Reported Speech in the Transition from Orality to Literacy

Reported Speech in the Transition from Orality to

Literacy

(to appear in Glotta: Zeitschrift für griechische und lateinische Sprache)

Emar Maier

University of Groningen

https://sites.google.com/site/emarmaier

Abstract

In ancient Greek the line between direct and indirect discourse appears blurred.

In this essay I examine the tendency of Greek writers to slip from indirect into

direct speech. I explain the apparent difference between modern English and

ancient Greek speech reporting in terms of a development from orality to

literacy.

Direct and indirect speech

It is traditionally assumed that there are two distinct modes of reported speech: direct and

indirect discourse. A number of linguistic characteristics can be used to tease them apart,

even in the absence of prosody and modern, written punctuation. For instance, in indirect

speech we adjust pronouns and other context dependent expressions to fit the reporting

context, while in direct speech we simply copy the originals. That is, Yesterday, Otto said,

“I’m going there tomorrow” in indirect discourse becomes Yesterday, Otto said that he was

coming here today.

But not all forms of reported speech are straightforwardly classifiable as either

direct or indirect. In languages so diverse as Amharic and Catalan Sign Language, for

instance, linguists have struggled with reported speech constructions where some elements

behave as in direct speech and others as in indirect speech (Schlenker 2011). I will

demonstrate below that in ancient Greek too the line between direct and indirect discourse

is blurred.

I follow Maier's (2012) linguistic analysis of the ancient Greek switches in terms

of “mixed quotation”. The aim of this paper is to explain the difference between ancient

Greek and, say, modern English in the way they allow mixing of direct and indirect

discourse. The answer, I claim, starts with the observation es to the idea that Classical

Greek literature, and the reading culture in particular, retains some key elements of an

oral performance culture.

Reported speech in Ancient Greek

How do we distinguish between direct and indirect discourse in ancient Greek? In general,

there are two types of clues: (i) morphosyntactic surface features, such as a special

subordinating conjunction (like English that) or word order (as in German, for instance) to

indicate indirect speech, and (ii) semantic/pragmatic features such as the interpretation of

pronouns, demonstratives, and other deictic elements. Let’s start with the surface

characteristics.

The direct discourse reporting mode is presumably a linguistic universal: in ancient

Greek, as in any other language, we can report someone’s utterance by repeating

it and,

to avoid misunderstanding, adding a frame, i.e. something meaning x uttered these words.

This frame may be fronted, or added parenthetically as an afterthought or interjection

(e.g. “By Zeus,” said he, “what have I done?”). It is important to keep in mind that original

Greek texts have no actual quotation marks, so this important signal is lacking. However,

older authors like Homer and Herodotus often indicate the range of a direct quotation on

both ends, by prefacing it with a saying clause (Achilles said the following), and appending

a closing formula at the end (So spoke Achilles).

Ancient Greek indirect speech is overtly marked as such in a variety of ways. The

first type of syntactic indirect speech marking involves a verb of saying and a finite clause

introduced by ὅτι or ὡς (‘that’) – essentially the construction we use for indirect speech in

English. One minor difference between English and ancient Greek is that Greek, like e.g.

modern Russian, leaves all verb tenses as if they were still direct speech, rather than adjust

them to the current utterance situation. So, He said that he was ill becomes, in Greek,

literally He said that he is ill. In addition, in Classical Greek, the that-clause of an indirect

discourse may be marked with a special mood, the oblique optative.

A rather different way of marking indirect reporting is the Accusativus cum

Infinitivo (AcI) construction. In the AcI there is no that. Instead, the subject of the

embedded clause, if overtly realized, gets accusative case and the verb is in the infinitive.

In English we still find this construction with certain idioms (she expected him to be home

vs. she expected that he was home), but in ancient Greek, this was the preferred

construction to mark any type of indirect discourse (Gildersleeve 1906).

In most cases the morphosyntax of indirect speech, as described above, will

prevent ambiguity when interpreting a report construction in a text. If the reported speech

complement’s main verb is in the infinitive, or if it's introduced by ὅτι or ὡς, we are

probably dealing with indirect speech; if not, it’s probably direct speech. Note that this

classification is not foolproof because, for instance, in indirect questions, an interrogative

pronoun replaces that (cf. John asked who was there), and ὅτι and ὡς have a number of

uses distinct from the use as subordinating conjunction that.

More robust characteristics to distinguish the two modes are the seemingly

universal semantic differences. These include, first and foremost, the interpretation of

context-dependent expressions. In direct discourse the context of interpretation is shifted

to the original utterance context. In indirect discourse every expression has its regular

denotation. So, in colloquial, spoken English Otto said I’m a fool could be a direct or

indirect speech report, but these two options lead to divergent interpretations of the

embedded first person pronoun. On a direct speech reading, I refers to the speaker of the

Or at least pretending to do so. A note on terminology and theoretical background may be in order

here. I'm assuming that, as a matter of grammar, a report construction (x said (that) … ) always

purports to report a speech event, and hence I will refrain from using more general terms like

constructed dialogue (Tannen 1989).

reported context, i.e. Otto; on an indirect speech reading I refers to the current narrator,

i.e. the author of the report. Some knowledge about the context in which the report is

situated will likely disambiguate which reading is the most likely, and hence whether we

are dealing with direct or indirect speech.

A second useful heuristic for teasing apart direct and indirect speech semantically,

is the fact that the complement of an indirect report has to express a proposition, i.e. a

thought or content that is either true or false (Frege 1892). It follows that indirect speech

cannot contain non-propositional contributions to speech acts like exclamations (Hey

there!), vocatives or imperatives. Direct speech, on the other hand, being merely a more or

less verbatim reproduction of the original words is by no means restricted to grammatical

assertions. Hence, the mere occurrence of vocatives and imperatives in a reported speech

are clear signals of direct discourse.

Slipping from indirect into direct

The tendency to switch between direct and indirect discourse is relatively well studied

phenomenon in ancient Greek philology. Typically, the switch goes from indirect to

direct, in which case it has also been described as “slipping” (Richman 1986). Below is an

example from the opening of Homer's Odyssey. Zeus is speaking about how he had sent

the messenger Hermes to Aigisthus to warn him against Orestes' revenge if he, Aigisthus,

would kill Agamemnon. Note: In the following I’m underlining all the relevant linguistic

clues I rely on to determine whether something is a direct or an indirect speech report,

including the main reporting verb (i.e. the verbum dicendi in the frame), that, accusatives

and infinitives, and some context dependent expressions and vocatives. I will leave out the

quotation marks that modern editors have added to the Greek texts, but in translations I

represent the apparent mode switches with quotation marks.

Hom. Od. 1.35

[35] ὡς καὶ νῦν Αἴγισθος ὑπὲρ μόρον Ἀτρεΐδαο

γῆμ᾽ ἄλοχον μνηστήν, τὸν δ᾽ ἔκτανε νοστήσαντα,

εἰδὼς αἰπὺν ὄλεθρον, ἐπεὶ πρό οἱ εἴπομεν ἡμεῖς,

Ἑρμείαν πέμψαντες, ἐύσκοπον ἀργεϊφόντην,

μήτ᾽ αὐτὸν κτείνειν μήτε μνάασθαι ἄκοιτιν:

[40] ἐκ γὰρ Ὀρέσταο τίσις ἔσσεται Ἀτρεΐδαο,

ὁππότ᾽ ἂν ἡβήσῃ τε καὶ ἧς ἱμείρεται αἴης.

ὣς ἔφαθ᾽ Ἑρμείας, [...]

Even as now Aegisthus, beyond that which was ordained, took to himself the

wedded wife of the son of Atreus, and slew him on his return, though well he

knew of sheer destruction, [37] seeing that we told him before – by sending

Hermes, the keen-sighted Argeiphontes – [39] to neither kill him nor to seduce his

wife [40] “because there will be vengeance from Orestes for the son of Atreus,

when once he has come to manhood and longs for his own land.” So spoke

Translation based on A.T. Murray, Homer: The Odyssey. Cambridge: Harvard University Press, 1919.

Hermes.

The passage reports Zeus’ warning, as relayed by Hermes. It starts with a clear case of

infinitival indirect speech (‘we told him not to kill or seduce’ [37-39]).

In 40, we are still

listening to Zeus reporting how he, via Hermes, warned Aigisthus that killing Agamemnon

was a bad idea. Crucially, the main verb ἔσσεται in 40 is not an infinitive, but a finite

verb, “there will be vengeance”, which looks rather like a verbatim copy of what Hermes

(might have) literally said. The same holds for the next verb, ἱμείρεται (‘he would desire’).

Thus, the lack of indirectness markers (that or infinitive) strongly suggest direct speech for

40-41. The closing formula “So spoke Hermes” in 42 confirms the direct speech status of

the preceding passage. We infer that a switch from indirect to direct discourse has

occurred between 39 and 40.

The narrative effect of switching to direct in this way seems clear enough: the

author condenses the perhaps less interesting parts of Hermes’ speech by paraphrasing it

in indirect discourse, but toward the end he switches to direct to present a more vivid

rendition of the stern warnings, drawing the audience into the scene of Hermes addressing

Aegisthus. The closing formula then can be seen as a way to shift the focus back to the

narrator.

To our modern eyes, such a seemingly fluent change of construction within a

single report seems strange, perhaps strictly ungrammatical. Indeed, dropping the

quotation marks from the translation would effectively turn the whole passage into indirect

speech. This is not to say that in modern writing we cannot change from indirect to direct

speech in the middle of a report, just that we really need explicit quotation marks to mark

such a shift, as shown in the translation presented.

Overtly marking the switch to direct

Interestingly, even without quotation marks, these lapses into direct discourse, can be

overtly marked rather precisely on the surface. Herodotus, for instance, uses an interjected

ἔφη λέγων (‘he said, saying’) for this purpose (Kieckers 1916).

Hdt. 1.118.1-2

μετὰ δὲ ὣς οἱ ἐπαλιλλόγητο, κατέβαινε λέγων ὡς περίεστί τε ὁ παῖς καὶ τὸ

γεγονὸς ἔχει καλῶς. τῷ τε γὰρ πεποιημένῳ ἔφη λέγων ἐς τὸν παῖδα τοῦτον

ἔκαμνον μεγάλως, καὶ θυγατρὶ τῇ ἐμῇ διαβεβλημένος οὐκ ἐν ἐλαφρῷ

ἐποιεύμην. […]

Then, after repeating it, [Harpagus] ended by saying that the boy was alive and

Another possibility is that the infinitives in line 39 are independent main clause infinitives, which

Homer does indeed use to express commands. In that case we would not be forced to assume a switch,

just a direct speech report, “We told him, ‘Don't kill his son or seduce his wife, because from Orestes

revenge shall come!’” There are, however other indirect – direct switches in Homer described in the

literature, identified on the basis of a variety of more or less convincing clues (Kühner & Gerth 1904;

Kieckers 1916).

Translation based on A.D. Godley, Herodotus. Cambridge: Harvard University Press, 1920.

that the matter had turned out well. “For,” he said, “I was greatly afflicted by what

had been done to this boy, and it weighed heavily on me that I was estranged from

my daughter. […] ”

Herodotus starts his report of Harpagus’ speech in the indirect mode, this time marked

with a saying verb plus finite that-clause (λέγων ὡς περίεστί ‘saying that he was alive’).

In the next sentence, Harpagus explains why he thinks “everything turned out well.” His

reason is presented as, again, a finite clause, but the first person pronouns and affixes (I

was afflicted … weighs heavily on me … my daughter) show beyond a doubt that the

author has slipped into direct discourse – this is Harpagus speaking about his guilt and his

daughter, not the actual narrator’s (Herodotus).

The interjection of the formulaic saying frame ἔφη λέγων (‘he said’) in Herodotus

is an indicator of the transition to direct discourse. Note that parenthetical he said

interjection is typical of direct reporting, even in modern English (as shown in the

translation).

In Attic Greek, we find similar switches. Like Herodotus, Xenophon sometimes

marks these with an interjected ἔφη (‘he said’):

Xen. Anab. 1.8.12

καὶ ἐν τούτῳ Κῦρος … τῷ Κλεάρχῳ ἐβόα ἄγειν τὸ στράτευμα κατὰ μέσον τὸ

τῶν πολεμίων, ὅτι ἐκεῖ βασιλεὺς εἴη: κἂν τοῦτ᾽, ἔφη, νικῶμεν, πάνθ᾽ ἡμῖν

πεποίηται.

At this instant, Cyrus … called aloud to Clearchus to advance against the enemy's

centre, for there the king was to be found: “And if we strike home at this point,”

he said , “our work is finished.”

But ἔφη (‘he said’) is not a very reliable indicator. First, it is not sufficient, as it may also

introduce indirect discourse, and second, it is not necessary, as there are examples of

shifts that are not so marked. Both phenomena are illustrated in the passage below. We

have two typical infinitival indirect discourse constructions headed by ἔφη, before the

author slips into direct mode, as evidenced by the use of the first person plural.

Xen. Anab. 1.3.20

πρὸς τοῦτον οὖν ἔφη βούλεσθαι ἐλθεῖν: κἂν μὲν ᾖ ἐκεῖ, τὴν δίκην ἔφη

χρῄζειν ἐπιθεῖναι αὐτῷ, ἦν δὲ φύγῃ, ἡμεῖς ἐκεῖ πρὸς ταῦτα βουλευσόμεθα.

It was against him, that he said he desired to march. And, if he were there, he said

that he wished to inflict due punishment upon him, “but if he has fled, we will

deliberate about the matter then and there.”

It is not entirely clear where the direct discourse begins, but arguably in this, like all other

examples discussed so far, it is right between two juxtaposed main clauses.

Translation (of this and subsequent Xenophon examples) based on C.L. Brownson, Xenophon.

Cambridge: Harvard University Press, 1922.

Subclausal switching: consequences for syntax and semantics

Syntactically speaking, on the basis of the data up to this point, we could maintain that

there is always a full indirect discourse segment that is closed off before a new sentence in

direct discourse is started. We can even posit an elided or otherwise covert frame for the

free (=unframed) direct discourse clause to get the semantics right. The underlying

“logical form” of our last example, would then be, roughly, He said that he wished to

inflict punishment. And then he said, “If he has fled, we will deliberate.”

The advantage of such an analysis would be that we can maintain the basic

assumption underlying the received view in linguistics: that there are just two types of

reports, direct and indirect. Note that covert frames, or, more neutrally, free direct

discourse, is common in English as well. The seemingly trivial difference between English

and Greek would be the need to always mark direct discourse overtly in writing by means

of quotation marks.

The reality of indirect-to-direct switch in ancient Greek is, however, more

complicated than that. Shifts are not restricted to coordinate clause boundaries. A few

paragraphs above the last passage quoted, we find a clear shifts at a relative clause

boundary:

Xen. Anab. 1.3.14 (cf. also 1.3.16)

εἷς δὲ δὴ εἶπε … στρατηγοὺς μὲν ἑλέσθαι ἄλλους … πέμψαι δὲ καὶ

προκαταληψομένους τὰ ἄκρα, ὅπως μὴ φθάσωσι μήτε Κῦρος μήτε οἱ Κίλικες

καταλαβόντες, ὧν πολλοὺς καὶ πολλὰ χρήματα ἔχομεν ἀνηρπακότες.

One man in particular … proposed to choose other generals … and likewise to

send a force to occupy the mountain heights in advance, in order that they be

forestalled by neither Cyrus nor the Cilicians “of whom we have many in our

possession as well as a lot of their property that we have seized as plunder.”

The first person plural we have in our possession in this context indicates direct discourse,

but it occurs inside a relative clause (of whom…), embedded in a subordinate purpose

clause (in order that…), inside an infinitival indirect discourse (proposed… to send…).

There is simply no way to place the supposed shift between two coordinated main clauses.

Hence, a paratactic semantic analysis in terms of two separate indirect and direct reports,

as sketched above, fails to capture the right dependencies (cf. ??He proposed to send a

force so they wouldn’t be forestalled by the Cilicians. And then he said “Of whom we have

many in our possession”).

Another example, this time with a shifted first person in a because-clause,

embedded in an AcI:

Aristoph. Eccl. 821

ἀνέκραγ᾽ ὁ κῆρυξ μὴ δέχεσθαι μηδένα

Translation based on B.B. Rogers, Aristophanes. Cambridge: Harvard University Press, 1924.

χαλκοῦν τὸ λοιπόν ἀργύρῳ γὰρ χρώμεθα.

[just as I was holding out my sack], the herald shouted that nobody should accept

copper in the future “because we need silver”

In this case, the quoted fragment appears to be a main clause itself, but coordinated with

an infinitival clause, in such a way that both conjuncts logically seem to depend on the

saying frame (the herald shouted). Note how difficult it would be for a modern reader of

English to get the right interpretation if there were no quotation marks. Given the context,

the we in this last example just wouldn’t make sense: ??Just as I was holding out my sack,

the herald shouted that nobody should accept copper because we need silver.

What the examples above show is that the phenomenon cannot be reduced to a

mere concatenation of indirect and direct discourse. Below I will introduce the linguistic

notion of mixed quotation as a way to properly capture the syntax and semantics of mode

switching. But first, I want to chart the boundaries of the phenomenon under discussion.

This will lead us to consider more complex switches in and out of direct discourse, which

will be amenable to the same semantic treatment.

Some notes on the diachronic and crosslinguistic spread of switching

Transitions from indirect to direct discourse within a single report are common to all

eras, genres, and dialects of ancient Greek writing. So far, we’ve seen examples from

Homer (c.800BC), Herodotus (c.450BC), Xenophon (c.400BC), and Aristophanes

(c.390BC). But similar switching occurs much later, e.g. in Polybius (c.150BC, cf.

Usher 2009), and Josephus (c.95AD, cf. Richards 1939). In the New Testament,

especially Luke and Acts, the phenomenon is well studied (Cadbury 1929; Buttmann

1859). Let’s consider one of these late examples.

In the passage below, cited by Cadbury, we see an indirect speech report in AcI

dependent on the framing verb charged. Further along, the AcI has evidently slipped into

the direct mode, as shown by the local person forms, you told me, that are not meant to

refer to the narrator and his audience.

Acts 23.22

ὁ μὲν οὖν χιλίαρχος ἀπέλυσε τὸν νεανίσϰον παραγγείλας μηδενὶ ἐϰλαλῆσαι

ὅτι ταῦτα ἐνεφάνισας πρὸς ἐμέ.

So the commanding officer let the young man go, charging him to tell no one “that

you have told these things to me.”

As for the crosslinguistic aspect, I would just note that the phenomenon of unmarked

switching from indirect to direct discourse is attested in a number of other “old languages”

like Aramaic (Richards 1939) and Old English (Richman 1986).

More generally, the

phenomenon is sometimes described as characteristic of oral storytelling (Rajić 2008). I

Arguably, Late Egyptian has some form of mode switching too, although Kammerzell & Peust

(2002) classify some of the relevant examples as a special type of indirect discourse.

return to the relation between switching and orality in the final sections of this essay.

The translations above show that modern English has no problem with switching

between direct and indirect discourse either, even subclausally. The difference is that, in

writing, we now really need the quotation marks to be felicitous.

Narratological aspects of switching

As for the stylistic, narratological effects of mode switching, I will not go much further

than what I already noted. Direct discourse presents a more vivid picture of the speech act

reported. Indirect speech on the other hand allows the narrator to significantly shorten and

summarize the less important parts. In some cases the compromise is to start in indirect

discourse but end with a few vivid key phrases in direct speech. The result is the kind of

switching we've seen above.

Now, the question naturally arises, can we also switch the other way, from direct to

indirect speech? Clear examples of this are hard to find, but Buttmann (1859) list a

number of plausible cases from the New Testament:

Acts 23.23-24

εἶπεν Ἑτοιμάσατε στρατιώτας διακοσίους ὅπως πορευθῶσιν ἕως Καισαρίας,

καὶ ἱππεῖς ἑβδομήκοντα καὶ δεξιολάβους διακοσίους, ἀπὸ τρίτης ὥρας τῆς

νυκτός, κτήνη τε παραστῆσαι ἵνα ἐπιβιβάσαντες τὸν Παῦλον διασώσωσι

πρὸς Φήλικα τὸν ἡγεμόνα,

He said, “Prepare two hundred soldiers to go as far as Caesarea, with seventy

horsemen, and two hundred men armed with spears, at the third hour of the night”

[and told them] to provide animals, that they might set Paul on one, and bring him

safely to Felix the governor.

The saying verb εἶπεν (‘he said’) in the first line introduces a direct speech Prepare two

hundred soldiers (second person plural imperative). But the last part of the very same

command, provide animals to put Paul on, appears to have been cast in indirect discourse,

as signaled by the infinitive main verb παραστῆσαι (‘to provide’).

A phenomenon very closely related to the rare shift from direct to indirect, is the

switching between direct discourse and parenthetical remarks that we find in particular

with some of the Attic orators. These are cases where a direct quote, is interspersed with

commentary from the narrator himself. This can lead to some rather intricate

constructions. Take the following example of Aeschines quoting a law (he refers to the

legislator), while adding his own polemically charged commentary.

Aeschin. 1.19–20

ἄν τις Ἀθηναίων, φησίν, ἑταιρήσῃ, μὴ ἐξέστω αὐτῷ τῶν ἐννέα ἀρχόντων

γενέσθαι, ὅτι οἶμαι στεφανηφόρος ἡ ἀρχή, μηδ’ ἱερωσύνην ἱερώσασθαι, ὡς

This example discussed by Dover. Similar examples from Demosthenes and Aeschines are discussed

by Bers (1997).

Translation based on C.D. Adams, Aeschines. Cambridge: Harvard University Press, 1919.

οὐδὲ καθαρεύοντι τῷ σώματι, μηδὲ συνδικησάτω, φησί, τῷ δημοσίῳ, μηδὲ

ἀρξάτω ἀρχὴν μηδεμίαν μηδέποτε, μήτ’ ἔνδημον μήτε ὑπερόριον, μήτε

κληρωτὴν μήτε χειροτονητήν: μηδὲ κηρυκευσάτω, μηδὲ πρεσβευσάτω, μηδὲ

τοὺς πρεσβεύσαντας κρινέτω, μηδὲ συκοφαντείτω μισθωθείς, μηδὲ γνώμην

εἰπάτω μηδέποτε μήτε ἐν τῇ βουλῇ μήτε ἐν τῷ δήμῳ, μηδ᾽ ἂν δεινότατος ᾖ

λέγειν Ἀθηναίων.

“If any Athenian,” he says, “shall have prostituted his person, he shall not be

permitted to become one of the nine archons,” because, I think, that official wears

the wreath;

“nor to discharge the office of priest,” as being not even clean of body;

“nor shall he act as an advocate for the state,” he says, “nor shall ever hold any

office whatsoever, at home or abroad, whether filled by lot or by election; nor shall

he be a herald or an ambassador”—nor shall he prosecute men who have served as

ambassadors, nor shall he be a hired slanderer— “nor ever address senate or

assembly,” not even though he be the most eloquent orator in Athens.

There are very few concrete linguistic signals of switching here. We just have two

apparently interjected cases of φησί (‘he said’), which signal that we’re dealing with a

direct reported speech. Yet, the content makes clear that some of the clauses are not part

of the law quoted. The clearest signal is the first person οἶμαι (‘I think’), which we can

only interpret as belonging to a “clarifying” interpretation by Aeschines of the law that he

is quoting. The translation reveals the rapid back and forth switching in the rest of the

passage.

Note that this switching serves a rather different purpose from the more literary

switches discussed earlier. Aeschines’ intention is not to liven up his narrative, but rather

to trick the jury into interpreting the law the way it suits him. As Dover (1989:24) points

out, regarding this passage, “we have to remind ourselves that if a speaker in court thought

it helpful to his case to confuse the issue while professing to clarify it, he would do his

best to confuse it.”

The examples in this section show that indeed the switch can go the other way

around, and moreover, that an author may have various pragmatic reasons for switching,

ranging from literary stylistics, to rhetorical trickery.

Towards a linguistic analysis of mode switching as mixed quotation

Maier (2012) provides a uniform linguistic analysis of the mode switching phenomena

discussed above, building on recent advances in the study of so-called mixed quotation in

linguistics and philosophy of language. Mixed quotation is a form of speech reporting that

is typically associated with factual, written genres of text, such as newspapers and

scientific writing. On the surface, it looks like an overtly marked mix of direct and

indirect speech, best defined by example:

Quine says that quotation “has a certain anomalous feature”

This mixed quote is both an indirect discourse report, informing us that Quine said that

quotation has a certain anomalous feature, and at the same time also a verbatim direct

quote of a specific phrase uttered by Quine.

The direct discourse aspect is brought out when we consider context dependent

expressions:

He said that during those moments “my ass was Uncle Sam’s.”

The intended reading is clearly one where the first person my refers to the reported

speaker, he (some general Boyd), rather than the actual author of the sentence, the

reporter for Time. It follows that mixed quotation is not “mere punctuation” but has a

genuine semantic impact. Without the quotation, this example could only mean that

general Boyd said that the Time reporter’s life belonged to the military. Quotation marks

are required in written English to mark the intended, mixed quotation reading of the

sentence, where the first person possessive is interpreted as spoken literally by, and hence

referring to, the general.

But on the other hand, mixed quotation is not just a verbatim reproduction. Both in

form and in meaning it resembles indirect discourse, in which every word is used in its

ordinary sense. Note for instance that we also infer from the original example that Quine

says that quotation has an anomalous feature. Moreover, as Davidson (1979) argues, if

mixed quotation were pure quotation, the quoted phrase would be a referential term, a

noun phrase referring to the very words quoted, but that simply doesn’t fit the

grammatical slot filled by the quotation (a verb phrase in the Quine example, a full clause

in the general Boyd example).

In sum, beside overtly mixing some surface characteristics of direct and indirect

discourse, mixed quotation also mixes the underlying semantic characteristics, use and

mention, of the two modes.

Maier (2012) goes on to propose a formal semantic analysis of the phenomenon of

mixed quotation that unites both the direct and indirect discourse aspects discussed above.

In a nutshell, this account analyzes our mixed quotation as follows:

Quine said that quotation has the property that he refers to with the words has an

anomalous feature.

Maier spells out this rough paraphrase in the theoretical linguistic framework of

truth-conditional semantics, the subdiscipline of linguistics that deals with meanings in

terms of reference and truth (also known as formal, or model-theoretic semantics, and to

be distinguished from cognitive semantics). For our current purposes what's important is

that in his analysis (i) the actual words are literally part of the truth conditions and (ii), in

addition, the property he (Quine) referred to with these words (presumably, something

close to the property of having an anomalous feature) is also part of the truth conditions.

These two properties correctly derive both the directness and the indirectness

characteristics observed with mixed quotation. For a more fleshed out formal semantic

analysis of this idea along with a demonstration that it indeed adequately captures the

Report in Time about General Boyd's speech, found at

http://swampland.time.com/2010/05/20/wearing-the-uniform-with-pride/

linguistic facts about mixed quotation, I refer to Maier (2012) and references therein.

Returning to ancient Greek, the idea is simple: Ancient Greek, unlike modern

English, allows unmarked mixed quotation within indirect discourse complements. In

other words, both English and Greek can freely switch from indirect to direct discourse by

means of mixed quotation, but written English requires overt quotation marks to achieve

this, while written Greek does not. The main claim here is that the underlying semantic

mechanism to achieve such a switch is the same in both languages, viz. mixed quotation.

The difference between English and Greek can now be restated as follows: written

English overtly realizes mixed quotation in the surface form, but written ancient Greek

does not. Note already that in spoken language there may not be much of a difference at

all: both English and Greek speakers can mark (mixed) quotations paralinguistically, i.e.

with gestures, intonation and pauses. The difference between oral and written

communication will be the topic of the next section.

From orality to literacy

The question left wide open by Maier's (2012) semantic analysis is, why do ancient Greek

texts allow mode switching by covert mixed quotation, where English requires overt

quotation marks?

My explanation starts from the idea that this difference between the two languages

resides wholly in the modality of writing and reading. In direct, face-to-face

communication, modern speakers also switch back and forth within a single reporting

clause. For example, it’s easy to imagine spoken versions of the English mixed quotes in

the previous section occurring in a conversation, without the direct analogues of quotation

marks, i.e. “spoken punctuation” like saying quote—unquote or using “fingerdance

quotes.”

However, when uttering a mixed quotation aloud, speakers do tend to use some

kind of marking of the perspective shift – a slight pause, a special intonation, or perhaps

even a different voice or accent mimicking some peculiarity of the reported speech act

(Kasimir 2008). I hypothesize that, to facilitate successful communication, speakers in a

direct communicative situation will always try to mark quotation boundaries whenever the

context leaves any room for ambiguity. For this marking, speakers, now and in antiquity,

rely on an array of more or less subtle paralinguistic means. Prosodic and gestural

“role-playing” thus allows speakers to mark mode switches in reported speech fluidly yet

effectively.

In modern, written communication, the relation between sender and receiver is

radically different. The reader has become far removed from the author and can no longer

rely on any extra- or paralinguistic perspective marking. To find out exactly whose voice

we’re hearing in a written report, the modern reader would benefit from a strict, easily

recognizable distinction between direct and indirect speech. In modern writing we do

indeed find direct and indirect discourse always clearly marked as such, by various forms

of punctuation. More subtle mixes involving subclausal switches are also still possible – in

fact quite common in some genres – but they strictly require overt quotation marks.

On the current view, quotation marks are the modern way of disambiguating

reported speech in a medium that cannot convey the gestural and prosodic clues that serve

this purpose in face-to-face communication. Borrowing an example from Johnson (2000),

compare the invention of written quotation marks to the recent invention of emoticons.

Just as :) in chatty email conversations stands in for the paralinguistic irony marking of

oral communication, quotation marks stand in for the paralinguistic marking of

perspective shift.

Where do our ancient texts fit in? In purely oral cultures we may expect to find

that reliance on fluid, paralinguistic disambiguation has made a rigidly marked separation

between direct and indirect speech superfluous.

As for Greek, we might expect a strong

influence of orality in the epic poetry of Homer, which, although committed to writing at

some point during or after composition, is now commonly characterized as essentially an

oral narrative (Parry 1971). Havelock (1963) has famously argued that Greek literature

remained oral until an abrupt transition to literacy starting with Plato in 4

century BC

Athens. That might already explain some of our examples as instances of oral writing, but

what about Xenophon, or Josephus and Luke (1

century AD)?

Nowadays, Havelock's black and white picture, although taken over by Goody &

Watt (1963), Ong (1982) and further work in anthropology and cultural history, is

typically rejected by classicists (cf. e.g. Solmsen 1966). In its place came a much more

nuanced picture emphasizing the coexistence of different forms and degrees of orality and

literacy. For instance, as Thomas (1989) points out, the fact that many Athenians in the 5

and 4

century could read and write their name, graffiti, and records of sales is perfectly

consistent with a primarily oral culture, where narratives were told and re-told rather than

read from books. On the basis of this nuanced picture, I will explain below the seemingly

problematic lack of quotation marking in Greek writing in terms of its roots in oral

performance culture. This line of argument is reminiscent of, for instance, Slings'

(1992) analysis of another apparent anakoluthon, the dangling participle (nominative

absolute). Slings points out that, although this phenomenon seems somehow incorrect or

puzzling from a modern reader's perspective, it is easily explained when viewed from the

perspective of oral communication and its specific demands on information structuring.

Although in a sense also an anakoluthon, the lacking mixed quotation marking under

discussion here differs from the dangling participle in that it cannot be viewed as a

pragmatic iformation structuring device from oral communication: as I've pointed out

above, oral communication demands a marking of the mixed quotational perspective shift

as much as the written version, be it typically of a paralinguistic nature. Following

Johnson (2000), my derivation of unmarked switches from oral performance goes via an

examination of the reading and writing culture in the long period of transition from orality

to modern literacy. This route also explains why in this case, oral influence extends

beyond what Slings refers to as “quasi-spoken writings” (dialogues and plays).

Ancient writing, reading and performance

Cf. Everett (2010) for an example of an oral culture (the Pirahã of the Amazon) that don't seem to make

a clear direct/indirect speech distinction. We may expect similar direct—indirect mixing in signed

languages, which also lack a writing system and which are known to rely heavily on gestural support

in direct communicative situations. Some evidence of this can be seen in recent studies of Role Shift,

commonly viewed as the sign language equivalent of direct quotation, is shown to exhibit

characteristics of both direct and indirect discourse (Herrmann & Steinbach 2007).

In a sense it’s obvious why the mixed quotations in our Greek examples were not

marked with quotation marks – the Greek writing system simply did not have any

punctuation whatsoever. In fact, until well into the Middle Ages, texts like these were

written and copied in scriptio continua, i.e.

WITHOUTANYWORDSPAC

ESORSENTENCEBREAKSIN

LONGNARROWCOLUMNS

OFCAPITALLETTERS

The real question is, why did this seemingly impractical writing mode survive for so long?

Johnson (2000) points out the stark contrast between (i) the slow adoption of word

separation and punctuation in ancient writing, and (ii) the fast spread of the smiley face in

modern electronic communication. Clearly, if the ancients had felt a need for quotation

marks or commas, in the same way that internet users of the 1990's felt a need for

textually marking irony, it wouldn't have taken so many centuries to invent a way to mark

it.

Let's take a closer look at this seemingly impractical scriptio continua. Modern

accounts call on the neurophysiology of reading to make sense of it. Saenger (2000), for

instance, argues that this type of writing may have been hard to read silently, but does

allow, even necessitates, reading aloud, which was indeed common practice until well into

the Middle Ages. Taking into account both the physiology and sociology of reading,

Johnson turns it around: he argues that the particular practice of reading aloud and

performing literary works actually favored the writing style. Unlike modern emailers and

chatters, the ancients had no desire to make their writing easily accessible. The reason is

that the social practice of reading in antiquity was an altogether different affair from

reading today. Silent reading to oneself surely was an option for the literate Greek, but for

literary text in particular, reading was primarily a social activity – a rehearsed

performance to entertain the elite and their guests.

Just as speeches and plays are meant to be rehearsed and then performed in front

of a live audience, rather than read to oneself in silence, so too reading prose and poetry

typically involved a dramatic performance. A well-known metaphor is that the ancients'

reading of scriptio continua was like a conductor's reading a musical score in a

performance: studied, rehearsed and mostly memorized in advance. The reader, a skilled

professional, did not seek an efficient intake of information, but merely a good mnemonic

recitation aid. The narrow columns, which according to Johnson contain precisely so many

letters as to be captured and processed in a single eye fixation, may have been nearly

optimal for this purpose. In support of this view, not also that, just as we might expect in a

conductor’s sheet music, we find occasional remnants of what appear to be a reader's

preparations in the margins of manuscripts. Mostly these are paragraph or sentence

breaks, but in some cases the onset of a direct quotation is marked as well.

In sum, the remnants of oral performance in ancient reading of literature meant

that ancient text served a fundamentally different purpose from modern text. An ancient

reader was a performer. He had plenty of time to parse, prepare and rehearse his oral

delivery of a text. In particular, this advance preparation included figuring out who said

what in a report, on the basis of various contextual and grammatical clues. Furthermore,

as these reading performances were reserved for the elite, writing did not have to facilitate

efficient dissemination of information.

Conclusion

In this essay I have shown that a strict dichotomy between direct and indirect discourse is

untenable for ancient Greek literature. I have presented various forms of switching

between direct and indirect discourse within a single speech report. The prototypical case

is the switch from indirect to direct discourse, which can occur between two coordinated

clauses or subclausally, and may or may not be marked with a parenthetical he said. To

detect an unmarked switch we can rely on the intended referents of context dependent

expressions like pronouns, as well as independent indicators of indirect speech (e.g.

infinitive main verb, subordinating that) or direct speech (vocative, exclamation,

imperative).

This type of switch is well studied in philology, and is typically understood as a

literary device to liven up a speech report by condensing the boring parts, but ending with

a vivid direct quote. The phenomenon is typical of a wide range of genres, eras and

authors (from Homer to the New Testament).

Switches of the unmarked, subclausal variety are particularly challenging for

modern readers of English prose, where quotation marks are vital to getting the right

interpretation. They also a pose a problem for the semantics, as the meaning of such a

report cannot be analyzed as a conjunction of two independent reports, one indirect and

one direct. For this reason I choose to model mode switching as mixed quotation, a

phenomenon that is actively being studied in philosophy and linguistics. In mixed

quotation, any constituent in an indirect report can be quoted, yielding a two-dimensional

interpretation where both form and meaning play a role. Moreover, direct discourse

reports can be analyzed as a limiting case, i.e. mixed quotation where the full report

complement is mixed quoted.

Presumably, speakers of ancient Greek and English are alike in marking the

semantic shift inherent in a mixed quotation/mode switch by a variety of prosodic and/or

gestural means. However, their writing systems are significantly different in this respect:

English writing requires quotation marks, but Greek doesn’t have those, nor any other

form of punctuation, or even word spacing to represent the paralinguistic marking we

expect in direct, oral communication.

To explain this difference I turn to the ancient Greeks’ reading culture. Recent

accounts of reading suggest that text in scriptio continua is perfect as a mnemonic tool for

a rehearsed, oral performance. And this was, indeed, the primary use of literary texts at

the time. The sociology of reading in antiquity thus explains a linguistically puzzling

difference between Greek and English with respect to quotation marking in reported

speech.

References

Bers, V., 1997. Speech in Speech: Studies in Incorporated Oratio Recta in Attic Drama and

Oratory, London: Rowman & Littlefield.

Buttmann, A., 1859. Grammatik des neutestamentlichen Sprachgebrauchs, Berlin:

Dümmler.

Cadbury, H.J., 1929. Lexical Notes on Luke - Acts IV: On direct quotation, with some

uses of ὅτι and εἰ. Journal of Biblical Literature, 48(3), pp.412–425.

Dover, K.J., 1989. Greek Homosexuality, Cambridge: Harvard University Press.

Everett, D.L., 2010. You drink. You drive. You go to jail. Where’s recursion? In U Mass

Recursion Conference 2009. http://ling.auf.net/lingBuzz/001141.

Frege, G., 1892. Über Sinn und Bedeutung. Zeitschrift fur Philosophie und philosophische

Kritik, 100(1), pp.25–50.

Gildersleeve, B.L., 1906. Notes on the Evolution of Oratio Obliqua. The American

Journal of Philology, 27(2), pp.200 – 208.

Goody, J. & Watt, I., 1963. The Consequences of Literacy. Comparative Studies in Society

and History, 5(3), pp.304 – 345.

Herrmann, A. & Steinbach, M., 2007. Wenn ’ich ' nicht ich ist: Redewiedergabe in

Gebärdensprachen. In E. Brendel, J. Meibauer, & M. Steinbach, eds. Zitat und

Bedeutung. Hamburg: Helmut Buske Verlag, pp. 20–40.

Johnson, W.A., 2000. Toward a Sociology of Reading in Classical Antiquity. The

American Journal of Philology, 121(4), pp.593 – 627.

Kammerzell, F. & Peust, C., 2002. Reported speech in Egyptian: forms, types and

history. In T. Güldemann & M. von Roncador, eds. Reported Discourse: a Meeting

Ground for Different Linguistic Domains. Philadelphia: John Benjamins, pp.

289–322.

Kasimir, E., 2008. Prosodic correlates of subclausal quotation marks. ZAS Papers in

Linguistics, 49, pp.67 – 77.

Kieckers, E., 1916. Zur oratio recta in den indogermanischen Sprachen II.

Indogermanische Forschungen, 36, pp.1–70.

Kühner, R. & Gerth, B., 1904. Ausführliche Grammatik der Griechischen Sprache: II. Teil:

Satzlehre, Hannover: Hahn.

Ong, W.J., 1982. Orality and Literacy: the technologizing of the word, New York:

Methuen.

Parry, A., 1971. The making of Homeric verse: The collected papers of Milman Parry,

Oxford: Clarendon Press.

Rajić, L., 2008. Mixing oratio recta and oratio obliqua: a sign of literacy or orality? In

Oral Art Forms and their Passage into Writing. Copenhagen: Museum Tusculanum

Press, pp. 203–208.

Richards, G.C., 1939. The Composition of Josephus’ Antiquities. The Classical Quarterly,

33(1), pp.36–40.

Richman, G., 1986. Artful slipping in Old English. Neophilologus, 70(2), pp.279–291.

Saenger, P., 2000. Space between Words: the Origins of Silent Reading, Stanford: Stanford

University Press.

Schlenker, P., 2011. Indexicality and De Se Reports. In K. von Heusinger, C. Maienborn,

& P. Portner, eds. Handbook of Semantics. The Hague: Mouton de Gruyter.

Slings, S.R., 1992. Written and spoken language: an exercise in the pragmatics of the

Greek sentence. Classical Philology, 87(2), pp.95–109.

Solmsen, F., 1966. Review of E. Havelock “Preface to Plato”. The American Journal of

Philology, 87(1), pp.99–105.

Tannen, D., 1989. Talking Voices: Repetition, Dialogue, and Imagery in Conversational

Discourse, Cambridge: Cambridge University Press.

Thomas, R., 1989. Oral Tradition and Written Record in Classical Athens, Cambridge:

Cambridge University Press.

Usher, S., 2009. Oratio Recta and Oratio Obliqua in Polybius. Byzantine Studies, 49,

pp.487–514.