You are on page 1of 11

OUP CORRECTED PROOF – FINAL, 25/5/2016, SPi

CHAPTER 1

Latin as a source for the


Romance languages
JAMES CLACKSON

1.1 Chronological and spatial scope stylized poetry of Vergil (Publius Virgilius Maro, 70-19 BC),
of Latin Horace (Quintus Horatius Flaccus, 65-68 BC), and Ovid (Pub-
lius Ovidius Naso, 43 BC-17/18 AD) written in quantitative
metres based on Greek models, and sometimes employing
The traditional date for the founding of Rome is 753 BC, but
extreme hyperbaton and recondite vocabulary; through
our earliest records for the Latin language date from around
speeches and historical works written in prose, which can
200 years later. Only a few Latin texts survive from before
combine multiple subordinate clauses to achieve highly
200 BC, but after this date there is a wealth of material, both
elaborate periodic sentences; to dramatic comedies, satires,
through inscriptional sources and transmitted manuscripts
and novels that sometimes appear to imitate different
of literary and sub-literary compositions. The corpus of
speech styles of individual characters. Texts, usually surviv-
surviving Latin texts up until AD 476, the time of the depos-
ing in manuscript form, which treat technical matters, or
ition of Romulus Augustulus, the last Roman emperor in the
instruction manuals, such as medical or veterinary trea-
west, is greater than ten million words, with well over
tises, works on military tactics or land surveying, cook-
100,000 surviving inscriptions. Documents or inscriptions
books, and the like, are usually written in less elaborate
written in Latin have been found throughout the Roman
styles of Latin, and sometimes referred to as ‘sub-literary’
empire, as far north as Hadrian’s Wall in Britain (an import-
texts. Epigraphic and documentary evidence also encom-
ant new source of Latin documents written on wooden
passes a broad range, comprising carefully composed offi-
tablets) and as far south as Philae in Egypt. Texts in Latin
cial inscriptions recording laws or imperial decrees; poorly
have been found in all the regions now populated by
scratched or painted graffiti from walls at Pompeii and
speakers of varieties of Romance in the Romània continua
other Roman settlements; laundry lists, letters, and
and also in other areas, including Britain, north Africa, and
accounts from military camps and the like; and countless
central Europe, where the modern languages do not derive
examples of funerary epitaphs left in memory of patrons,
from Latin. Indeed, many of the most important Latin
relatives, and friends (see Cooley 2012 for a recent survey).
authors of the imperial and later periods were born outside
To add to this, there is a rich tradition of grammatical
the domain of modern Romance languages: St Augustine
literature in Latin, which, though largely prescriptive,
(Aurelius Augustinus, AD 354-430) was a native of what is
sometimes can give insights into the speech of individuals
now Algeria; St Jerome (Eusebius Hieronymus, AD 347-420),
and social groups.
whose translation of the Bible into Latin is still in use, was
born in the Roman province of Dalmatia, probably in what is
now Slovenia; Priscian (Priscianus Caesariensis, fifth–sixth
century AD), author of the most extensive and most import-
ant ancient grammar of Latin, was born in north Africa and 1.2 Classical and vulgar Latin
spent his working life in Constantinople (now Istanbul);
Tribonianus (died AD 541 or 542), one of the compilers of Latin evolved into the Romance languages, and it is natural
the Latin codification of laws under Justinian, was a native to seek antecedents to linguistic developments in Romance
of Side, now in southern Turkey. within Latin. The huge range of evidence for Latin offers
The chronological and spatial spread of Latin is conse- many possibilities to the researcher. In some accounts
quently vast. The corpus of surviving Latin texts also pre- (e.g. Solodow 2010:107, see also other references in Adams
sents an enormous variety of material written in different 2013:7), the Roman world in the last century of the republic
registers. Literary texts range across genres: from the and during the empire is presented as diglossic. According

The Oxford Guide to the Romance Languages. Adam Ledgeway and Martin Maiden (eds)
This chapter © James Clackson 2016. Published 2016 by Oxford University Press. 3
OUP CORRECTED PROOF – FINAL, 25/5/2016, SPi

JAMES CLACKSON

to this view, so-called ‘Classical Latin’ was the H(igh) var- Italy also take place in the transition from Latin to Romance
iety, which was also the written form taught at school and (see Coleman 2000 for a survey of various phonological
employed by members of the Roman elite. This is the form changes). These include the palatalization of velars before
in which nearly all literary texts and official inscriptions are front vowels (observable in Umbrian); loss of final conson-
composed. Classical Latin is itself the result of a period of ants (found in Umbrian); the monophthongization of diph-
linguistic standardization which took place some time thongs (again found in Umbrian); and a series of vowel
between 150 BC and AD 100, and is associated with precepts mergers, including between long e and short i (found in
and models laid down by authors such as Cicero (Marcus Oscan, but assumed for Umbrian as well). In §1.3, I shall
Tullius Cicero, 106-43 BC) and Caesar (Gaius Julius Caesar, discuss in more detail the case of the vowel mergers and the
100-44 BC) (see further Clackson 2011b). Followers of the monophthongization of the diphthong ae.
notion of a Latin diglossia often use the term ‘Vulgar Finally, it is assumed that some vulgar Latin forms are
Latin‘ to designate the L(ow) variety, which is presumed to found in Latin texts, because they have ‘leaked’ into the
be the spoken idiom of the lower classes (note however that compositions of particularly unskilled, or undereducated,
other scholars attribute other senses to the term ‘Vulgar individuals, who did not have sufficient competence or
Latin’, see Adams 2013:7-27 for a recent survey of some of training to manipulate the Classical Latin idiom. Palmer
the different senses accorded to the term; cf. also §2.5). (1954:149) uses a geological metaphor to convey this view
Vulgar Latin is held to be the ancestor of the Romance of vulgar Latin occasionally breaking through the inert
languages, but, according to the diglossia theory, its subor- crust of the Classical language: ‘[t]here are, as it were, in
dinate status usually prevented its representation in the the dead landscape of literary Latin, seismic areas where
written record. Consequently, in order to access the actual occasional eruptions reveal the intense subterranean activ-
source of Romance languages, different strategies may be ity which one day will make a new world of language.’
explored. The model of vulgar Latin as the ‘submerged’ language is
First, it is possible to examine texts written before the apparently supported by some Latin texts. Of these one of
process of standardization. The bulk of surviving literary the most important is the Satyrica, a Roman novel which
texts composed during the period before the standardiza- only survives in fragments, written in the first century AD by
tion of the Classical language are the comedies of Plautus Petronius (Gaius Petronius Arbiter, c. AD 27-66). The longest
(Titus Maccius Plautus, d. c.184 BC) and Terence (Publius extant portion of the novel, now known as the Cena Trimal-
Terentius Afer, d. 159 BC); these are dramatic works often chionis, describes in extravagant detail a dinner party hosted
based on Greek originals. Plautus and, to a lesser extent, by a wealthy freed slave, Trimalchio. Unusually among
Terence exhibit more variation in their language than later ancient writers of literary works, Petronius represents the
Classical authors, and consequently some of their linguistic speech of Trimalchio, his wife, and the band of fellow
features have been assumed to have been stigmatized in the freedmen at dinner as markedly different, in vocabulary,
Classical language and to have disappeared from the written sentence structure, and word-endings, from the speech of
record, only to emerge in Romance languages 1,000 years other characters in the novel and from the language of the
later (Pulgram 1958:321). An example from the lexicon is surrounding prose narrative. This can be illustrated from a
given in §1.5. section of a longer speech given by the character Echion,
Other possible sources for the pre-standardized spoken identified as a collector of rags. In this passage Echion
Latin, and also for the vulgar Latin of the regions outside addresses Agamemnon, a pedantic teacher of rhetoric,
Rome, have been found in the surviving inscriptions written who has been silent up to this point in the dinner:
in the sister languages of Latin from Italy, especially
Umbrian, attested from a region roughly corresponding to uideris mihi, Agamemnon, dicere: ‘quid iste argutat molestus?’
modern Umbria, and Oscan and the so-called ‘minor dia- quia tu, qui potes loquere, non loquis. non es nostrae fasciae, et
lects’ spoken in central and southern Italy until the first ideo pauperorum uerba derides. scimus te prae litteras fatuum
esse. quid ergo est?
century BC (see Marchesini 2009 for a recent survey). As
[Agamemnon, you look like you are saying ‘What is this bore
speakers switched from these varieties into Latin during
going on about?’ Because you, who can speak, don’t speak. You
the Romanization of Italy, it is possible that they continued are not one of our gang, and so you laugh at the words of us ’umble
to incorporate features of their native phonology into their folk. We know that you are mad for learning. What’s it all about
pronunciation of their second language; and hence by ana- then?]
lysing these languages, it may be possible to unearth some
detail about the local Latin spoken in Italy in the period Agamemnon is presented as ‘able to talk’, meaning that
from 200 BC to AD 100. Some of the developments which are he can speak correct or standard Classical Latin, but holding
seen in Oscan, Umbrian, and other languages once spoken in back from the conversation, and instead laughing at the

4
OUP CORRECTED PROOF – FINAL, 25/5/2016, SPi

LATIN AS A SOURCE

language of the freedmen, who belong to a different social the form a non b, i.e. ‘a not b’. For example pauper mulier non
group. It is certainly true that the narrator of this section is paupera mulier, which might be translated ‘[the correct form
poking fun at the language of the freedmen, since the is] pauper mulier “poor woman” not paupera mulier’. The
introductions and conclusions to their speeches include statement thus indicates that PAUPER is an adjective declined
ironic statements about the quality of their language such according to the Latin third declension class, and does not
as ‘Such sweet conversation . . . ’. It is also true that the group with adjectives such as INTEGER ‘whole’, which are
freedmen’s speeches contain infringements of the prescrip- declined in the first/second declension class, with a femin-
tions of the grammarians (and deviations from the vast ine INTEGRA. The phrase paupera mulier thus indicates that the
majority of literary usage). Some of these departures from adjective has made the same switch of declension class as
the norms are further mirrored in sub-literary documents that which was seen in Echion’s pauperorum, in the passage
and graffiti (see Adams 2003b for a recent collection) and of Petronius cited above. Palmer (1954:154) took the Appen-
many of them also win out over the Classical forms in the dix Probi to be ‘notes on current errors in speech’ and
later Romance languages. Two of the non-Classical features ascribed it to the third or fourth century AD, and likewise
of Echion’s speech will suffice to exemplify these features. Pulgram (1958:317), who dated it even earlier. However,
First, the word pauperorum is assigned to a different declen- consideration of the full text means that the compilation
sion class than is normal in Classical Latin (it should be dates from the fourth century or later, although some of the
PAUPERUM ‘of the poor’ the genitive plural of the third- observations may repeat grammatical precepts that have
declension adjective PAUPER), and we know that the gram- their origin as far back as the first century BC. More import-
marians censured this switch of declension (Palladius in Keil antly, it is clear that the list is concerned not with the
1855-80, IV:83f.) and it appears in the Appendix Probi (on correction of spoken errors but with written forms of
which see below). Some of the continuations of the word in Latin (as already noted by Grandgent 1908:5 ‘a list of good
Romance languages, such as It. povero (with F povera) show and bad spellings’, a statement changed in later editions, see
the same change in declension class (although in other also Clackson 2011c:518f.). Some of the written forms may
Romance varieties the change has not taken place: compare also have been castigated when used in speech (as in the
e.g. Sp. pobre). Secondly, Echion conjugates the verb mean- case of paupera mulier), but we cannot assume that the forms
ing ‘to speak’, LOQUI, as an active rather than deponent for which replacements are recommended necessarily
(hence loquere in place of infinitive LOQUI, and loquis in place occurred in the spoken language, and it is certainly incor-
of second person singular LOQUERIS). In Romance languages rect to imagine that all of these items were characteristics
the marking of verbs with active meaning through passive of a single dialect of spoken Latin.
morphology was also lost, and the beginnings of the process Petronius is therefore unique for the representation of
are observable in Latin literary and sub-literary texts writ- non-Classical Latin alongside the Classical variety of his
ten in later centuries (cf. also §60.3). But the view expressed narrative. Moreover, the evidence of Petronius does not
by some scholars that the deponent is avoided in spoken necessarily support the diglossia model of Classical and
Latin as early as the first century AD (see e.g. Pulgram vulgar Latin. Indeed, the freedmen themselves are not rep-
1978:223) is open to doubt, since there are in fact very few resented as all speaking a uniform dialect: characters
examples of the transfer of conjugation of deponent verbs exhibit various different non-Classical features (see the
to the active in the speeches of the freedmen in Petronius. collection of material gathered by Boyce 1991), and individ-
As Flobert pointed out (1975:308), most ‘vulgar’ writers in ual freedmen have idiosyncratic speech styles, such as the
this period are comfortable with deponents, whereas trans- freedman Hermeros, who code-switches between Latin and
fers from the deponent to the active are found in authors of Greek (see Adams 2003a:21); Petronius also seems to have
Latin of the highest literary pretension. drawn partly on literary models in constructing the speech
The representation of the speech of the freedmen in of the freedmen (see Leiwo 2010). Furthermore, failure to
Petronius therefore contributes to the conclusion that follow the precepts of grammarians in speech is not limited
some features of Latin were censured by grammarians and to Petronius’ freedmen, or indeed to any particular class of
avoided in the literary register, but present in speech. One Roman society. Several ancient sources remark, for
grammatical work in particular has also been often cited as example, on the impatience of the Emperor Augustus
a parallel for Petronius’ apparent direct citations of every- (63 BC–4 AD) with the precepts of the grammarians (some
day speech, the so-called Appendix Probi. This is actually the are collected at Adams 2007:16f.). Augustus made use of
third of five grammatical appendices to the Instituta Artium prepositions with the names of towns and had a preference
of Probus, found in a seventh–eighth-century manuscript for syncopated forms such as caldus in place of CALIDUS
now in Naples (and re-edited in Powell 2007). The Appendix ‘warm’. From much later in the history of Latin there is an
consists of a collection of over 200 word pairs, expressed in explicit statement of a grammarian, Pompeius (late fifth to

5
OUP CORRECTED PROOF – FINAL, 25/5/2016, SPi

JAMES CLACKSON

early sixth century AD) that a phrase in use in speech among In the remainder of this chapter I will exemplify these
the educated of his day (inimicus tuus est ‘he is your enemy’) points by consideration of three different areas in which
is not Latinus ‘(Classical) Latin’ (see further Adams 2013:13- attempts have been made to find the origin of Romance
15 for this and other examples and discussion). formations within Latin. These will be the changes in the
There is indeed little to support the view that the vowel system (cf. also §4.2.1, §25.1); the development of new
language situation during either the Roman republic or forms of the future tense (cf. also §46.3.2.2); and develop-
empire was one of diglossia, and statements in the sources ments in the lexicon (cf. also Ch. 32). More detailed work on
imply that Latin speakers of all classes varied their speech these topics, and many other aspects of the origins of
in different contexts, and that for all classes and at all Romance features within Latin can be found in the recent
periods there may have been a disjuncture between forms compendious work of Adams (2013), to which I shall make
used in the written language and those used in speech. frequent reference.
Cicero, for example, refers in one of his letters to himself
as using sermo plebeius, which can be literally translated as
‘plebeian language’, but this is better translated as ‘collo-
quial style’, and does not refer to a separate Latin variety.
Similarly, Quintilian (Marcus Fabius Quintilianus, c. AD 35-95)
1.3 Changes in the vowel system
refers to the conversational language of the Roman elite from Latin to Romance
between friends and family both as ‘everyday speech’
(sermo cotidianus) and as sermo uulgaris, a phrase normally Latin distinguished five short and five long vowels, which
rendered into English as ‘vulgar Latin’. When Latin authors will be distinguished in what follows through the use of a
of the later empire, including grammarians and Christian macron on the long variants: i ī e ē a ā o ō u ū. Roman authors
writers such as Jerome and Augustine, differentiate and scribes did not make a systematic distinction between
between educated forms which are used by the uulgus, i.e. long and short vowels in writing, although long vowels are
the ‘masses’, with what is Latinus, i.e. Classical Latin, they sporadically marked by various devices at different stages in
may be contrasting social distinctions of language, but they the history of the language, including doubling of the letter
may also be referring to words found in the written lan- representing the vowel (e.g. inscriptional paastores for Latin
guage (and not used in speech even by the educated) with PASTORES ‘shepherds’), the placement of a small mark called
spoken forms. an apex over the letter, and, in the case of the vowel ī, the
Since features which were stigmatized by grammarians use of an elongated form of the letter in inscriptions. Vowel
were in fact found in the language of all classes of society, it length is also revealed through metrical compositions,
is necessary to adjust the view that vulgar Latin was the which rely on patterned alternations between long and
submerged language of the underclass, driven underground short vowels. In the earliest Latin inscriptions there is
after the second century BC, and only able to break free from evidence for a series of diphthongs, including ai, au, ei, oi,
the veneer of Classical Latin at the end of the Middle Ages. It and ou, but a series of monophthongizations from the third
is certainly true that some spoken features were avoided in to the first century BC eliminated ei, oi, and ou, which mostly
the written language (most clearly certain vocabulary merged with inherited ī and ū (see Weiss 2009:100-104, 142f.
items, as will be discussed in §1.5), and that in the later for details of these changes). By the first century BC, there-
periods of the language the gap between what was said and fore, Latin only had two diphthongs attested in a large
what was written was considerable. However, speakers of all number of words, which are written ae and au (with oe, eu,
classes deviated from the written norms of Classical Latin, and ui restricted to a few lexical stems); in Latin verse,
and there is no a priori reason to assume that all of the diphthongs were scanned as equivalent to a long vowel.
features in Romance originated as ‘change from below’. The vowel system of Latin was to undergo substantial
Some of the linguistic developments within Latin may changes in the Romance languages. Length was lost as a
indeed have begun amongst the more wealthy or better distinctive feature of vowels, and instead vowel length
educated, and spread from there to all speakers; I shall became concomitant with the place of the accent (see
discuss an example in §1.4. In the words of Adams Loporcaro 2011a:53-8 and Adams 2013:43-51 for recent
(2013:858): ‘ “Submerged” does not necessarily mean “vul- accounts and surveys of the literature). Short vowels appear
gar”.’ The conclusion must therefore be that in assessing the to have been lengthened under the stress accent, and
origin of the Romance languages within Latin, it is import- unstressed long vowels shortened. A characteristic set of
ant to consider the widest range of Latin texts as possible in mergers takes place for stressed vowels. In general (exclud-
order to gain insight into how the written standard itself ing individual developments in peripheral areas such as
developed over time, and what lay behind it. Sardinia, Sicily, and in the east), ē merges with i and ō

6
OUP CORRECTED PROOF – FINAL, 25/5/2016, SPi

LATIN AS A SOURCE

merges with u; neither ī nor ū merges with any other vowel; assessing the impact of Oscan speakers on the Latin
the low vowels ā and a merge together. This seems to reflect language. This is because of the eruption of Vesuvius in 79
a qualitative difference between the long and short mid AD, and the consequent burial of several coastal communi-
vowels which is noted by some ancient writers on language: ties under the volcanic ash and lava flow, which led to the
the long mid vowels were pronounced more close than the preservation of large amounts of sub-literary Latin,
short mid vowels, and the short high vowels were pro- scratched or painted on the walls and buildings, or, remark-
nounced comparatively open. A further change was the ably, preserved in carbonized wax-tablets. There is a great
monophthongization of the diphthong represented by the deal of surviving documentary material, around 11,000
digraph ae, which merged in quality with the short vowel e. inscriptions in total, including around 8,000 graffiti and
The net effect of these changes is the establishment of a dipinti (as the painted inscriptions are known) from Pompeii
vowel system with distinctive high-mid and low-mid vowels and surrounding areas. It should not be assumed that these
in place of one with length distinctions but only three inscriptions present a unified linguistic picture, or that they
distinctive vowel heights. were all composed by members of the lower classes (see
As I mentioned in the previous section, there is good Kruschwitz and Halla-aho 2007 and Kruschwitz 2010 for
evidence that developments analogous to the Romance recent surveys of Pompeian graffiti and their value for the
changes took place in Oscan. In Oscan, vowel length can be linguist). However, inscriptions reveal that Pompeii was an
expressed by doubling of the letter used to represent the Oscan-speaking town before it became a Roman colony in
vowel, but in native texts the doubling is almost entirely the first century BC, and a few Oscan inscriptions from
restricted to initial syllables of words, supporting a hypoth- Pompeii date from the first century AD (Crawford et al.
esis that all words were stressed on their first syllable, and 2011:39; the latest datable inscription appears to have
that in non-stressed syllables the length distinction was been written in the decade before the eruption, Crawford
lost. Non-initial syllables show a three-way contrast in et al. 2011:702f.). Furthermore, the quantity of non-literary
front vowels, which are represented in the native alphabet Latin surviving from Pompeii and surrounding areas allows
of Oscan by the signs transcribed i, í, and e. The high-mid the researcher to make more widespread use of statistical
vowel, represented by the letter í, derives historically from techniques than work on historical linguistics normally
both *i and *ē. On the back axis, Oscan also shows merger of allows.
inherited *ō and *u, with both vowels normally written as u, In order to ascertain whether speakers have already
the accented ú representing the outcome of inherited *o. made adjustments to the Latin vowel system, scholars
Hence, for example, the writing dunúm ‘gift’ < *dōnom. It is have looked at what I shall call mis-spellings. If a word
usually assumed that Umbrian has undergone the same set containing the vowel ē is represented in script by a spelling
of vowel mergers as Oscan, although the writing systems with i, or the vowel i is represented by the vowel sign used
used for Umbrian never innovated a consistent way of for e, then this may indicate that the composer of the text
representing the triple distinction of vowel height, and has merged the two vowels in his or her everyday pronun-
the picture is obscured by a rule lowering u before nasals ciation. Väänänen (1966:20-22) collected the examples of
(Meiser 1986:51f., 120-22). As for the diphthongs, Oscan such mis-spellings from Pompeii. The picture is complicated
generally maintained the set which it had inherited from by various orthographic factors. For example, in the cursive
its parent language (i.e. ai, ei, oi, au, and ou); in Umbrian, all alphabet used for many of the low-level texts in Pompeii,
inherited diphthongs had been monophthongized by the the letter <e> is often written with two upright strokes, II,
third century BC. leading to the possibility that a careless or hasty attempt to
Similarities between the Sabellian (as the language group write <e> may result in something which is read as <i>.
which encompasses Oscan and Umbrian is now known) and Furthermore, in some cases writers select archaic orthog-
the Romance developments of the vowel system inevitably raphies of words, in which the long high front vowel ī was
lead to the question of whether there is any ground for spelt with the letter <e>, reflecting an early Latin pronunci-
connecting the two. Does Romance reflect, in the termin- ation of the outcome of monophthongized *ei (Weiss
ology of Seidl (1994:368), a ‘lingua romana in bocca sabellica’ 2009:101). Once these factors are taken into account, there
(‘Roman language in the mouths of Sabellian speakers’)? are hardly any good examples of mis-spellings of the
Much previous scholarship assumes it does (see Eska expected type within accented syllables. The only good
1987:147 for references to those who have accepted the remaining case is the word for ‘lucky’, FELIX (which is also
notion of an Oscan substrate acting on Latin, to which can frequent as a personal name), which is once written filix for
be added Coleman 2000). Unlike many other cases where fēlīx; however, the same word is also attested written felex,
language change has been hypothesized to reflect periods of which shows a highly anomalous representation of the vowel
prehistoric linguistic contact, there is better evidence for ī by the letter <e>. It is likely that in this word speakers tended

7
OUP CORRECTED PROOF – FINAL, 25/5/2016, SPi

JAMES CLACKSON

to enunciate the two vowels with the same quality, some- vowel of matching quality. The monophthongization of ae
times choosing to repeat the vowel of the initial syllable, to a long low-mid e-vowel seems to have been widespread in
sometimes that of the second syllable. In comparison, Latin judging from inscriptional and documentary evidence.
Väänänen found many more examples of the diphthong Knowledge of when to write ae and when to write e appears
ae spelt with a simple <e> in Pompeii, and reverse spel- to have been a mark of good education in the later empire,
lings, with an etymological ē (or ĕ) represented by the but we cannot be certain whether any speakers aimed to
digraph <ae> (Väänänen 1966:24; see also Coleman preserve the diphthong in speech (Adams 2013:80f.).
1971b:185-90 for a wider collection of such spellings in For the first good evidence for the loss of distinctive
Latin inscriptions, and Adams 2013:75-7 for a survey of vowel quality in Latin, it is necessary to look at later texts,
other recently discovered evidence for this change from and here it is more rewarding to consider the work of
the early empire). grammarians and other commentators on Latin, combined
The conclusion from these spellings would appear to be with the analysis of the metrical practices found in literary
that, in the first century AD, the confusion between ē and i authors, rather than inscriptional or documentary evidence.
was not yet at a stage where it permeated into speakers’ Both Loporcaro (2011a:530-58) and Adams (2013:43-51) treat
writing habits (so also Eska 1987; Adams 2013:58f.), whereas this material in far more detail than is possible here, so
the monophthongization of ae was already under way. I will merely present a couple of pieces of evidence in order
These two developments run counter to the theory of an to exemplify what the sources can reveal. First, some
Oscan substrate in the Latin spoken around the Bay of writers on Latin state that there is a tendency to lengthen
Naples, since Oscan speakers had preserved the diphthong, short vowels when they are in a stressed syllable. For
but operated with a distinction between high-mid and low- example, the fifth-century grammarian Consentius (Publius
mid vowels. Indeed, the vowel system of the Latin spoken at Consentius) identifies as a linguistic vice the practice of
Pompeii seems to be in step with other evidence for the African speakers of Latin of pronouncing the word for
pronunciation of spoken Latin in other areas of the early ‘pepper’, PIPER, with a long vowel in the first syllable, and
Roman empire. Letters and administrative documents from elsewhere he castigates the African practice of pronouncing
Egypt and north Africa, from the Roman forts at Vindo- unaccented long vowels as short, as in the word ŌRĀTOR
landa on Hadrian’s wall (now in northern England), and ‘speaker’, in which the unaccented initial ō was pronounced
from Vindonissa (now in the canton of Aargau in Switzer- short. Since Augustine refers to similar processes of length-
land) also fail to show confusion between accented ē and i ening of vowels under the accent (although without the
(see Adams 2013:51-8 for a survey). In the same documents, same prescriptivism observable in Consentius), it has been
there is good evidence for the monophthongization of ae usual to associate this change specifically with African Latin
(Adams 2013:73-5). Vowels in final syllables of polysyllabic (thus Loporcaro 2011a:54f.). But Adams (2013:47-51) cites
words (which were never under the stress accent in the work of grammarians and authors outside Africa to
speech) do show alternations in writing between e and i, show that other speakers in the empire were already uncer-
but here the merger affects short e and short i, and need tain about vowel length in the third century AD. Of interest
not be related to the changes which will later affect here is Adams’s reference to the work of Holmes (2007) on
Romance vowels. the use of rhythmical endings of periodic sentences
If the distinctive merger of ē and i is not attested in the (so-called clausulae) in the author Vegetius (Publius Flavius
early empire, what can be said about the loss of distinctive Vegetius Renatus, c.400 AD). Vegetius strove to emulate
vowel quantity? As already mentioned above, the Pompeian Classical models by employing a restricted set of clausulae,
spellings sometimes write ae where we expect both etymo- but he seems to have been uncertain about which vowels in
logical ē and ĕ. Pulgram (1978:229) took this as a clear open syllables were traditionally scanned long and which
indication that in spoken Latin, length was no longer dis- were scanned short. The evidence for Vegetius is significant,
tinctive, but that the oppositions between vowels were since it involves the work of an elite Roman (Vegetius was
solely based on quantity. But these criticisms had already in the imperial administration, and addressed his work on
been countered by Coleman (1971b; 1974). In Coleman’s military organization to the emperor), and thus indicates
view, the vowel which resulted from monophthongized ae that the collapse of length distinctions may have been
shared its length with inherited ē, but agreed with ĕ in its widespread among the Latin of all classes by this date.
quality, as a more open vowel than the inherited ē. When Of all the different sources for discovering when the
writers chose to replace an original ē with ae, as in aegisse for vowel system of spoken Latin moved closer to that of the
ēgisse ‘to have led’, they selected a writing which matched in Romance languages, inscriptional evidence can only tell us
length but not quality; but when they wrote ae for ĕ as in so much. In the case of the diphthong ae, inscriptional
Saecundae for the personal name Sĕcundae they chose a evidence is particularly useful, since it allows the precise

8
OUP CORRECTED PROOF – FINAL, 25/5/2016, SPi

LATIN AS A SOURCE

dating of the early appearance and the spread of spellings SCRIBES, was, for some speakers, indistinguishable from the
that reflect a change in pronunciation. For the loss of vowel second person singular of the present, SCRIBIS; the third person
quantity, however, inscriptional spellings are not as helpful singular of the future of the verb ‘to deny’, NEGABIT, overlapped
as explicit statements by grammarians and Latin authors with the third singular of the perfect, NEGAUIT (see Adams
and the careful analysis of metrical features. It is true that it 2013:653f. for references to textual examples using these
is possible to count up cases of metrical inscriptions where exact forms).
vowel quantities differ from the norm in different areas Indications of this uncertainty over the correct morph-
(Loporcaro 2011a:56f. cites a study of Herman 1982 which ology of the future indicative can be seen from attempts to
does exactly this). But here it is not clear whether the replace the analytic future tense with other constructions.
results actually reveal the spread of a phonological phe- Thus Echion, one of the fictional freedmen whose language
nomenon, or the differing expertise and education of verse is represented by Petronius (part of which was cited at §1.2),
composers in different places and different times. In gen- uses the following forms in the course of a long speech:
eral, such quantitative studies of inscriptional material have habituri sumus for HABEBIMUS ‘we shall have’, daturus est for
revealed rather less about the changing nature of spoken DABIT ‘he will have’, and persuadeam in place of PERSUADEBO
Latin than is sometimes claimed (see Adams 2007:629-35 for ‘I shall persuade you’. The first two forms make use of a
a critique of some studies of this sort). periphrasis of the verb ESSE ‘to be’ with the future participle,
and the third employs the subjunctive in place of the future
(note that in the third and fourth conjugations, the first
person singular of the subjunctive is identical with the first
1.4 Development of the future tense person singular of the future, hence REGAM means both ‘I will
rule’ and ‘I may rule’; Echion may have generalized this
In Classical Latin the four main conjugation classes into pattern to verbs of the second conjugation, into which
which all regular verbs fall differ in the formation of the PERSUADEO falls). The periphrasis with the future participle
future tense. In the first and second conjugation, and in and the verb ‘to be’ is also found in other sub-literary
early Latin also in the fourth conjugation, the future is documents. For example Claudius Terentianus, a Roman
formed through the addition of a morpheme incorporating soldier in Egypt whose correspondence is a rich source of
b (hence first person -BO, as in AMABO ‘I will love’ formed from non-standard Latin forms, once writes missurus es in place of
the verb AMARE) to the stem; in the third and fourth conju- MITTAS ‘you will send’ and once daturus est in place of DABIT ‘he
gations, a different formation is used, incorporating a mor- will give’; Terentianus also twice substitutes the future
pheme ē (except in the first person), giving a paradigm REGAM tense with a verb conjugated in the present (Adams
‘I will rule’, REGES ‘you will rule’ from the verb REGERE. The 1977:49). It is worth noting in passing that, although the
Latin future tense was replaced in Romance by a number of use of the present tense with future meaning is paralleled in
different constructions, all of which have their origin in other Latin texts including the comedies of Plautus, it is not
periphrases. Thus a periphrasis of infinitive and HABEO in itself indicative of a socially marked speech variant, but
‘have’ is found in most of western Romance, a periphrasis rather of conversational and colloquial language; and the
of HABEO AD ‘have to’ þ infinitive or DEBEO ‘owe, must’ þ use of the present with future meaning in both Plautus and
infinitive in Sardinian, and UOLO ‘want’ þ infinitive in Roma- Petronius can be explained in terms of pragmatics or com-
nian. The change from the Latin future to the Romance future municative effect (Leiwo 2010:287-91; Adams 2013:666-72).
is one of the perennial topics of the historical development of Despite the apparent avoidance of the Classical Latin
Romance (see Adams 2013:652-4 for a short survey of some future forms, there is no clear indication in either Petronius,
recent literature on the topic, as well as §46.3.2.2). sub-literary documents and letters, or other texts associ-
What light can Latin sources shed on this issue? There ated with a lower register of Latin that the construction of
are some indications from our sources that Latin speakers in infinitive þ HABEO was coming to be used in place of the
the early empire were uncomfortable with the future future. Indeed, the infinitive þ HABEO construction does not
morphology of Classical Latin. Phonological changes, includ- occur with an unambiguous future meaning in any of the
ing the changes in the vowel system discussed in the previ- texts collected in Rohlfs (1969b) earlier than the fifth cen-
ous section and the merger in some positions of the voiced tury AD. The early seventh-century Chronicon Fredegarii (II 62)
bilabial stop (represented in Latin orthography by the letter has often been cited for the first evidence for the contracted
<b>) with the labiovelar approximant (represented by <u>), form of infinitive followed by the verb habeo in a literary
had led to an overlap between some forms of the future text. The specific passage involves a folk etymology for the
with either present or perfect forms. Thus the second name of the town Daras, reproducing a fictional conversa-
person singular of the future of the verb meaning ‘write’, tion between the sixth-century Byzantine emperor

9
OUP CORRECTED PROOF – FINAL, 25/5/2016, SPi

JAMES CLACKSON

Justinian and a Persian king (Herman 2000:74). The Persian same order as in the later Romance futures, Sacerdos pre-
king says to Justinian non dabo ‘I will not give [up the town]’ sents perhaps the earliest incontestable example of the
employing the Classical Latin synthetic future DABO, but construction with a future meaning in Latin; but there are
Justinian replies daras ‘you will give [it up]’ i.e. the result parallels from the third century onwards, in other educated
of a contraction of DARE HABES ‘give.INF have.2SG.PRS.SBJV’. The and generally high-status authors. In the words of Adams
contracted form of the future occurs only once, in this (2013:657f.) ‘the periphrasis seems to have had some cur-
instance, in all literary Latin of the period, where it is rency among the educated as an exponent of deontic and
necessary for the sake of the pun—and, incidentally, alethic modality when it was convenient to convey an
where it occurs in a citation of supposed direct speech. ambiguity, with an idea of futurity also present’.
The use of the contracted form here suggests however The infinitive þ HABEO construction is therefore instruct-
that it was recognized and employed by speakers of all ive for the researcher into the sources of Romance within
classes by this time; it is appropriate in the mouth of an Latin. Although the texts gathered under the label of ‘vulgar
emperor. Latin’, particularly in the early empire, may reveal aspects
Although the infinitive þ HABEO construction does not of the breakdown of the Classical Latin system of expressing
appear in texts normally considered to be ‘vulgar Latin’ futurity, they give no indication of the construction which
with future meaning, it is not entirely absent from all would eventually win out in Romance. Indeed, the best
Latin texts before the fifth century AD. Indeed, it is found source for this construction is not found in sub-literary
as early as Cicero in Classical Latin, but here with a different documents or texts written by or intended for those with
meaning (and in both orders infinitive þ HABEO and HABEO þ little education, but is found in the works of grammarians
infinitive). In Cicero, a phrase such as HABEO DICERE ‘I.have say. and others who are usually thought of as linguistic purists.
INF’ (more common in Cicero with the infinitive placed after If the construction did exist in the speech of the uneducated
the modal than in the order DICERE HABEO) means ‘I can say’ and illiterate, it was evidently not stigmatized by grammar-
(Adams 1991:155f.). In later authors the construction can ians, but was also present in their written and presumably
also mean ‘I have to say’ or ‘I must say’. Indeed, Adams in their spoken language too (see Adams 1991:135 on Pom-
(1991) detected a pattern in the order of the constituents in peius’ composition technique and style, and for the conclu-
the fifth/sixth-century AD grammarian Pompeius, who has sion that ‘his treatise is bound to preserve characteristics of
38 examples of the infinitive þ HABEO construction in his the spoken language’).
commentary on the Ars of the earlier grammarian Aelius
Donatus. Pompeius uses the order HABEO þ infinitive when
the context supports an interpretation of either possibility
or obligation, but when he has the order infinitive þ HABEO, 1.5 Lexicon
the context usually excludes a reading of possibility, and the
construction must be understood to mean ‘have to’ or The final area of Latin and its continuation into Romance
‘must’. The link between obligation and futurity is found that I shall consider is the lexicon. In many respects the
in many languages (the Sardinian future is an obvious par- lexicon is a straightforward area of language to track his-
allel for a former modal verb expressing obligation becom- torically, since it is possible to isolate individual words and
ing grammaticalized as a future exponent; cf. §17.4.2). States detect their first (or last) appearance in the written record,
of affairs which are obligatory must come into effect at a and digital corpora make it possible to search quickly
future point in time, and hence there is a natural slide from through millions of words of text for lexical items. Contrast
verbs which express obligation to markers of futurity. It is with this the domain of syntax, where much more labour is
not surprising that some of the examples from Pompeius required in order to search for and detect changing patterns
can also be rendered into English by a future tense. over time; the researcher also needs to take into account a
Indeed, it appears that in other grammarians there are host of other factors, for example, the disruption of natural
much earlier examples with the HABEO þ infinitive construc- word orders through the metrical constraints imposed by
tion (in this order) having a meaning which can only be verse or by clausulae in prose (discussed in §1.3). However, it
translated by a future. Pinkster (1987:207) cites a passage of is no surprise that the maxim ‘every word has its own
the third-century grammarian Sacerdos (Marius Plotius Sa- history’ (or ‘chaque mot a son histoire’) is usually attributed
cerdos) discussing the difference between the past, present, to the Romance linguist Jules Gilliéron (1854-1926) (although
and future. There are some people, Sacerdos says, who deny it does not appear in his published work; see Christmann
that the present tense exists because everything has either 1971). Lexical developments from Latin to Romance are
been done or will be done: aut factus esse aut habere fieri (Keil extremely varied, with many idiosyncratic and unparalleled
1855-80, VI:432.13). Although the constituents are not in the developments of individual words, alongside wider, more

10
OUP CORRECTED PROOF – FINAL, 25/5/2016, SPi

LATIN AS A SOURCE

general patterns of semantic change and replacement (see even when discussing punishments to slaves. In the first
Stefenelli 2011 and Dworkin 2011 for recent surveys of century AD, COLAPHUS reappears in literary texts, but there it is
lexical stability and lexical change from Latin to Romance, generally associated with the language of the underclass. In
as well as Ch. 32). the works of Quintilian, it only appears in a piece of
In the investigation of the origins of the Romance reported speech, deemed ‘unworthy of a gentleman’, and
vocabulary, it is important to consider all the evidence in the narrative section of Petronius it occurs with the
from Latin sources, and not focus in on certain registers or original field of reference when describing the beating
text types; prior assumptions about the nature of vulgar administered to a slave. It is also evident that the word
Latin (as discussed in §1.2) can be misleading. Indeed, the had also by this time come to have a wider sense in the
replacement of lexical items is also still frequently seen in spoken language, since Petronius has one of the freedmen
terms of the vulgar Latin of the lower classes winning out utter a derived form, the verb PERCOLOPABANT, which is on its
over Classical Latin forms used by the elite (see e.g. way to becoming the general word for ‘hit’ (Adams
Stefenelli 2011:583 for a recent expression of this view), or 2007:439). The second example is a word for ‘mouth’ (on
of colourless terms being replaced by those which were which see now also Adams 2013:782; cf. also §32.3.4). In
more ‘expressive’ (so Stefenelli 2011:572f.). These views Classical Latin BUCCA is restricted to the meaning ‘cheek’,
can be bolstered by lists of lexical winners and losers: TESTA but it replaces the Classical Latin term OS ‘mouth’ in western
‘pot’, generally replaces CAPUT with the meaning ‘head (Fr. Romance (It. bocca, Sp./Pt. boca, Fr. bouche, etc.). The mean-
tête, It. testa, but in Naples and surrounding dialects testa still ing ‘mouth’ is however attested in the Latin sources, once in
means ‘earthenware pot, vase’, and the word for ‘head’ is the speech of a freedman in Petronius, and the biographer
still capo); PORTARE ‘carry a burden’ and LEVARE ‘lift’ oust FERRE in Suetonius (Gaius Suetonius Tranquillus, c. AD 69–22) reports
the meaning ‘carry’ (It. portare, Fr. porter, Sp./Pt. portar, Sp. that the emperor Augustus used the term in his letters.
llevar, and Pt. levar); CABALLUS ‘workhorse’ becomes the every- Furthermore, something close to the meaning ‘mouth’
day word for ‘horse’ as opposed to EQUUS (It. cavallo, Fr. cheval, appears to lie behind the phrase IN BUCCAM UENIT ‘come to
Sp. caballo, Pt. cavalo, Ro. cal; note that Lat. EQUA ‘mare’ mind’ which occurs even in the correspondence of Cicero.
survives in some languages, e.g. Sp. yegua, Ro. iapă; cf. These examples show that, although some words may
discussion in §33.3.3); CASA ‘hut’ wins out against DOMUS or have gained an association with lower registers, and seemed
AEDES in the meaning ‘house’ (Ro. casă, It./Sp./Pt. casa) and to have been either unconsciously avoided in high-register
many others (see the useful material gathered in Stefenelli compositions or to have been actively stigmatized and ridi-
2011). However, in some of these cases alternative explan- culed (as the statements of Quintilian and Suetonius sug-
ations are possible. Already 60 years ago, Benveniste gest), they were still used by some members of the educated
(1954:256) argued that in late Latin TESTA could be used in classes in writing as well as speech, and could be deployed
medical texts as a term for the skull (compare English for effect even in a written literary work. The spoken and
brainpan), and that the medical sense may be the origin of written language were part of a single continuum, and there
the Romance term, although of course the semantic shift was no gap between the two. It is important to note also
from a meaning ‘container’ to ‘head’ can be widely paral- that the patterns of word avoidance in Classical Latin are
leled (Dworkin 2011:590f.). not always shared across all literary genres. Thus for
It is certainly the case that one feature of Classical Latin example, as Housman (1930) noticed, Latin poets do not
was the avoidance of words which could be associated with generally use ASINUS to refer an ‘ass’ or ‘donkey’, but instead
certain registers, and that some lexical items that occur in prefer the diminutive ASELLUS (or a Greek term); and ASINUS is
early Latin are systematically avoided by writers of the indeed the term which wins out in Romance (It. asino,
Classical period. In Clackson (2011b: 52f.) I discussed two Fr. âne, Sp./Pt. asno, etc.), with ASELLUS surviving in Italian
examples. First, the Latin word meaning ‘a punch’ or ‘a with the sense of ‘woodlouse’, asello. But the avoidance of
thump’: COLAPHUS (also spelt COLAPUS), which was to give the ASINUS only takes place in Latin poetry. In historical prose,
standard word for a ‘blow’ or ‘hit’ in Romance languages ASINUS is attested, even in the stylist Tacitus (Publius Cornelius
(e.g. Fr. coup, It. colpo). The Latin word originates as a loan Tacitus, AD 56-117), whose delicacy is such that he notoriously
from Greek, and came into the language probably through prefers to use a periphrasis rather than call a spade a PALA
the speech of slave traders from the south of Italy, where ‘spade’, instead referring (at Annals 1. 65) to ‘that by which
Greek was the lingua franca (Adams 2003a:351 n.100). Plau- earth is tilled or turf is cut’ (as noticed by Palmer 1954:142).
tus and Terence and other comedians use the word freely, The case of ASELLUS and ASINUS is further noteworthy, since
especially in the description of beatings given to slaves. ASELLUS is in origin a diminutive of ASINUS, and therefore offers
However, Cicero and other writers in the first century BC a counterexample to a widespread pattern whereby a
employ other Latin terms, ICTUS or PLAGA, to refer to blows, diminutive replaces its base noun, which has led scholars

11
OUP CORRECTED PROOF – FINAL, 25/5/2016, SPi

JAMES CLACKSON

to think that ‘[d]iminutives must have been not only widely documents such as those from Vindolanda, which included
used but thought of as a colloquial feature’ (Herman lists of everyday items, no doubt more such terms in use in
2000:99). It is true that in high registers of Latin diminutives the spoken language would be found.
are frequently avoided, and they sometimes surface in texts
that are often thought of as more low-level or colloquial.
However, not all diminutives in Latin, even those which are
found only in low-level texts, are destined to survive into 1.6 Sources of the Romance languages
Romance. Consider, for example, the class of diminutive
adjectives that occur in Cicero’s letters. The adjective BELLUS Latin speakers of all classes during the empire, and probably
‘fine, pretty’, in origin a diminutive form of BONUS ‘good’, is during the last centuries of the republic, did not write
attested 24 times in Cicero’s correspondence, but only 13 exactly as they spoke. Even Cicero, the model for later
times in his more voluminous prose speeches and philo- language purists, states that he would rather say posmeri-
sophical works, and it does not occur in high-register poetry dianus ‘in the afternoon’ with omission of the medial t in the
or the Latin historians. Both BELLUS and BONUS survive into written Classical form POSTMERIDIANUS (Orator 157, cited by
Romance (It. bello, Fr. beau, etc.; Ro. bun, It. buono, Fr. bon, Adams 2013:153). Anyone writing down Latin was in some
etc.), and generally with the same meanings as in Latin. sense actually ‘writing up’ (to borrow the term of Langslow
Cicero also uses other diminutive adjectives in his letters: 2000:412)—that is to say, using literary or educated models
INTEGELLUS ‘unharmed’ (beside CLat. INTEGER ‘whole’), LONGULUS for the spellings, syntactic constructions, and vocabulary.
‘long’ (beside LONGUS ‘long’), and even BARBATULUS ‘a bit Although inscriptions and graffiti may reveal clues about
bearded’ (beside BARBATUS ‘bearded’) (see Leumann et al. divergent phonological developments, learning how to
1977:308 for these formations). However, unlike the adjec- write was also a process of learning how to spell, and it is
tive BELLUS, none of these forms is continued into Romance. important not to confuse the author of a ‘misspelt’ Latin
Furthermore, some of the lexical items that occur in appar- text with an ancient amateur phonetician, attempting to
ently more colloquial or everyday registers, such as letters, record precisely the sounds of speech. Some sub-literary
but not in more literary works need not necessarily reflect texts abound in deviations from the classical norm, but
the spoken language that would win out in the end. Here these were also the result of a meeting between everyday
again the equation between ‘colloquial’ or ‘vulgar’ and the speech and a literary education. Even the female author of
material that is to survive into the Romance languages is far the Itinerarium Egeriae (known in older scholarship as the
from adequate. Perigrinatio Aetheriae), a chatty account of a trip to the Holy
Despite the caveats of the preceding paragraphs, many Land written in the fourth century AD (Väänänen 1987:8),
words which were to become winners in Romance scarcely long prized as a treasure trove of ‘vulgar’ forms’, wrote up
appear at all in the record of written Latin, whether in her Latin in the hope of emulating the norms of her day
literary, sub-literary, or inscriptional genres. Some of the (for which the model was probably the Latin Bible trans-
finds of ‘new Latin’ in the last 30 years, such as the docu- lation rather than Cicero). Thus Egeria employs parti-
ments written on wood from the Roman fort at Vindolanda, ciples, the ablative (and in some cases accusative or
or papyrus or ostraca from north Africa and the Near East, nominative) absolute construction, and subordinating
have revealed how patchy our record of the Latin which was or complement clauses which she has little control over,
in day-to-day use actually is. Adams (2013:777) cites three but which reveal her desire to make her account more
words now attested from texts from Roman Britain around ‘literary’ (see Clackson and Horrocks 2007:286-92 for fuller
100 AD, baro ‘man’, uectura ‘wagon’, and cimusa ‘cloak’, which discussion).
are scarcely attested in any other source for Latin with Given that all surviving Latin is mediated through a
these meanings before the fourth century AD, and even written form, and that the origins of the Romance languages
thereafter largely avoided in Latin texts of all types. All lie in the (irrecoverable) spoken language (with, of course,
three of these words have reflexes in Romance, and it is occasional feedback from educated norms or ‘corrections’
likely that they continued in use throughout the Roman incorporated into speech), there will always be a mismatch
empire, but were largely avoided in all genres with any between the Latin sources and the parent of the Romance
pretence of literary merit. Since the Roman educational languages. However, there is a great deal of surviving Latin,
system taught literacy largely through the inculcation of written over a long timescale and across a wide geograph-
Classical literary texts and models (see Clackson ical spread, in very disparate genres and registers. In this
2011b:241f.), anyone who had learned how to write became chapter I have shown some case studies of what can be
accustomed to avoiding terms which were not found in learned from careful consideration of the totality of the
literary works. If there were a greater surviving portion of written evidence, without preconceived notions about

12
OUP CORRECTED PROOF – FINAL, 25/5/2016, SPi

LATIN AS A SOURCE

which are more likely sources for the Romance languages. texts, both those that are traditionally classed as ‘Vulgar’
Inscriptional evidence, such as the surviving graffiti and and those from higher or more technical registers, are
dipinti from Pompeii, may be helpful in establishing chron- essential for the understanding of developments in morph-
ologies of sound change, but should be combined as far as ology, syntax, and lexicon. But in some cases, such as the
possible with other sources. The non-Latin native languages infinitive þ HABEO construction, or in the history of individ-
of Italy do not seem to have had a major impact on the ual lexical items, the sources may not (yet) reveal what was
spoken ancestor of Romance. Literary and sub-literary actually taking place in ancient speech.

13

You might also like