13.1 Introduction
Welsh is one of the descendants, alongside Breton and Cornish, of the Brythonic (Brittonic, British) Celtic language spoken across almost all of the island of Britain before and during the Roman period (see also Russell, this volume). It was used for a full range of functions in medieval Wales up until the English conquest of 1282–83, and has a vigorous written record beginning in the eighth century. English rule brought with it use of English for administrative and legal purposes and increasing use of English in other domains too. Nevertheless, up to the nineteenth century, Welsh was the dominant language of Wales: in 1801, perhaps 90 per cent of a Welsh population of around 587,000 spoke Welsh and some 70 per cent of the population was monolingual (Jenkins, Suggett and White Reference Jenkins, Suggett, White and Jenkins1997:47–8). During that century, however, migration from England and Ireland brought large numbers of non-Welsh speakers into the northeast and southeast who could not be assimilated linguistically. Furthermore, from the mid nineteenth century, education was taken into the hands of the British state, which pursued a policy of linguistic eradication via an English-medium education system. Ravenstein (Reference Ravenstein1879:620–1) estimated that 66.2 per cent of the population of Wales, some 934,530 people, spoke Welsh in 1871. The proportion of Welsh speakers in the population, although not their absolute number, continued to fall, to 49.9 per cent (930,000 people) in the census of 1901. During the twentieth century, absolute numbers fell too, reaching 18.9 per cent (504,000 people) in the census of 1981. Grassroots campaigns in the 1960s and 1970s led to a gradual reversal of policies to suppress the language, including its reintroduction into public administration and the emergence of a strong network of Welsh-medium and bilingual schools. Welsh-medium education has encouraged significant growth in the number of young people able to speak the language in areas where it was weak or moribund, notably in the northeast and southeast. As a result, overall numbers of speakers have stabilised since the 1980s, although in- and out-migration remains a serious threat to the maintenance of Welsh as a community language in many areas. Current estimates of the number of Welsh speakers range from 538,300 (17.8% of the population of Wales) in the 2021 Census, and 18 per cent in the 2022–23 National Survey for Wales,Footnote 1 to 892,200 (29.5%) in the 2021 Annual Population Survey. As of January 2023, 23.9 per cent of primary schoolchildren and 23.0 per cent of middle and secondary schoolchildren in Wales attended Welsh-medium schools.Footnote 2 Welsh government policy (Cymraeg 2050) now aims to achieve one million Welsh speakers living within Wales by 2050, although it remains to be seen whether the radical policies needed to achieve this aim will be implemented.
13.2 Phonology
13.2.1 Vowel System
13.2.1.1 Vowel Inventory and Quality
Contemporary northern varieties of Welsh have seven short vowels /ɪ ɨ ʊ ɛ ɔ ə a/ and six long vowels /iː ɨː uː eː oː aː/. Southern varieties have merged the high front and high central vowels, leaving six short vowels /ɪ ʊ ɛ ɔ ə a/ and five long vowels /iː uː eː oː aː/. It has traditionally been claimed that qualitative differences between the long and short members of the paired vowels are substantial in the south but minor in the north (G. E. Jones Reference Jones, Ball and Jones1984:57). However, recent experimental work (Mayr and Davies Reference Mayr and Davies2011) has found substantial qualitative differences in both varieties. The northern system is shown in Table 13.1, with the usual orthographic correspondences of each vowel given in angle brackets.
Table 13.1 Vowel phonemes of northern Welsh
| Front | Central | Back | |
|---|---|---|---|
| High | ɪ, iː <i> | ɨ, ɨː <y, u> | ʊ, uː <w> |
| Mid | ɛ, eː <e> | ə <y> | ɔ, oː <o> |
| Low | a, aː <a> |
Note that the letter <y> represents /ɨ/, /ɨː/ and /ə/ and that /ɨ/ and /ɨː/ are sometimes represented by <y> and sometimes by <u>. The reasons for this are historical: historically, /ə/ results from the reduction of /ɨ/ and /ʊ/ in non-final syllables (syllables that were unstressed in Old Welsh); and the difference between <y> and <u> once represented the difference between unrounded and rounded high central vowels that had merged by the early modern period. Today, the orthographic difference between <y> and <u> serves to represent a morphophonological distinction between those instances of /ɨ/ or /ɨː/ that alternate with /ə/ (e.g. hyd ‘length’ /ˈhɨːd/ ~ hydoedd ‘lengths’ /ˈhədɔɨð/), represented as <y>, and those that do not (e.g. hud ‘magic’ /ˈhɨːd/ ~ hudol ‘magical’ /ˈhɨːdɔl/), represented as <u>. As well as arising through these regular phonological alternations, schwa is also found in English loanwords, where it corresponds to English /ʌ/, as in cyt ‘cut’ /ˈkət/. In northern mid-Wales (Merionnydd and Montgomeryshire) and in the traditional dialects of the southeast, /aː/ is a front vowel [æ̝ː] (Rees Reference Rees2018).
13.2.1.2 Vowel Length
In stressed monosyllables, vowels are generally short before /p t k m ŋ/ and clusters, and long in word-final position (lle /ɬeː/ ‘place’) and before /b d g f v θ ð s x/. Before /n r l/, vowels in stressed monosyllables may be short or long, with minimal pairs such as llen ‘curtain’ /ˈɬɛn/ ≠ llên ‘literature’ /ˈɬeːn/, cor ‘dwarf’ ≠ côr ‘choir’ and tal ‘tall’ ≠ tâl ‘payment’. Furthermore, loanwords frequently violate these generalisations, so that minimal pairs for vowel length can be constructed even for vowels preceding the first group of consonants, for instance, mẁg ‘mug’ /ˈmʊg/ ≠ mwg ‘smoke’ /ˈmuːg/ or cot ‘cot’ /ˈkɔt/ ≠ côt ‘coat’ /ˈkoːt/.
There is north–south dialect variation in vowel length in certain configurations. Before /ɬ/, non-high vowels are mostly long in the south and always short in the north (coll ‘lost’ southern /ˈkoːɬ/, northern /ˈkɔɬ/); but before /ɬt/, the reverse is the case, and all vowels are long in the north and short in the south (gwyllt ‘wild’ northern /ˈgwɨːɬt/, southern /ˈgwɪɬt/). In all words where the vowel is /ɪ/ or /ʊ/, and in some cases where it is /ɨ/, the vowel is short before /ɬ/ in all dialects (sill ‘syllable’ /ˈsɪɬ/, hyll ‘ugly’ /ˈhɨɬ/, twll ‘hole’ /ˈtʊɬ/). The distribution before /ɬ/ and /ɬt/ leads to the unexpected result that, in northern dialects, a consonant cluster is associated with a longer vowel than a single consonant alone (Hannahs Reference Hannahs2013:36–7). Southern dialects manifest the more expected association between the consonants in the syllable coda and vowel length. Before /sp/, /st/ and /sk/, vowels are long in the north and short in the south (clust ‘ear’ northern /ˈklɨːst/, southern /ˈklɪst/).
Long vowels are marked using a circumflex in the orthography; however, this marking is not consistent: circumflexes are not used where length can be deduced from the general rules given at the start of this section (tad ‘father’ /ˈtaːd/, llif ‘flood’ /ˈɬiːv/), in some very common words (dyn ‘man’ /ˈdɨːn/, hen ‘old’ /ˈheːn/), and in words where length varies according to dialect. Vowels which are unexpectedly short, generally in loanwords, are, in some but not all cases, marked with a grave accent, for instance, mẁd ‘mud’ /ˈmʊd/.
In polysyllabic words, vowel length is not contrastive in northern varieties, all vowels outside final syllables being short. Southern varieties do, however, maintain a partial distinction. Even in these varieties, underlyingly long vowels are shorter in the stressed, penultimate syllable of polysyllabic words, and are often described as ‘half long’. Thus, a minimal pair of monosyllabic words, like ton ‘wave’ /ˈtɔn/ ≠ tôn ‘tone’ /ˈtoːn/ remains distinct in inflected forms in the south, as with the plurals tonnau ‘waves’ [ˈthɔnɛ] ≠ tonau ‘tones’ [ˈthoˑnɛ]. In the north, the two vowels in these words fall together, rendering them homonyms, both [ˈthɔna].
Southern speakers have eight diphthongs in formal speech, while northern speakers have thirteen. The differences derive from the presence of the vowel /ɨ/ in the north, and its corresponding absence in the south. As with the plain vowels, the orthography represents the more conservative, northern system. The two systems are given in Table 13.2.
Table 13.2 Diphthong systems of northern and southern formal spoken Welsh
| South | North | |||||
|---|---|---|---|---|---|---|
| Diphthong | Orthography | Example | Diphthong | Orthography | Example | |
| /ʊɪ/ | <wy> | llwy ‘spoon’ | ||||
| /əɪ/ | <ei, eu> | lleiaf ‘least’, lleuad ‘moon’ | /əɪ/ | <ei> | lleiaf ‘least’ | |
| /ɔɪ/ | <oi, oe> | lloi ‘calves’, toes ‘dough’ | /ɔɪ/ | <oi> | lloi ‘calves’ | |
| /aɪ/ | <ai, au, ae> | llai ‘less’, cau ‘close (v.)’, cae ‘field’ | /aɪ/ | <ai> | llai ‘less’ | |
| /ɪʊ/ | <iw> | lliw ‘colour’, llyw ‘rudder’ | /ɪʊ/ | <iw> | lliw ‘colour’ | |
| /ɛʊ/ | <ew> | llew ‘lion’ | /ɛʊ/ | <ew> | llew ‘lion’ | |
| /əʊ/ | <yw> | llywydd ‘president’ | /əʊ/ | <yw> | llywydd ‘president’ | |
| /aʊ/ | <aw> | llaw ‘hand’ | /aʊ/ | <aw> | llaw ‘hand’ | |
| /ɨʊ/ | <yw> | llyw ‘rudder’ | ||||
| /ʊɨ/ | <wy> | llwy ‘spoon’ | ||||
| /əɨ/ | <eu> | lleuad ‘moon’ | ||||
| /ɔɨ/ | <ou> | toes ‘dough’ | ||||
| /aɨ/ | <au> | cau ‘close (v.)’ | ||||
| /aːɨ/ | <ae> | cae ‘field’ | ||||
In southern varieties, monophthongisation of diphthongs ending in /ɪ/ may reduce the number of diphthongs still further in ordinary speech, with /ʊɪ/ merging with /uː/, /əɪ/ merging with /iː/, /ɔɪ/ merging with /oː/ and /aɪ/ merging with /aː/. In all varieties, there are reductions of post-tonic diphthongs: post-tonic /aɪ/ is reduced to /ɛ/ in the south, and the equivalent diphthongs in the northern system are reduced to /a/ in the northwest and /ɛ/ in the northeast, hence darnau ‘pieces’, northwestern [ˈdarna], elsewhere [ˈdarnɛ].
13.2.2 Consonants
The inventory of consonantal phonemes of present-day Welsh, along with their conventional orthographic representation in angled brackets, is given in Table 13.3.
Table 13.3 Consonant phonemes of Welsh
| Labial | Dental | Alveolar | Lateral | Post-Alveolar | Palatal | Velar | Glottal | |
|---|---|---|---|---|---|---|---|---|
| Voiceless stops | p <p> | t <t> | k <c> | |||||
| Voiced stops | b <b> | d <d> | ɡ <g> | |||||
| Voiceless affricates | tʃ <tsi, ti> | |||||||
| Voiced affricates | dʒ <di> | |||||||
| Voiceless fricatives | f <ff, ph> | θ <th> | s <s> | ɬ <ll> | ʃ <si, sh> | x <ch> | h <h> | |
| Voiced fricatives | v <f> | ð <dd> | (z <z>) | |||||
| Voiceless liquids | r̥h <rh> | |||||||
| Voiced liquids | r <r> | l <l> | ||||||
| Voiceless nasal | m̥h <mh> | n̥h <nh> | ŋ̊h <ngh> | |||||
| Voiced nasal | m <m> | n <n> | ŋ <ng> | |||||
| Glides | w <w> | j <i> |
Voiceless stops are aspirated in syllable onsets. The affricates /tʃ/ and /dʒ/ occur in loanwords, such as tsiênj /ˈtʃeːndʒ/ ‘change’, and arise in native words in some southern dialects from the sequences /tj/ and /dj/, for instance, diwrnod /ˈdjʊrnɔd/ > /ˈdʒʊrnɔd/ ‘day’. The case for treating them as single phonemes rather than as sequences of two segments comes from their distinctive behaviour within the mutation system (see below).
The voiced fricative /z/ is limited to loanwords in southern dialects, such as sŵ ‘zoo’ northern /suː/ or southern /zuː/, and the loaned English plural marker ‑s, sometimes applied to native items, such as the jocular bonddigions ‘aristocrats’ /bɔnˈðɪɡjɔnz/.
The voiceless fricative /x/ may be velar [x] or uvular [χ] depending on dialect (G. E. Jones Reference Jones, Ball and Jones1984:47).
The voiceless nasals and voiceless /r̥h/ are phonetically partially voiced and strongly aspirated, and could therefore be treated as sequences of sonorant plus /h/. This view is supported by the fact that many southern dialects lack /h/, and also lack voiceless nasals and /r̥h/. This would be more easily accounted for by a two-segment analysis according to which these dialects simply lack the single phoneme /h/ in all its manifestations. However, the voiceless nasals function within the mutation system (see below) as the nasal counterpart to the voiceless stops, and /r̥h/ undergoes soft mutation to /r/. Thus, treating them as separate phonemes leads to a more straightforward analysis of the mutation system, and this is the approach usually adopted.
The phoneme /r/ is mainly realised as a voiced alveolar trill [r] or a voiced alveolar tap [ɾ]. After /d/ and /t/, it is an alveolar approximant [ɹ], voiced after /d/ and voiceless after /t/. Approximant pronunciations in other positions are becoming more common, particularly in areas of revitalisation with imposition from English (Morris, Mayr and Mennen Reference Morris, Mayr, Mennen, Durham and Morris2016). Uvular pronunciations as [ʁ] are not unheard of in parts of the north.
13.2.3 Stress
Stress is generally on the penultimate syllable. A few words have final stress where a suffix beginning in a vowel is attached to a vowel-final stem (e.g. rhyddhâd ‘liberation’ (< rhydd- ‘free’ + causative ‑ha- + abstract-noun suffix ‑ad)), in some loanwords (e.g. carafán ‘caravan’ or syrpréis ‘surprise’), and in some place names stressed as two words (e.g. Caerdydd ‘Cardiff’ or Caerlŷr ‘Leicester’). Such unexpected positioning of stress is usually marked in the orthography with an acute accent if the vowel is short, and a circumflex if it is long.
13.2.4 Initial-Consonant Mutations
Welsh makes extensive use of changes to word-initial consonants, termed (initial-consonant) mutations. While these were originally conditioned phonologically by the final segment of the previous word, they are no longer phonologically predictable and are instead triggered by the lexical or grammatical environment. For instance, certain prepositions trigger a mutation on the initial consonant of the following noun phrase:
| rhag ‘before’ + radical consonant | rhag tŷ ‘before a house’ |
| am ‘about’ + soft mutation | am dŷ ‘about a house’ |
| yn ‘in’ + nasal mutation | yn nhŷ’r deintydd ‘in the dentist’s house’ |
| â ‘with’ + aspirate mutation | â thŷ ‘with a house’ |
Other mutations are triggered by more complex grammatical environments. For instance, a feminine singular noun undergoes soft mutation after the definite article (cath ‘cat’ > gath), as does an adjective immediately following a feminine noun (bach ‘small’ > fach) or a feminine adjective (du ‘black’ > ddu):
(1)
y gath fach ddu the cat small black ‘the small black cat; the black kitten’
The present-day system of initial-consonant mutations is given in Table 13.4. From there, it can be seen that the voiceless nasals are treated as the nasal mutation counterpart to the voiceless stops. In some northern dialects, mutations are applied to other consonants. Specifically, /tʃ/ may become /dʒ/ under soft mutation and either /n̥hʃ/ or /n̥hj/ under nasal mutation; and /m/ may become /m̥h/ in contexts that trigger addition of h- before a vowel (e.g. ei mham ‘her mother’ and eu mham ‘their mother’ for ei mam and eu mam in other dialects).
Table 13.4 Phonological alternations involved in initial-consonant mutations
| Radical | Soft | Nasal | Aspirate |
|---|---|---|---|
| p | b | m̥h | f |
| t | d | n̥h | θ |
| k | g | ŋ̊h | x |
| b | v | m | |
| d | ð | n | |
| g | ∅ | ŋ | |
| m | v | ||
| ɬ | l | ||
| r̥h | r |
A related process triggers addition of /h/ before an initial vowel. The contexts that trigger this partially overlap with those for aspirate mutation, but the correspondence is by no means perfect. Thus, ei ‘her’ triggers both aspirate mutation and initial /h/ (ei chath ‘her cat’ and ei hafal ‘her apple’ < afal ‘apple’), but tri ‘three’ triggers only the former (tri chath ‘three cats’ but tri afal ‘three apples’) and eu ‘their’ triggers only the latter (eu cath ‘their cat’ but eu hafal ‘their apple’). Addition of /h/ is of course not possible in those southern dialects that lack /h/, and is becoming obsolete in speech more generally.
Mutations occurred historically word-internally and across word boundaries within the phonological phrase. They therefore also figure in the derivational morphology. Some prefixes trigger certain mutations: the negative prefix an- triggers a nasal mutation, hence an- + datod ‘dissolve’ > annatod ‘indissoluble’, while hy- ‘easy to, ‑able’ triggers a soft mutation, hence hy- + car- ‘love’ > hygar ‘amiable’. Compounding leads to a soft mutation on the second element, hence ysgol ‘school’ + tŷ ‘house’ > ysgoldy ‘schoolhouse’.
For further details of this, and other aspects of Welsh phonetics and phonology, see G. E. Jones (Reference Jones, Ball and Jones1984) and Hannahs (Reference Hannahs2013).
13.3 Morphology
13.3.1 Nominal and Adjectival Morphology
Welsh nouns inflect for number, but not case. There are two genders, masculine and feminine, with some remnants of a former neuter in the demonstrative system. As we have seen, gender is marked in the mutation system, with feminine singular nouns undergoing soft mutation after the definite article and adjectives undergoing soft mutation after a feminine noun (see (1)).
Plural marking on nouns is diverse and lexically idiosyncratic. Most commonly, the plural is marked by addition of suffixes, the most common of which is ‑(i)au, as in mam ‘mother’, plural mamau. Other suffixes include -(i)on (dyddiadur ‘diary’, plural dyddiaduron), ‑od (llew ‘lion’, plural llewod), ‑i (eglwys ‘church’, plural eglwysi), and ‑ydd (heol ‘street’, plural heolydd). Choice of suffix is largely an arbitrary property of the noun, although there is a degree of semantic conditioning: nouns denoting animals tend to have plurals in ‑od, for instance, draenog ‘hedgehog’, plural draenogod, while those denoting groups of people have either ‑(i)aid or ‑(i)on, for instance, estron ‘stranger’, plural estroniaid, or swyddog ‘officer’, plural swyddogion. Of these, ‑(i)au is fully productive for new lexical items, while ‑od, ‑(i)on, ‑i and ‑(i)aid have some degree of productivity within their semantic class. Other, less common, suffixes are no longer productive for new lexical items.
Some nouns form their plurals via a vowel alternation, for instance, bardd ‘poet’, plural beirdd. Nouns that have a vowel alteration with no addition of a suffix are the result of early sound changes triggered by a suffix containing /i/ or /j/ that has since disappeared (‘i-affection’, cf. Germanic umlaut), as with the example just cited. In other cases, a vowel alternation accompanies the addition of a suffix in the plural. In these cases, the alternation may be due to i-affection in the plural (e.g. mab ‘son’, plural meibion), where the suffix contains or at one time contained /i/ or /j/. In rare instances, there is i-affection in the singular in items whose singular formerly had a suffix containing /i/ or /j/ and the plural suffix did not (adain ‘wing’, plural adanedd).
Processes relating to the reduction of non-final vowels in earlier Welsh are also a major source of vowel alternations in the plural. Stress shifted from the final to the penultimate syllable in Old Welsh, but not before unstressed (i.e. non-final) vowels had undergone various reduction processes. These leave their mark in such plurals as dyn ‘man’ /ˈdɨːn/, plural dynion /ˈdənjɔn/; saer ‘carpenter’ /ˈsaːɨr/, plural seiri /ˈsəɪri/; sain ‘sound’ /ˈsaɪn/, plural seiniau /ˈsəɪnjaɨ/; cenhedlaeth ‘generation’ /kɛnˈhɛdlaɨθ/, plural cenedlaethau /kɛnɛˈdləɪθaɨ/. These alternations are a productive part of Welsh phonology, not limited to nominal phonology, and similar alternations are also found when derivational suffixes are added to roots, as with twf ‘growth’ /ˈtuːv/ ~ tyfu ‘grow (v.)’ /ˈtəvɨ/ or saeth ‘arrow’ /ˈsaːɨθ/ ~ saethydd ‘archer’ /ˈsəɨθɨð/.
Typologically, Welsh has attracted interest for the fact that a sizeable group of nouns leave the plural unmarked but add a singulative suffix, masculine ‑yn or feminine ‑en, in the singular, as with moch ‘pigs’, singular mochyn or clêr ‘flies (insects)’, singular cleren. These show some differences from ordinary plurals in that they can act as the input to derivation and compounding, for instance coed ‘trees’ > coediog ‘wooded’, and, on this basis, have been argued to be a category of ‘morphological collectives’ distinct from the plural (Nurmio Reference Nurmio2019). Some nouns have a singulative suffix in the singular and a plural suffix in the plural, for instance, cwningen ‘rabbit’, plural cwningod.
Other points of interest are nouns with double marking of plurality, as with neges ‘message’, whose plural is negeseuon, with addition of two suffixes, ‑au and ‑on, each of which is sufficient to mark plurality alone in other nouns. Some nouns have a different derivational suffix in the plural from the singular; for instance, the plural of posibilrwydd ‘possibility’ is posibiliadau, where the abstract-noun suffix ‑rwydd is replaced by a different abstract-noun suffix ‑iad. Some nouns form their plural from a derived form that is not found in the singular; for instance, the plural of cais ‘attempt’ is ceisiadau, with the abstract-noun suffix ‑iad appearing between the root and the plural suffix in the plural. A few irregular forms show no straightforward relationship between singular and plural, for instance, ci ‘dog’, plural cŵn; tŷ ‘house’, plural tai.
Welsh has also attracted attention for a few cases where derivational morphology appears to be found further from the root than plural inflectional morphology: the pejorative suffix ‑ach and diminutive suffix ‑os attach outside inflectional suffixes, as in dynionach < dyn ‘man’ + ‑ion ‘plural’ + ‑ach or merchetos < merch ‘girl’ + ‑ed ‘plural’ + ‑os.
Attributive adjectives inflect for number and gender in literary Welsh. This was only ever possible for some adjectives, but even for these, such inflection has now largely died out in speech and in some writing. Thus, in formal writing, gwyn ‘white’ has a feminine gwen, with vowel alternation, and a plural gwynion with a plural suffix ‑(i)on.
Comparison of adjectives is mostly via suffixes, with three degrees of comparison. For a regular adjective such as oer ‘cold’, there is an equative in ‑ed, (cyn) oered ‘as cold, so cold’, a comparative in ‑ach, oerach ‘colder’, and a superlative in ‑af, oeraf ‘coldest’. Some adjectives, particularly those with more than one syllable, form their degrees of comparison analytically using mwy ‘more’ or mwyaf ‘most’. All adjectives can also form the equative analytically using mor ‘as’.
Personal pronouns do not inflect for case, but some of them vary in form according to syntactic position. A weak form occurs in environments where there is corresponding agreement (e.g. a verb agreeing with its subject, as in (3); or a noun with a pronominal possessive, see (46)). A strong form occurs in isolation, in focus position and in other syntactic environments. Contrast the pairs fi ~ i, chdi ~ ti and fo ~ o in the following examples from spoken northern Welsh:
(2)
Fi/chdi/fo sy ’ma. 1sg.str/2sg.str/3msg.str be.prs.rel here ‘It’s me/you/him that’s here.’
(3)
Welais i / welaist ti / welodd o ddamwain. see.pst.1sg 1sg.wk see.pst.2sg 2sg.wk see.pst.3sg 3m.sg.wk accident ‘I saw/you saw/he saw an accident.’
There are also genitive proclitics that optionally double weak pronouns as possessors of nouns and as objects of non-finite verbs (1sg. fy triggers nasal mutation, gweld > ngweld and cath > nghath, cf. above):
(4)
Mae Megan wedi fy ngweld i. be.prs.3sg Megan perf 1sg.gen see.vn 1sg.wk ‘Megan has seen me.’
(5)
fy nghath (i) 1sg.gen cat (1sg.wk) ‘my cat’
In speech, there is a tendency to replace doubling with a genitive proclitic and a post-head weak pronoun (the pattern in (4)) with a single pronominal element, a strong pronoun in post-head position (see section 13.5, especially example (48)).
Very formal literary Welsh uses accusative clitics for pronominal objects of finite verbs, but these have died out in speech; contrast literary (6) with spoken (7).
(6)
Fe ’th welodd Mair. aff 2sg.acc see.pst.3sg Mair ‘Mair saw you.’
(7)
Welodd Mair ti/chdi. see.pst.3sg Mair 2sg.str ‘Mair saw you.’
13.3.2 Verbal and Prepositional Morphology
There are significant differences between spoken and literary Welsh in the domain of verbal morphology. In spoken Welsh, regular verbs have three main synthetic paradigms, namely future, past and conditional, plus an imperative. In literary Welsh, the three main paradigms have rather different semantics, so that they are often referred to as present, preterite and imperfect respectively. In literary Welsh, there is also a pluperfect formed using ‑as- between the root and the conditional (imperfect) endings, and a present subjunctive, remnants of which survive only in fixed expressions in speech. The main forms for a regular verb cysgu ‘sleep’ are given in Table 13.5. The major spoken forms are given there, with literary forms given in parentheses where they differ considerably. The future (present) paradigm expresses a modal future in speech ‘will sleep, is willing to sleep’, while it can be used as a plain present in literary Welsh. Various verbal periphrases using aspectual particles express other tense–aspect combinations, particularly in speech. In spoken Welsh, subject pronouns are generally obligatory, while they may be omitted in a null-subject pattern in literary Welsh.
Table 13.5 Sample paradigm of a regular verb cysgu ‘sleep’ in present-day Welsh
| Future (present) | Past (preterite) | Conditional (imperfect) | Imperative | |
|---|---|---|---|---|
| First sg. | cysga(f) | cysgais | cysgwn | – |
| Second sg. | cysgi | cysgaist | cysget | cysga |
| Third sg. | cysgiff (south) cysgith (north) (cysga) | cysgodd | cysgai | (cysged) |
| First pl. | cysgwn | cysgon (cysgasom) | cysgen (cysgem) | cysgwn |
| Second pl. | cysgwch | cysgoch (cysgasoch) | cysgech | cysgwch |
| Third pl. | cysgan (cysgant) | cysgon (cysgasant) | cysgen (cysgent) | (cysgent) |
| Impersonal | (cysgir) | (cysgwyd) | (cysgid) | (cysger) |
In spoken Welsh, only affirmative imperatives are available, and negative imperatives are formed with paid (singular) and peidiwch (plural), the imperative forms of the verb peidio ‘stop, cease’. Note that this no longer has a specifically inhibitive sense (negating an action that has already begun) and can be used to negate any imperative.
A notable feature of Welsh is the existence of impersonal forms of the verb used to express a generic human subject:
(8)
Gellir chwarae pêl-droed yma. be.able.prs.impers play.vn football here ‘One/you can play football here; it is possible to play football here.’
This is mainly a feature of the literary language.
There is only one non-finite form in the verbal paradigm, traditionally termed the verbnoun. Morphologically, the verbnoun is generally formed using a lexically unpredictable suffix attached to the bare root (‑io in syrthio ‘fall’, ‑u in gwenu ‘smile’, ‑o in hyrwyddo ‘faciliate’, ‑ed in yfed ‘drink’, ‑i in croesi ‘cross’, ‑yll in sefyll ‘stand’, ‑an in clebran ‘chatter’ etc.), or it may be suffixless (dal ‘hold’, atal ‘prevent’, cyrraedd ‘arrive’). For new verbs, ‑io is by far the most productive suffix.
Many verbs have a deverbal adjective in ‑(i)edig, such as printiedig ‘printed’ or berwedig ‘boiling’. Although for transitive verbs this is superficially similar to a past participle in other languages, it can only be used as an attributive or predicative adjective, and, unlike a past participle in other languages, it does not participate in any verbal constructions, such as compound tenses. Furthermore, the form is not available for all verbs, with very many idiosyncratic gaps. It is thus more comparable to other derivational affixes that form adjectives from verbs, such as ‑adwy ‘-able’ (deall ‘understand’ > dealladwy ‘understandable’) or hy- ‘easy to, ‑able’ (plygu ‘bend’ > hyblyg ‘flexible’).
A characteristic of all the modern Celtic languages is the presence of person–number inflection on prepositions. Most prepositions agree with pronominal objects, as with am ‘about’ > amdani hi ‘about her’. A few prepositions are invariant, such as gyda ‘with’ > gyda hi ‘with her’.
13.4 Syntax
Present-day Welsh is a verb-initial language, although, at an earlier stage, it had a verb-second rule in main clauses like Breton and Cornish (Willis Reference Willis1998). The usual word order in main clauses where the verb is synthetic rather than periphrastic is verb – subject – object (VSO):
(9)
Gwelodd Mair gath. see.pst.3sg Mair cat ‘Mair saw a cat.’
In speech, only the past, future and conditional are expressed using synthetic verb forms. Other tense–aspect combinations are expressed periphrastically using the verb bod ‘be’ and a number of aspectual particles before the non-finite verbnoun. The result is the pattern auxiliary – subject – aspect marker – verb – object (AuxSVO):
(10)
Mae Mair wedi gweld cath. be.prs.3sg Mair perf see.vn cat ‘Mair has seen a cat.’
There is also a tendency, particularly in northern varieties and for medium- and low-frequency verbs, to replace synthetic forms of all but the commonest verbs with a periphrasis using the verb gwneud ‘do’:
(11)
Naeth Mair weld cath. do.pst.3sg Mair see.vn cat ‘Mair saw a cat.’
In the negative, there is an especially sharp contrast between literary Welsh and spoken Welsh. In literary Welsh, sentential negation is expressed in main clauses using the preverbal particle ni(d), while, in speech, the post-verbal marker ddim is used:
(12)
Ni neidiodd Mair. neg jump.pst.3sg Mair ‘Mair did not jump.’
(13)
Neidiodd Mair ddim. jump.pst.3sg Mair neg ‘Mair didn’t jump.’
Historically, this is the result of a negative (Jespersen) cycle, with ddim (originally < dim ‘thing, something’) having grammaticalised as a strengthener of the original negative particle ni(d), eventually replacing it (Willis Reference Willis, Willis, Lucas and Breitbarth2013).
In the present tense (including periphrases), most southern varieties use a negative auxiliary, variously smo, sa/so, simo, sana/sano; contrast southern (14) with (15) as found elsewhere:
(14)
So Mair wedi gweld cath. neg.3sg Mair perf see.vn cat ‘Mair hasn’t seen a cat.’
(15)
Dydy Mair ddim wedi gweld cath. neg.be.prs.3sg Mair neg perf see.vn cat ‘Mair hasn’t seen a cat.’
Historically, this derives from an existential construction nid oes dim o ‘there isn’t any of’, with grammaticalisation and extension beyond the original existential context. This form is a marker, and to some extent even a stereotype, of southern colloquial speech and is showing some signs of spreading within the south beyond its original distribution. In traditional varieties of the far southwest (the western half of Pembrokeshire), this auxiliary also has an imperfect form wena or ana (Awbery Reference Awbery, Ball, Fife, Poppe and Rowland1990).
Negative concord is a pervasive feature of the negative system. Some instances have already arisen: the verb bod ‘be’ has special negative-concord forms for use in negative clauses; contrast dydy in the negative clause in (15) with mae in the corresponding affirmative clause in (10). Negative-concord items, such as neb ‘no one’ or dim byd ‘nothing’, may express a negative in isolation, for instance, in sentence fragments, in (16), but may also trigger multiple negative concord elsewhere in the sentence, as in (17), where both dydy and ddim show negative concord with dim byd.
(16)
Pwy oedd yno? Neb. who be.impf.3sg there n.one ‘Who was there? No one.’
(17)
Dydy Mair ddim wedi gweld dim byd. be.prs.3sg Mair neg perf see.vn n.thing ‘Mair has not seen anything.’
Like main clauses, embedded clauses also manifest VSO or AuxSVO order. In literary Welsh, embedded complement clauses are introduced by the particle y(r):
(18)
Mae Geraint yn credu y gweli di gath. be.prs.3sg Geraint prog believe.vn prt see.fut.2sg you cat ‘Geraint believes that you will see a cat.’
This particle is omitted in speech and may be replaced by a soft mutation in some varieties.
Finite verb forms are somewhat restricted in embedded complement clauses. The verb in (18) is future in meaning, and finite future forms are permitted in such clauses. However, where the verb would be in the past tense (preterite), an affirmative embedded clause is replaced by a non-finite clause with the subject marked using the preposition i ‘to’ (an ‘i-clause’) and the overall word order is SVO (Tallerman Reference Tallerman1998). Thus, (9) and (11) embed as (19).
(19)
Mae Geraint yn credu [i Mair weld cath]. be.prs.3sg Geraint prog believe.vn to Mair see.vn cat ‘Geraint believes that Mair saw a cat.’
Where the verb would be in the present tense or imperfect (including periphrases), an affirmative embedded clause is replaced by a clause introduced by non-finite forms of the verb bod ‘be’, corresponding to an AuxSVO word order. Thus, (10) embeds as (20).
(20)
Mae Geraint yn credu [bod Mair wedi gweld cath]. be.prs.3sg Geraint prog believe.vn be.vn Mair perf see.vn cat ‘Geraint thinks that Mair has seen a cat.’
From the discussion so far, we have seen that VSO word order alternates with certain other patterns: AuxSVO and i + SVO. These alternations have been interpreted as evidence that VSO word order in Welsh should be derived from underlying SVO by movement of the verb to an inflectional head preceding the subject. Where this occurs, we see an inflected verb in VSO order. Where this does not occur, we see either an inflected auxiliary followed by SVO or else an uninflected structure with a verbnoun.
Typologically, Welsh is a fairly strongly head-initial language, and word order within phrases reflects this. Adjectives, demonstratives, possessors and some quantifiers follow their head noun, while the definite article, numerals, some quantifiers and a few adjectives precede:
(21)
y gath ddu hon the cath black dem ‘this black cat’
(22)
holl gathod Steffan all cats Steffan ‘all Steffan’s cats’
(23)
y tair hen gath i gyd the three old cat all ‘all the three old cats’
A major feature of Welsh syntax is the use of initial-consonant mutations. Morphophonological aspects of the mutations were dealt with in Section 13.2.2. Many mutations are triggered by a given word on the immediately following item within the same phrase. For instance, soft mutation is triggered by the pronominal elements dy ‘your (sg.)’ and ei ‘his’, by numerals dau ‘two (masc.)’ and dwy ‘two (fem.)’, by quantifiers holl ‘all’ or ychydig ‘few’, by prepositions am ‘about’ and o ‘from’, among many others (soft mutations in bold):
(24)
am glustiau dy ddwy gath di about ears 2sg two cat you ‘about the ears of your two cats’
Of all the Celtic languages, Welsh is the one that has gone furthest in integrating mutation into the syntax, and a number of mutation triggers are quite abstract. For instance, the direct object of a finite verb undergoes soft mutation, as in (25) (cath > gath), as does a noun phrase following a prepositional phrase, as in (26) (cathod > gathod).
(25)
Welodd Mair gath. see.pst.3sg Mair cat ‘Mair saw a cat.’
(26)
Mae ’n gas gan Mair gathod. be.prs.3sg pred hateful with Mair cats ‘Mair hates cats.’
Such patterns have been interpreted in various ways, either as an indication that Welsh uses mutation to mark accusative case (Roberts Reference Roberts2005), or as an indication that mutation marks structural boundaries between phrases (the central insight of the XP-Trigger Hypothesis, Borsley and Tallerman Reference Borsley and Tallerman1996).
Welsh has an elaborate series of clause-initial particles used to indicate clause status (main or embedded), polarity (negative or affirmative), interrogative vs. declarative and to indicate narrow focus on a fronted element. The literary system is given in Table 13.6 (superscripts indicate mutation effects: S = soft mutation, AS = combination of aspirate and soft mutation).
These particles are generally analysed as complementisers, meaning that Welsh uses complementisers to introduce some main clauses as well as subordinate clauses. Particularly striking is the availability of the complementiser fe to mark affirmative main clauses (cf. example (6), and also use of mi in this function in some varieties of spoken Welsh below):
(27)
Fe welodd Mair gath. aff see.pst.3sg Mair cat ‘Mair saw a cat.’
Table 13.6 Clause-initial particles in literary Welsh
| Main | Embedded | |
|---|---|---|
| Affirmative declarative | ø, feS | y(r) |
| Negative declarative | ni(d)AS | na(d)AS |
| Affirmative interrogative | aS | aS |
| Negative interrogative | oni(d)AS | oni(d)AS |
| Affirmative focus | ø | mai |
| Negative focus | nid | nad |
| Aff. inter. focus | ai | ai |
| Neg. inter. focus | onid | onid |
There are both affirmative and negative question particles. As would be expected typologically (initial question particles are found in prepositional languages, Greenberg Reference Greenberg and Greenberg1963:47), these are clause-initial:
(28)
A welodd Mair gath? qu see.pst.3sg Mair cat ‘Did Mair see a cat?’
(29)
Oni welodd Mair gath? neg.qu see.pst.3sg Mair cat ‘Didn’t Mair see a cat?’
Narrow focus is indicated by fronting of the focus element. What follows this element is formally identical to a relative clause:
(30)
Mair a welodd gath. Mair prt see.pst.3sg cat ‘It was Mair that saw a cat.’
Narrow-focus negation of the fronted element is indicated by the particle nid:
(31)
Nid Mair a welodd gath. neg.foc Mair prt see.pst.3sg cat ‘It wasn’t Mair that saw a cat.’
If the fronted element is questioned, this is marked by the particle ai, or, in the negative, onid:
(32)
Ai Mair a welodd gath? qu.foc Mair prt see.pst.3sg cat ‘Was it Mair that saw a cat.’
(33)
Onid Mair a welodd gath? neg.qu.foc Mair prt see.pst.3sg cat ‘Wasn’t it Mair that saw a cat.’
Where an element has been fronted for focus in the embedded clauses, this is marked using the particle mai, or its negative counterpart nad:
(34)
Gwn mai Mair a welodd gath. know.prs.1sg foc Mair prt see.pst.3sg cat ‘I know that it was Mair that saw a cat.’
(35)
Gwn nad Mair a welodd gath. know.prs.1sg neg.foc Mair prt see.pst.3sg cat ‘I know that it wasn’t Mair that saw a cat.’
This system of particles has broken down to some extent in speech. Some of the particles have been eroded. The question particle a is no longer used in speech, and only its mutation effect remains. Similarly, fe, along with its spoken northern counterpart mi, are frequently omitted in speech, although their mutation effects often remain. As we have already seen, the sentential-negation marker ni(d) is not used in speech.
Some of the particles have different variants in speech: thus ai is replaced by (d)ife in much of the south, and omitted entirely in the northwest; mai is replaced by taw in the far south, and by na in the northwest; the focus negative marker nid is replaced by dim or by the sentential-negation marker ddim in speech.
Finally, some of the particles have been re-formed from their constituent semantic parts: thus nad, which represents embedded focus and focus negation, is replaced by mai nid, a combination of particles used elsewhere on their own to express those values (36); in embedded clauses, ai is replaced by os ‘if’ + embedded focus marker mai (37).
(36)
Dwi ’n gwybod mai dim Mair welodd gath. be.prs.1sg+I prog know.vn foc neg.foc Mair see.pst.3sg cat ‘I know that it wasn’t Mair that saw a cat.’
(37)
Dwi ’n ansicr os mai Mair welodd gath. be.prs.1sg+I prog uncertain if foc Mair see.pst.3sg cat ‘I’m unsure if it was Mair that saw a cat.’
The full spoken system is given in Table 13.7.
Table 13.7 Particles in spoken Welsh
| Main | Embedded | |
|---|---|---|
| Affirmative declarative | miS, feS, ø, øS | ø, øS, y(r) |
| Negative declarative | ø(A)S … ddim | na(d)(A)S, ø(A)S |
| Affirmative interrogative | øS | os |
| Negative interrogative | on’(d), ø(A)S … ddim | on’(d), ø(A)S … ddim |
| Affirmative focus | ø | mai, taw, na |
| Negative focus | nid, dim, ddim | nad, mai nid, mai dim, mai ddim |
| Aff. inter. focus | ai, (d)ife, ø | ai, os mai |
| Neg. inter. focus | on’(d), nid, dim, ddim | oni(d) |
Welsh shows agreement in person, number and, in the third-person singular, gender, between heads and pronominal elements, but not with lexical ones. Thus, verbs agree with subject pronouns, but remain in the default, third-person singular form with all lexical subjects, including plural ones, thus daeth y merched ‘the girls came’, but daethon nhw ‘they came’. Prepositions similarly agree with pronominal objects, but use a bare form with lexical objects, thus am ferch ‘about a girl’ but amdani hi ‘about her’.
There are two main copular structures. The first is a predicational one, in which a predicate nominal or adjective is marked using the predication marker yn in a verb-initial order, as in (38). The predication marker yn is also used to mark adverbs and secondary predicates.
(38)
Mae Mair yn feddyg / dalentog. be.prs.3sg Mair pred doctor talented ‘Mary is a doctor/talented.’
The second copular structure is primarily an identificational one, in which the new element (in terms of information structure) is fronted in a verb-medial structure:
(39)
Caerdydd yw prifddinas Cymru. Cardiff be.prs.3sg capital Wales ‘The capital of Wales is Cardiff.’
For further discussion of Welsh syntax, see Borsley, Tallerman and Willis (Reference Borsley, Tallerman and Willis2007).
13.5 Sociolinguistics
13.5.1 Diglossia and Standardisation
At least until the second half of the twentieth century, Welsh manifested diglossia, with literary Welsh used in high contexts, notably for religious purposes, journalism, formal writing and literature, and, in slightly modified form, for scripted speech, and the dialects used in low contexts, for all everyday unscripted speech and some folk literature. There were considerable phonological, morphological and syntactic differences between the two, some of which have been outlined above.
More recently, there has been some convergence of the two varieties, and a stylistic continuum has been emerging. Example (40) shows a sentence in traditional literary Welsh. A parallel contemporary written Welsh version, typical, for instance, of much journalism today, is given in (41). This removes more archaic features of literary Welsh not shared with speech, notably the preverbal particle y, the first-person plural prepositional ending ‑om, the verbnoun form gweled ‘see’ and plural inflection on adjectives. This example also illustrates some of the ways in which written Welsh has begun to adopt some features of spoken varieties, namely the prepositional ending ‑on and the inclusion of the pronoun ni ‘us’ as the object of the preposition.
(40)
Y mae arnom eisiau gweled eich teigrod gwynion.
(41)
Mae arnon ni eisiau gweld eich teigrod gwyn. prt be.prs.3sg on.1pl 1pl want see.vn 2pl tigers white.pl ‘We want to see your white tigers.’
These changes bring the written form close to a usage possible in careful speech, and, to some extent, also bridge the gap with less formal spoken forms, such as the northern (42) and the southern (43). Some regional features now also appear in relatively high contexts, such as print and broadcast journalism and government communication.
(42)
’Dan ni eisia gweld eich teigyrs gwyn chi
(43)
’Ŷn ni moyn gweld eich teigyrs gwyn chi. be.prs.1pl 1pl want see.vn 2pl tigers white 2pl ‘We want to see your white tigers.’
In (42) and (43), the experiencer is marked as the subject of the verb ‘be’ rather than with ar ‘on’; the form of the verb ‘(we) are’ is dialectally specific (rather than formal literary yr ydym or less formal but still written rydyn ni); the lexical item eisiau ‘want’ is replaced in the south by a different item altogether, namely moyn (cf. the literary cognate ymofyn ‘fetch’); with the loanword teigyr ‘tiger’, the morphological plural in ‑od is replaced by the English loan morpheme ‑s; and the preposed possessive construction is replaced by a bipartite embracing construction (eich … chi).
Conversely, features of scripted speech, particularly those represented in the orthography, have begun to appear in non-scripted contexts. For instance, educated southern speakers now adopt a phonology based on the orthography in free conversation, replacing long monophthongs with diphthongs, for example, thus pronouncing cae ‘field’ as [ˈkhaɪ] rather than as [ˈkhɑː].
More colloquial registers also exist. Examples of the same sentence in normal casual speech are given in northern (44) and southern (45). Here, the verb ‘be’ is dropped entirely in the south, as is the prenominal possessor marker eich in both north and south.
(44)
’Dan ni ’sio gwel’ teigyrs gwyn chi.
(45)
Ni moyn gweld teigyrs gwyn chi. be.prs.1pl 1pl want see.vn tigers white.pl 2pl ‘We want to see your white tigers.’
Linguistic revitalisation has also had an impact on this emerging continuum. The question of which variety of Welsh should be taught as a second language (L2) has consumed considerable energy since the 1960s. Until then, learners were generally taught literary Welsh, leaving them able to read and write Welsh, but not to speak it outside of the most formal environments. A new form intended to be taught as a spoken variety to L2 learners, Cymraeg Byw ‘Living Welsh’, was planned and developed by the Welsh Joint Education Committee in the 1960s and 1970s. This variety was condemned as artificial by conservatives who stressed the existing use of a variety close to literary Welsh in highly formal, unscripted conversation and advocated this as a learner variety (Thomas Reference Thomas1982). Nevertheless, it was influential in the late twentieth century.
Today, highly localised dialect features are on the decline among all speakers, with general southern and northern colloquial varieties emerging and beginning to serve as de facto spoken standards within their regions. Adult L2 learners are now generally taught one of these two varieties. In immersion schools, the variety used by teachers with children is typically a regional spoken standard with some prescriptive influence from literary Welsh, and the variety acquired is this, often mediated by an English-language substrate, leading to a further decline in use of local dialect features (M. C. Jones Reference Jones1998).
13.5.2 Contemporary Variation and Change
Studies of variation in spoken Welsh in a quantitative framework have been undertaken since the 1970s. They have identified age, style or register, linguistic background and linguistic identity as significant factors in determining variation. There seems to be little significant social-class variation, although this has generally been assumed as an informal observation rather than demonstrated. Absence of social-class variation might indeed be expected, partly due to the traditional diglossic situation described above and the consequent absence of a sociolinguistic continuum, and partly because the high variety, which might influence speech, was replicated through religious and cultural institutions to which access was restricted by network rather than social class (Thomas Reference Thomas1987:100).
Current topics of interest include variation and change in the expression of pronominal possessors and in omission of the verb bod ‘be’ as auxiliary and in other contexts, and recent work in these areas will be briefly sketched here.
Traditionally, a pronominal possessor is expressed by a clitic that precedes the noun, optionally doubled by a pronoun following it:
(46)
ein cath (ni) 1pl cat (we) ‘our cat’
While these patterns are found in speech, they have been joined by another pattern, where only the postnominal pronoun is present:
(47)
cat ni cat we ‘our cat’
This pattern dates back to the nineteenth century in L2 Welsh (Willis Reference Willis2016) and was perhaps originally used by first-language (L1) speakers with personal names (Twm ni ‘our Tom’) (Watkins Reference Watkins1977:157). In contemporary Welsh, the postnominal pattern is found much more frequently in the language of younger speakers (Davies Reference Davies, Durham and Morris2016). In the 1990s, it was reported that the incidence of the postnominal variant was very high in the speech of primary schoolchildren, with L2 children having higher proportions of non-standard variants than L1 ones (B. M. Jones Reference Jones1990a). The emergence of this pattern means that Welsh more consistently adopts noun–possessor word order, increasing its adherence to the head-initial language typology. This can also be thought of as an analogical extension of the word order found with nominal possessors (cf. cath Steffan ‘Steffan’s cat’), or as an extension of a parallel change that began with the object of non-finite verbs (verbnouns). Direct objects of verbnouns can be expressed using preverbal clitics with optional doubling of a post-verbal pronoun, cf. (4). In speech, this pattern is often replaced with a solely post-verbal one:
(48)
Mae Megan wedi gweld fi. be.prs.3sg Megan perf see.vn 1sg.str ‘Megan has seen me.’
This change has progressed further than that with possessors.
The role of English here has been disputed. While, on the one hand, there is no analogue of the postnominal structure in English, it has been argued that language contact promotes regularisation and simplification (Trudgill Reference Trudgill2011) and that this is what has been happening in this case in Welsh (see B. M. Jones Reference Jones, Ball, Fife, Poppe and Rowland1990b).
Deletion of forms of ‘be’ is another feature of contemporary spoken Welsh. Thus, in (49), ‘be’ is omitted, and ∅ marks the place where an overt form of it might be expected to appear.
(49)
∅ Ni ’n dal y bws nawr. (be.prs.1pl) we prog catch.vn the bus now ‘We[’re] catching the bus now.’
While, on the one hand, this parallels be-deletion in spoken English interrogatives (You ready yet?), introducing an SVO pattern that resembles English, and is found in nineteenth-century representations of L2 speakers (Willis Reference Willis2016, Reference Willis2020), some aspects of it are not so readily explained as contact-induced. All varieties of Welsh manifest this phenomenon (often termed ‘auxiliary deletion’, although it applies to both auxiliary and main-verb ‘be’), yet it shows complex patterns of variation by person, number and dialect that do not closely match factors relevant in English. Deletion in the first-person plural, for example, as in (49), is largely restricted to southern varieties. Auxiliary deletion is mostly found in clauses with pronominal subjects (although deletion with lexical subjects is not completely unheard of). The auxiliary does not need to be in absolute sentence-initial position, and, indeed, can be in a subordinate clause. Another view (Davies Reference Davies2010:323–8) thus takes it to be a phonologically motivated language-internal change subsequently accelerated by the isomorphism with English.
Given the current close relationship with English brought on by language revitalisation in the northeast and southeast, and continued pressure on the language in the heartland areas of the west, sociolinguistic issues involving language contact are likely to continue to feature prominently in the study of Welsh linguistics.