Content uploaded by Catherine Atherton
Author content
All content in this area was uploaded by Catherine Atherton on Dec 29, 2019
Content may be subject to copyright.
david blank and catherine atherton
12 The Stoic Contribution to
Traditional Grammar
The phrase ‘traditional grammar’ refers to the body of knowledge
about the correct use of word-forms and syntax transmitted in the
West at least since the early Middle Ages for the study of Latin and
Greek and whose categories were used as a template for the study of
other languages. It has long been recognised that traditional gram-
mar shares numerous terms and concepts with the linguistic stud-
ies of the Stoics, and this chapter examines the relations between
them.1
1. sources
As is so often the case with Hellenistic philosophy, the dearth of
reliable, high-quality, first-hand material is a serious obstacle to re-
constructing Stoic thought in this area. No Stoic grammatical trea-
tise of any period survives; indeed, only one text with what can be
called, broadly, grammatical interests is extant in even something
like its original form; in any case this book of Chrysippus’ Logical
Questions (PHerc 307) belongs rather in what moderns would call
philosophical logic and the philosophy of language – although this
overlap is significant in itself (see Section 2.2).
What is known of the general character and organisation of Stoic
grammatical work and of its position within Stoic philosophy comes,
first, from doxographical sources, the most important being the
lengthy report of Stoic dialectic in Diogenes Laertius (DL VII 41–83)
and the bibliography of Chrysippus’ works on linguistic topics, again
in Diogenes (DL VII 190–2). Such sources, typically, can boast only
1On the development of grammar, see, e.g., Pinborg (1975) and D. J. Taylor (1986).
310
The Stoic Contribution to Traditional Grammar 311
a limited understanding of the material they are handling, which
will itself be of widely variable quality; fortunately, the relevant
portion of Diogenes’ report seems to go back to Chrysippus’ pupil
Diogenes of Babylon (second century b.c.). A major source for the
Stoic-influenced grammar of Tauriscus, a pupil of Crates of Mallos
(mid-second century b.c.), is a Pyrrhonist (sceptical) philosopher,
Sextus Empiricus, who probably wrote in the late second century
a.d. (MI).2Information otherwise emerges indirectly from whatever
particular theses survive in or are reported and typically criticised
by non-Stoic grammarians and philosophers (see Section 2.2). The
details of Stoic grammatical theories have likewise to be extracted
from later authors, such as Apollonius Dyscolus (mid-second century
a.d.) or the Byzantine commentators on the grammatical Technˆ
eat-
tributed, almost certainly wrongly, to Dionysius Thrax (c.100 b.c.);
these writers are bound by different allegiances and have different
aims than those of the theorists whose views they are reporting.
An additional complication is a general lack of information about
grammar outside the Stoa in the crucial period: third through the
first century b.c. No grammatical text survives complete before the
works of Apollonius Dyscolus; the earliest school grammars or gram-
matical handbooks are Roman. Further, evidence exists to show that
the earliest stages of grammar as something like a distinct discipline
were marked by deep disagreements about the scope, contents, and
internal structure of grammar.3The later, extant, texts are marked
by comparable, if less violent, variations. Peripatetic texts of the
fourth century b.c. and later, such as Theophrastus’ On diction (or
On style,Peri lexe ˆ
os), which may well have influenced grammatical
theorising, have likewise disappeared (although of course Aristotle’s
Poetics and Rhetoric have survived). The consequence is that distinc-
tions between and comparisons of Stoic and non-Stoic contributions
to traditional grammar are at best a hazardous undertaking, whether
in terms of organisation and general content (see Section 2.2)orof
particular theses (see Section 4).
It cannot even be assumed that classical antiquity produced an in-
dependent discipline of ‘grammar’ akin to modern linguistics. Two
important differences between modern and ancient grammatical
2See the translation and commentary in Blank (1998).
3See, e.g., Blank (2000).
312 david blank and catherine atherton
studies must be noted. First, studies we would call ‘grammatical’
were from the start housed within or were seen as contributing to
a variety of disciplines, themselves often owing allegiance to phi-
losophy, broadly construed: poetics, rhetoric, stylistics, literary and
textual criticism, and dialectic or logic. A broad intra-disciplinary
distinction was eventually arrived at in the Hellenistic period (cf.
e.g., Varro On the Latin Language VIII 6,S.E.MI57f., 91ff., 248ff.,
DT Sch. 135.7ff.) between ‘parts’, ‘tools’, or ‘offices’ of ‘grammar’
dealing on a mostly ad hoc unsystematic basis with the philolog-
ical and the exegetical-cum-literary-critical study of texts, on the
one hand (say, the authenticity of a text, or the explication of reli-
gious practices, or topographical features, mentioned in it), and, on
the other, ‘technical’ grammar, the division or function of grammar
which dealt systematically with the elements of spoken and written
language and their appropriate combination, and with the formal and
syntactic properties of the parts of speech. Second, technical gram-
mar was practiced for various purposes. On the level of research, it
was intended to explain various usages as deriving in a codifiable
way from original, correct norms, thus justifying the existence of
a scientific study of language and establishing a framework for the
use of grammar in the study of literature. In the schools, it was not
a descriptive discipline, but rather formulated a set of prescriptions
for writing and public speaking.
One piece of cargo traditionally carried by the historiography of
grammar must quite definitely be jettisoned, however: that a Stoic,
philosophical, anomalist grammar, the creation of a Pergamene
‘school’, waged war on an Alexandrian, philological, analogist gram-
mar. Further, the assumption that Stoic philosophy helped slow the
emergence of a ‘pure’ linguistic grammar from other disciplines rep-
resents a profound misunderstanding of the nature of ancient gram-
matical discourse (see Sections 2.2and 3).
Finally, on a general note; the notion of ‘influence’ is notoriously
slippery; thus, for example, failure of an accredited Stoic thesis to
survive does not show that such a thesis did not have a role in the
shaping of the orthodoxy, a role which may be exercised as much by
provoking counter-argument and counter-theorising as by earning
approval, whether universal or qualified (see Section 4). The follow-
ing survey, of necessity highly selective, focuses on topics in which
more definite influence, positive or negative, can be discerned.
The Stoic Contribution to Traditional Grammar 313
2. different perspectives
2.1 Ancient versus Modern
A first crucial distinction to be made in this area is between what
was important and influential work in antiquity, and what happened
to survive into medieval, early modern, and ultimately ‘traditional’
grammar. It is through the work of the greatest technical grammar-
ian of the ancient West, the Alexandrian Apollonius Dyscolus – who,
despite his implicit or explicit rejection of particular Stoic theses,
adapted the Stoic approach to syntax (see Section 4.1) – that Stoic
theorising in this area found its way into the medieval tradition. It
did so, however, not directly, but in modified form, and at two re-
moves: for Apollonius’ treatises were themselves filtered through
the work of his Latin acolyte, Priscian (sixth-century a.d.).4Another
important influence on the development of Latin grammar was the
Roman polymath Varro (first century b.c.). Fragments of his work
show a decidedly Stoic cast, for example in the division of differ-
ent types of pronominal expressions by the ‘definiteness’ of their
reference, or in his work on syntax (On the Latin Language Books
XIV–XXV), the sole fragment of which corresponds to the subject
matter of the Stoic ‘syntax of sayables’ in DL. In contrast, the most
copied and (to judge by the number of scholia on it) most used text in
technical grammar, the handbook attributed to Dionysius Thrax al-
ready mentioned (the authenticity of which was already questioned
in later antiquity; e.g., DT Sch. Vat. 160.24ff., and which is in large
part almost certainly a product of the period after the second century
a.d.) is clearly often incompatible with Stoic theorising. Indeed, that
what little is known of Dionysius’ own genuine work is strongly
marked by Stoic theory is an argument against the authenticity of
the Handbook.5Yet this does not mean that Stoics did not exercise
profound influence on technical grammar in its formative period (see
Section 4). Less clear is how far elements or fragments of Stoic theo-
rising became known in the most widely diffused manifestations of
4In the Byzantine East, a similar role was played by Georgios Choiroboskos (ninth
century a.d.) and other commentators on the supposititious ‘Technˆ
e’ of Dionysius
Thrax, as well as scholars such as Maximus Planudes (twelfth century a.d.).
5On Dionysius Thrax, see Linke (1977); on the (in-)authenticity of the treatise trans-
mitted under his name, see, e.g., Di Benedetto (1958–1959) and Lallot (1997).
314 david blank and catherine atherton
‘grammar’ in antiquity – first, the teaching to the very young of the
basic skills of reading and writing (grammatistikˆ
e) and second, in-
struction of (´
elite male) children approximately eight to twelve years
old in the more advanced arts of reciting and paraphrasing canonical
literary texts, and of basic composition as a preparation for rhetorical
performance proper.
2.2 Stoic versus Autonomous Grammar
Not only can nothing like a distinct, autonomous discipline we
might call ‘Stoic grammar’ be securely identified, but also theoris-
ing of the kind which would later be called ‘grammatical’ had its
home in at least two, and possibly four, different areas within Stoic
philosophy.
First, the ‘dialectical’ part of Stoic ‘logic’ analysed and described
the properties of logos – rationality, ratiocination, rational thought,
and discourse – relevant to the philosopher’s function as construc-
ter and arbiter of human reasoning activity, the perfecting of which
is integral to human happiness. It thus comprises (using the most
widely accepted division of ‘logic’) parts of psychology; epistemol-
ogy; the theory of the elements of speech and of the parts of the
sentence; stylistics, poetics, and music theory; the study of defini-
tions and of ambiguities; semantics and syntax; philosophical and
formal logic; and the theory of fallacies. The pioneering work of
M. Frede (1987a, b) has shown that whereas the parts of the sentence,
usually called the ‘parts of speech’ (mer ˆ
e tou logou), were classified
within the subsection of dialectic, which is concerned with ‘voice’
(ph ˆ
onˆ
e) or ‘signifiers’ (sˆ
emainonta), syntax was understood prima-
rily as a relation at the level of items ‘signified’ (sˆ
emainomena)by
(complexes of) the parts of speech, and hence was the subject of
a different subsection of dialectic.6Stoic metaphysics insists that
sounds and words are material objects (primarily, bits of air shaped by
speakers; secondarily, their written representations), whereas signi-
fications constitute one of four species of incorporeal: lekta, literally
‘sayables’ (or pragmata, ‘things (done)’, and ‘significations’, as they
are also labeled).7Sayables are defined as dependent (in a way that,
6On Stoic syntactic theory, see Egli (1986).
7See Ch. 8, Brunschwig, this volume.
The Stoic Contribution to Traditional Grammar 315
it must be admitted, is not entirely clear) on rational thoughts, but
seem to be the contents of those thoughts and of the sentences that
express them; they are thus shareable and communicable by mem-
bers of a linguistic community, in a way that private psychological
states are not.8
Sayables themselves are either complete or incomplete. Of the
former, the most important species is that of the prime bearers of
truth and falsity, ‘propositions’ or axi ˆ
omata. Surviving accounts ap-
proach syntax by identifying the proposition’s two principal con-
stituents: the ‘case’ or pt ˆ
osis (e.g., ‘man’, ‘Socrates’), and the ‘pred-
icate’ or katˆ
egorˆ
ema (e.g., ‘writes’, ‘sees’) which is defined as an
incomplete lekton. (Syntactic analyses of other complete sayables
have not survived.) A ‘non-compound’, verbal, predicate (e.g., ‘walks’)
or the verbal portion of a ‘compound’ predicate (e.g., ‘hits’) is what
is signified by a ‘verb’ (rh ˆ
ema), one of the five Stoic parts of speech.
The character and ontological status of the ‘case’, albeit clearly re-
lated in some way to the nominal parts of speech (i.e., the proper
name [onoma], the appellative [prosˆ
egoria], and the pronoun), is far
more problematic and is discussed in Section 4.3.
Crucially for Stoic dialectic and philosophy generally, the struc-
ture of complete sayables allows the articulating of distinctions –
above all that between a physical object and its properties – which
have no physical correlate (in that the properties are physical parts
of the object or are reducible to it without metaphysical remainder).
The forms of words in a sentence will provide information about
the contribution of their correlates at the level of the sayables to
the meaning(s) of the sentence as a whole. Further, complexes of
propositions go to make up arguments, which, when sound, articu-
late real (e.g., causal) connections obtaining in the physical world;
hence, the conjunctions that link propositions must in some way
have meanings that can also capture such real-world relations. The
non-accidental properties of sayables are presumably immune from
change, in a way that words and their properties are not; perfect iso-
morphism between the two levels is also compromised by solecism,
synonymy, and ambiguity. The dialectician’s interest in grammar is
stimulated by the need to minimise these defects in his own usage,
8On key concepts of Stoic linguistic theory, see M ¨
uller (1947), Long (1971a), M. Frede
(1974,1987a, b, 1994a, b), and Atherton (1993).
316 david blank and catherine atherton
and to detect and compensate for them in others’. Correct language
embodies as close as possible a match between the constituents and
structures of sentences, and those of the sayables which those sen-
tences express, and which constitute the contents of impressions.
But since proper control of assent to one’s impressions is essential
for building up a system of correct beliefs and for governing correct
behaviour, sensitivity to linguistic correctness will be a useful – and
may be an essential – asset for the Stoic philosopher.
A third home of ‘grammatical’ theorising within Stoic philosophy
may well have been the study, within the topic of ‘voice’ or linguis-
tic signifiers, of the ‘virtues and vices of speech’ (aretai kai kakiai
tˆ
es lexe ˆ
os).9Unusually, this portion of the Stoic theory seems to
have been adapted from Theophrastus, with the (characteristic) ad-
dition of a virtue of ‘conciseness’ (suntomia) to those of ‘purity’ (for
Greek: hellˆ
enismos; later, for Latin, latinitas), ‘clarity’ (saphˆ
eneia),
‘appropriateness’ (to prepon), and ‘elegance’ (kataskeuˆ
e). ‘Vices’ or
‘defects’ are or include barbarism and solecism. A solecism is de-
fined as a word-complex (logos) that is ‘incongruent’ (akatall ˆ
elos);
the key concepts of ‘congruence’ (to katall ˆ
elon) and its opposite we
will meet again in Section 4.1. Again, we have very little informa-
tion about which criteria of linguistic purity were endorsed by the
Stoa, but the standard association of Stoics with an ‘anomalistic’
grammar is incorrect (see Section 3). Further, although etymological
researches seem to have been of interest to the Stoa as part of a gen-
eral philosophical account of the origins of language, demonstrating
the natural suitability of signifiers to significations (cf. Augustine
de dialectica Ch. vi), the most that can be said is that this does not
preclude the application of these principles in prescriptive grammar,
or in a theory of tropes and metaphors as the prerogative of poetic
creativity working to extend the boundaries of language (Quintilian I
viii 14ff., VIII vi 34ff.; Servius Comm. in Don. 447.5ff.).
It has been claimed that Stoic grammatical theorising found its
way directly into Roman grammar, bypassing the Greek tradition
entirely (Barwick 1922). Not only, however, is the sharp distinction
on which this thesis is grounded – between a Stoic text or texts on the
virtue of Hellenism and a non-Stoic Grammatical handbook (Tˆ
echnˆ
e
9See Stroux (1912) and Barwick (1957), though the latter attributes too much to the
Stoics and too confidently.
The Stoic Contribution to Traditional Grammar 317
grammatikˆ
e), as the respective sources for Roman and Greek gram-
mar – in itself deeply implausible given even the little we know of the
fluidity and diversity of the Hellenistic tradition, but there is in fact
no hard-and-fast rule that Roman grammarians do and Greek gram-
marians do not append a study of the virtues and vices of speech to
the study of correctness. Furthermore, the thesis that this supposed
stylistic portion of Roman grammars replaced a treatment of syntax
(to the detriment of the later history of the subject) is at odds with
the evidence that syntax was indeed dealt with by Stoics, only (as
we have seen) not – or not primarily – in the context of a theory of
Hellenism or of the parts of speech.
Lastly, Stoic rhetorical theory, which constituted the other ma-
jor subdivision of Stoic ‘logic’, could have borrowed this stylistic
construct, and in some way adapted it to the demands of insti-
tutionalised, formal discourse. The convention that the stylistic
virtues apart from purity would be dealt with by rhetoricians (cf.
Ad Herennium IV 17; cf. Quintilian inst. rhet. Iix6) could have
been adopted by Stoics.10
Early Stoic interest in the criticism of literary texts, at least from
an ethical perspective, can be seen in the titles of some of Zeno’s (On
listening to poetry,Homeric problems, DL VII 4) and Chrysippus’
works (e.g., How to listen to poetry), and Stoics are traditionally
linked to the ‘allegorical’ interpretation of such texts, although here
again scholars disagree about the extent of their interest.11 A more
substantial literary grammar was associated with Crates of Mallos,
a pupil of Diogenes of Babylon, who is said to have introduced the
study of grammar (in the sense of textual and literary criticism) to
Rome (Suetonius On grammarians and rhetoricians 2). What is strik-
ing is that his discipline of ‘criticism’ (kritikˆ
e), which includes the
orthodox ‘grammarian’s’ expertise in such matters as accentuation
and the literary lexicon (S. E. MI79), but requires familiarity with
the whole of ‘logical knowledge’ (epistˆ
emˆ
e logikˆ
e), appears – at least
in theory – to be as dependent on (albeit not as embedded in) Stoic
dialectic as is the theorising on ‘grammatical’ topics evidenced by
10 On Stoic rhetoric, see Atherton (1988).
11 An idea of Chrysippus’ ethical approach to poetry may be gleaned from Plutarch’s
How to listen to poetry; the relation of this approach to grammarians’ claims about
the ethical value of poets may be seen in S. E., M I 270 ff. and Blank’s commentary
ad loc. On allegory, see Long (1992) and Most (1989).
318 david blank and catherine atherton
Diogenes Laertius’ report. Crates’ pupil Tauriscus divided ‘criticism’
into ‘a rational part concerned with diction and the grammatical fig-
ures’; an empirical part ‘about the dialects [of Greek] and the different
forms or types of style’; and a historical part ‘about the preexisting
unordered raw material’ (S.E. MI248f., tr. Blank). The original con-
tents and form of the (or a) Stoic Handbook on voice (Technˆ
e peri
ph ˆ
onˆ
es) have been the object of extensive study and considerable
scholarly disagreement;12 the basic text is again the summary of this
portion of Stoic dialectic in Diogenes Laertius, but possible indirect
clues are also found in the earliest structural divisions of grammar
which are reported to us, the ‘critical’ expertise of Crates of Mallos,
and the ‘expert part’ of the grammar of Asclepiades of Myrlea.13
To summarise: it remains unclear, and may perhaps always remain
unclear, what precisely found its way from which portion of Stoic
philosophy into which treatise or monographs of the Hellenistic
Stoic-influenced authors writing on grammatical topics, and un-
clear too what survived of either of these sources (Stoic or Stoic-
influenced) in the handbooks and monographs of professional gram-
marians. The fact of Stoic influence, however, cannot be doubted.
3. schools of grammar and the
anomalist/analogist controversy
An important theoretical distinction to be observed is that between
formal and semantic anomaly and analogy. Chrysippus is known
to have referred to cases of semantic anomaly; that is, of inconcin-
nity between form and signification: e.g., the inappropriate privative
form of the Greek adjective athanatos (deathless, immortal), as if
death were something of which the gods were deprived (Simplicius
In Ar. Cat. 396.3ff.); or the gendered forms of different names for the
moon deity, mˆ
en and selˆ
enˆ
e, when the one immanent deity is in fact
genderless (Philodemus Piet. 11,=SVF 2.1076). The general albeit
sometimes disappointed expectation is thus of similarity of some
sort between words and what they name (which supports the Stoic
project of semantic etymology). In contrast, the traditional associ-
ation of the Stoics with something called ‘anomalistic grammar’ is
both unfounded (there being no such thing as ‘anomalistic grammar’,
12 See Ax (1986)240–252; Schenkeveld (1990).
13 On the divisions of grammar by Tauriscus and Asclepiades, see Blank (2000).
The Stoic Contribution to Traditional Grammar 319
only different attitudes to the status of the principle of regularity in
language) and deeply misleading.
In the nineteenth century, scholars developed a picture of the early
development of grammar, including literary studies, centered on the
clash of two schools: the Alexandrian ‘grammarians’, starting with
Zenodotus and Aristophanes of Byzantium in the third century, and
reaching a zenith with Aristarchus of Samothrace and his pupils in
the second; and the Pergamene ‘critics’ founded by Crates of Mallos,
a contemporary of Aristarchus.14 This picture was based on the pre-
sentation by Varro, in his On the Latin Language VIII–X, of sets of
arguments for and against resolving questions of grammatical us-
age (e.g., which of two possible forms of the genitive of a particu-
lar noun should be used) by choosing forms that are ‘analogous’ –
that is, ‘similar’ – to those of words that are themselves arguably
‘similar’ in relevant ways to words whose forms are agreed upon,
as opposed to forms that are ‘dissimilar’ to such analogous forms
(‘anomalous’). Varro cites Crates as arguing against Aristarchus and
analogy (VIII 63–5,68–9;IX2) and as relying on Chrysippus, who (as
we saw previously) had spoken of ‘anomalies’ in the relation between
names and what they designate. Further, because Crates15 is known
to have studied with Chrysippus’ student Diogenes of Babylon, and
because he interpreted the Homeric poems in accordance with Stoic
physics, scholars assumed that Crates’ school had ‘anomalist’ doc-
trines which were derived from the Stoics.
This view cannot be adopted, however. Varro himself had already
pointed out (IX 3) that Aristarchus did not want to recommend the
use of ‘analogical’ forms which common usage would abhor. The
extreme ‘anomalist’ position, that there are no regularities in lan-
guage, would be equally implausible, especially for a Stoic, since
that school viewed language as a manifestation of the divine reason.
We do not know whether, besides the semantic anomalies noted by
Chrysippus, the Stoics actually spoke about formal irregularities as
well, but if they did, they can only have pointed to them against the
overwhelming background of the logical and regular nature of lan-
guage. The ‘analogy-anomaly controversy’ is rather part of a much
larger ancient dispute between rationalists and empiricists, in this
14 For details and bibliography, see Fehling (1956–1957) and Blank (1982,1994).
15 See the forthcoming edition of the fragments of Crates by M. Broggiato. On his
poetic theory, see Janko (2000).
320 david blank and catherine atherton
case over the character and use of linguistic rules: should one gov-
ern one’s usage with reference to rules purporting to derive from an
understanding of the nature of language, or rather by observing the
usage of others in relevant geographical, social, and technical con-
texts? Sextus Empiricus (MI) clearly argues for the latter, and in so
doing summarizes the ‘anomalist’ case.
As the Stoics and Crates will evidently have been on the ration-
alist side of this debate, there can be no question of opposing a Stoic or
‘philosophical’ Pergamene anomalist grammar to a ‘technical’ gram-
matical school, perhaps influenced rather by Peripatetics, in Alexan-
dria. Philologists, both Alexandrian and Pergamene, were interested
mostly in literary texts, the establishment of correct texts and their
interpretation. As a preliminary to such work, they had regard to
rules about linguistic usage and these were used, mainly by others,
for instruction in reading, writing, and rhetoric. In their work on
language, grammarians took over what they found useful from the
Peripatetics, and especially from the Stoics, applying it in their own
field and also modifying it as they saw fit. Accordingly, the assump-
tion of a divergence among philologists over the relation of philoso-
phy and grammar must not be allowed to obscure the indebtedness
of Alexandrian grammar and, later, of traditional grammar to Stoic
philosophy.
4. case studies
Four particular areas have been selected, on grounds of their central-
ity to grammatical theory, to illustrate the relation between Stoic
and traditional grammar.
4.1 Syntax
It has already been remarked how Apollonius Dyscolus adopted a ver-
sion of the Stoic conception of syntax, taking it to hold, not at the
level of words (i.e., word-forms as written or spoken entities), but
rather at the level of the ‘intelligibles’ (noˆ
eta)(Synt. I ii, p. 2.3ff.).16
16 Cf. Baratin and Desbordes (1981) and Blank (1982); on Apollonius, see also Sluiter
(1990) and Ildefonse (1997), along with the new edition, translation, and commen-
tary of Lallot (1997).
The Stoic Contribution to Traditional Grammar 321
These can be identified as what modern linguistics would call the
morphosyntactic properties of word-forms (ph ˆ
onai, lexeis), such as
case, number, gender, voice, and mood, and, where relevant, their se-
mantic and functional properties (e.g., the syntax of adverbs or cer-
tain kinds of verbs and pronouns). Further, syntactic relations are,
strictly speaking,always correct or ‘congruent’ (katall ˆ
ela), inconcin-
nities or solecisms occurring only at the level of expressions or word-
forms; what is more, they can be shown to be ‘abnormalities’ (path ˆ
e)
departing in a rule-governed and rationally explicable way from the
correctly formed logos isomorphic with the noetic structure.
It is highly plausible that the distinction between imperfect lin-
guistic forms and perfect intelligible structures is again borrowed
from Stoic linguistic theory, which assigned such semantic and struc-
tural defects of language as ambiguity and solecism to the level of ar-
ticulate utterance (lexis) or discourse (logos); in contrast, non-atomic
lekta such as propositions, questions, or commands are always per-
fectly formed constructs from the basic building blocks of predicates
and cases (and other elements too, such as conjunctions). The Stoic
definition of ‘solecism’ (see Section 2.2) suggests a rational, rule-
governed notion of syntax, distinguished by its internal ‘congruity’,
rather than appeal to usage or authority to explain and justify what
is taken to be syntactically correct and to outlaw what is not. But
it must be admitted that little is known of the general criteria of
Hellenism admitted by the Stoa (the grammatical orthodoxy estab-
lished four or five: analogy, which sometimes includes etymology,
history, usage, and authority). Apollonius apparently combined this
distinction with the method of ‘pathology’ (the identification and
classification of path ˆ
e) familiar from the older discipline of orthogra-
phy to construct his own characteristic procedure for demonstrating
the basic rationality of language. Other grammarians tended to take
the basic unit of grammatical analysis to be an item on the level of
expression, viz. the complete logos or sentence, consisting of parts of
speech (merˆ
e tou logou) or words (lexeis) (e.g., DT Sch. Mel. 57.12ff.),
but traces of the Stoic view can perhaps be found (e.g., DT Sch. Vat.
211.27ff., Marc. 354.7f.).
Another, less welcome, legacy of Stoic syntactic theory seems to
be the exclusive focus on syntax as grouping (in particular, the group-
ing of a case and a predicate to form a proposition). The limitations
of this conception of syntax can be seen, for example, in the two
322 david blank and catherine atherton
surviving Stoic classifications of ambiguity types (Galen De sophis-
matis Ch. 4Kalbfleisch, XIV 591.1ff. K ¨
uhn; Theon Prog. 81.30ff.).
Both contain a category for expressions whose ‘significant parts’
(roughly, parts of speech: see Section 4.2) can be variously grouped
to produce sentences with different meanings, but neither can easily
accommodate ambiguities where what is responsible for the multi-
plicity of possible meanings is the multiplicity of grammatical rela-
tions into which two (or more) parts can enter: their entering into
some sort of syntactic grouping is not at issue. Thus, in the Greek
sentence touto hor ˆ
ai, ‘this one sees’, what is unclear is (using the ter-
minology of traditional grammar) whether this one is the subject or
the object of the verb. Here a Stoic could identify this one either (1)as
a nominative case grouped with a ‘direct’ (orthˆ
e; roughly, transitive)
‘sub-predicate’ (literally ‘less-than-predicate’; i.e., the verbal portion
of the predicate lacking the oblique case required to make a predi-
cate), or (2) as an accusative case grouped with the same sub-predicate
to form a predicate, which in turn needs syntactic association with
a nominative to make a proposition. But although such analysis is
clearly possible in Stoic terms, what is striking is that no room is
made for its application in the lists of ambiguity kinds; indeed, the
notion of governance, which was to become so central to traditional
grammatical analysis, to all appearances has no part in Stoic gram-
matical theory, although a rough equivalent in logic (e.g., the relation
between a negative and a proposition) was well established.
This example does show, however, that the notions of transitivity
and intransitivity were, in some form, familiar to Stoic grammati-
cal theorists, who (to simplify the rather difficult sources) appear to
have distinguished between ‘direct’ (ortha =transitive), ‘reversed’
(=passive, huptia), and ‘neutral’ (oudetera =intransitive) predi-
cates or sub-predicates, as well as ‘reflexive’ (antipeponthota) ones
corresponding to what professional grammarians called ‘middle’
(mesa) verb forms (DL VII 64). The concept of the ‘passing over’
of the action signified by a verb to another ‘person’ in a syntac-
tic grouping (or to the same person in ‘middle’ constructions) was
to become a fixture in grammar, such verbs being typically desig-
nated ‘active’ (energˆ
etika, drastika; e.g., DT Sch. Lond. 548.34ff.),
although subject/object syntax was not an ancient ‘discovery’. It
has been argued that these distinctions among predicates originally
reflected the distinctions of Stoic aitiological theory; but evidence
The Stoic Contribution to Traditional Grammar 323
for systematic correlation is obscure and controversial – unfortu-
nately so, since clearer understanding of the philosophical underpin-
nings of Stoic work in the field would greatly assist accurate esti-
mation of what did and did not survive of it into grammar outside
Stoicism.
4.2 The Parts of Speech
According to our main sources, Stoics identified at first four, later
five, and still later six, parts of speech (mer ˆ
e tou logou), as opposed to
the later standard eight (e.g., DT 23.1f., DT Sch. 354.4f.). Surviving
definitions are strongly semantic and/or functional, but with some
formal elements. Of the single original category of noun, two types
were distinguished by Chrysippus, each signifying a different type of
quality (qualities being corporeals, according to Stoic metaphysics):
the proper name (onoma), signifying particular qualities, and the ap-
pellative (prosˆ
egoria), common ones. This distinction was rejected by
professional grammarians, who preferred one between the ‘general’
(koinon) and the ‘particular’ (idion) nominal (e.g., DT 23.2f., 24.4f.;
DT Sch. Mel. 58.21ff.). The verb (rhˆ
ema) was, by some, defined as
‘not inflecting for case’ (apt ˆ
ota) as well as signifying (parts of) pred-
icates. The conjunction (sundesmos) is also uninflected, and serves
to combine other parts of speech: a puzzling definition, since known
uses of conjunctions are to combine complete lekta (e.g., the conjunc-
tion if combines propositions to form a conditional proposition). Fur-
ther, it remains unclear whether conjunctions have their own proper
significations at the level of the lekton; are regarded as significant,
not in isolation, but rather because of the contribution they make
to the meanings of sentences; or are to be classed as functional, not
semantic, parts of speech. The article (arthron), which does inflect
for case, was subdivided into the ‘definite (hˆ
orismenon) article’ –
what came to be called the pronoun (ant ˆ
onumia) – and the ‘indefi-
nite (aoriston, aorist ˆ
odes) article’, which corresponds to the definite
article of traditional grammar (Apollonius Pr. 5.13ff., DT Sch. Marc.
p. 356.12f.). The last part of speech, the ‘mean’ (mesot ˆ
es) may be the
adverb (cf. DT 74.3; DT Sch. Heliod. 30.2ff., Vat. 75.5ff.), but other
sources report that participles (metochai) were classed with nomi-
nals or verbs, adverbs (epirrhˆ
emata) with verbs, and prepositions as
conjunctions.
324 david blank and catherine atherton
The puzzle is intensified by the report just alluded to that con-
junctions can combine complete lekta, as if conjunctions themselves
belong to that level, rather than to the level of the signifying logos.
They are as puzzling in this regard as are cases (see Section 4.3),
some of which are ‘demonstrative’ (deiktikai) (e.g., this [male]
one [houtos]), these apparently being identical with or signified by
demonstrative and (some) personal pronouns. The grammarians’ pro-
nouns are, roughly speaking, Stoic definite articles. In our principal
source here, Diogenes Laertius VII 58, definite articles are charac-
terised in purely formal and functional terms (as distinguishing num-
ber and gender of proper names), a definition which, as it stands, is
plainly unsatisfactory and, furthermore, is at odds with the Stoics’
having labeled the ‘definite’ article ‘indefinite’ because of its gener-
alising implication (e.g., ‘The walking person moves’ is equivalent
to ‘Whoever walks, moves’; cf. Apollonius Synt. I111,p.94.11–13. II
32,p.149.8ff.). Puzzlingly, this is to ignore or overlook the anaphoric
use of definite articles perfectly familiar to later grammarians.
Apollonius himself favours a semantic definition (a pronoun signi-
fies the being of something, without its qualities), which may have
Stoic origins.
A comparable disagreement divides Stoic and non-Stoic grammat-
ical theory over the proper noun (onoma) and the common noun or
appellative (prosˆ
egoria). Chrysippus seems to have been the first to
treat these as separate parts of speech (DL VII 57f.), primarily for se-
mantic reasons, although formal considerations were also appealed
to (DT Sch. Vat. 214.18ff.); such formal/semantic isomorphism will
have characterised language’s pristine state. The division was re-
jected by the grammatical tradition, which did, however, accept the
Stoics’ inclusion in a single category – that of the ‘appellative’ – of
what traditional grammar would come to call common nouns and
adjectives.
4.3 Cases and Morphosyntactic Properties
It is clear that Stoics, as compared with Aristotle (Poetics 1457a18ff.;
Rhetoric 1364b34f., Topics 114a34ff.), narrowed considerably the
scope of the technical term ‘case’ (pt ˆ
osis); as in non-Stoic grammar,
it is used only in connection with nominals. But beyond that ba-
sic starting point, scholarly disagreement about the Stoic view(s) is
The Stoic Contribution to Traditional Grammar 325
rife. An important preliminary is that Stoic metaphysics imposed a
sharp distinction between physical objects, which ‘exist’ (einai)on
the one hand, and incorporeals on the other, which merely ‘subsist’
(huphistanai); that is, depend for their being on corporeal individ-
uals (as lekta ‘subsist in accordance with’ rational impressions).17
Accordingly, a sharp cut-off point must hold between those mate-
rial objects which are words (and their written representations) and
the incorporeal meanings of sentences, and of at least some sentence
parts. However, and rather oddly, our sources do not always clearly
indicate the ontological status of certain key items, such as, and in
particular, the ‘cases’ which go to constitute (certain kinds of) predi-
cates as well as propositions and (presumably) other complete lekta.
Scholars are divided as to whether cases are (1) another kind of in-
complete lekton complementing the predicate (which is suggested
by their figuring in Diogenes Laertius’ report of the ‘significations’
part of Stoic dialectic, and by their combining with predicates [VII
64]); (2) the qualities which are the significations of proper names
and appellatives (whether these qualities should be classed as corpo-
real or incorporeal is the object of a further dispute); and (3) words –
that is, words in certain forms: just as, for Stoics, properties of bodies
are simply reducible to those bodies in certain states, so cases, being
forms and hence properties of words, will be reducible to those words,
which are physical objects formally differentiated according as their
role in a sentence changes. For the grammarians generally, no such
metaphysical constraints operate; and here a gap could well have
opened up between them and Stoic grammarians. Thus Apollonius
seems to have plumped for something like (3), to judge by his defini-
tion of the case, whereas a commentator on Dionysius Thrax states
that ‘the five cases belong to the meanings, not to the word-forms’,
citing in support the variety of the latter as opposed to the singleness
of the former (DT Sch. Lond. 551.10ff.).
Stoic grammar does not seem to have recognised a ‘vocative’
(klˆ
etikˆ
e) case (cf. DL VII 63 with, e.g., DT 31.5–7, DT Sch. Marc.
384.16ff.), even though Stoic dialectic identified a complete sayable
used for ‘addressing’ (prosagoreutikon) which could comprise cases
alone, without a predicate (DL VII 67). Stoic influence can, however,
be felt in the general acceptance of the thesis that the nominative
17 DL VII 63. See Ch. 8, Brunschwig, in this volume.
326 david blank and catherine atherton
case, called by Stoics the ‘direct’ (orth ˆ
e,DLVII64; so also DT 31.5)
and by other grammarians also the ‘straight’ (eutheia) or ‘nomina-
tive’ case is itself a case (ct. Aristotle Int. 16a31-b5). The motivation
reported for the Stoic position is confused and especially hard to
interpret (Ammonius Int. 42.30ff.; cf. DT Sch. Marc. 383.5ff.), but
may reflect what for moderns would be a distinction between the
pragmatic and the semantic: although a nominative is formally in-
distinguishable from its decontextualised nominal, its being used in
the context of a sentence by a particular speaker to refer intention-
ally to an object (to ‘the thought of it we have in the soul’, Amm.
Int. 43.10f.) sets it apart from the nominal itself, the enduring se-
mantic properties of which have other origins, perhaps in the larger
community of speakers. But this is largely speculative.
Interpretation of this and other Stoic texts is not served by what
look to be deficiencies in the Stoic semantic apparatus, which – to
judge by available evidence – may well have failed to distinguish sys-
tematically among modes of meaning, and in particular to separate
(features of) linguistic items having lexical meaning and referring
to corporeals or their properties or states on the one hand, from, on
the other, ones signifying non-corporeal items (lekta) or contributing
syntactically (as with conjunctions) or in virtue of their form (e.g.,
by indicating tense, mood, case) to the semantic content of an entire
expression. Most importantly for present purposes, no comprehen-
sive analysis seems to have been developed of the relation between
a word(-form), a case (if a case is not a word or word-form), or a
predicate, and their morphosyntactic properties (if any). As noted,
for the Stoic grammarian syntactic (in)congruity must be at least
to some extent a matter of (mis)matches of parts of sayables, but
it is unclear if the relevant properties – say, a singular third-person
predicate’s indicative mood and present tense; perhaps a case’s be-
ing singular and nominative – are formally distinguished at the level
of the lekton as they are at the linguistic level and, if not, whether
their various contributions to the semantic whole are conceived of
differently. Predicate classification certainly requires formal distinc-
tions between cases (DL VII 64: e.g., ‘direct’ sub-predicates require
an oblique case to form a complete predicate; passive predicates are
construed with the ‘passive particle’, i.e., the preposition ‘by’ [hupo]
and, presumably, the genitive case).
There is indeed some evidence that the concept of ‘implicature’
([par]emphasis) may have helped to distinguish lexical or referential
The Stoic Contribution to Traditional Grammar 327
from other sorts of meaning (as ‘signify’ [sˆ
emainein] and ‘denote’
[dˆ
eloun] were used of appellatives and proper nouns, respectively
[DL VII 58]). Thus, according to Chrysippus, ‘habitual privatives’
such as shoeless ‘signify some implicature’ (paremphasis) that the
thing lacked habitually belongs to the thing which lacks it (Simpli-
cius In Ar. cat. 395.12–14). Certain syllables constituting a word can
also (fail to) ‘implicate’ local separation (Galen PHP II ii, 15.11ff.,
especially 16.1). Chrysippus wrote a work On implicature, which is
classed with works On the grammatical cases and On enunciations
definite as regards the subject (DL VII 192); his Logical questions
uses emphainein for the ‘revealing’ of a sayable (col. XII 17; that the
usage is not exclusively technical is shown by XI 29 [27 Marrone]),
and also mentions ‘things said implicitly (paremphatik ˆ
os) in rela-
tion to something else’ (fr. iii 22f. [21 Marrone]). Definite articles in
certain constructions ‘implicate’ a plurality of persons (Apollonius
Synt.I111,p.94,11–13,II32,p.149.8ff.); the Stoic provenance of
the term in this context is highly probable, though not certain, just
as it is in a report of the Stoic verbal tense system (DT Sch. Steph.
250.25ff.) which states that someone who uses a present-tense verb
‘implicates’ (emphainei) that his action extends into the past and the
future, the same being true, mutatis mutandis, of imperfects.
That Apollonius Dyscolus also employs the ‘implicature’ termi-
nology – and against the background of an apparently vague and un-
systematised semantic vocabulary – perhaps argues Stoic influence
in this area too. Thus, the distinction between the ‘implicature’ by
possessive pronouns of possessions as well as of possessor(s) (Synt. III
112,p.366.5ff.) may have its origin in the type of problem discussed
in Logical questions col. VI (Is ‘our’ [hˆ
emeteros] to be classed as sin-
gular or as plural?). The labeling of the infinitive as the verb ‘lacking
implicature’ (aparemphaton) – that is, of person and mood though not
of tense or voice (e.g., Synt. III 60,p.326.1f.) – may perhaps be Stoic
in origin (though other sources are against it [cf. Apollonius Synt. I
50,p.44.14ff.; Clement Strom. VIII ix 26.4f.). But unfortunately we
can only speculate that the Stoics took some word(-forms) to (fail to)
‘implicate’ gender (e.g., Apollonius Dyscolus Pr. 11.23,28.14,61.10)
or number (Synt. II 152,p.247.6), that the tenses ‘implicate’ differ-
ences of time (Dionysius of Halicarnassus De compositione uerbo-
rum VI 4), or that some prepositions ‘implicate’ a spatial relation
(a.d. Adv. 204.19–23).
