A Shibboleth upon Their Tongues: Early English /r/ Revisited

Abstract

This article discusses the pronunciation of the rhotic phoneme /r/ in early English. The traditional belief that the dominant pronunciation in Old and Middle English was [r] (an apical trill) is still supported by some authors, but there is growing consensus that there was a fairly wide range of /r/ realisations already in early Germanic, and that the pronunciation of /r/ in Old English was about as variable as it is in present-day English. The article defends this view and goes a step further, suggesting that the modern distribution of variant rhotic pronunciations in British English reflects to some extent the distribution of very similar sounds in Old English.

Full text

Studia Anglica Posnaniensia 42, 2006
A SHIBBOLETH UPON THEIR TONGUES:
EARLY ENGLISH /r/ REVISITED
PIOTR GSIOROWSKI
Adam Mickiewicz University, Poznań
… the Natives of this Country, of the an-
tient original Race or Families, are distin-
guished by a Shibboleth upon their
Tongues in pronouncing the Letter R…
(Defoe 1724-27: 232-233).
ABSTRACT
This article discusses the pronunciation of the rhotic phoneme /r/ in early English. The traditional
belief that the dominant pronunciation in Old and Middle English was [r] (an apical trill) is still
supported by some authors, but there is growing consensus that there was a fairly wide range of
/r/ realisations already in early Germanic, and that the pronunciation of /r/ in Old English was
about as variable as it is in present-day English. The article defends this view and goes a step
further, suggesting that the modern distribution of variant rhotic pronunciations in British English
reflects to some extent the distribution of very similar sounds in Old English.
1. Introduction: What makes a sound rhotic?
The notion of “rhotic” is notoriously controversial and hard to formalise. A
variety of segments with diverse articulations, including those symbolised as [r,
ɹ, , ɻ, , , , , , χ, , ̯, ̯, ...], with or without diacritics, may be described as
rhotic, though it is difficult to see what properties unify them as a natural class –
if rhotics constitute such a class in the first place. To aggravate the taxonomic
difficulties, there is a good deal of overlap (and of cross-linguistic ambiguity)
between rhotics and other sounds. For example, the apico-alveolar tap [],
though usually regarded as a rhotic, may function as an allophone of apical
stops (/t/ or /d/) in some languages (including American English), and sounds

P. Gsiorowski
64
such as [ɹ, , ] may well function as rhotic “liquids” in one language but as
fricatives in another.1 There have been attempts to define rhotics in terms of
shared physical correlates, but the similarity between them, while hard to deny,
is frustratingly elusive. For example, while many rhotics exhibit a conspicu-
ously lowered third formant (F3) of their acoustic spectrum, such a “dip” is not
characteric of rhotics in general:
Uvular r-sounds have a high third formant, sometimes close to the fourth formant.
Dental r-sounds also have a relatively high third formant, though not so high as
the uvulars (Lindau 1985: 165).
Some authors flatly deny the existence of any phonetic basis for defining
rhotics; that is what Lindau emphasises in her seminal article:
But there is no physical property that constitutes the essence of all rhotics
(Lindau 1985: 166).
One desperate solution, in the face of such difficulties, would be a cover fea-
ture with a circular definition (in brief, a rhotic is [+rhotic]), amounting in fact
to an admission of defeat. According to Lindau (1985: 167), what underlies the
class of rhotics is their “phonological behaviour” and “family resemblance”.
Ladefoged – Maddieson (1996: 215) content themselves with a shockingly in-
formal description of rhotics as a class of sounds unified mostly by historical
connections and the choice of the letter r (or Greek ρ, or the phonemic symbol
/r/) to represent them.
Wiese (2001) proposes to define rhotics in terms of phonotactic patterning,
as a point on the sonority scale between laterals and glides (semivowels), but,
apart from the question whether such a definition is methodologically accept-
able, it is far from evident that all glides are inherently more sonorous than all
rhotics; many of the latter have glide realisations such as [̯, ̯] etc. Worse still,
the universal sonority hierarchy may be at odds with the language-specific pho-
notactic behaviour of “glides” vs. “liquids”; in Old English, for example, /wr/
and /wl/ are permissible word-initial clusters. Oostendorp (2001) argues that
rhotics have no fixed sonority value and are instead characterised by “chamele-
onic” adaptability: the more consonantal position a rhotic occupies, the less
sonorous it is.
1 This ambiguity is by no means unique to rhotics, and is the inevitable consequence of trying to
reconcile a binary phonological contrast [±sonorant] with a scalar dimension (the openness of
stricture and the degree of airflow obstruction). Other approximants similarly “shade into” frica-
tives as the constriction of the vocal tract is narrowed. The means by which IPA distinguishes
degrees of constriction are very inconsistent: in narrow phonetic transcription [ɹ, , ] are symbols
for fricatives, whereas [ɹ, , ] stand for the corresponding approximants.

A Shibboleth upon Their Tongues … 65
Perhaps, then, the relative colourlessness of rhotics is precisely what charac-
terises them. Thus defined, they do not constitute a natural class;2 “rhotic” is
merely a cover term for minimally specified consonantal sonorants3 – a subset
of those that do not fit into the classes of nasals and laterals and so lack any
manner specification.4 That puts them on a par with typical “semivowels” (or
“glides”) such as [j] and [w]. The latter are distinguished by the same features
as the close vowels to which they are related (i.e. [+high, +front] and [+high,
+round]), while typical rhotics lack those salient combinations of features and
are phonetically akin to unrounded non-front vowels such as [ɨ, , , , , , ].5
The old description of /r/ as littera canina ‘dog’s letter’ (i.e. a “growling”
sound) alludes at the same time to the central-vowel timbre of many rhotics and
to the frequent occurrence of trills among them.
Because of their underspecification, rhotics rarely contrast with one another
within the same system; I shall therefore follow the usual practice of employing
/r/ as a convenient phonemic transcription if the language in question has only
one “rhotic” phoneme, no matter what its realisation. It must be noted that,
thanks to the articulatory diversity of “basal” sonorants, phonetic realisations of
/r/ will often vary allophonically and dialectally as well as idiolectally.
2. The traditional view of early English rhotics
The apical trill [r] is commonly regarded as the prototype of the whole category.
It is also rather common cross-linguistically; hence the widespread tendency to
believe that any other phonetic realisations of /r/ must be modern corruptions of
an original trilled pronunciation.6
2 That is, in a system with two or more rhotics they will not pattern together to the exclusion of
other sonorants (but see fn. 4).
3 Or, as Oostendorp (2001: 121) puts it, “segment[s] with a phonologically (almost) empty
specification”.
4 This will be true of most phonological inventories; however, [±trill] should perhaps be intro-
duced as a manner-of-articulation feature in systems that contrast trilled rhotics with non-trilled
ones. Czech distinguishes trilled and non-trilled fricatives, just as some languages distinguish
lateral fricatives from non-lateral ones. If accepted, such a feature would make trills (but not
“rhotics” in general) a well-defined natural class.
5 Of course, palatalised and labialised rhotics occur in some systems, but such secondary ac-
companiments should be representationally distinguished from [j] and [w] as independent seg-
ments.
6 See Erickson (2003) for a modern variant of this position. Erickson argues that an original [r]
would be capable of producing all the modern types of English /r/ through perfectly natural de-
velopments, which is no doubt true, but the same, in principle, could be claimed of many other
rhotic pronunciations, and Erickson’s analysis does not rule out other possibilities.

P. Gsiorowski
66
It is still commonly believed that /r/ “ought” to be rolled in English, a belief no
doubt stemming from the Italianate tradition in the teaching of singing…
(O’Connor 1973: 150).
In studies of English historical phonology the assumption that the modern
realisations of /r/ have developed from a trill is often simply taken for granted,
the only question being how long ago /r/ became “untrilled”. Many, perhaps
most, historical phonologists would probably favour a trilled articulation as the
expected value of Proto-Indo-European and Proto-Germanic *r – in deference to
traditional scholarship. Jespersen (1909) ascribes it to Old and Middle English
as well:
The OE and ME /r/ was probably a strongly trilled point-consonant everywhere.
The first indications of a weakening of /r/ are found towards the end of the 16th c.
… Ben Jonson († 1639) is the first to recognize a difference according to its posi-
tion…; initial /r/ he distinctly describes as point-trilled
(Jespersen 1909: 318 [§11.11]).
In the early 17th c. r was probably a trilled point-// [= IPA [r] – PG] (like the
Scotch) before a vowel, and before a consonant an untrilled consonantal /r/ very
much like the sound now given to r before a vowel in South England
(Jespersen 1909: 358 [§13.21]).
The much-cited passage by Ben Jonson is also the foundation of Sweet’s
(1888) opinion on the pronunciation of early Modern English /r/ (trilled ini-
tially, untrilled in rhyme positions):
R is the Dogs letter, and hurreth in the sound; the tongue striking the inner palate,
with a trembling about the teeth. It is sounded firme in the beginning of the words,
and more liquid in the middle, and ends: as in rarer. viper. and so in the Latine
(Sweet 1888: 263-264 [§900]).
According to Sweet, there seems no reason why Jonson’s description should
not be taken literally. However, it may appear odd that there should be no doubt
about this particular report while on the same page Sweet rejects Cooper’s tes-
timony to the effect that final /r/ was trilled:
But here the mention of the vibration seems to be nothing but a part of the tradi-
tional definition of r. It it remarkable how people cling even now to the idea that
the E. r is trilled, probably confounding trilling with the voice-vibration in the
glottis. Walker even imagines a trill of the root of the tongue in one of his pronun-
ciations of r
(Sweet 1888: 264 [§901]).

A Shibboleth upon Their Tongues … 67
Dobson (1968: 326) makes an observation that gives an ironic twist to the
dispute. He remarks that Jonson’s celebrated description “seems most interest-
ing, but it is merely translated from Ramus” (i.e. the French grammarian
Ramée). Further on (Dobson 1968: 946 [§370, fn.]), Dobson dismisses the
whole description as worthless and notes again that Jonson “merely plagiarizes
Ramée”. Dobson’s own opinion is that at some indeterminate time in the devel-
opment of “Standard English” (but not later than the late fourteenth century)
there probably was
… a change in the nature of r from a point-trilled consonant to the PresE point-
fricative, which has strong guttural quality and is closely allied to the vowel [ə];
but in intervocalic position it commonly remained either a trilled consonant or the
PresE ‘flap’ [r] Dobson 1968: 945 [§370]).
As for the earliest stages of English, Sweet (1888: 136 [§506]) states: “The
OE r was no doubt a strong point trill as in the present Scotch dialects”.7 A
more cautious opinion is expressed in his Anglo-Saxon Primer:
r initially was probably trilled, as in Scots: rœ̅d ‘advice’, rīdan ‘ride’. Finally and
before a consonant it was probably made with the tip of the tongue curved back,
as in south-western dialects of Mn.E. and in American: ār ‘mercy’, eard ‘coun-
try’, feorh ‘life’ (Sweet 1882 [1952]: 3 [§3]).
This remains the standard handbook summary of the Old and Middle English
situation. Positional retroflexion is supposed to account for some of the rhotic-
conditioned sound changes in Old English (especially pre-/r/ breaking), for
rhotic metathesis, and for the wide distribution of [ɻ] in modern times (Mossé
1945). Reszkiewicz (1953) and Fisiak (1967) argue that the “light” (trilled)
rhotic (IPA [r]) and the “dark” (retroflex) rhotic (IPA [ɻ]) represent different OE
phonemes contrasting in initial position, where ‹r› /r/ is distinguished from ‹wr-›
/ɻ/; and if the spelling ‹hr› is taken to represent voiceless [r̥], up to three rhotic
phonemes can be posited for Old English.8 However, their defective distribution
and limited contrastive potential, and the fact that Old English alliteration rules
treat ‹wr-› and ‹hr-› as clusters, militate against such an interpretation; conse-
quently, Reszkiewicz’s theory has not gained wide acceptance.
7 Actually, the famous Scottish trill “tends to be restricted to formal or declamatory styles”
(Wells 1982: 411), and the mainstream realisations of /r/ in Scotland are a tap [] and a postalveo-
lar approximant or fricative [ɹ] (preferred in coda positions and frequent initially). The situation in
Sweet’s times was not much different.
8 And, similarly, three lateral phonemes, /l, ɫ, l̥/, represented word-initially as ‹l-›, ‹wl-› and ‹hl-›.

P. Gsiorowski
68
3. Other opinions
In recent decades the realisation of early English (and, more generally, early
Germanic) /r/ has become a matter of some controversy, and the traditional
reconstruction has been challenged by several authors. The problem is interest-
ing because of the role of /r/ as the conditioning factor of several important
sound changes in the history of the Germanic languages. If one wishes to ac-
count for those changes in terms of natural phonetic processes conditioned by
the presence of a rhotic segment, then an articulation known to cause similar
effects in modern languages is obviously preferable as a reconstruction. Thus,
Lass – Anderson (1975: 85-89) argue in favour of a back articulation of Old
English /r/ (a uvular fricative [] or trill []) on the grounds that Old English
front-vowel diphthongisation (breaking) is explicable as a coarticulatory effect
caused by members of the natural class of back consonants.9
More recently, Lass (1983) attempts to explain not only breaking but a
whole variety of disparate sound changes in Old, Middle and early Modern
English (vowel retraction, raising, lowering, rounding and centralisation) by
reconstructing /r/ as a “multifocal” bundle of components, each of them respon-
sible for some of the observed effects. Lass’s /r/ is thus a combination of labial,
coronal, velar, palatal and pharyngeal subcomponents that may, at particular
times, be selected as the phonological “focus” eliciting a particular sound
change. The kind of rhotic that best meets such a description is, in Lass’s opin-
ion, virtually identical with the so-called “bunched” /r/ prevailing in Modern
American English, i.e. a dorso-prevelar approximant, pronounced with a secon-
dary pharyngeal constriction and often with some lip-rounding (the tongue-
blade may be raised as well).10 Lass speculates that other realisations of /r/
known from various accents of Modern English (alveolar trills and taps, apical
or uvular approximants and fricatives) are mostly post-sixteenth-century inno-
vations (Lass 1983: 82).
9 Assuming further that /l/ in syllable codas was velarised ([ɫ]) and therefore also counted as
back. The role of backness, however, seems to have been overestimated. Studies of vowel
changes in the modern Germanic languages show that uvular rhotics do not tend to produce the
desired effect, i.e. the diphthongisation of a preceding front vowel with the retraction of its second
portion, while coronal and dorso-prevelar approximants often do (Howell 1991).
10 The chief articulatory component of this consonant is an approximant that roughly corre-
sponds to the high central vowel [ɨ]; Lass (1983) describes it as “velar”, which is not quite accu-
rate and potentially confusing (the actual articulation can be described as prevelar or, depending
on one’s preferred terminology, as retracted palatal, midpalatal or “central”). Since IPA lacks a
convenient symbol for such a sound, I shall use the ad hoc notation [ȝ] (Latin small letter yogh) in
this article to refer to “bunched” /r/ and to distinguish it from the genuinely retroflex (i.e.
subapico-prepalatal) variety. Laver (1994) uses [ψ] instead, but in my opinion the symbol pro-
posed here has a mnemonic advantage because of its resemblance to [] and [], the typical re-
flexes of vocalised rhotics.

A Shibboleth upon Their Tongues … 69
It should be noted that Lass’s reconstruction fails to explain why OE /r/ con-
sistently patterns with coronals. The cluster /rd/ counts as homorganic, produc-
ing vowel lengthening, e.g. in OE beard › bēard › ME bęrd ‘beard’. Given the
protean character of Lass’s /r/, one would expect it to behave like a dorsal as
well – in fact, presumably more like a dorsal than anything else – yet we find
neither homorganic lengthening before medial -rg- /r/ nor Anglian smoothing
before -r(r) (cf. dēor ‘animal’ vs. nēh ‘near’).
According to Hogg (1992: 40 [§2.74]), the West Germanic merger of *r and
*z and the present-day pronunciation of English /r/ suggest, as the most likely
possibility, a coronal fricative, flap or tap in Old English, with a “velarised”
allophone in coda positions – in other words, articulations similar to those found
in some varieties of Modern English.
Scepticism about the archetypical status of the alveolar trill has been ex-
tended well beyond the Old English chronological horizon. Catford (2001) ar-
gues that sound changes in some of the early Indo-European languages are more
compatible with the assumption of an approximant articulation, possibly the
“bunched” variety ([ȝ]). Denton (2003) examines the various effects of /r/ in
early Germanic dialects and concludes that, whereas Proto-Germanic *r may
have begun as an apical trill at least in onset positions, it was definitely weak-
ened in postvocalic positions in North and West Germanic, developing ap-
proximant allophones. As for Old English /r/, Denton’s conclusions are similar
to Hogg’s, although she is more specific about the pronunciation of the ap-
proximant variant:
In fact, from the Old English and pre-Old English sound changes considered here,
there is not compelling evidence that Old English ever had an apical trill in any
environment. On the contrary, the Old English evidence points to a possible apical
tap/flap in strong syllable positions and more certainly to a rounded central ap-
proximant with a relatively high tongue position in rhyme position, an articulation
which may also have been present in Old Saxon and parts of Dutch
(Denton 2003: 39).
Both Denton and Hogg emphasise the natural variability of rhotics. Denton
remarks that “the rhotics of the modern European Germanic languages include
virtually every conceivable rhotic articulation from labial fricatives to uvular
trills and pharyngealized approximants” (2003: 40) and goes on to say that it
would be unrealistic to expect the same phonetic value of Germanic *r across
all early dialects. Hogg makes a similar point about internal variation within
English:

P. Gsiorowski
70
One must also accept that it is unlikely that all OE speakers had the same or very
similar pronunciations of the central approximant, given the variability in the pro-
nunciation of the sound both in later periods of the language and in PDE [= pre-
sent-day English – PG] (Hogg 1992: 41 [§2.74, n.1]).
Tristram (1995) analyses the distribution of retroflex [ɻ] in the modern ac-
cents of England and its diffusion across the English Channel into the Breton
and French accents of Trégor. She finds that the area where [ɻ] is, or used to be,
a typical realisation of /r/11 coincides with the extent of the power of Wessex in
the late 9thc. (including West Mercia). Tristram conjectures that retroflex [ɻ] was
an innovation that arose spontaneously in West Saxon Old English and became
the ordinary West Saxon pronunciation of /r/ in all positions. It freely spread
throughout the area controlled politically by Wessex but was checked at the
border of the Danelaw. Tristram refrains from detailed speculation about the
Danelaw pronunciation in Old English times, but notes that the eventual devel-
opment of /r/ in that area in modern times has been the apico-(post)alveolar
approximant or fricative [ɹ] with a tendency to be weakened and lost in coda
positions (as in Received Pronunciation and other non-rhotic varieties of Eng-
lish). The occurrence of dorso-prevelar [ȝ] in British English is not discussed by
Tristram, and as far as I am aware has not been systematically investigated by
phoneticians. However, given the fact that [ɻ] and [ȝ] are nearly indistinguish-
able even to a well-trained ear,12 and that [ȝ] had gone practically unnoticed
until the 1950s despite being by far the dominant variant in American English,
it would hardly be surprising if [ȝ] turned out to be a frequent realisation of /r/
in the English West Country (and possibly elsewhere) as well. As Alwan – Na-
rayanan – Haker (1997) point out, the “canonical” retroflex and bunched rhotics
may be treated as two extremes of a continuous spectrum of tongue shapes
rather than discrete articulatory categories. Indeed, it is hard to see what could
compel speakers to select consistently only one of a number of possible articu-
latory strategies that produce practically the same acoustic effects; therefore,
inter-subject and allophonic variability encompassing [ȝ] and [ɻ] as well as any
intermediate articulations involving both dorsal and coronal gestures is probably
their natural mode of existence.
11 Especially in coda positions. Word-initially, retroflex [ɻ] seems to have lost much ground to
the apical approximant [ɹ] of the standard accent, and is now confined to Hampshire, Wiltshire,
Dorset, Somerset, Devon and Cornwall (LAE; Wells 1982: 342). Full or variable rhoticity with
final or preconsonantal /r/ realised as the retroflex colouring of vowels occurs throughout south-
western England, according to the LAE.
12 Except, perhaps, for their different coarticulatory effects (for example, the presence of [ȝ] can
be suspected if the /r/ in question has a distinct palatalising influence on a preceding coronal stop,
making the /t/ in /tr/ sound like [ʧ] – a type of pronunciation which is definitely widespread in
Britain.

A Shibboleth upon Their Tongues … 71
4. Regional rhotics in Old English?
4.1. West Saxon vs. Anglian
If Old English /r/ was variable, we should consider the possibility that different
variants were preferred in different dialectal areas. In order to test this hypothe-
sis, let us see if it offers any insights into dialect-specific sound changes involv-
ing rhotics or conditioned by their presence. Pre-r breaking seems to have been
more consistent in West Saxon than it is in the Anglian dialects. In particular,
breaking before r-clusters failed in Anglian (but not in West Saxon) if there was
an *i or *j in the next syllable, as in afirran ‘drive out, remove’ vs. WS afyrran
with y ‹ *io affected by i-umlaut (Hogg 1992: 90 [§5.24], 133 [§5.83]). This
suggests that the quality of West Saxon /r/ was more antagonistic to the antici-
patory effect of palatal segments. Interestingly, it seems that breaking occurred
regularly even in Anglian before *rr ‹ *rz, as in Angl. iorre ‘anger’ ‹ *irzija-.
Denton (2003: 25-27) argues that the Old English rhotic reflex of *z (conjectur-
ally reconstructed by her as a “central approximant”, i.e. my [ȝ]) was initially
different from inherited *r, and that the difference was still relevant in the an-
cestor of Anglian at the time of breaking. If the original contrast was between
an anterior (apico-alveolar) sound, [r ~ ] or [ɹ], and a posterior (dorso-prevelar
or retroflex) one, [ɻ ~ ȝ], it is imaginable that the merger of the two rhotics
yielded different results in different dialects, with West Saxon generally favour-
ing the posterior quality (as argued by Tristram 1995) and Anglian retaining the
anterior articulation at least in prevocalic positions. Such a scenario would be
consistent with the modern dialectal distribution of the realisations of /r/ in Brit-
ish English as described by Wells (1982: 341-343, 368, 372) and mapped in
LAE.13
The degree of lip-rounding accompanying the articulation of /r/ (lowering
the F3 and thus enhancing the auditory cues for a retroflex or “central” ap-
proximant) was perhaps greater in West Saxon as well, hence the frequent u-
colouring of short vowels and diphthongs (especially eo) as a combinative ef-
fect of a preceding /w/ and a following /r/ in Late West Saxon (sweord ~ swurd
etc., cf. Hogg 1992: 205-207 [§§5.183-187]).
The transparency of Anglian /r/ with respect to the influence of a following *i
or *j suggests a relatively “colourless” rhotic. This explains both why /r/ itself did
not cause Anglian smoothing and why an intervening /r/ did not block the mo-
nophthongising effect of a following velar (/k, x, /), as in Angl. werc ‘work’ or
ðwerh ‘crooked’. If Anglian smoothing is interpreted as the rightward spread of a
palatalising prosody (cf. Hogg 1992: 143-145 [§5.93]), this and the failure of
breaking before *rCi/j can be viewed as two aspects of the same phenomenon.
13 Especially maps Ph224, Ph225, Ph244 and Ph245.

P. Gsiorowski
72
The lowering and retraction of Middle English /e/, which yields /a/ before
preconsonantal or final /r/, was initiated in the North ca. 1300 or earlier (Wełna
1999: 56-57, 60). The diffusion of the change into the southern dialects and
London English took place in the course of the fifteenth century. Since a pha-
ryngeal constriction tends to pull vowels in the direction of [] by lowering their
F2 and raising their F1, the effect could be attributed to a pharyngealised rhotic
in the source area of the innovation. The retraction of *æ to OE // before /r/,
catalysed by the vicinity of a labial consonant (warþ ‘he became’, harm
‘harm’), is also Anglian (found in Northumbrian and in the early Mercian glos-
saries, cf. Hogg 1992: 92 [§5.29]).14 Residual pharyngealisation retained after
the loss of the coronal gesture may be responsible for the phonetic alterations
(lowering, retraction, diphthongisation with a fairly open [ə]-glide) affecting
vowels originally followed by tautosyllabic /r/ in non-rhotic accents of English.
It is perhaps significant that vowel lengthening before OE /rd/ never oc-
curred in distinctively Late West Saxon forms such as swurd ‘sword’ or wurd
‘word’ (here u was indubitably a short vowel, cf. Hogg 1992: 213 [§5.203]). If
the lengthening effect of /rd/ was stronger in the North, it may have been so
because of the stronger articulatory affinity of /r/ and /d/ in the Anglian dialects.
4.2. Northumbrian
The “Northumbrian burr” must be an old phenomenon. As early as the 1720s
the inhabitants of Northumberland clearly regarded it as a time-honoured fea-
ture of their dialect and were proud of it precisely by reason of its presumed
antiquity. So much, at any rate, can be inferred from the earliest known descrip-
tion of the burr, left by Defoe:
I must not quit Northumberland without taking notice, that the Natives of this
Country, of the antient original Race or Families, are distinguished by a Shibbo-
leth upon their Tongues in pronouncing the Letter R, which they cannot utter
without a hollow Jarring in the Throat, by which they are as plainly known, as a
Foreigner is in pronouncing the Th: this they call the Northumberland R, or
Wharle; and the Natives value themselves upon that Imperfection, because, for-
sooth, it shews the Antiquity of their Blood
(Defoe 1724-27: 232-233; quoted in Påhlsson 1972).
Påhlsson’s (1972) monographic study of the Northumbrian burr confirms
that the usual modern pronunciation of burred /r/ is a voiced uvular fricative []
in all environments. A voiced velar fricative is used by some speakers, and
other occasional variants include a uvular tap or even a uvular trill, although the
14 Note also the widespread retraction of early Modern English /æ/ to // before tautosyllabic /r/.

A Shibboleth upon Their Tongues … 73
latter is very rare. The burr is nowadays giving way to the mainstream realisa-
tions of /r/ in the north of England. It has already receded from the metropolitan
Geordie accents of Tyne and Wear, where it used to occur. Its present-day range
approximately coincides with the modern county of Northumberland, extending
into northern County Durham and a coastal strip of the southern Borders region.
The accents that have switched to apical [ɹ] have generally become non-rhotic,
whereas those with uvular [] retain it in non-prevocalic positions, where it
tends to coalesce with the preceding vowel, making it uvularised. This kind of
coarticulation is easy to achieve only if the vowel is back, hence the strong re-
tracting effect of non-prevocalic []. Wells (1982: 396) cites [bɔʶːdz] birds and
[wɔʶːmz] worms as characteristic Northumbrian forms, and LAE (maps Ph244
and Ph245) records [ɔʶ] as the normal pronunciation of unstressed -er in north-
ern Northumberland. Tyneside accents, although no longer rhotic, show similar
effects in work [wɔːk], first [fɔːst] etc. (Wells 1982: 374).
Interestingly, in the Northumbrian dialect of Old English (and only there) *e
was regularly retracted to /o/ in the sequence /wer/ when followed by a coronal
or labial consonant: *werþa- › worþ ‘property’ = non-Nbr. weorþ (cf. Hogg
1992: 93 [§5.30]). The similarity to the modern rounding (which is not condi-
tioned by the consonants preceding the vowel or following the /r/) may be for-
tuitous, but there is another Old English change that suggest a special pronun-
ciation of Northumbrian /r/: rhotic metathesis in /Vrxt/ sequences (berht › breht
‘bright’). It occurs sporadically in West Saxon (e.g. wrohte for worhte
‘worked’), but is otherwise a Late Northumbrian speciality (Hogg 1992: 303
[§7.95]). It is also the only kind of regressive rhotic metathesis found in Old
English dialects; the process normally targets prevocalic /r/ (as in *βrannjan- ›
bærnan ‘kindle, burn’). As an alternative to full transposition, Northumbrian
spellings sometimes show an anaptyctic vowel that breaks up the sequence /rx/
(also in early texts: berec(h)t ‘bright’, uyrihta ‘maker’).15 It seems that the /rxt/
cluster, while tolerated elsewhere in Old English, was for some reason avoided
in Northumbrian. This would be entirely natural in a dialect with /r/ realised as
uvular [] or a similar sound, whose combination with a directly following /x/
would be very awkward both articulatorily and perceptually.16 The Northum-
brian metathesis or anaptyxis affecting /Vrxt/ can thus be explained as a process
akin to dissimilation, eliminating a phonetically difficult cluster. As there is no
reason to believe that /xt/ was pronounced differently in different Old English
accents, the awkwardness of the /rxt/ sequence in Northumbrian is likely to
have been caused by a distinct regional pronunciation of /r/.
15 We are probably dealing with successive stages of a single process: berht › bér[ə]ht › b[ə]réht
› breht.
16 Cf. Blevins – Garrett (2004, fn. 14).

P. Gsiorowski
74
5. Conclusion
The picture that emerges from the discussion above is as follows: The pronun-
ciation of /r/ was not uniform across the network of Old English dialects. The
prevalent West Germanic value of inherited *r (an apical trill or tap prevocali-
cally, a weakened approximant [ɹ] finally and preconsonantally) was preserved
most faithfully in Anglian (and perhaps Kentish). Against this background there
arose two local innovations: a posterior realisation of /r/ as [ɻ ~ ȝ], which spread
across southern England from its West Saxon birthplace, and a uvular or velar
pronunciation in the northeast – the ancestor of the Northumbrian burr. If we
assume that by Late Old English times both innovations had already reached the
status of regional “shibboleths” well entrenched in their respective home areas,
quite a number of phonological peculiarities characterising Old English dialects
become easier to explain. While it would be extremely naive to treat a modern
linguistic atlas of England as a reliable guide to Anglo-Saxon dialectal geogra-
phy, one should avoid the other extreme, which would be to deny that there is a
strong element of regional continuity despite all the linguistic changes and
boundary shifts that have taken place since the Old English period. In fact, there
is sufficient evidence to justify the belief that some dialectal phenomena, in-
cluding the distribution of certain characteristic allophones across the regional
accents of English, may be older than traditionally believed and surprisingly
stable in the historical perspective.
REFERENCES
Alwan, Abeer – Shrikanth Narayanan – Katherine Haker
1997 “Towards articulatory-acoustic models of liquid consonants. Part II: The rhotics”,
Journal of the Acoustical Society of America 101/2: 1078-1089.
Blevins, Juliette – Andrew Garrett
2004 “The evolution of metathesis”, in: Bruce Hayes – Robert Kirchner – Donca Steriade
(eds.), 117-56.
Catford, J. C.
2001 “On Rs, rhotacism and paleophony”, Journal of the International Phonetic Associa-
tion 31/2: 171-185.
Davenport, Michael – Eric Hansen – Hans Frede Nielsen (eds.)
1983 Current topics in English historical linguistics. Odense: Odense University Press.
Defoe, Daniel
1724-1727 A tour thro’ the whole island of Great Britain. London: [No indication of publisher.]
Denton, Jeanette Marshall
2003 “Reconstructing the articulation of Early Germanic *r”, Diachronica 20/1: 11-43.

A Shibboleth upon Their Tongues … 75
Dobson, Eric J.
1968 English pronunciation 1500-1700. 2 vols. Oxford: Clarendon Press.
Erickson, Blaine
2003 “On the development of English r”, in: Donka Minkova – Robert Stockwell (eds.),
183-206.
Fisiak, Jacek
1967 “The Old English ‹wr-› and ‹wl-›”, Linguistics 32: 12-14.
Fisiak, Jacek (ed.)
1995 Linguistic change under contact conditions. Berlin – New York: Mouton de Gruyter.
Fromkin, Victoria A. (ed.)
1985 Phonetic linguistics: Essays in honor of Peter Ladefoged. New York: Academic
Press.
Hall, Tracy Allan (ed.)
2001 Distinctive feature theory. Berlin: de Gruyter.
Hayes, Bruce – Robert Kirchner – Donca Steriade (eds.)
2004 Phonetically based phonology. Cambridge: Cambridge University Press.
Hogg, Richard M.
1992 A grammar of Old English 1: Phonology. Oxford: Blackwell.
Howell, Robert B.
1991 Old English breaking and its Germanic analogues. Tübingen: Max Niemeyer.
Jespersen, Otto
1909 A Modern English grammar on historical principles. Vol. 1: Sounds and spellings.
Heidelberg: Carl Winter.
Ladefoged, Peter – Ian Maddieson
1996 Sounds of the world’s languages. Oxford: Blackwell.
LAE = Orton, Harold – Stewart Sanderson – John Widdowson
1978 The linguistic atlas of England. London: Croom Helm.
Lass, Roger – John M. Anderson
1975 Old English phonology. Cambridge: Cambridge University Press.
Lass, Roger
1983 “Velar /r/ and the history of English” in: Michael Davenport – Eric Hansen – Hans
Frede Nielsen (eds.), 67-94.
Laver, John
1994 Principles of phonetics. Cambridge: Cambridge University Press.
Lindau, Mona
1985 “The story of /r/”, in: Victoria A. Fromkin (ed.), 157-168.
Minkova, Donka – Robert P. Stockwell (eds.)
2003 Studies in the history of the English language. A millenial perspective. Berlin – New
York: Mouton de Gruyter.
Mossé, Fernand
1945 Manuel de l’anglais du moyen âge 1: Vieil-anglais. Paris: Éditions Montaigne.
O’Connor, J. D.
1973 Phonetics. Harmondsworth: Penguin Books.
Oostendorp, Marc van
2001 “The phonology of postvocalic /r/ in Brabant Dutch and Limburg Dutch”, in: Hans
van der Velde – Roeland van Hout (eds.), 113-122.

P. Gsiorowski
76
Påhlsson, Christer
1972 The Northumbrian burr: A sociolinguistic study. Lund: CWK Gleerup.
Quirk, Randolf – C. L. Wrenn
1957 An Old English grammar. New York: Holt, Rinehart & Winston.
Reszkiewicz, Alfred
1953 “The phonemic interpretation of Old English digraphs”, Biuletyn Polskiego To-
warzystwa Językoznawczego 12: 179-187.
Sweet, Henry
1888 History of English sounds (from the earliest period). Oxford: Clarendon Press.
1952 Anglo-Saxon primer. (9th edition, revised by Norman Davis.) Oxford: Clarendon
Press.
Tristram, Hildegard L. C.
1995 “Linguistic contacts across the English Channel: The case of the Breton retroflex ‹r›”,
in: Jacek Fisiak (ed.), 291-314.
Velde, Hans van der – Roeland van Hout (eds.)
2001 Sociolinguistic, phonetic and phonological characteristics of /r/. Bruxelles:
ILPV/Université Libre de Bruxelles.
Wells, John C.
1982 Accents of English 2: The British Isles. Cambridge: Cambridge University Press.
Wełna, Jerzy
1999 “‘Downs and ups’ of short [e] before nonprevocalic [r], or Late Middle English e-
Lowering”, Studia Anglica Posnaniensia 34: 55-72.
Wiese, Richard
2001 “The phonology of /r/” in: Tracy Alan Hall (ed.), 335-368.