John Benjamins Publishing Company

John Benjamins Publishing Company
This is a contribution from Synchrony and Diachrony. A dynamic interface.
Edited by Anna Giacalone Ramat, Caterina Mauri and Piera Molinelli.
© 2013. John Benjamins Publishing Company
This electronic file may not be altered in any way.
The author(s) of this article is/are permitted to use this PDF file to generate printed copies to
be used by way of offprints, for their personal use only.
Permission is granted by the publishers to post this file on a closed server which is accessible
to members (students and staff) only of the author’s/s’ institute, it is not permitted to post
this PDF on the open internet.
For any other use of this material prior written permission should be obtained from the
publishers or through the Copyright Clearance Center (for USA: www.copyright.com).
Please contact [email protected] or consult our website: www.benjamins.com
Tables of Contents, abstracts and guidelines are available at www.benjamins.com
Gradual change and continual variation
The history of a verb-initial construction in Welsh
Oliver Currie
University of Ljubljana
This article contrasts two different analyses – a diachronic Construction
Grammar (CxG) approach and the Principles & Parameters approach of Willis
(1998) – of the development of a verb-initial construction, Absolute-initial
verb (AIV) order, in Early Modern Welsh. The P&P approach attributes the rise
of AIV order in Early Modern Welsh to an abrupt and discrete change in the
grammaticality of V1 following the resetting of the V2 parameter. We argue, on
the basis of a detailed corpus study of the period c.1550–c.1750, that the historical
data shows a gradual increase as well as significant sociolinguistic variation in the
frequency of use of AIV order. We further argue that a diachronic Construction
Grammar approach can better account for gradual syntactic change and syntactic
variation, since, unlike P&P approaches, it does not seek to model gradual
historical data in terms of discrete grammars and grammatical categories, but has
a gradient conception of grammaticality and grammatical categories and can thus
propose gradual mechanisms of change and integrate sociolinguistic variation
directly in grammatical description.
1. O
verview
The prominence of “gradualness” in studies of syntactic change stems in a large
part from the apparent paradox that grammatical theory, in particular generative
grammatical theory, typically envisages syntactic change as discrete and abrupt,
while observable syntactic change in language use over time appears to be gradual.
Accordingly, much of the debate on gradualness in syntactic change has centred on
reconciling the discrete mechanisms of syntactic change posited by generative grammatical theory with the gradual patterns of syntactic change evident in historical
linguistic data.
In this article we will approach the question of gradualness from a somewhat different angle and examine whether it is necessary to posit primarily discrete mechanisms of
syntactic change in the first place. The discussion will be based on a case study of word
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
order change in Welsh: the increase in use of a verb-initial construction, ­Absolute-initial
verb (AIV) order, where a finite verb comes in absolute initial position in a positive
declarative main clause, in Early Modern Welsh (c.1500–c.1800) prose. Absolute-initial
verb order was rare in Middle Welsh (c.1100–c.1500) prose, which had a characteristic
verb-medial word order, but came to be used more frequently in prose texts from the
second half of the sixteenth century; however for more than two c­ enturies after that we
find significant variation in the frequency of use of the construction, mostly between different writers (some use AIV order as the dominant word order, others appear to avoid
the construction and others still show various ­intermediate ­patterns of use), but also to
lesser extent within individual writers.
We will contrast a Principles and Parameters (P&P) account of this change in Welsh
word order by Willis (1998), which seeks to explain the rise of AIV order in terms of a
discrete and abrupt change in the grammaticality of the construction, but which does
not in itself seek to account for the variation in the use of the construction, with an
alternative Construction Grammar-based (CxG) approach, where we posit a gradual
mechanism of syntactic change and seek to propose an integrated syntactic and sociohistorical account of the change and variation in the use of AIV order. In the Principles
and Parameters approach, Middle Welsh is analysed as a V2 language, where unmarked
VSO is ungrammatical in positive declarative main clauses, and the emergence of
grammatical unmarked VSO in Early Modern Welsh is attributed to the resetting of
the V2 parameter, resulting in an abrupt change from a V2 grammar with ungrammatical unmarked VSO to a non-V2 grammar with grammatical unmarked VSO. In
the Construction Grammar approach, Absolute initial-verb order is analysed as being a
grammatical, but weakly motivated construction in Middle Welsh prose and the gradual
increase as well as the variation in its use is analysed in terms of changing and ­competing
motivations (both structural/syntactic and stylistic) for its use over time.
The rest of this article is structured as follows. Section 2 outlines the theoretical
differences between the P&P and CxG approaches and how these differences impinge
on the treatment of the key issues discussed here: the gradualness vs. discreteness of
syntactic change and the understanding of variation and its relationship to syntactic
change. Section 3 presents an overview of Middle Welsh word order, focusing on the
status of the main construction under investigation, AIV order. Section 4 presents an
overview of the Early Modern Welsh data, focusing on the evidence of the increase
in use of the AIV order compared to Middle Welsh prose and the variation in the
­frequency use of AIV order. Finally, Section 5 presents the conclusion.
The main aim of this article is to argue that it is possible to account for the observed
gradualness of the syntactic change and for syntactic variation using a diachronic
Construction Grammar approach. The alternative P&P approach of Willis (1998) is
presented primarily by way of contrast; it is not within the scope of this article, for
reasons of space, to provide a detailed exposition or critique of the P&P approach.
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
2. Gradualness and variation
2.1 The gradualness paradox
Opening their landmark 1968 article, Weinreich, Labov and Herzog made the striking
and challenging observation that “[…] structural theories of language, so fruitful in
synchronic investigation, have saddled historical linguistics with a cluster of paradoxes
that have not been fully overcome” (1968: 98). Although Weinreich, Labov and Herzog
focused on sound change, their observation is no less relevant to the field of syntactic
change. The paradox which is the starting point for the present discussion concerns
the question of the gradualness of syntactic change; on the one hand observable syntactic change – changing patterns of language use in texts or discourse over time –
appears to be gradual, on the other hand, the putative mechanisms of syntactic change,
based on synchronic theories of grammar (especially, though not exclusively, generative
­grammar), are often inherently discrete (i.e. non-gradual), in that they involve changes
from one discrete grammar (e.g. parameter setting) to another or from one discrete
formal category to another (van Kemenade & Vincent 1997: 3; Roberts 2007: 295).
Since the discrete nature of grammatical change is assumed to be a given in
­Principles and Parameters analysis, it is generally argued within Principles and Parameters approaches, in order to reconcile the discrete conception of syntactic change
with the gradual historical data, that discrete mechanisms of change in speakers’ internalised grammar (I-language) underlie the apparently gradual patterns of change in
language use (E-language). For example, variation between new syntactic structures
resulting from a parametric change and structures associated with the old parameter
setting in written texts can be attributed to stylistic conservatism, cf. Willis (1998: 44):
“writers will have access to a considerable body of earlier literature in their language
[…] If the literary tradition is a strong and continuous, there is no reason why the conservative pattern should not be maintained for many centuries.” Further, it is possible
to argue for a discrete and abrupt actuation but gradual diffusion of syntactic change
and propose models within the Principles and Parameters framework to account for
the gradual diffusion of syntactic change and variation associated with it in terms of
discrete mechanisms, such as competition between grammars (Kroch 1989; Pintzuk
2003), lexical diffusion (Roberts 2007: 297–300), microparametric change (Roberts
2007: 300–305) and formal optionality (Roberts 2007: 305–309). It can, therefore, be
argued that gradualness in the manifestation of syntactic change does not necessarily
provide counter-evidence to the discrete, parametric model of syntactic change. The
gradualness paradox nevertheless highlights a potential mismatch between linguistic theory and historical linguistic data, which can undermine the explanatory power
of principles-and-parameters’ approaches to syntactic change. Although the primary
goal of principles-and-parameters approaches is to explain changes in “I-language”
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
rather than provide detailed accounts of syntactic variation and change observable
in historical data (“E-language”) (Faarlund 1990: 5–6), the evidence for any putative
parametric or other changes in I-language must necessarily derive from such historical data; if it is possible to argue that a putative discrete grammatical change can be
compatible with quite different patterns of change evidenced in historical data, the
problem then arises of how we can validate such an analysis from an empirical (as
opposed to a theory-internal) point of view.
Here we propose to approach the gradualness paradox from a different angle:
rather than asking whether observed gradualness can be reconciled to a discrete
mechanism of change, we shall examine whether it is actually necessary to posit a
discrete mechanism of change and argue that, in a Construction Grammar-based
approach, it is possible to propose gradual (though not necessarily exclusively g­ radual)
mechanisms of syntactic change. A key theoretical difference between P
­ rinciples
and ­Parameters and Construction Grammar which has a bearing on our discussion of the gradualness of syntactic change, is that unlike Principles and ­Parameters
theories of grammar, Construction grammar does not have a binary conception of
­well-formedness or have necessarily discrete grammatical categories, but allows for
clines and continua. In Construction Grammar, as in Cognitive Grammar, the wellformedness (or conventionality) of a grammatical construction is usage-based and
is determined by the extent to which the construction is motivated (cf. G
­ oldberg
1995: 67–73; ­Goldberg 2006: 217–219) by use (e.g. frequency, context, stylistic/
social connotations) and by its relationship (e.g. formal and functional similarity)
to other constructions in the language (cf. Bybee 2010: 214). Therefore, a diachronic
­Construction Grammar approach can account for changes in the frequency of use of
a construction in terms of motivation, and in turn relate changes in well-formedness
to frequency of use, e.g. in the emergence of new constructions and obsolescence of
existing ones. Crucially in such an approach, gradual changes in the frequency of use
of a construction can be argued to reflect gradual changes in motivation.
2.2 Gradualness and variation in Diachronic Construction Grammar
The Construction Grammar approach used in this article corresponds most closely to
what Goldberg (2006: 214–215) terms “Cognitive Construction Grammar (CCxG)” –
as opposed to “Unification Construction Grammar” (e.g. Fried 2009) – in that the
approach is usage-based, the construction is the basic unit of syntactic analysis, motivation is used a key explanatory tool and formal notation is not used as a central part of
the analysis. The two principal (interrelated) theoretical issues discussed in this article –
the gradualness of syntactic change and synchronic variation and its relation to ­syntactic
change – have received considerable attention in recent D
­ iachronic ­Construction (and
Cognitive) Grammar studies, though the approach here is s­ omewhat different in its
emphasis, primarily because of the nature of data in this case study.
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
The change under investigation involves the increase in use of a construction and
we argue that the increase in use is gradual, in the sense of continuous as opposed
to discrete, in that it happens progressively over a protracted period, reflecting the
fact that the factors motivating the change (and variation) in its use are continuously
available during the period under investigation. Gradualness is defined in a different sense in Traugott and Trousdale (2008) as “a sequence of micro-steps affecting
various aspects of the use and structure of a linguistic sign”; the micro-steps are themselves “discrete and therefore abrupt for individual speakers”, but change, in particular grammaticalisation in Traugott and Trousdale’s discussion, is gradual on a macro
and speech community level, since it comprises a series of micro-steps. These two
conceptions of gradual change are not necessarily incompatible, but rather reflect a
difference of emphasis. In this study were are not dealing with grammaticalisation
or change in the meaning or function of a construction but an increase – and variation – in use motivated by sociolinguistic and stylistic factors as well as structural
analogy (i.e. the increase in use of AIV order is motivated by the increase in use of
another, formally similar construction). Similarly to Traugott and Trousdale (2008),
Fried (2008 and 2013) and Bergs and Diewald (2008: 10), which also discuss gradualness in a Diachronic Construction Grammar framework, deal with grammaticalisation or ­semantic or functional changes in constructions and understand gradualness
as ­step-by-step changes in the features of a construction.
In the discussion of variation in the present case study, there is a two-fold difference
in emphasis compared to recent Diachronic Construction Grammar studies. First, diachronic Construction Grammar studies have tended to focus on synchronic syntactic
variation as a manifestation of earlier diachronic change (e.g. Traugott & Trousdale 2008;
Barðdal 2011) and thus as a means of detecting and understanding diachronic change as
well as in turn of providing a potential basis for syntactic reconstruction (Barðdal 2013).
There is also significant evidence in the present study of synchronic variation appearing
to be both the result of previous diachronic change (in the case of the variation between
rare AIV order in Middle Welsh prose and common AIV order in Middle Welsh poetry,
cf. Section 3 below) and ongoing diachronic change (in the case of the variation in the
frequency of use of AIV order between different Early Modern Welsh prose texts, cf.
Section 4 below). However, we further argue that the variation observed is not just a
manifestation of diachronic change, but itself a potential driver of diachronic change, as
the variation in the use of AIV order between Middle Welsh prose and poetry and the
potential association of the construction with poetic style provided a resource which
later Early Modern Welsh prose writers could use in extending the use of AIV order (cf.
Section 4.5 below). Second, compared to recent Construction or C
­ ognitive G
­ rammar
studies of syntactic variation (e.g. Gries 2003, Grondelaers, Speelman & Geeraerts
2008, Stefanowitsch & Gries 2008; Cappelle 2009a & 2009b, Haser & ­Kortman 2009; De
Sutter 2009, Berg-Olsen 2009; Rosenbach 2008; Szmrecsanyi 2010), there is a greater
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
emphasis here on sociolinguistic, specifically stylistic factors as determinants of syntactic variation as opposed to linguistic ones, such as lexical, semantic, discourse pragmatic
or prosodic factors. The particular significance of stylistic variation in the present casestudy of Early Modern Welsh word order may be attributable to the fact that the construction under investigation, AIV order, is highly schematic and abstract and appears
to have been used interchangeably with the two main constructions with which it was in
competition without affecting meaning. As a result, semantic, discourse pragmatic and
lexical factors appear to be relatively less significant in this case.
3. Overview of Middle Welsh data
3.1 Overview of Middle Welsh prose word order
Middle Welsh prose had a predominantly verb-medial word order in positive declarative main clauses (PDMCs), but a predominantly VSO word order in subordinate
clauses and negative main clauses. This article deals only with the word order of
PDMCs, which has changed significantly in the history of Welsh; the word order of
subordinate clauses has, in contrast, been more stable.
Typically in PDMCs in Middle Welsh prose, at least one constituent precedes
the verb in clause-initial position, most often only one constituent, but the fronting of multiple constituents is also common. Poppe (2000: 42) posited the following
abstract schema for Middle Welsh prose word order, with the choice of the preverbal
constituent(s) determined largely by discourse-pragmatic factors:
(C4/3/2) C1 P V (S) (Onom) (A)
[Key to abbreviations: C1-4 = fronted Constituents; P = preverbal Particle (a/y);
V = finite Verb; S = Subject; Onom = nominal Object; A = adverbial phrase]
A preverbal particle comes between the preverbal constituent and the verb, either
a, where the preverbal constituent is a subject or direct object, or y, where the preverbal constituent is an adverbial phrase or prepositional phrase. These particles are
formally identical to, and historically derived from, relative particles: a is the particle
used in direct relative clauses (where the antecedent is a subject or direct object) and
y in oblique relative clauses (where the antecedent is an adverbial or prepositional
phrase). According to statistical word order analyses carried by Poppe (1991), there
is no single dominant word order in PDMCs in Middle Welsh prose. The commonest
verb-medial order types are illustrated in Examples (1) to (5) below.
(1) Pronominal Subject + a + verb
mi a wn-af
dy gynghordi
I prtdo-prs.1sgyouradvice 2s
‘I will act on your advice’
© 2013. John Benjamins Publishing Company
All rights reserved
(Thomson 1972: 15, 402–3)
The history of a verb-initial construction in Welsh 
(2)Nominal Subject + a + verb
Ekennadeu a ae-thant ar ol Matholwch
themessengersprtgo-pst.3plafterMatholwch
‘The messengers went after Matholwch’
(Williams 1930: 33, 26)
(3) Nominal Object + a + verb
Ymarch a gym-erth
thehorse prttake-pst.3sg
‘He took the horse’
(Thomson 1972: 8, 217)
(4) Verb noun object + a + AUX gwneuthur “to do” (‘gwneuthur-inversion’)
Kynhewi a orucPwyll
Fall-silent
prt do.pst.3sgPwyll
‘Pwyll fell silent’
(Thomson 1972: 12, 323)
(5) Adverbial phrase + y(d) + verb
Yna y rod-es
Arawny furufa’y
dryche hun
Then
prtgive-pst.3sgArawnhisformand hislook hisown
yPwyll
toPwyll
‘Then Arawn gave Pwyll his own appearance back.’ (Thomson 1972: 6, 138)
According to the word order frequencies calculated by Poppe (1991) for seven
­Middle Welsh, the frequency of finite verb-initial constructions varies from zero
(in the case of two texts) to 9.6%. Examples of three verb-initial constructions are
given in (6) to (8) below; in (6) and (7) a preverbal particle precedes the verb in
clause-initial position; in (11), AIV order, the verb comes in absolute-initial position. The commonest verb-initial construction in Middle Welsh prose is y + verb
order (8), with the preverbal particle y, which is also used when an adverbial phrase
is fronted as in (5) above.
(6) Y + verb order
‘Yd af
i yn agel y gyt
ac wynt,’ heb y Peredur
prtgo.prs.1sgIpredangeltogetherwiththem said Peredur
‘I shall become an angel with them,’ said Peredur
(Goetinck 1976: 8, 18)
(7) The dummy subject construction
efa doeth
taraneu a mellt
It
prtcome.pst.3sgthunderandlightning
‘There came thunder and lightning’
(Roberts 1975: 107, 1)
(8) AIV order
Gwel-sant niuer Otgareu meint
see-past.3plretinueOtgartheirstrength
‘Otgar’s retinue saw their strength’
(Bromwich & Evans 1997: 37, 1043)
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
3.2 AIV order in Middle Welsh poetry
AIV order was rare in Middle Welsh prose – there are no examples at all in several
texts and often only one or two in those where the construction is attested – yet it is
common in Middle Welsh poetry. An interesting illustration of the difference between
Middle Welsh prose and poetic word order is provided by the mixed prose and poetic
text Gwassanaeth Meir (edited in Roberts 1961), a fourteenth century Welsh translation of the Latin liturgical text Officium Parvum Beatae Mariae Virginis, including short extracts from the Vulgate gospels and several psalms. The poetic sections
(translations of Psalms and hymns), which make up the majority of the text, contain
numerous examples of AIV order (48 out of 179 PDMCs, i.e. 27%), whereas the prose
sections (mostly translations of short extracts from the gospels) contain no examples
of AIV order (out of 48 PDMCs). Moreover, we find examples where verb-initial constructions in the Latin original are translated using AIV order in the Welsh poetry, as
in Example (9), but using other verb-initial constructions than AIV order in the Welsh
prose, such as the dummy subject construction in Example (10) and the gwneuthurinversion in Example (11), where a verbal noun object of the auxiliary verb gwneuthur
“to do” is fronted; the gwneuthur-inversion is a construction particularly associated
with Middle Welsh narrative prose and was typically used in sequences of narrated
events. Examples (9) to (11) suggest that the author of Gwassanaeth Meir was sensitive to verb-initial order in the Latin original and used formally similar constructions
in the Welsh translation, but consciously avoided using AIV order in the prose, while
favouring it in poetic sections.
(9) Verb-initial in Latin > AIV order in Welsh poetry
magnific-av-it Dominusfacere no-bis-cum
Magnify-pst-3sgLord
do-infwe-abl-with
Mawrha-awd yr Arglwydyn gwneuthur
magnify-pst.3sgtheLord
ourmake-vn
‘The Lord performed a wonder making us.’
(Vulgate, Psalms 125: 3; Roberts 1961: 27, 22–23)
(10) Verb-initial in Latin > dummy subject + verb in Welsh prose
Da-bit
ill-i Domin-usDe-us sed-em
Give-fut.3sghe-dat.sgLord-nom.sgGod-nom.sgseat-acc.sg
Davidpatr-is e-ius
David-genfather-gen.sghe-gen.sg
Efa ryd
yr ArglwydDuwy-daw eistedua
Itprtgive.prs.3sgthelord
God to-3sg.mthrone
Dauydy tat
David hisfather
“The Lord God will give him the throne of David his father.”
(Vulgate, Luke 1:30, Roberts 1961: 7: 14–15)
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
(11) Verb-initial in Latin > non-finite verb initial construction in Welsh prose
et a-it
angel-use-i
Andsay-pst.3sgangel-nom.sgshe-dat.sg
netim-eas
Maria
negfear-pres.subj.Mary.voc
A dywetuta oruc
yr aghelwrth-i: Nac
Andsay.vn prtdo.pst.3sgtheangelto-3sg.f neg
ofnh-aa,Ueir
fear-imp.2sgMary
“And the angel said to her ‘Do not fear, Mary!’”
(Vulgate, Luke 1: 30; R
­ oberts 1961: 7, 10)
To give an idea of the frequency with which AIV order could occur in a non-translated
Middle Welsh poetic text, out of 100 PDMCs in the cywyddau (100–126) of Dafydd ap
Gwilym (1320–1380) (Parry 1979), there are 37 examples of AIV order.
3.3 The status of AIV order in Middle Welsh prose
Understanding the status of AIV order in Middle Welsh is crucial for any explanation of the subsequent increase in its use in Early Modern Welsh. The P&P approach
of W
­ illis (1998) analyses Middle Welsh as having a strict V2 requirement, where
“unmarked/true” VSO is ungrammatical, and explains the emergence of unmarked
VSO as a result of the loss of V2 following the resetting of the V2 parameter; the
instances of V1 in Middle Welsh prose as having a null topic operator, such that they
conform to the V2 constraint:
“[…] Middle Welsh does not have verb-initial clauses in the syntax. Even if we
consider only the surface ordering, we must accept that verb-initial ordering
is highly marked, occurring only in contexts of narrative continuity. It is thus
entirely different from the neutral VSO order of Contemporary Welsh. Once we
have posited a null topic operator in the topic position of apparent instances of
V1, it is possible to maintain a strict V2-requirement for Middle Welsh.”
(Willis 1998: 129)
For a P&P framework, reconciling the presence of V1 with a V2 analysis is a general
theoretical problem, since Middle Welsh is not an isolated case of a language analysed
as V2 but which also allows V1. Old English, Old French, Old Portuguese, Icelandic and
Yiddish are all similar in this respect. Different analyses of V1 in V2 grammar have been
proposed in a P&P framework. One approach, like that of Willis (1998), is to assimilate
the V1 structures to the dominant V2 pattern and in this way maintain the obligatory
V2 constraint. Roberts (1993: 96), for example, analyses the V1 structures in declarative
clauses in Old French as having a ‘null operator’ in the clause-initial slot, so that they
conform to the V2 constraint. Junker (1990), however, adopts a very different approach
in her analysis of V2 in Middle French, specifically arguing that the occurrence of V1
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
in declarative clauses shows that the V2 topicalisation of a preverbal constituent is not
obligatory. Fontana (1997: 210), argues more generally against a strict V2 requirement,
citing as evidence against strict V2 the occurrence of “declarative V1 constructions” in
V2 languages such as Old Norse, Old English, Yiddish and Icelandic.
Since V1 can quite commonly occur as a variant order in declarative sentences in
languages which have another pattern as the dominant order or equally in languages
with flexible word order (Luraghi 1995) – especially when the language also has prodrop features – there would seem to be no a priori reason why V1 should not also
occur in languages which show verb-second effects. In fact, it seems that V1 can occur
more or less frequently in ‘V2 languages’, ranging from a marginal use of V1, at one
extreme, to a relatively common use, at the other. The same may be true diachronically: a language may over time develop progressively more or less systematic V2 effects,
with the frequency of V1 declining as V2 is generalised and conversely increasing as
V2 patterns become less generalised. Indeed, it can be argued that the earlier history
of Welsh (Old to Middle Welsh) provides such an example, where V2 word order patterns spread at the expense of V1. The evidence of Old Welsh texts (prose and poetry),
earlier Middle Welsh prose texts such as Kulhwch ac Olwen and Middle Welsh poetry
suggests that AIV order had been significantly more frequent. The Middle Welsh verbmedial topicalisation system using the preverbal particles a and y, which are formally
identical to relative particles, is likely to have originated as a cleft focus system which
became extended and shifted to a topicalisation system (cf. Willis 1998: 100).
In a usage-based and bottom-up approach like Construction Grammar, the examples of AIV order in Middle Welsh prose can be accepted at face value as instances of
an independent, albeit rare verb-initial construction; it is not necessary to subsume
them in a V2 system as in a top-down, P&P approach, or to reject them as ungrammatical (cf. Currie 2000). The question whether AIV order or V1 in Middle Welsh
prose is “marked” or not – Willis (1998: 129) describes it as “highly marked, occurring
only in contexts of narrative continuity” – underscores a further theoretical difference
between the P&P and CxG approaches. P&P theory presupposes the autonomy of syntax and assumes that it is possible to account for the grammaticality of constructions
without reference to their meaning; since Willis distinguishes between V1 order, which
is characterised as having a specific, restricted discourse-pragmatic function, and V2
order, which does not, V1 is “marked” in relation to V2 order. In a CxG framework,
where constructions are by definition pairings of form and meaning, every construction is deemed to have a function (e.g. semantic, pragmatic, stylistic) associated with
it. Further, in a CxG or functional syntax framework, a “narrative continuity” function
would not be considered as marked in discourse-pragmatic terms, but rather as neutral, compared, for example, to topic shift or focus. Semantically or pragmatically, the
instances of AIV order in Middle Welsh prose, e.g. (8), (12a) and (13a), do not appear
to be marked. In (12) and (13) we have apparent instances of interchangeability of AIV
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
order with more frequent constructions: the dummy subject construction in (12) and
pronoun subject + verb and y + verb constructions in (13).
(12) Variation in different manuscript versions of a Middle Welsh prose text
a. Version I – AIV order
Damwein-aud hagen gythy-auo vn or meib-ion
Happen-pst.3sghoweverpush-vnbyoneof theboy-pl
yllallyr ar yllofft y–r llaur yny
theotherfrom onthegalleryto thegroundso-that
vu
varw
be.pst.3sgdead
‘And it happened that one of the boys pushed the other out of the
gallery onto the ground so that he died’
(Williams 1912: 230, 26)
b. Version II – dummy subject construction
efa damweini-awd digwyd-awun ormeib-iongan
it prthappen-pst.3sgfall-vn
oneofboy.plby
y uwrw o unarallor
llofft
hisstrike.vn byoneotherfrom theloft
y-r llawryny uu
uarw
to-thegroundso-thatbe.pst.3sgdead
‘One of the boys, after being shoved by another, fell from the loft to the
ground and died.’
(Williams 1912: 193)
(13) Variation in the same idiom in a Middle Welsh prose text
a. Diolch-af
inheuyDuw
thank-prs.1sg i.conjtoGod
‘I thank God’
(Goetinck 1976: 39, 24–26)
b.Y diolch-af
yDuw […]
prtthank-prs.1sgtoGod
‘I thank God […]’
c.‘Mia diolch-af
y duw […]
I prtthank-prs.1sgtoGod
‘I thank God’
(Goetinck 1976: 50, 15)
(Goetinck 1976: 178: 30)
4. Increase and variation in use of AIV order in Early Modern Welsh
4.1 Evidence of the increase in use of AIV order
The Early Modern Welsh data is drawn from a corpus of texts spanning the period
from c.1550 to 1772 and comprising nearly 10,000 PDMCs. The composition of the
corpus was designed to reflect as far as practically possible the range of texts available
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
for this period, with both original and translated texts, and is subdivided into the following sections: narrative prose (printed and manuscript), expository prose (printed
and manuscript), sermons (all autograph manuscripts), extracts from the sixteenth
century Bible translations, Slander case records and popular drama in verse. The Corpus also includes five pairs of texts (“parallel texts”), where we have different versions
of the same text or a similar text, and which provide important evidence about the
variation in the use of AIV order:
a. William Salesbury’s translation Mark in the 1567 Welsh New Testament and
­William Morgan’s revised translation in the 1588 Welsh Bible.
b. William Salesbury’s translation of the Psalms in the 1567 Welsh Book of Common
Prayer and William Morgan’s revised translation in 1588.
c. Two Welsh translations of Thomas Gouge’s Principles of the Christian Religion, the
first by William Jones in 1676 as Principlau neu Bennau y Grefydd Ghristianogol,
and the second, of a revised and expanded edition of Gouge’s work, translated by
Charles Edwards in 1679 as Gwyddorion y Grefydd Gristianogol.
d. Two Welsh translations of John Bunyan’s Grace Abounding (Helaethrwydd o Râs)
by John Einnion in 1737 and by Rhys Thomas in 1763;
e. Two different 16th century manuscript versions of the morality play Y Gwr
Kadarn, one (GK I) in Cardiff MS. 2.83 (copied c.1550) and the other (GK II) in
Peniarth 65 (copied c.1600). GK II does not appear to be a copy of GK I but rather
an independent version.
Examples of AIV order from this period are given in (14) and the frequency of the AIV
order in early corpus texts from c.1550 to 1610 is shown in Table 1. The most clearcut evidence of an increase in use of AIV order comes from the 1567 and 1588 Bible
translations, in particular Isaiah and the Psalms, where AIV order occurs in more
than 20% of PDMCs. William Salesbury’s 1567 translation of the Psalms is the earliest
prose text in the corpus which shows a significantly more frequent use of AIV order
compared to Middle Welsh prose; William Morgan’s 1588 translation of the Psalms, in
part a revision of William Salesbury’s 1567 translation, extends further the use of AIV
order, using it twice as a frequently as Salesbury, and, occurring in 41% of PDMCs,
AIV order is the single most frequent word order in the 1588 Psalms (1–20).
(14) Examples of AIV order from corpus texts up to 1610
a.Gofynn-asochy-m pan oe-ddwn
ynGhaerdydd […]
ask-pst.2pl to-1swhenbe-imperf.1sgin Cardiff
‘You asked me when I was in Cardiff […]’
(Letter from William Midleton to Siôn Dafydd Rhys, 1583;
Jones ed. 1956: 52, 17)
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
b.Dyl-em
bawbi gofleid-io
ought-cond.1plall hisembrace-vn
‘We should all embrace him’ (Drama,Y Gwr Kadarn I, Cardiff 2.83: 64)
c.A llef-odd
y naillwrthy llall
Andcall-pst.3sgtheone to theother
‘And one called to the other’
(1588 Bible – Isaiah 6: 3)
Table 1. Frequency of AIV order in early corpus texts c.1550–1610
Text
Date
AIV order
(n0/PDMCs)
AIV order
%
Bible translations
William Morgan, Psalms
1588
114/278
41.0%
William Morgan, Isaiah
1588
56/226
24.8%
William Salesbury, Psalms
1567
55/268
20.5%
William Morgan, Esther
1588
18/191
9.4%
William Morgan, Mark
1588
14/217
6.5%
William Salesbury, Mark
1567
5/219
2.3%
Printed & MS Prose (except Bible
translations)
Maurice Kyffin, Deffyniad y Ffydd
1595
4/170
2.4%
Anon., Gesta Romanorum
late C16th
2/148
1.4%
William Griffith, Sermons
c.1600
2/192
1.0%
Robert Gwyn, Gwssanaeth y Gwyr Newydd
1580
2/198
1.0%
Richard Davies, Epistol at y Cembru
1567
1/130
0.8%
Evan Morgan, Sermons
1610
1/154
0.6%
mid C16th
0/82
0.0%
1583
0/207
0.0%
Anon., Darn o’r Ffestifal
mid C16th
0/213
0.0%
Slander case records (aggregated)
1593-1610
0/45
0.0%
Anon., Y Marchog Crwydrad
Roland Puleston, Llefr yr Eglwys Crhistnogedd
Drama (verse)
Anon., Y Gwr Kadarn I, MS. Cardiff 2.83
mid C16th
33/86
38.4%
Anon., Y Gwr Kadarn II, MS. Peniarth 65
late C16th
30/117
25.6%
Anon., Y Dioddefaint
mid C16th
17/126
13.5%
The evidence from the Corpus, which shows that parts of the sixteenth
c­ entury Bible translations are innovative in having a frequent use of AIV order,
contradicts the traditional but unsubstantiated view in Welsh scholarship
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
(e.g. Fife & King 1991: 144–5; Fowkes 1993 and repeated in Willis 1998: 197) that
the 1588 Bible was conservative in terms of word order. In contrast to the Bible
translations, the frequency of AIV order in the other corpus prose texts from
c.1550–1610 ranges from 0% (three texts) to 2.4%. The use of AIV order in the
five prose corpus texts with only one or two examples of AIV order out of 130 or
more PDMCs is comparable to that of Middle Welsh prose texts where AIV order
is found. The dramatic texts (two manuscript versions of a morality play, Y Gwr
Kadarn, and a passion play, Y Dioddefaint) all show a significantly more frequent
use of AIV order compared to Middle Welsh prose, ranging from 13.5% to 38.4%.
However, since these texts are all verse, the use of AIV order could be interpreted
either as a continuation of Middle Welsh verse usage, where AIV order was common, or as evidence of a general change in the use of AIV order, whether as a result
of parameter resetting or another process.
70%
60%
50%
40%
30%
20%
10%
Ffe c.1550
Ep 1567
M1 1567
S1 1567
Cr c.1575
Gw 1580
LlEG 1583
M2 1588
S2 1588
Es 1588
Is 1588
De 1595
GR 1600
Go 1615
Ed 1629
Ca 1631
Ll 1653
Ys 1675
Prin 1676
Ffy 1677
GG 1679
Yb 1691
Be 1693
GBC 1703
TB 1715
HBA 1721
Pe 1735
He1 1737
Prof 1750
MF 1750
He2 1763
Fa 1772
0%
Graph 1. Frequency of AIV order in corpus prose texts over time
The frequency of AIV order in prose texts over the whole corpus period
from c.1550–1772 is shown in Graph 1, which shows a continuation of the increase
in use of AIV order in prose in the seventeenth and eighteenth centuries in two key
respects:
a. we find more texts with a very frequent use of AIV order, including for the first time
texts where AIV order occurs in over 50% of PDMCs such as Charles Edwards’ Y
Ffydd Ddi-ffuant (1677) with 63% AIV order and James Owen’s T
­ rugaredd a Barn
(1715 2nd ed.) with 54% AIV order;
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
b. over time we find relatively fewer texts with no or only one or two examples of
AIV order. In the period 1610–1700, four out of eleven narrative and expository
prose texts in the corpus have a frequency of AIV order of less than 1% of PDMCs,
including three texts with no examples of AIV order. In the period 1700–1772,
there are no narrative or expository prose texts out of eight in the corpus with
a frequency of AIV order of less than 1% of PDMCs, though two texts have a
­frequency of AIV order of less than 5%.
However, the increase in use of AIV order is far from uniform: the frequency of AIV
order does not increase steadily and consistently over time, rather, we find more of
a zigzag pattern, since there is significant synchronic as well as diachronic variation
throughout the Corpus period.
Table 2. Variation in use of AIV order between contemporary texts of similar genre
Text type
Author 1 Text 1
Date
% AIV Author 2 Text 2
order
0.0%
Oliver
(0/148) Thomas
Expository Robert
prose
Lloyd
Pregeth
am …
edifeirwch
1629
Manuscript Anon.
sermons
MS., Cardiff
2.226
c.1660 0.9%
John
(1/107) Piers
Expository Rondl
prose
Davies
Profiad yr
Ysprydion
1675
Date
% AIV
order
Carwr y
Cymru
1630
16.7%
(21/126)
MS., NLW
12205A
mid 10.1%
C17th (18/179)
0.0%
Charles Ffydd Ddi(0/243) Edwards ffuant
Early 3.1%
Anon.
Manuscript Samuel MS.,
C18th (8/258)
sermons
Williams Cwrtmawr
253B, 121-15
1677
63.1%
(246/390)
MS., Bangor
362, 5
1717
22.0%
(26/118)
Interludes
William
Roberts
Ffrewyll y
1745
Methodistiaid
4.4%
Richard
(8/183) Parry
Cyndrigolion
y Deyrnas
Hon
1737
25.6%
(71/277)
Narrative
prose
(1st
person)
Henry
Lloyd
Profiad … o
Nefoedd ag
Uffern
4.4%
Rhys
(6/138) Thomas
Helaethrwydd 1763
o ras
29.8%
(42/141)
1750
Not only do we find significant variation in the frequency of use of AIV order
between near contemporary texts, but we also find significant variation between contemporary or near contemporary texts of a similar text type, as illustrated in Table 2.
The evidence from the parallel texts, summarised in Table 3, confirms this pattern of
variation: not only do we have variation in the frequency of AIV order between similar
text types, but between different versions of the same text.
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
Table 3. Variation in use of AIV order between “parallel texts”
Text
Text type
Author 1
Date
Psalms
Bible (trans. <
Hebrew)
William
Salesbury
1567
Mark
Bible (trans. <
Greek)
William
Salesbury
1567
Y Gwr
Kadarn
Morality play
(verse)
(original
Welsh MS.)
Principles of
the
Christian
Religion,
Thomas
Gouge
Expository
prose
(trans. <
English)
Grace
Abounding,
John
Bunyan
% AIV
order
Author 2
Date
% AIV
order
20.5% William
(55/268) Morgan
1588
41%
(114/278)
2.3%
(5/219)
1588
6.5%
(14/217)
Anon., MS. c. 1600 25.6% Anon.,
Peniarth
(30/117) MS.
65
Cardiff
2.83
mid
C16th
38.4%
(33/86)
William
Jones
1676
0.0%
(0/138)
Charles
Edwards
1679
40.0%
(80/200)
Narrative
John
prose (trans. < Einnion
English)
1737
9.4%
Rhys
(13/138) Thomas
1763
29.8%
(42/141)
William
Morgan
4.2 A discrete or gradual syntactic change?
Now, does this evidence of the increase in use of AIV order in the sixteenth to the
eighteenth centuries represent an instance of gradual or discrete syntactic change? If
we view the phenomenon as one of a change in frequency of use of AIV order, then it
would appear prima facie to be a case of gradual syntactic change, since it is a development spread over a protracted period – of at least the two centuries covered by the
Corpus. Willis (1998: 196–7, 261), however, views the change as discrete and sudden,
positing an abrupt change in the grammaticality of AIV order (unmarked VSO in his
terminology) underlying the observable increase in use of AIV order in Early Modern
Welsh texts.
It is evident from the data presented in Table 1, that there is no across the board
increase in the use of AIV order in Early Modern Welsh prose, in fact, apart from
the Bible translations, AIV order is uncommon in sixteenth century prose texts. If
we posit a “sudden” innovation of “general unmarked VSO” in the sixteenth century,
how are we to explain the usage of the Corpus texts where AIV order is rare or unattested, not only in the sixteenth century when Willis posits the parametric change, but
also, for example, in texts from the second half of the seventeenth century by writers
such Rondl Davies (1675), William Jones (1676) and John Griffith (1680s)? Logically,
there are two options: (a) we could argue that sixteenth and seventeenth century texts
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
­ ithout AIV order reflect the parameter change but happen not to have examples of
w
AIV order, and alternatively, (b) the “diffusion argument”, that texts without AIV order
do not reflect the parametric change, on the assumption that there was a progressive
actuation of the parameter change in the speech community and that the authors of
these texts had not yet reset the V2 parameter.
Taking these two options in turn, one argument advanced by Willis (1998: 197)
to account for texts with no or rare AIV order, is that such texts with no or infrequent
AIV order may reflect a conservative (V2) literary style. While it is important to recognise the potential impact of literary tradition when interpreting textual evidence, it
seems methodologically questionable to disregard as unrepresentative, at least without
empirical justification in specific cases, the evidence of certain literary texts which
appears not to conform to theoretical predictions, while admitting that of others which
does. Indeed, the clearest evidence of innovation in the use of AIV order is in literary
prose texts, both in that they are the earliest texts which show a significant increase in
the use of AIV order compared to Middle Welsh prose and in that they are the only
corpus texts where AIV order is used as the dominant word order (in 40% or more of
PDMCs); literary prose texts also show the greatest range of variation in the frequency
of use of AIV order (from 0% to 63% of PDMCs). On the other hand, the most popular
text-type, the Slander Case records (Suggett 1983), albeit fragmentary in nature, since
they do not represent continuous texts and typically comprise no more than one clause
at a time, show the least frequent use of AIV order with only one instance of AIV order
out of 418 PDMCs in the corpus. The verse interludes (c.1660–1745), the next most
popular text-type, also show on the whole less frequent use of AIV order than literary
prose texts (ranging from 4% to 26% of PDMCs).
In principle, since the P&P approach posits a change in the grammaticality of AIV
order – ungrammatical before the parameter resetting and grammatical after – it is
also possible to argue that AIV order being grammatical means that the construction
was available for use but not that it would necessarily be used. So, on this basis, one
could argue that although the authors of texts with no instances of AIV order did not
use the construction in these texts, they could have done (indeed may have actually
done so in other non-extant texts or in texts not included in the corpus). However,
it would seem to be difficult within the P&P approach to reconcile such an argument
with the claim that, after the resetting of the V2 parameter in the sixteenth century,
there was “general unmarked VSO” and that Welsh was “by this stage a VSO language”,
albeit with more frequent SVO order than in Modern Welsh (Willis 1998: 205).
According to the “diffusion argument”, which Willis (1998: 48) outlines in general theoretical terms but does not attempt to apply to the Early Modern Welsh data,
a parametric change could spread progressively through a speech community. Thus
from the perspective of an individual speaker, the parameter resetting is discrete, but
from the perspective of the speech community, the cumulative effect resulting from
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
many discrete individual parameter resettings may be gradual. However, there would
appear to be two problems in applying such a model to the variation we see in the use
of AIV order in Early Modern Welsh. First, while such a model could offer a plausible
explanation of variation between sixteenth century texts which show AIV order and
those which do not, it would seem to be less plausible for similar variation evident in
the second half of the seventeenth century, more than a century after the proposed
commencement of the parametric change. Second, such a model of the spread of parametric change can only account for binary variation in the grammaticality of AIV
order, i.e. between its occurrence (as an “unmarked” construction) or not, and not
variation in its frequency of use. The observable variation in data appears to suggest a
continuum in the frequency of use of AIV order, at one end of which is zero use of AIV
order and at the other almost systematic use of the construction.
The key difference between the P&P analysis of the development of AIV order
as a discrete syntactic change and the CxG analysis of the phenomenon as a gradual
change hinges crucially on how the change under investigation is defined. In the P&P
approach, the change is defined as a discrete and abrupt change in the grammaticality of AIV order in the sixteenth century and subsequent variation and change in its
frequency of use has to be treated as a separate phenomenon. In the CxG approach,
the change is one of a gradual increase frequency of use of AIV over an extended
period and the variation in the frequency of its use is understood as an integral part
of the change.
4.3 Competition between AIV order and other constructions
AIV order in Welsh went from being a marginal construction in Middle Welsh prose
to the dominant word order in several Early Modern Welsh prose texts (up to 63% of
PDMCs in Charles Edwards’ Y Ffydd Ddi-ffuant). A key factor which enabled the construction’s dramatic expansion was that it was a highly schematic construction which
could appear in a wide range of syntactic environments, for example when the subject
of the verb was a personal pronoun or noun, as well as in impersonal constructions;
there was no other single Welsh construction which could appear in all these syntactic environments. We have grouped these syntactic environments into two main
types – those where there is a (grammatical or logical) pronominal subject (“PRO
clauses”) and those where there is a (grammatical or logical) nominal subject (“NOM
clauses”) – and calculated the relative word order frequencies in such in PRO and
NOM clauses excluding PDMCs with clause-initial adverb-phrases, where by definition AIV order cannot occur. Table 4 summarises the patterns of variation we find
in PRO and NOM clause in a sample of nine prose corpus texts from the fifty year
period 1653–1703, while Table 5 summarises the variation in the prose parallel texts.
In PRO clauses, there is essentially a two-way pattern of variation between AIV order
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
Table 4. Word order in “PRO” and “NOM” clauses: prose Corpus texts 1675–1703
1691 1693
1700 1703
1653 1675–6
1675
1677 1683–5
John William James Samuel
Ellis
Morgan Anon. Rondl Charles
Llwyd NLW 3 Davies Edwards Griffith Thomas Owen Williams Wynne
B 95
Yb
Be Cw 253 GBC
LlTA Serm. Ys. Exp.
Ffy
Exp. Exp.
Serm. Narr.
Exp.
Narr./ Serm.
Exp.
PRO
Clauses (n0)
222
58
53
111
72
% AIV
0.9%
10.3%
0.0%
92.8%
% PSV
110
57
87
82
0.0%
30.0% 54.4%
2.3%
59.8%
90.1%
74.1%
84.9%
1.8%
84.7%
61.8% 43.9%
86.2%
36.6%
% Dummy
7.2%
10.3%
1.9%
0.0%
11.1%
3.6% 1.8%
5.7%
1.2%
% Other
1.8%
5.2%
13.2%
5.4%
4.2%
4.5% 0.0%
5.7%
2.4%
93
62
115
162
57
74
29
NOM
Clauses (n0)
% AIV
102
93
0.0%
4.8%
0.0%
79.6%
0.0%
29.4% 64.5%
6.8%
58.6%
% Dummy
55.9%
11.3%
12.2%
0.0%
77.2%
56.9% 9.7%
31.1%
3.4%
% NSV
43.0%
79.0%
81.7%
19.8%
19.3%
11.8% 22.6%
59.5%
37.9%
% Other
1.1%
4.8%
6.1%
0.6%
3.5%
2.0% 3.2%
2.7%
0.0%
Table 5. Word order in “PRO” and “NOM” clauses: prose “parallel” texts 1567–1763
1567
1588
1567
1588
1676
1679
1737
1763
William William William William William Charles
John
Rhys
Salesbury Morgan Salesbury Morgan
Jones Edwards Einnion Thomas
M1
M2
S1
P2
Prin.
GG
H1
H2
PRO
Clauses (n0)
101
98
135
154
43
53
51
59
% AIV
3.0%
1.0%
37.8%
67.5%
0.0%
67.9%
23.5%
67.8%
% PSV
71.3%
87.8%
41.5%
25.3%
88.4%
30.2%
60.8%
22.0%
0.0%
0.0%
0.0%
0.0%
7.0%
0.0%
5.9%
1.7%
25.7%
11.2%
20.7%
7.1%
4.7%
1.9%
9.8%
8.5%
60
62
79
77
43
81
19
22
3.3%
16.1%
5.1%
10.4%
0.0%
46.9%
5.3%
9.1%
% Dummy
10.0%
3.2%
3.8%
0.0%
48.8%
2.5%
21.1%
13.6%
% NSV
66.7%
77.4%
84.8%
88.3%
44.2%
46.9%
68.4%
77.3%
% Other
20.0%
3.2%
6.3%
1.3%
7.0%
3.7%
5.3%
0.0%
% Dummy
% Other
NOM
Clauses (n0)
% AIV
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
and pronominal subject + verb order, which together account for 84%–98% of word
orders in Tables 4–6. In NOM clauses, there is essentially a three-way pattern of variation between AIV order, the dummy subject construction and nominal subject + verb
order, which together account for 94%–100% of all word orders. The variation in word
order both clause types spans almost the full range possible in the prose texts. In PRO
clauses we go from 0% AIV order and over 90% pronominal subject + verb order at
one extreme to over 90% AIV order and only 2% pronominal subject + verb order at
the other extreme; and in NOM clauses we go from 0% AIV order and 77% dummy
subject + verb order at one extreme to 79% AIV order and 0% dummy subject at the
other extreme.
What seems to have happened is that AIV order appears to have been perceived
by Early Modern Welsh writers as functionally equivalent to and interchangeable with
pronominal subject + verb order and dummy subject + verb order. This meant that
as the use of AIV order expanded in Early Modern Welsh prose, it could be used
­everywhere where the pronominal subject and dummy subject + verb orders were
used, but at the same time need not be used at all, since pronominal subject + verb and
dummy subject + verb remained well established and productive constructions, and, if
we can extrapolate from the evidence of the Slander case records and interludes, may
have remained the dominant constructions in spoken discourse.
Evidence of the potential interchangeability of AIV order, on the one hand, and pronominal subject and dummy subject + verb orders, on the other hand, can be found as
early as Middle Welsh and in (12) and (13) a couple of examples were noted from Middle
Welsh prose. Such examples are rare in Middle Welsh prose, as AIV order itself was rare,
though at the same time AIV order was common in Middle Welsh poetry and this contrast between prose and poetic usage, as well as the use of AIV order side by side with
the pronominal subject and dummy subject + verb orders in poetry could have provided
further evidence for a perceived interchangeability between AIV order and the pronominal subject and dummy subject + verb orders. More substantial evidence of the potential
interchangeability can be found in the parallel texts of the Early Modern Welsh corpus,
spanning the full corpus period from the mid sixteenth to the mid ­eighteenth century,
in prose and verse texts. In four of the parallel texts – the two versions of the Psalms, of
Gwr Kadarn, of Helaethrwydd o Râs, and of t­ he Principles of the Christian ­Religion – we
have evidence of significant variation between AIV order and pronominal subject + verb
order (cf. Tables 5 and 6). We also have a number of instances of directly comparable
PDMCs, with AIV order in one version and ­pronominal subject+verb and/or dummy
subject+verb order in the other, exemplified in (15–18).
(15) Psalms 5: 6
a.William Salesbury 1567 – pronominal subject + verb order
Ti
ddestryw-y
y rei y ddywed-antgelwydd
You.2sgdestroy-prs.2sgtheonesrelsay-prs.3pl obj/lie
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
b.William Morgan 1588 – AIV order
Difeth-i
y rhai a ddywed-antgelwydd
Destroy-prs.2sgtheonesrelsay-3pl.prs obj/lie
‘You destroy those who tell lies’
(16) Helaethrwydd o Ras
a.John Einnion 1737, 18 – pronominal subject + verb order
mia ddichwel-ais yn ffromwylltatfy
I prtreturn-pst.1sg advbrazen tomy
chwariaethdrachefn
play
back
‘I returned brazenly to my play’
b.Rhys Thomas 1763, 8 – AIV order
dychwel-ais atfy Ynfydrwydddrachefn
return-pst.sgtomyfolly
back
‘I returned to my folly’
(17) Mark 1.40
a.William Salesbury 1567 – dummy subject + verb order
Ac e ddaeth
at-aw ddynclavrllyt
Anditcome.pst.3sgto-3sg.mman leprous
b.William Morgan 1588 – AIV order
Adaeth
att-o ef[vn] gwahan-glwyfus
Andcome.pst.3sgto-3sg.mhe[one]leprous
‘And a leper came up to him’
(18) Helaethrwydd o Ras
a.John Einnion 1737, 1 – Dummy subject
feryng-odd
boddi Dduwi osod yneu
itplease-pst.3sgwill toGod toset-vnin their
Calonn-aunhwyim dodi i mewnYscol
heart-pl3pl to-myput-vninto school
i ddysg-u darllena Scrifenn-u
tolearn-vnread.vn andwrite-vn
b.Rhys Thomas 1763, 8 – AIV order
rhyngodd
boddi’r
ARGLWYDDi roi
yn
please-pst.3sgwill to theLord
toput.vnin
eu Calonnaui’m dodi mewnYscol i ddyscu
theirhearts to-myput-vninto schooltolearn-vn
darllena ’scrifenn-u
read-vnandwrite-vn
‘It pleased God to set it in their hearts to put me in school to learn to
read and write’
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
4.4 Syntactic motivation for the use of AIV order
4.4.1 Link between the “loss” of the preverbal particle y and rise of AIV order
Evans (1968) first noted that the preverbal particle y used in Adverb + y + verb constructions began to be lost in Early Modern Welsh and hypothesized a link between
this loss and the increase in use of AIV order. Willis (1998: 188) described the loss
of the preverbal particle y “as the single most important development precipitating
the breakdown of verb-second”; the loss of y is attributed to “phonological erosion”
(Willis 1998: 204) and it is claimed that the “omission of y after adverbs was certainly
widespread by the second half of the sixteenth century” (Willis 1998: 188). In the P&P
approach, the loss of y is one of several leading changes posited which reduced the
evidence in the trigger material supporting the acquisition of V2; the other changes
were the loss of the other preverbal article a used when subjects and direct objects
were fronted (cf. (1–4) above), a decline in object topicalisation and the extension of
the dummy subject construction. The link posited between the loss of y in the P&P
approach is therefore an indirect and abstract one: the loss of y contributed to the
reduction of evidence for V2 in the trigger material, the V2 parameter was reset to
negative and, as result of the resetting of the V2 parameter, unmarked VSO became
grammatical. Insofar as the parametric change is discrete, the relationship between the
loss of y and the re-emergence of AIV order is also necessarily discrete.
4.4.2 Evidence from the Early Modern Welsh corpus
There is no quantitative empirical analysis in either Evans (1968) or Willis (1998) to
support their account of the loss of y or the link posited between the loss of y and
rise of AIV order in Early Modern Welsh prose. As part of the corpus study, we analysed the frequency of the various adverb-initial constructions – Adverb + y + verb,
Adverb + Verb (i.e. the evidence of the loss of y) and Adverb + XP + verb order – and
examined whether there was empirical evidence of a link between Adverb + verb order
and AIV order in the form of statistical correlation between the frequency of use of the
two ­constructions in individual texts.
It is apparent from the corpus study that it is over simplistic to speak of a loss of y in
the sixteenth century; it would be more appropriate to speak of a decline in the use of y
during the Early Modern Welsh period, since side by side an increase in use of Adverb +
verb constructions, there is an increase in frequency of Adverb + XP + verb constructions (e.g. Adverb + pronoun subject/dummy subject/nominal subject + verb orders) at
the expense of Adverb + y + verb order. Adverb + XP + verb constructions were already
common in Middle Welsh prose. We find variation between Adverb + Verb order and
Adverb + XP + verb order, just like we do between AIV order and pronominal subject +
verb and dummy subject + verb orders. The decline in use of y also appears to be gradual
over the corpus period, as is shown in Graph 2, which plots the percentage of Adverb +
y + Verb order out of total adverb-initial constructions over time in prose corpus texts.
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
100%
90%
80%
70%
60%
Ffe c.1550
Ep 1567
M1 1567
S1 1567
Cr c.1575
Gw 1580
LlEG 1583
M2 1588
S2 1588
Es 1588
Is 1588
De 1595
GR 1600
Go 1615
Ed 1629
Ca 1631
LlTA 1653
Ys 1675
Prin 1676
Ffy 1677
GG 1679
Yb 1691
Be 1693
GBC 1703
TB 1687
HBA 1721
Pe 1735
H1 1737
Prof 1750
MF 1750
H2 1763
Fa 1772
50%
40%
30%
20%
10%
0%
Graph 2. Frequency of Adverb + y + Verb order in prose texts
Moreover, there is no evidence of a widespread loss of y in the sixteenth ­century;
indeed, as shown in Graph 3, Adverb + Verb order is relatively uncommon in ­sixteenth
century prose. The corpus study also shows that there is evidence of a correlation in
frequency between Adverb + Verb order and AIV order. Graph 3 plots for prose texts
the frequency over time of AIV order as a percentage of PDMCs without clause-initial
adverbial phrases against the frequency of Adverb + Verb order as a percentage of
Adverb-initial clauses. Apart from the significant exception of the sixteenth century
Bible translations, there a consistent correlation in both the prose and verse of between
the use of Adverb + verb order and AIV order: not only do the two constructions
co-occur in the same texts, but they also co-occur with similar frequency.
100%
90%
80%
70%
60%
50%
Adverb + Verb
order
40%
AIV order
30%
20%
10%
Ffe c.1550
Ep 1567
M1 1567
S1 1567
Cr c.1575
Gw 1580
LlEG 1583
M2 1588
S2 1588
Es 1588
Is 1588
De 1595
GR 1600
Go 1615
Ed 1629
Ca 1631
LlTA 1653
Ys 1675
Prin 1676
Ffy 1677
GG 1679
Yb 1691
Be 1693
GBC 1703
TB 1687
HBA 1721
Pe 1735
H1 1737
Prof 1750
MF 1750
H2 1763
Fa 1772
0%
Graph 3. Correlation in frequency of use between AIV order and Adverb + Verb order
The correlation between Adverb + verb order and AIV order is part of a wider
pattern of parallelism in word order patterns between adverb-initial clauses and
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
clauses without clause-initial adverbial phrases, since we also find correlations of
­co-occurrence and frequency between dummy subject and adverb + dummy subject
constructions (Graph 4) as well as between pronominal subject + verb and adverb +
pronominal subject + verb constructions (Graph 5). In other words, the variation in
word order in adverb-initial PDMCs mirrors that of PDMCs without clause-initial
adverbial phrases.
35%
30%
25%
Adverb + Dummy
subject + Verb
20%
15%
Dummy subject + Verb
10%
5%
Ffe c.1550
Ep 1567
M1 1567
S1 1567
Cr c.1575
Gw 1580
LlEG 1583
M2 1588
S2 1588
Es 1588
Is 1588
De 1595
GR 1600
Go 1615
Ed 1629
Ca 1631
LlTA 1653
Ys 1675
Prin 1676
Ffy 1677
GG 1679
Yb 1691
Be 1693
GBC 1703
TB 1687
HBA 1721
Pe 1735
H1 1737
Prof 1750
MF 1750
H2 1763
Fa 1772
0%
Graph 4. Correlation in frequency of use between Dummy subject + Verb an Adverb +
Dummy subject + Verb constructions
80%
70%
60%
50%
40%
Adverb + Pron
subject + Verb
30%
Pronoun subject + Verb
20%
10%
Ffe c.1550
Ep 1567
M1 1567
S1 1567
Cr c.1575
Gw 1580
LlEG 1583
M2 1588
S2 1588
Es 1588
Is 1588
De 1595
GR 1600
Go 1615
Ed 1629
Ca 1631
LlTA 1653
Ys 1675
Prin 1676
Ffy 1677
GG 1679
Yb 1691
Be 1693
GBC 1703
TB 1715
HBA 1721
Pe 1735
H1 1737
Prof 1750
MF 1750
H2 1763
Fa 1772
0%
Graph 5. Correlation in the frequency of use of Pronoun subject + Verb and Adverb +
­Pronoun subject + Verb constructions
4.4.3 Significance of the correlation between Adverb + verb order and AIV order
The existence of a correlation between the use and frequency of Adverb + verb
order and that of AIV order suggests that the (partial) loss of y in Adverb + y +
Verb order contributed to the increase in use of AIV order in Middle Welsh, but
the mechanism proposed by Willis (1998) does not fit the data from the corpus in
three respects.
First, rather than indirect and abstract link between Adverb + Verb order and AIV
order mediated by a parametric change, there appears to be direct and transparent link
between the use of the two constructions, which can be analysed in a CxG framework
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
in terms of motivation. There seems to have been a perceived p
­ arallelism between
PDMCs with a clause-initial adverbial phrase and those without; any c­ onstruction
which could follow a clause-initial adverbial phrase could also come in absolute clauseinitial position and vice-versa, e.g. Adverb + XP + Verb or Adverb, XP + Verb constructions motivated XP + Verb constructions and vice versa, and similarly Adverb +
Verb or Adverb, Verb constructions motivated AIV order, as indicated in the schema
in Table 6. Returning to the question of the rarity of AIV order in Middle Welsh prose,
the use of AIV order in Middle Welsh prose could not be motivated by Adverb + Verb
order, because Adverb + Verb order hardly ever occurred – it was even rarer in Middle
Welsh prose than AIV order.
Table 6. Parallelism between Adverb-initial and non-adverb-initial constructions
#y + Verb…#
⇐ motivates ⇒
#Adverb + y + Verb…#
#Pronominal Subject + a + Verb#
⇐ motivates ⇒
#Adverb, Dummy Subject + a + Verb…#
#Subject + a + Verb…#
⇐ motivates ⇒
#Adverb, Subject + a + Verb…#
#Object + a + Verb…#
⇐ motivates ⇒
#Adverb, Object + a + Verb…#
#Verb noun object + a + Aux…#
⇐ motivates ⇒
#Adverb,Verb noun object + a + Aux…#
#Verb…# (AIV order)
⇐ motivates ⇒
#Adverb, Verb…#
The basis for this motivational relationship is the formal similarity between the
respective pairs of constructions and the fact the clause-initial adverbial phrase could
be analysed as a clause connector, separate from the verbal phrase, so that the ­following
construction – XP + verb or verb – could be perceived as clause-initial.
Second, the correlation between Adverb + verb order and AIV order (as well as
between Adverb + XP + verb and XP + verb orders) and the motivational relationship between them holds throughout the corpus period, not just at the initiation of
the change, when the increase in the use of AIV order in Early Modern Welsh prose
is first observed. This suggests that the motivational relationship is a continuous one,
rather than a discrete one occurring at a particular point in time (e.g. the parameter
resetting), and is therefore more compatible with a gradual mechanism of syntactic
change in the CxG approach rather than the discrete mechanism posited in the P&P
approach.
Third, while the decline in use of the preverbal particle y and associated rise of the
Adverb + verb construction appears to have been a motivating factor in the increase in
use of AIV order in Early Modern Welsh prose, we cannot say that the decline in the
use of y necessarily caused the increase in use of AIV order, since there were alternative Adverb + XP + verb constructions available as well as Adverb + verb order, just
as there were alternative XP + verb constructions which could be used instead of AIV
order; while some writers used Adverb + Verb and AIV orders, other contemporary
writers preferred Adverb + XP + verb and XP + verb orders.
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
4.5 Stylistic motivations for the use of AIV order
4.5.1 Potential influence of poetic style in the prose Bible translations
The sixteenth century Bible translations represent an exception to the correlation
between the use of Adverb + verb order and AIV order, since there are no examples of
Adverb + verb in the four extracts from William Morgan’s 1588 Bible translations in the
corpus (Mark, Esther, Isaiah and the Psalms) while AIV order is common (cf. Table 1),
and to a lesser extent in William Salesbury’s 1567 Psalms where Adverb + verb and
AIV order co-occur but AIV order is used disproportionately more frequently than
Adverb + verb order. The use of AIV order is significantly higher for both writers in
the parts of the Bible translated from poetry (the Psalms and Isaiah) than in the parts
of the Bible translated from prose (Mark and Esther). It is possible that the frequent
use of AIV order by Salesbury and Morgan in the Psalms in particular was motivated
by a desire to recreate a poetic effect in a Welsh prose translation of the Hebrew poetry
by incorporating a linguistic feature, AIV order, which, by virtue of being common in
contemporary and earlier Welsh poetry but rare in prose, could have been associated
with poetry.
The motivation for poetic stylistic influence was potentially especially strong for
the Psalms, since the Psalms were not only a poetic text in relation to the source
culture, but also in respect of the contemporary target culture, since the Psalms were
sung in worship and poetry was the medium par excellence of song. The fact that the
Psalms were translated from Biblical Hebrew, which has a predominantly verb-initial
word order, could have provided a further motivation for the use of AIV order, not
simply because the use of AIV order permitted a closer rendering of the Hebrew VSO
sentence structure in general, but also because it gave the Welsh translators more
scope and flexibility in recreating the parallelism characteristic of the Biblical Hebrew
poetry, of which word order was a key element. In addition, William Salesbury’s use
of AIV order in the Psalms could also have been influenced by the use of AIV order
in the Middle Welsh verse translations of the Psalms in Gwasanaeth Meir; Mathias
(1970) and Thomas (1988) have shown that William Salesbury knew of and used
Gwasanaeth Meir.
How significant might the use of AIV order in poetry and its potential influence
on prose have been in the overall increase in use of AIV order in Early Modern Welsh
prose? The fact that the Bible translations are the earliest corpus text with a frequent
use of AIV order means that the influence of poetic style on prose could have been
a significant factor in the initiation of the syntactic change, complementary to the
gradual loss of the preverbal particle y. Further, the Bible translations were particularly
influential texts and a frequent use of AIV order in the Bible could have provided a
motivation for later writers to use AIV order; indeed, it is possible that, without the
frequent use of AIV order in parts of the Bible translations, potentially motivated by
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
the use of AIV order in contemporary and earlier poetry, AIV order may never have
become widely used in Early Modern Welsh prose. The fact AIV order was well established in poetry may have also provided a more general and indirect motivation for the
use of AIV order simply by providing writers with an additional source of exposure to
the construction. More specifically the association of the AIV order with poetry may
have contributed to a perception of AIV order as a prestigious stylistic feature, and
some writers may have chosen to use the construction more frequently for this reason.
4.5.2 Stylistic nature of the variation
The variation in the use of AIV order does not appear to correlate with any pre-­
determined social or textual parameters such as social class, dialect, register or genre
in a traditional sociolinguistic variationist (e.g. Labovian) paradigm, since we find significant variation between texts of a very similar nature (in terms of genre, content and
discourse type) and written by authors (all men) of a similar social status, geographical
origin and educational background. Rather, the main parameter of variation appears
to stylistic choice by individual writers. The gradual loss of the preverbal particle y is
likely to have increased the incidence of AIV order in discourse and this, combined
with existing synchronic variation in the use of AIV order between prose and poetry
and the perceived the interchangeability of AIV order with dummy subject and pronominal subject + verb order, provided Early Modern Welsh writers with a linguistic
resource which they used in very different ways. Some Early Modern Welsh prose writers such as Charles Edwards and James Owen systematically extended the use of AIV
order and developed a new prose style with AIV order as the dominant word order (in
over 50% of PDMCs). Other prose writers, such as Morgan Llwyd, Rondl Davies, William Jones and John Griffith, appear to have avoided the construction. Other writers
still, such as Thomas Williams and Simon Thomas, show an intermediary pattern of
usage (22%-24% of PDMCs); Thomas Williams, for example, uses AIV order and the
dummy subject construction interchangeably in some idioms, while Simon Thomas
prefers AIV order in some idioms but the dummy subject construction in others. William Salesbury and William Morgan, on the other hand, show style-shifting in the use
of AIV order in different parts of the Bible translations, with frequent AIV order in the
some texts (Psalms, Isaiah) and less frequent in others (Mark, Esther).1 The theoretical
framework which seems best able to accommodate the pattern of individual stylistic
variation we see here is Coupland’s concept of styling,2 where speakers “can frame the
. See Currie (2000) for a more detailed discussion of the stylistic variation.
. Hollmann and Siewierska (2011) use Coupland’s concept of styling within a theoretically similar framework to Construction Grammar – Cognitive Sociolinguistics – to explain
variation in definite article reduction in contemporary English dialects.
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
linguistic resources available to them in creative ways, making new meanings from old
meanings” (2007: 84).
The pattern of synchronic variation and diachronic change in the use of AIV
order is characteristically contingent. By contingent, we understand language variation and change which is neither a necessary nor predictable; it cannot be described as
necessarily following from a given set of causes, nor can its subsequent course be said
to follow a predictable pattern. Thus we can posit only potential motivations for the
increase in use of AIV order, but not causes, since they do not necessarily lead to an
increase use of the construction in individual writers. The key Construction Grammar
concept of motivation, which we have used here to analyse both the syntactic and sociolinguistic/stylistic aspects of the synchronic variation in the use of AIV order in Early
Modern Welsh, is therefore also applicable diachronically in trying to understand the
contingent nature of the language change evidenced in this case study.3
5. C
onclusion
We have compared a diachronic Construction Grammar analysis of the development
of Absolute-initial verb order in Early Modern Welsh with a Principles and Parameters analysis and showed that a Construction Grammar approach has a very different
perspective on the question of the gradual vs. discrete nature of syntactic change as
well as on the treatment of syntactic variation. The Principles and Parameters conception of grammars and grammatical categories as discrete entails a discrete conception
of ­syntactic change, which has to be reconciled with the gradual patterns of change
observable in historical data. The paradox of a mismatch between the discrete theory,
on the one hand, and gradual patterns of change in the data, on the other hand, need
not arise in a diachronic Construction Grammar approach, as Construction Grammar
permits gradient categories as well as a gradient conception of grammaticality.
In the corpus-based case-study we showed that the pattern of the increase in use
of AIV order in Early Modern Welsh appears to be gradual in that it happens progressively over the two century corpus period from the mid sixteenth to the mid eighteenth
century and in that there is significant variation in the use of AIV order throughout
this period. We argued that a Construction Grammar-based analysis of this change
in Welsh order as a gradual increase in the frequency of use of AIV order seems to fit
the historical data better than the Principles and Parameters analysis of the change as
an abrupt and discrete change in the grammaticality of AIV order (absolute V1) as a
. Cf. the use of “motivation” in Hopper and Traugott (1993: 63) in the sense of an enabling
or potential factor – as opposed to an absolute factor – in language change.
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
result of the resetting of a V2 parameter. We further showed that it is possible to posit
a gradual mechanism of syntactic change within a Construction Grammar framework
to account for the gradual pattern of change in usage observable in the historical data,
using the Construction Grammar concept of motivation. We argued that there were
both syntactic factors (the perceived parallelism between constructions with and without clause-initial adverbial or prepositional phrases, i.e. between Adverb + Verb and
AIV order, on the one hand, and Adverb + XP + Verb and XP + Verb order on the
other hand) and sociolinguistic/stylistic factors (i.e. the perceived interchangeability
with and competition between AIV order and dummy and personal pronoun subject +
Verb constructions, the association of AIV order with poetic style and the fact that the
dummy and personal pronoun subject + verb constructions were well-established in
written prose and probably also spoken discourse) which motivated the use – as well
as the avoidance – of AIV order. We adduced empirical evidence of these factors and
further showed that they appear to have operated throughout the corpus period.
The apparent mismatch between discrete theoretical models and gradual historical data in Principles and Parameters approaches potentially also makes modelling
and explaining sociolinguistic syntactic variation more problematic. In this particular
case, while the Principles and Parameters approach of Willis (1998) does not seek to
address the variation in the use of AIV order in Early Modern Welsh, the analysis
and understanding of the variation is central to the Construction Grammar approach.
Sociolinguistic/stylistic and syntactic motivations for the use of AIV order form an
integral part of the description of the synchronic variation in Early Modern Welsh
word order in the Construction Grammar approach; moreover, the synchronic variation, is understood not just as side-effect of diachronic change in the use of AIV order,
but also a motivating and enabling factor for the change in the first place.
Focus on the dynamic interface between synchrony and diachrony
This article explores using a case study of word order change in Early Modern Welsh
the interrelationship between synchrony and diachrony from several different angles:
a. How synchronic variation may reflect earlier diachronic change (in the case the
variation between rare AIV order in Middle Welsh prose and common AIV order
in Middle Welsh poetry) or ongoing diachronic change (in the case of the variation
in the use of AIV order between different Early Modern Welsh prose texts). From
a methodological perspective, such synchronic variation enables us to identify
historical syntactic change.
b. How synchronic variation can provide the resources and motivation for subsequent
diachronic change. The variation in the use of AIV order between Middle Welsh
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
prose and poetry and the potential association of the construction with poetic
style provided a resource which later Early Modern Welsh writers could use in
extending the use of AIV order in particular in the prose translations of parts of
the Old Testament, such as the Psalms and Isaiah, which were translated from
Hebrew poetry.
c. How synchronic variation can interact with the mechanisms of diachronic change.
The progressive loss of the preverbal particle the preverbal particle y in Adverb +
y + verb and in y + verb constructions seems to have been a significant factor in
increase in use of AIV order in Early Modern Welsh, however, on its own it cannot be said to have necessarily given rise to the increase in use of AIV order since
alternative, well-entrenched and competing constructions (the dummy subject
and pronominal subject + verb constructions) were also available. The frequent
use of AIV order in parts of the influential sixteenth century Bible translations,
possibly inspired by Welsh poetic style, also seems to have been a significant factor in the propagation of the change, since the Bible were the earliest innovative
Early Modern prose texts in having frequent AIV order did so before evidence of
a widespread loss of the preverbal particle y.
d. The significance of synchronic and diachronic variation in terms of the debate on the
gradualness or discreteness/abruptness of syntactic change: the existence of variation over a protracted period (at least two centuries in the present case study) is a
key piece evidence in favour of the gradualness of syntactic change.
e. The potential of Construction Grammar as framework to analyse and model
­synchronic syntactic variation and gradual diachronic change. The broad, semantic
and usage-based nature of Construction Grammar enables it to provide an integrated analysis of language variation, incorporate a wide range of factors (e.g. syntactic, prosodic, semantic, discourse pragmatic, lexical, idiomatic, sociolinguistic
or stylistic), which may motivate the use of particular linguistic constructions or
variants. Further the fact that Construction Grammar allows clines and continua
enables it to model gradual change, whereas the exclusively discrete categories
of Principles and Parameters frameworks impose a discrete analysis of syntactic
change, irrespective of the observable patterns of change in historical data.
Abbreviations
aiv
conj
Dummy
Exp.
Narr.
Absolute initial-verb order
Conjunctive pronoun
Dummy Subject + Verb order
Expository prose corpus text
Narrative prose corpus text
© 2013. John Benjamins Publishing Company
All rights reserved
nom
nsv
pdmc
pro
prt
psv
Serm.
vn
The history of a verb-initial construction in Welsh 
Clauses (PDMCs) with a nominal subject
Nominal Subject + Verb order
Positive Declarative Main Clause
PDMCs with a personal pronoun grammatical or logical subject
Preverbal particle
Pronominal Subject + Verb order
Manuscript sermon corpus text
Verbal noun
References
Primary sources
Middle Welsh
Bromwich, Rachel & Evans, Simon D. 1997. Culhwch ac Olwen. Caerdydd: Gwasg Prifysgol
Cymru.
Goetinck, Glenys Witchard. 1976. Historia Peredur vab Efrawc. Caerdydd: Gwasg Prifysgol
Cymru.
Parry, Thomas (ed.). 1979. Gwaith Dafydd ap Gwilym. 3rd ed. Caerdydd: Gwasg Prifysgol
Cymru.
Roberts, Brinley F. 1961. Gwassanaeth Meir. Caerdydd: Gwasg Prifysgol Cymru.
Roberts, Brinley F. 1975. Cyfranc Lludd a Llefelys. Dublin: Institute for Advanced Studies.
Thomson, R.L. 1972. Pwyll Pendeuic Dyuet. Dublin: Institute for Advanced Studies.
Williams, Ifor. 1930. Pedeir Keinc y Mabinogi: allan o Lyfr Gwyn Rhydderch. Caerdydd: Gwasg
Prifysgol Cymru.
Williams, Mary. 1912. Llyma Vabinogi Iessu Grist. Revue Celtique 33: 184–248.
Early Modern Welsh corpus
16th century Bible translations
EsMorgan, William. 1956 [1588]. Llyfr Esther. In Rhyddiaith Gymraeg. Yr Ail Gyfrol. Detho­ aerdydd:
lion o Lawysgrifau a Llyfrau Printedig 1547–1618, Thomas Jones, 85–104. C
Gwasg Prifysgol Cymru.
IsMorgan, William. 1987 [1588]. Llyfr Eseia. In Y Beibl Cyssegr-lan 1588, 261a–264a.
Aberystwyth: Llyfrgell Genedlaethol Cymru.
M 1 Salesbury, William. 1967 [1567]. Llyma Cyssecrsanct Euangel Iesu Christ yn ol Marc.
In Detholion o Destament Newydd 1567, Thomas Parry (ed), 1–12. Caerdydd: Gwasg
­Prifysgol Cymru.
M 2Morgan, William. 1987 [1588]. Yr Efengyl yn ôl Marc. In Y Beibl Cyssegr-lan 1588,
454b-457a. Aberystwyth: Llyfrgell Genedlaethol Cymru.
S 1William Salesbury. 1965 [1567]. Psallwyr neu Psalmae Dauidd. In Llyfr Gweddi ­Gyffredin
1567, ii–xii. Caerdydd: Gwasg Prifysgol Cymru.
S 2Morgan, William. 1987 [1588]. Psalmau Dafydd. In Y Beibl Cyssegr-lan 1588, 218a–221a.
Aberystwyth: Llyfrgell Genedlaethol Cymru.
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
Drama
Dioddef Y Dioddefaint, MS., BM Add. 14986, 10b-33b.
GK I
‘Y Gwr Cadarn’. MS., Cardiff 2.83, 59–78.
GK II‘Y Gwr Cadarn’/ Yr ymddiddan afy Ryng yr effeiriad ar gwr bonheddig. MS., Peniarth
65, 40–72.
CynRichard Parry. Enterlute neu chwaryddiaeth Ar Destun Odiaethol yn dangos pa Drigolion a fu’n Preswulo yn y Deyrnas hon […] o wnaethuriad R. P. MS., NLW 833B.
[­Copied 1737].
Ffre
Roberts, William 1745. Ffrewyll y Methodistiaid neu Buttein-glwm Siencyn
ac Ynfydog.
Expository prose
Be
Owen, James. 1693. Bedydd Plant o’r Nefoedd. Llundain. 1–126.
CaBallinger, John (ed). 1930. Carwr y Cymry. Caerdydd: Gwasg Prifysgol Cymru. (First
published 1631).
DePrichard Williams, W. 1908. Deffynniad Ffydd Eglwys Loegr a gyfieithwyd i’r Gymraeg,
o Ladin Esgob Jewel yn y flwyddyn 1595, gan Maurice Kyffin. Bangor: Jarvis & Foster.
vi–xix, 1–105.
EdLloyd, Robert. 1629. Pregeth dduwiol yn traethu am iawn ddull, ac agwedd gwir
­edifeirwch. Llundain.
EpDavies, Richard. 1976 [1567]. Epistol Episcop Menew at y Cembru. In Rhagymadroddion
1547–1659, Garfield H. Hughes (ed.), 17–43. Caerdydd: Gwasg Prifysgol Cymru.
FfyCharles Edwards. 1936 [1677]. Y Ffydd Ddi-ffuant sef Hanes y Ffydd Cristianogol a’I
Rhinwedd gan Charles Edwards. Agraffiad cyfatebol o gopi yn Llyfrgell Salisbuy (Y Trydydd Argraffiad, 1677). Caerdydd: Gwasg Prifysgol Cymru.
GorsParry, Thomas (ed). 1930. Theater du Mond (Gorsedd y Byd) gan Rhosier Smyth.
­Caerdydd: Gwasd Prifysgol Cymru. 1–110. (First published 1615).
GwsBowen, Geraint (ed). 1970. Gwssanaeth y Gwyr Newydd Robert Gwyn 1580. Caerdydd:
Gwasg Prifysgol Cymru. 4–39.
GG
Edwards, Charles. 1679. Gwyddorion y Grefydd Gristianogol. Llundain. 1–53.
LITALlwyd Morgan. 1899 [1653]. Llyfr y Tri Aderyn. In Gweithiau Morgan Llwyd o Wynedd,
Thomas E. Ellis (ed.), 157–266. Bangor: Jarvis & ­Foster.
PrinJones, William. 1676. Priniplau neu Bennau y Grefydd Ghristianogol. Translated by
­William Jones. Llundain: A. Maxwell.
YbWilliams, Thomas. 1691. Ymadroddion Bucheddol Ynghylch Marvvolaeth. Rhydychen:
Thomas Jones. 1–84, 200–359.
Ys
Davies, Rondl. 1675. Profiad yr Ysprydion. Rhydychen.
1st Person Narrative Prose
Cr
Y Marchog Crwydrad. Edited in: Parry-Williams, T.H. 1988. Rhyddiaith Gymraeg. Y
Gyfrol Gyntaf. Detholion o Lawysgrifau 1488–1609. Caerdydd: Gwasg Prifysgol Cymru.
96–105.
FaRisiart, D. 1772. Hanes Bywyd a Marwolaeth y Parchedig Mr. Fafasor Powel. ­Caerfyrddin.
6–21.
GBCWynne, Ellis. 1976 [1703]. Gweledigaethau y Bardd Cwsc. Caerdydd: Gwasg Prifysgol
Cymru. 5–49.
H 1
Einnion, John. 1737. Helaethrwydd o Ras i’r Gwaelaf o Bechaduriaid. Caerfyrddin. 3–30.
© 2013. John Benjamins Publishing Company
All rights reserved
H 2
Prof
The history of a verb-initial construction in Welsh 
Thomas, Rhys. 1763. Helaethrwydd o ras i’r pennaf o bechaduriaid. Caerfyrddin. 1–17.
Lloyd, Henry. 1750. Profiad Tufewnol o Nefoedd ag Uffern. Brista. 11–32.
3rd Person Narrative Prose
GR
Gesta Romanorum, Story 42, Mab y Fforestwr. Edited in: Parry-Williams, T.H. (ed.).
1988. Rhyddiaith Gymraeg. Y Gyfrol Gyntaf. Detholion o Lawysgrifau 1488–1609.
Caerdydd: Gwasg Prifysgol Cymru. 122–130.
HBAThomas, Simon. 1718. Hanes y Byd a’r Amseroedd. Y Mwythig. 1–48, 59–67, 91–103,
140–9.
LlEGPuleston, Roland. 1583. Llefr o’r Eglwys Crhistnogedd. MS., NLW 716B, 1a–25a,
149a–153b, 161a–168a.
MFDefoe, Daniel. 1750. Hanes y ffortyn ar anffortyn, a fu i Mal Flanders. Y Mwythig.
[Abridged and translated by Arthur Jones?]
Pe
Thomas, Simon. 1735. Histori yr Heretic Pelagius. Y Mwythig. 13–73, 137–141.
TB
Owen, James. 1715. Trugaredd a Barn. Llundain. 1–30.
Sermons
B362,5 Robert Wynne. 1 sermon. MS., Bangor 362, 5. Preached 1717 in Llanddeiniolen.
B 95
John Griffith. 9 sermons. MS., Bangor 95. Preached 1683–5 in Llanelian.
C2.226Anon. 3 sermons. Cardiff 2.226. Preached late C17th.
EMEvan Morgan. 4 sermons, dated 1610. Edited in: Morgan, Glyn. 1969. Pregethau Cymraeg William Griffith (?1566–1612) ac Evan Morgan (c.1574–1623). Traethawd M.A.,
Prifysgol Cymru, Bangor. 578–646.
NLW3Anon. 6 sermons. MS., NLW 3B, 23–97. Preached 1675–6.
JP
John Piers. 5 sermons. mid C17th. MS., NLW 12205.
WGWilliam Griffith. 3 sermons. C16th/C17th. Edited in: Morgan, Glyn. 1969. Pregethau
Cymraeg William Griffith (?1566–1612) ac Evan Morgan (c.1574–1623). Traethawd
M.A., Prifysgol Cymru, Bangor. 70–91, 144–153, 314–331.
Other texts
Suggett, Richard F. 1983. Early Modern Welsh Defamation Suits. SSRC Final Report (HR 6979).
Jones, Thomas, 1956. Rhyddiaith Gymraeg. Yr ail gyfrol, Detholion o lawysgrifau a llyfrau
­printiedig, 1547–1618. Caerdydd: Gwasg Prifysgol Cymru.
Secondary sources
Barðdal, Jóhanna. 2011. The rise of dative substitution in the history of Icelandic: A diachronic
construction grammar approach. Lingua 121(1): 60–79. Special issue Semantic Aspects of
Case Variation, Klaus von Heusinger & Helen de Hoop.
Barðdal, Jóhanna. 2013. Construction-based historical-comparative reconstruction. In The
Oxford Handbook of Construction Grammar, Graeme Trousdale & Thomas Hoffmann (eds)
438–457. Oxford: OUP.
Bybee, Joan. 2010. Language, Usage and Cognition. Cambridge: CUP.
Berg-Olsen, Sturla. 2009. Lacking in Latvian: Case variation from a construction grammar perspective. In The Role of Semantic, Pragmatic, and Discourse Factors in the Development of
Case [Studies in Language Companion Series 108], Johanna Barðdal & Shobhana Lakshmi
Chelliah (eds), 181–202. Amsterdam: John Benjamins.
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
Bergs, Alexander & Diewald, Gabriele. 2008. Introduction: Constructions and language
change. In Constructions and Language Change [Trends in Linguistics. Studies and
Monographs 194], Alexander Bergs and Gabriele Diewald (eds), 1–22. Berlin: Mouton
de Gruyter.
Currie, Oliver. 2000. Word order stability and change from a sociolinguistic perspective: The
case of Early Modern Welsh. In Stability, Variation and Change in Word-order Patterns over
Time [Current Issues in Linguistic Theory 213], Rosanna Sornicola, Erich Poppe & Ariel
Shisha-Halevy (eds), 203–230. Amsterdam: John Benjamins.
Cappelle, Bert. 2009a. Can we factor out free choice? In Describing and Modeling Variation
in Grammar [Trends in Linguistics. Studies and Monographs 204], Andreas Dufter, Jürg
Fleischer & Guido Seiler (eds), 183–202. Berlin: Mouton de Gruyter.
Cappelle, Bert. 2009b. Contextual cues for particle placement: Multiplicity, motivation, modeling. In Context in Construction Grammar [Constructional Approaches to Language 9],
Alexander Bergs & Gabriele Diewald (eds), 145–192. Amsterdam: John Benjamins.
Coupland, Nikolas. 2007. Style: Language Variation and Identity. Cambridge: CUP.
De Sutter, Gert. 2009. Towards a multivariate model of grammar: the case of word order variation in Dutch clause final clusters. In Describing and Modeling Variation in Grammar
[Trends in Linguistics. Studies and Monographs 204], Andreas Dufter, Jürg Fleischer &
Guido Seiler (eds), 225–250. Berlin: Mouton de Gruyter.
Evans, D. Simon.1968. The sentence in Early Modern Welsh. Bulletin of the Board of Celtic
­Studies 22: 311–337.
Faarlund, Jan Terje. 1990. Syntactic Change: Toward a Theory of Historical Syntax [Trends in
Linguistics, Studies and Monographs 50]. Berlin: Mouton de Gruyter.
Fife, James & King, Gareth. 1991. Focus and the Welsh ‘abnormal sentence’: A cross-linguistic
perspective. In Studies in Brythonic Word Order [Typological Studies in Language 83],
James Fife & Erich Poppe (eds), 81–153. Amsterdam: John Benjamins.
Fontana, Josep M. 1997. On the integration of second position phenomena. In Parameters of Morphosyntactic Change, Ans van Kemenade & Nigel Vincent (eds), 207–249. Cambridge: CUP.
Fowkes, R.A. 1993. The standard Welsh of the 1588 Bible. Language Sciences 15(2): 141–53.
Fried, Mirjam. 2008. Constructions and constructs: Mapping a shift between predication and
attribution. In Constructions and Language Change [Trends in Linguistics. Studies and
­Monographs 194], Alexander Bergs & Gabriele Diewald (eds), 47–79. Berlin: Walter de
Gruyter.
Fried, Mirjam. 2009. Construction Grammar as a tool for diachronic analysis. Constructions and
Frames 1(2): 261–290.
Fried, Mirjam. 2013. The principles of constructional change. In The Oxford Handbook of Construction Grammar, Graeme Trousdale & Thomas Hoffmann (eds), 419–437. Oxford: OUP.
Goldberg, Adele. 1995. Constructions. A Construction Grammar Approach to Argument Structure. Chicago IL: University of Chicago Press.
Goldberg, Adele E. 2006. Constructions at Work: The Nature of Generalization in Language.
Oxford: OUP.
Grondelaers, Stefan, Speelman Dirk & Geeraerts, Dirk. 2008. National variation in the use
of er “there”. Regional and diachronic constraints on cognitive explanations. In Cognitive S­ociolinguistics: Language Variation, Cultural Models, Social Systems [Cognitive
­Linguistics Research 39], Gitte Kristiansen & Rene Dirven (eds), 153–203. Berlin: M
­ outon
de Gruyter.
Gries, Stefan T. 2003. Multifactorial Analysis in Corpus Linguistics: A Study of Particle Placement.
London: Continuum Press.
© 2013. John Benjamins Publishing Company
All rights reserved
The history of a verb-initial construction in Welsh 
Haser, Verena & Kortmann, Bernd. 2009. Agreement in English dialects. In Describing and
Modeling Variation in Grammar [Trends in Linguistics. Studies and Monographs 204],
Andreas Dufter, Jürg Fleischer & Guido Seiler (eds), 271–296. Berlin: Mouton de Gruyter.
Hollmann, Willem B. & Anna Siewierska. 2011. The status of frequency, schemas, and identity in Cognitive Sociolinguistics: A case study on definite article reduction. Cognitive
­Linguistics 22(1): 25–54.
Hopper, Paul J. & Traugott, Elizabeth Closs. 1993. Grammaticalization. Cambridge: CUP.
Junker, Marie-Odile. 1990. L’effet V1: Le verbe initial en moyen français. Canadian Journal of
Linguistics 35(4): 351–371.
van Kemenade, Ans & Vincent, Nigel. 1997. Introduction: Parameters and morphosyntactic
change. In Parameters of Morphosyntactic Change, Ans van Kemenade & Nigel Vincent
(eds), 1–25. Cambridge: CUP.
Kroch, Anthony S. 1989. Reflexes of grammar in patterns of language change. Language Variation and change 1: 199–244.
Luraghi, Silvia. 1995. The pragmatics of verb initial sentences in some ancient Indo-European
languages. In Word Order in Discourse [Typological Studies in Language 30], Pamela
Downing & Michael Noonan (eds), 355–386. Amsterdam: John Benjamins.
Mathias, W. Alun. 1970. William Salesbury – ei ryddiaith. In Y Traddodiad Rhyddiaith G. Bowen
(ed.), 54–78. Llandysul: Gwasg Gomer.
Pintzuk, Susan. 2003. Variationist approaches to syntactic change. In The Handbook of Historical
Linguistics, Brian D. Joseph & Richard D. Janda (eds), 509–528. Oxford: Blackwell.
Poppe, Erich. 1991. Word order in Cyfranc Lludd a Llefelys: Notes on the pragmatics of
­constituent-ordering in MW narrative prose. In Studies in Brythonic Word Order [Current
Issues in Linguistic Theory 83], James Fife & Erich Poppe (eds), 155–204. Amsterdam:
John Benjamins.
Poppe, Erich. 2000. Constituent order in Middle Welsh: The stability of the pragmatic principle. In Stability, Variation and Change of Word-order Patterns over Time [Current Issues
in ­Linguistic Theory 213], Rosanna Sornicola, Erich Poppe & Ariel Shisha-Halevy (eds),
42–51. Amsterdam: John Benjamins.
Roberts, Ian. 1993. Verbs and Diachronic Syntax. Dordrecht: Kluwer.
Roberts, Ian. 2007. Diachronic Syntax [Oxford Textbooks in Linguistics]. Oxford: OUP.
Rosenbach, Anette. 2008. How synchronic gradience makes sense in the light of language
change (and vice-versa), In Gradience, Gradualness and Grammaticalisation [Typological
Studies in Language 90], Elizabeth Closs Traugott & Graeme Trousdale (eds), 149–179.
Amsterdam: John Benjamins.
Stefanowitsch, Anatol & Gries, Stefan T. 2008. Channel and constructional meaning: A collustructional case study. In Cognitive Sociolinguistics: Language Variation, Cultural Models,
Social Systems [Cognitive Linguistics Research 39], Gitte Kristiansen & René Dirven (eds),
129–152. Berlin: Mouton de Gruyter.
Szmrecsanyi, Benedikt. 2010. The English genitive alternation in a cognitive sociolinguistics perspective. In Advances in Cognitive Sociolinguistics [Cognitive Linguistics Research 45], Dik
Geeraerts, Gitte Kristiansen & Yves Peirsman (eds), 141–166. Berlin: Mouton de Gruyter.
Thomas, Isaac. 1988. Yr Hen Destament Cymraeg 1551–1620. Aberystwyth: Llyfrgell ­Genedlaethol
Cymru.
Traugott, Elizabeth Closs & Trousdale, Graeme. 2008. Gradience, gradualness and grammaticalisation: How do they intersect? In Gradience, Gradualness and Grammaticalisation
[Typological Studies in Language 90], Elizabeth Closs Traugott & Graeme Trousdale (eds),
19–44. Amsterdam: John Benjamins.
© 2013. John Benjamins Publishing Company
All rights reserved
 Oliver Currie
Weinreich, Uriel, Labov William & Herzog, Marvin I. 1968. Empirical foundations for a theory
of language change. In Directions for Historical Linguistics, Winfred P. Lehman & Yakov
Malkiel (eds), 97–195. Austin TX: University of Texas Press.
Willis, David W.E. 1998. Syntactic Change in Welsh. A Study of the Loss of Verb-Second. Oxford:
Clarendon Press.
© 2013. John Benjamins Publishing Company
All rights reserved
`