ArticlePDF Available

Where Do English Sibilant Plurals Come From?



Very early in Middle English, texts especially in the North and East, tend to use an orthographic suffix-(e)s for noun plurals, in Southern and Western texts the plural suffix-(e)n of the Old English weak declension at first spreads, but then by 1300 also yields to-(e)s. This essay first shows that on phonological and phonetic grounds this-(e)s, which remains the productive plural in Modern English, must, as a vocabulary item, be lexically specified as +Voice; it is not voiced by any progressive assimilation process in synchronic derivations. The source of this underlying voiced sibilant-z, completely absent in Old English, is to be found in the genealogical ancestor of Middle English, Proto-Scandinavian, whose plural in all non-neuter declensions is precisely this segment (Haugen 1982). The present essay argues that this form was an integral part of the Norse brought to England by the earliest Scandinavian settlers in the 9th c. In all likelihood, the later change in Mainland Scandinavian of this-z to-r, completed in the 12th c., failed to establish itself in the Anglicized Norse of England, due to sociolinguistic factors akin to those set out in the classic paper of Labov (1963).
Where Do English Sibilant Plurals Come From?
Joseph E. Emonds,
In Joseph Emonds, Markéta Janebová, and Ludmila Veselovská, eds. Language Use
and Linguistic Structure. Proceedings of the Olomouc Linguistics Colloquium
2018. Olomouc: Palacký University, 201 9.
Abstract: Very early in Middle English, texts especially in the North and East, tend to
use an orthographic suffix (e)s for noun plurals, in Southern and Western texts the
plural suffix (e)n of the Old English weak declension at first spreads, but then by 1300
also yields to (e)s. This essay first shows that on phonological and phonetic grounds
this (e)s, which remains the productive plural in Modern English, must, as a vocabulary
item, be lexically specified as +Voice; it is not voiced by any progressive assimilation
process in synchronic derivations. The source of this underlying voiced sibilant z,
completely absent in Old English, is to be found in the genealogical ancestor of Middle
English, Proto-Scandinavian, whose plural in all non-neuter declensions is precisely this
segment (Haugen 1982). The present essay argues that this form was an integral part of
the Norse brought to England by the earliest Scandinavian settlers in the 9th c. In all
likelihood, the later change in Mainland Scandinavian of this z to r, completed in the
12th c., failed to establish itself in the Anglicized Norse of England, due to
sociolinguistic factors akin to those set out in the classic paper of Labov (1963).
Keywords: Common Scandinavian; English plurals; Middle English inflection; Old
English plurals; Proto-Scandinavian; Voicing Assimilation; Vowel syncope
1. Middle and Modern English Noun Plurals
The Modern English noun suffix, spelled (e)s, became the regular and productive way
to form plurals in (early) Middle English (ME).
Other than in conservative southern and
western dialects, which were a closer continuation of Old English (OEalso known as
West Saxon), this usage was already established from 1200 onwards; this was noted
already in White (1852, xxii). The detailed summary of Baugh and Cable (2013, chap. 7)
merits reproduction in full.
In early Middle English only two methods of indicating the plural remained fairly
distinctive: the s or es from the strong masculine declension and the en (as in
oxen) from the weak (see § 41). And for a time, at least in southern England, it
would have been difficult to predict that the s would become the almost universal
sign of the plural that it has become. Until the 13th c. the en plural enjoyed great
favour in the south [the productive, default, so-called “weak” OE plural—JE],
being often added to nouns which had not belonged to the weak declension in Old
I am particularly grateful to Kristina Smejová for discussions on Section 3.4. I thank Simin
Karimi for organizing a presentation at the University of Arizona in January 2018, and the
audiences there, at the Fourth Olomouc Linguistics Conference, and at the 20th International
Conference on English Historical Linguistics in Edinburgh for helpful commentary.
Baugh and Cable’s passage ends with a Darwinian flourish. There is of course no non-circular
reason to consider s as “more fit” than –n for survival as a plural suffix. The metaphor reflects the
fact that the authors find no internal linguistic motivation for the change.
English. But in the rest of England the s plural (and genitive singular) of the old
first declension (masculine) was apparently felt to be so distinctive that it spread
rapidly. Its extension took place most quickly in the north. Even in Old English
many nouns originally of other declensions had gone over to this declension in the
Northumbrian dialect. By 1200 s was the standard plural ending in the north and
north Midland areas; other forms were exceptional. Fifty years later it had
conquered the rest of the Midlands, and in the course of the fourteenth century it
had definitely been accepted all over England as the normal sign of the plural in
English nouns. Its spread may have been helped by the early extension of s
throughout the plural in Anglo-Norman, but in general it may be considered as an
example of the survival of the fittest in language.
This view is not modified, certainly not in any essentials, in more recent work such as
Fulk (2012). This use of the spelling s to mark plurals can be seen in the book The
Ormulum (c. 1200) and other 13th c. work (Watts 2011, 110).
As is well known, this same plural suffix in Modern English, call it Z, has three
allomorphs, which are without exception conditioned by the final segment of a noun
(1) Allomorphs of the plural morpheme Z: /əz/ after final sibilant segments,
then /z/ following final voiced segments and /s/ following final unvoiced
As far as I know, it is reasonable to assume that the phonetic alternation between voiced
and unvoiced allomorphs has been present from the earliest Middle English uses of Z.
Analyses of Modern English are quite aware of the fact that three other
inflectional morphemes have exactly the same phonetic forms as (1):
(2) (a) The third singular agreement suffix on present tense verbs, referred to
here as Z, has the same allomorphs as the noun plural Z.
(b) The contracted form s of the third singular copula is has the same
allomorphs as the noun plural Z.
(c) The possessive s, referred to here as Z, has the same allomorphs as
the plural Z.
Verb forms such as chooses, holds, and thinks exemplify (1) for (2a). One can easily
exemplify the same patterns for (2bc):
(3) (a) Contracted and possessive /əz/ after final sibilants:
The Church’s still fixated on the past.
The Church’s strong fixation on the past
(b) Contracted and possessive /z/ after voiced segments:
The Cardinal’s still fixated on the past.
The Cardinal’s strong fixation on the past
The voiced sibilant ending on English possessive pronouns (his, hers, its, whose, (y)ours, theirs)
could as well be spelled ’s, since this allomorphs are exactly those in (3b).
(c) Contracted and possessive /s/ after final unvoiced segments:
The Pope’s still fixated on the past.
The Pope’s strong fixation on the past
Moreover, almost all analyses agree that the underlying form of all these forms should
be the same. For an overview of their arguments, including a minority position about
(2c) unrelated to concerns here, see Zwicky (1975).
2. The Lexical Representation of the English Plural
The underlying phonological form of the English plural Z (and of Z’ and Z as well)
must be voiced z, rather than either unvoiced s or a “neutralized” sibilant unspecified
for voicing. Several papers rather conclusively argued for this lexical z in the 1970s,
Lightner (1970), Sloat and Hoard (1971), and Shibatani (1972), so this conclusion seems
established. For concreteness, I formulate here three arguments that unequivocally
support this conclusion, including one which I do not think has previously been made in
strong enough or general enough terms.
A first argument is that the phonetic behaviour of plurals (2a) is exactly the
same as the contracted allomorphs of the free morpheme is (2b). The copula’s final
consonant, in its uncontracted lexical form, is always voiced. Contraction consists in
simply dropping the vowel, yielding the Cardinal is
the Cardinal’s (no change in the
underlying voiced sibilant). But when the preceding consonant is Voice, then devoicing
must change the phonetic z to s: the Pope is
the Pope’s. If the lexical forms of Z,
Z’, and Z are all +Voice, the exact same analysis (phonetic devoicing of underlying z)
accounts for their allomorphs as well, with no added stipulation.
A second argument concerns the several irregular plurals of nouns ending in f:
calves, hooves, knives, leaves, loaves, scarves, selves, shelves, wolves, etc. 3rd singular
verbs and possessives are unaffected: she loafs around; a wolf’s fur. Mossé (1952, 39)
and other researchers have hypothesized that the final f of these roots was voiced
between vowels in ME. However, the vowel in the ending was dropped by 1400 at the
latest, resulting in irregular morphemes with an f/v lexical alternation for singulars vs.
plurals (Lass 2006, 5960). Given the many centuries that no vowel has followed these
v, today’s synchronic (and still learnable) analysis must be different.
To begin, today’s alternation must be lexically stipulated with these roots. Over
the centuries there has been no general tendency for the voiced allomorphs to generalize
phonetically, either before vowels/ sonorants in (4i) or in plurals (4ii):
(4) (i) leafy, stuffy, beefy, goofy, toughie, selfish, loafer, loafing, oafish
(ii) bluffs, briefs, cliffs, cuffs, foodstuffs, puffs, reefs, spoofs, toughs
Now, if the underlying plural segment in the irregular pairs were either unvoiced or
unspecified for voicing, these plurals would be completely irregular, since the voicing of
the final consonant sequence (vz) could not be related to any other source in English
phonology. However, this voiced sequence can be related to an underlying . . . fz by
regressive voicing assimilation, While not productive in English, this universal tendency
is sporadically found elsewhere in the language (and often reflected in spelling) in e.g.
halv-ed, lous-y, spas-m, fif-th, fif-teen, lef-t, twelf-th, leng-th.
Before continuing to the third argument, these first arguments suggest a lexical
entry for the English plural morpheme. (Because the notion “alveolar” may well combine
more than one phonological feature. I do not write it with ± in this paper.)
(5) Lexical entry for the English plural z.
PLUR, N____, [Alveolar, +Continuant, +Voice, Sonorant]
The third argument for the Voice feature in (5) concerns a cross-linguistic restriction on
voicing assimilation. Many texts, looking for an instance of “progressive assimilation”
readily give English plurals as an example of a rule that spreads the voice feature of a
stem-final segment rightward to a bound suffix’s first (or only) consonantal segment.
But more generally, progressive assimilation, particularly of the value +Voice,
is quite rare in the world’s languages (Lombardi 1999; Borowsky 2000). In fact, English
excludes progressive voicing entirely in any compounds or any suffixes other than the
inflection under discussion and the parallel regular past tense d (see again note 3).
(6) No rightward phonetic spreading of +Voice in English:
him-self (*himzelf), special-ty, frail-ty (*special-dy, *frail-dy), lambkin;
four-th, nin-th, ten-th, leng-th, wid-th, tru-th, heal-th (all exclude a voiced th);
contain/content, restrain/restraint, high/height, weigh/weight (never final -d);
spoon-ful, hand-ful, dread-ful, care-ful (ful never assimilates to voiced *vul);
Bingham-ton, Washing-ton, Middle-ton (ton never becomes *don);
John-son, Richard-son, William-son (son never assimilates to *zon);
hand-some, fear-some, loath-some, cumber-some (some never becomes
The diverse sources of the morphemes in these combinations testify to the fact that
Middle and Modern English have never had any phonetic “tendency, even slight, to
spread voicing of a final segment to a following morpheme in the same word.
These data strongly suggest that, throughout history, the voicing in the English
noun plural (5) has been underlying (i.e. in a lexical entry) rather than due to a
derivational process. Minkova (2014, 89) argues that similarly, voicing of the alveolar
stop of the regular English past tense is due to its lexical entry. But if both these
inflections (z and d) are underlyingly voiced, the robust data pattern in (6) essentially
forces the following conclusion:
(7) Progressive voicing ban. No progressive assimilation in English introduces
Regressive voicing assimilation is widespread in the world’s languages, though the relatively few
instances in Modern English appear to be isolated remnants of earlier sound changes: Current
English has plenty of contexts, even with bound morphemes, where no voicing assimilation
happens: childhood, dreadful, dukedom, Falklands, handsome, handful, and Scotland are a few of
many possible examples.
In this paper, * before a form uniformly means “ill-formed” rather than “unattested.”
As in all other positions in English, bound morphemes have voiceless s as their initial lexical
segment (self, son, some) rather than the voiced z in (5).
On the other hand, it might still appear that the other feature value Voice can spread
rightward in English, so as to account for the voiceless allomorph /s/ of Z, Z’ and ‘Z, as
well as the voiceless allomorph /t/ of the regular English Past Tense. However, we can
show that this is also a misconception.
We have seen that the English regular plural morpheme has long contained an
underlying voiced sibilant z. Voicing on this plural morpheme disappears only if the
final segment of the noun is voiceless: cats, naps, cliffs, rocks. However, this devoicing
is not due to some morpheme-particular “rightwards de-voicing.” The lack of voicing in
this context on all the Z morphemes has its source in a more general, probably universal
restriction which is moreover bi-directional. Consider for example clauses which begin
with an optionally contractible singular copula is. The second column is a (perhaps not
standard) spelling of the contraction, and the third represents it phonetically:
Is Dave coming back?
’s Dave coming back?
/z/ Dave coming back?
Is Beth coming back?
’s Beth coming back?
/z/ Beth coming back?
Is Ann coming back?
’s Ann coming back?
/z/ Ann coming back?
Is Ed coming back?
’s Ed coming back?
/z/ Ed coming back?
Unsurprisingly, all these contracted forms retain their lexical feature +Voice.
But now what happens when the subject begins with an unvoiced segment? The
fully contracted form (with no vowel) must be unvoiced:
’s Ted coming back?
/s/ Ted coming back?
*/z/ Ted coming back?
’s Fanny coming back?
/s/ Fanny coming back?
*/z/ Fanny coming back?
The following general restriction, plausibly valid across at least a range of languages,
suffices to describe the loss of voicing in the contracted English copula is, regardless of
whether it precedes or follows a host morpheme in the same phonetic word:
(10) Cross-linguistic Voicing Restriction. Voicing is not realized in positions
separated from all Sonorant segments in the same word by a voiceless
For a recent general justification of the feature Sonorant, see Kaisse (2011). I am not
assuming that the “sonoricity” of phonological segments must be graded along a scale;
the feature Sonorant as used here can as well be purely binary; i.e. vowels, glides, and
voiced liquids and nasals are sonorants and other segments are not.
This formulation (10) is designed to make my use of it later transparent.
Nonetheless, this restriction might still be a special case or corollary of some Sonority
Sequencing Principle as in Clements (1990), which forbids a more sonorant segment
being external in a syllable to a less sonorant one.
Though there are debates as to how
The contractions discussed in this section are not separate words, since they have no vocalic
nucleus, as in (8)(9). Generally an English contraction must be part of a preceding word, but
when contraction is allowed clause-initially, it becomes part of the following word.
If English voicelessness results from the feature Spread Glottis, as argued in Iverson and
Salmons (1999), then Restriction (10) is equivalent to saying that when this feature is present in a
syllable, it impedes any voicing external to it (in either direction).
I am taking for granted here rather traditional uses of these feature labels. It may be that the
English lexically voiced suffixes are voiced phonetically only by virtue of a neighbouring voicing.
voicing relates to sonority, essentially all accounts claim that voiced fricatives, which are
what concerns us here, are more sonorous than any voiceless segment.
The Voicing
Restriction therefore blocks realization of Voice on a fricative (i.e. the lexical z) in e.g.
cats/ coughs/ tricks. It is irrelevant that these segments are adjacent to a voiced segment
inside a following word.
The Voicing Restriction (10) now automatically explains the devoicing not only
of all contracted English singular copulas. It equally well accounts for the voiceless
allomorphs of the English plural morpheme Z and its homophones Z’ and Z, provided
they are all lexically specified as +Voice, This completes the third argument for +Voice
in the lexical entry (5).
Notice further that the Voicing Restriction applies regardless of the host being
on the left or right. It is more general than any prohibition of voicing formulated
explicitly or implicitly as a constraint on left to right (or right to left) scanning. For
example, any constraint formulation in terms of “turning on” or “turning off” voicing
during such scanning fails to capture the generalization that includes the voiceless
prefixation seen in (9).
Independent support for this analysis is that it also allows us to generalize (7):
(11) Ban on progressive voice assimilation rules. English has no rightward
phonetic assimilation to either value ±Voice.
This principle is thus an alternative to the apparent “progressive (de-) voicing
assimilation” in both the plural and past tense inflections of English. Such phonetic rules
are banned by (11).
In this case, (10) would be a special case of a more general phonetic property. This issue appears
related to ultimately determining which laryngeal feature should be used to characterize English
obstruent voicing. For discussion see Iverson and Salmons (1999).
Thus, the restriction as formulated in (10) does not itself depend on Sonority Sequencing, and is
even consistent with the claim in Henke, Kaisse, and Wright (2012) that such sequencing is an
epiphenomenon: . . . the patterns attributed to Sonority Sequencing are the result of a few broad
perceptually-motivated constraints which interact with other constraints and language-specific
lexical contrasts to yield the phono-tactics of particular languages.”
If the direction of scanning/ production of syllables is left to right, devoicing the prefixed
contractions in (9) should count as “turning voice off.” Then, not resuming voicing would imply
that an entire syllable with a devoiced prefixal onset would be expected, counter to obvious facts.
The Voicing Restriction (10) accurately avoids an implication of directionality.
3. A Diachronic Source for the English Plural z
3.1 Why the Source Is Not (Anglo-Norman) French
With regard to appearance of final voiced continuants such as z in ME, mention is
sometimes made of their presence in Anglo-Norman French. However, the lexical
entries for the latter were not borrowed in any significant number before the late 13th
century (Jespersen 1912; Classen 1919; Watts 2011, 110111), later than the appearance
in early ME of the sibilant plural. The serious influence of French vocabulary on English
thus occurs too late for this language to have been the source of something as
grammatically central as the ME plural.
There is a second and more telling reason why the ME plural suffix cannot be
ascribed to French. Despite its huge influence on later ME vocabulary, the fact remains
that English borrowed no French inflections (or grammatical free morphemes) at all.
More generally, borrowing of any inflection into a living, expanding language under
even intense language contact situations is extremely rare.
The idea that early ME
speakers in especially the north of England would borrow one of its most basic
inflections from French even before any open class vocabulary is linguistically
On the other hand, a general fact about noun plurals in current French can serve
to undermine a frequent presupposition about why ME z so quickly replaced OE n
plurals. Since final -n tended to drop in northern ME, it is sometimes speculated that
English somehow “needed” a new productive pronounced plural, a need filled by z. But
there is no general “need” for a productive inflectional noun plural (cf. their lack in
Chinese, Japanese); even in Indo-European which generally has them, Modern French
no longer has such a morpheme. On nouns its plural s is purely orthographic, not even
pronounced in liaison with a following initial vowel, e.g. in les magasins outverts the
stores open”. There is thus no structural reason why ME nouns, even if they had lost all
others, had to have a new pronounced plural.
3.2 Why the ME Plural z Is Plausibly Proto-Scandinavian
A motivated and more plausible source of the lexical z of the ME noun plural is the
Norse language brought to England by Scandinavian settlers between the early Viking
raids (before 800) and the Norman Conquest of 10661090.
Their language was in the
North Germanic (NG) branch of Indo-European. In contrast, the conclusion of
essentially all analysts is that Old English (West Saxon) was a West Germanic (WG)
By 837, today’s England was divided into West Saxon and Danish kingdoms
(the “Danelaw”). Scandinavian immigration into the latter region was extensive; see
The grammatical free morphemes very and much might seem exceptions, but very derives from
the French open class vrai “true” and much has a Proto-Germanic source.
For instance, in today’s American Southwest, there is not even a hint of its Spanish borrowing
any English verbal inflection, or vice-versa.
There are a few phonetically distinct irregular noun plurals in French, but none of them involve
pronouncing an s: chevaux, vitaux, yeux, oeufs, etc. have final vowels. Their number is comparable
to that of English plurals with vowel changes, e.g. feet, geese, mice, women.
For the demographics and economics of this extensive and continuous settlement, see Woodruff
(1974); Wood (1986); Townend (2002), and Kershaw (2017).
Map 1 for its density and location. Danelaw Scandinavians were numerous and
prosperous, reclaiming farmland from marshland (Lincolnshire) and establishing new
currencies and economic centers (East Anglia), some as far west as Leicester (Wood
1986; Kershaw 2017).
As is generally agreed, West Saxon (OE) and Norse co-existed in England well
into the 12th c. However, a century later, as far as surviving texts are concerned, Middle
English (ME) (with its disparate “dialects”) was the country’s sole surviving native
Germanic tongue.
In contrast to OE, ME in its syntax is typologically a North Germanic (NG), i.e.
Scandinavian language (Gianollo, Guardiano, and Longobardi 2008). On the basis of
evidence from over twenty syntactic constructions, Emonds and Faarlund (2014) argue
further that, counter to previously accepted classifications, ME descends directly from
NG Scandinavian, modified over the centuries in England to include extensive West
Saxon (OE) vocabulary. They refer to this branch of NG as “Anglicized Norse” (AN),
which is then synonymous with Early ME. The beginning of written AN is probably best
dated in the late 12th c., perhaps starting with the first book in AN, the monk Orm’s
Ormulum of c. 1200. This book is notable among other things for its wide use of the
nominal plural inflection s. For dating and discussion of the sharp ME break with OE,
i.e. West Saxon, see Watts (2011, chap. 34).
Following traditional scholarship, Emonds and Faarlund note that AN/ME lacks
much of the bound morphology of both Old English and Old Norse (ON). Thus, ME
quickly lost most agreement (except for 2nd and 3rd singular verbs).
these authors give four inflection-based reasons for deriving ME from an NG source.
(12) Norse sources for Modern English Inflections
(a) Both ON and ME replaced synthetic comparison on longer adjectives
(Germanic er and est) with analytic grading (English more, most; ON
meir, mest).
(b) The ME nominalizing suffix ing comes directly from ON (e.g. viking
walking) rather than from OE ung.
(c) WG infinitives are marked by a bound prefix (OE to, Dutch te, German
zu), while NG has free morphemes (ME to), i.e. NG infinitives can be
“split” by adverbs.
(d) ME and Mainland Scandinavia both eventually develop phrasal rather than
word-based genitive suffixes.
Given that verbal past stems in ON and OE are quite similar (Strang 1970, Ch.
IV), the only remaining inflection that might distinguish ME from Scandinavian is in
Here are the accepted names for stages of the languages germane to this paper. Note that Old
Norse corresponds to the time period of early Middle English.
Proto-Scandinavian (PS) until c. 850. Only the sparse evidence of runic inscriptions.
Common Scandinavian (CS), c. 8501150. More runic evidence, but still unwritten.
Old Norse (ON), written in Latin script, from 1150 onwards.
Old English or West Saxon (OE), written in mostly Latin letters, until 1150.
Middle English (ME), written in Latin script, from 1150 onwards.
For example: “In the North, the endings e and en on finite verbs are lost after the earliest
texts” (Fulk 2012, 74). Those ME plural verbs in (e)n that remain plausibly derive from the PS
3rd plural agreement (Haugen 1982, 122125). OE speakers did apparently import the now lost
2nd singular suffix st into ME.
fact the noun plural. The rest of this essay addresses this issue and concludes that its ME
form (e)z derives not from late OE but from an NG source. As in note 17, the relevant
contemporary of later OE, from 8501100, is not ON, but instead the earlier “Common
Scandinavian” (CS) rather sparsely documented in runic inscriptions. This stage of NG
in turn immediately follows reconstructed Proto-Scandinavian (PS), which is
contemporary with earlier OE. The sequence is thus Proto-Germanic PS (coeval with
early OE) CS (coeval with late OE) ON.
Haugen (1982) describes both these successive NG stages in some detail, and
his tables in chap. 45 (1982, 9091; 122125) unfailingly represent the PS plural as a
voiced sibilant z, which I write here as ż. We can now review why Haugen’s practice is
uncontroversial in NG scholarship.
(13) Justifications for taking CS ż as a PS Voiced Alveolar Sibilant
(i) The CS rune ż for both noun plurals and 2nd sing. verbs uniformly
corresponds to the Est Germanic Gothic endings z.
(ii) The final CS ż runes occur precisely after unstressed vowels, where
throughout Germanic they are predictably voiced by Verner’s Law.
(iii) CS ż has fully expected unvoiced non-Germanic cognates s in 2nd
sing. Agreement, Czech –š and Spanish s. Similarly for the cognate
Spanish noun plural s.
(iv) When CS ż dissolves into allophones of other ON phonemes (12th c.),
all of them are coronal and (except for s) all are voiced: d, n, l, r, s
(Haugen 1982, 62).
(v) During the CS period, the rune for nominal plurals and 2nd sing.
agreement was entirely distinct from runes for either r or s. This
“pitch-fork” rune (for a phonemic voiced continuant) persisted in
certain regions into the 12th c. (Haugen 1982, 5762)
Haugen’s tables of Proto-Scandinavian nominal inflections (1982, 9091) also indicate
that the most common nominal plural in non-neuter nominatives and accusatives (the
same forms used by traditional histories of English for ancestors of the noun plural), is
by far the same mono-segmental voiced z of Modern English plurals:
(5) Lexical entry for the Proto-Scandinavian and English plural z
PLUR, N____, [Alveolar, +Continuant, +Voice, Sonorant]
I therefore propose:
(14) Genealogical source of the English voiced plural
The productive English noun plural z descends directly from Proto-
Scandinavian z.
We have now established that both the Proto-Scandinavian plural and the ME z of entry
(5) are alveolar voiced continuants. Nonetheless, the productive Common Scandinavian
nominative plural suffix, also a descendant of PS final z, eventually became a Latin
alphabetic r in ON (1150 onwards). The development of ON thus involved a change
All the “strong nouns” in Haugen’s tables have this form, except that some masculine nouns
take n in the accusative. All non-neuter nominatives and all feminines take a z plural.
that distinguishes ME z from ON; the single feature difference between the two
segments is that ON r is sonorant, while the earlier (more archaic) z retained in ME is
3.3 Later Development of Proto-Scandinavian z
The change from the PS plural suffix z to a standard r in later ON and Mainland
Scandinavian, if one is too quickly influenced by orthography, may seem unrelated to
the English s. But since this s is just a spelling for a lexical z, and moreover
rhoticization (z
r) is widely attested in both North and West Germanic, it is not so
surprising that if Proto-Germanic final z in plurals could develop into r.
From this perspective, the pre-history and history of Middle and Modern
English plurals seems to be as follows:
(15) Step-by-step history of English plurals
(i) The modern noun plurals in English z and Scandinavian r (differing
only by ±Sonorant) both originate in I.-E. case/ number inflections that
contained s preceded by a long or lengthened vowel, e.g. I.-E. e:s
and o:s, etc.
(ii) These I.-E. inflectional long vowels on nouns were most often
(iii) After this, when Germanic stress moved to initial syllables; all final
sibilants in plurals become voiced, because of Verner’s Law but
possibly also by some “analogical levelling”.
(iv) No later than when NG short vowels dropped due to vowel syncope/
apocope in the 7th and 8th c. (Haugen 1982, 2829), voicing of the
plural sibilant z became distinctive, i.e. a lexical feature, as in the
lexical entry for the plural morpheme (5).
This last step preceded the bulk of Scandinavian settlement in England (c. 8501066).
That is, the settlers brought with them to England a noun plural inflection that was some
kind of coronal voiced continuant. The one uncertainty, to be discussed below, is: what
was its mode of articulation? Was it a fricative, a sonorant or something with features of
both? Whatever the answer, one can conclude that in NG languages, the final alveolar
continuant (with possibly some allophonic variation) that marks noun plurals has never
lost its voicing. By 1150, this continuant became r in ON (written in Latin script) and
current Mainland Scandinavian, but it remained an unchanged z (with allophones) in
Middle and Modern English.
3.4 The Proposed OE Precursor of the English Plural z
As can be inferred from the cited summary from Baugh and Cable (2013, chap. 7), no
possible OE ancestor of the ME noun plural is or contains the segment z. The process
of deriving z must then involve changes in representing the plural, which don’t arise for
the hypothesis (14). According to this simple proposal, the ME z is identical to the
same phoneme in PS, i.e. nothing happened to noun plurals between PS and ME.
This is transparently reflected in Latin descendants of I.-E. A two syllable noun has initial stress,
even if the second syllable has a long vowel. One can observe many unstressed long vowels in
final syllables in the Latin inflectional tables in Henle (1945, 213).
This obvious hypothesis has not previously found supporters among historians
of English (perhaps never crossing their minds). As remarked in Emonds and Faarlund
(2014), all detailed studies of ME assume without argument that outside of lexical
borrowing, essentially all characteristics of ME find their source in OE.
Despite this assumption, these historians have not succeeded in finding a
convincing OE source for the voiced plural suffix z. This is not for lack of trying, and
in fact most analysts have settled on (and firmly believe in) an impressionistic scenario
consistent with the assumption that OE ME. Upon investigation, we will see that this
scenario, which can be called “re-lexicalization of n as z, is badly flawed.
A first and brutal formulation of re-lexicalization (16) assumes a preliminary
reduction of unstressed OE short vowels to e (Minkova 1991, 5) and leaves aside the
vowel after noun stems ending in sibilants.
(16) Traditional diachronic change leading from OE to ME z:
PLUR, {en, es, e} z / N___
On the face of it, such an arbitrary (but pervasive) change is quite implausible. No doubt
to soften the blow (to the revered ancestral status of OE), scholars have divided it into
four less drastic intermediate steps and discussed diverse conditioning factors for
deleting e, such as vowel reduction and loss in (a few) unstressed final syllables; see
Lass 2006, 102105; 109111 and others he cites). In addition, it is generally assumed
that (16) implies two separate changes, one into a mono-segmental s and then a second
step whereby s
But no matter how complex the interplay of factors such as region, number of
syllables, preceding consonants, and poetic meter, orthographic (e)s must emerge as the
only competitor for productively marking the ME noun plurals. Revealingly, with
respect to the voicing in (16), scholarship has chosen to debate when the sound change
from es to (e)z took place, rather than why; consequently, this voicing is not
systematically related to other ME properties or developments. And independently of all
this, what also must be explained is the initial come from behind” victory of (e)s over
(e)n as the regular plural; see again the summary in Baugh and Cable (2013, chap. 7).
The scenario required by (16) remains ad hoc, no matter many intervening steps it
supposedly results from (all moreover taking place in not much more than a century,
In more detail, this basic scenario consists of four steps that derive ME plurals
from OE nominative/ accusative plurals.
(17) (a) Various OE plural morphemes in non-productive declensions consist
of vowels that reduce to early ME short e; the non-productive (“strong”) as
reduces to es (Lass 2006, 152; Algeo and Butcher 2014, 137140).
productive (“weak”) plural en remains.
The possible non-linguistic sociological, religious and historical motivations for this (probably
unconscious) assumption are too many and too obvious to merit space here. A few studies have
proposed, with sketchy and unsystematic arguments and definitions, that ME is a “creole” (i.e.
derives from multiple sources). This is indisputable only with respect to the lexicon.
Sometime in the pre-history of the OE suffix as, which derives from I.-E. “unstressed vowel +
sibilant”, Verner’s Law should have voiced the sibilant, as it did in both PS (North Germanic) and
(b) Final short e, often considered to be a schwa, deletes. Minkova (1991)
covers many facets of this process, including contexts that specify numbers of
syllables, vowel lengths, optionality, regional variants, borrowings from French,
relation to syntax, etc.
(c) After an early 12th c. spread of productive OE en from the South,
(e)s inexplicably replaces it as the productive ME plural first in the North
around 1200, and then spreads from North to South (Baugh and Cable 2013,
Ch. 7; Lass 2006, 111).
(d) This last change consists of two phonetic steps: e drops (except after
sibilants), and final s becomes z. That is, es becomes z. Each step should be
considered separately.
Thus, the changeover from OE plurals, whose last texts are about 1140, to a general ME
plural (c. 1250) involves four rules, or sound changes.
According to Bech and Walkden
(2016, section 2.1), nothing can be more important than sound change in determining the
history of a language, so I will consider the plausibility of (17) as abbreviated in (18i
(18) (i) Final short e deletes.
(ii) en is relexicalized as es.
(iii) Short e deletes in “some” closed final syllables, in particular in noun
(iv) Progressive voicing applies to “some” final s and f.
I critically examine in turn the plausibility of steps (18iiv) given in the traditional
histories. We will see that none of them express generalizations with the scope expected
of “regular sound changes;” none of them really has any general or explanatory force.
Gothic (East Germanic). However, even though a WG language, OE loses this voicing in final
obstruents, as described in the cited sources.
This traditional consensus scenario of four steps is summarized in a Wikipedia entry, which
however ignores the change from OE voiceless s to the voiced z of Modern English
The English plural s is the only survivor of a much more complicated Old English
nominal declension system. . . The plural ending for the Nominative and
Accusative of “strong masculine nouns” was as, and as the Old English nominal
system broke down, this ending was generalized to all nouns in all cases. By
Middle English we only have the ending es for all nouns, and in Modern English
the e has disappeared (except in spelling in some cases), giving us the plural s.
I do not subscribe to these authors’ claim that phonological sound change should remain today
the only sure foundation for linguistic genealogy. The ground-breaking papers in Battye and
Roberts (1995) demonstrate that syntax is on a par with phonological inventories as a source for
uncovering a language’s past.
3.4.1 Deletion of Final Short e
Rule (18i) describes the deletion of final short e, probably a schwa. This first step
accounts for the loss of OE plural allomorphs that consist of only a vowel. This rule, at
least when formulated as optional, seems to be general from exactly the period first
identified as ME around 1200, e.g. the language of the Ormulum.
That is, final “Schwa
loss during the ME period is axiomatic in all standard descriptions of the history of
English” (Minkova 1991, 36).
However, Minkova’s further claim (1991, 9) that “there is no parallel
development in Scandinavia” is almost certainly wrong. Short vowel deletion in final
position, including e-deletion, was endemic in earlier North Germanic, practically its
hallmark (Haugen 1982, 2829). Since the question in this essay is exactly whether ME
and Scandinavian are related, i.e. share their history, it is circular to use a dating
difference in schwa-deletion which has been determined by assuming that they are not
related. If, as Emonds and Faarlund (2014) argue, ME is simply a successor of PS, it is
no wonder that evidence of general final e-deletion, i.e. resembling NG syncope, is
found only in ME and not yet in the OE period. Many forms with final schwas that
appear to “delete” in early ME were possibly words whose final short e had deleted
earlier in NG.
So given this possibility, there is no safe conclusion about when final e-deletion
starts in England; we can only conclude that it was not fully productive in OE. There is
in fact a parallel in ME and ON (i.e. from 1150 onwards): neither language exhibits final
short e for any inflections. By this period, some ON inflections had again acquired short
a, i, and u, but not e. This is clear from the many ON inflectional tables in Faarlund
(2004, chap. 3).
Since traditional histories of English have not fully investigated relating final e-
deletion (a particular short vowel) to the more general short vowel apocope in
Scandinavian, rule (18i) is not general enough to merit what is usually meant by “regular
sound change.” It is rather a description of an ME vocabulary artificially isolated from
its possible roots.
3.4.2 Re-lexicalization of en as es
Leaving aside outright irregular plurals (formed with umlaut, null morphemes, etc.),
regular OE plurals were constructed within different noun classes with several different
plural suffixes: en, as, e, a, and u. There is no linguistic reason, other than a vague
appeal to “frequency,” why out of these five endings, only as should have become the
only productive survivor. A century ago, Classen (1919) showed the frequency factor
favored en, not as.
The logic of the traditional scenario thus is not based on
linguistic plausibility or independently justified aspects of ME phonology. Rather, by
assumption (not argument), among the five OE non-productive plurals, the choice is as
because it “looks like” ME s more than do the others.
The date of the onset of schwa loss is debated, but some authors put it in the 12th c. (Minkova
1991, 24) on the basis of some words in early documents without certain final e.
Classen argues for a hybrid analysis involving OE and Norse: OE speakers in the Danelaw
borrowed many Norse “weak declension” nouns, and then due to similarities in oblique cases, they
misanalysed them as OE “strong declension” nouns, so that as plurals became (for only Danelaw
speakers) more frequent, while OE en plurals remained more frequent in the South. As a result,
the North generalized as and the South en. The argument seems to depend on OE speakers
recognizing the (foreign) oblique case endings of the Norse weak declension, which is a shaky
sociolinguistic assumption on which to base a sound change.
The fact is, the traditional choice of an OE precursor depends on “looks like” (in
orthography) rather than on the appropriate “sounds like”, which involves comparing
phonological features, not graphemes. When we do this, there is no affinity between OE
as and ME z. In OE, as eventually in other WG languages (e.g. Dutch and German)
non-sonorants including continuants (f, th, s. χ) were unvoiced in word-final position
(Strang 1970, 288; Mitchell and Robinson 1992, 15; Lass 2006, 5761).
Since final
voicing was not a possibility, the relexicalization step in the traditional scenario has no
basis whatever in either frequency or phonetics; it is purely arbitrary.
3.4.3 Short e Deletion in Closed Final Syllables
The traditional scenario for noun plurals needs the (sporadic) ME “sound change” (18iii)
in order to delete the short e in the newly productive descendant es of the OE strong
plural as.
I first note that in other Middle and Modern English inflections of similar form,
no productive process of “short e deletion” has ever happened: neither to Proto-
Germanic short e in superlatives (slowest, truest, highest, greyest do not rhyme with
toast, boost, Christ, taste), nor to its short e in comparatives or agent nouns: the pairs
rower/roar, lower/lore and mower/more are not homonymous. The ME 3rd singular
suffix eth never productively lost its vowel (grow-eth, show-eth, stay-eth do not rhyme
with growth, both, faith), nor has vowel deletion ever affected the pervasive unstressed
suffix ing.
Outside inflections, there are some instances of ME schwa deleting in final
closed syllables. Yet according to Fulk (2012, 50), “Unstressed /ǝ/ in final syllables is
never lost when the result would be a final consonant cluster in which the sonority of the
final consonant is greater than that of the preceding consonant.” In this same passage, the
author’s logic crucially uses the “high sonority of fricatives.” By this reasoning then, the
vowel in the OE plural as/es should never be lost after a stop, yet (except after
sibilants) it always is. This general fact renders this deletion of e discussed by Fulk
irrelevant to the history of the plural, even if extended to (a few) inflections. In fact, Fulk
(2012, 5960) also claims via metric analysis of poetic texts that some ME medial e are
indeed purely orthographic. For instance he scans sinnes sins from the poet Richard
Rolle (c. 1325) as monosyllabic; the plural consists of only a consonant.
In sum, an important advantage of this essay’s history of English plurals is that
it dispenses with the need for the ad hoc rule (18ii).
Since voicing in sibilants was non-distinctive in OE, occurring phonetically only in intervocalic
contexts, the s in the OE suffix as was always unvoiced.
If regular sound changes derived the ME plural from a late OE es, by deleting schwa before a
sibilant and voicing s, then one should also find shyness phonetic shines, oneness
ones, baroness
phonetic barons, and illness rhyming with kilns.
As southern speakers adopted AN/ ME, initially as a second language (nonetheless close to their
native West Saxon), they could have felt that its plural z corresponded to the Saxon as/es, not
realizing that the AN plural was mono-segmental. In this way, some southern ME speakers might
have used it in e.g. poetry as they would a final weak syllable plural in West Saxon, which they
doubtless still also spoke. This study’s hypothesis (14) thus predicts that any evidence for deriving
ME –s from “Vowel + s” should be from southern ME dialects.
The much earlier NG loss of short vowels in final syllables (7th and 8th c.) was a regular and
productive sound change, namely the short vowel apocope that is almost this family’s defining
characteristic (Haugen 1982, 2829). But the text here concerns ME after 1200.
3.4.4 Progressive Voicing in Plurals
The traditional scenario requires finally a spontaneous and ad hoc voicing of an OE
word-final s. Though authors often fail to mention this, Honeybone (2012) realizes the
isolated nature of this voicing, moreover occurring in very few contexts: “English is odd
in this regard. It seems to feature a case of final obstruent voicing, which is essentially
unheard of in the history of languages” (2012, section 3.4). This final voicing cannot
therefore be related to any “progressive voicing tendency” in any stage of English, since
it would contravene the general Ban (11) argued for in detail in Section 2.
Lass (2006, 5961) suggests that distinctive voicing of ME z in both initial
position and in sibilant plurals was present from the beginning of ME, a view with which
I concur:
Be that as it may, by around 1250, /v/ and /z/ were separate phonemes in foot-
initial position . . . The development of a final voice contrast is tied to the loss of
final /ǝ/ [reference omitted], which probably began in the north and north
midlands in the twelfth century [before 1200; my emphasis, JE], and then spread
southwards . . .
However, this dating of initial and final voiced segments in ME still leaves open
the issue of a motivated source for this new peripheral ME z. In this regard Lass makes
two further points: (i) He favors an account in which phonemic distinctness in both final
and initial positions reinforce each other (his account is in terms of weak and strong
(ii) He finds a source for distinctive voicing of initial v and z only in some
non-productive borrowings of southern dialect forms (e.g. vixen vs. fox, etc.). The first
point seems broadly correct, but the second is very weak. I suggest instead that only the
long standing voicing in CS noun plurals provided a robust springboard for a extending a
contrastive ME z to both initial and final positions.
Note that this view is consistent with the sequencing in the above quote from Lass
(2006): the voiced plural (1200) precedes the establishment of an initial sz contrast. The
only motivated source for the ME voiced plural is thus in CS; the voicing cannot be
convincingly squeezed out of OE or its dialects.
The overall points of Section 3 can be summarized: Deriving the voiced ME
plural z from one OE plural inflection (among many) requires that it comes from e +
voiceless s, via two unmotivated, ad hoc rules: vowel deletion in (very few) final closed
syllables and progressive voicing assimilation. Moreover, this productive plural z has to
spontaneously replace a late OE tendency by which the southern (West Saxon) plural n
was spreading, as indeed expected in the West Germanic languages. In the end there is
no actual evidence for postulating the sound change (16) or the developmental sequence
(18) proposed by traditional historians of English. This sound change, no matter how
many steps it is decomposed into, amounts to nothing more than just what they have to
(and do) say to maintain a priori that main properties of ME grammar, including its noun
plurals, must originate in OE.
For a scholar working in the traditional framework, to situate a ME “innovation” prior to 1200 is
equivalent to making it part of what I claim is the changeover from OE to AN.
We should avoid any “intuition” that distinctive consonantal phonemes always enter a language
in word-initial position. The English voiced palatal continuant phoneme zh is a counter-example
(cf. the medial contrasts in lesion, reason, lotion and occasion, station).
One remaining question concerns the earliest typical ME spelling es of the
plural, claimed here to be a phonetic z in most contexts. Does the spelling suggest a
different phonetics? Of course, the exact same question applies to 500 years of the same
Modern English spelling, and here the answer is, scribes/ printers are not linguists;
spelling is by far more influenced by the ambient scribal/ printing tradition, which
almost without fail overrides phonetic accuracy.
What was different for early ME is that it was essentially being newly written,
so its first scribes like the monk Orm could draw on only Latin and OE writing
traditions. In Latin, most plural case forms end in Vowel + s, as does the only sibilant
plural in OE. Hence, we cannot take the ME combination of a Vowel + s in plurals as
phonetic evidence, unless it is corroborated by evidence such as metric scanning of
poetry, as cited here in Section 3.4.3.
4. Overall History of the North Germanic Plural
4.1 Proto-Scandinavian and Common Scandinavian Plurals
In the first millennium, NG inscriptions are found in an alphabet of phonological
“runes”. Spurkland (2005) is a detailed scholarly treatment of this writing system and its
stages. The considerations listed earlier in (13) motivate the consensus in NG studies
that the PS noun plural was a voiced phonemic z, written as the 3-pronged pitchfork rune
represented here as ż.
For these reasons, Haugen (1982, chap. 45) is justified in
systematically transcribing PS ż in his tables as a voiced sibilant z.
The period in which ON and its descendants were written with the Latin
alphabet, starts about 1150. ON still had several different inflectional classes of nouns,
with four often distinct cases in both the singular and plural; they are given with
examples and sources in Faarlund (2004, 2433).
Inspection of these paradigms shows
that non-neuter nominative plurals no longer terminate in z but in r; this change is
typically called rhoticization; the accusative plural counterparts are either identical to the
nominative or simply lack the r. That is, except for one class of neuter nouns, ON had no
other overt allomorphs in nominative and accusative that compete with r as a noun
The transition to Modern Norse then consisted simply in generalizing the ON
nominal plural ‘(vowel) + r. Since the distinctive features of r are [Alveolar,
+Continuant, +Voice, +Sonorant], the “phonemic distance” of the modern inflection
from the PS and lexical ME plural z is minimal; they differ in only one distinctive
During the period of Common Scandinavian, the rune ż used for the nominal
plural and 2nd singular agreement remained entirely distinct from the runes for either r
or s. This indicates that despite (probably undecidable) debates about its exact phonetic
quality, the CS inflectional ż must have remained a phoneme separate from the phoneme
r, which it eventually would join. According to Haugen (1982, 5762), this separate rune
Like some other runes, those for s, m, and h, later runic script modified its form; in the case of z,
the “pitchfork” was inverted but quite recognizable.
Faarlund’s descriptive grammar of ON predates by several years serious consideration that
English might be North Germanic, so his study is definitely not skewed in that direction.
Modern writers who use the small Latin capital R for this rune, written here with ż, are
graphically anticipating their knowledge of its linguistic future: that ż will later merge with r.
and the phonemic voiced sibilant it represented persisted in certain regions into the 12th
To construct a timeline, we can date the end of significant Scandinavian
immigration to England in 1066, at the Danish defeat at Stamford Bridge and the
imminent arrival of William the Conqueror. Consequently, during most of, and perhaps
all of, the period of Scandinavian settlement in England, their noun plural was more akin
phonemically to its origin as a fricative than to its future as a sonorant. For these reasons,
I conclude that the z that became an inflectional r in ON was not yet actually part of
that phoneme well into the Common Scandinavian period (c. 8001100).
This then provides the source of the voicing of the ON plural and an even more
transparent one for voicing in the plural of its Anglicized Norse (=ME) sibilant
counterpart. At least for some time after 800, Norse in England had a phoneme written
here as ż, in contrast to r, and this was the spelling of the plural on nouns.
represented exactly the distinctive features of what we recognize as a phonemic z in ME,
spelled as a word-final sound with s (as are modern is, was, as, these, because, phrase,
rise, rose, vase, etc.); those features were and are: Alveolar, +Continuant, +Voice,
Sonorant. Today’s productive English plural is therefore an unchanged continuation
from Proto-Scandinavian.
4.2 Common Scandinavian Splits into Anglicized Norse (ME) and Old
I thus propose that the main diachronic structural event affecting Proto-Scandinavian and
Common Scandinavian z occurred not in England or the history of English but in
Scandinavia. CS (written only in runes) underwent a phonemic change apparently
starting around 900 and completed in the 11th c.
(19) Old Norse Rhotic Merger
The phoneme z loses the feature value distinguishing it from the phoneme r.
On the face of it, this process merged z and r in one abrupt step. But there is an
additional factor. In its history z appears to have somehow “rhotacized” (become a
sonorant) before the merger (19) in early ON made it an r like any other.
According to Thöny (2016), a first stage of rhoticization occurred early in
Proto-Scandinavian. The insertion of a rhotic feature (for which I am using +Sonorant)
exempted ż from later devoicing of final obstruents such as z. If his scenario is correct,
the Common Scandinavian phoneme z was already +Sonorant (rhotic), at least in the NG
branch that became ON. This suggests that something like (20) produced an allophone of
z, whose features still differed from the feature content of “true r. For convenience, I
label it ř.
(20) CS Allophonic Rhoticization of ż.
[Alveolar, +Continuant, +Voice] +Sonorant
Essentially all historians of English agree that Norse continued to be spoken in England into the
12th c. For discussion, see Emonds and Faarlund 2014, sections 1.3 and 2.1. Since ON runic
inscriptions also continued to appear in England into the 11th c., so distinctions in that alphabet
almost certainly reflect those in spoken AN.
To me the most parsimonious, least convoluted account is that the PS “true r”,
like the two very distinct r of today’s French and English, was not alveolar. That is, (20)
was an allophonic rule that began in the PS period, but did not bring about merger with
the phonemic r (which had a separate rune). In order for ON ż and r phonemically to
merge as in (19), they first had to lose distinctive specifications for the feature
specification Alveolar.
Turning now to the destiny of PS z in England, one need only say that
Allophonic Rhoticization (20) (z
ř) was not (permanently) implemented in AN/ME;
instead, the earlier z was uniformly retained. It might be asked, if this process began on
the Mainland by some (bit not all) accounts as early as the 7th century, why would
Scandinavian speakers in England not adopt it, and then transmit it to ME? The fact is,
sociolinguistics frequently describes phonetic innovations in a language’s homeland or
central area that do not develop in its colonies or overseas extensions. Thus, French in
Canada is often more conservative than in France; several innovative changes in English
phonology have not occurred in e.g. Ireland, Scotland and the United States (one thinks
of the loss of post-vocalic r). Along the same lines, post-vocalic s is retained more in
Spanish-speaking countries farthest from Spain (Mexico).
Such divergence (using or not using an allophonic variant) may sometimes be
random, but as Labov (1963) persuasively argues, it can also reflect social distinctions
between populations whose cultural allegiances are split, in the case at hand between an
“old world” (Scandinavia) and a “new world” (the Danelaw), even when they speak the
same language, Common Scandinavian (8001050). Let us keep in mind that the
Scandinavian colonists in the Danelaw were prospering (by the standards of the time)
in fact, it must have been the improved opportunities that attracted settlers to England in
the first place. According to Wood’s (1986) account of their economic and legal status,
such as the ability to own and bequeath land, they came on average to surpass the Anglo-
Saxon peasantry.
Probably because of such success, Anglicized Norse was slowly replacing
Anglo-Saxon dialects, from North to South in the Danelaw, as the island’s predominant
Germanic tongue. This tendency must have been reinforced and accelerated by the
Danish conquest of all of England in 10131016. The subsequent rule of England by
Norsemen, continuous into the late Middle Ag (except for 10 months in 1066), sealed the
fate of OE (West Saxon). But at the same time, the settled English Scandinavians, far
from remaining poor immigrants who identified with their ancestral country, were better
off than more recently arriving immigrants. It is thus sociolinguistically natural to
propose that in the 10th and 11th c. Danelaw the older, conservative Norse of established
settlers, which retained z as a plural morpheme, was more prestigious than that of
immigrants and successive generations of Viking raiders, whose speech could be
identified by Mainland innovations such as the ř of (20).
This situation in 9th and 10th c. England calls to mind that on another island a
millennium later, the dialectal differences on the island of Martha’s Vineyard off the
New England coast, as analysed in Labov’s (1963) classic sociolinguistic study. He
uncovered social correlates of the unconscious differences in allophones of their English
dialects. (In the following quote, “the model” refers to the speech of the oldest English
stock fishing families on the island; “centralization” to their non-standard allophones of
certain diphthongs.)
Languages can have two r sounds that differ by the feature Alveolar. Current Czech
orthographic r is an alveolar trill, while Czech orthographic ř is palatal, not alveolar.
If someone intends to stay on the island, this model will be ever present to his
mind. If he intends to leave, he will adopt a mainland reference group, and the
influence of the old-timers will be considerably less. The differential effect in the
degree of centralization used is a direct result of this opposition of values . . . In
summary, we can then say that the meaning of centralization, judging from the
context in which it occurs, is a positive orientation towards Martha's Vineyard.
(Labov 1963, 305306)
Replacing “(the degree of) centralization” with “a non-sonorant sibilant plural”,
i.e. non-application of (20), I propose that for the Scandinavian settlers, the
meaning of a non-sonorant z plural signified a positive orientation towards living
in England.
Especially in the 10th c., when Allophonic Rhoticization (20) was spreading
on the mainland, English Scandinavians strongly identified with being permanently
settled in England, and very likely as a population, they rejected or never seriously
considered severing links with their established island home. In fact, English
Scandinavians are known to have often sided with the Anglo Saxons in the 10th and
again in the late 11th c. in efforts to ward off ever renewed Norse incursions.
thus had social reasons for not identifying with their newly arrived aggressive
“cousins”. Instead, while retaining and spreading their mother tongue AN/ ME inside
England, they freely adopted West Saxon vocabulary. In the same vein, they
unconsciously resisted Mainland linguistic innovations such as Allophonic
Rhoticization in the noun plurals.
As a lasting result, English has steadfastly adhered to older Proto-
Scandinavian hallmarks such as the voiced sibilant plural z. The torturous derivation
of the ME noun plural z from the non-productive West Saxon as must be
Appendix: Labov’s scenario for prestigious archaism
Labov (1963) lays out five steps in (21) that lead to an archaic prestige dialect becoming
predominant in a region. They fit not only the situation on Martha’s Vineyard c. 1970
(the left column follows his exposition), but also I submit in the Danelaw 10001100
years earlier. These steps in the right column led to the dominance of the prestige plural
z of Anglicized Norse, the dialect that rejected the Mainland use of the rhotic plural:
Labov’s “centralization” refers to a conservative rejection of final stage diphthongs ai and au in
the English vowel shift.
The Norman Conquest itself was simply the last and most devastating of these. The rulers of
Normandy were a war-prone clan of Scandinavian descent who felt that Anglo-Saxons were
wrongly taking control of England after the Danish King Canute and his stepson Edward the
Confessor were left without heirs (both ruled all of England from 1016 to 1066).
A fortiori, AN never adopted some even later Scandinavian innovations, such as a definite
enclitic en, which appeared in Mainland runes around 1100 (Haugen 1982, 173174).
According to this essay, the An/ME noun plural is closer to Proto-Germanic than OE. A
hypothetical parallel can be drawn in the history of Romance. Standard French is solidly
established as a daughter of some version (perhaps spoken) of Latin, though it lacks the Latin
feature of unstressed final syllables. Suppose Provençal were only recently proposed as related to
French. Then, discovery of Provençal’s unstressed final syllables would place it between Latin and
French, and be hailed as confirming the Comparative Method. This essay’s analysis of the English
sibilant plural likewise places ME between Proto-Germanic and ON.
(21) Labov’s five sociolinguistic steps (15 in italics cited from Labov 1963, 307):
1. On Martha’s Vineyard, c. 1970
2. In the 9th10th century Danelaw
3. 1. A language feature used by a group A is marked by contrast with another standard
Island fishing families (group A) older
“central” diphthongs. Standard Mainland
English has the full vowel shift diphthongs
au and ai.
Early Scandinavian immigrants
(group A) arrive in England with z
plurals, Mainlanders start replacing it
with ř.
4. 2. Group A is adopted as a reference group by group B, and the feature is adopted
and exaggerated as a sign of social identity in response to pressure from outside
Islanders who identify with a life and future on the island (group B) adopt the older
pattern of Group A, in response to the possibility of a life on the Mainland model.
This holds for both Martha’s Vineyard and the Danelaw.
5. 3. Hypercorrection under increased pressure, in combination with the force of
structural symmetry, leads to a generalization of the feature in other linguistic units
of group B.
Minority island communities also adopt the
older diphthongs, as the fishermen become
the model for Group B’s “independent life
on the Island.”
All Scandinavians in England adopt
the older plural z; settled successful
farmers become the model for group
B’s “life in England”.
6. 4. A new norm is established as the process of generalization levels off.
“Down-island” speech keeps the Mainland
dialect, which lacks centralization.
All of Mainland Scandinavian adopts
r plurals.
7. 5. The new norm is adopted by neighbouring and succeeding groups for whom group
B serves as a reference group.
“Up-island” speech becomes the prestige
dialect on the island, with centralization.
Anglicized Norse with z plurals
becomes standard in ME.
Whether the plural of Anglicized Norse was phonetically simply z or an allophonic
rhotic z, we cannot know with certainty. If the latter, the rhotic quality was lost in
England by the time ME was written, e.g. late 12th c. But we know that this voiced
sibilant plural spread southward exactly in the way and at the time of several dozen
other “Norsifications” of early ME (Thomason and Kaufman 1988), just as ME became
a written language. In contrast to the blanket devoicing of West Germanic final non-
sonorants, the North Germanic languages of ME and Modern English have ever since
used the Proto-Scandinavian voiced final sibilants for their noun plurals.
In sum, returning to the general question of whether Modern English inflection
is North Germanic, not only some but basically all productive Modern English
inflections (s, d, ing, er, est) have ancestral lineages traceable to Proto-
Works Cited
Algeo, John, and Carmen Butcher. 2014. The Origins and Development of the English
Language. 7th ed. Boston: Wadsworth.
Battye, Adrian, and Ian Roberts, eds. 1995. Clause Structure and Language Change.
Oxford: Oxford University Press.
Baugh, Albert, and Thomas Cable. 2013. A History of the English Language. 6th ed.
London: Routledge and Kegan Pau.
Bech, Kristen, and George Walkden. 2016. English Is (Still) a West Germanic
Language. Nordic Journal of Linguistics 39 (1): 65100.
Borowsky, Toni. 2000. Word-Faithfulness and the Direction of Assimilations.
Linguistic Review 17: 128.
Classen, Ernest. 1919. S and N Plurals in Middle English. Modern Language
Review 14 (1): 9496.
Clements, George. 1990. The Role of the Sonority Cycle in Core Syllabification. In
Papers in Laboratory Phonology 1: Between the Grammar and Physics of Speech,
edited by M. Kingston and M. Beckman, 283333. Cambridge: Cambridge
University Press.
Emonds, Joseph, and Jan Terje Faarlund. 2014. English: The Language of the Vikings.
Olomouc: Palacký University Press.
Faarlund, Jan Terje. 2004. The Syntax of Old Norse. Oxford: Oxford University Press.
Fulk, Robert D. 2012. An Introduction to Middle English. London: Broadview Press.
Gianollo, C., C. Guardiano, and G. Longobardi. 2008. Three Fundamental Issues in
Parametric Linguistics. In The Limits of Syntactic Variation, edited by Theresa
Biberauer, 109142. Amsterdam: John Benjamins.
Haugen, Einar. 1982. Scandinavian Language Structures. Tuebingen: Niemeyer Verlag.
Henke, Eric, Ellen Kaisse, and Richard Wright. 2012. “Is the Sonority Sequencing
Principle an Epiphenomenon?” In The Sonority Controversy, edited by Steve
Parker, 65100. Berlin: Mouton de Gruyter.
Henle, Robert. 1945. Latin Grammar. Rev. ed. Chicago: Loyola University Press.
Honeybone, Patrick. 2012. Lenition in English. In The Oxford Handbook of the
History of English, edited by T. Nevalainen and E. Traugott, Chapter 60. Oxford:
Oxford University Press.
Iverson, G., and J. Salmons. 1999. Glottal Spreading bias in Germanic. Linguistische
Berichte 178: 135151.
Jespersen, Otto. 1912. Growth and Structure of the English Language. Leipzig: B. G.
Kaisse, Ellen. 2011. The Structure Features. In The Blackwell Companion to
Phonology, edited by M. van Oostendorp, C. Ewen, E. Hume and K. Rice, 288
310. Oxford: Wiley-Blackwell.
Kershaw, Jane. 2017. An Early Medieval Dual-Currency Economy: Bullion and Coin in
the Danelaw. Antiquity 91: 173190.
Kiparsky, Paul. 2009. Syncope, Umlaut, and Prosodic Structure in Early Germanic.
Unpublished manuscript, Stanford University.
Labov, William. 1963. The Social Motivation of a Sound Change. Word 19 (3): 273
Lass, Roger. 2006. “Phonology and Morphology.” In The Cambridge History of the
English Language, Volume II (10661476), edited by Norman Blake, 23155.
Cambridge: Cambridge University Press.
Lightner, Theodore. 1970. On Regular Inflectional Endings in English. Papers in
Linguistics 3 (3): 503519.
Lombardi, Linda. 1999. Positional Faithfulness and Voicing Assimilation in Optimality
Theory. Natural Language and Linguistic Theory 17: 267302.
Minkova, Donka. 1991. The History of Final Vowels in English: The Sound of Muting.
Berlin: Walter de Gruyter.
Minkova, Donka. 2014. A Historical Phonology of English. Edinburgh: Edinburgh
University Press.
Mitchell, Bruce, and Fred Robinson. 1992. A Guide to Old English. 5th ed. Oxford:
Mossé, Fernand. 1952. A Handbook of Middle English. Translated by James Walker.
Baltimore: Johns Hopkins Press.
Shibatani, Masayoshi. 1972. The Phonological Representations of English Inflectional
Endings. Glossa 6 (1): 117127.
Sloat, Clarence, and James Hoard. 1971. The Inflectional Morphology of English.
Glossa 5 (1): 4756.
Spurkland, Terje. 2005. Norwegian Runes and Runic Inscriptions. Translated by Betsy
van der Hoek. Woodbridge: Boydell Press.
Strang, Barbara. 1970. A History of English. London: Routledge.
Thöny, Luzius. 2016. The Chronology of Final Devoicing and the Change of *z to ʀ in
Proto-Norse. Futhark: International Journal of Runic Studies 7: 4762.
Thomason, Sarah, and Terrence Kaufman. 1988. Language Contact, Creolization, and
Genetic Linguistics. Berkeley: University of California Press.
Townend, Matthew. 2002. Language and History in Viking Age England: Linguistic
Relations between Speakers of Old Norse and Old English. Turnhout: Brepols
Watts, Richard. 2011. Language Myths and the History of English. Oxford: Oxford
University Press.
White, Robert. 1852. The Ormulum with Notes. Oxford: Oxford University Press.
Wood, Michael. 1986. Domesday: A Search for the Roots of England. London: Guild
Woodruff, Douglas. 1974. Alfred the Great. London: Weidenfeld and Nicolson.
Zwicky, Arnold. 1975. Settling on an Underlying Form: The English Inflectional
Endings. In Testing Linguistic Hypoheses, edited by D. Cohen and J. Wirth, 120
185. Washington: Hemisphere.
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
Abstract: English as North Germanic: Modern English is Modern Norse It is well known that Middle English (and its descendent Modern English) has a large number of words of Scandinavian origin. This is conventionally attributed to language contact and heavy borrowing of Scandinavian words into Early Middle English (not into Old English). However, this alleged borrowing was not limited to lexical words, counter to the normal case in contact situations; grammatical words and morphemes were also borrowed. This is unusual, and calls for an explanation. The explanation argued for here is that the roots of Middle English (and therefore Modern English) are North Germanic, with large borrowings from the Old English lexicon, rather than the other way around, as generally assumed, and that the fusion of the two lexicons dates back not to early Scandinavian settlement in England, but about 200 years later, especially the 12th c. during the full impact of the Norman Conquest. Even more problematic is the fact that Middle English and Modern English syntax is Scandinavian rather than West Germanic. The languages share numerous syntactic properties (e.g. word order, P-stranding, infinitival and directional particles, auxiliaries, infinitival constructions, participles and case inflections), which reflect a deep and typologically significant relation of Scandinavian with Middle/ Modern English. With respect to all these characteristics Middle/ Modern English groups with North Germanic rather than West Germanic.
Schwa has always been the poor relation of the stressed vowels in English historical phonology. The academic searchlights have been directed on the developments of the vowels in stressed position. Most studies in the field do make some reference to the changes of unstressed vowels, yet the supposedly predictable genesis and unidirectional fate of the ubiquitous schwa-type vowels have attracted very little attention. Seen in a wider context, however, the reduction of unstressed vowels to schwa and its subsequent loss in final position has had a more profound effect on the phonology, prosody, and grammar of the language than any individual qualitative or quantitative change of the stressed vowels. This book is an attempt to bring together traditional philological records and knowledge of the ways and means of schwa loss and reexamine them in relation to the entire language system.
Metal detecting in England has recovered a large number of Viking Age single finds that have been reported to the Portable Antiquities Scheme. These reveal that silver bullion of Scandinavian origin was used as currency throughout the Danelaw between AD 865 and 940. Standardised weights of copper alloy were an integral part of this metal-weight economy. Bullion was not the sole means of silver payment during this period: coinage had long been used in the occupied Anglo-Saxon territories and continued to be minted under the Vikings. The resulting dual-currency economy may have facilitated trade with neighbouring Scandinavian territories, but the two currencies also served as markers of cultural identity, offering a choice of monetary media.
A thorough and fascinating exploration of the evolution of English' phonological structure, this book traces the history of individual sounds and their representation through Old, Middle, Early Modern and Present Day English. Written in an engaging and accessible style, the book covers the sounds of English, consonantal histories, Middle English dialects, vowel quality and quantity in Early Modern English, the English stress system and Early English verse forms to demonstrate how the present form of the language is indebted to its past.
This book aims to deconstruct the myths that are traditionally reproduced as factual accounts of the historical development of English, and to reveal new myths that are currently being constructed. Using concepts and interpretive sensibilities developed in the field of sociolinguistics, discourse analysis, and sociocognitive linguistics over the past 40 years, the book unearths these myths and exposes their ideological roots. Based on the assumption that conventional histories of English are histories of standard English rather than histories of the varieties of English, he sets his goal as being not to construct an alternative discourse, but rather to offer alternative readings of the historical data. It defines what we mean by a linguistic ideology and shows how language myths, rather than simply being untruths about language, are derived from conceptual metaphors of language and are crucial in the formation of hegemonic discourses on language. He argues, in effect, that no discourse-a hegemonic discourse, an alternative discourse, or even a deconstructive discourse-can ever be free of ideology. The book argues that a naturalized discourse is always built on a foundation of myths, which are all too easily taken as true accounts, and is a call to study alternative ways in which the full range of "Englishes" may ultimately be accounted for historically. But the book also issues the warning that, whatever new histories are proposed, they, too, will ultimately need to undergo a thorough investigation with regard to the myths that may underlie them.