The Relevance of Evidentiality for Ancient Greek: Some Explorative Steps through Plato

The present paper aims at drawing renewed attention to the relevance of evidentiality for Ancient Greek by means of a number of case studies taken from two of Plato’s works (namely the Apologia Socratis and Crito). First, I briefly identify the conceptual framework within which the main analysis of Attic evidential phenomena occurs. Then, I provide a preliminary overview of (possible) linguistic means used in marking evidentiality in Ancient Greek (formal aspect). I also explore the way in which evidential values are conveyed (semantic aspect). Certain Attic particles (e.g., ara, dḗpou), functional oppositions in complementizing patterns (e.g., hoti vs. hōs), defective verbal forms (e.g., ēmi), and “auxiliaries” (e.g., dokeō) are revealed as evidential markers or “strategies”. These are able to express inferential, presumptive, reportative, quotative, visual, and participatory evidentiality. The oblique optative is suggested to have evidential overtones as well. In summary, the paper endeavors to show the importance of “evidentiality” as an integrative conceptual frame for the descriptive analysis of certain Ancient Greek phenomena.


Introduction
Speakers of any language are able to express the information source on which they rely in uttering a certain proposition. The information source is prototypically encoded by verbal morphemes.1 For example, if a speaker of Jamul Tiipay (a Yuma language spoken in Mexico and the United States) notices that one of his friends is sweating and seems nervous while inventing an implausible excuse for not paying him back, he will tell you: (1) tew-ii-kex-a 3sg-lie-infr-emp 'He must be lying.' example taken from miller 2001: 192 Based on the attitude of his friend and the implausibility of his account, the speaker infers that he must be lying (expressed by the suffix -kex). Clearly, Ancient Greek in general and fourth-century bc Attic in particular do not have such evidential systems of verbal morphemes. Nevertheless, Attic relies on other linguistic means to express evidentiality. These include disputed, heterogeneous grammatical phenomena (such as particles, contrastive complementizer constructions, and the oblique optative), for which the conceptual domain of evidentiality may prove to be a useful integrative approach. There may be evidential subsystems (e.g., a particle cluster around dḗ and its derivatives dḗpou and dêta; cf. sub 4.1.) and evidential extensions of certain verbal constructions (consisting of contrasting subsystems; cf. sub 4.2. & 4.3.). Moreover, although it seems that the linguistic encoding of this conceptual domain can take varying shapes in Ancient Greek, no attempts have been made to con-Journal of Greek Linguistics 16 (2016) 3-46 nect all these (possibly) evidential features through the conceptual domain of evidentiality. The present paper does not seek to provide a comprehensive study of evidentiality in Attic, but rather to offer: i. a concise introduction to the phenomenon of evidentiality and to the conceptual framework that constitutes the analytical starting point, ii. a discussion of its relevance to the Ancient Greek language as it appears in two of Plato's works (comparable to Cuzzolin's 2010 exploration of evidentiality in Latin), and iii. a number of case studies that consider (possible) evidential morphemes and strategies, so as to reinterpret some Attic phenomena as (partially) evidential.
After presenting some provisional conclusions, the paper rounds off with an outlook for future research. Due to the explorative nature of this contribution, the discussion of the evidential morphemes and strategies draws on a close analysis of two texts, the Apologia Socratis and the Crito by Plato (cf. sub 2).

2.1
Definition Several definitions of the conceptual domain of evidentiality are available.2 A majority of scholars would likely agree that evidentiality is the explicit linguistic encoding of the information source on which the speaker relies in uttering a certain phrase. Evidential markers or strategies make the "evidence" for the contents of a proposition explicit.
Because evidentiality concerns an epistemic stance toward the contents of a proposition, it is bound to interfere with epistemic modality (and even with 2 Cf. Diewald-Smirnova (2010: 1). The term "conceptual domain" refers to a certain idea constituting a coherent entity within human experience that can be encoded linguistically (with more grammatical and/or more lexical elements). This conceptual domain needs to be carefully distinguished from its grammatical realization by means of morphological systems. Until very recently, the latter has been the main focus of attention for many linguists studying evidentiality (cf. sub 2.3.). Although evidentiality may be considered a linguistic universal in that the information source for a proposition can be rendered in every language, evidential morphemes are as little a universal as temporal morphemes (Aikhenvald 2003(Aikhenvald : 1 & 2004. indirect speech; Volkmann 2005). However, the exact nature of the relationship between these two conceptual domains remains a subject of discussion (cf. Cornillie 2009;Plungian 2010; studies in Aikhenvald-Dixon 2014). Another conceptual domain closely linked to evidentiality (and, more specifically, to inferential markers) is that of "mirativity", which refers to the semantic field expressing the speaker's surprise regarding the information conveyed (Aikhenvald 2004: 195-215; see also Feuillet 2006: 333 et sqq. for a discussion of notions cognate to evidentiality). Evidentiality also interferes with deixis, as evidential morphemes can "point to" discernable information within the reach of the speakers and/or hearers (cf., e.g., Mushin 2001: 33-34;see Aikhenvald 2004: 275-276 for the deictic origins of certain evidentials). This is, however, not the place to go further into these matters; they will be touched upon when relevant to the discussion.

Classifying Evidential Values: The Semantic Perspective
Drawing largely on Plungian (2010: 37), who, in his turn, elaborated upon Willett (1988) and Aikhenvald (2004), I used the classification of evidential values presented in Table 1 to approach the Attic linguistic encoding of evidentiality. As is common in semantics, these evidential values are not to be viewed as clearly demarcated categories, but rather as fields that overlap with one another (cf. the different, but cognate evidential meanings of dokéō sub 4.7.). Because evidentiality concerns the signaling of the information source for a proposition, the sentences that prototypically contain evidential values are of a declarative nature (although evidentiality can also appear in imperative, interrogative, and other types of clauses; cf. Aikhenvald 2004: 242-256). Therefore, I focus on declarative sentences.
Journal of Greek Linguistics 16 (2016) 3 -46   table 1 Classification of evidential values according to Plungian (2010: 37)3 direct (ergo personal) -participatory: participation in the action -endophoric: internal sensation, will -common knowledge -sensory -visual (with subtypes: e.g., close, far, present, past) -non-visual: other senses (with subtypes: e.g., audition) -revelative: dreams indirect (a) personal -inferential: deduction based on personally observed results -presumptive: deduction based on plausible reasoning (b) non-personal -reportative (or hearsay) -quotative -common knowledge The distinction between direct and indirect evidentiality is fundamental within this semantic classification.4 Speakers express direct evidentiality when they have direct access to the information source of a proposition, i.e. via sensory, visual or non-visual, perception of certain information or via direct participation in a certain action or situation (participatory). Endophoric markers are used in some languages when speakers offer information about their own psychic or physiological status (direct access to the information by means of their own internal feeling). Revelative evidentiality concludes the list of direct evidential values (not mentioned by Plungian 2010). It designates information taken from dreams or other forms of visions (cf. Kratschmer-Heijnen 2010: 333).
Whereas direct evidentiality is always personal, indirect evidential values fall into two main categories: personal and non-personal. Inferential evidentiality is indirect and personal; in order to state something about a certain situation that they did not perceive directly, speakers rely on observable data to which they have personal access. The same goes for presumptive evidentiality; speakers employ their expectations and "logical reasoning" (= personal) to describe situations that are not directly perceived. Reportative evidentiality is, on the other hand, indirect and non-personal; an external center of consciousness constitutes the information source of the proposition (i.e. a report from someone else, also termed hearsay). Quotative evidentiality is a sub-form of the previous value; it signals that a quotation is being cited verbatim (formally, most often a morpheme or a particle).5 A last, rather exceptional evidential value is common knowledge, which can be placed in both the direct and the indirect/non-personal categories.6

2.3
Classifying Evidential Morphemes: The Formal Perspective Initially, only verbal morphology and morphemes with evidentiality as their principal function or semantics were taken into account in general linguistics. Such an approach largely excluded other means to express evidentiality from a thorough linguistic study (cf. Aikhenvald 2003;. Recently, however, less grammaticalized (and even lexical) evidential constructions have increasingly become part of the research program of linguists (cf. Cornillie 2007 for evidential values of Spanish modal auxiliaries; see also Diewald-Smirnova 2010& Hennemann 2013). These constructions are generally referred to as "evidential strategies", for their evidential value mainly comes about through contextualcollocational "side effects" (cf. Aikhenvald 2004: 105-152 for a discussion of evidential strategies). This is motivated by the fact that not every language relies on verbal morphology to express evidential values; in addition, the distinction between grammatical and lexical means is not to be regarded as a dual polarity, but rather as a gradual continuum ranging from "highly grammaticalized" over "less grammaticalized" to "lexical" (cf. Wiemer 2010: 63). Moreover, certain linguistic elements may evolve from more lexical to more grammatical over time (grammaticalization; very rarely in the reverse direction). Wiemer's (2010: 63) Journal of Greek Linguistics 16 (2016) 3-46 figure 1 Linguistic encoding of evidentiality according to Wiemer's (2010: 63) lexico-grammatical continuum. Only the horizontal direction is meaningful; the vertical dimension is arbitrary in this case.
lexico-grammatical continuum is employed as the basis in formally analyzing Ancient Greek evidential constructions (cf. Figure 1).

Evidentiality and Extinct Languages
Researchers who specialize in the study of extinct languages, such as Ancient Greek, and who rely on corpus-based methods when studying linguistic phenomena are compelled to rely chiefly on literary texts (cf. Van Hal 2010). Moreover, the corpus of elaborate low-register texts is substantially smaller than that of literary texts. Therefore, researchers need to be aware of the fact that differences may exist between written and spoken language. Chafe (1986), for example, points out that English evidential constructions occur more frequently in written academic texts than in colloquial speech. This is probably related to the fact that colloquial speech leaves much more unsaid than does academic language. For example, some pieces of information may be conveyed by facial expressions or gestures. Nonetheless, in the case of Attic, such a comparison of spoken and written language is impossible. Thus, I have opted to analyze two texts that exhibit both oral and written features (cf. sub 3.).
Until a few years ago, the notion of evidentiality and its linguistic encoding had not been directly addressed in the study of the so-called classical languages. In fact, Mišeska Tomić (2008: 212) even went so far as to explicitly deny that evidentiality could be encoded linguistically in Latin and Ancient Greek. Consequently, several linguists mentioned in this paper do not use the concept "evidentiality" in their study of certain relevant aspects of Ancient Greek (e.g., Basset 1984;. Hence, investigating evidentiality in classical languages is a recent development. Nevertheless, so far it has remained a smallscale, punctual endeavor. For example, Méndez Dosuna (1999) and Faure (2010;2014) investigate, among other options, the possibility of an evidential value for the oblique optative in Ancient Greek (cf. sub 4.6.). In addition, the "logicalinferential" value of the periphrastic construction with méllō as well as the Greek future tense has already received some attention from E. Bakker (1997: 17-23) and S. Bakker (2002), respectively. Cuzzolin (2010) offers some preliminary remarks on evidentiality in Latin. Revelative constructions in Greek and Latin are briefly touched upon by Kratschmer-Heijnen (2010). Moreover, Joseph (2003) not only takes into account extinct Indo-European languages, but also explores the possibility of evidential marking in Proto-Indo-European itself. The results of such earlier studies are discussed in relevant sections of this paper (sub 4.). Whereas most of the abovementioned scholars concentrate on one specific construction, the present contribution aims at offering a more integrative approach.

Methodology
This research takes a case study approach to investigating the existence of evidential markers and strategies in Plato's Apologia Socratis ("Apology of Socrates"; as) and Crito (Cr), which are both Attic writings from the beginning of the fourth century bc. These constitute an adequate and complementary starting point. The as is an argumentative text, containing Socrates' orations at his 399 bc process, as they are presented by Plato. However, it also contains dialogic passages (cf. the anecdote about Callias and his sons and the dialogue between Socrates and Meletus). Because it comprises both argumentative and dialogic elements, it is taken as the main text. In order to include more dialogic passages in the sample, I also consider a short Platonic dialogue (Cr) that encompasses additional evidential features not present in as. It stages Socrates and his friend Crito, who informs him that the day of his execution is near and offers him a last-minute escape from Athens. Socrates refuses, as he does not want to commit any injustices. The corpus thus consists of two early works of Plato.
Occasionally and when relevant, references are made to later works by Plato and to other, mostly Attic, writers. Consequently, the conclusions drawn in this paper concern, in the first place, the Attic speech variety of that time (and Plato's idiom in particular). Their applicability to other varieties of the Greek language (ancient, medieval, and/or modern) remains to be determined by further research.7 The case studies of evidential markers and strategies have been drawn from the whole spectrum of Wiemer's (2010: 63) continuum (cf. Figure 1 above), from particles (situated at the left, lexical end) to the extension of a modality paradigm (oblique optative; situated at the right, grammatical end). This approach allows for an adequate exemplification of the variety of means Plato's Attic can rely on to indicate the information source of a proposition. Each section generally consists of the following three components: 1) A succinct overview and, if applicable, a discussion of relevant secondary literature (dividable in two groups: studies explicitly referring to the notion of evidentiality and studies which indirectly touch on it; cf. sub 2.4.). 2) An interpretation of the formal features discussed, as they are found in our sample and starting from the following central research questions: a. Does the formal feature have evidential meanings or side effects (evidential strategies)? b. Which evidential value is expressed or generated by the formal feature in question? c. Is the evidential value part of the semantic core of the feature (= evidential marker), or is it generated epiphenomenally (= evidential strategy)? Where possible, the conditions under which this side effect occurs are taken into consideration. d. To which extent do the evidential values of these formal features interfere with other conceptual domains (e.g., modality, mirativity)? This research question is addressed only peripherally. 3) Where possible, I formulate generalizations on the evidential status of the linguistic means in question.

Evidentials in Modern
Greek have already received some attention. See, e.g., Friedman (1999) for evidentiality in the Balkan Sprachbund. In addition, see Ifantidou (2005) and Markou (2011) for the reportative value of the particles ðiθen and taxa, "supposedly". The former also seems to have had evidential meanings in Ancient Greek (dêthen). However, it only appears once in a later work by Plato (cf. Politicus, 297c), where a presumptive value seems to be combined with emphatic irony (cf. Denniston 1954: 264-266;esp. 266

Particles
Greek is well-known for its abundance of particles, which have received extensive attention. In fact, one of them has already been explicitly connected to evidentiality: ára (Bakker 1997: 17-23). Other particles discussed in this section are usually interpreted without being linked directly to evidentiality (cf., e.g., some of the more intuitive analyses of Denniston 1954). However, based on my reading of the corpus texts, at least five particles qualify for an evidential value: ára, dḗpou, dêta, oukoûn, and pou. A sixth, dḗ, may be added, although its precise status remains disputed. Table 2 provides an overview. The particle ára is inferential; it signals that Socrates' utterance is based on data he personally perceived (in (2), it is based on Meletus' argumentation).9 The inferential value seems to be intensified by the use of the parenthetical phrase "hōs éoiken".10 This mitigating ("hedging") and evidential collocation occurs approximately 50 times in the entire Platonic corpus and a few times in the works of other Attic writers (Aristophanes, Euripides, Sophocles, Xenophon; once in each), which seems to point to an idiosyncrasy of Attic (or even of Plato). Specifically, it suggests a lower degree of epistemic certainty on the speaker's part about the contents of the proposition.11 In addition, the particle may have a mirative extension (cf. sub 2.1. for mirativity), as possibly in (2); Socrates is (ironically) surprised to hear that he alone is corrupting youngsters and that all others are making them better. This seems to indicate that ára is an evidential marker with inferential evidentiality as its semantic core, interfering with a hedging function (indicating lesser epistemic certainty) and possibly having a mirative extension. 9 For an example from Sophocles, cf. Oedipus Coloneus 1744-1746: "[Antigone] mógos ékhei.
The value of dḗ is problematic. Although it may generate an inferential evidential value, it certainly has a discursive-deictic function. It recapitulates what precedes the clause in which it appears (Denniston 1954: 239), or indicates that the evidence for the proposition is shared by the speaker and hearer (it is "obvious").12 As in (3), the evidence for the necessity of Socrates' action to defend is accessible to both the speaker and hearer, for it is based on previous elements in his process and his speech. Here, there is a thin line between inference and recapitulation of information known to the speaker and hearer. Not only is the need for a defense obvious to both; it also constitutes a piece of information that is inferred on the basis of the external circumstances. In this case, however, inference appears to be a contextual side effect of the deictic function of the particle (= possible evidential strategy).13 A third particle, dḗpou, clearly points to presumptive evidentiality. In (4), Socrates is proposing a hypothesis in the irrealis. He presumes that if he were a stranger, his hearers would forgive him. There is no direct evidence to support his proposition; he is basing this notion on his own expectations. This passage (along with the other 8) clearly establishes dḗpou as an evidential marker. Formally, it consists of the particle dḗ and the relativizing enclitic pou, which mitigates the obviousness of the information expressed by dḗ and can also have a presumptive meaning itself (cf. infra). A fourth particle also has the element dḗ in it: dêta. It expresses an unexpected inference (and even indignation), which points to a mirative use (cf. Denniston 1954: 272). This use, however, seems to be mainly (but not exclusively) limited to interrogative contexts, referring to an inference that will be made in what follows (Denniston 1954: 269 et sqq.). In declarative sentences, it signals a strong confirmation of the information provided in the proposition.14 Therefore, the mainly deictic particle dḗ not only may have an inferential value epiphenomenally, but also constitutes the formal basis for two other markers, with evidentiality as (a part of) their semantic core.

12
I thank one of the anonymous reviewers for pointing out the importance of deixis and information status for the value of this particle. The interaction between deixis, information status, and evidentiality is in need of further investigation. 13 Inference is not always present, cf. 27c: "ei dè daimónia nomízdō, kaì daímonas dḗpou pollḕ anágkē nomízdein mé estin: oukh hoútōs ékhei? ékhei dḗ: títhēmi gár se homologoûnta, epeidḕ ouk apokrínēi." Here, dḗ merely recapitulates information that is suggested in the preceding question and evident to both speaker and hearers. 14 Cf. Cr 49b, where Crito is confirming Socrates' inference "oudamôs ára deî adikeîn" with the reply "ou dêta".
Apart from indicating recapitulation, the interrogative particle oukoûn (< negation ouk + oûn) also denotes an inference within an interrogative context.15 In (6), Socrates is inferring Meletus' statement on the basis of the latter's words. Often in such cases, the speaker considers the contents of his question to be true (cf. Denniston 1954: 434, who considers this use "strictly inferential").
The original meaning of the enclitic pou is clearly local ("somewhere"). Among other things, a presumptive meaning developed out of this, which was already present in Homer.16 In (7), Socrates is relying on his own reasoning and expectations to assert that he has to die the day after the ship arrives. This enclitic particle probably conveys a lower degree of epistemic certainty than does the cognate particle dḗpou, as it lacks the parameter of "obviousness" conveyed by dḗ. This indicates that the evidential values of pou and dḗpou are interconnected, but with different degrees of epistemic certainty, pointing to interference of modality with evidentiality (cf. sub 2.1.). The presumptive meaning of both particles appears to block their collocation with the confirmative particle oûn (cf. Denniston 1954: 493).
Thus, the abovementioned findings seem not only to establish some particles (ára, dḗpou, oukoûn, pou) as evidential markers but also to confirm the link with mirativity as characteristic of inferential evidential morphemes crosslinguistically (in the case of ára and dêta). The Attic data are well suited to the general typology of evidentiality.17 The evidential value of dḗ seems to be a side effect of its main discursive-deictic function (evidential strategy). Two particles (dêta and oukoûn) are only used evidentially in interrogative contexts.18 The ways in which the several inferential particles differ remain to be determined. Nonetheless, the status of the proposition is clearly of importance in this regard (statement vs. question).19 15 The particle oûn is not taken to be evidential on the basis of the instances in the sample; it functions mainly as a connective or as a confirmatory particle (cf. also Denniston 1954: 415 et sqq.). 16 Cf. Ilias 3.308-309: "Zdeùs mén pou tó ge oîde kaì athánatoi theoì álloi / hoppotérōi thanátoio télos peprōménon estín." 17 Cf., e.g., the inferential suffix -l'el in Kolyma Yukaghir, a linguistic isolate (or possibly a Uralic language) from East Siberia, which also displays a mirative extension (Maslova 2003: 173). 18 According to Rijksbaron (1976: 77 et sqq.), the conjunction epeí, in some of its uses, can also be regarded as an inferential particle. However, it remains to be determined whether this assertion holds and to what extent this connotation can be viewed as functionally evidential rather than semantically causal. 19 I thank one of the anonymous reviewers for drawing my attention to this question.

4.2
Complementizer Strategies: Participial Clauses A number of well-known constructions may denote several evidential values that are intertwined not only with the semantics of the main predicate, but also with the way it is complemented syntactically. Two main semantic groups to be dealt with here are verbs of perception and declarative verbs. The former are treated in this section, and the possible complementizing patterns of the latter category are presented in Section 4.3.
In 22c, Socrates repudiates the poets because of their boasting and their fallacious claims of wisdom. Because this is a fact he himself experiences in the streets of Athens, he uses a first person verbal form and a genitive and participle construction (genitivus cum participio; abbreviated as GcP) to express his personal observation of their arrogance: Exactly the same verb as in (8) is used to relate a different mode of knowing. Whereas Socrates himself had witnessed the information expressed by the GcP in (8), his evidence for Evenus' stay in Athens (9) is indirect and based on hearsay, which is rendered by the use of an accusative and participle (accusativus cum participio; abbreviated as AcP). These constructions are well-known to grammarians of Ancient Greek, who describe them adequately, but fail to identify them as evidential strategies conveying different evidential values.
The clearest example of a verb of perception with different complementizing patterns in the sample is aisthánomai: "to observe personally (+ GcP); to obtain information indirectly through hearsay (+ AcP)". Depending on the case of the noun and predicative participle following the main verb, the construction expresses a direct, sensory (i.e. visual) evidential value (genitive; cf. example (8)) or hearsay (accusative; cf. example (9)), provided that the main verb itself is constructed in the first person.20 An example from Cr shows that the direct, perceptual mode of knowing can also be expressed by an indirect question after aisthánomai.21 A verb with a similar semantico-syntactic pattern is akoúō: "to hear directly (+ GcP); to hear via an intermediary (+ AcP)". Therefore, the genitive is used for direct, sensory modes of knowing, whereas the accusative is reserved for indirect, hearsay modes of knowing with verbs of perception. Other constructions with akoúō can also convey evidential meanings.22 20 If not, the construction does not refer to the information source for the proposition, but rather reports an observation (genitive) or an acquisition of indirect information by the subject of the main predicate (accusative). 21 Cf. Cr 43b: "allà kaì soû pálai thaumázdō aisthanómenos hōs hēdéōs katheúdeis." For a perception verb with completive hōs, cf. sub 4.3.2. and-more specifically-example (12). 22 Cf., e.g., akoúō + accusative of the object + genitive of source (hearsay) in Homer, Odyssea 12.389: "taûta d'egṑn ḗkousa Kalupsoûs ēükómoio." Journal of Greek Linguistics 16 (2016) 3-46 A similar case might be detected in the different constructions after verbs with sensory meanings other than audition. However, no examples of this construction were found in the sample. A quick search through the tlg reveals that there indeed exists a participial construction for visual evidence, i.e. the verb horáō, "to see", followed by an AcP, constituted by an accusative noun and a predicative present participle. See, e.g., the following sentence from Aristophanes' Ecclesiazusae 27-28: In contrast, the same verb with an aorist participle appears to be an evidential strategy, because it expresses an inferential value epiphenomenally (being a side effect caused by the semantics of the main verb and the aorist (perfective) aspect of the participle; cf., e.g., Euripides' Hecuba 733-734: "tín'ándra tónd'epì skēnaîs horô / thanónta Trṓōn?", where Agamemnon infers on the basis of the dead body that the man in question has died there, near the tents). The abovementioned participial phrases clearly have evidential side effects; they are evidential strategies.
Summary: Different complementizers with verbs of perception appear to have evidential side effects (evidential strategies). A genitive and participle construction points to direct observation, whereas the accusative counterpart epiphenomenally expresses hearsay.

4.3
Complementizer Strategies: The Functional Opposition between hóti/háte and hōs The opposition hóti-hōs on the one hand and háte-hōs on the other comprises three different functional oppositions: i.
the contrast between completive hóti and hōs; ii. the contrast between non-completive hóti and hōs, and iii. the contrast between háte and hōs as conjunctions introducing participial clauses.

4.3.1
Completive hóti and hōs In the past decades, the exact value of completive hóti and hōs has received much attention. Table 3 provides a schematic overview of the most important hypotheses that have been designed to account for this opposition and which are often based on the study of only one author or on some arbitrarily selected examples. The present section aims to investigate (i) the value of the opposition between these two Attic complementizers, which is inextricably intertwined with the semantics of the main predicate, and (ii) whether this much-studied contrast has an evidential semantics. In general, linguists seem to agree that hóti denotes an actual fact in a neutral way, whereas hōs is connected-however vaguely-with subjectivity. It is clear that there is a certain functional contrast between the two complementizers, the precise nature of which still remains rather opaque and probably differs diachronically and dialectally. In what follows, the two most recent hypotheses acknowledging the contrast between hóti and hōs will take center stage, i.e., that of  and that of Cristofaro (2008). According to , hōs signals that the subject of the main verb itself constitutes the source of information for the contents of the completive phrase (= "subjective" value); hóti, on the other hand, presents information taken over from others and/or based on external facts. The author bases her hypothesis principally on a study of Herodotus, but also offers a brief verification of her hypothesis by means of examples from Homer, Aeschylus, Sophocles, Euripides, Thucydides, and Plato. Ultimately, she concludes that Attic has undergone a semantic shift; after verbs of speaking, hōs points to a lack of the speaker's commitment (abbreviated as sc). Cristofaro (2008: 589), on the other hand, argues that the distinction between the two complementizers is largely pragmatic; hóti-clauses express information "with a high communicative value" ( focus or rhema), whereas the communicative value of hōs-clauses is said to be low. Therefore, hōs-clauses have two distinct, but related, meanings: (i) for expressing information already known (topic or theme) and (ii) for indicating that the speaker does not want to take the responsibility for the contents of the completive clause. Thus, hōsclauses do not contribute to "the further development of the communicative process" (Cristofaro 2008: 588). If hóti and hōs are interchangeable without a difference in meaning, this is said to be due to the pragmatically intermediary status of the situations described. In what follows, the two hypotheses are tested against the evidence of the sample and discussed within the framework of evidentiality. The data of the sample reveal that the information conveyed by completive hóti-clauses receives the speaker's commitment (as in (13)). It is also the typical van rooy Journal of Greek Linguistics 16 (2016) 3-46 complementizer following verbs of knowing and considering (cf., e.g., 21d and 29b), which are never completed by hōs-clauses in the sample. This seems to be a general tendency. A preliminary search in the texts included in the Perseus under PhiloLogic database reveals only 162 examples of the lemma oîda + hōs, in which cases hōs mostly functions as an indirect interrogative adverb rather than as a complementizer. In comparison, there are 1272 instances of oîda + hóti.23

Analysis and Discussion
There are clear instances in which hóti contrasts with completive hōs. In certain cases, hōs indicates that there is no sc, occasionally provoking an epistemic extension of uncertainty. This connotation only seems to occur when the main predicate is a verb of saying or a verb of perception. See the analysis of Neuberger-Donath (1982) discussed above and (11) In this case, the perception verb + hōs construction not only indicates that the information expressed in the hōs-clause is directly heard by the subject of the verb of perception from another center of consciousness. It also seems to signal a lack of sc, as the words "oudè toûto alēthés" make explicit (Socrates does not agree that he is trying to educate people, while asking money for it).
With regard to the instances in the sample, it remains unclear which of the two hypotheses is correct. Nevertheless, a tentative answer may be possible; see, e.g., (11)  When a speaker expresses his source of information based on sensory (visual, auditory, etc.) observation and, in some cases, on hearsay (e.g., with aisthánomai + AcP; see (9)), it is logically necessary that the first person is closely connected to the main verb of the sentence in order to demonstrate a truly evidential value (most often as the subject of the main verb or as a dative pronoun with impersonal verbs). This is, however, not the case with akēkóate in (12).  (13) with (11), hōs seems to be used (i) to offer information that is already known and does not help forward the action (in accordance with Cristofaro 2008) and (ii) to mark the subject of the predicate governing the hōs-clause as the source of information for the contents of the completive clause (in accordance with Neuberger-Donath 1982). As discussed above, hōs also implicates a lack of sc (Neuberger-Donath 1982 and Cristofaro 2008), resulting in an epistemic extension of uncertainty and a connotation of indirect reportative evidentiality. These two nuances are limited to verbs of speaking, whereas the lack of sc also appears after perception verbs + completive hōs (as is clear from our analysis of (12)).25 When hōs follows a verb of showing, it indicates that the subject of the verb governing the completive clause constitutes the information source for the contents of the hōs-clause; see (15) and . In addition, the information in the hōs-clause has a topic value (Cristofaro 2008). On the other hand, when hóti follows a verb of showing, it indicates that the information has not sprung from the subject of the main verb, but is founded on external factors (in (14) the actions of Meletus; Neuberger-Donath 1982). However, the information introduced by hóti in (14) does not seem to have a high communicative value (pace Cristofaro 2008), as the hóti-phrase discusses information already mentioned by Socrates. This may suggest that the focus value of hóti-clauses is a frequent pragmatic effect of this complementizer.
Following : hōs indicates that the subject of the main predicate is the source for the information offered in the completive clause. The connotation of a lack of sc is, however, blocked by the first person subject. Neuberger-Donath's (as well as Cristofaro's) hypothesis needs to be corrected in this regard; lack of sc only arises in non-firstperson contexts. ii. Following Cristofaro (2008): hōs marks the introduction of information with topic value that is already known from the accuser's speech (in addition, the presence and deictic function of "taûta" signals that the hōsclause conveys known information).

Conclusion
The few instances present in the sample cannot yield a definitive answer as to which of these hypotheses is the correct one or whether they are tenable at all. However,  seems to have penetrated more deeply into the functional value of the hóti-hōs distinction. Cristofaro's (2008) explanation only holds true for a pragmatic effect. A tentative description of the interaction between the complementizers on the one hand and the several semantic types of verbs on the other is offered in Table 4 below. hóti becomes the typical complementizer after verbs of knowing and considering (thus initiating a setback of hōs as a complementizer in later times; see Jannaris 1968: 412). Broadly speaking, the hóti-hōs opposition shares common ground with evidentiality in that the complementizers are able to signal the nature of the relationship between the information expressed by the completive clause and the subject of the main verb. In particular, when verbs of saying are combined with these two complementizers, this construction conveys both the evidential value of hearsay and different degrees of epistemic commitment (lack of sc with hōs in non-first-person forms).27 Thus, it seems justified to state that the distinction between these conjunctions is to be situated on the interface between epistemic modality, information status (pragmatics), and evidentiality.
Other approaches may also be useful in interpreting the opposition between hóti and hōs, such as that of the "privative opposition" (cf. Clinquart-Isebaert 1984, where it is successfully applied to Ancient Greek linguistic features other than hóti and hōs). Because hóti is diachronically the first complementizer, and hōs only gradually obtained a complementizing function, there may have existed a so-called opposition privative between hóti and hōs. Within this framework, hóti would be the unmarked neutral term, solely expressing subordination, and hōs would be the marked term, which is used when there is a relationship of "interiorization" between the information presented in the completive clause and the subject of the verb governing it (evoking a reportative evidential value combined with a lack of sc after verbs of speaking in non-first-person forms). This approach can be employed to map out certain Ancient Greek syntactic structures, as proven by the analysis of consecutive phrases. These either contain an infinitive verb form (the older structure & the neutral term), which expresses both factual and non-factual consequences,28 or a finite verb form (the more recent structure & the marked term), which only denotes factual consequences.29 20: "légousi dé tines iētroì kaì sophistaí hōs ouk éni dunatòn iētrikḕn eidénai hóstis mḕ oîden hó tí estin ánthrōpos." 27 In first person forms, the subject of the main verb of saying constitutes the source of information. This blocks a possible connotation of "lack of sc". 28 The infinitive is thus indifferent to the factuality of the consequence. 29 Consult Clinquart-Isebaert (1984) for further details and other examples of oppositions privatives in Ancient Greek.

hóti and Other Quotative Markers
As a side note, I must record one clearly evidential value of hóti, derived from its completive use. In the sample, there are ten instances where hóti is used as a quotative marker: In (19), the fact that the speaker is citing himself shows that quotative evidentiality does not always have to be indirect and non-personal; it can also be direct (ergo personal) in past contexts. Quotative evidentiality is part of the semantic core of this defective and highly reduced verb, which is subject to a number of syntactic restrictions, including the following: i.
The use of this verb is strictly limited to quotations and it cannot be accompanied by an accusative and infinitive clause nor by quotative/ 30 It has a limited number of forms-the first person present form ēmí and the past tense collocations ên d 'egṓ, ê d'hós, and ê d'hḗ. Journal of Greek Linguistics 16 (2016) 3-46 completive hóti, probably marking that the defective verb has absorbed the quotative function. ii. If ēmí is used in the past (imperfect) tense, an Attic speaker is obliged to mark the grammatical person in order to give more substance to the verb form, which is reduced to one syllable in the first person (ên) or even to one vowel in the third person (ê). The personal pronoun is always preceded by the elided adversative particle d' (< dé). As a rule, the ensemble [d' + personal pronoun] follows the verb form.
Summary: hōs marks a specific relation between the subject of the main verb and the information relayed in the completive clause. The subject of the main verb is presented as the source for this information and can, in non-first-person forms, convey a lack of the speaker's commitment (mainly after verbs of speaking). hóti, on the contrary, does not signal such an evidential overtone. As a side note, the quotative values of hóti and ēmí are briefly discussed.

háte and hōs & hóti and hōs Introducing Adverbial Clauses
A study of the semantics of the conjunctions háte and hóti on the one hand and hōs on the other makes clear that this distinction is less ambiguous than that between completive hóti and hōs. This is also suggested by the limited amount of secondary literature on this topic.31 Muchnová (1990Muchnová ( -1992 discusses the causal-completive use of hóti with verbs of feeling. The most extensive treatment of these conjunctions is found in Rijksbaron (1976). He asserts that non-completive hōs-clauses do not express a real reason or cause; rather, they indicate indirect discourse, occasionally giving rise to a causal connotation (depending on the context).32 Causal hóti-clauses, just as completive hóticlauses originating from a relative pronoun meaning "the fact that" (Monteil 31 Some grammars do not even discuss a potential semantic contrast between háte and hōs + participle (e.g., Humbert 1960). 32 See Rijksbaron (1976: 120-121). hōs + finite verb is said to have developed by analogy with causal hóti (Monteil 1963: 358-359). The original fundamental meaning of hōs was clearly an instrumental one ("the way in which, the means by which"; cf. Monteil 1963: 341-342), from which the reportative as well as the completive use were derived via so-called bridging contexts (i.e. contexts in which hōs may be ascribed both an instrumental and a reportative/completive value; cf. Monteil 1963: 354-355). The methodological principle of bridging contexts is a useful means for analyzing the diachronic evolution of semantic and syntactic evolutions; cf., e.g., Bloem (2008 1963: 248-251), appear to possess four characteristics that are lacking in hōsclauses with finite verbs (Rijksbaron 1976: 146). hóti-clauses can be i. used to answer questions introduced by tí ('why?'), ii. coordinated with other "causal adjuncts", such as prepositional clauses, iii. accompanied by an anaphoric or a cataphoric prepositional phrase, and iv. followed by a parenthetical phrase relativizing the contents of the clause.33 Rijksbaron (1976: 151 & 205, note 45) also accounts for the contrast between háte and hōs + participle. He argues that the latter construction first marks that someone else's words are reported, with the causal connotation being a contextual side effect. hōs + participle signals that the information is based on the opinion-or rather the knowledge-of the subject of the main predicate, which may be fallacious (Rijksbaron 1976: 153-154 & 205, note 49). To conclude, Rijksbaron (1976: 155) mentions a semantic restriction on the use of hōs + participle that supports his analysis: only main verbs that have a subject with the semantic role of agent can be accompanied by this participial phrase.34 háte + participle, on the other hand, refers to factual causes recounted from the point of view of the speaker or narrator.35 Oguse (1962: 204-210) offers 33 Rijksbaron (1976: 147) also sees a pragmatic distinction between hóti and hōs; non-completive hōs introduces a new phase in the story, whereas causal hóti provides information that is already known. This is an inversion of the pragmatic distinction Cristofaro (2008) suggests for completive hóti and hōs (she does not mention Rijksbaron's study; cf. sub 4.3.1.). It remains to be determined whether these two approaches are reconcilable. Initially, Rijksbaron's thesis, which is based on a partial analysis of Herodotus' Historiae, does not seem tenable; compare Herodotus 1.34.1: metà dè Sólōna oikhómenon élabe ek theoû némesis megálē Kroîson, hōs eikásai, hóti enómise heōutòn eînai anthrṓpōn hapántōn olbiṓtaton (in which the hóti-clause offers new information) with 1.79.2: enthaûta Kroîsos es aporíēn pollḕn apigménos, hṓs hoi parà dóksan éskhe tà prḗgmata ḕ hōs autòs katedókee, hómōs toùs Ludoùs eksêge es mákhēn (in which the hōs-clause contains information already known). 34 Rijksbaron (1976: 155 & 206, note 57) refers to the (lack of) grammaticality of the following English sentences: Supposing that the rain had stopped he rushed out into the garden. vs. *Supposing that the rain had stopped he fell into the garden. The subject of the last sentence is an experiencer and not an agent; consequently, the participle construction supposing that, the English equivalent of hōs + participle, cannot be used within this context. Therefore, the use of this construction is much more limited than that of háte + participle (Rijksbaron 1976: 156). 35 háte + participle owes its causal value to the originally comparative and appositive use Journal of Greek Linguistics 16 (2016) 3-46 a similar explanation for the value of háte + participle. However, he adds that the meaning of háte is different when used with an irrealis or a potentialis. In this case, háte + participle expresses a premise-or at least a certain form of supporting information-which is needed in order to follow the hypothetic reasoning in question. In the next paragraphs, the instances of non-completive hóti and hōs and the conjunction háte as they appear in the sample are discussed.
Although hōs + finite verb does not appear in the sample (predominantly used in older Attic and Ionic; cf. Jannaris 1968: 408), causal hóti is found six times; it is followed by the indicative in all instances. In each case, the speaker commits himself to the cause he states: háte + participle occurs three times in the sample (as 23d, 39b (2 ×)). It indicates that the speaker takes responsibility for the reason expressed in the participial clause, which is reinforced by the parenthetical use of oîmai ("methinks") after háte in 23d, indicating that the speaker is the source of the information:  (Monteil 1963: 244-245).
'Thus, since they, methinks, are ambitious, impetuous, and numerous, and since they speak vigorously as well as persuasively about me, they have filled your ears both long ago and whilst accusing vehemently.' 23d-e The sample contains 12 examples of hōs + participle. It denotes in all instances that the speaker distances himself from the information offered by the participle, and that he attributes it to the subject of the main predicate (e.g., 17a-b) as long as he is not the subject of the main predicate (as in 40a: humîn gàr hōs phílois oûsin epideîksai ethélō, where hōs is used to stress that the speaker is the source of the information conveyed in the participial clause): In (22), the reportative status of the evidence is confirmed by the use of a verb of speaking followed by completive hōs, indicating a lack of sc. Rijksbaron's (1976) suggestion that this construction first signals reportative evidence (and not a cause) clearly seems to coincide with the data from the sample. Nonetheless, it must be noted that this is only the case when the main subject's identity does not coincide with that of the speaker.
With non-first-person main predicates in Attic, hōs + participle is consistently used when the information is ascribed to the subject of the main predicate (for which the speaker bases himself on indirect evidence: hearsay ["claiming that"] or inference ["based on the actions of the subject, they must be believing that"]) and the speaker does not agree with the information offered (lack of sc). With first person main predicates, the speaker himself constitutes the information source. Non-completive hóti and háte generally indicate that the speaker commits himself to the information, without necessarily having an evidential side effect (evidential strategy). Table 5 provides an overview of the use of these conjunctions by Plato.
Journal of Greek Linguistics 16 (2016) 3-46 As shown, the use of different cases in complementizing patterns has evidential implications (4.2.). Similarly, the oppositions hóti vs. hōs + finite verbs and háte vs. hōs + participle also seem to have evidential extensions in certain contexts (mainly reportative, but also inferential). The precise extent of the evidential value(s) of these conjunctions remains to be determined by more extensive corpus research. phrase, invoking a reportative evidential value and a lack of sc with nonfirst-person main verbs. háte + participle refers to causes represented as factual by the speaker. hóti + finite verb presents a cause neutrally or possibly indicates the sc; no clear example of hōs + finite verbs was found.

4.6
A Difficult Case: The Oblique Optative Revisited Many analyses of the value of the oblique optative have been suggested. Méndez Dosuna (1999) and Faure (2010;2014) independently discuss the possibility of an evidential value for the oblique optative. The latter seems to exclude it, whereas the former seeks to confirm it. Both refer to the work of Basset (1984;, without further elaborating upon his views. Some earlier twentiethcentury grammarians have arrived at an intuitive consensus regarding the reportative character of the oblique optative. For reasons explained in Section 2.4., they were not yet able to link this with the notion of evidentiality. Smyth (1916: 336), for example, informs us that "causal clauses denoting an alleged or reported reason take the optative after secondary tenses".37 Kühner-Gerth (1966: 548) offers a similar explanation for causal clauses. According to these authors, the oblique optative is used to express "past thoughts and speeches", the original "potential semantics" of the optative being "completely obscured" (Kühner-Gerth 1963: 254-255, where a parallel with the German Konjunktiv ii is drawn). Neuberger-Donath (1983), on the other hand, suggests that the oblique optative has a mirative value, a semantic category close to evidentiality (cf. sub 2.1.). The Greek oblique optative is purported to indicate that the speaker is surprised-post factum-at the information offered in an instance of indirect speech. Because of this post factum character of the speaker's surprise, the oblique optative can only be used following the past tense.38  Basset (1984; states that an oblique optative marks the source for the whole proposition. Specifically, this modal construction delimits the source as particular (personal judgments, arguments, and information), whereas other moods mark more general sources (imposed judgments, natural arguments, and public information) (Basset 1986: 110-111). Because the oblique optative always occurs in subordinate clauses governed by past-tense main verbs, it is described as an "escape from nynegocentrism" (< nûn and egṓ ;Basset 1986: 111). For it allows the speaker to put the contents of the proposition in a situation removed in time from the speaker (an escape from the current moment: nûn) and, possibly, entirely detached from his own person (an escape from the self: egṓ). Thus, this use of the optative can be seen as a deictic strategy with possible evidential implications.
Starting from an analysis of the sample (17 instances of oblique optatives),39 I aimed to investigate the extent to which Basset's hypothesis is tenable. His suggestion (Basset 1986: 110-111) that the oblique optative expresses specific personal information, thoughts, or judgments appears to be an adequate approach to the passages in the sample; cf., e.g., (23) The two oblique optatives are used to articulate the particular stance of the speaker, which is clearly not shared by the person to whom the pronoun "autôi" refers. On the one hand, the oblique optative seems to delineate the source for the proposition; it delimits certain information from others and endorses its particular character, whereas the indicative and the subjunctive 39 Cf. as 21a (2×), 21c (2 ×), 22a, 22b (2 ×), 22c (2 ×), 22d, 22e, 27e, 32c, 34c, 36c, and 40a and Cr 45b. 40 As one of the anonymous reviewers rightly pointed out, it would be worthwhile to investigate whether there is a correlation or interaction between the usage of optative oblique constructions and the distribution of the complementizers hóti and hōs. This, however, exceeds the scope of the present paper.
generally convey commonly accepted views. On the other hand, the oblique optative constitutes a means for the speaker to distance himself from the words of the protagonist, which Basset (1984) calls "la dissociation énonciative". The speaker ("le centrage locutoral") renounces the utterances of the protagonist ("le centrage protagoniste"), who differs at least situationally from the speaker and possibly on the basis of identity (for the terms centrage locutoral and centrage protagoniste, see Basset 1999). Indeed, this seems to be an important restriction on the use of the oblique optative; this mood can only be used when the contents of the proposition are represented from the point of view of the protagonist as particular to someone else or to himself in a different (i.e. past) situation (cf. Basset 1984).
The instances in the sample seem to point out that the oblique optative constitutes a deictic strategy, "delimiting" the information, as is suggested by Basset in several of his publications. The speaker distances himself in time (and possibly in identity) from the subject of the main predicate. Because it is often impossible to identify the speaker with the protagonist, a number of early twentieth-century grammarians have attributed to the oblique optative a value having a close resemblance to reportative evidentiality. This appears to be too radical a generalization, which is based on a limited group of similar examples. The semantics of this specific construction have a higher complexity and do not simply denote reportative evidentiality. Thus, at most, this value may be regarded as a contextually bound evidential connotation of the oblique optative, but not as being part of its fundamental function (possible evidential strategy). The complex interaction between the conceptual domain of evidentiality and the use of the oblique optative as a narrative device is, however, still in need of a thorough corpus-based investigation. Moreover, in analyzing the instances in the sample, it is important to account for phenomena such as the consecutio modorum and to determine the counterpart of the oblique optative in case of a present tense main verb (indicative, subjunctive with án, etc.).41 However, this does not lie within the scope of the present paper. In any case, it is clear that the oblique optative signals that the speaker/narrator in the hic et nunc moment of speaking/narrating does not vouch for the information expressed by others or by himself in past contexts. This may be regarded as an evidential overtone of the deictic function of the oblique optative. 41 I thank one of the anonymous reviewers for pointing out the relevance of the consecutio modorum for this discussion.
Journal of Greek Linguistics 16 (2016) 3-46 Summary: The oblique optative is said to be evidential in as far as it signals that the speaker/narrator in the hic et nunc moment of speaking/narrating does not commit himself to the information expressed by others or by himself in past contexts.

Evidential Values of Attic Auxiliaries
Just as is the case for Spanish auxiliaries (cf. Cornillie 2007 for a detailed discussion & 2009: 51-54 for a case study of the hearsay meaning of parecer, 'to seem'), Attic auxiliary constructions are commonly employed to express evidential values.42 This more grammatical means is prominent in the sample. Table 6 presents an overview of the evidential use of the Attic auxiliaries dokéō, éoika, and phaínomai, whose evidential values are intertwined with the construction in which they appear.
First, the inferential use seems to constitute the semantic core of dokéō (cf. (24); it is the only one present in Homer) and is also attested in parenthetic phrases. In this case, dokéō is always constructed impersonally with a first person dative personal pronoun (moi) and with an infinitive (if dokéō is not utilized parenthetically). Second, the reportative use seems to be mainly limited to attributive participial constructions with dokéō in the sample (cf. (25); see also the intuitive analysis in Liddell-Scott-Jones 1940: sub voce). When expressing this evidential value, dokéō is always followed by an infinitive, but does not have a dative pronoun. Third, presumptive dokéō is attested five times (cf. (26)), but it also appears in main clauses, where it is followed by an infinitive. Here, dokéō is always accompanied by a dative personal pronoun in the first person, which is necessarily co-referential with the subject of dokéō [= the speaker].
Inferential éoika is always followed by an infinitive in main clauses (with the exception of elliptical parenthetic phrases; cf. (27)). However, if it is construed with a dative, it commonly has the meaning "to resemble". It can be used personally and impersonally in its evidential-inferential meaning. The construction [phaínetai (+ moi) + infinitive] (which is used personally and impersonally) both expresses inferential evidentiality and indicates that the contents of the following complement possess a higher degree of certainty compared to the construction [dokeî + first person dative personal pronoun + infinitive]. If, on the other hand, phaínomai is used in the first person, it receives a participa-tory value (with the evidence for the proposition being one's own participation in the action), the "demonstrative" semantics of the verb being predominant. The verb is not an auxiliary in this value.
In summary, the semantics of éoika seem to allow only one evidential value, i.e. inference. Moreover, phaínomai, depending on the grammatical person in which it is used, encodes inferential and participatory evidentiality. dokéō can have three different indirect evidential values, to wit, inferential, presumptive (both personal), and reportative (non-personal). In the sample, the inferential value is used most frequently, which probably stems from the argumentative nature of the text. The speaker uses the inferential auxiliary structure with dokéō as the discursive strategy par excellence to relativize his own statements.
Inferential evidentiality is part of the semantic core of these verbs-for which typological parallels can be found in other, non-Indo-European languages; see, for example, the "secondary" verb awine/awa, "to seem", in Jarawara.43 The other evidential values of dokéō and phaínomai may be derived from the original inferential meaning. Diachronic research may shed light on the evolutions these auxiliaries went through and on the interactions between their different meanings.
Summary: dokéō, éoika, and phaínomai all express inferential evidentiality. Inferential dokéō and phaínomai signal different degrees of epistemic certainty. In addition, dokéō is used to mark presumptive and reportative evidentiality, depending on the construction in which it figures. phaínomai also reflects participatory semantics in certain contexts. 43 See Dixon (2004: 232-233) for the inferential value of this verb. Jarawara is an Arawá language spoken in Amazonas, a western Brazilian state. The so-called secondary verbs in Jarawara are part of the predicate, acting as grammatical elements that complete the main verb but preserve an independent phonological status (Dixon 2004: 226).

Conclusions and Outlook
By focusing on a number of particles, complementizer strategies, the oblique optative, and auxiliaries in Attic, this paper has attempted to demonstrate the importance of evidentiality as an integrative approach for the descriptive analysis of Ancient Greek. Both evidential markers (e.g., ára and dḗpou) and strategies (e.g., the uses of dokéō, complementizer strategies) are shown to have been available in Attic. My tentative approach to evidentiality in Ancient Greek is mainly based on the analysis of Socrates' orations at his process, as recounted in Plato's as, an argumentative text with a peculiar ironic character, in which inference is a prominent evidential value (expressed by several particles and auxiliaries; cf. Bakker 2002: 213). However, other evidential values are also present: presumptive (particles, an auxiliary), reportative (completive constructions, an auxiliary, a conjunction + participle construction), quotative (a complementizer, a defective verb), visual (completive constructions), and participatory (a verbal construction) evidentiality. The oblique optative can also be regarded as an evidential strategy in that the information it expresses is attributed to another center of consciousness, different from the speaker at least in time and possibly also in person (= epiphenomenon of its deictic function). It is, however, difficult to label this interpretation with an existing evidential value. In order to formulate more generally valid statements about the linguistic encoding of evidentiality in Ancient Greek on the whole, it is necessary to investigate additional types of texts. Furthermore, it is essential to start from a well-defined methodology and to involve the diachronic aspect in the research, as this can aid in illuminating the origins of evidential morphemes/strategies and in mapping out their evolutions. In summary, I have attempted to sketch a linguistic-conceptual interpretative and integrative framework with which to approach certain Attic morphemes and constructions that are often intuitively understood by grammarians and linguists but only seldom connected with the conceptual frame of evidentiality.