		dhan
		be.past	a.lot.of	sweet.pl.gen	perf	3pl	give.vn			to.3sg.f
		a	chloinn.
		the.dat	children.dat
		‘A lot of sweets were given to the children.’

(77)	b.	Bha	tòrr	shiùcairean	gan	toirt	dhan
		be.past	a.lot.of	sweet.pl.gen	at.3pl	give.vn	to.3sg.f
		a	chloinn.
		the.dat	children.dat
		‘A lot of sweets were given to the children.’

There is further evidence of the verbal status of verbal nouns in raising and control constructions, as we must assume that an understood argument of the predicate represented by the verbal noun in the subordinate clause either appears in the main clause in the case of raising or is determined by an argument of the predicate of the main clause in the case of control. Thus, on the one hand, an co-dhùnadh seo ‘this decision’ is an argument of the predicate dùisg ‘arouse’ but appears in the main clause in (77) and, on the other hand, a first person singular argument is shared by both predicates iarr ‘want’ and dùin ‘close’ but it is explicitly realized only in the main clause in (78). In view of this, the complex of which the verbal noun forms part in these constructions must be considered clearly verbal:

(77)	Tha	an	co-dhùnadh	seo	buailteach	air	fearg
	be.pres	the	decision	dem.prox	likely	on	anger
	a	dhùsgadh.
	inv	arouse.vn
	‘This decision is likely to arouse anger.’

(78)	Dh’iarr	e	orm	an	uinneag	a	dhùnadh.
	ask.past	3sg.m.s	on.1sg	the	window	inv	close.vn
	‘He asked me to close the window.’

Although it is not so clearly verbal, further evidence of the verbal nature of the verbal noun can be found in grammatical contexts similar to exceptional case marking (ECM) constructions, where it appears to function in the same way as the present participle in other languages, such as English:38

(79)	Chan	fhaca	mi	thu	a’	coiseachd.
	neg	see.past.dep	1sg.s	2sg	prog	walk.vn
	‘I didn’t see you hiking.’

Something similar occurs in causative constructions, where the verbal noun resembles the English bare infinitive:

(80)	Thug	mi	air	ruith	gu	luath.
	give.past	1sg	on.3sg.m	run.vn	adv	fast
	‘I made him run fast.’

Further differences between the behaviour of nouns and verbal nouns can be observed in cleft constructions. Firstly, unlike a finite verb, the verbal noun can only be regularly fronted in emphatic sentences because Scottish Gaelic follows the Insular Celtic VSO word order, which is characterized by commencing an unmarked simple sentence (basic word order) with a finite verb. Besides this, the genitive object of a non-finite verb may be fronted in a cleft construction (81), a possibility that is not available to genitive complements within complex noun phrases (82). Finally, it is noteworthy that, of the two strategies employed by Scottish Gaelic to form cleft constructions, the emphasis of nominal phrases, as well as of nominal clauses, is realized through the first strategy, which uses a postcopular clefted pronoun e ‘it’ (83), whereas verbal nouns make use of the second strategy, which requires the presence of the non-inflecting preposition ann ‘in’, a pattern also observed when adjectival, adverbial, and prepositional phrases are clefted (84):

(81)	Is	e	an	taighe	sin	a	tha	i
	be.pres	3sg.m	the.gen	house.gen	dem.dist	rel	be.pres	3sg.f
	a’	ceannach.
	prog	buy.vn
	‘It is that house that she is buying.’

(82)	* Is	e	an	taighe	sin	a	tha	ùidh
	be.pres	3sg.m	the.gen	house.gen	deic.dist	rel	be.pres	interest
	agam	ann	an	ceannachd.
	at.1sg	in.3sg.m	in	buy.dat
	‘It is of that house that she is thinking about the purchase.’39

(83)	Is	e	cèic	a	rinn	mo	mhàthair.
	be.pres	3sg.m	cake	rel	do.past	1sg.poss	mother
	‘My mother made a cake.’ (lit. ‘It is a cake that my mother made.’)

(84)	’S	ann	a’	dèanamh	cèic	a	rinn	mo	mhàthair.
	be.pres	foc	prog	do.vn	cake	rel	do.past	1sg.poss	mother
	‘My mother made a cake.’ (lit. ‘It is making a cake that my mother did.’)

Further evidence supporting the assumption that verbal nouns are verbs rather than nouns is provided by the fact that non-finite clauses in which verbal nouns occur (85) can be transformed into finite clauses (86 & 87) without altering the meaning of the construction:

(85)	Bu	mhath	leam	Mìcheal	a	bhuannachadh.
	be.past	good	with.1sg	Michael	inv	win.vn
	‘I would like Michael to win.’40

(86)	Bhiodh	e	math	nam	biodh	Mìcheal	a’
	be.cond	3sg.m	good	if	be.cond.dep	Michael	prog
	buannachadh.
	win.vn
	‘It would be good that Michael wins.’

(87)	Bhiodh	e	math	nam	b’	urrainn	do	Mhìcheal
	be.cond	3sg.m	good	if	be.past	ability	to	Michael.dat
	buannachadh.
	win.vn
	‘It would be good if Michael could win.’

While this evidence is not so clearly verbal, it is of note that verbal nouns can also occur as the only verb in constructions not involving strictly aspectual uses of prepositions before verbal nouns. A verbal noun may also be the only verb in a tenseless clause, typically introduced by a conjunction like gus ‘in order to’ or gun ‘without’:41

(88)	Gus		a		pàrantan	a	cuideachadh,		chuir
	in.order.to		3sg.f.poss		parent.pl	purp	help.vn		put.past
	i	romhpe		(a)	stad	a	bhith	ag	obair.
	3sg.f	before.3sg.f		inv	stop.vn	purp	be.vn	prog	work.vn
	‘In order to help her parents, she decided to stop working.’

(89)	Chaidh	e	tarsainn	air	an	t-sràid	gun
	go.past	3sg.m	across	on	the.dat	street.dat	without
	choimhead.
	look.vn
	‘He crossed the street without looking.’

As shown in the previous section, the verbal noun usually appears in aspectual constructions after an element resembling a preposition at least formally. The putative preposition is generally formally identical to a particular locative preposition, which would probably indicate that the verbal noun is nominal in character42. However, it would seem more appropriate to analyse these aspectual markers as aspectual particles rather than prepositions for several reasons, proving that the original locative preposition has now grammaticalized as an aspect marker.

One of the reasons for this is their pronunciation, as the aspectual particle and the preposition may not share the same phonological form (90). This is the case with the particle used to express the progressive aspect, namely ag, which has two allomorphs (ag preceding verbal nouns beginning with a vowel (and sometimes ràdh ‘say’) and a’ preceding verbal nouns beginning with a consonant) and is weakened in pronunciation, being consistently realized as/əg/or simply/ə/, respectively, in contrast to the preposition from which it is diachronically derived, aig ‘at’, which only has one allomorph and is pronounced/ɛgj/(92). A second distinction between aspectual particles and prepositions lies in their interaction with possessives: while the former allow contraction with possessives (gam ‘at my’ in (91)), the latter must always appear separately from them (aig mo ‘at my’ in (90)43). Finally, whereas the aspectual particle must be repeated before each verbal noun in coordinated constructions (92), the preposition may, in contrast, be omitted in the same context (93):

(90)	Tha	mi	ag/ǝg/	ionnsachadh	seo	aig/ɛgj/	mo
	be.pres	1sg	prog	learn.past	dem.prox	at	1sg.poss
	sheanmhair.
	mother.dat
	‘I am learning this from my grandmother.’

(91)	Chan	eil	thu	gam	chuideachadh.
	neg	be.pres.dep	2sg	at.1sg	help.vn
	‘You are not helping me.’

(92)	Tha	mi	a’	leughadh	agus	a’	sgrìobhadh	litrichean.
	be.pres	1sg	prog	read.vn	and	prog	write.vn	letter.pl
	‘I am reading and writing letters.’

(93)	Chunnaic	mi	i	aig	a’	bhùth	agus	(aig)	an
	see.past	1sg	3sg.f	at	the.dat	shop.dat	and	at	the.dat
	taigh-dhealbh.
	cinema.dat
	‘I saw her at the shop and the cinema.’

Further support for a verbal analysis of verbal nouns is provided by the observation that, when the emphasis on the pronominal object is required, it may be expressed either through a discontinuous emphatic possessive (e.g., mo^L (+consonant)/m’^L (+vowel)…sa, do^L/d’^L…sa, a^L/Ø…san, a/a h-…se, ar/ar n-/…ne, ur/ur n-…se, an/am…san) (94) or through an emphatic personal pronoun (e.g., mise, thusa, esan, ise, sinne, sibhse, iadsan) (95) ‘1sg-emph, 2sg-emph, 3sg.m-emph, 3sg.f-emph, 1pl-emph, 2pl-emph, 3pl-emph’:

(94)	Theab44	e	mo	mharbhadh	sa	cuideachd.
	def	3sg.m	1sg.poss	kill.vn	emph	too
	‘He almost killed ME too.’

(95)	Theab	e	mise	mharbhadh	cuideachd.
	def	3sg.m	1sg.poss.emph	kill.vn	too
	‘He almost killed ME too.’

Finally, it is also possible to find a verbal noun—càradh ‘repair’—functioning as the sole predicate in a non-finite subordinate clause—a’ càradh an lanntair ‘repairing the lantern’, which is a context where the non-finite verb form is typically interpreted as a converb modifying the main clause verb and acquiring clausal status. That is, the verbal noun performs an adverbial role by expressing relations of time, cause, or manner:

(96)	Bha	am	bodach	na	shuidhe	air	being	a’	càradh
	be.past	the	old.man	in.3sg.m	sit.vn	on	bench	prog	repair.vn
	an	lanntair.
	the.gen	lantern.gen
	‘The old man was sitting on a bench repairing the lantern.’

6. Discussion

An analysis of grammatical contexts involving verbal nouns in Scottish Gaelic shows a complex interplay of nominal and verbal characteristics, thereby justifying the continued debate surrounding their categorical classification. Further evidence would indicate that, while, in certain contexts, the verbal noun is either nominal or verbal, in other contexts this is less clear-cut, owing either to exceptions or because the context reflects properties of both categories. This therefore highlights the absence of a definitive boundary between them.

Thus, certain contexts allow for a straightforward identification of verbal nouns as either nominal or verbal. The clearest nominal contexts are those in which verbal nouns function as arguments of predicates (3–5), show distinctions in case (6–8), gender, and number (9–12), appear with determiners and numerals (13 & 14), require possessive pronouns instead of personal pronouns to express their arguments (26 & 27), and bear the genitive case when functioning as the complements of other nouns (21 & 22). Additional features reinforcing their nominal profile include the position of possessives (46) and PP complements (48), the formation of compound nouns (23 & 24), and their co-occurrence with the light verb dèan ‘do’ in periphrastic constructions (49 & 50).

Similarly, a number of contexts show the verbal nature of verbal nouns. These include their prototypical association with eventive semantics—such as actions, processes, and states (54 & 55), their morphological divergence from deverbal nouns, and their use as the non-finite complements of modal verbs and in impersonal constructions involving nominal or adjectival predicates (66 & 68). Other clear verbal properties are illustrated by the fact that verbal nouns also serve as the sole predicate in periphrastic aspectual constructions (69–73)—despite the lenition effects that possessives have on the verbal noun in this context—and passive constructions (75 & 76), and occur in control, raising, exceptional case marking (ECM), and causative constructions (77–80). Finally, other clearly verbal contexts include their behaviour in cleft constructions (81–84), the substitutability of verbal noun clauses with finite clauses (85–87), and their presence as a converb (96) in adverbial subordinate clauses.

On the other hand, a substantial number of contexts resist such a straightforward classification. For instance, regarding the morphology of the verbal noun, although it is evident that its root is clearly verbal, its inflection is unquestionably nominal. Verbal nouns may allow both adjectival and adverbial modification, which adds ambiguity (15 & 16 vs. 64 & 65). It is also unclear whether the positioning of the complements of verbal nouns can help determine whether these elements are more nominal or verbal in character, considering that their complements may either precede or follow the verbal noun in different constructions (63 vs. 18–20). Additionally, the case assigned to NP complements varies: while the genitive case is still often used in the expression of the progressive, prospective, and purposive aspect (18 & 20)—and this occurs usually when the object is definite, many constructions today favour the use of the basic or common case (63). Also, the co-occurrence of verbal nouns with relative clauses (30) is not frequently attested in contemporary usage. Moreover, the fact that verbal nouns are negated differently from finite verbs (31 & 32) does not necessarily indicate a nominal nature, as similar phenomena are observed cross-linguistically in non-finite clauses; hence, this does not appear to be sufficient grounds for precluding the interpretation of the verbal noun as retaining verbal properties in this context. Likewise, although prepositions typically select for nominal complements (42–44), it is plausible that in these cases the complement is the entire clause containing the verbal noun, rather than the verbal noun itself, which retains a predicative function. Furthermore, the lack of dependent/independent formal distinctions (36), a hallmark of finite verbs in Scottish Gaelic, should not automatically be taken as evidence of nominal status, as such distinctions could well be tied to tense (the imperative form does not show this distinction either). The absence of voice alternations (38) also need not imply nominal status; rather, it may reflect one particular language-specific strategy for expressing the passive voice in Scottish Gaelic, one that diverges significantly from the garden-variety passive construction observed, for example, in English. Additional areas of ambiguity include the use of reflexive pronouns with verbal nouns (41)—a phenomenon also attested with adverbs and adjectives—and the presence of obligatory and unambiguous arguments (28 & 29), a property not exclusive to verbs, as many nouns also generally have arguments with an unambiguous syntactic and semantic function. Finally, the distinction between prepositions and aspectual particles (90–93)—apart from a few well-established cases such as aig vs. ag/a’, ann, and even air in progressive, stative, and perfect constructions, respectively—is not always a reliable way of determining diagnostic purpose.

In summary, it seems clear that the behaviour of verbal nouns in Scottish Gaelic straddles the traditional divide between noun and verb and, rather than slipping neatly into a single category, verbal nouns have a dual nature that depends heavily on the grammatical contexts in which they appear.

While it is true that this type of confusion may be more prevalent in Celtic languages than in other branches of the IE family, due to the fact that Celtic languages possess only a single non-finite form—the verbal noun, leaving the verbal adjective aside, the difficulty in distinguishing between nominal and verbal properties in non-finite forms also occurs across other language families, as with the gerund in English (Quirk et al., 1985, p. 1291; Huddleston, 2002, pp. 1220–1222; Aarts, 2007, pp. 142–144; Malchukov, 2019; Keizer, 2023, pp. 186–187). Indeed, in many languages, the same non-finite form may appear in different syntactic constructions, reflecting a distinctly lexical nature. This is the case with the Italian infinitive, which can function either as a subject or an object (e.g., Fumare fa male ‘Smoking is bad’) and can also serve as the sole lexical predicate in constructions involving modal verbs (e.g., Lui deve studiare oggi ‘He must study today’). It is also possible for two distinct non-finite forms to appear within the same syntactic context, such as the infinitive and the gerund in English when functioning as the subject or object of a sentence (e.g., I started to do/doing).

The root of this problem could be that, at least now, the distinction among non-finite forms is primarily based on their morphological characteristics (e.g., -ar/-er/-ir for infinitives, -ando/-endo for gerunds, -nte for present participles, and -ado/-ido for past participles in Spanish), rather than on the verbal, nominal, adjectival, or adverbial function they currently perform or may have performed in the past. Although it is not possible to trace the original morphology of verbal nouns further back than Proto-Indo-European, the solution to this categorial problem may be found through an analysis of the historical development of non-finite forms and the diachronic developments they have undergone up to the present day, not only within the Celtic languages but also across other languages, and more specifically, within those branches of the IE family most closely related to Celtic.

As has been widely acknowledged (Shields, 1992; Beekes, 1995; Szemerényi, 1996; Clackson, 2007), the reconstruction of the verbal system constitutes the most intricate aspect of PIE linguistics, owing both to its internal complexity and the considerable dialectal variation that distinguishes verbal morphology from other major categories. Notwithstanding this difficulty, there is a degree of scholarly consensus that the infinitives in PIE were formerly verbal nouns or action nominals. Nevertheless, it remains difficult to determine with certainty whether they were originally either nouns incorporated into the verbal system or were formed by attaching nominal morphology to a verbal root (Zeuss, 1871; Windisch, 1878; Delbrück, 1893–1900; Brugmann, 1886–1900/1897–1916; Pedersen, 1909–1913; Vendryes, 1910; Lewis & Pedersen, 1931; Disterheft, 1980, 1981). In addition, this protolanguage also made use of various participles—which may also have originated from a verbal root—that appear to have been adjectives assimilated into the verbal system, such as the present participle, the verbal adjective, which was semantically similar to a past participle, the gerundive, which was used to indicate need, obligation or possibility, and the absolutive (also referred to as a gerund but in a different sense), which had an adverbial function.

New forms gradually emerged from the ancient categories, acquiring non-personal verbal functions. Thus, subsequently, the infinitive constructions found in the classical IE languages appear to descend from fixed case forms of PIE verbal nouns (Buck, 1933; Hahn, 1966; Rosén, 1981; Beekes, 1995; Sihler, 1995). For example, the Classical Latin and Ancient Greek infinitives are believed to have developed from PIE s-stem verbal nouns marked for the locative or dative case. Similarly, the Latin gerund (as well as the supine, which was employed to express purpose or to complement adjectives) is thought to have originated from the same element, although it was used in all nominal cases, excluding the nominative, which remained the domain of the infinitive. Verbal adjectives such as the gerundive, which expressed necessity, obligation, or propriety regarding the execution of an action, and participles, which occurred in the active, middle, and passive voice, are also frequently used. However, unlike the infinitive and the gerund, these forms adopted an adjectival morphology (Whitney, 1924; Sihler, 1995).45

Over time, the extensive inventory of non-finite forms of Latin was reduced to three primary types in the subsequent Romance languages: two hybrid categories, namely the infinitive and the gerund, which combine nominal and verbal features, and one mainly adjectival (although also adverbial and verbal) form, namely the participle, which appears in two different forms: the present and past participle. In most Romance languages, the functions of the gerund and the present participle eventually merged, with one of the forms emerging as the prominent variant—it is mainly used to express the progressive aspect and adverbial subordination—and the other being relegated to highly sporadic uses—as is the case with the evolution of the old present participle in -ns/-nt (in the genitive)—as an adjective in many of these languages.

The Germanic languages developed the same three principal non-finite forms, also fusing the original gerund and present participle forms, thus introducing significant ambiguity in determining the syntactic and semantic function of each (Ringe, 2006). For instance, although the infinitive commonly serves as the subject or object of a verb and as the complement of a preposition or noun, it is also widely used to express various aspectual distinctions, such as the prospective or predictive, and to function as the sole predicate in constructions involving modal verbs and impersonal clauses with adjectival or nominal predicates. A comparable ambiguity arises from the coalescence of the gerund and the present participle, now a single category that may serve a verbal role (as in the progressive aspect), an adjectival role (as a noun modifier), or an adverbial role (functioning as a converb in adverbial subordinate clauses) in many modern Germanic languages. The past participle presents a similar challenge, as it may act as a verb (in the perfect aspect or in passive constructions) or as an adjective (when modifying nouns).

Finally, the Celtic languages appear to have undergone fewer changes than other IE branches with regard to the evolution of non-finite verb forms, as they have retained the categories of the verbal noun and the verbal adjective46.

In light of this diachronic explanation, it is reasonable to assume that the reduction and convergence of non-finite forms appears to be one reason to explain the confusion surrounding the identification of their lexical category. This issue is, of course, even more pronounced in the Celtic languages, where a single non-finite form like the verbal noun can serve a nominal, verbal, and even adjectival or adverbial function.

Another reason, which may account for the hybrid character of certain non-finite forms, is the presumed fuzzy nature of categories. If more information were available regarding the origin of verbal nouns prior to PIE, two different hypotheses could be reasonably proposed. Firstly, if verbal nouns derived from verbs, they might still be reinterpreted and reanalysed as nouns when their function is more closely aligned with a reference to actions or states, rather than with predication, that is the description of the event in which the participants are involved. Secondly, by contrast, if verbal nouns were originally nominal in nature, they could also be reinterpreted as verbs when their function is more closely associated with the depiction of specific events rather than with an explicit reference to the participants engaged in those events. In either case, this issue appears to invite a broader reflection on whether it is the verbal noun itself that should be regarded as a mixed or hybrid category, or whether such hybridity is more appropriately attributed to the fuzziness displayed by the categories of verb and noun when considered from a more diachronic perspective and even from a more conceptual standpoint.

With this in mind, the issue of categorization may not be fully resolved, but it can at least be simplified if, like Croft (1990, 2001, 2007), we consider that the two core pragmatic functions, reference and predication, are only prototypically aligned with the two major semantic classes, object and action—or, more accurately entity and event, and, consequently, with the two primary lexical categories, the noun and the verb47. Thus, while the relationship between entity and reference, on the one hand, and between event and predication, on the other, seems to be grounded in basic cognitive processes, as objects typically serve as referents and actions or states are primarily conveyed through predicates in such a way that both reference and predication underpin propositional content, allowing utterances to express states, actions, or processes as well as the participants involved in them, it is entirely possible for an entity like ‘beer’ not to be referential (e.g., ‘He drinks beer’) and for an event like ‘smoke’ to be non-predicating (e.g., ‘To smoke is bad’). Even more commonly, each of the two lexical categories, noun and verb, may be capable of fulfilling either of the two pragmatic functions. For instance, a noun such as ‘dog’ can function referentially (e.g., ‘My dog is good’) or predicatively (e.g., ‘It is a dog’), and a verb such as ‘read’ can serve a predicative function (e.g., I read a book) or a referential function (e.g., ‘Reading is funny’).

This perspective may suggest that, at an early stage of language development48, words emerged, independently of any fixed lexical category49, as symbols—their sounds or letters signify only by cultural convention since the link between the sign and the signified could have been arbitrary—standing for both the entities and the events in which they are involved. Their subsequent classification into specific lexical categories would thus depend on the specific pragmatic function being expressed, which in turn necessitated a corresponding syntactic construction. In accordance with Haspelmath (2007)’s scepticism about the existence of pre-established grammatical categories and their purported semantic basis (Haspelmath, 2023, p. 35), these source categories may then have originated not as fixed lexical classes, but rather as conceptual categories and, while in contemporary grammar each of these lexical categories tends to be associated with particular semantic properties and core pragmatic functions—namely noun with entity and reference and verb with event and predication, these correlations remain clearly distinct only in the most prototypical cases, which obviously do not include non-finite forms. This situation therefore appears to suggest that, in practice, the boundary between the categories of noun and verb is often blurred in non-finite forms, suggesting that the most defining feature of their interrelationship is their gradient nature and that the distinction is even less evident in languages with a more limited number of non-finite forms, as is the case with the Celtic languages. Thus, the gradience observed in Celtic verbal nouns may in fact be an epiphenomenon, reflecting a blurred boundary between two lexical categories, noun and verb, due to the difficulty in distinguishing between the functions of reference and predication in certain syntactic contexts. This gradience has been inherited, in a particularly salient way, by the hybrid category known as the verbal noun, and the fact that it typically appears in a single, morphologically undifferentiated form in Celtic languages further complicates the identification of its lexical status.

In light of the lack of early historical evidence, the only viable approach to understanding its lexical status and functional behaviour lies in examining and analysing it synchronically within the specific grammatical contexts in which it occurs. Regardless of whether their origin was verbal or nominal, it is evident that verbal nouns in the Celtic languages are employed across a significantly broader range of syntactic contexts than might be expected for either prototypical nouns or verbs, as they occur in environments where PIE and late IE would have used a genuine verbal noun, various types of participles, the absolutive or gerund, and the gerundive, each of which has different properties. Accordingly, the position of the verbal noun within the nominal-verbal continuum in the Celtic languages can be understood to vary across time and constructions50. This assumption that this category may be construction-specific and even language-specific is therefore consistent with Cristofaro’s (2009) hypothesis that accounts for the cross-constructional and cross-linguistic diversity of individual categories and relations and is also aligned with a constructional view of categorization (Croft (2001); Goldberg (2006); among others). Accordingly, categories may be epiphenomenal, insofar as they depend on constructions, which are considered to be primitives, and word classes are defined by their distributional properties in different constructions.

On the one hand, it is important to bear in mind that that the properties of many contexts are currently in a state of flux, for example the grammaticalization of some prepositions as particles, the shift from possessives to personal pronouns, the case assigned to some NP objects, and the positioning of other NP objects, etc. This would suggest that the position occupied by these contexts within the continuum between clearly nominal and clearly verbal domains may not be fixed. Consequently, this could entail a reanalysis of the properties of the verbal noun in the future. In this light, the analysis of contexts involving verbal nouns in Scottish Gaelic may be better understood as part of an ongoing process of reanalysis, rather than as evidence of a completed categorial classification. Hence, the evolution of these constructions and the future categorization of verbal nouns remain subject to diachronic empirical investigation.

On the other hand, in the case of Modern Scottish Gaelic particularly, if these contexts were to be arranged along a continuum, contexts would be clearly nominal at one extreme and clearly verbal at the other. The former would include those contexts when the verbal noun functions as a subject or object, when it exhibits gender, number, and case distinctions, when it is modified by determiners and numerals, when it requires possessive pronouns instead of personal pronouns to express the arguments, and when it bears the genitive case as a complement to another noun. The latter would encompass contexts when the verbal noun unquestionably expresses an event, when it is different from a deverbal noun, and when it occurs in aspectual constructions or as a complement of modal verbs and adjectival or nominal predicates, etc. Accordingly, a differentiation between verbal nouns in their nominal and verbal uses in these contexts would seem logical.

However, it is arguably within the intermediate zone that the intersective gradience of the verbal noun becomes most evident. It is precisely in this middle ground that the analytical challenges are most pronounced, as this area involves hybrid contexts that oscillate between nominal and verbal syntactic behaviour. As a result, the verbal noun emerges as a flexible part of speech in such liminal cases and resists a clear-cut classification, as neither its morphological features nor its syntactic distribution provide definitive clues to its function or categorial status. Thus, this case of the convergence of nominal and verbal properties in these instances renders the verbal noun a genuinely hybrid category, as it includes contexts in which it is not possible to ascertain whether it functions as a noun or a verb. In summary, the current difficulty in determining whether the category traditionally referred to as the verbal noun in Celtic languages and predominantly exhibiting nominal or verbal properties may be attributed to the primordial categorial indeterminacy of words and, subsequently, to the inherent gradience of the categories from which it historically derives, primarily the noun and the verb, as well as the convergence of a wide range of both nominal and verbal functions into a single non-finite form. This illustrates the fact that verbal nouns are a graded, rather than a discrete, category whose categorical identification can only be achieved through the analysis of each specific grammatical context in which they appear, taking into account the fact that the properties of the construction may vary diachronically and across languages.

7. Conclusions

Non-finite constructions in Scottish Gaelic are intriguing because they use a specific non-finite verbal form traditionally referred to in Celtic linguistics as the verbal noun, which, as the term implies, has both verbal and nominal properties. Accordingly, the categorial status of verbal nouns in the Celtic languages—specifically, whether they are fundamentally nominal or verbal—has long been the subject of extensive debate within the field of Celtic linguistics.

The present study has sought to provide an analysis of the properties of the verbal noun in Modern Scottish Gaelic, adopting a functional-typological and multidimensional perspective. This approach emphasizes a holistic examination of this distinctive feature of Celtic languages, addressing the syntactic, semantic, and pragmatic dimensions of the constructions in which this hybrid category appears. It cannot be denied that the analysis of the contexts in which the verbal noun currently appears in Scottish Gaelic offered in this study reveals a rather ambiguous situation, insofar as it does not allow us to identify the lexical nature of this category in the language clearly, except in a few cases. This implies that, as with verbal nouns in other Celtic languages, it remains impossible to determine whether the verbal noun is nominal or verbal in nature. However, this should not be seen as a disappointment, nor should it be regarded as an impediment to gaining a deeper understanding of the language. Rather, it may be interpreted as a confirmation of an evident reality. Thus, it makes little sense to attempt to assign exclusively nominal or verbal properties to this category, as it has inherited its versatile character from the categorial indeterminacy of word classes and, subsequently, from the fuzzy nature of verbs and nouns, so its categorial properties depend on the specific grammatical context in which it occurs. Additionally, these properties may also be subject to diachronic variation and may differ across languages.

With the historical information on the Celtic languages available to us, and even their most distant ancestor, PIE, we can only hypothesize as to whether the verbal noun originally belonged to the nominal or the verbal system. Subsequently, and as with other non-finite verb forms across languages, it gradually came to be used in a wide range of constructions in which it could acquire properties typically associated with verbs, nouns, and even adjectives and adverbs. The issue, however, is that, while it is generally straightforward to correlate many types of non-finite forms with specific lexical categories, the participle is often equated with a verbal adjective and the converb with a verb of adverbial nature, for example, forms such as the infinitive and the gerund resist such correlation, as they frequently have both nominal and verbal uses. This is especially controversial in the Celtic languages, where a single non-finite form, namely the verbal noun, is used to cover a range of functions typically expressed through the infinitive and gerund, and even the present participle—as in the case of the progressive aspect or when it is used as a modifier of another noun or a main verb, and the past participle—as in the passive construction and perfect aspect—in other languages.

This complex behaviour shows that identifying the verbal or nominal character of this category is essentially unnecessary, as its status appears to depend on the construction in which it occurs. Rather, this study highlights the need for a nuanced, construction-based approach when examining the use of verbal nouns in Scottish Gaelic. Consequently, an analysis of their syntactic distribution might be more reasonable, since what may constitute nominal or verbal is usage rather than the verbal noun itself. An explicit and systematic distinction between the constructions in which the verbal noun participates and a description of its function in each of them are therefore necessary.

As a result of the analysis of the distribution of verbal nouns in Modern Scottish Gaelic offered in this study, it is evident that the verbal noun has gradually and unquestionably become a highly versatile syntactic constituent since it can now behave either like a noun or a verb in different grammatical environments. However, it has also become evident that a distinction between verbal nouns used either nominally or verbally can be drawn only in a limited number of contexts. The main challenge therefore lies in the fact that not all of the contexts in which the verbal noun appears allow for a clear identification of whether the verbal noun functions as a noun or as a verb. It is precisely in these contexts that the intersective gradience between these two categories—that is, the intersectionality between the properties of the nominal and verbal categories—becomes most apparent. The two categories of noun and verb have traditionally served to represent syntactically the correlation between the semantic classes of entity and event, and the pragmatic functions of reference and predication. However, due to the evident intersectionality between reference and predication in some grammatical contexts and to the convergence of non-finite forms into a single form in Scottish Gaelic, it is only possible to ascertain whether a verbal noun functions as a noun (converging on the class of nouns) or as a verb (converging on the class of verbs) in contexts where the categories of noun and verb clearly correlate with their corresponding semantic class and pragmatic function, but, apart from these prototypical instances, the inherent hybridity of the verbal noun becomes apparent in the other syntactic environments. This provides compelling evidence that, given that a clear-cut distinction between nominal and verbal characteristics is problematical, we should analyse constructions including verbal nouns as a continuum from clearly nominal contexts to clearly verbal contexts with intermediate contexts where the twofold character of the verbal noun is evident and, consequently reflects an instance of intersective gradience. In summary, it is in these intermediate contexts that this hybrid non-finite form reflects the intersection between the two word classes of noun and verb, making it possible to confirm that categories are inherently gradient in nature and that gradience is a pervasive phenomenon.