Professional Documents
Culture Documents
inflectional future (2a) with an auxiliary + infinitive construction (2b; cf. discussion
in }4.3.1.5); and (iii) the accusative and infinitive (AcI) construction, in which the
sole marker of subordination lies in the ‘exceptional’ accusative marking of the
infinitival subject (3a),5 with a (non-)finite subordinate clause introduced by an
overt complementizer (3b):
1
Meillet ([1937] 1964: 439), Bourciez (1956: 23), Harris (1978: 15–16), Tekavčić (1980: 15), Schwegler
(1990), Posner (1996: 156–7).
2
For an historical overview of the use of the terms ‘synthetic’ and ‘analytic’, see Schwegler (1990: ch. 1).
3
See, for example, Schwegler (1990), Vincent (1997a: 102–3), Bauer (2006).
4
Wüest (1998: 92), Zamboni (2000: 121–2); cf. also }5.3.1.2.
5
The ‘exceptional’ accusative case of the infinitival subject is determined by the construction as a whole,
as highlighted by the fact that CREDO ‘I believe’ canonically assigns dative to its complement (see }3.4.1,
}4.4.1.1, }5.4.2.1).
Syntheticity and analyticity 11
Indeed, so deep-rooted is this transformation that Tekavčić (1980: 15) considers it ‘the
guiding principle underlying the whole of Romance morphosyntax and undoubtedly
its deepest and most important characteristic’.
12 From Latin to Romance
2.2 Problems
This traditional interpretation of the synthesis-analysis cycle proves, however, problem-
atic on a number of accounts, as does the fundamental typological distinction on
which it crucially rests.6 This was first noted by Humboldt (1836), who considered
contrasts between Latin and Romance such as those exemplified in (1)–(5) above as
purely superficial, external differences: though admittedly not as rich as Latin in
their inflectional specifications, at a deeper, internal level the Romance languages are
of the same inflectional type as the ancestor language, and hence ultimately constructed
according to the same structural principles, allowing Humboldt to conclude (pp. 288–9)
that, although individual inflectional forms were lost, inflection as a whole was not lost
(‘Es sanken Formen, nicht aber die Form’). While this idea might be difficult to maintain
in its strongest form in view of the undeniable typological differences between Latin and
Romance to be discussed in the following chapters, it does highlight how the passage
from Latin to Romance cannot be characterised tout court in terms of a shift from
synthetic to analytic structures as long as Romance continues to display extensive
syntheticity involving nominal and verbal inflections. Moreover, as Coseriu (1987: 58)
observes, there is little justification for lumping all presumed Romance analytic devel-
opments into a single all-encompassing principle, for they represent a collection of
extremely heterogeneous changes, the diatopic distribution, individual chronologies and
details of which vary enormously from one construction to another.7
6
Schwegler (1990: 4–5), Bauer (1995: 10–11, 138, 166), Vincent (1997a: 99–100, 105).
7
See, for example, the comparative discussion of HABERE ‘have’ in the formation of the perfect and future
periphrases in Pinkster (1987) and Bauer (2006), and the discussion in }}4.3.1.4–5, }7.3.1.1, }7.4.1.1 below.
8
Coseriu ([1971] 1988: 210; 1987: 56), Schwegler (1990: 28), Vincent (1997a: 99).
9
Highly exceptional are cases such as NOLO ‘I do not want’, NESCIO ‘I do not know’, NEQUEO ‘I am not
able’, and NEGO ‘I say no, deny,’ (< NE(C) ‘not’ + UOLO / SCIO / QUEO / AIO ‘I.want / I.know / I.am.able / I.say’),
which belong to a small number of lexicalized cases in which the negator NE(C) has univerbized with the
category it modifies (Ernout and Thomas [1953] 1993: 149–50).
Syntheticity and analyticity 13
passive periphrasis consisting of ESSE ‘be’ + PPtP (6b), and a number of overt markers
of subordination, including the subjunctive complementizers UT/NE ‘that (/not)’ (6c):
(6) a quem amat, amat; quem non amat, non amat (Lat., Petr. Sat. 37.5)
whom she.loves she.loves whom not she.loves not she.loves
‘Whom she likes, she likes; whom she does not like, she does not like’
b arma [ . . . ] sunt humi inuenta (Lat., Cic. Diu. 1.74)
weapons.N.PL are ground.LOC found.PASS.PFV.PCTP.N.PL
‘the weapons . . . have been found on the ground’
c Pompeius suis praedixerat ut Caesaris impetum
Pompey.NOM his.DAT.PL had.foretold that Caesar.GEN charge.ACC
exciperent ne ue se loco mouerent (Lat., Caes. B.C. 3.92.1)
they.received.SBJV that.not=or selves place.ABL they.moved.SBJV
‘Pompey had told his men beforehand to receive Caesar’s charge and not
to move from their position’
In fact, Wüest (1998: 91) goes so far as to claim that early and classical Latin, despite
systematically favouring synthetic forms, present, for instance, just as rich an array of
future expressions as modern French. Thus alongside the synthetic paradigms in -BO
(e.g. UOCA-BO ‘call-FUT.1SG’) and -AM (e.g. SCI-AM ‘know-FUT.1SG’), we also find the
grammaticalized periphrasis built on ESSE ‘be’ and the future active participles (e.g.
UOCATURUS ‘call.FUT.ACT.PTCP’ + SUM ‘I.am’: ‘I am about to call’), as well as various
modal + infinitive constructions,10 not to mention the widely attested use of the
present with future value (Bennett 1910: 18–22).
Conversely, in Romance number and gender marking on nouns and adjectives is
still typically suffixal (7a),11 as are person/number (7b) and temporal, aspectual, and
modal categories (7c) on finite verbs:
(7) a o(s)/a(s) nos-o(s)/-a(s) veciñ-o(s)/-a(s) ruidos-o(s)/-a(s) (Glc.)
the.M(PL)/the.F(PL) our-M(PL)/-F(PL) neighbour.M(PL)/F(PL) noisy- M(PL)/-F(PL)
‘our noisy neighbour(s)’
10
The modals with presumed future uses include INCIPERE ‘to begin’, POSSE ‘can’, UELLE ‘to want’, and
DEBERE ‘must’ (cf. Hofmann and Szantyr 1965: }175; Fleischman 1982; Haverling 2010: 395–9), although
Pinkster (1989) finds no evidence for this claim.
11
In many Occitan varieties the original synthetic marking of plural number has optionally been
reinforced with the creation of doubly marked inflectional plurals (Arnaud and Morin 1920: 273; Wheeler
1988b: 256) such as pial ‘hair’ > pial-s ‘hair-PL’ (> pials-es ‘hair.PL-PL’), amì ‘friend’ > amì-s ‘friend-PL’
(> amìss-es ‘friend.PL-PL’), molin ‘mill’ > molin-s ‘mill-PL’ (> moliss-es ‘mill.PL-PL’). Moreover, suffixal
marking of feminine gender (viz. -a, predominantly realized as [ɔ], more rarely [a] (e.g. Auv, Niç., Mtp.,
Ros.) or [e] (e.g. Bea., EGsc.)) is so robust in Occitania that it is productively extended not just to
professional titles (e.g. coconièr ‘egg.seller’ > coconièr-a ‘egg.seller-FSG’ (‘egg-seller’s wife; female egg-seller’),
tabaton ‘tobacconist’ > tabaton-a ‘tobacconist-FSG’ (‘tobacconist’s wife; female tobacconist’)), but also to all
masculine proper names (Arnaud and Morin 1920: 272) such as Paul-on ‘Paul-DIM’ > (la) Paulon-a ‘(the)
Paul.DIM-FSG (= Paul’s wife)’, Guston ‘Auguste’ > (la) Guston-a ‘(the) Auguste-FSG (= Auguste’s wife)’.
14 From Latin to Romance
b partiv-i/-as/-a/-am/-atz/-an (Gsc.)
leave.-IND.PST.IPFV.1sg/2SG/3SG/1PL/2PL/3PL
‘I/you etc. was/were leaving’
c parr-u/-ava/-ai/-assi (Sic.)
speak-1SG.PRS.IND/1SG.PST.IPFV.IND/1SG.PST.PFV.IND/1SG.PST.SBJV (/COND)
‘I speak / was speaking / spoke / spoke (/would speak)’
To these we can add the wide-spread use of evaluative suffixes (e.g. It. mur-etto ‘wall-
DIM’, Lgd. pintair-às ‘drinker-PEJ (= drunkard)’, Pt. mulher-ão ‘woman-AUG’, Ro.
frumuş-el ‘beautiful-DIM’, Ven. gat-on ‘cat-AUG’),12 superlative and residual compara-
tive forms (e.g. Cat. alt-íssim ‘high-SUPERL.MSG’, fred-íssim ‘cold-SUPERL.MSG’, mill-or
‘good-COMPR.SG’, pitj-or ‘bad-COMPR.SG’, It. sicur-issimo ‘sure-SUPERL.MSG’, noios-issimo
‘boring-SUPERL.MSG’, super-iore ‘high-COMPR.SG’, infer-iore ‘low-COMPR.SG’, Pt. felic-
íssimo ‘happy-SUPERL’, mai-or ‘big-COMPR.SG’, men-or ‘small-COMPR.SG’), and the fre-
quent retention of case distinctions in pronouns (e.g. MLuc. iə ‘1SG.NOM’, (a) mì ‘(P.
ACC-DAT.MRK) 1SG.ACC-DAT’, (pə) mévə ‘(for) 1sg.OBL’, (cu) mìchə ‘(with) 1SG.COM’; Ro.
eu ‘1SG.NOM’, (pe) mine ‘(P.ACC.MRK) 1SG.ACC’, mie ‘1SG.DAT’, meu ‘1SG.GEN’). If we
consider furthermore that many of the Romance languages also display a whole
host of innovative synthetic structures (Coseriu 1987: 59), notably the future and
conditional, derived from an erstwhile infinitival periphrasis in conjunction with
HABERE ‘have’ (8a),
13
sentence and manner adverbs in -men(t(e)) (8b),14 internal
morphophonological alternations such as metaphony and related types of stressed
vowel alternation to mark number and/or gender in nominal categories (8c–e),15 and
inflected non-finite verb forms such as the old Neapolitan inflected infinitive (8f),16
the presumed inexorable driving force underlying Romance analyticity appears
seriously undermined.
(8) a MONSTRARE + HABEO / HABEBAM > Fr. (je) montrerai / montrerais
show.INF I.have / I.had (I) will.show.1SG / would.show.1SG
b APERITA + MENTE > It. aperta-mente
open.ABL mind.ABL open-ly
‘with (an) open mind’ > ‘openly’
12
In modern French these are generally no longer productive (e.g. fill-ette ‘girl-DIM’, but not
**chaussur(e)-ette ‘shoe-DIM’, which has to be expressed by the analytic petite chaussure ‘little shoe’).
13
See the discussion in }4.3.1.5, as well as the treatments in Valesio (1968; 1969), Coleman (1971), Harris
(1978: ch. 6), Fleischman (1982), Green (1987), Pinkster (1987), Adams (1991), Maiden (1996), Loporcaro
(1999), Nocentini (2001), Bourova (2005), La Fauci (2006; 2011b), Bourova and Tasmowski (2007).
14
See the discussion in }2.2.2 and }5.3.1.5, as well as Karlsson (1981), Bauer (2001b; 2003; 2006), Ricca
(2010: 181–5).
15
Hilty (1991), Maiden (1991), Savoia and Maiden (1997), Chitoran (2002), Loporcaro (2011a: }1.2).
16
See the discussion in }6.3.2, as well as Maurer (1968), Loporcaro (1986), Raposo (1987), Jones (1992;
1993: 78–82), Vincent (1996; 1998b), Ledgeway (1998: 41–6; 2000: 109–14; 2007c; 2009a: 585–90), Mensching
(2000), Pires (2002), Sitaridou (2002).
Syntheticity and analyticity 15
In short, considerations such as these reveal that, if they are to be employed at all, the
terms ‘synthetic’ and ‘analytic’ should not be predicated of languages but, at best, of
individual constructions. Thus, to the extent that any generalizations can usefully be
reached in relation to the synthetic and analytic typology, they must be made
in relation to specific construction types, rather than individual languages. Exemplary
in this respect is the often overlooked observation (though see Meyer-Lübke 1894: 2–3;
Coseriu 1987: 58; La Fauci 1998: 524–5) that the fate of Latin inflection shows a
differential treatment in the nominal and verbal domains: whereas all but a few isolated
residues of the rich Latin nominal declension—early Gallo-Romance, Ræto-Romance,
and Romanian aside (cf. }7.3.2.2)—have survived into Romance, the Latin verbal
conjugation formally continues largely intact, even accommodating, in some cases,
several new additions such as the future and the conditional (}4.3.1.5; see Maiden 2011a,
b). This observation apparently explains why analytic developments occurring within
the nominal domain typically assume a substitutive function replacing earlier syn-
thetic structures (e.g. PAUL-O ‘Paul-DAT’ > Sp. a Pablo ‘to Paul’, NOU-IOR ‘new-COMPR.’ >
Ro. mai nou ‘more new (= newer)’, INDE ‘thence’ > It. di là ‘from there’), unlike in the
verbal domain where, rather than replacing earlier synthetic structures, they usually
come to mark new functions that complement those already marked synthetically
within the system.17 For example, Romance reflexes of FACTUM HABEO (‘do.PFV.PASS.PTCP
I.have’) do not replace, at least not initially, the synthetic perfect FECI ‘I did, I have
done’, but variously come to assume novel ancillary functions of the perfect (Harris
1982; Squartini and Bertinetto 2000), ranging from a present resultative (9a; Alfonzetti
1998) to a present perfect with marked iterative function (9b; Leal Cruz 2003: 132),
before ultimately encroaching upon (some of) the functions proper of the original
perfect, witness its canonical present perfective reading in standard peninsular Spanish
(9c; Penny 2000: 159) and its full-fledged punctual perfective value in modern spoken
French (9d; Harris 1970: 79–83).
17
Replacement of the synthetic imperfective passive (e.g. AMAT-UR ‘love.PRS.IND.3SG-PASS’) with the
analytic (AMATUS EST ‘love.PFV.PASS.PTCP he.is’; cf. Fr. il est aimé ‘he is loved’) is not a counterexample to
this generalization, since it represents the functional extension of an already existing periphrasis of the
classical language.
16 From Latin to Romance
18
See Kayne (1975), Zubizarreta (1985), Burzio (1986), Alsina (1992), Guasti (1993).
Syntheticity and analyticity 17
The greater autonomy of such negative structures is, however, only apparent. For
example, limiting ourselves just to French, we must note that with finite verbs pas is
19
See, among others, Price (1962; 1986), Posner (1984; 1985a, b), Schwegler (1990), Zanuttini (1997: ch. 3),
Parry (1997), Veny (1982: 51); see also the discussion in }5.2.1.
18 From Latin to Romance
always immediately postverbal and can never, for example, be separated from the
finite verb by a nominal (12a), participial (12b), or infinitival (12c) complement.
Indeed, the only other elements that may precede pas are certain other adverbs,
themselves occurring in a fixed order (cf. Cinque 1999: ch. 1) such as peut-être
‘perhaps’ (12d).
(12) a il (ne) mange pas la viande (**pas) (Fr.)
he NEG eats pas the meat pas
‘he doesn’t eat meat’
b il (n’) a pas mangé (**pas) la viande (Fr.)
he NEG has pas eaten pas the meat
‘he hasn’t eaten the meat’
c il (ne) veut pas manger (**pas) la viande (Fr.)
he NEG wants pas eat.INF pas the meat
‘he doesn’t want to eat (the) meat’
d il (ne) mange peut-être pas la viande (Fr.)
he NEG eats perhaps pas the meat
‘he perhaps doesn’t eat meat’
This lack of autonomy manifested in the rigidly fixed positioning of pas is hardly
what would be expected a priori of a supposedly analytic structure; on the contrary,
pas and similar postverbal negators display a high degree of bonding with respect to
their associated verb.
Analogous issues arise from a consideration of the Romance sentential and
manner adverbs in -men(t(e)) and the Romance compound perfect discussed in
(8b) and (9b–d) above, respectively. Beginning with the former, we note, on the
one hand, how the evidence of coordination from modern varieties such as French,
Italian, and Occitan points to the synthesized suffixal nature of the original ablative
nominal MENTE ‘with (a) mind’ (e.g. ModFr. humblement et doucement ‘humbly and
sweetly’, ModIt. villanamente e aspramente ‘cruelly and bitterly’, Lgd. doçament
e simplament ‘quietly and simply’), whereas the same coordination facts in modern
Ibero-Romance, on the other hand, reveal how -ment(e) retains a greater degree of
analyticity (e.g. Cat. dolçament i suau ‘sweetly and soft(ly)’, Pt. intensa e constante-
mente ‘intense(ly) and constantly’, Sp. impensada pero providencialmente ‘unex-
pected(ly) but providentially’), parallel to old French and old Tuscan (OFr./OTsc.
humeles e dulcement / umile e dolcemente (CDR 1163 / Nov. 4) ‘humbly (lit. ‘humble’)
and gently’). Even greater autonomy can be seen in old Occitan where not only is the
reflex of MENTE (variously -men(s), -menz, -ment) usually attached to just one of the
adjectives in coordinated structures (usually the first, e.g. cruelmen et amara ‘cruelly
and bitter(ly)’, but also suau e bellament ‘soft(ly) and beautifully’; Diez 1874: 429), but
Syntheticity and analyticity 19
the two constituent parts of the adverbial structure can be also separated by inter-
vening material (e.g. epsa . . . ment ‘identical . . . ly’; Grandgent 1909: }105).
Turning now to the so-called analytic perfect, this is characterized by converse
patterns of syntheticity and analyticity: whereas in modern French, Italian, and
Occitan varieties the component parts continue to display considerable autonomy,
witness the interpolation of such elements as adverbs of varying sizes (13a–c;
Monachesi 2005: 134), in modern Ibero-Romance auxiliary and participle, despite
continuing to be written as separate words, never allow interpolation of any sort
(13d–f):
(13) a il a énormément mangé (Fr.)
he has enormously eaten
‘he’s eaten a great deal’
b avevo continuamente sbagliato (It.)
I.had continuously erred
‘I had continuously made mistakes’
c avèm pas encara enviadas de cartas postalas (Lgd.)
we.have not yet sent of cards postal
‘we haven’t yet sent any postcards’
d jà han (**jà) contestat (Cat.)
already they.have already replied
‘they’ve already replied’
e não tinha (**nunca) tocado nunca o saxofone (Pt.)
not he.had never played never the saxophone
‘he had never played the saxophone’
f siempre habías (**siempre) mentido (Sp.)
always you.had always lied
‘you had always told lies’
Examples like these, coupled with the thorny problem of how one is to correctly
measure the autonomy of linguistic units (Schwegler 1990: ch. 2), frequently obscured
by conventional, yet non-systematic, orthographic representations of the ‘word’
(14a–e),20 lead Schwegler (1990: 193) to conclude, somewhat unsatisfactorily, that
Cf. also Fr. est-ce que . . . ? [ɛs(ə)kə] lit. ‘is-this that’ (‘is it that . . . ?’), an erstwhile morphosyntactically
20
complex interrogative cleft which, synchronically, has been reduced in normal speech to a morphosyn-
tactically simplex polar interrogative particle [(ɛ)sk]:
(i) est-ce qu’ il fume? [(ɛ)ski(l)fym]
Q he smokes
‘Does he smoke?’
20 From Latin to Romance
the labels synthetic and analytic can, at best, be understood as nothing more than ‘the
rough measure of the morphemic interdependence of speech units’ [italics A.L.].21
(14) a ¿por qué? (Sp.) / per què? (Cat.) vs porque (Sp.) / perquè (Cat.)22
for what for what for.what for.what
‘why?’ vs ‘because’
b da capo / daccapo (It.)
from head / from.head
‘from the beginning’
c bien que (Fr./Sp.) vs benché (It.)
well that well.that
‘although’
d tal vez (Sp.) vs peut-être (Fr.) vs chissà (It.) / benlèu (Gsc.) / talvez (Pt.)
such time can-be.INF who.knows well.soon such.time
‘perhaps’
e não trabalhando (mais) (Pt.) vs ne(mai)lucrând (Ro.)
not work.GER (anymore) not.(anymore).work.GER
‘not working (anymore)’
Yet even adopting a relativized interpretation of the traditional usage still fails to
make any intuitive sense of many developments. For example, in the wake of
Schwegler (1990: xv), Vincent (1997a: 99–100) proposes a scalar definition of the
synthetic-analytic distinction in terms of the degree of phonological and morpho-
syntactic autonomy borne by the constituent grammatical properties of a given
construction. On this view, however, one of the most important consequences of
the presumed synthetic to analytic drift, manifested in the gradual replacement of an
original ‘free’ word order with a ‘fixed’ (S)VO order (cf. }3.3.2, }5.4.2.4), must now,
despite the obvious contradiction, be treated as a synthetic development. In particu-
lar, the remarkable syntactic autonomy and independence of the core constituents of
the Latin sentence which could, in accordance with pragmatic principles, not only
occur in all possible permutations (see }3.2.1, }3.3.1), but whose internal structure,
when complex, could, in certain cases and in specific registers, be scattered discon-
tinuously across the sentence (see }3.2.1.1, }3.3.1.1, }5.4.3), must be taken as an indica-
tion of greater analyticity. By the same token, the greatly reduced positional
autonomy, coupled with the increased semantico-syntactic interdependence, of the
21
Indeed, such principles lead Bauer (2006: 287–8) to consider the Romanian definite article, an enclitic
conventionally written as a single word together with its associated noun or adjective (e.g. Ro. câine=le
‘dog=the’), a synthetic development, whereas its proclitic counterpart in the other Romance varieties,
where it happens to be conventionally written as a separate word, is described as an ‘analytic article’ (e.g. Fr.
le chien ‘the= dog’).
22
Cf. also Fr. parce que ‘for.this which’ vs Occ. per ço que ‘for this which’ (both ‘because’).
Syntheticity and analyticity 21
core constituents of the Romance sentence which can now only be interpreted
relative to each other, and whose constituent parts are cohesively bound together,
are to be understood within the current approach as a synthetic development.
23
Grandgent (1907: 42–8), Muller and Taylor (1932: 65), Löfstedt (1959: 126), Väänänen (1966: 115–19;
1982: 195–7), Lakoff (1972: 189), Adams (1976b: 49), Bauer (1995: 137–9), Molinelli (1998: 147).
24
Vennemann (1974; 1975), Bichakjian (1987: 89), Bauer (1995: 5–6).
25
Sasse (1977), Harris (1978: 9), Bauer (1995: 7–6), La Fauci (1997: 41), Adams (forthc. a: }3.1).
22 From Latin to Romance
26
This conclusion is further confirmed statistically by Pinkster (1993b: 242) who finds the proportion of
NPs vs adpositional phrases in Caesar’s prose to be 85:15, whereas an examination of the first 306 nominal
phrases of the fourth-century Peregrinatio revealed only a slight increase in the number of adpositional
phrases, namely 77 (25%).
27
Adams (1977: 36–7), Bauer (1995: 138), Clackson and Horrocks (2007: 290).
28
See, for instance, Lindsay (1907: 20, 83), Bauer (1995: 137), Molinelli (1998).
Syntheticity and analyticity 23
free variants but were generally in complementary distribution, the ablative con-
struction typically occurring, for example, in proverbial expressions, negative struc-
tures, and rhetorical questions (see }5.3.1.2). Remaining with the comparative, we may
also note that alongside the synthetic formation in -IOR (as well as the superlative in
-ISSIMUS) analytic formations with PLUS and MAGIS ‘more’ (and MAXIME or PLURIMUM
‘most’ in the case of the superlative) are also attested since the earliest period (e.g.
CLARIOR vs MAGIS/PLUS CLARUS, CLARISSIMUS vs MAXIME/PLURIMUM CLARUS; Lindsay 1907:
38; Wüest 1998: 92), and indeed were obligatory with those adjectives ending in -VUS
(e.g. PLUS / MAXIME IDONEUS ‘fitter / fittest’). In both cases, we know that it is the
29
Adams (1976a: 98), Bichakjian (1987: 89), Roberts (1997).
30
Koch (1974), Miller (1975), Sasse (1977), Bichakjian (1987: 89), Bauer (1995: 7), Magni (2009: 247).
31
Bennett (1914: 292–7), Adams (1976a: 83–6), Bauer (2009a: 263), Magni (2009: 243–6).
24 From Latin to Romance
analytic variant that wins through in Romance, but this was not an immediate or
foregone conclusion in the Latin period.
Similar early synthetic-analytic variations are found in other areas, including the
expression of the future and the present perfect, where from an early date the
synthetic forms were, to some extent, in competition with periphrastic HABERE
formations (Pinkster 1987; Bauer 2006: 289). In the case of the future it is the analytic
structure (itself eventually synthesizing) which ultimately triumphs (}4.3.1.5), but in
the case of the perfect the original synthetic construction lives on and the competi-
tion between the two forms is far from over (Harris 1982), with the division of labour
between the two available paradigms resolved differently across the Romània (cf.
discussion of examples (9a–d) above). The case of the future considered here also
raises the related issue of competing analytic structures: assuming that there was a
general drift towards analyticity, how are we then to explain the fact that the already
existing periphrastic future forms of the classical period (cf. discussion in }2.2.1 of SUM
‘I am’ + future active and passive participle) were themselves replaced by the new
32
HABERE future periphrasis (Pinkster 1987: 221)?
In short, we have observed that it is demonstrably false to portray the passage from
Latin to Romance in terms of a simple and unified linear development from the
synthetic to the analytic. In many cases, synthetic and analytic developments
occurred in parallel and co-existed over long periods of time as contrasting and/or
stylistic variants, in some cases up until the present-day. Where analytic structures
have won through, these typically show a gradual development rather than a cata-
strophic or saltational change and co exist alongside many original synthetic struc-
tures that have survived into Romance from the ancestral language. Furthermore, in
some cases these same analytic innovations have ousted earlier classical periphrases
and/or have gone on themselves to become synthetic, developments which are
patently incompatible with the traditional thesis of an inexorable shift away from
syntheticity towards ever-greater analyticity.
32
For discussion of the variation in the formation of adverbs (e.g. MODO/MENTE ‘manner.ABL/mind.ABL’ +
ADJ), see Wüest (1998: 93) and Bauer (2003; 2006: 296–300).
Syntheticity and analyticity 25
In syntax too, verbal and nominal heads, once frequently preceded by such modifiers
as direct objects/genitives and manner adverbs/adjectives, come instead to precede all
such modifiers:
(17) a Lat.MORTEM METUIT > It.teme la morte
death.ACC he.fears he.fears the death
‘he fears death’
b Lat.MORTIS METUS > It.il timore della morte
death.GEN fear.NOM the fear of.the death
‘fear of death’
c Lat.LIBERE UIUIT > It.vive liberamente
freely he.lives he.lives freely
‘he lives freely’
d Lat.LIBER HOMO > It.un uomo libero
free man a man free
‘a free man’
33
von Wartburg ([1934] 1971: 256), Harris (1978: 6), Bauer (1995: 10, 24, 166), Oniga (2004: 52, 75). The
reversal in the head parameter is so pervasive that it equally surfaces in the area of derivational morphology
(cf. Lat. SILUICOLA lit. ‘forest.inhabitant’ (‘forest dweller’) vs Cat. guardabosc lit. ‘watch.forest’ (‘forester’)),
although admittedly at a later date than in the areas of inflectional morphology and syntax (Oniga 2004:
52–3).
26 From Latin to Romance
These latter examples, which clearly do not involve analyticity, therefore highlight
that the relevant change in linear order in syntax is consistent with that observed in
inflectional morphology, ultimately both derivable from a single integrated and
comprehensive structural change.34
Perhaps the only genuinely convincing attempt to invest the traditional synthetic-
analytic dichotomy with some explanatory power is to be found in Coseriu (1987),
who, however, reinterprets the distinction as one between ‘internal’ and ‘external’
structures, respectively, an opposition ultimately going back to his earlier work
(cf. Coseriu [1971] 1988: 211). According to this view, the Romance languages are
not characterized by a tendency towards analyticity, but, rather, by a tendency
to distinguish between external and internal determination and between relational
and non-relational functions. Although it is undoubtedly true that the internal (viz.
synthetic) determination of syntactic structures predominates in Latin, Coseriu
makes the original observation that this is not the distinctive structural characteristic
of Latin; rather what sets Latin apart from Romance is its failure to make any
formal distinction between relational and non-relational functions. By contrast,
Romance stands out, not by its admittedly steady increase in the number of exter-
nally determined (viz. analytic) structures, but, rather, by its consistent formal
distinction between relational and non-relational functions, the former aligned
with externally determined (= analytic) structures and the latter with internally
determined (= synthetic) structures. Thus, the real change in the passage from
Latin to Romance is not to be seen as a move from syntheticity to analyticity, but
is to be identified with the emergence of a new formative principle of mapping
relational and non-relational functions respectively onto externally and internally
determined structures, as consistently and uniformly attested across a whole series
of otherwise superficially puzzling and unrelated changes from Latin to Romance
(cf. also Coseriu [1971] 1988: 224).
It therefore follows, for example, that the internally determined (paradigmatic)
functions of the Latin case system or comparative are replaced by externally deter-
mined (syntagmatic) expressions in Romance, since they both involve relational
functions which establish a relation between a given term and one or more other
terms (e.g. X is ‘of ’ Y, or X is ‘bigger than’ Y). By the same token, number and gender
marking on nominal categories are argued to instantiate non-relational functions and
thus continue to be marked paradigmatically by internally determined expressions.
As for verbs, Coseriu (pp. 61–2) maintains that simplex verbs convey non-relational
34
Harris (1978: 16) too sees the emergence of a specified-specifier order as central to the developments
in the syntax (e.g. SVO) and morphology (e.g. loss of inflection) of Romance. Yet, he does not try to
subsume the shift from synthetic to analytic within this linear change but, rather, continues to treat it as an
independent, albeit isolated, phenomenon, ultimately part of a general tendency towards more explicit
structures (see also Bourciez 1956: 23).
Syntheticity and analyticity 27
functions in that they pick out a single moment in time or temporal space, and hence
are internally determined structures, whereas compound verb forms establish a
relation between two points in time or space, and hence display externally deter-
mined structures.35
A further revealing case concerns the fate of the Latin synthetic passive, which
Coseriu (p. 63) claims to subsume three distinct functions: (i) the objective passive
(18a); (ii) the middle (18b); and (iii) the impersonal passive (18c).
(18) a ab eis ita amantur (Lat., Cic. Amic. 27)
by them.ABL thus they.love.PRS.PASS
‘they are thus loved by them’
b aut cum Fabricius, aut Aristides iustus nominatur (Lat.,
or when Fabricius.NOM or Aristides.NOM just.NOM call.PRS.PASS.3SG
Cic. Off. 3.16)
‘or when Fabricius or Aristides is called “the just”’
c sic itur ad astra (Virg. Aen. 9.641)
thus go.PRS.PASS.3SG to stars.ACC.PL
‘so it is man rises to the stars’
Significantly, of these three functions, it is only the first that expresses a relational
function between a PATIENT and an AGENT, which gives rise to the Romance periphras-
tic construction (19a), while the other two non-relational functions are aligned with
apparently internally determined structures (19b–c):36
(19) a ils sont aimés par eux (Fr.)
they are loved by them
b Fabrice s’ appelle le juste (Fr.)
Fabrice self= calls the just
c c’ est ainsi qu’ on atteint les astres (Fr.)
this is thus that one attains the stars
35
While this analysis might be true for many of the Romance languages, especially in earlier periods, it
runs into difficulties when we consider those varieties that still express the present perfective, a relational
function, through an internally determined (viz. synthetic) structure such as southern Calabrian (cf. also
the meaning of compound form in (9a) above):
(i) ancora non mi chiamau (SCal.)
yet not me= he.called
‘he hasn’t rung me yet’
Equally problematic are examples such as (9d) from spoken French, where the compound form can now
also mark the apparently non-relational function of the punctual perfective.
36
As pointed out to me by G. Salvi (p.c.), in reality examples (19b–c) appear to be just as analytic as
example (19a).
28 From Latin to Romance
Coseriu ([1971] 1988: 219–20; 1987: 62) also applies this model to sentential
word order, arguing that changes in this area cannot be appropriately characterized
as a shift from a ‘circular’ (viz. SOV) to a ‘linear’ (viz. SVO) arrangement, but
are to be viewed once again in terms of the emergence in Romance of a formal
distinction between internal and external functions that was not systematically made
in Latin. Specifically, functionally internal determinations such as adjectives
and modals which modify exclusively the subject and the verb, respectively,
come to be placed in Romance ‘inside’ the sentential core (understood here as the
syntagm consisting of the subject and verb) (20a–b), whereas functionally external
determinations such as complements come to be placed ‘outside’ of the sentential
core (21a–b):
(20) a BONUS HOMO scribere DEBET (Lat.)
good.NOM man.NOM write.INF must
b el hombre bueno debe escribir (Sp.)
the man good must write.INF
‘the good man must write’
(21) a HOMO SUO AMICO LITTERAS SCRIBIT (Lat.)
man.NOM his.DAT friend.DAT letter.ACC.PL he.writes
b el hombre escribe una carta a su amigo (Sp.)
the man writes a letter to his friend
‘the man writes a letter to his friend’
2.2.6 Grammaticalization
By way of conclusion, we should also note that the postulation of an independent
synthetic-analytic parameter is further undermined by the observation that all
presumed cases of analytic development can otherwise be independently subsumed
within the general theory of grammaticalization (Hopper and Traugott 1993: 17),
integrating synthetic forms as those having achieved the highest degree of gramma-
ticalization (Klausenburger 2000: 105, 152; Bauer 2006: 288).37 More precisely, the
analytic developments witnessed in the history of Romance are clearly not in any way
exclusive to the Romance family, but simply exemplify a crosslinguistic tendency for
37
By contrast, Bauer (2006: 301) argues that syntheticity and analyticity in themselves are not an
indicator of the degree of grammaticalization of a given structure, insofar as the synthetic or analytic
outcome of individual grammaticalization processes is determined by branching patterns: left-branching
structures produce synthesis whereas right-branching structures yield analysis. Superficially, at least, there
are, however, numerous counterexamples to this generalization including, for example, the Latin typically
left-branching perfective passive (e.g. SCRIPTUS EST write.PRF.PASS.PTCP is: ‘it has been written’; cf. }5.3.1.5) and
the (apparently) Romanian right-branching enclitic definite article (e.g. bărbatul ‘man=the’: ‘the man’;
}4.2.2.2).
Syntheticity and analyticity 29
synthetic structures, once weakened through phonetic erosion or other forces within
the system, to be progressively replaced by new competing structures which ‘given
the nature of syntactic change, cannot help but be analytic’ (Vincent 1997a: 101).38
Once again, analyticity turns out to be a secondary or epiphenomenal development,
ultimately the manifestation of a deeper change but not, significantly, its cause.
38
Thoroughly unconvincing is the position taken by Wüest (1998: 96–7), who argues that the largely
analytic tendencies of Romance are the result of a dramatic morphosyntactic restructuring in the Latin
period due to language contact, rather than spontaneous internal change. In particular, he argues that
during the period of Romanization a sizeable section of the population of the Empire would have had only
a very fragmentary knowledge of Latin, speaking a rudimentary code he terms an ‘inter-language’. Their
speech, he presumes, would have been characterized by such features as the imperfect mastery of
morphology, which, consequently, they would have tended to replace with periphrases that would
subsequently become established in the speech of the entire population. Apart from the fact that there is
no evidence, direct or indirect, for this presumed source of analyticity or the presumed inter-language itself
(cf. Adams 2003), among the many other objections to this theory it will suffice here to recall that many of
the original morphological structures of Latin (together with many of their irregularities), especially in the
verbal domain (Maiden 2011a), have been retained in Romance (see }2.2.1), a situation which is clearly
incompatible with the inter-language scenario postulated by Wüest.