You are on page 1of 23

Linguistic Society of America

Where Does Latin Sum Come from?


Author(s): Martti A. Nyman
Source: Language, Vol. 53, No. 1 (Mar., 1977), pp. 39-60
Published by: Linguistic Society of America
Stable URL: http://www.jstor.org/stable/413054 .
Accessed: 27/01/2011 14:45
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .
http://www.jstor.org/action/showPublisher?publisherCode=lsa. .
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access to Language.

http://www.jstor.org

WHERE DOES LATIN SUM COME FROM?


MARTTIA. NYMAN

Universityof Helsinki
The derivation of Lat. sum, es(s), est from IE *esmi, *esi, *esti involves methodo-

logical problems. The data, as they appear, e.g., in early drama, are well known, but
they have been incorrectlydescribed.At face value, the paradigmof Plautus contains
allomorphicvariation-sum, ess ss, est st-which has traditionallybeen explained
away by 'aphaeresis'. This has been used as a synchronic morphophonemic rule,
generatingthe surfacevariantsfrom the underlyingforms /ess/ and /est/ respectively,
and with the implicationthat the IE model is relateddirectlyto these underlyingforms.
But there is evidence that 'aphaeresis' has no linguistic basis in Latin, and the IE
paradigm described must be in terms of the allomorphic variation observable in
Plautus. It is claimed here that the development of sum from *esmi is related to the
origin of the variation est st (< *esti). The study is primarilyconcerned with the
mechanism of this remodeling process, but some chronological suggestions are also
made.*

The Latin copula has been a stumbling block for students attempting to relate its
present indicative paradigm (1) to the Indo-European model paradigm (2):
(1) sum, es(s), est, sumus, estis, sunt
(2) *esmi, *es(s)i, *esti, *smos, *ste(s), *senti (N *s6nti)1
Relating 1 to 2 apparently presupposes more than mere operation of sound laws.
However, recourse to analogy as an explanatory principle has been shunned by
some scholars; as Bonfante (1932:114) says, 'on ne voit absolument pas comment
un paradigme du type sum, ess, est, sumus, [estis], sunt aurait pu naitre par analogie.'2 This conclusion is understandable, since analogical development may
reasonably be expected to have produced a paradigm more regular than 1, something like that (re)constructed by Varro (Ling. 9, 100): esum, es, est.
To resolve the controversy, Bonfante constructed the model paradigm on the
Latin basis, while van Wijk 1905 posited IE by-forms to which the Latin paradigm
could be related by means of sound laws (e.g.

*s5mos

>

sumus). These lines of

reasoning have been criticized by Szemerenyi (1946, 1964:191-5), who gives an


insightful account of the processes apparently involved in the development of the
'unexpected' Latin paradigm from the IE model. Like most scholars, Szemerenyi
* I am grateful to Philip Baldi, Jaakko Frosen, Esa Itkonen, Andrew Sihler, and Oswald
Szemerenyifor readinga preliminarydraft of this paper.Theircriticalremarkshave contributed
invaluably to the final version. The responsibilityremains, of course, entirely my own.
1 The model paradigmis based on Szemerenyi(1970:288; cf. 1964:191,n. 4) and on Watkins
(1969:25-6). As a result of comparativereconstruction,the variationit shows is naturalenough.
To quote Hall (1960:203): 'Ever since the beginningsof the comparativemethod, it has been
evident that ... every proto-language has to be reconstructed as non-uniform, i.e. showing
dialectal variations.'
2
Ernout (1953:175) motivates the anomalous characterof the copula by invoking the frequency of use: 'La verbe signifiant"etre" est le plus irregulierde la langue latine. En raison de
frequence de son emploi, il a echappe en grande partie aux actions analogiques et a conserve
sa structurecompliquee.'
39

40

LANGUAGE,

VOLUME 53, NUMBER

1 (1977)

accepts 2 as a point of departure. It is also the starting point of the present study.3
Adopting 2 as our point of departure challenges us to find a conceivable way of
relating it to the Latin facts. First, in ?1, I shall review the path opened by Szemerenyi, not because it necessarily leads to the truth, but because it is the most
thorough study so far.
1. SZEMERENYI'S
ANALYSIS.According to Szemerenyi 1964, the remodeling pro-

cess started from certain phonetic difficultiesconnected with the voicing of s before
voiced consonants, with the forms *smos and *esmi developing into *mos and *emi
respectively. This development would have destroyed the transparency (Kiparsky
1971) or iconicity (Anttila 1972) of the paradigm.4 Instead of using one or two
(synchronic) base forms (/es-/ and /s-/), from which the surface forms could have
been derived, speakers would have had to memorize a list of suppletive variants:
*emi, *essi, *esti, *mos, *stes, *sonti. The attested form sumus, which comes from the

earlier *somos, indicates that the speakers chose to salvage the synchronic derivability of the paradigm;5 they separated the endangered sequence sm by the anaptyctic vowel o, 'whose timbre was determined by the labial character of m' (191).
3 Let it be mentioned, however, that paradigm2, which rests primarilyupon the Sanskritand
Greek evidence, would not be accepted as basic by everyone. Philip Baldi (privatecommunication) has drawn my attention to the proposal of Schmalstieg 1972 which, if accepted, would
force us to reconsiderthe possibility that the Latin paradigmmay representthe earliest stage,
with Skt. asmi, stha as well as Gk. eimi ( < esmi) being innovations. On the basis of fragmentary
data collected from various IE languages, Schmalstieg reconstructs 'an ancient accentual
mobility which consisted of putting the stress on the final syllable in the first singular, first
plural and third plural but the penultimatesyllable in the second and third singular and the
second plural' (136). This accentual pattern generates the ablaut grades characteristicof the
presentindicativeparadigmof the Latin 'athematic' verbs sum'I am', fero 'I carry', edo 'I eat',
and volo 'I wish', as well as the Gothic third weak class etc. Such a reconstructionhas the
theoretical property of deriving the more predictable from the less predictable, since it is
obviously easier to explain the Sanskritand Greek paradigmsby the principleof paradigmatic
leveling than to derive an apparentlyirregularparadigm(like that in Latin; cf. Bonfante, 114)
from a more regular one (e.g. Skt. asmi, asi, asti). Substantially, of course, Schmalstieg's
conclusion is not cogent, becausethe linguisticphenomenaservingas input to the reconstruction
can be, and have been, explained as internal developments(cf. Jasanoff 1973 on the Germanic
third class, and Szemerenyi1964:198-9 on Lat. fers, fert etc.)
4 The point Szemerenyiis making can be expressedin these terms, although he does not use
them.
5 The conflict between synchronic derivability(transparency)and list character(opacity) is
the everydaylife of language, and linguists attemptingto write psychologicallyreal grammars
must take account of this fact. In the case of the Latin copula, the change *smos > *somos
reflects the victory of synchronicderivabilityover listing of variants(cf. *mos and *emos). But
the Romance languages show that lists have won in the end, despite the workings of analogy
(cf. Roth 1965). Again, compare the Avestan forms ahmi, asti, mahi, hanti to the respective
Sanskrit forms asmi, asti, smaki, santi: here too, lists have won. The Greek paradigm is an
instructiveexampleof these conflictingtendencies.At the outset, the full root es- was generalized
to the whole paradigm,as is seen in the Mycenaeangraphemicsequence e-e-si 'they are', i.e.
/e(h)ensi/ < *es-enti(Strunk 1960, cf. Watkins27) or /?(n)si/ (J. Frosen, p. c.). Then the initial
and intervocalic s became h and subsequently disappeared,producing the opaque paradigm
eimi, ei, esti, eimen, este, eisi, which had to be analysedin a differentmanner than before. The
forms of the paradigmcould no longer be synchronicallyderivedfrom a single underlying/es-/.
Instead, the variantses- and ei- were abduced and listed in the lexicon, as can be seen in hypercharacterizedforms like Ionic eis 'you (sg.) are' (i.e. ei 'you are' plus 2sg. -s; e.g. II. 16.515,

WHERE DOES LATIN SUM COME FROM?

41

The therapeutic effect could also, and even more naturally, have been achieved by
generalizing es- to the whole paradigm; but Szemerenyi points out that '*esmos
would have presented the same difficulties as *esmi' (192), i.e. *esmos > *ezmos >
*emos. (This is the way it went in Greek.)
Having explained why es- was not generalized, Szemerenyi remarks (193) that
'the form *smos became difficult, and therefore called for remedy, long before the
form *esmi. This means that the new *somos was created when *esmi was still in
use.' It is evident that initial sm became difficultearlierthan medial sm. The evidence
adduced by Szemerenyiis strengthenedby the form co-smitterefrom VerriusFlaccus
(as cited in Paul. Fest. 67), who had access to this obsolete form but not to the
primeval simplex *smittere (or the like). Besides, the inscription of Duenos has
mitat (not *smitat), although it has cosmis (for comis).6 There are, moreover,
parallel developments suggesting that phonetic innovations involving cluster
simplification tended to begin in the initial position.7
At this point, however, a trifling inconsistency appears. If initial sm in *smos
became difficult long before medial sm in *esmi, there of course was no principled
reason why the situation COULD not have been remedied by introducing the full
stem *esmos. But sumus leads us to conclude that es- was not generalized. Instead,
s and m, which were felt as belonging to different constituents, were separated by
means of the vowel o. Szemerenyi thinks that the development in question was
phonetic in character, viz. some kind of therapeutic anaptyxis. However, the
'anaptyctic' o can be and has been interpreted in a different way (see, e.g., Leumann 1963:310).
As soon as medial sm also began to create difficulties, the endangered *esmi was
gradually replaced by *esomi. According to Szemerenyi (192), this change cannot
have taken place much earlier than the Latin rhotacism, i.e. about 350 B.C. In historical Latin, the sg. form is sum, which can easily be traced to *som(i). But why
was *esomi replaced by *somi?
Szemerenyi suggests three contributing factors. First, he states that *esomi gave
way to *somi 'perhaps partly under the influence of *somos' (192). The influence of
*somos has been invoked many times before, but not via the intermediate *esomi
(cf. Sommer 1914:528, Leumann 1963:310, who arrive at *som directly through
Hdt. 3.71.3); this was a deductiveinnovation resultingfrom an abduction (for the terminology,
see Andersen 1973). The mechanismis clear enough; cf. the Herodotean forms ea-s 'you (sg.)
were' (1.187.5) and ea-te 'you (p1.)were' (4.119.2) in which the stem ed- is originally the impf.
lsg. morpheme (cf. Chantraine 1961:206). These examples show that sound laws produce
paradigmatic irregularity,creating lexicalized allomorphs which may be used for creative
purposes (cf. Anttila 1975). The innovation essi (e.g. Od. 1.175, Pind. 01. 6.90) reflects,on the
other hand, an attempt to make the root es- iconic by means of internalreconstruction(which,
in this case, actuallyrestoresthe early IE situation). In Attic, however, it was ei that survivedin
the 2sg. form, while eimen was replaced by esmen. The curious fact for the item-and-process
linguist is that eimi was not replaced by esmi. This shows, again, that lists tend to prevail,
especially in the most frequentlyused forms.
6 See Kent 1926, Goldmann 1926.
7 Cf. gnoscere : cognoscere > noscere: cognoscere(class.) > noscere : connoscere(later; cf.
Fr. connaitre).Accordingto Fisher (1948:156), connoscereis attested in Plaut. Amph.822E and
Truc. arg. 9B; but these are surely clerical, not Plautine, reflexes.

42

LANGUAGE, VOLUME 53, NUMBER 1 (1977)

the proportion x: *somos = *esiam:*esdmos). Second, from Szemerenyi's treatment of the Oscan sim, the principle of paradigmatic pressure can be extracted:
*esomi gave way to the disyllabic *somi to fit into the same series with the other
disyllabics *essi and *esti (194). Third, 'the reduction of *esomi to *somi was no
doubt speeded up by such phrases as *nesomi (ne-esomi), ego6(e)somi'(195).
Szemerenyi adduces sentence-phonetic factors, but confines discussion to the Isg.
form.
2. THE LATIN FACTS AND THEIRDESCRIPTION.
All attempts to relate the Latin

paradigm (1) to the IE model paradigm (2) have been based on the assumption that
es(s) and est directly continue the IE forms *es(s)i and *esti respectively. Accordingly only sum has been regarded as problematic:
IE

(3)

LATIN

*esmi -?-

sum

*es(s)i ->
*esti ->

es(s)
est

This current misconception stems from an incorrect or uncompleted synchronic


description of the philological data. Adoption of this conventional description
prevented even Szemerenyi from going far enough in his analysis.
Before our first documents, there was a period during which the 2sg. and 3sg.
forms had only the forms *ess(i) and *est(i) respectively. But as soon as we arrive
at Archaic Latin and written records, we face the variations (2sg.) ess ss and
(3sg.) est

st. The following examples are from Plautus:

(4) Cas. 1007 non iratass? 'you're not angry?'


Mil. 615 quis homo sit magis meus quam tuss? 'who's more a man after
my own heart than you are?'
Amph.836 mulieress, audacteriuras'you're a woman, you swear boldly'
Amph.937 iam nunc irata non ess? 'you're not angry anymore?'
Cist. 120

idem mihist ... uitium 'I've got the same fault'

Cas. 587 quodfactost opus 'what needs to be done'


Aul. 147 quid est id, soror? 'what is it, sister?'
Men. 1128 quod tibi nomen est, fecit mihi 'he gave me your name' (lit.
'the name which is to you').
The distribution of the 'full' and 'reduced' forms can be stated, roughly, as follows
(for details, see Nyman 1974:7-8, Brinkmann 1906):
(5) ess, est after consonants;
after vowels.
ss, st
The change, if expressed by means of diachronic correspondence between IE and
Plautine Latin, was not (e.g.) *esti > est, but *esti > est st. Consequently, the
problem is not that summarized in 3, but rather that of relating *esmi to sum,
*es(s)i to the variants ess ss, and *esti to est st:
-

(6)

IE

*esmi
*es(s)i->

LATIN
->

sum
ess

*esti ---> est

'

ss

st

WHERE DOES LATIN SUM COME FROM?

43

This variation in Plautus was, of course, a synchronic fact of his language, and-we
may assume-of Archaic Latin in general. It would thus be interesting to try to
describe how the variation ess ss, est st was structured (in the sense of Linell
1974:45) in the synchronic grammar of Plautus. At least three possibilities can be
envisioned:
(7) The form est (ess) was taken as basic, and st (ss) was synchronically

derived from it by 'aphaeresis':


e-0/V

sC/

This is the traditional approach represented by a whole series of respectable


scholars (Brinkmann 1906, Sommer 1914:293, Leumann 1963:174, Safarewicz
1953:98, Soubiran 1966). According to this analysis, the paradigm is representedas
sum, ess, est, sumus, estis, sunt.

(8) The copula had two lexicalized paradigms, a 'full' one and a 'reduced'
one:
(sum, ess, est,
sunt
sum, ss, st, sumus, estis,

This kind of analysis was suggested in passing by Havet 1884. The selection of one
paradigm or the other was determined by the phonological context, as in 5.
(9) The form st (ss) was taken as basic, and the 'full' form est (ess) was
derived from it synchronically by the application of a morphophonemic
sandhi rule characterizableas 'prothesis':
0
e//C#
sC#
This analysis is suggested by Nyman 1974. Accordingly, the Plautine paradigm is
represented as sum, ss, st, sumus, estis, sunt.8

However, the synchronic description of the Plautine system is not our primary
concern here, although some discussion will be devoted to it below. The relevant
question for the present purpose is this: How did the variants ss and st come into
being? The commonly accepted answer has been 'aphaeresis', regarded as caused
8 The question concerninglexical storagecan, of course, also be put in a more subtle
manner.
It can be asked whetherthe presentindicativeparadigmof the copula was stored as a paradigm
consisting of 'word-size' units, or whetherthe speakerswere able to segment the 'words' into
smaller building-blocks ('morphs'). It has been observed above (fn. 5) that morphological
evolution can be viewed as a constant tug of war, with lists of suppletives, or surface variants
tied to syntagmaticassociationsand languageuse, opposed to processesfor derivingthe surface
forms from basic or underlyingroots (synchronicderivability).Accordingto one suggestion(cf.
Kehoe & Whitaker 1974), words of great frequency are stored as 'word-size' units, while
infrequent words are stored as smaller units; but this hypothesis is at best only a rough
approximation.In the case of Latin, the replacementof *smosby *somosseems to indicate that
speakers were aware of the morpheme boundary between s- and -mos, although *smos was
phonetically monolithic. It also reflects the victory of synchronic derivability over listing of
paradigm forms (*smos would have resulted in *mos, if listed without internal constituent
analysis). It seems that the synchronic derivabilityof the present indicative of the copula had
already begun to fade at the time of Plautus (with the introduction of estis).

44

LANGUAGE, VOLUME 53, NUMBER 1 (1977)

by enclisis. While the enclitic nature of the copula is beyond question in itself,
inferring 'aphaeresis' as a phenomenon consequential upon it has involved an
epistemological 'salto mortale' which has not been very successful (cf. Nyman
1974). This problem will be my main topic in ?3, which contains a review of the
processes conceivably involved in the metamorphosis of the IE model paradigm
into Latin, when examined in pursuance of the time-honored research (including
Szemerenyi's).
3. INFEASIBILITY
OF'APHAERESIS'.
According to Szemerenyi, the replacement of
IE *esmi by *esomi was caused by the phonetic difficulties created by the sm
sequence. This explanation is very attractive, although it suggests that Oscan sumwhich, from one perspective of comparative reconstruction, seems to have a
common origin with Lat. sum (cf. Sommer 1901:347, 1914:528)-is either an
indigenous development, or the result of Latin influence (Szemerenyi 1964:194,
n. 3). Subsequently *esomi developed into *som(i), with possible contributory
factors as enumerated above. Although the influence of *somos has been generally
invoked, this explanation suggests 'petitio principii', since it seems to apply only
in this particular case.9 Also, the alleged tendency toward parisyllabism within the
singular paradigm needs supplementary evidence. Considerations of sentence
phonetics have not been carried out to the logical conclusion: if the form *nesomi,
whose constituent boundary was opaque, was resegmented as *ne-somi, why was
not the equally ambiguous*nesti (see Brugmann 1904:210) segmented as*ne-sti?
Obviously the way in which the Latin paradigm is usually given in grammars and
handbooks (sum, es, est) has made this question inconceivable.
3.1. Seemingly, Lat. est (ess) is the direct continuation of IE *esti (*es(s)i).
It has therefore been easy to consider est (ess) as the basic alternant, from which st
(ss) was derived by 'aphaeresis'. Invoking either phonetic or semantic weakness of
the copula, together with its enclitic nature, has not led to clarification of problems.
On the contrary, unnecessary problems have been created.
From the diachronic point of view, why was e phonetically weak only in (e)ss and
(e)st-but not, e.g., in esse (Nyman 1974:8-11)? And if the suppression of e in
(e)st was caused by the meaninglessness of the copula (e.g., Ubi est liber?=> Ubist
liber?; cf. the total deletion of the copula in Russian Gdejest' kniga?=> Gdekniga ?),
then why was est reduced even when used 'vi substantiva' (cf. Plaut. Cist. 735 est
quidamhomoqui ... ait, but Mil. 1012 homoquidamstqui scit)? It seems that neither
phonetic nor semantic factors can be held responsible for the 'aphaeresis' of e in
est (ess). In general the reduction has been seen as an effect of the enclitic nature of
the copula. However, it is incomprehensible why the enclitic effect should be so
strong, especially in monosyllabic forms; the opposite might be expected.
From the preceding it can be seen that it is very difficult to motivate 'aphaeresis'
historically. Moreover, in regard to terminological content, Latin 'aphaeresis' is a
9
Cf. the 1sg. future (faci)am, in spite of pl. (faci)emus. It is true that the 'expected'
desinence -em occurs in the manuscripts (e.g. Plaut. Mil. 676 accipiem [-am Nonius], Truc. 963
sinem); but the occurrences are very few, and are generally regarded as clerical errors (Leumann
1963:326; cf. Sommer 1914:525).

WHERE DOES LATIN SUM COME FROM?7

45

phenomenon categorically different from the Greek aphaeresis (see Nyman


1974:9-18); and thus Leo was justified in stating that "'Apharesis" [ist] fur das
Lateinische kein grammatisches Begriff' (1912:285). His explanation 'Synalophe
mit Enklisis' is, however, equally difficult. In what follows, one more problem will
be taken up to show how 'aphaeresis' as an explanatory principle can only
complicate the issue.
3.2. In the data presented in 4, one unexpected and therefore difficult type was
left out. Consider the following examples from Plautus:
(10) Mil. 574 sed satine oratuss? :: abi 'but have I begged your pardon
enough? :: Off with you'
Mil. 825 eho tu sceleste, qui illiisubpromuss,eho10'ha, you rascal, you're
his under-butler, so see here'
Mere. 1004 nihil opust resciscat 'there's no need of her learning'
Amph. 615 geminus Sosia hicfactust tibi 'this Sosia has been made your
twin'
Pseud. 248 mortuost quifuit; qui sit ussust 'a past one's a dead one; a
present one's needed'.11
How are forms such as oratuss 'you have been begged', opust 'it is needed',
mortuost 'he's dead' etc. to be explained historically? And how are they analysed
synchronically?
Three principles have been invoked in order to explain the provenance of the
forms in -uss and -ust:
(11) a. Haplology (e.g. Sommer 1914:293): opus(es)t.
b. Loss of s, with subsequent contraction of vowels (Leo, 279-88): opus
est > opu est > opust.
c. Aphaeresis (e.g. Brinkmann 1906:73-4, Juret 1913:95): opus est >
opus st > opust.
The feasibility of haplology has been discredited by Nyman (1974:30-31), and
Leo's explanation has been generallyrejected(see e.g. Brinkmann,45-74, Leumann,
175). Aphaeresis seems to furnish the only serious solution; but even this explanation is very difficultto justify, because it runs counter to our metatheory of phonetic
naturalness. In principle, it is quite conceivable that 'aphaeresis' could take place
before vowels; but why should s, which is a consonant, behave like a vowel?
Brinkmann'sphonetic explanation, according to which the weak e was lost between
identical consonants (73-4), is problematic;12and after all, it is very improbable
that the change -us est > -ust was phonetic in character. On the other hand, the
possibility of analogical development remains distant as long as it is unclear what
could have provided the model for extending the 'aphaeresis'
10 Cui tu
suppromusseho, B

analogically to

(see Leo, 281).


difficult problem of -est for -is est will not be considered in this study (see Nyman
1974:33-5, Touratier 1975).
12
For example,it fails to motivatethe non-occurrenceof forms such as *salust(for salussest)
(see also Nyman 1974:31).
1' The

46

LANGUAGE, VOLUME 53, NUMBER 1 (1977)

-us(e)st (as suggested by Juret), for the vowels do not form a natural class with the
sole consonant s.
3.3. From the synchronic point of view, deriving -ust from -us est involves
serious problems, which weaken the 'traditional' approach outlined so far.
Technically, it is not difficult to manipulate the environment of 'aphaeresis' (7) to
'account for' the complication created by the -ust cases:
e-^
(12)

sC

v{ _s

But, to be sure, this is no linguistically significant generalization-whether it is


judged by counting the distinctive features required (as suggested by Halle 1961),
or by observing the functioning of the combination 's & vowels' as a natural class
(in the spirit of Campbell 1974).
Let us consider, e.g., the form opust. Should it be segmented as opu'st or as
opus't? Our first guess is that opu'st is the 'correct' segmentation. (This is also the
option of most editors.) This assumption enables us to make the following
important observation:
(13) From the morphophonological point of view, the final s seems to behave
as though it preceded a word beginning with a consonant.
This is to say that final s seems to disappear as if conditioned by the same factor
responsible for the loss of final s in instances such asfacturus sum -+facturu' sum,
opus sunt -? opu' sunt, Venus mi -. Venu'mi etc. Cf. the evidence provided by the
Plautine prosody:
(14) Asin. 376 dico hercle ego quoqueutfacturu' sum 'and by the Lord, I'm
telling how I'm going to do it'
Cap. 164 iam maritumiomnes milites opu' sunt tibi' and furthermoreyou
need every sea soldier'
Mere. 38 eodem quo amorem Venu'mi hoc legavit die 'it was my legacy
from Venus on the same day she gave me my love'.
What entitles us to interpret the -ust type as a sub-case of the tendency of final s
to drop before consonants? First, both phenomena are facultative in Plautus:
Bacch. 705 sed nunequantillumusu'st auri tibi Mnesiloche?
(15) a. -u'#st:
die mihi 'well now, Mnesiloche, what's the paltry sum you
need? tell me'
-us#est: Bacch. 706 militi nummisducentisiam usus est pro Bacchide
'I need two hundred pounds at once to pay the captain
for Bacchis'
b. -u'#C-: Epid. 535 credo ego illi hospitio usu' uenit 'she's in need of
hospitality, I dare say'
-us#C-: Bacch. 763 sed nunc truculentomi atque saeuo usus senest
' but now I must make the old man feel fierce and savage'.
Moreover, both phenomena have similar 'frequency profiles'. The statistics provided by Brinkmann (51-2), concerning the ratio between -ust and -us est,13 con13 -us
est, 248; -ust, 440 (-us ess, 67; -uss, 133). The 'frequency profile' is just the opposite in
Terence.

WHERE DOES LATIN SUM COME FROM?

47

verge with the general characterization by Lindsay: 'In the earlier poetry it is the
rule, not the exception, that final s before an initial consonant does not lengthen a
preceding short vowel by "position"' (1915:40, n. 3).
Second, as a rule, s disappears only when preceded by a SHORTvowel. This statement holds true with regard to both phenomena. There is no -Vs#C-

'
*-V'#C-

variation in Plautus, nor has the -uis#esttype any *-u'#st counterpart.14


Third, the short vowel in question is prevailingly u (etymologically *o) in both
cases. This suggests a diachronic relationship. Note Brinkmann's statement
concerning the -ust cases, as compared with Proskauer's observations concerning
the loss of s:
Verum enim vero ea ipsa re, quod -us terminatio multo crebrius quam -is et -es in sermone
Romanorum adhibebatur,evenire iam poterat, ut litterarum-s est in -st contractio a compositione -us est initium caperet.- Itaque syllabarumin -us cadentiumcum est contractionemab
altera et tertia personissing. perf. pass. profectamesse aio: primofactus es, factus est, similia in
factus factust contracta sunt, quae contractio postmodo omnia alterius declinationis nomina
arripuitatque postremo usque eo propagataest, ut quamlibet formam, quae in brevem u cum
s simpliciexibatcum es est contrahereliceret.(Brinkmann,50-51; originalemphasisremoved.)15
In ein- und derselbenInschrift ist mehrmals-s nach -o weggelassen,wahrendes nach anderen
Vokalen geschriebenist. - -s fehlt ausser nach -o nur einmal nach -e (Proskauer1909:15) ...
Das indogermanische-s kam ... hinterdem offeneno-Laut in Wanken.- Fehlt -s nach anderen
Vokalen, so ist das der Analogie des besonders haufigen-o zuzuschreiben(38).

These statements imply *o as the point of departurefor analogical extension to other


vowels 16
Fourth, the -ss est type has no -st counterpart. There is no such variation as
hospess est *hospest. If interpreted from the perspective adopted in this section,
this circumstance turns out to be an instance of the fact that -s is not suppressed
if it is morphophonemically -ss. In other words, there are no such instances as
*hospe'fuit (for hospessfuit).
Fifth, the -ust type seems to have a 'fatal' unity not with the 'normal' aphaeresis,
but with the loss of final s. While the -st variant continues to be used after vowels
during the classical period (see e.g. Siedow 1911), the PRODUCTIVITYof the -ust type
can be seen as DECLINING
CONVERGENTLY
with the productivity of the loss of final
-s. The correctness of this statement is not, however, obvious without qualificatory
notes; consider Table 1 (p. 48).
14 It is, however, highly probable that this second rule reflects only a general norm or tendency (cf. the discussion in Nyman 1975, Perini 1974:131-42).
16 'To be sure, for the very reason that the -us ending was used much more frequentlythan
-is and -es in the speech of the Romans, it might have happened that the contraction of -s est
into -st originatedfrom the -us est combination. - Thereforemy claim is that the coalescence
of the syllables ending in -us with est originatedfrom the 2nd and 3rd sg. perf. pass.: factus es,
factus est etc. were first contractedinto factus, factust. This contraction then affectedall nouns
of the 2nd declension, and finally had such an extension that it was possible for any form
terminatingin a short u and a simple s to coalesce with es and est.'
16 Juret's
explanation (91-5) also presupposesthis. In general,grammariansseem to generalize the environmentalstatement to include ALL short consonants (e.g. Sommer 1914:303-4)without sufficientscreeningof philological data, to my mind.

LANGUAGE,VOLUME53, NUMBER 1 (1977)

48

CATULLUSCICERO
PLAUTUS TERENCE ENNIUS LUCRETIUS
+

-V'#st

+a

-V SC+
+
+
+
+b
+
TABLE 1. Occurrence of -V'#st and -V'#C- in six authors (Cicero: translation
of Aratus' Phaenomena. '+' = occurring; '-' = not occurring).
a

The fragmentsof Ennius contain only three instances of -V'#st, all of which

are -ust.
b Catullus employs this device only once, viz. in the last line of the last poem
(166,8), and notably after i: dabi(s).

The table shows that the -ust type is not employed by the time of Lucretius,
Catullus, and Cicero; and even the fragments of Ennius have only three cases of
-V'#st, all of which are -ust (Ann. 129 datust, 252 paratust, 306 dictust). Thus, at
first sight, Table 1 seems to contradict the above statement; but consider Table 2.
3rd
4th
2nd
5th
1st
8
8
5
15
58
Enn. Ann.
6
2
4
2
35
Lucretius
TABLE2. The frequencyof the suppressionof final s, as distributed
according to the metrical feet in the hexametricpoetry of Ennius
and Lucretius. The table is based on the statistics in Bailey
(1947:124).

Table 2 shows that final s is usually suppressed on the fifth foot, which by
metrical convention is constrained to be of dactylic form. (On the other feet there
is a free choice between dactylic and spondaic forms.) On the basis of his statistics,
Bailey concludes: 'whereas in Ennius' day the suppression was regarded as normal,
by the time of Cicero and Lucretiusit was considered as archaic licence'. Moreover,
he states that' Ennius seems to be ready to use the suppression in any foot where
he wishes for it.' No attention is paid to the prevalence of the fifth foot both in
Ennius and in Lucretius, although the statistics clearly indicate that the suppression
of final s was a device for creating the requireddactylic prosody. We are entitled to
conclude that, even in Ennius, most instances of the suppression falling on the fifth
foot (and quite probably on the other feet, too) belonged to metrical conventions
specific to dactylic poetry. In this way the apparentdisproportion of the 94 instances
of -V'#C-, vs. the mere three cases of -u'#st, is extenuated, perhaps eliminated
altogether. According to Proskauer (15, 38), -s was re-introduced to the final
position about 200 B.C., in connection with the change of o to u in final syllables (cf.
Juret 1913:92, Hamp 1959:170-71). The usage of Ennius (and even that of Plautus)
differed already from colloquial reality (later restored by the 'poetae noui'; cf.
Proskauer, 39, and Hamp, 171). Thus, in Ennius, some cases of suppression belong
to the general poetic usage, which was in the process of taking on an archaic
flavor, while part belong to the specific conventions of dactylic poetry.
All these considerations strongly suggest thot the -ust type must be interpretedas
a sub-case of the suppression of final s in Latin between 600 B.C. and 200 B.C. This
association enables us to infer further hypotheses which, if confirmed, accumulate
as supporting evidence for the correctness of this assumption. For example, we
could expect that the -ust type has a ' (phono-)stylistic profile' similar to that of the

WHERE DOES LATIN SUM COME FROM?

49

suppression of final s. Dressier (1973:132) suggests that, in the more familiar and
negligent style, suppression is less restricted. If this is true, we ought to expect
similar behavior of the -ust type.
The above discussion amounts to establishing the -ust type as a sub-case of the
general tendency of final s to be suppressed under certain conditions. If this
generalization is valid, then we must conclude that, at a certain period of time, the
Latin paradigm was structured or 'memorized' as represented in 9.
4. DERIVATION OF THE LATIN PARADIGM FROM THE IE MODEL. The IE model

paradigm, as presented in 2, results from the application of the comparative method


to the IE daughter languages, especially Sanskrit and Greek. It contains synchronic
variation that may, in principle, be sociolinguistically interpreted.17Our starting
point will be a paradigm which represents one form of linguistic continuity derivable from it. Anticipatorily, and somewhat sloppily, we could call it the Latinizing
continuity:
(16) *esmi, *essi, *esti, *smos, *ste(s), *sonti

This paradigm is one surface structurewhich can be derived from the variation in 2.
It is a piece of sociolinguistic reality, since it reflects a certain dialectal choice. Our
proto-paradigm 16 ties up with linguistic continuity by representing one split-off
point of further linguistic evolution toward Latin (i.e. one node in the genealogical
tree).
We may envision (with Szemerenyi) that the first split-off relevant to our discussion concernedlpl. *smos: a variant *somos was created as in Figure 1.
*smos > *smos

*somos

or
*smos
- *SOios
*smos
FIGURE1.

The process may be described in terms of the practical syllogism model proposed
by Itkonen (1974:298):
(17) A intends to bring about
p.
A considers that he cannot bring about p unless he does a.
Therefore A sets himself to do a.
When we apply this teleological model to our data, we get the following picture:
(18) Some speakers intended to bring about
p.
They considered that they could not bring aboutppunless they created and
used *somos.
Therefore they created *somos.

Creating *somos was an intentional act motivated by p. That this innovation


received social acceptance and brought about the sociolinguistic variation *smos ^
*somos (tradition vs. innovation), shows that it was somehow called for. This
'7 By yielding a diasystem of surface structures (see Anttila 1972:275), the comparative
method gives us exemplifications of sociolinguistic reality.

50

LANGUAGE, VOLUME 53, NUMBER 1 (1977)

appears also in the fact that the variant *smos was ousted by 'natural selection'.
Now what was it the speakers intended to bring about? In other words: What
was the content of p? We are already familiar with Szemerenyi's explanation,
according to which initial sm began to create difficulties by threatening to destroy
the transparent constituent structure of the paradigm (this happened in Avestan,
which has mahi for Skt. smah). Szemerenyi suggests that the perilous situation was
healed by inserting o between s and m. The 'anaptyctic' o vowel has been interpreted
by many scholars (e.g. Safarewicz 1953:243, Ernout 1953:176, Leumann 1963:310)
as the theme vowel-which, at this time, was o before nasals. Thus the form *somos
can be regarded as a partial thematization of the copula, in order to maintain the
synchronic recoverability of its constituents.
In addition, Isg. *esmi was thematized, but we are unsure of the exact chronology. There are at least three possibilities; but before reviewing them, some
points of relative (as well as absolute) chronology must be made. First, apocope of
i (e.g. *esti > est) antedated rhotacism. This may be concluded from the 2sg.
ending -is (e.g. leg-is) < *-es < *-esi. Had the final i not been apocopated before
rhotacism, the process would have looked like this: *legesi > *legeri > *leger (cf.
Goetze 1923:88-9). Moreover, in the inscription of Duenos, which dates from a
period before rhotacism,18the apocopation of i is already a fait accompli (cf. mitat
and Kent's remarks on it, 1926:211-12). Second, rhotacism took place approximately in the middle of the 4th century B.C. (for a competent discussion of rhotacism, see Safarewicz 1932). Third, rhotacism antedated the fading of s (or z) with
compensatory lengthening of the vowel before a voiced consonant. For example,
the process dusmo (= dumoso; Livius Andronicus fr. 31 Mor.) > dumo postdates
rhotacism (cf. Goetze, 114).
Now we are ready to compare three conceivable thematization processes of
*esmi. Let us first consider the chronology implied in Szemerenyi'sview that 'when
*esmi (or*ezmi) began to create difficulties, it followed the model of the existing
*somos' (1964:193). If this is true, the thematization did not take place until after
rhotacism. Let us also assume that the final i in *esmi had been apocopated before
rhotacism, together with other verbs such as *legesi > legis (although, as has been
pointed out to me by Szemerenyi, this does not need to be the case). Because of
functional reasons, disyllabic forms obviously tended to resist apocopation (cf.
ante, mare etc.; but also *essi, *esti > ess, est). The process may be visualized as
in Figure 2.
Time

ca. 350 B.C.

-*

*esmi

( J

*ezm

( )

*esom

FIGURE2. Visualization of the possibility that *esom postdates rhotacism.

It is, however, hard to believe that an apocopated variant of *esmi, viz. *esm
(or
*ezm), would have survived and superseded the non-apocopated variant-which,
to be sure, was easier to pronounce. It is improbable that even one generation of
18 According to Kent (1926:222), 'slightly

before 350 B.C., perhaps even earlier'. Goldmann


(32) posits the date somewhat earlier, viz. 450-400 B.C.

51

WHERE DOES LATIN SUM COME FROM?

people would have tolerated a form like *ezm with its heavy final cluster. Therefore
it seems more plausible to think that, after the apocopation of i, the resulting *ezm
was somehow immediately remodeled into *som, without the intervening *esom.
This may have happened during the period of rhotacism. According to this line of
thought, *esmi was remodeled to *som because rhotacism would have produced a
variant such as *erom. This process is visualized in Figure 3.
Time -

ca. 350 B.C.

*esmi

FIGURE

*ezm

*som
(*erom...0)

3. Remodeling of *esm(i) into *som.

A third possibility would be to think that *esmi was thematized or 'regularized'


before the apocopation of i. This assumption puts the chronology somewhere between the 6th and 5th centuries, as visualized in Figure 4.
Time ->
*esmi

( )

*esomi
^

FIGURE

ca. 350 B.C.

5th cent.
(

*esom
J

*som

~~~~~~
(*erom..0)

4. Visualizationof the possibilitythat *esmiwas thematizedbefore the apocopation of i.

According to this view, the variant *esom had been ousted by 350 B.C., either
antedating rhotacism or coinciding with it. The latter alternative would explain the
remodeling *esom > *som: The traditional form *esom was rhotacized into *erom,
which fell outside the paradigm. Therefore *esom was re-interpretedon the basis of
*somos. The result was the variation (*esom >) *som
*erom, whose latter
member did not survive.
It is not necessary to make an absolute decision as to which of the three possibilities in Figs. 2-4 is closest to the truth. In the present study we are interested
of change, not in its exact chronology; and the mechprimarily in the MECHANISM
anism is reasonably similar in all three cases. (As far as chronology is concerned,
I hope to have elaborated the matter to such an extent that a detailed chronological
study can be attempted.)
Let us adopt the third possibility (Fig. 4) for our purposes.'9 According to it,
*esmi was thematized to *esomi, either simultaneously with the thematization of
*smos to *somos, or later-perhaps in connection with the apocopation of final i.
If we assume the latter possibility, the thematization can be motivated in a conceivable way. The process may, again, be described by means of a practical
syllogism:
19Szemerenyinow has somewhat modifiedhis position. He suggests (p.c.) a twofold remedy
for the endangered *smos. The speakers could either create *somos after *sonti (cf. *legomos:
*legonti), or *esmos after *estes (which was created to achieve parisyllabism with *sonti). He
assumes that both forms existed side by side for a time; during that period *esmi acquiredthe
by-form *somi (cf. *esmos: *somos = *esmi: *somi). Later the forms with es- were dropped in

the first person forms.

52

LANGUAGE, VOLUME 53, NUMBER 1 (1977)

(19) The speakers intended to avoid the sequence *ezm which (would have)
resulted from the apocopation of final i, because final zm was awkward
to pronounce.
The speakers considered that they could not carry out this intention
unless they manipulated the heavy cluster by 'thematizing' *ezm.20
Therefore they set themselves to create *esom on the model of *somos.
The apocopation, which took place somewhere between 700 and 450 B.c.,21
produced both *ezm and the therapeutic innovation *esom. It may be noted that
this corresponds also to the Varronian esum (Ling. 9, 100)-which, to be sure, has
usually been regarded as suspicious.
We are now at the point where the present indicative paradigm of the copula was
of the following shape:
(20) *esom, ess, est, *somos, *stes, *sont
This state of affairs prevailed until the rhotacized and non-rhotacized variants
began to co-exist. At this transitory period-during which, if we adopt the chronology of Kent 1926, the inscription of Duenos was written-the traditional form
*esom tended to be replaced by the new *erom. The latter form represented the
regular development, although it caused irregularity within the present indicative
paradigm-a situation known as Sturtevant'sparadox. *Esom was doomed to give
way to *erom; but this innovation was too extraordinary to be adopted into the
paradigm, the defining characteristic of which was the presence of s. A visualization is given in Figure 5.
DIFFERENT

om
s

s
s

omos
tes

ont

SAME
FIGURE 5.

(The visualizationis adopted from Anttila 1975:11-12.)

The paradox was resolved by creating the new variant *som. It would be tempting
to say that, once again, *som was modeled on *somos; but this would be somewhat
inexact. It seems not implausible to think that *som resulted from a flash of
20
The 'thematization' of the lsg. form was, strictly speaking, a secondary or non-proper
thematizationon the model of somos, which representsa true thematization.The properform
would have been *es6, but it was already in use (cf. Szemerenyi1964:192, n. 1).
21

The Carmen Saliare seems to attest the non-apocopated

desinence -nti (in tremonti for

tremunt).According to the tradition,the badly transmittedCarmenSaliaredates from the time


of Numa Pompilius, the second king of Rome (715-673 B.C.). The inscriptionof Duenos already
has apocopated forms.

WHERE DOES LATIN SUM COME FROM?

53

abduction (cf. Andersen), guided and aided by the existence of *somos. At some
phonostylistic level at least, *ne esom 'I am not' and *ego esom 'I am' were of the
shape *nesom and *egosom (cf. Szemerenyi 1964:195). Now we can see that *som
resulted from the only reasonable way of analysing the constituent structure of
expressions of this sort:
(21) *ne-som, *ego-som
This abductive segmentation resulted in a new piece of knowledge, statable roughly
as follows:
(22) The Isg. form is som, not esom or erom.
A new intuition was created. When this 'knowledge' was applied in practice, the
form *som was generalized to other contexts as well; e.g., it could be used in the
beginning of a clause. This is (in Andersen's terms) the deductive stage. In this way
the innovation was submitted to social control; and, as can be seen in Lat. sum,
it passed this inductive stage. The form *som was not sensitive to rhotacism, even
when placed after a word ending in a vowel (e.g. ego-som), because the word
boundary suspended the rhotacizing effect. Thus s could continuously serve as the
index of the paradigm (cf. Fig. 5). Besides, *som was consistent with *somos.
Now we have reached the end of the current Handbuch information (except
for some refinements of detail such as *som > sum, *stes > estis). However,
as appears from the Latin data elaborated above, it was not only the Isg. form
that was remodeled. THE ENTIRESINGULARPARADIGM
WAS RE-INTERPRETED.
The

result of the re-interpretationcan be conceived as a morphological rule:


(23) Let the present-tense morpheme of the copula be /s/.
This rule can be thought of as generating the following underlying system:
(24) /s-om/

/s-omos/

/s-s/

/s-tes/

/s-t/

/s-ont/

The character of the remodeling process can be representeddiagrammaticallyas in


Figure 6.
INDICATIVE

SG.

SUBJUNCTIVE

eS70
-0
r >

PL.

s-

S-

FIGURE 6.

The type of change was, of course, paradigm leveling, and one important aspect
of its inner motivation is visualized in the diagram: elimination of purposeless
variety (note that number was signaled by personal endings anyhow), or the old
principle 'One meaning-one form' (cf. Anttila 1972, passim). This principle

54

LANGUAGE, VOLUME 53, NUMBER 1 (1977)

covers both Kiparsky's 'paradigm coherence' (i.e., 'allomorphy tends to be minimized in a paradigm' 1971:598-9) and the 'Existenz gleich strukturierter Paradigmen ohne Alternation' mentioned by Schindler (1974:4)-and, in the case of
Latin, emphasized by Nyman (1974:20-21). The possibility of renderingthe presentttense system of the copula more iconic, by eliminating purposeless variation,22was
a factor that guided the abductive association as well as contributed to the social
acceptance of the development diagrammed in Fig. 6.
That the 'knowledge' of the forms ss and st had really been abduced by speakers
appears from their deductive reflexes in historical Latin(tuss, Plaut. Mil. 615;
mihzst, Plaut. Cist. 120; etc.; cf. also the -ust cases, and even the admittedly uncertain reading hic'st in Plaut. Poen. 1333A, see Lindsay 1922:76), which have
usually been interpreted as instances of aphaeresis. However, by asserting the
reality of the underlying representations in 24, I do not imply that the variants ess
and est disappeared altogether. What I am claiming is that their status was reinterpreted on the systematic level. The knowledge of ss and st was readily applicable when these forms were to be placed after words ending in a vowel; but as soon
as they were to be placed after words ending in a consonant, the result would have
been unsatisfactory. We can say that forms such as *mulierss (for mulier ess),
*nomenst (for nomen est), etc., did not pass the social control (cf. also the abovementioned instance of hic'st). After words ending in a consonant, the full forms ess
and est continued to be the norm. In order to adhere to the norm, speakers abduced
an adaptive morphophonemic rule of prothesis (9 above), which adjusted their
pronunciation to the received norm. Thus, e.g., mulierest was derivedsynchronically
as follows:
(25) /mulier#st/
(Underlying form)
muler#est

(Rule 9)

The model of the prothetic e came, of course, from the traditional ess and estwhich perhaps did not disappear from all styles. We can imagine that ess and est
were borrowed back from some sociolinguistically higher or more conservative
style(s). The important matter is, however, that they were re-interpretedas resulting
from the application of the adaptive prothesis rule.
5. THE PLAUTINE SYSTEM. We have now finished the characterization of what
might be called the pre-Plautine system of the copula. According to the foregoing
Note that the same tendency to paradigmaticleveling is to be seen, somewhat later, in the
presentsubjunctiveparadigmof the copula. At the time of the inscriptionof Duenos the system
was as follows:
22

s-ie-m
s-ie-s

s5-i-mos
s-i-tes

s-ie-d
s-i-ent
Plautus belongs to the period of transition:
(s-ie-m
(s-ie-s
(s-ie-t

')
%)
.)

s-i-m
s-i-s
s-i-t

s-i-mus
s-i-dis
s-i-nt ( . s-i-ent)

The allomorph -i-, which originally signalled plurality, has here to a great extent ousted -ie-,
the use of which in Plautusis confined almost entirelyto the verse-end(cf. Hodgman 1907:108).
In classical Latin the remodeling process can be seen as completed.

WHERE DOES LATIN SUM COME FROM?

55

discussion, the surface forms


(26) *som, ss

ess, st

est, *somos, *stes, *sont

were derived from the underlying morphophonemic system (24) by the application
of the 'adaptive' prothesis rule. In this section, we shall discuss the system of
Plautus. On the surface, it can be represented as follows:
(27) sum, ess

ss, est

'

st, sumus, estis, sunt

This representation displays so much difference from the pre-Plautine one that it
cannot be synchronically derived from the same underlying system. When comparing these paradigms, we can sense both the operation of sound laws (e.g.
*som > sum) and the re-organization of the morphological system (*stes > estis).
The problem of the internal constituent structure of the paradigm forms has been
kept to the background in this study. This does not, however, imply that speakers
were unable to perceive a similarity between the lexical items listed, or were unable
to 'figure out' the structuring principles of the paradigm. On the contrary, it has
been shown that the remodeling of *smos into *somos, and of *esmi into *som,
presupposes both awareness of the manner in which constituents go together and
intention to maintain the transparencyof the constituent structure. Fig. 5 diagrams
this iconic tendency from one point of view, rule 23, from another. Note that the
existence of rules governing the morphological organization of the present paradigm does not entail non-existence of paradigm representation of words, particularly those of frequent use (cf. Kehoe & Whitaker).
In what follows we shall trace the remodeling process resulting in the Plautine
paradigm. Two obviously 'analogical' processes are observable, viz. (*somos >)
*somus > sumus and (*stes >) *stis > estis,23 and consideration of them helps us

to recover some details of the synchronic principles of morphological organization


(cf. Kiparsky 1972). Displaying these principles both describes the mechanism of
the remodeling process and serves as a synchronic description of the Plautine
system.
5.1. THE FIRSTPERSONPLURAL.During the 3rd century B.C., a short 6 changed
into ii in a closed final syllable (cf. Sommer 1914:141). We can imagine that this
change affected the paradigm of sum as follows:
(28) *som > sum
(e)ss
(e)st
*somos > somus
*(e)stis
> sunt
*sont

Obviously the change *somus > sumus was analogical, not sentence-phonetic (as
suggested by Leumann, 310); but I think we should realize, in the spirit of Kiparsky
(1972:280), that what is implied in this change is not just an occasional pairwise
association ('unter dem Einfluss von sum', Sommer 1914:528; 'after 3rd pl.', Kent
1946:106), but rather a 'deeper' reformulation of a morphological spell-out rule.
23 For ease of presentation, we here choose the relative chronology
*stes > *stis > estis.
The real order of the process has no bearing on the present discussion.

56

LANGUAGE, VOLUME 53, NUMBER 1 (1977)

The 'punctum saliens' is that it was the 'paradigmatic' alternation u o which was
felt to render the constituent structure opaque-not, e.g., the m n alternation,
which was equally real at face value; see Figure 7.
The change proceeded
(A) from

to

not (B) from

sumsuum
sum
*s o mus
s u mus
s u nt
s u nt

c
sum

*so m us
su n t

to

d
sum
*so m us
*su m (p)t

FIGURE7.

From this we can see why o in *somus was changed to u. As it was 'spelled out'
by the same rule which, after the above-mentioned change o > u, spelled out the u
in sum and sUnt,24the u was also generalizedto the Ipl. form, for reasons too obvious
to be specified here. Seen in this light, the analogical change *somus > sumusshows
the reality of the rule formulated by Foley (1965:61).25
(29) s + [+nasal]->. s + u + [+nasal]
As implied in the above discussion, the associative track, the existence of which was
indicated by the analogical change *somus > sumus, can be represented in two
ways: first, as in Fig. 7; second, as in 29. Obviously the change in question was
first implemented as the variation *somus sumus. The latter variant, which
representsthe innovative usage of some speakerswho had abduced rule 29, was then
generalized to the whole linguistic community. We could speculate furthermorethat
generalizing u was not the only remedy (though it was the successful one) for the
somewhat annoying situation illustrated in 28 and in Fig. 7(A). One could imagine
that some speakers tended to generalize the o from *somus to the lsg. and 3pl.
forms as well, thus excluding them from the operation of the sound law which
changed o to u in a closed final syllable. The isolated form sont in CIL 12, 1529,
row 3, may be cited as evidence for this kind of reversedchange (i.e. *sont > sunt >
sont), because at the date when this inscription was written (somewhere between
134 and 90 B.C.), the change o> u was already complete (r. 1: Betilienus;6: campum;
9: balinearium;etc.; cf. esp. r. 7: ludunt,which shows regularvocalism in the 3pl.
ending). Perhaps another bit of evidence may be seen in the form so 'I am' in a
vulgar inscription (CIL 10,2070 [3rdcent. A.D. ?]) Obviously the graphemicsequence
so was phonetically [so], which presupposes a morphophonemic /som/.26
24
That the 3pl. ending is to be analysedas -nt is evidentfrom the change sient /s-i-ent/ > sint
/s-i-nt/ (cf. Sommer 1914:529, and n. 22).
25
Foley is at fault (1) for not backingup his analysis of the presentindicativeparadigmwith
philological evidence which could have been drawn from early drama (see Nyman 1974); and
(2) for strivingto make do with only one underlyingformative/s-/ in the whole presentsystema strategywhich led him to invent some unjustifiedsynchronicprocesses. Foley's critics have,
however, ignored the stimulative element in his approach: presentation of an alternative
synchronic analysis, which works quite well as far as the present indicative paradigm is
concerned.
26
Hardly /sum/, unless we assume that it was realized with a nasalized vowel, [su] or [so],
and that so was an attempt to expressthis pronunciationin writing. For a similarsuggestionin
regard to archaic inscriptions,cf. Safarewicz 1934 (other explanationsin Sommer 1914:528)

WHERE DOES LATIN SUM COME FROM?

57

5.2. THE SECONDPERSONPLURAL.The change *stis > estis was also analogical.

The model came from the sg. forms ess and est (e.g. Leumann, 310). However, what
is neglected by Latin scholars is the fact that this change took place at a period
during which ess and est had the 'reduced' variants ss and st, respectively, and these
could not serve as models. Moreover, in the pre-Plautine system sketched in ?4, ss
and st were the 'normal' variants; whereas, from the logical point of view, ess and
est had only a derivative existence. How, then, could they serve as models ?
Now it is clearly the case that our method, which looks upon language as a static
ERGON,is preventing us from capturing the reality.27Whatever has been uttered
existence is secured only by (successful) use, and
exists in a sense; but LINGUISTIC
it is the very use which may change functional relations between linguistic elements,
reverse markedness values of allomorphs, provide them with new semiotic functions, etc. In the present case, ess and est acquired their linguistic existence through
continued use. They were listed in the lexicon as equipollent allomorphs of ss and
st. This implies that the prothesis rule became superfluous; thus it hardly belongs to
the synchronic system of Plautus. From the point of view of synchronic description,
the two-paradigm approach (8) comes quite close to the proper representation;
but note that the creation and existence of estis (as well as the surprisingly rapid
disappearance of *stis) presupposes that the 'full' forms had acquired the status of
dominant, normal variants. This situation may be formally expressed by means of
two partially overlapping morphological spell-out rules-viz., rule 23 defining the
present (indicative as well as subjunctive) as /s-/, and the following:
COPULA
(30)
E
es /___+ + [+coronal^
(3)[-perfective]e/
[-nasal
J
The transient character of the prothesis rule is emphasized by the fact that its
replacement by rule 30 was a grammar simplification: now rule 30 stated the
distribution of es- in the whole non-perfective system, while it had formerly been
constrained to apply only to the non-present sub-system (i.e. es-se-s, es-to etc.)
Rule 30 took precedence over rule 23, because the variants which had the vocalic
nucleus were more natural and thus more frequent.
It has been suggested above that the unexpected form estis (cf. Leumann, 310) is
an index of re-organization of the morphological system, whereby the variants ess
and est acquired dominant status in comparison to the 'reduced' variants ss and st.
Is it possible to sustain this claim by something more substantial than the fact that
the creation of estis seems to relate to it in a rather natural way ? I think so, though
the point can be only sketched here. Consider the following line in a metrical
inscription (CIL 12,882):
(31) Raptusqu(e)afatis conditushoc tumulost.
(Cf. also CIL 10,4427,2.) There is a differencein graphemicrepresentationaccording
to whether elision or 'aphaeresis' is involved. In the case of elision, the elidable
vowel is manifested in writing. This indicates that the vowel in question is present
on the systematic or morphophonemic level, although it is doomed to be reduced or
dropped altogether on the phonetic level. (In this respect it may be compared to
27
Cf. Coseriu's warnings (1974:11-12) against the fallacy 'transitus ab intellectu ad rem',
i.e. taking what are requirementsof method as belonging to the essence of language.

58

LANGUAGE, VOLUME 53, NUMBER 1 (1977)

the graphemic/morphophonemic representation urbs, pronounced [urps]). In the


case of 'aphaeresis', the absence of the stem vowel in the graphemic representation
(0)st indicates the absence of this vowel in the underlying representation. (For a
similar argumentin favor of the underlyingabsence of final s in archaicinscriptions,
see Hamp, 167-8.) This implies that the 3sg. form had two underlying representations, viz. /est/ and /st/. Consequently 'aphaeresis' has no linguistic basis in Latin.
According to the majority of metricians and grammarians, 'aphaeresis' takes
place whenever es(s) or est is placed after a word terminating in a vowel. This view
also precludes elision of the preceding vowel; e.g., vi(a) est, for viast, is judged
impossible. Expressed in the terminology used here, this implies both that the
'reduced' variant is chosen unexceptionally after a vowel, and that this should be
reflected in the orthography. Some scholars regardthe metricians' rule as too strict.
Soubiran (1966:165-79) considers cases where elision, and consequent use of the
'full' variants es(s) and est, seem to be preferable on syntactic/semantic grounds.
His point is that where the copula is separated from the preceding word by a
syntactic cut, elision is more likely (cf. also Marouzeau 1908). This is especially
clear in the case of sentence boundary; e.g., Plaut. Trin. 934 Arabiast in Ponto?
Est 'Arabia is in Pontus? It is.' But a minor constituent boundary may also condition elision (Soubiran 1966:178-9; Marouzeau, 295-8). These are, it is true, only
intuitive judgments. However, Soubiran is able (1966:177) to cite one indisputable
case from Lucilius (1238 Marx: 0 Publ(i), o gurges Gallon(i), es homomiser, inquit)
in which the 'full' variant es is placed after a vowel, and what takes place is elision.
On the grounds of the above discussion, I suggest that, in historical Latin, it was
always possible to spell es and est even after a vowel (while the 'reduced' variants
could not be used after a consonant); and whenever this was done, it was elision
that was supposed to take place. In the example cited from Lucilius, there is, to be
sure, a syntactic cut between Gallon! and es; but the expression mentules, i.e.
mentul(a) es, in a Pompeian graffito (CIL 4,8931), warrants the belief that the
reduced variants tended to be ousted from everyday speech altogether. This
development is particularly understandablein the case of the 2sg. form, where use
of the reduced form was a potential source of ambiguity (cf. Brinkmann, 13); but a
few statements by Roman grammarians seem to indicate that st also tended to be
ousted: Marius Sacerdos (3rd cent. A.D.) adduces licitum est as an example of a
phrase ending in a monosyllable (6,493,14 Keil); and Bede (673-ca. 735) in his De
arte metrica states that via est is scanned vi(a) est (7,274,1 Keil). If es(s) and est
were indeed the dominant variants in historical Latin,28 the creation of estis
becomes entirely understandable.
6. EPILOG.In spite of what I have said about the creation of estis, this development remains problematic. It is somewhat surprising that there seems to be no
trace of *stis in our texts (cf. Nyman 1974:28, where an attempt is forced to
motivate the non-occurrence of *stis on syntactic grounds). We cannot rule out the
28
Soubiranhas shown that, as a rule, 'Virgile place est apres consonne ou au debut du vers
lorsqu'il possede un sens fort; il place est apres finale elidable [i.e., he uses the variant st]
lorsqu'il n'est qu'une copule ou un outil grammatical'([1957] 1966:166).In the case of Vergil,
the variants est and st seem to have developed a semiotic function which influencesthe word
order.

WHERE DOES LATIN SUM COME FROM?

59

possibility that the root vocalism in estis represents the most archaic type, as
suggested by Watkins (32-4)-although the argument ultimately rests on the
controversial question of how much weight must be given to the Hittite evidence
(cf. Wyatt 1972:689 and Kurylowicz 1958, with subsequent discussion). The same is
implied by Schmalstieg (see fn. 3, above). Proto-paradigm 2, which serves as our
point of departure, represents the 'conservative' type of reconstruction; and one
must confess that reconstructing the IE model paradigm still involves problems of
principle. Consequently, our discussion is more an experiment than a final answer.
REFERENCES
H. 1973. Abductive and deductive change. Lg. 49.765-93.
ANDERSEN,
ANTTILA,R. A. 1972. An introduction to historical and comparative linguistics. New
York: Macmillan.
-- . 1975. The indexical element in morphology. (Innsbrucker Beitrage zur Sprachwissenschaft, Vortrage, 12.) Innsbruck: Institut fur Sprachwissenschaft.
BAILEY,C. 1947. Prolegomena to Titi Lucreti Cari De rerum natura libri sex, vol. I.
Oxford: Clarendon.
G. 1932. Lat. sum, es, est, etc. Bulletin de la Societe de Linguistique de Paris
BONFANTE,
99.111-29.
0. 1906. De copulae ESTaphaeresi. Diss. Marburg.
BRINKMANN,
K. 1904. Kurze vergleichende Grammatik der indogermanischen Sprachen.
BRUGMANN,
Strassburg: Trubner.
L. 1974. Phonological features: problems and proposals. Lg. 50.52-65.
CAMPBELL,
P. 1961. Morphologie historique du grec. 2nd ed. Paris: Klincksieck.
CHANTRAINE,

E. 1974. Synchronie, Diachronie, und Geschichte. Transl. by H. Sohre.


COSERIU,
Munchen: Fink.
W. 1973. Pour une stylistique phonologique du latin: a propos des styles
DRESSLER,

negligents d'une langue morte. Bulletin de la Societe de Linguistiquede Paris


68.129-45.
ERNOUT,A. 1953. Morphologie historique du latin. 3rd ed. Paris: Klincksieck.
FISHER,I. 1948. Lat. connoscere. Bulletin Linguistique 16.156.
FOLEY,J. 1965. Prothesis in the Latin verb sum. Lg. 41.59-64.
GOETZE,A. 1923. Relative Chronologie von Lauterscheinungen im Italischen. Indogermanische Forschungen 41.78-149.
GOLDMANN,E. 1926. Die Duenos-Inschrift. (Indogermanische Bibliothek, 3:8).

Heidelberg:Winter.
HALL,R. A., JR. 1960. On realism in reconstruction. Lg. 36.203-6.

HALLE,M. 1961. On the role of simplicityin linguistic descriptions.Structureof language and its mathematicalaspects (Proceedingsof Symposiain Applied Mathematics, 12), ed. by Roman Jakobson, 89-94. Providence.
HAMP,E. 1959. Final -s in Latin. Classical Philology 54.165-72.
HAVET,L. 1884. sum, s, st. Memoires de la Societe de Linguistique de Paris 5.158.

A. W. 1907. Verb forms in Plautus. ClassicalQuarterly1.97-134.


HODGMAN,
ESA. 1974. Linguistics and metascience. (Studia philosophica Turkuensia, 2.)
ITKONEN,
Kokemaki: Societas Philosophica et Phaenomenologica Finlandiae.
J. R. 1973. The Germanic third weak class. Lg. 49.850-70.
JASANOFF,
JURET,C. 1913. Dominance et resistance dans la phonetique latine. Heidelberg: Winter.
1974. Lexical structure disruption in aphasia: a
KEHOE,W. J., and H. A. WHITAKER.
case study. Psycholinguistics and aphasia, ed. by H. Goodglass & S. Blumstein,
267-79. Baltimore: Johns Hopkins University Press.
KENT,R. G. 1926. The inscription of Duenos. Lg. 2.207-22.
-.
1946. The forms of Latin: a descriptive and historical morphology. Baltimore:
LSA.

60

LANGUAGE, VOLUME 53, NUMBER 1 (1977)

P. 1971. Historical linguistics. A survey of linguistic science, ed. by W. 0.


KIPARSKY,
Dingwall, 576-649. College Park: University of Maryland Linguistics Program.
-- . 1972. Review of Watkins 1969. Foundations of Language 9.277-86.
J. 1958. Le hittite. Proceedings of the 8th International Congress of
KURYLOWICZ,
Linguists, 216-43. Oslo: University Press.
LEO, FR. 1912. Plautinische Forschungen zur Kritik und Geschichte der Komodie.
2nd ed. Berlin: Weidmann.
LEUMANN,M. 1963. Lateinische Laut- und Formenlehre. (Handbuch der Altertumswissenschaft, 11:2.1.) Munchen: Beck.
LINDSAY, W. M. 1915. A short historical Latin grammar. 2nd ed. Oxford: Clarendon.
1922. Early Latin verse. Oxford: Clarendon.
---.
LINELL,P. 1974. Problems of psychological reality in generative phonology: a critical
assessment. Reports from Uppsala University, Department of Linguistics, 4.
J. 1908. Sur l'emploi de la graphie -st = est. Revue de Philologie 32.291-9.
MAROUZEAU,
NYMAN,M. A. 1974. Ubi est and ubist: the problem of Latin aphaeresis and the phonology of esse. (Annales Universitatis Turkuensis, B:128; Opera ex Institute Philologiae Classicae Universitatis Turkuensis edita, 3.) Turku: Turun Yliopisto.
1975. 1st der rest-Typus moglich? Arctos (Acta Philologica Fennica) 9.61-73.
--.
PERINI, G. B. 1974. Due problemi di fonetica latina. (Ricerche di storia della lingua
latina, 12.) Roma: Edizioni dell'Ateneo.
C. 1909. Das auslautende -s auf lateinischen Inschriften. Strassburg:
PROSKAUER,
Trubner.
ROTH,W. 1965. Beitrage zur Formenbildung von lat. 'esse' im Romanischen. (Romanische Versuche und Vorarbeiten, 17.) Bonn: Romanisches Seminar.
J. 1932. Le rhotacisme latin. Wilna.
SAFAREWICZ,
-.
1934. A propos de i'm final latin. Eos 35.133-8.
--.
1953. Zarys gramatyki historycznej jezyka lacifiskiego: Fonetyka historyczna;
fleksja. Warszawa: Panstw. Wydzial Nauk.
J. 1974. Fragen zum paradigmatischen Ausgleich. Die Sprache 20.1-9.
SCHINDLER,
W. R. 1972. Denominative verbs with the suffix -j-. La Linguistique
SCHMALSTIEG,
8.123-36.
SIEDOW,A. 1911. De elisionis, apharesis, hiatus usu in hexametris Latinis ab Ennio
usque ad Ovidii tempora. Diss. Greifswald.
F. 1901. Zur italischen Flexion des Ind. praes. von esse. Paper read at the '46.
SOMMER,
Versammlung deutscher Philologen und Schulmanner in Strassburg', Oct. 4.
[Summary in IF Anzeiger 12.347.]
- -.
1914. Handbuch der lateinischen Laut- und Formenlehre. Zweite und dritte
Auflage. Heidelberg: Winter.
J. 1957. L'apherese du est chez Virgile. Pallas 5.43-61.
SOUBIRAN,
--.
1966. L'elision dans la poesie latine. (Etudes et commentaires, 63.) Paris:
Klincksieck.
STRUNK,K. 1960. Die 3.pl. des Verbum substantivum im Griechischen. Glotta 38.203-9.
0. 1946. Latin sum. Egyetemes Philologiai Kozlony 69.82-91 (with an
SZEMERENYI,
English summary).
. 1964. Syncope in Greek and Indo-European and the nature of Indo-European
accent. Naples.
1970. Einfuhrung in die vergleichende Sprachwissenschaft. Darmstadt: Wissen--.
schaftliche Buchgesellschaft.
CHR. 1975. Review of Nyman 1974. Revue des Etudes Latines 52.436-7.
TOURATIER,
WATKINS,C. 1969. Formenlehre 1: Geschichte der Indogermanischen Verbalflexion.
(Indogermanische Grammatik, 3:1.) Heidelberg: Winter.
WIJK, N. VAN. 1905. Zur Konjugation des Verbum substantivum. Indogermanische
Forschungen 18.49-59.
WYATT,W. F. 1972. Review of Watkins 1969. Lg. 48.687-95.
[Received 11 March 1976.]

You might also like