You are on page 1of 17

Late Latin: the evidence

Roger Wright
University of Liverpool
roger.wright@liv.ac.uk

ABSTRACT
Late Latin is not a special case; it can be studied in the same way and according to the
same principles as other living languages, in the areas where it was the native speech (Italy,
The Iberian Peninsula and France), during the period before the arrival of the “Carolingian”
reforms into a speech community. Since the direct surviving evidence provided for us by
those who spoke it is in written form, the relations between speech and writing are an
important aspect of such investigations. We need in particular to take into account the ideas
and prejudices of teachers, which often prevent the written text from being a direct
representation of the spoken equivalent. The traditionally correct spellings, and aspects of the
morphology, continued to be taught and learnt, but new features of syntax and vocabulary
were easy to represent within the old writing systems. Thus in essence, what Late Latin (Early
Romance) native speakers were doing was essentially the same as we all do now, in Romance
Language communities and elsewhere; we speak our contemporary form of the language, and
aim to write it according to older principles.

I’ve spent much of my academic life suggesting that Late Latin isn’t a
special case, but can be studied by historical linguists and philologists as a
language operating in its native speaking community in the same way as any
other does. With the phrase “Late Latin” I’m referring to Latin in the years
before it came to be fossilized, and elevated to the status of a special
language for use in the Church, by the ninth-century scholars in France. The
language standardized at that time was what we now call Medieval Latin,
which is indeed a special case, having in effect no native speakers. In areas
where Latin wasn’t the usual native speech, such as in the Germanic and
Celtic speaking areas, it had been a foreign language all along; but before the
ninth century, among the native speakers in most of Italy, in the Iberian
Peninsula and in much of France, it seems reasonable for us to study the
language according to the same principles and ideas as other native
languages are studied. This is not a new thought on my part; I’ve said before
most of what I’m going to say today, particularly in Wright (2013).
Naturally, we need to consider the nature of the evidence. The two main
sources of evidence for the features of spoken Late Latin, the language
130 R. WRIGHT

which can also refer to as Early Romance, are reconstruction, which involves
arguing backwards from later Romance, and the texts of the time which were
written by those who spoke it as native speakers, in Italy, France and the
Iberian Peninsula. The main problem with using reconstructions as evidence
concerns the near impossibility of dating the reconstructed features in any
way other than approximate; the main problem with using the texts as
evidence is that texts are not direct representations of speech, because
anybody who has learned to write has, by definition, been taught to do so,
and the ideas and prejudices of the teachers are normally there in the writer’s
mind, presenting a barrier between their natural speech and its written
counterpart. In the case of Late Latin (Early Romance), these barriers
between their natural speech and the texts will usually have taken the form
of a pedagogical desire to emulate the admired writers of the past; but the
barriers can also involve more ad hoc rules of thumb and even
misconceptions. If writers make what seem to us to be mistakes because they
are following a teacher’s precepts which happen to be inaccurate, this can
confuse our evidence, but even in these cases, this evidence can still be of
sociophilological interest, illuminating what teachers and writers were doing.
The question of what it is that textual sources actually attest is a slippery
one for synchronic linguists investigating their own native language, so it is
unsurprising that it should also be a difficult one for the diachronic linguists
attempting to analyse linguistic states of the past. The question has come to
seem harder than it used to, because of the development over the last fifty
years of sociolinguistics, which has led to the general understanding that
variation is not just a normal feature of a living language but a necessary
one. One consequence of this is that no text is likely to attest more than a
small subset of the linguistic variants available in speech. And variation
applies in many ways. For example, there always exists, in any community,
not merely diatopic variation between different ways of speaking in different
geographical areas, which has traditionally been privileged by historical
linguists as the apparently most salient kind of variation, but also variation
between the ways of speaking of different people in the same area; this kind
of variation was often analysed in the late twentieth century through a quasi-
Marxist analysis of speech communities in unsubtle terms of social classes.
In addition, as is increasingly being appreciated, there is usually variation
between the way the same individual speaks and writes in different
circumstances; for example, the nature of the person we are talking to, or
writing for, can fundamentally affect the way we speak or write. The further
realization that features of speech and writing are always to some extent
different from each other, even within the same speaker/writer, has seemed
of no great import to the synchronic investigator interested in the idealized
speaker/hearer, but has provided considerable food for thought for those who
Late Latin: the evidence 131

hope to recreate linguistic details of the distant past, for which the only
direct evidence is written evidence. These problems have been well
discussed in the Romance field by Wulf Oesterreicher (e.g. Oesterreicher
2004), but I’m not sure that Latinists have taken them fully on board.
Given that individuals regularly vary with each other, even when
incontrovertibly speaking the same language, it has become normal to allow
for such variation to exist within our synchronic descriptions of a single
language viewed as a whole; the only alternative would be for each idiolect
to be studied separately. Ontologically, such a procedure, studying one
speaker at a time, is justified in itself, and many conscientious studies have
been made of the language of Latin-speaking and Romance-speaking
individuals writing in the past, sometimes without any detailed discussion of
how far the individuals studied should be seen as representative of the
language of their age. Indeed, since most of the writers chosen for this kind
of analysis stand out for their overall individuality and untypical linguistic
skill, it can be argued that literary sources such as theirs are the wrong data
to turn to for evidence and enlightenment about the contemporary language
in general. For example, it seems improbable that the style of the voluble
seventh-century writer Valerio del Bierzo could be representative of the Late
Latin / Early Romance speech of his neighbours in the Bierzo; these
unusually articulate individuals and skilled stylists are special cases, while
still fitting within the wide-ranging variation of their community’s single but
multifaceted language.
So it has seemed sensible to many historical Romanists and Late Latinists
to investigate data of a more humdrum nature than the great literary stylists,
such as those administrative and legal documents that appear to preserve
verbatim accounts, wills designed to be read aloud to interested parties,
diaries and journals, inscriptions and graffiti, or letters (all of which usually
have the great advantage of being precisely datable and locatable, unlike
many more literary manuscripts). The evidence of poetry is often
disregarded by historical analysts interested in the speech community of the
poet, on the grounds that poetry is not a natural register, tending to include
archaisms, unnatural word orders and diction, either for ostentation, or more
prosaically metri gratia; and yet syllabic verse, such as Early Medieval Latin
‘rhythmic’ verse, can be a help when studying syllabification, and a careful
consideration of rhyme can give direct phonetic indications as well. In a
previous study I showed how Late Latin verse composed by Romance-
speakers differed significantly in linguistic details from Late Latin verse
composed by Celtic or Germanic speakers (Wright 2003: chapter 7). Most of
the available source texts of whatever kind, even if they happen to be
anonymous now, were composed by an individual, and to some extent
132 R. WRIGHT

administrative and legal texts also represent an idiolect rather than a


generality. When studying legal texts of the Iberian tenth century, for
example, it is worth knowing who the scribe was, if we can find out, and
then we can look at his production before subsuming it into a greater mass of
data produced by many others. This is what Wendy Davies has been doing
recently, with illuminating results (Davies 2013).
The general dichotomy between speech and writing has been useful to
explore over the last few years, but it can be less stark than it is sometimes
made to seem. For there are many kinds of spoken register and many kinds
of written register. The distinction is porous. Written texts can be, and in the
Early Middle Ages usually were, read aloud, and when they were read aloud
it was normal for them to be intelligible even for listeners who couldn’t
themselves read. Conversely, spoken styles can be, and often have been,
written down, with greater or lesser degrees of faithfulness; dramatic works,
for example, can be, though often aren’t, close to genuine speech. In the
modern world, there are now media which seem to be both at once, such as
text messages. The features of spoken language and those of written
language cannot be separated with any clarity in many cases. It’s true, for
example, that speech uses more deictics than writing, and that writing uses
more subordinate clauses than speech, but there are deictics in writing and
subordinate clauses in speech; as a consequence, the strict diglossic
relationship which has sometimes been envisaged as necessarily existing
between spoken language and written language in a literate community is as
much an idealization, if postulated as existing within the competence of the
ideal speaker/writer, as the ideal speaker/hearer is. Diglossia is a concept
which has often been applied to the Early Middle Ages, and it has probably
been a useful concept for use in the analysis of bilingual societies, as
opposed to individuals. But I don’t like the concept of diglossia myself. It’s
now apparent that under the laxest of the available definitions most
monolingual literate societies would be classified as diglossic, given the
unavoidable presence of wide variation and several registers in monolingual
literate societies anyway; and even under a stricter definition of diglossia, all
the societies which have been described as diglossic are in fact rather
different from each other. The widespread use of the term ‘diglossia’ may
thus have done little more than create confusion in the linguistic analysis of
individuals operating in complex monolingual speech communities and of
the textual evidence which they have provided for us, which is why diglossia
will not be mentioned again here today. I have, however, previously
discussed the value of the concept of diglossia in the context of Medieval
Iberian culture at considerable length (Wright 2010a).
Late Latin: the evidence 133

There have, naturally, been valiant attempts to overcome these potential


hurdles. Many investigators don’t feel constricted by any need to allow for
the varying vagaries of idiolects. The existence in the modern world of huge
computerized linguistic corpora has meant that it is possible to circumvent or
ignore the problem of the necessarily idiolectal nature of all evidence by
amassing vast accumulations of detail. If we have available millions of
words provided by thousands of speakers, we can at the least draw out a
highest common factor. We can still allow for the necessity of variation,
without regarding it as a problem. That is, variation isn’t a problem, it’s a
fact. But unfortunately, among linguists and grammarians of all periods,
there has always existed a human tendency to want to decide between two
competing variants, and depressingly often an instinct to want to call one
variant ‘right’ and the other variant ‘wrong’ rather than just accepting that
both exist. If there are two ways of expressing the same concept in a modern
language, and even in Late Latin, and if we choose our questions with care,
Google, for example, can tell us which of the variants is the commoner, and
we can, if we are of a mind to be impressed by statistics, choose that one as
belonging to a supposed norm for the period being investigated; even so, it
often seems fairer not to feel compelled to finger either of the variants for
such an honour. Such search engines operating over the whole Internet may
soon even supplant the largest dedicated corpora for those investigators who
are only interested in synchronic states of the present, being so easy to use.
The available evidence from the past can also be amassed into such
searchable stores, and is being so, as with Maurilio Pérez González’s
Lexicon of medieval Latin texts from León before 1230.
In practice our direct sources and our direct evidence for the past have to
depend on written manifestations of particular idiolects, even when the
collective data are vast enough to allow us to offer statistically valid
generalizations. But generalizations might not in total apply to any one
individual. And despite the caveats just mentioned, there are intrinsic gaps
between speech and writing which need to be taken on board. Not just
because the two registers always and inevitably manifest specific
differences; as well as the obvious contrasts between phonetic media, which
involve intonation, pitch, rhythm, sandhi, and other phenomena
unrepresented in writing systems, and graphic media, which involve
punctuation, spaces between words, capitals, fonts, etc., there are the large
statistical morphosyntactic distinctions referred to above, such as the greater
presence of deictics and diminutive suffixes in speech, and of subordinate
clauses and third-person verbs in writing. But the greatest difference
between the two modes is one which in practice has often not been taken
into account by the diachronic Latinist or Romanist; that is, under normal
circumstances, speech comes to us all naturally, but, in any literate society,
134 R. WRIGHT

writing needs to be taught. Writing at all, even if it’s done badly and
clumsily and unintelligibly and inaccurately, needs the writer to have been at
some point taught how to do it. However incompetent written evidence may
be, it still comes from a literate source. And to the eternal disappointment of
all sociolinguists and philologists, and of the sociophilologists who try to
combine both kinds of expertise and insight, elementary students have not
usually been taught to write their own linguistic habits in a phonetic script.
In all societies, teachers always aim to have their pupils writing their words
in their traditionally correct graphic shape, even in societies where the
standard orthographical systems were established in a long-distant past time
which pre-dated any number of phonetic changes. Indeed, in the pedagogical
traditions of most societies, any graphical variant which reflects such
phonetic changes is ipso facto by definition wrong and to be chastised. In
this respect, writing disguises speech rather than providing a photograph of
it.
This means that as a general rule we can take spelling mistakes as
potentially being evidence of evolved phonetic features, but correct spelling
is not to be regarded in itself as evidence of unchanged phonetics. Such
orthographic conservatism has to be taken into account by the analyst. A
letter h, for example, in Romance and in Latin, has often failed to attest a
sound [h], both when found as itself (e.g., in such words as French huit,
Spanish hallar) and as part of a digraph; for instance, the modern Portuguese
and French digraph ch represents the sound [ʃ], the Italian ch represents the
sound [k] and the Spanish ch represents the sound [tʃ], but nowhere does ch
represent the sounds [kh], and in none of these cases does h represent [h].
Such digraphs were popular when Latin borrowed Greek words containing
chi, theta, phi or rho, and the digraphs ch, th, ph and rh came to be used
when writing those words in a Latin text; whether the Latin pronunciations
of such words ever included the same aspirations as the Greek is rarely clear,
but even if they did the aspirations certainly did not continue into Late Latin,
whether or not the written form with one of these digraphs was still being
used.
The misguided advice of teachers has often been invoked to explain
common features of written Late Latin, but it could still come as a surprise to
realize how many details this helps to explain. Sometimes the new form
became a kind of alternative correct form; for example, we’re used to the
fact that in Late Latin the word mihi is often written with the letters ch rather
than h, michi. The digraph wasn’t included because there was an aspirated
velar in the writers’ spoken Romance; it’s there merely because they were
told to write the word like that. We’re also used to the idea that scribes must
regularly have been told to write the open [e] sound with the letters ae as a
Late Latin: the evidence 135

general rule, seeing that they so often did this in words where the word with
the open [e] vowel had never included a diphthong at all. It’s clear from the
concordances prepared at Santiago de Compostela that the seventh-century
writers in the Iberian Peninsula often wrote the word iste as ste (without the
initial letter i-) and Hispania as Spania, (without the initial Hi-); probably
because they were told not to write any letter corresponding to the initial
epenthetic vowel of such Romance words as [estar], originally [stare]
(stare), and mistakenly identified the words pronounced as [este] and
[espaɲa] as coming into that category. Carmen Pensado pointed out many
years ago (1991/1996) that the spelling of eglesia for ecclesia is so common
in tenth-century León that the scribes must have been taught the word that
way. But textual change can happen even when phonetic change hasn’t.
Twenty years ago I studied a remarkable example from tenth-century Galicia
(Wright 1995: chapter 12), that of Latin SOBRINUM ( meaning “cousin”),
which became eventually Galician sobrinho (meaning “nephew”); this
phonetic trajectory seems to leave no scope for confusion in the first four
letters, since the first four sounds in this word did not change, they were
always [sobr-], but the word turns up in the documentation from tenth-
century Galicia more often with super- or supr- than with any other initial
combination of letters. We can hardly postulate that it was said with [supr-],
and this written form can only have been chosen because the Ibero-Romance
word [só-bre] was written every day, correctly, as super (both as a free word
and as a prefix), and the forms of SOBRINUM written with super- or supr-
must be the result of misapplied intelligence from scribes or their teachers
who wrongly, but intelligently, identified the initial sounds of the word,
[sobr-], as a prefix. Such misdirected intelligence is at times a more plausible
explanation of ostensibly absurd written forms than is the capricious
stupidity and barbarous ignorance often imputed to their writers by
generations of smugly unsympathetic modern Romanists.
There are similar phenomena as regards morphology and syntax. For
example, the prevalence of sequences of nouns with apparently randomly
chosen morphological endings in many Late Latin texts seems best
explained by a pedagogical instruction to scribes that they should not use as
many prepositions in writing as they used in speech, where prepositions had
taken over many of the semantic functions of the nominal endings. This
instruction, naturally, assumes that the scribe can achieve appropriate
morphological endings. But the Late Latin evidence includes many
sentences in which neither appropriate endings nor appropriate prepositions
appear. In a study of the nominal morphology of the Acts of the 839 Council
of Córdoba, for example, I considered the following passage (folio 2r, lines
5-6); conubia fidelium quum infideles uarietat inpietas crimina moribus
inserentes. We can deduce from the context that this criticism of the heretics
136 R. WRIGHT

who are the subject of the Council is intended to mean that “with their
customs they have introduced marriage of Christians with non-Christians
and a variety of other unacceptable crimes”. There is one preposition in the
text (quum, for cum), and there are four in the English or Spanish
equivalents, and probably also in the speech of the time. All the words used
in the sentence are genuine traditional Latin words and also, probably, part
of the normal ninth-century vocabulary, at least among Church people. And
the inflections themselves are also genuine; it’s the syntax that binds them
together which is lacking on paper, and which I deduce to be the result of
some kind of pedagogical preference for not using as many prepositions in
writing as they had used when preparing this comment orally (Wright 2005:
500).
That’s an example in which I suggest the scribes have been told not to
represent on paper a feature that was normal in their speech. Conversely,
they could also be encouraged to add features to written texts which were no
part of their speech at all. There seems, for example, to be no doubt that
scribes were encouraged in some places to use as many verbs with ostensibly
passive morphology as they could, on the grounds that it looked good; this
certainly seems to have been the case in the successive versions of the
Chronicle of Alfonso III prepared in Oviedo (Wright 2014). There is, for
example, in the Rotense manuscript, chapter 10 (Gil et alii 1985), this
sequence of synthetic historic present forms at an account of the start of a
battle: arma adsumunt, eriguntur fundiuali, abtantur funde, migantur enses,
crispantur aste hac incessanter emittuntur sagitte…. There is no need for
these verbs to be passive at all, and the soldiers could have been presented
throughout as the subject of active verbs, as they are of adsumunt at the start
of the sequence; and the easiest interpretation of the choice of verb forms
seems to be that the scribes have been encouraged to write that way. It’s
certainly a more likely interpretation than the idea that all these verbs have
suddenly become systematically deponent, which would be Pierre Flobert’s
analysis (Flobert 1975). One corollary of this recommendation of synthetic
passive morphology in writing could well have been that the scribes were at
the same time discouraged from using, in writing, reflexive grammar with
se, a word which is surprisingly rare in written Late Latin even with its
original syntactic function.
And there are individual cases which seem at first inexplicably capricious
till we consider the possible effect of the scribe’s teacher. There is an
extraordinary example in a document of the year 991 found in the archives
of the Monastery of Vairão in Galicia (Wright 2010b). In the middle of the
document there is a list of the witnesses to a previous agreement, where we
see the word ante followed by the full names of five individuals, who are
Late Latin: the evidence 137

also as it happens the five main witnesses to the document in question, and
these in turn are followed by the phrase et aliorum multorum filio bonorum.
It seems to be meant to mean “in front of these five witnesses and many
other good people”. But literally it says “the son of many other good
people”. Who is this “son” with many parents? I don’t want to deduce that
the scribe has gone temporarily mad; I would prefer to find a reason for the
use of these words. Well, Menéndez Pidal, in his vocabulary of the Poema
de Mio Cid (1908: 691), under the word fijodalgo, mentions the phrase multi
filii bene natorum in a document of 1020 from Palencia, meaning “good
solid people”. So there seems to have been a phrase filius boni, “son of a
good man”, meaning “a good person” in the same way as fijo d’algo later
did, “the son of something”, which became the single word hidalgo;
normally the plural of the phrase would have been filios boni, but I suggest
that our scribe here, conceiving of the phrase as being a single word, and
knowing from his education that the plural of boni would be bonorum,
deduces from this that the plural of filioboni should be filiobonorum, and
conscientiously gets the two adjectives to agree with that as well; for even
though the -orum ending cannot have been live in the speech of tenth-
century Galicia, it was a genuine entity in the scribal mind as a result of their
training. There is such a small space between filio and bonorum in the
document that this one-word analysis is plausible. A modern Latinist, or
Romance philologist, will be tempted to dismiss this phrase as a mere
whimsical act of crass stupidity; I prefer to see it as misapplied intelligence
from a scribe trying his best to do as he was told.
There are other potential sources of evidence for our analyses, such as
explicit comments on linguistic matters. And yet similar problems can also
be present in otherwise potentially helpful metalinguistic comments. That is,
the grammarians of Late Antiquity can make illuminating remarks on
contemporary practice, but they tend to tell us what they thought ought to
happen rather than what actually did. The literate learnt to write in the Early
Middle Ages from teachers who knew the tradition, and perhaps
occasionally even the text, of the Ars Minor of the fourth-century scholar
Aelius Donatus (who taught Jerome); but Donatus hadn’t known that he was
founding a long pedagogical tradition, or that he was educating the
apprentice scribes of the future, since he was writing for contemporaries
wishing to read the classics of the past. Speech was hardly on his radar at all,
and, in so far as it was, his data are likely to have been out of date even at the
time he wrote. Some grammarians are more helpful; Velius Longus in the
second century, for example, and the fifth-century grammarian Consentius,
probably working in Africa, sometimes tell us about speech. Consentius is
regarded by József Herman (2000), for this very reason, as his favourite
grammarian. The metalinguistic source need not be a specialist linguist at
138 R. WRIGHT

all; Augustine, for one, gives a few revealing explicit insights into his speech
(the Latin of Africa around 400 A.D.). One kind of source that has so far
hardly been exploited by modern Romanists, being so difficult to analyse, is
that of the monolingual Latin Glossaries, particularly those which derive
from the seventh-century Hispanic tradition, although I have made initial
attempts to use them (Wright 2008). It is also true, unfortunately, that not
everybody who comments on language has sensible things to say, and the
analyst always needs to remain wary.
The point being made is that the gap between speech and writing is
largely created, or at least exaggerated, by education. When we’re trying to
work out the nature of spoken Late Latin (Early Romance), therefore, the
way in which scribes were trained needs to be studied explicitly when it can
be. Unfortunately, there is no necessary direct connection between textual
change and language change, and the relationship is not the same in every
linguistic aspect. Even languages which are written with alphabets tend to
have teachers and speakers with logographic rather than phonographic
attitudes to writing individual words. Spelling reform, however desirable it
may seem to the outsider, always meets obstacles in practice, because in any
language, for those who have learnt to read, it is naturally easier to read, as
well as to write, the traditionally correct form, rather than a novelty inspired
by phonetic transcriptions; while those who cannot read at all cannot read
any form, so a form based on a phonetic transcription would be of no more
use to them than the traditional counterpart.
The consequence is that phonetic change can progress a long way without
evidence for it being directly attested in a written source. The realization of
this dissociation was the main motivation for the twentieth-century
reconstruction of ‘Proto-Romance’ on the basis of extrapolating backwards
from Medieval Romance written evidence (taken to be largely
phonographic) rather than by examining the written texts provided for us by
those who supposedly spoke it. This enterprise was modelled on the previous
reconstruction of the “Proto-Indo-European” spoken five thousand years ago
or more, wherein there is perforce no contemporary written evidence to take
into account (Hall 1976, 1983; De Dardel 1996). The changes implied in the
Proto-Romance reconstructions cannot be dated by this method, but Hall and
his colleagues effectively established most of the phonetic features of spoken
Latin. Reconstruction of an ancient state of a language on the basis of later
data has helped analysts work out details of many languages all over the
world, and is enthusiastically defended by historical linguists (such as Lass
1997), but it is also always recognized to be a faute-de-mieux kind of
procedure; the specialists in Proto-Indo-European themselves are delighted
when new ancient written evidence is discovered, a delight which presents a
Late Latin: the evidence 139

marked contrast with, for example, De Dardel’s insistence on never using


written sources, apparently on the grounds that preferring written testimony
of the Late Latin period to reconstructions based on later Medieval Romance
is “unscientific”.
Another problem with reconstruction theory has been that until recently it
has seemed unable to accept variation; in particular, any variant
reconstructed or attested in the past had in that theory to be allotted to a
separate dialect or language in which that was the canonical form, rather
than being allowed to be a simple variant in a single multivariable
vernacular. Thus the many sub-branches of Proto-Romance, in particular as
envisaged in geometric tree diagrams by Hall (1974), including various
kinds of “Gallo-Romance”, etc., are of necessity dated so early (during or
even before the Roman Empire) as to seem unconvincing to the classicists
and historians who know that even if Latin during the Roman Empire varied
diastratically and geographically, it was still one language.
Even so, despite all the caveats, phonetic evidence can be attested in
writing. It is particularly worth considering such possibilities at length in the
Late Latin context, in the event, where the evidence is most thorny. It is at
times possible to deduce differences in the vernacular pronunciation of
different places from the different errors made by scribes, as Herman (1990)
could do from epigraphic inscriptions, despite the reservations expressed by
Adams (2007: chapter X), and as Janson (2008) was able to do concerning
vowels in the Merovingian and Lombard areas. Sometimes the attestation of
evolved phonetics in a written text is the consequence of a deliberate act by
the scribe. Occasionally, though rarely, this results from a wholesale
decision to write everything in a text in a new more phonographically based
manner; this is what happened with the Strasbourg Oaths and the other early
complete texts in Romance (see Ayres-Bennett 1996). Previously, and more
commonly, such intentional novelties resulted from an attempt to represent
in writing a single word that had no standardized written form, rather than a
whole text. In Late and Medieval Latin this could apply particularly to words
borrowed from other languages, such as from Germanic in Gaul or, later,
from Arabic in the Iberian Peninsula. Personal and place names often offer a
way into the problem for the modern investigator, although such extraneous
considerations as popular etymology can make even these processes less
than transparent. Usually the problem addressed by the scribe was how to
write a non-Latinate name using the symbols available to him in the Roman
alphabet, but occasionally new letters could be adopted from elsewhere. The
originally Greek letter zeta was adopted to represent the affricate sound [dz];
it is particularly common in the patronymic suffix of witnesses to tenth-
century legal documents in the Iberian Peninsula, usually written as -iz or -
140 R. WRIGHT

izi; this led to Modern Spanish -ez, as in common names such as Sánchez.
Otherwise new letters were not adopted, not even from the Arabic alphabet
in order to represent Arabisms. The name just mentioned was occasionally
written as Sanggiz, but not because it was ever pronounced with double [gg];
the scribes and their teachers were casting around for a way to represent the
sound [tʃ] with the symbols that they had, given that Latin had had no
equivalent, and gg was one possibility bruited before the general recourse to
the digraph ch.
The earliest sources usable as a whole as evidence for Romance are those
in which the spelling system has been deliberately adapted in a novel way to
represent the colloquial sound of existing words via new combinations of
existing letters (as in the Strasbourg Oaths of 842). Thus in such textual
evidence the novelty lies mainly in new written forms of words. Direct
evidence for other aspects of Romance had been attested in writing already
for a long time. Romance syntax (including word order), Romance
morphology (including the loss of obsolete inflections, unless scribes had
been explicitly told to use them), Romance semantics (in which old words
acquired new meanings) and Romance vocabulary (whether borrowed from
elsewhere or internally created by normal derivational mechanisms) could all
be represented in the normal traditional writing systems before the time of
the Strasbourg Oaths. There had been no problem there; the teachers had not
needed to change their pedagogical practice at all. For example, when the
various forms of ILLE came to be used in speech as what we would think of
as definite articles, as well as with their older function of pronouns, there
was no need to change any of the written representations of the word;
whether article or pronoun, whether clitic or not, the four letters of the word
ille, and the other graphical forms of the lexeme, were taught and learnt and
written and still thought of as correct. When HABERE came increasingly to
be used as an auxiliary, there was no corresponding need to change the
written form; whether auxiliary or full verb, and however pronounced,
HABEO could continue to be written as habeo. When grammatically
reflexive SE increasingly found itself being used with non-reflexive non-
agentive semantics, there was no need to change its written form. When DE
changed meaning and function to rival and then overtake those of genitive
inflections, it could continue to be written as de. And so on: as in all modern
languages, so in Late Latin, new grammatical and semantic features could
easily be accommodated within existing spelling systems.
Thus features of morphology, syntax, semantics and vocabulary which
we would now prefer to call ‘Romance’ rather than ‘Latin’ were often
attested in writing long before the deliberate decisions were made to change
the spelling systems as a whole. This presence of Romance features in
Late Latin: the evidence 141

ostensibly Latin textual evidence has become obvious now in the evidence
from seventh- and eighth-century Gaul, most notably as the result of the
illuminating and groundbreaking research of Michel Banniard and his
colleagues at Toulouse, and of Marieke Van Acker at Ghent (e.g., Banniard
1992, 1993; Verdo 2010; Van Acker 2007). Francesco Sabatini’s work in the
1960s on Romance elements in sixth-century Italian documentation was
illuminating in itself (e.g., Sabatini 1968; see also Sabatini 1983), and
provided a lightbulb moment for António Emiliano’s perspicacious and
acute studies of similar, though rather later, evidence in texts from Portugal
(Emiliano 1999, 2003; Emiliano and Pedro 2004). I’ve prepared similar
studies on Ibero-Romance Latinate evidence, usually with reference to
particular documents, although also based on the concordances of the texts
in the Cathedral Archive at León (e.g. Wright 1995, 2003, 2004). The only
conclusion that can be drawn is that, all over the Romance-speaking world,
most of the texts of the centuries preceding the arrival of the Carolingian
Reforms into a community, apart from those of the most deliberately
antiquarian writers such as Avitus of Vienne, can be exploited as valuable
sources of evidence for Romanists, if used with care, and above all if not
‘emended’ by modern editors who assume that all scribes were trying to
imitate Cicero (cp. Lass 2004: 31, “an amended text is a falsehood”). Even
relatively well-known Late Latin and Medieval Romance texts are likely still
to contain linguistic data as yet imperfectly understood, which are in danger
of being lost to scholarship if a modern editor decides not to let his readers
see them; this invisibility of the linguistically interesting arose partly
because several Late Latin works have been studied mainly, or even only,
for their authors’ knowledge of earlier sources, which is of little relevance to
us, for it is precisely those aspects that have no source which are likely to be
of interest to linguists. Lindsay’s work on glossaries is unhelpful to us for
this reason.
Texts written in Gaul can be used as evidence of speech up until the late
eighth century. Then written and spoken Latin were integrally involved in
the “Carolingian Reforms” of around 800 A.D. Their educational component
aimed to standardize linguistic usage and pedagogy along antiquarian lines
(Wright 1982: chapter 3). As a consequence, the relation between written
evidence and natural spoken usage became more opaque, and the idea that
this newly reestablished educated Latin was a separate language from the
normal vernacular began to spread after that time. These reforms spread to
most of the Romance-speaking world outside the Carolingian realms later,
with the intellectual movements summarized as the “Twelfth-Century
Renaissance”. Thus the earlier, pre-Reform, textual sources are evidence of
speech to a greater extent than those written later. Syntactically, there is little
in the pre-Reform textual evidence which cannot be accommodated in a
142 R. WRIGHT

plausible account of the grammar of the speech of the time (unless the author
was deliberately imitating a model from the distant past). For example,
nowadays people tend to think that the main feature of Latin syntax was that
verbs came at the end of their sentence. This word order was perhaps
statistically the most common during the Roman Empire, but it was in no
sense obligatory, as several studies of Latin word order have shown (e.g.
Pinkster 1991/96; Spevak 2010). 19th-century schoolboys were told to put
the verbs at the end of Latin sentences, but Roman grammarians and teachers
didn’t tell their pupils to postpone the verbs in this way, and from the textual
evidence such an instruction doesn’t seem to have formed a part of most
scribal training. It is true that verbs in modern Romance, particularly in
Spanish, often appear earlier in their sentence than they would have done in
the first century A.D., and earlier than they would in English, but that is also
true of much of the documentary evidence from the Late Latin of the pre-
Reform period, including in both eighth-century France and tenth-century
Iberia. József Herman’s succinct account of Vulgar Latin demonstrates how
such non-Classical grammatical features in particular were attested in
writing (Herman 2000).
The Late Latin texts tend to manifest old and new features together. For
example, the same document can include both inflected genitives and
prepositional phrases with de, with similar meaning; both accusative plus
infinitive constructions and synonymous constructions with quod.
Yet later Romance evidence allows us to reconstruct the continuing
presence of the older feature in the speech of the time. Cases other than the
nominative and accusative had dropped from nouns and adjectives before the
time of the earliest Romance texts (and even this distinction survived only in
Gallo-Romance and Ræto-Romance), but oblique cases (other than vocatives
and ablatives) had not disappeared from pronouns, so the genitive itself was
not alien to the speakers: Italian loro and French leur derive from
ILLORUM, for example, and still survive, and although leur is now best
analysed as an adjective meaning “their”, with its own plural form in leurs,
loro still seems to be a pronoun meaning “of them”, with no feminine **lora
nor plural forms in **lori and **lore. Proper names could still have genitive
forms in Old French as well as in the Late Latin texts (e.g., in the phrase
ecclesia Sanctae Mariae). And the accusative plus infinitive construction has
not entirely gone from Romance either even now, particularly with verbs of
perception (e.g., Italian ci vede venire, “she sees us arrive”). So even though
genitive inflections and accusative plus infinitive constructions may indeed
have been recommended in class, and were thus on the list of those features
which were commoner in writing than in speech, they were not as alien to
speech as we might suppose. It would be altogether misleading to see such
Late Latin: the evidence 143

data as evidence of a mixture of different languages; theirs was one single


language still at the time we are considering, which had competing variants
as all literate languages do, some of which were newer in origin than others.
The two were only going to be conceptually separated out at a later time
when the general idea arose that there were two languages involved here
rather than one, Latin and Romance, a conceptual recategorization which it
seems can only have followed the development of new written Romance
modes rather than preceded them.

Conclusion
Late Latin is not a special case. Such a combination as we often see
attested in the evidential sources from the Late Latin pre-Reform
documentation, of contemporary semantics, vocabulary, morphology and
grammar with traditionally correct orthography, is exactly what literate
speakers and writers of English, French, Spanish, Portuguese, Italian,
Catalan, Romanian, Galician, etc., do every day of the week. It is what one
would naturally expect in any language community in which the spoken
language changes, unavoidably, and the graphical forms taught to apprentice
scribes remain the same; which may well be all such language communities.
So it is hard to see why this perspective is sometimes thought to be
controversial when applied to Early Romance. Reconstruction is valuable,
but written evidence can and should be used as the prime source for our
analysis of speech, as long as we continually bear the context of production
in mind. There is no text without context, as the sociolinguists say. And
diachronically, this kind of analysis is the role of sociophilology, which aims
to combine sociolinguistic insight with philological concentration.

REFERENCES
Adams, James N. (2007): The Regional Diversification of Latin, 200 BC - AD 600,
Cambridge: University Press.
Ayres-Bennett, Wendy (1996): A History of the French Language Through Texts, London:
Routledge.
Banniard, Michel (1992): Viva Voce: communication écrite et communication orale du IVe au
IXe siècle en Occident latin, Paris: Institut d’Études Augustiniennes.
— (1993): “Latin tardif et français prélittéraire: observations de méthode et de chronologie”,
Bulletin de la Société Linguistique de Paris 88: 131-154.
Davies, Wendy: “Local priests and the writing of charters in northern Iberia in the tenth
century”, in Escalona, Julio and Hélène Sirantoine, eds (2013), Chartes et cartulaires
comme instruments de pouvoir: Espagne et Occident chrétien (VIIIe-XIIe siècles), Madrid:
CSIC, 29-43.
144 R. WRIGHT

De Dardel, Robert (1996): A la recherche du protoroman, Tübingen: Niemeyer.


Emiliano, António (1999): “O mais antigo documento latino-portugués (882 A.D.) - edição e
estudo grafémico”, Verba 26: 7-42.
— (2003): Latim e Romance na segunda metade do Século XI, Lisbon: Fundação Gulbenkian.
Emiliano, António and Susana Pedro (2004): “Da Noticia do Torto”, Zeitschrift für
Romanische Philologie 120: 1-82.
Flobert, Pierre (1975): Les verbes déponents latins des origines à Charlemagne, Paris: Les
Belles Lettres.
Gil, Juan, José-Luis Moralejo and Ignacio Ruiz de la Peña, Ignacio, eds (1985): Crónicas
asturianas, Oviedo: Universidad.
Hall, Robert A. Jr. (1974): External History of the Romance Languages, New York: Elsevier.
— (1976): Proto-Romance Phonology, New York: Elsevier.
— (1983): Proto-Romance Morphology, Amsterdam: John Benjamins.
Herman, József (2000): Vulgar Latin, Pennsylvania: Penn State Press.
Janson, Tore: “The vowel systems of Late Latin”, in Wright Roger, ed. (2008), Latin vulgaire
~ latin tardif VIII, Hildesheim: Olms-Weidmann, 345-350.
Lass, Roger: “Ut custodiant litteras: editions, corpora and witnesshood”, in Dossena, Marina
and Roger Lass, eds (2004), Methods and Data in Historical Dialectology, New York:
Peter Lang, 21-48.
Menéndez Pidal, Ramón (1908): Cantar de Mio Cid: Vocabulario, Madrid: Espasa-Calpe.
Oesterreicher, Wulf: “Textos entre inmediatez y distancia comunicativas. El problema de lo
hablado escrito en el Siglo de Oro”, in Cano, Rafael, ed. (2004), Historia de la lengua
española, Barcelona: Ariel, 729-769.
Pensado, Carmen: “How was Leonese Vulgar Latin read?’, in Wright, Roger, ed.
(1991/1996), Latin and the Romance Languages in the Early Middle Ages, London:
Routledge (reprint, Pennsylvania: Penn State Press), 190-204.
Pérez González, Maurilio (2010): Lexicon Latinitatis Medii Aevi Regni Legionis (s. VIII-
1230) Imperfectum, Turnhout: Brepols.
Pinkster, Harm: “Evidence for SVO in Latin?”, in Wright, Roger, ed. (1991/1996), Latin and
the Romance Languages in the Early Middle Ages, London: Routledge (reprint,
Pennsylvania: Penn State Press), 69-82.
Sabatini, Francesco (1968): “Dalla scripta latina rustica alle scriptae romanze”, Studi
Medievali 9: 320-358.
— (1983): ‘Prospettive sul parlato nella storia linguistica italiana’, in Leoni, F. A., et alii, eds
(1983): Italia linguistica: Idee, Storia, Strutture, Bologna: Il Mulino, 167-201.
Spevak, Olga (2010): Constituent Order in Classical Latin Prose, Amsterdam: John
Benjamins.
Van Acker, Marieke (2007): Ut quique rustici et inlitterati hec audierint intellegant:
hagiographie et communication verticale au temps des mérovingiens (VIIe-VIIIe siècles),
Turnhout: Brepols.
Verdo, Remy (2010): “Reconfiguration du latin mérovingien sous les carolingiens: étude
sociolinguistique des diplômes royaux et des réécritures hagiographiques (VIIe - IXe
siècle)”, in Positions des thèses soutenues par les élèves pour obtenir le diplôme
d’archiviste paleographe, Paris: École des Chartes, 257-265.
Wright, Roger (1982): Late Latin and Early Romance in Spain and Carolingian France,
Liverpool: Francis Cairns.
— (1995): Early Ibero-Romance, Newark [Delaware]: Juan de la Cuesta.
— (2003): A Sociophilological Study of Late Latin, Turnhout: Brepols.
— (2004): “La representación escrita del romance en el Reino de León entre 1157 y 1230’, in
Fernández Catón, José María, ed. (2004), Orígenes de las lenguas romances en el reino
de León, siglos IX-XII, León: Archivo Histórico Diocesano, 273-291.
Late Latin: the evidence 145

— (2005): “Nominal morphology in the Acts of the Council of Córdoba (839): a


sociophilological investigation”, in Archer, Robert et alii, eds (2005), Antes y después del
Quijote, Valencia: Generalitat Valenciana, 497-505.
— (2008): “The monolingual Latin glossaries of the Iberian Peninsula: can they help the
Romanist?”, in Van Acker, Marieke et alii, eds (2008), Latin écrit - roman oral? De la
dichotomisation à la continuité, Turnhout: Brepols, pp.137-158.
— (2010a): “Bilingualism and diglossia in Medieval Iberia (350-1350)”, in Cabo
Aseguinolaza, Fernando, et alii, eds (2010a), A Comparative History Of Literatures In
The Iberian Peninsula, Vol.I, Amsterdam: John Benjamins, 333-350.
— (2010b): “Gontigius, Sagulfus, Domitria y el hijo de muchos otros buenos”, in Cano
González, Ana María, ed. (2010b), Homenaxe al Profesor Xosé Luis García Arias, Vol.I,
Oviedo: Academia de la Llingua Asturiana, 407-420.
— (2013): “Evidence and Sources”, in Maiden, Martin et alii, eds (2013), The Cambridge
History of The Romance Languages, Vol.II, Cambridge: University Press, 125-142.
— (2014): “Passive morphology in Late Latin”, in Garcea, Alessandro et alii, eds (2014),
Polyphonia Romana: Hommages a Frédérique Biville, Vol.I, Hildesheim: Olms, 81-90.

You might also like