Competition in the Complementation of Old English Control Verbs with Oblique Marking: A Corpus Analysis

Abstract

The aim of this article is to explain the syntactic competition found in the complementation of Old English Prevent verbs. The competition on argumenthood involves linked verbal predications and linked nominal predications. Evidence is gathered for continuity both between finite and non-finite linked verbal predications as well as between non-finite and nominalised linked predications. This evidence points to a diachronic development: finite clause > non-finite clause > nominalisation. The main conclusion of the article is that the Interclausal Relation Hierarchy predicts the replacement of the finite clause complementation with non-finite clause complementation in such a way that the syntactically tighter noun phrase involving a deverbal nominalisation constitutes the next step of syntactic development.

1. Aims and Scope

This article deals with Old English syntax and semantics and, to be more precise, with competition in the complementation of the verb. Competition can be defined as an alternation on the synchronic axis that may have consequences on the diachronic axis. Competition results from structural redundancy, which arises when two units serve the same function. In synchronic analysis, competition involves an asymmetry between two functionally redundant structural units that show different degrees of morphological markedness or syntactic complexity. Throughout diachronic evolution, one of the structures in competition is likely to replace the other, which may decay or get lost. This is the case with the structural units competing on argumenthood that this study considers: a linked finite clause vs. a linked non-finite clause, on the one hand, and a linked verbal predication vs. a linked deverbal nominalisation, on the other.

Of all the areas that could be relevant for the analysis of syntactic competition in Old English, this article focuses on the verbs from the Prevent class. A verbal class (Levin 1993) is a set of verbs that share meaning components and grammatical behaviour. With Old English Prevent verbs, linked finite clauses, linked non-finite clauses, and nominalisations enter competition on verbal complementation in such a way that we come across instances comparable to *They prevented me that I help you, *They prevented me to help you and *They prevented my help of you. Prevent verbs are object control verbs (Sag and Pollard 1991, p. 65) with oblique marking. Linked finite clauses are found with this function in Old English, but not in Present-Day English, thus [LawCn 1020 5] & þæt hæbbe [ic] mid Godes fultume forene forfangen, þæt eow næfre heononforð þanon nan unfrið to ne cymð, þa hwile þe ge me rihtlice healdað & min lif byð ‘And with the help of God, I have taken measures to prevent hostility ever from this time forth coming upon you from that quarter, as long as you support me loyaly and my life lasts.’

For Sag and Pollard (1991, p. 66), the semantics of object control verbs consist of states of affairs of influence, which, in the case of Prevent verbs, depict an action that the influenced participant is influenced to not perform. Take, as illustration, an instance like They prevented me from helping you (Rohdenburg 1995, p. 86). The decline of the formally distinctive subjunctive, which must be considered against the wider setting of the simplification of inflections and the loss of verb endings, led to the demise of that-clauses of the type *They prevented that I should help you from the later Middle English period onwards (Iyeiri 2010, p. 198). According to Rohdenburg (1995, 2006), the competition between finite clause complementation (as in *They prevented that I should help you) and oblique nominalisation complementation (as in They prevented me from helping you) continued until around 1800. Even in Present-Day English, there is diatopic variation between They prevented me from helping you and They prevented me helping you, although the latter is far less frequent, particularly in American English (Kaunisto and Rudanko 2019, p. 112).

Against this background, the aim of this article is to explain the syntactic competition that arises in the complementation of Old English Prevent verbs. Previous research has considered the competition between finite and non-finite clauses as complements of control verbs but has ignored the role played in this context by noun phrases based on a deverbal nominalisation. The literature has found reasons why complementation patterns like *They prevented me that I help you are not possible in Present-Day English but has not explained why others like *They prevented me to help you cannot be found either; or what They prevented me from helping you has to do with other patterns such as *They prevented my help of you. The focus has been on the Middle English and the Early Modern English periods because the rise of oblique nominal phrases with verbal nouns is a late phenomenon of which no direct evidence can be found in Old English. Nevertheless, the synchronic explanation for the competition on argumenthood with Prevent verbs in Old English adds a new perspective to the study of the form and function of English complementation. This article intends to open a new avenue of research in the origin of the English gerund, with respect to which the role of morphological nominalisations with the suffix -ung/-ing has been considered, but the part of the syntactic nominalisations discussed in this article has not been taken into account. The methodology of the research can be described as a corpus analysis of Old English, comprising three steps: the selection of the meaning components of the verbal class under scrutiny, the compilation of the corpus of analysis, and the discussion of the linking between semantics and syntax.

The article is organised as follows. Section 2 reviews previous studies in the competition found in verbal complementation. Section 3 presents the relevant aspects of Role and Reference Grammar, the linguistic theory on which the analysis of competition carried out in this article is based. Section 4 describes the methodological steps and the sources and data of the analysis. Section 5 is devoted to competition with Prevent verbs. Section 6 discusses the limits of the method based on verbal classes and the perspectives on diachronic evolution offered by the synchronic analysis of Old English. The main conclusions of the article are summarized in Section 7.

2. Infinitives and Clauses in Competition

According to Molencki (1991, p. 91), complement clauses with a non-finite verbal form (infinitive or participle) are less frequent in Old English than in Present-Day English because Old English had not yet developed the system of perfective and progressive passive infinitives and participles. For this author (Molencki 1991, p. 129), the most outstanding difference between the complementation of Old English and Present-Day English is the fact that finite clauses were used where, in Present-Day English, the infinitive, the gerund, or the participle are required.

Callaway (1913) addresses the question of competition in verbal complementation from the perspective of the variation between the uninflected infinitive and the inflected one (as in to leornian). Callaway (1913, p. 266) draws the conclusion that the inflected infinitive is found as the subject more frequently than the uninflected infinitive.

Denison (1993, p. 172) deals with infinitive constructions (VOSI, Verb+Object/Subject+Infinitive) and finite clause alternatives. The finite clause alternatives to infinitive complementation (Denison 1993, p. 179) include the V+NP+finite clause, as in ÆCHom I.1.16.3 and het ða eorðan þæt heo sceolde forðlædan cuce ‘and ordered the earth to bring forth live animals’; V+finite clause in Or 140.11 he forbead ofer ealne his onwald þæt mon nanum cristenum men be abulge ‘he forbade throughout his whole dominion that anyone should offend a Christian man’; and finite clause coordinated with VOSI, as in Or 59.14 siþþan gelicade eallum folcum þæt hie Romanum underþieded wære, & hiora æ to behealdanne ‘then all the peoples were content to be subjected to the Romans and to observe their law’.

For Los (2005), Old English verbs complemented by infinitives can belong to three types: AcI (accusativus cum infinitivo) verbs, monotransitive subject control verbs, and ditransitive object control verbs. In AcI verb constructions, the subject of the matrix clause and the subject of the infinitive clause are different. AcI verbs are verbs of perception and causation that select the bare infinitive (Ringe and Taylor 2014, p. 484). In monotransitive subject control verb constructions, the subject is shared by the matrix and the infinitive clause. Monotransitive subject control verbs are verbs of intention, aspectual verbs, and pre-modal verbs. Whereas the pre-modals are always followed by the bare infinitive, the others can also be complemented by a to-infinitive (Ringe and Taylor 2014, p. 486). In ditransitive object control verb constructions, the object of the matrix clause is shared with the subject of the infinitive clause. Ditransitive object control verbs include the classes of commanding, permitting, persuading, and enticing and usually take an inflected infinitive (Ringe and Taylor 2014, p. 489). According to Los (2005) and Ringe and Taylor (2014), competition mainly holds between the þæt-clause with the subjunctive and the infinitive in instances such as Lk(WSCp)14.23 Ga geond ðas wegas and hegas and nyd hig ðæt hig gan in ‘go along the roads and hedges and urge them that they go in’ vs. ÆHom II 376 Ga Geond wegas and hegas, and hyd hi inn to farenne ‘go along the roads and hedges and urge them to come in’ (Los 2005, p. 68).

3. Competition in Role and Reference Grammar

Role and Reference Grammar, hereafter RRG (Foley and Van Valin 1984; Van Valin and LaPolla 1997; Van Valin 2005), is a typological theory of language that is focused on clausal relations. Overall, the emphasis of RRG on clausal relations (thus the name of the theory) and typological adequacy results in a semantic and pragmatic approach to areas of grammar that other theories consider from a strictly syntactic point of view. In RRG, the different roles played by verbal arguments are explained on the basis of hierarchies that rank the different candidates for a function. This includes relations central to the theory like macroroles, privileged syntactic arguments, and juncture–nexus types. In the remainder of this section, these questions are presented in the wider context of the projection–realisation apparatus of the theory, called linking. Linking is the correspondence between syntax and semantics, which operates in both ways: from semantics to syntax (production) and from syntax to semantics (comprehension). The direction of linking followed in this article is semantics–syntax. This section draws on Van Valin and LaPolla (1997) and on the overview of RRG available from: http://linguistics.buffalo.edu/people/faculty/vanvalin/rrg/RRG_overview.pdf (accessed on 16 March 2023).

3.1. Macroroles

In RRG, the semantic interpretation of verbal arguments is based on two generalised semantic roles or macroroles called the Actor and Undergoer. The assignment of a macrorole requires the previous projection of the lexical representation of a verbal predicate onto a logical structure. This takes two steps, namely the assignment of Aktionsart type and the unfolding of a logical structure.

The typology of Aktionsart adopted in RRG consists of four classes: State, Activity, Achievement, and Accomplishment. States and activities constitute the basic types whereas achievements are punctual changes of state and accomplishments are durative changes of state. Van Valin and LaPolla (1997) and Van Valin (2014) also distinguish the class of active accomplishments (the telic use of activity verbs) and the causative version of all Aktionsart classes. Van Valin (2005, p. 32) also proposes the class of semelfactives, which represent punctual events (both non-causative and causative). Aktionsart types are defined in terms of the set of features shown in Figure 1. An illustration of each type is also provided.

Logical structures relate clausal semantics to clausal syntax. Logical structures constitute an essential component of the semantics–syntax linking because they relate unrealized predicates (represented by means of lexical items and Aktionsart features) to realized predications (represented by projections of arguments and operators at core, clause, and sentence levels). Figure 2 presents Aktionsart types and the corresponding logical structures. The main distinction holds between the stative (predicate’) and non-stative (do’) part of logical structures. The variables x, y, and z stand for verbal arguments. The abstract semantic predicates INGR(essive), SEM(e)L(factive), BECOME, and CAUSE indicate ingressives, semelfactives, accomplishments, and causatives, respectively.

Macroroles (Actor and Undergoer) make grammatical generalisations across argument types and structures. In a transitive predication, the Actor is the first argument and the Undergoer the second argument of the verb. In an intransitive predication, the only argument can get an Actor or Undergoer depending on the semantic properties of the verb. For instance, the first argument of Jill is ill is an Undergoer because be is a stative verb whereas the first argument of Sam is walking is an Actor because walk is an active verb. The relation between verbal arguments and macroroles constitutes an area of competition. This relation abides by the Actor–Undergoer Hierarchy, which stipulates that the leftmost argument in the hierarchy in Figure 3 will be the Actor and the rightmost argument in the hierarchy will be the Undergoer. The hierarchy is governed by the principle of markedness and is, ultimately, asymmetrical. The arrows in Figure 3 indicate an increasing markedness of the realisation of an argument as a macrorole in such a way that the leftmost argument in a logical structure is always the Actor whereas the rightmost argument is only the default choice for the Undergoer (Van Valin 2005, p. 58). For example, the first argument of an activity like ‘jump’ performs the thematic role Mover and gets the macrorole Actor while the argument of a state like ‘see’ plays the thematic role of Patient and receives the macrorole Undergoer.

In RRG, there is no third macrorole available for ditransitives like someone giving something to someone else. The third argument is called the non-macrorole direct core argument. For instance, in They gave the first year students a warm welcome, the Effector they is the first argument and gets the macrorole Actor, the Theme a warm welcome receives the macrorole Undergoer, and the Patient, i.e., the first year students, is a non-macrorole argument.

3.2. Privileged Syntactic Argument

Turning to grammatical relations, a subject and object are not universal for RRG. Instead, RRG proposes the universal notion of Privileged Syntactic Argument (PSA). The PSA is a construction-specific relation that results from a restricted neutralisation of semantic roles and pragmatic functions for syntactic purposes. As happens in macrorole assignment, the other arguments in a clause are either direct or oblique core arguments. Two types of PSA can be distinguished interlinguistically depending on their function. The Pivot is the argument around which the complexity of the construction revolves. The Pivot, for instance, is the missing argument in a construction like These books are for you to read or the shared argument in a construction like I want you to leave. Controllers guarantee agreement, thus giving the interpretation for pivots. For example, in a coordinate subject construction, such as The computer shut down and went to sleep, the explicit traditional subject is the Controller whereas the implicit traditional subject is the Pivot. The selection of the PSA is governed by the hierarchy in Figure 4, which is based on argumenthood in logical structures.

In accusative constructions, the PSA is the highest-ranking direct core argument in terms of the hierarchy in Figure 4. In ergative constructions, the PSA is the lowest-ranking direct core argument in terms of this hierarchy. In some languages, macrorole arguments only can be PSAs whereas in others, such as Old English, non-macrorole direct core arguments can be PSAs. In languages like English, as well as in Old English, the controller of finite verb agreement is the highest-ranking core macrorole argument in terms of the hierarchy in Figure 4. Case assignment rules for direct core arguments in accusative languages like Old English are the following. The highest-ranking core macrorole (in terms of the PSA Selection Hierarchy in Figure 4) takes the nominative case. The other core macrorole takes the accusative case. Non-macrorole direct core arguments take the dative case. These rules exclude morphological case governed by preposition.

3.3. Juncture–Nexus Types

The RRG theory of complex sentences is based on two concepts, the juncture and nexus, in such a way that the type of unit (juncture) is independent of the type of relation (nexus).

The discussion of juncture types calls for an explanation of the structure of the clause in RRG. The Layered Structure of the Clause (LSC) is a hierarchical structure that consists of several semantic layers that are motivated by the scope of operators (grammatical features such as tense, aspect, modality, evidentiality, etc.). The components of the LSC are the Core, the Clause, and the Sentence. The Core comprises a verbal Nucleus and its arguments and its argument-adjuncts, as in drink soda and go to the countryside, respectively. The Clause is comprised of the Core and the Periphery, as in drink soda in the park. The Sentence consists of one or more units of the Clause level, as in I watch TV before going to bed. The arguments in a verbal core, as well as the governed elements in argument-adjuncts and peripheries, are noun phrases. RRG distinguishes several types of complex noun phrases, including noun phrases modified by relative clauses and nominalisations. Deverbal nominalisations are noun phrases headed by a derived noun that is morphologically related to a verb through a productive process of word formation. For RRG, nominalisations are derived from a unit of the clausal level with the verb on which the nominalisation is based. For instance, The arrest of John by FBI agents in New York City has the clausal correlate FBI agents arrested John in New York City in such a way that the noun arrest is converted (or zero-derived) from the verb to arrest. The existence of nominal correlates of elements of the clausal level is further demonstrated by the fact that the modifiers of the noun arrest correspond to the arguments and periphery of the clause: of John < John and in New York City < in New York City (Van Valin and LaPolla 1997, p. 186).

In the construction of complex sentences, the unmarked option is the combination of nuclei, cores with cores, clauses with clauses, and sentences with sentences. The term juncture makes reference to the types of units that enter the complex structure. Depending on the degree of complexity of the combining units, the levels of juncture are nuclear juncture, core juncture, clausal juncture, and sentential juncture. Nuclear junctures, for example, are complex constructions made up of more than one nucleus. For example, in John forced open the vault, two nuclei, force and open, can be found in the Core. Core junctures comprise two or more cores in a Clause, as in I asked John to force the vault open. In this type of core juncture, the two cores share a core argument, in this case the participant John. A clause juncture can be identified in more complex structures of the type John phoned Sue yesterday and Jill phoned her too. Further differences between the levels of juncture have to do with complementisers (to, from, that, etc.). Nuclear junctures do not include complementisers whereas core junctures may require them. As a result, the two nuclei can be adjacent in a nuclear juncture while they cannot be adjacent in a core juncture. In English, a nuclear juncture is only possible if the second predicate is intransitive (Van Valin and LaPolla 1997, p. 445).

The possible syntactic and semantic relations between the units in a juncture, called the nexus, include coordination and subordination. Subordination is divided into two subtypes: daughter subordination, when the Clause is an argument of the Sentence, as in That she arrived late shocked everyone; and peripheral subordination, when the Clause is a Periphery of the Sentence, as in Kim saw Pat after she had arrived at the party. Both subtypes of subordination are possible at the juncture levels of the Nucleus, the Core, and the Clause. For subordination to take place, it is a requirement that clefting and passivisation are possible. Thus, Mary regretted John’s losing the race is an instance of subordination because the cleft (It was John’s losing the race that Mary regretted) and the passive (For John to lose the race was regretted by Mary) are possible (Van Valin and LaPolla 1997, p. 445). RRG distinguishes a third nexus type, cosubordination. Cosubordination is dependent coordination. The dependence in cosubordination is due to the operators, given that the matrix predication and the linked predication must share at least one operator at the level of juncture. For example, in Mary sat playing the guitar, the operator of imperfect aspect has scope over both nuclei, sat and playing. The dependence in cosubordination also results from an argument shared by the matrix predication and the linked predication. In Mary sat playing the guitar, the first argument Mary is shared by sat and playing.

The semantic relation between the linked predication and the matrix predication is couched in terms of an inventory of semantic functions (Van Valin and LaPolla 1997, pp. 478–80) that includes, for instance, the first causative (the bringing about of one state of affairs directly by another state of affairs, as in Maggie pushed the door open), phase (when a separate verb describes the onset, continuation, or termination of a state of affairs, as in Hans finished writing the article), modifying subevents-manner (the manner in which an event is carried out, as in Bill entered the room whistling), direct discourse (the direct quotation of a speech event, as in Frank said, “Let’s get going”), circumstances (the spatial or temporal parameters of an event, as in Sam met Sally at the cafeteria after work), reason (the motivation of an event, as in The baby couldn’t sleep because she was hungry), etc.

Given the juncture–nexus types and the semantic relations reviewed above, the Interclausal Relations Hierarchy (IRH; see Figure 5) ranks juncture–nexus types on the basis of the tightness of the syntactic link between the units, on the one hand, and semantic relations on the basis of cohesion between the two propositions, on the other hand. On the syntactic part of the IRH, the degree of the integration of the two units is assessed: whether they are integrated into a single unit or remain two separate units. On the semantic part of the IRH, the semantic relations form a continuum expressing the degree of semantic cohesion between the propositional units linked in the complex structure—that is to say, whether they express a single action or event or discrete actions or events.

The IRH relates the cohesion of the semantic relation holding between the two propositions to the strength of the syntactic bond existing between the matrix predication and the linked predication. The IRH predicts that the closer the semantic relation between two propositions is, the stronger the syntactic link between the matrix predication and the linked predication must be. The semantic relations at the top of the IRH should be expressed by the linkage categories at the top of the syntactic part of the hierarchy, and the semantic relations at the bottom of the IRH should be expressed by the linkage categories at the bottom of the syntactic part of the hierarchy.

4. Method, Sources, and Data

The method that guides this research consists of three steps. The first is the selection of the relevant verbs and the identification of the meaning components of the verbal class under analysis. An initial inventory of Prevent verbs in Old English has been gathered with the data provided by the online version of the Thesaurus of Old English (Roberts et al. [1995] 2000). The Thesaurus of Old English has been searched for the lexical dimension that conveys the meaning ‘not doing something’ (Faber and Mairal 1999) and the lexical subdimension To cause somebody not to do something [prevent]. The provisional inventory of Prevent verbs includes āgǣlan, belēan, bewerian, forbēodan, forfōn, forhabban, for(e)sacan, and forwiernan. This inventory has been checked against the meaning definitions and citations provided by the Clark Hall ([1896] 1996) and Bosworth and Toller ([1898] 1973) Old English dictionaries, as well as by the Dictionary of Old English (Healey 2016) for the letters A–I. Even though Faber and Mairal (1999) do not make this distinction, the Old English data show that Prevent verbs (āgǣlan ‘to hinder’, belēan ‘to hinder’, forfōn ‘to take measures to prevent’, forhabban ‘to restrain’, and for(e)sacan ‘to refuse’) differ semantically and syntactically from Forbid verbs (bewerian ‘to prohibit’, forbēodan ‘to forbid’, and forwiernan ‘to forbid’). As instances of verbal polysemy may arise, all the fragments from both classes have been dealt with.

The second step in the methodology of this study is the compilation of the corpus of analysis. The data have been retrieved from the Dictionary of Old English Corpus (Healey et al. 2004). For the verbs beginning with the letters A–I, the data, including inflectional forms and meaning definitions, have been extracted from the Dictionary of Old English. The York-Toronto-Helsinki Parsed Corpus of Old English Prose (Taylor et al. 2003) has been used for the identification of the inflections of the verbs beginning with the letters L–Y and for the syntactic parsing of all the fragments (letters A–Y). A total of 155 fragments have been processed, which can be broken down by verb as follows: āblinnan (17), āgǣlan (4), belēan (6), bewerian (30), blinnan (13), for(e)sacan (5), forbēodan (63), forfōn (1), forhabban (24), forwiernan (22), geblinnan (3), and oflinnan (3). Of these verbs, Mitchell’s (1985) Syntax makes reference to forbēodan (§3723) and forwiernan (§847, §857) only. All the fragments have been provided with translations from available editions or directly translated with the help of the dictionaries by Sweet ([1896] 1976), Clark Hall ([1896] 1996), and Bosworth and Toller ([1898] 1973). The examples have been glossed with Leipzig interlinear morphosyntactic labels.

The third step of this research is the analysis of the relation between semantics and syntax. This specifically includes the description of the lexical representation and the logical structure of Prevent verbs and the analysis of the syntactic constructions in which these verbs are found. The aspects involved in the analysis of the relation between semantics and syntax include semantic roles, argument types, matrix vs linked predications of the clausal and nominal type, the status of the verbal form and the first argument of the linked predication, and the interclausal relation that holds between the matrix predication and the linked predication. Linked predications at the phrasal level draw special attention.

5. Competition with Prevent Verbs

The Aktionsart type of Prevent verbs is the Causative Activity. The x argument of the Activity, which plays the thematic role Agent and gets the macrorole Actor, impedes that the y argument of the linked predication performs an activity. This coincides with the y argument and the Undergoer of the matrix clause. The logical structure of Prevent verbs is given in Figure 6.

For example, in (1), the Patient hyra eagan ‘their eyes’ receives the Undergoer while the Theme is realised by a finite dependent clause (þæt hig hine ne gecneowun ‘that they would not recognise him’) that gets no macrorole.

(1)	[Lk (WSCp) 24.16]
	Soðlice hyra eagan wærun forhæfde þæt hig hine ne gecneowun.
	Soðlice	hyra	eagan	wærun
	indeed-ADV	he-GEN.3PL	eye-NOM.PL	be-PST.3PL
	forhæfde	þæt	hig	hine
	prevent-PST.PTCP	that-CONJ	he-NOM.3PL	he-ACC-3SG
	ne	gecneowun
	not-NEG	recognise-PST.3PL.SUBJV
	‘But they were prevented from recognising him.’

As can be seen in (2), Prevent verbs can be used absolutely.

(2)	[MtGl (Li) 006400 (3.14)]
	Soðlice foresoc ł forbead hine cueð ic from ðe rehtra is gefulwia & ðu cuom ł cymes to me.
	Soðlice	foresoc	ł	forbead
	truly-ADV	prevent-PST.3SG	and-CONJ	prohibit-PST.3SG
	hine	cueð	ic	from
	he-ACC.3SG	say-PST.3SG	I-NOM.1SG	by-PREP
	ðe	rehtra	is	gefulwia
	you-DAT.2SG	right-COMPAR	be-PRS.3SG	baptize-PST.PTCP
	&	ðu	cuom	ł
	and-CONJ	you-NOM.2SG	come-PST.2SG	and-CONJ
	cymes	to	me
	come-PRS.2SG	to-PREP	I-DAT.1SG
	‘Truly (John) prevented and prohibited him and said “it is more suitable that I am baptised by you and you come to me.”’

Far more frequently, however, the Theme of Prevent verbs is a linked predication realized by a finite clause. The linked verb can be an indicative form, but the general rule is that it is conjugated for the subjunctive, like geðeode ‘attach’ and gecyrre ‘convert’ in (3). The juncture–nexus construction in (3) is a clausal coordination because two units of the clausal level of juncture are involved, the first argument is not shared by the two clauses, and operators do not have scope over the two verbal predications.

(3)	[Bede 1 14.60.9]
	Ne we eow beweriað þæt ge ealle, ða þe ge mægen, þurh eowre lare to eowres geleafan æfæstnisse geðeode & gecyrre.
	Ne	we	eow
	not-NEG	I-NOM.1PL	you-ACC.2PL
	beweriað	þæt	ge
	forbid-PRS.1SG	that-CONJ	you-NOM.2PL
	ealle	ða	þe
	all-DAT.PL	then-ADV	which-REL
	ge	mægen	þurh
	you-NOM.2PL	may-PRS.2PL.SUBJV	by-PREP
	eowre	lare	to
	you-GEN.2PL	teaching-DAT.SG	to-PREP
	eowres	geleafan	æfæstnisse
	you-GEN.2PL	faith-DAT.SG	religion-DAT.SG
	geðeode	&	gecyrre
	attach-PRS.2PL.SUBJV	and-CONJ	covert-PRS.2PL.SUBJV
	‘Nor do we prevent you from attaching and converting to the religion of your faith, that you may, by your teaching.

In (4), the Theme is a linked predication realized by a clause with a non-finite form of the verb cuman ‘to come’. The juncture level of the construction is the core because the linked verb has arguments of its own (to me ‘to me’) and the nexus is coordination.

(4)	[Mt (WSCp) 19.14]
	Þa cwæð se hælend, lætað þa lytlingas & nelle ge hig forbeodan cuman to me.
	Þa	cwæð	se	hælend	lætað
	then-ADV	say-PST.3SG	the-NOM.SG	Saviour-NOM.SG	let-PRS.3PL
	þa	lytlingas	&	nelle	ge
	the-ACC.PL	child-ACC.PL	and-CONJ	will-PRS.PL.NEG	you-NOM.2PL
	hig	forbeodan	cuman	to	me
	he-ACC.3PL	forbid-INF	come-INF	to-PREP	I-DAT.1SG
	Jesus said, “Let the little children and do not prevent them from coming to me”.

In (5), the Theme is a phrasal predication that results from a nominalisation based on a verb. In (5a), the genitive weorca, morphologically related to weorcan ‘to work’, plays the thematic role Theme while the Patient is the PSA hio ‘she’, næs hio næfre weorca agæled ‘she was never prevented from working’ being a passive voice construction. In (5b), the genitive heora ‘their’ is the Patient while the accusative gedwyld ‘error’, which is morphologically related to dwolian ‘to err’, plays the thematic role Theme.

(5)	a.	[HomU 9 (Verc 4) 118]
		Ealle mine beboda hire wæron ieðe to donne; næs hio næfre weorca agæled.
		Ealle	Mine	beboda	hire
		all-NOM.PL	I-GEN.SG	command-NOM.PL	she-DAT.3SG
		wæron	Ieðe	to donne	næs
		be-PST.3PL	easy-NOM.PL	do-INF.INFL	be-PST.3SG.NEG
		hio	Næfre	weorca	agæled
		she-NOM.3SG	never-ADV	work-GEN.PL	hinder-PST.PTCP
		‘All my commands were easy for her to do. She was never prevented of working.’
	b.	[ÆCHom II, 5 44.83]
		Witodlice næs nan heahfæder. ne nan witega asend to hæðenum folce. þe heora gedwyld beloge ær drihtnes tocyme.
		Witodlice	Næs	heahfæder
		certainly-ADV	be-PST.3SG.NEG	patriarch-NOM.3SG
		nan	Witega	asend
		none-NOM.3SG	prophet-NOM.3SG	send-PST.PTCP
		ne	Hæðenum	folce
		not-NEG	heathen-DAT.3PL	people-DAT.3SG
		þe	Heora	gedwyld
		who-REL	she-GEN.3PL	sin-ACC.SG
		beloge	Ær	drihtnes
		prevent-PST.3PL.SUBJV	before-ADV	Lord-GEN.3SG
		tocyme
		advent-DAT.SG
		‘Certainly, no patriarch or prophet was sent to the heathen people who might prevent them from their sin before the Lord’s advent.’

Old English Prevent verbs such as āgǣlan, belēan, forfōn, forhabban, and for(e)sacan are characterised by the construction of coordination in juncture–nexus types of nuclear coordination, core coordination, and clausal coordination. In the three constructions, the first argument of the matrix predication is not shared by the linked predication. Figure 7 presents the constituent projection of the nuclear coordination construction, which comprises a deverbal nominalisation. The nominal nucleus þæs ganges ‘of going in’ and the verbal bewerede ‘prevented’ are adjacent. The accusative noun phrase me ‘me’ is the Patient while the genitive þæs ganges ‘of going in’ is the Theme. The PSA of the construction is the noun phrase controlling the agreement of bewerede ‘prevented’: þæt godcunda mægen ‘the divine power’. The macrorole Actor is assigned to the Agent while the Patient gets the macrorole Undergoer. In Figure 8, the constituent projection of the clausal coordination construction is provided. The complementizer þæt ‘that’ links the verbal predication realized by a finite clause hig ne sprecon faken ‘that they do not speak fake’. The linked predication is the Theme while the thematic role Patient is played by the implicit controller of the agreement of the imperative forhafa ‘(you) prevent’. As in (7), the Agent gets the Actor and the Patient receives the Undergoer.

Noun phrases, non-finite clauses, and finite clauses, therefore, can complement Prevent verbs in Old English. The competition on argumenthood affects the realisation of the thematic role Theme, which can be expressed by a nominalisation from a verbal predication or by a verbal predication.

6. Synchronic Perspectives on Diachronic Evolution

This section discusses the limits of the verbal classes at stake and the perspectives on diachronic evolution offered by the synchronic analysis carried out in this study. The implications for the IRH are also considered.

The description of the semantics and syntax allows us to draw distinctions between verb classes that have proven clear-cut enough to conduct the analysis and the discussion. For instance, forwiernan is a manipulative verb from the Forbid class in this analysis given that it takes a linked predication and a dative third argument, as in (6), in which the thematic role Theme is realized by the finite clause þæt he to hym ne gebæde ‘that he did not pray to him’ and the Patient is realized by the dative him ‘him’.

(6)	[ÆHom 30 79]
	Ac gyf hyt se witega wære, he wolde him forwyrnan þæt he to hym ne gebæde.
	Ac	gyf	hyt	se
	but-CONJ	if-CONJ	it-ACC.3SG	the-NOM.SG
	witega	wære	he	wolde
	prophet-NOM.SG	be-PST.SG.SUBJV	he-NOM.3SG	will-PST.3SG
	him	forwyrnan	þæt	he
	he-ACC.3SG	forbid-INF	that-CONJ	he-NOM.3SG
	to	hym	ne	gebæde
	to-PREP	he-ACC.3SG	not-NEG	pray-PST.3SG.SUBJV
	‘But if he were his lord, he would forbid him that he did not pray to him.’

The syntactic behaviour of Forbid verbs is clearly different from the Prevent verbs discussed in this article, which are found in nexus types of coordination. For this reason, forwiernan with a second argument in the genitive must be considered a Prevent verb, thus admitting a certain degree of leaking between verbal classes. Although instances like (7) are exceptional, verbal polysemy arises at this point.

(7)	[ChrodR 1 6.17]
	And gif se eard sy wynes wæstmbære, sylle man dæghwamlice ælcum breðer fif punda gewihte wines, gif þa unwedru his ne forwyrnað.
	And	gif	se	eard
	and-CONJ	if-CONJ	the-NOM.SG	earth-NOM.SG
	sy	wynes	wæstmbære	sylle
	be-PRS.3SG	wine-GEN.SG	fruitful-ACC.SG	give-SUB.PRS.3SG
	man	dæghwamlice	ælcum	breðer
	one-NOM.SG	daily-ADV	each-DAT.SG	brother-DAT.SG
	fif	punda	gewihte	wines
	five-NUM	pound-GEN.PL	weigh-GEN.SG	wine-GEN.SG
	gif	þa	unwedru	his
	if-CONJ	the-ACC.SG	bad-weather-ACC.SG	he-GEN.3SG
	ne	forwyrnað
	not-NEG	prevent-PRS.3SG
	‘And if the earth is devoid of wine, one must give daily to each brother five pounds of weighted wine, if the bad weather does not prevent him’

The analysis of the data turns out the results tabulated in

Table 1. Clausal and nominal linked predications.

	Simple Clause	Linked Clause	Nominalisation	Total
āgǣlan	1	2	1	4
belēan	2	3	1	6
bewerian	14	14	2	30
forbēodan	27	30	6	63
for(e)sacan	4	1	0	5
forfōn	0	1	0	1
forhabban	14	7	3	24
forwiernan	7	13	2	22
TOTAL	89	80	22	155

. From the quantitative point of view, the results indicate that the existence of complementation with deverbal nominalisations correlates with a higher number of instances of linked clauses than of simple clauses. This is the case with āgǣlan, belēan, forbēodan, and forwiernan. With other verbs, the occurrences of complementation with linked nominalisation seem to be a function of the relatively higher textual frequency of the verb. The results of bewerian and forhabban can be interpreted in this way. Overall, āgǣlan, belēan, for(e)sacan, and forfōn throw less than ten occurrences, which advises one to be cautious at this point. From a quantitative–qualitative perspective, all verbs with more than ten occurrences conveying the meanings under analysis are complemented by linked clauses and linked nominalisations.

It is possible to describe two types of diachronic continuity on the basis of the evidence gathered in Table 1, namely continuity between finite and non-finite linked verbal predication, illustrated by instances like (8a); and continuity between non-finite and nominalised linked predication, as in (8b). In (8a), forbeodan is complemented by the finite clause þæt he onfoe niwecumenum preostum ‘that he performs a service with new priests’ and by the non-finite clause & to gehælgenne ferunga ‘and to consecrate them afterwards’. In (8b), the same verb takes a deverbal noun phrase (unrihtwisnyssa ‘follies’) and a non-finite clause (yfel to donne ‘to do evil’).

(8)	a.	[MtMarg (Li) 10.14]
		Biscope is forboden þæt he onfoe niwecumenum preostum & to gehælgenne ferunga
		Biscope	is	forboden	þæt
		bishop-DAT.SG	be-PRS.3SG	forbid-PST.PTCP	that-CONJ
		he	onfoe	niwecumenum	preostum
		he-NOM.3SG	accept-PRS.3SG.SUBJV	new-coming-DAT.PL	priest-DAT.PL
		&	to gehælgenne	ferunga
		and-CONJ	consecrate-INF.INFL	afterwards-ADV
		‘It is forbidden that the bishop performs a service with new priests and to consecrate them afterwards.’
	b.	[ÆLet 6 210]
		…he us ne forbeode ealle unrihtwisnyssa and yfel to donne.
		he	us	ne	forbeode
		he-NOM.3SG	I-DAT.1PL	not-NEG	prohibit-PRS.3SG.SUBJV
		ealle	unrihtwisnyssa	and	yfel
		all-ACC.PL	unrighteousness-ACC.PL	and-CONJ	evil-ACC.SG
		to donne
		do-INF.INFL
		‘…he does not prohibit us all follies and to do evil.’

The possibility of inserting, at the same level and linked by a coordinating conjunction, a finite and a non-finite clause such as þæt he onfoe niwecumenum preostum & to gehælgenne ferunga in (8a) and a nominalisation and a non-finite clause like ealle unrihtwisnyssa and yfel to donne in (8b) constitutes additional evidence in favour of the competition between two units that perform the same function (verbal argument with the thematic role Theme) but belong to different levels of structure (linked clause vs. linked noun phrase). In Section 5, evidence for paradigmatic competition was gathered that involved a certain verb in various contexts. Instances like (8a) and (8b) represent additional evidence for competition, which arises on the syntagmatic axis because the structures in competition are found in a certain expression. Considering that the complementation with linked finite clauses and with linked infinitival clauses has not survived in Present-Day English, these instances point to a diachronic development of the following type: finite clause > non-finite clause > nominalisation.

The IRH predicts that, on the diachronic axis, these verbs are likely to replace finite clause complementation with non-finite clause complementation. Juncture–nexus types of clausal coordination of the kind found with Prevent verbs yield way to core or nuclear junctures of the same nexus types. It is remarkable in this respect that the semantic function and the nexus type remain stable on the diachronic axis while the type of juncture changes. From the point of view of a functional theory of language, this is a clear case of the priority of function over structure.

This explanation concurs with Rohdenburg (1995, 2006) and Los (2005), although the explanation based on the IRH is more motivated with semantics than Los (2005) because it takes into account fully semantic aspects such as event integration and more principled than Rohdenburg (1995, 2006) because it resorts to an exhaustive taxonomy of syntactic constructions, levels of juncture, types of nexus, and interclausal semantic relations. Moreover, Rohdenburg (1995, 2006) and Los (2005) do not consider the competition between units from the nominal and the clausal levels for argumenthood, which takes place, at least, with control, manipulative, and aspectual verbs.

This explanation is compatible with the development of the gerund from derived nominals suffixes with -ung/-ing, the present participle and the inflected infinitive. The only survival of these forms into Middle English was the -ing form (Lass 1992, p. 145), which acquired verbal properties such as the ability to take a direct object realized by a noun phrase and to be modified by an adverb (Fischer 1992). Visser (1963–1973, §1009) dates the first instances of the verbal gerund to the beginning of the 14th. century. In Early Modern English, the type prevent someone doing is attested from 1592 onwards (Visser 1963–1973, §1092). The earliest instance of the prevent someone from doing something type is dated by Visser (1963–1973, §2108) to the 18th century: It was my business to prevent him from enjoying a third chance [1748, Smollet, Rod. Random (Tauchn) XLIX p. 318]. Rohdenburg (2006) calls the changes in the complementation of the English verb The Great Complement Shift. Iyeiri (2010) makes a further distinction between the shift from that-clauses to to-infinitives on the one hand and the shift from infinitives to gerunds on the other, but, as has been remarked in Section 2, linked nominalisations are not considered.

This study allows us to draw the conclusion that the noun phrase comprised of a deverbal nominalisation can be described as syntactically tighter than the dependent non-finite clause, which is also tighter than the dependent finite clause. This conclusion may call for the enlargement of the IRH so that it includes noun phrases as the tightest syntactic option. Studies in linguistic typology, such as Dixon and Aikhenvald (2006), demonstrate the typological continuity of nominalisations and clauses as verbal complements. On the diachronic axis, the evolution from two clauses with two subjects to two clauses with one shared subject and, subsequently, to a clause with a linked noun phrase with verbal arguments as nominal modifiers is parallel to the evolution from paratactic verb complements (two subjects) to syntactic verb complements (shared subject) put forward by Heine and Kuteva (2007). This development is described with respect to nominalisations by Givón (2009, p. 68) as follows: from complex clause to complex word. For Givón (2009, p. 68), the degree of event integration results from referential integration (the sharing of referents between the two events), temporal integration (simultaneity or temporal adjacency of the events), and spatial integration (the two events share the location). Event integration explains the changes on the Finiteness Scale (Givón 2009, p. 68), which ranks expressions from the least finite (Her knowledge of mathematics) to the most finite (She knew mathematics well).

Despite the lack of lexical continuity of the Old English class of Prevent verbs, which disappeared and were replaced by Romance loanwords, including to prevent itself, in Middle English, it turns out that Present-Day English relies on a morphologically oblique noun phrase for the complementation of this verbal class. The fact that obliqueness is due to morphological case in Old English, whereas it results from prepositional government in Present-Day English, reflects the overall development of the language rather than representing a local phenomenon. From the point of view of verbal semantics, obliqueness can be regarded as a consequence of the combined presence of the thematic roles Patient and Theme in the complementation pattern of Prevent verbs, both in Old English and in Present-Day English.

7. Conclusions

This article has delved into the complementation of Old English control verbs with oblique marking from the Prevent class. Phrasal and clausal linked predications compete for argument status. This competition has been identified both on the paradigmatic axis (in verbal classes) and on the syntagmatic axis (in hybrid complementation patterns involving a finite and a non-finite clause and a clause and a noun phrase). This evidence suggests that there may have been a diachronic development: finite clause > non-finite clause > nominalisation. The IRH predicts that Prevent verbs are likely to replace finite clause complementation with non-finite clause complementation. The noun phrase based on a deverbal nominalisation is syntactically tighter than the non-finite clause and constitutes the next step of syntactic development. The IRH, therefore, may be enlarged to include noun phrases as the tightest syntactic option, but more research is needed in this respect. The acquisition of verbal properties by a former noun reinforces the parallelism between nominal and verbal predications and, ultimately, the applicability of the IRH to noun phrases. Finally, the only nominalisations that have been considered when dealing with the development of the English gerund are -ing/-ung verbal derivatives, but evidence like that gathered in this article indicates that other deverbal nominalisations must also have contributed to the generalisation and spread of the -ing verbal form. This aspect deserves more attention in future research too.