I

Social dialect, which can be defined negatively as dialect associated with variables other than geographic region, was hardly recognized as a linguistic category until the twentieth century. Although it has been recognized since antiquity that groups at the bottom of the socio-economic ladder speak differently from the elite, non-elite idioms did not merit serious investigation since they were regarded merely as corrupt or decadent approximations to the prestige variety. There is evidence that the Greeks also recognized gender as a variable in linguistic production. Age occasionally figures in discourse about language, but the association is vaguer since it was tangled up with the idea that earlier generations spoke a better or more authentic form of Greek.

Writing a grammar of a vernacular (or stigmatized spoken variety) is inconceivable in a social context in which such varieties are thought to be essentially ungrammatical; and it may even be offensive in a context in which the standard symbolizes the purity, heritage or identity of the community, perhaps by association with a body of culturally important texts, even (or especially) if the variety in question is spoken by a large chunk of the population. The idea that social dialects are mutations or modifications of the standard language was common in the ancient world, and is still widespread. It implies that the mutation has a cause which can be identified, as opposed to the standard, which exists as the legitimate unmarked variety and calls for no particular explanation, being the natural expression of the healthy organism (the body politic or its truly free citizens).Footnote ² This was not, so far as we can tell, the way in which regional dialects were conceived in the Greek world, since there was no supra-regional standard until the Hellenistic period. The status of the classical dialects (or, at least, of their literary standards) was remembered in the post-classical period; and even after the establishment of the Attic-based koinē local communities in the old Greek world continued to write Greek with elements of regional dialect until well into the Roman period, while a handful of regional koinai (coloured with West Greek features) competed with the Attic-based koinē.Footnote ³ The overlap between regional and social dialect is complex, and the distinction between them is relatively recent. A social dialect can be connected with a region, or with a neighbourhood in a city (such as the Piraeus); in many cultures the prestige dialect is connected with a specific urban centre, and regional dialects may be stigmatized. Analysis of the social varieties of Greek is closely connected with a general view of the Greek language and the regional dialects. Central to the perception that varieties of language (regional or social) are variations or mutations of the standard are the following positions:

1. a) as a general principle language moves from unity to diversity; and

2. b) a language variety is essentially connected with a defined group of speakers (such as an ethnic group).

The first point includes the (useful heuristic) supposition that a parent language is a unity, a homogeneous entity from which the daughter languages derive by innovation and divergence (either a reconstructed proto-language, such as Indo-European, or an attested language such as Latin). It also covers the vaguer notion that linguistic diversity and messiness represent a decline from an earlier period of purity, which may have been as recent as a generation or two ago. The written standard is almost always the representative and symbol of that earlier uniform and ordered language.

A strong version of point (b) has been abandoned in most recent academic discourse; the early view articulated by Kretschmer, for example, which proposed that the ‘Greeks’ arrived in Greece in three waves (first the Ionians, then the Achaeans—a term which included Aeolians—and finally the Dorians after 1200 b.c.e.),Footnote ⁴ was generally rejected by the mid twentieth century. Arguments over the Dorian invasion and other prehistoric ‘dialectal migrations’ rumbled on (the movement of Ionic speakers and Aeolic speakers from mainland Greece to the eastern Aegean, for example).Footnote ⁵ It is true that an ethnic group will often use language as part of the definition of its ethnicity. What is now generally disputed is whether it is possible to retroject the group and its ethnic identity back to an earlier undocumented or prehistoric period and then trace its migration into the historical period. Less often disputed is the status or definition of the language in question (the reality underlying the dialect referred to as Ionic or Doric, etc.); in most cases this is a projection of the mythical ethnos.Footnote ⁶

A version of (b) which has not been entirely abandoned is the connection between a language variety and a social group like a socioeconomic or occupational category, or gender. On the one hand, a social group can generally be more clearly defined than an ethnic group: women, men, slaves can be pointed to, while ethnic group membership is more complex and usually self-defined, or at least a negotiation between self-definition and external ascription. But there is a sliding scale, on which ‘essential’ categories are likely to be unstable when pressed. As Bourdieu argued,

All linguistic communities are equally messy at all periods. By ‘messy’ I mean that they have numerous competing variants at all levels (phonological, morphological, lexical), and subsume numerous varieties made up of intersecting isoglosses. The notion that an earlier stage of a language was uniform is at least partly the result of the culture of a corpus language; if only written records remain, these are likely to witness the written standard. Additional ideological value derives from the frequent association of a written standard with a prestigious literary canon (since a standard most often has that status by virtue of this connection).

Even though modern linguistics maintains a distinction between regional dialect and social dialect, it is worth observing that the Greek regional dialects (like many popularly defined dialects) are cultural and historical constructs. The term Doric, for example, has little substantial linguistic content. It covers a vast and not clearly defined range of territory from the Corinthian gulf to Rhodes, with few common elements that we can be sure of apart from a handful of inherited features (only common innovations constitute evidence of a meaningful dialect grouping).Footnote ⁸ The Greek dialects are in effect organizational devices in modern scholarship. The ancient ethnic / political terms were revived and expanded in the context of nineteenth-century linguistic and biological science. We have no control over our data, which in a circular fashion reflect and confirm the local standards that were largely created or defined by writing. The data comprise a relatively small number of urban standards, and almost no inscriptions at all from large parts of the Greek world until the Hellenistic period (including Laconia, Messenia, Thessaly, Lesbos and northern Greece).Footnote ⁹

II

The connection of a language with a group of speakers who migrated at a poorly documented period to the historical homeland is relevant to social dialect to this extent: it bolsters the idea of an underlying metaphysical pattern whereby a language exists, then splits.Footnote ¹⁰ For Greek, the pattern would start with proto-Greek, spoken by newly arrived ‘proto-Greeks’ in the Balkan peninsula; this then split into the historical dialects (via their shadowy second-millennium forerunners) as the speakers headed in different directions across what became the historical Greek territories. In the cities or regions the local dialect further diversified into standard and variations (the standard gave rise to variations).

The classical tradition of large migrations,Footnote ¹¹ combined with European colonial expansion during the Renaissance,Footnote ¹² is a central ingredient in the dominant European model of language change until the mid twentieth century. There are different ways, however, to account for the historical data. Speakers of a group of Indo-European dialects which contributed the basic morphological and phonological shape of Greek were probably established in the Aegean area by the late third millennium. We cannot know the numbers in which they arrived, but comparative evidence suggests that they could have been relatively few, rather than constituting a large migration which populated the region; there were certainly people already living in Greece and the islands (some of them may have been speaking dialects belonging to a different branchFootnote ¹³ of Indo-European), and the relatively high proportion of non-Indo-European and non-Greek lexicon in Greek suggests a period of cohabitation and bilingualism rather than rapid shift.Footnote ¹⁴ If this model is at least partly correct, it also suggests that the indigenous culture held some prestige in the eyes (and ears) of the newcomers (whereas in contact linguistics a ‘substrate’ language is often associated with a culture relatively lacking in prestige in the contact situation).

The new arrivals contributed a distinct linguistic ingredient which penetrated more or less the entire region; the historical results of this co-production we know collectively as Greek. There are parallels later in European history, of course, for a change in language associated with a new population group. It is striking, however, that social and linguistic processes in cases well within the historical period—the emergence of Old English in the wake of Anglo-Saxon settlement in Britain following the Roman-British collapse of the mid fifth century c.e., or the development of Spanish in the Iberian Peninsula—have been equally disputed, which suggests that models may be partly to blame (rather than simply a lack of data, which of course is always a problem). In the case of both English and Spanish, the question ‘How did it arrive?’ is confusingly mixed up with the history of treating the terms English and Spanish as a given, rather than as a social and historical construct (terms which refer, of course, to the standard language in both cases). In the cases of both Old English and Iberian Latin, immigrants speaking closely related idioms arrived over an extended period, from different regions, and encountered a range of local languages where they settled. They clearly exercised political and military power, and had cultural weight, though they are likely to have been well outnumbered by the existing population. The linguistic standards which emerged were not the result of linear development from a proto-language spoken by the putative genetic ancestors and brought from overseas; each represents ‘a discursively constructed political artifact’Footnote ¹⁵ which prevailed after a chaotic history of political struggle and cultural influence.Footnote ¹⁶ They forced regional and social competitors into the status of dialects or patois.

The historical details vary, but it is characteristic for a standard language, an artificial compromise of a dialect of power with a written chancellery standard, and unnatural in the sense that at times it may have had few or no native speakers, to create a non-authentic standing for the non-written varieties. Modern standard German had no native speakers before the end of the eighteenth century but—in a result quite at odds with received ideas of the lifeless ‘artificial’ language—rose to a status where it dominated the German dialects, many of which disappeared before it. It also sucked in neighbouring related languages (that is, reassigning them as dialects of German rather than as independent languages).Footnote ¹⁷ The term Hochdeutsch, High German, transformed itself from a descriptive regional term (the uplands of southern Germany, as opposed to Niederdeutsch) into an evaluative metaphorical term.Footnote ¹⁸ Standard languages have been described as ‘pathological in their lack of diversity’;Footnote ¹⁹ the ancient Greek situation was a little more complex. Apart from the fact that Greek was a pluricentric language with independent regional standards, there emerged a number of supraregional linguistic codes which were associated with high literature. A polity such as classical Athens did, therefore, allow a number of linguistic standards. It accepted at least some Greek dialects as valid forms of Greek: Aristophanes brings characters on stage speaking dialect (the rendition of the dialect is not parodic, despite the comic context), and certain types of poetry were produced in a form of Doric or Ionic.

The development of writing in Greece led to the evolution of a number of epigraphic standards, some of which were underpinned by a literary (prose) standard. With regional standards, and often radically different social and political structures across the Greek-speaking world, non-standard varieties (where they existed) must have been regional too. Even though the categories concerned were either universal (sex) or near universal (servile status), gender roles clearly differed from region to region in Greece, and so did the range of social and political statuses (slaves, serfs, metics and citizens) in the context of different notions of citizenship.

III

Detailed analysis of social varieties of language is associated with the growth of sociolinguistics after the Second World War and with the study of language in its political and social context: W. Labov's The Social Stratification of English in New York City (Washington, 1966) is widely regarded as the pioneering work in the new field. New recording techniques allowed researchers to identify variables in the speech of individuals, and to quantify them accurately. For the first time it was understood that linguistic variation is not random but structured, and that language change in a community grows out of synchronic variation.

A language can be thought of as a number of overlapping varieties, some of which are tied to region, others to social group, gender and age, and almost all to a combination of these. Conversely, a speaker is not a member of a single social group or linguistic community: speakers belong to social networks, and research has indicated that examining the nature and extent of the networks that a speaker forms part of can provide better explanations of the social and linguistic data than simple variables such as gender or social class considered separately.Footnote ²⁰ In addition, most speakers command a range of linguistic registers tied to the nature and formality of the setting, the interlocutors and the content of the utterance. Such registers are likely to include elements from different social varieties. Clearly we do not have access to this range of material in investigating social dialects in the ancient world, and most of our data comes from Attica.Footnote ²¹

IV

The social stratification of Attica cannot easily be compared with that of the industrialized nation states, in which much modern sociolinguistic research has been carried out. We shall concentrate on the central period for which we have evidence of the epichoric dialects (the sixth to the fourth centuries b.c.e.); in the centuries before and after this period the social structure of the region must have been rather different. In the citizen body social stratification was no doubt less extreme than in many modern political communities: nevertheless, there were clearly marked differences in wealth, education and prestige. In classical literature there is certainly evidence for snobbishness within the citizen body based on family, education and wealth. Male citizens comprised a small percentage of the population of Attica, a figure which is notoriously difficult to estimate (and the population seems to have declined in the fourth century). Most estimatesFootnote ²² put the number of male citizens in the range of 35,000–45,000, and then quadruple that figure to take account of women and children. To this one has to add slaves (perhaps as many as 100,000) and metics (estimated at 25,000–30,000): this gives a total population approaching 300,000, of which the male citizens formed around 10% or slightly higher. Even this restricted group cannot have been linguistically homogeneous, since it will have included (amongst others) the following large groups: i) smallholders working their land in rural locations across Attica, ii) the wealthy urban elite and iii) the urban poor (the thete class, who served in the navy).Footnote ²³ In an urban context the social networks will have been considerably more complex than such broad groupings suggest.

V

Evidence for linguistic variation associated with different sections of the population is scarce, since our two main sources, literature and inscriptions on stone, are themselves associated with very specific linguistic registers. High literature (which includes tragedy and the historical prose of Thucydides) is written in a special literary dialect, influenced by ‘foreign’ genres such as Ionic prose, and epic and lyric poetry. The prose of fourth-century speech-writers, on the other hand, who wrote for delivery in the law-courts or in political arenas such as the popular assembly, is designed to be an unadorned approximation to the educated speech of citizen jurors or voters. It is not poetic, and generally avoids overt stylization (especially the law-court speeches), and to that extent can give us an idea of what was considered suitable in a formal register of the highest social group. Comedy is an important source to which we shall return: the language of comic dialogue, like forensic prose, is clearly designed to sound close to (a high variety of) spoken Attic, even though it is written in verse, which means that some dialect features (for example the ‘Ionic’ dative plurals in -αισι / -οισι) are used merely for metrical convenience. Unfortunately for us, although comic playwrights imitate and poke fun at a variety of linguistic phenomena, they avoid mocking social dialects of Attic which sections of their citizen audience are likely to have used.

Inscriptions are generally written in a formal variety of Attic which is not identical to literary prose, but is nevertheless rather uniform and clearly not a close reflection of the vernacular, since it seems deliberately conservative in some respects.Footnote ²⁴ This ‘chancery’ language is characteristic of public (official) inscriptions; most inscriptions of any length that were put up by private individuals but were intended for public consumption are written in verse (epitaphs, dedications), and the language is a mixture of Attic and epic / elegiac literary dialect. For evidence of the way in which non-elite social groups spoke we rely on ‘informal’ epigraphic texts such as graffiti (incised on ceramic material), dipinti (painted on ceramics), ostraka and curse tablets (scratched on lead). These sources may give us glimpses of vernacular Attic, but the very fact of being written makes them uncertain evidence, since a) the writer must have had access to a reasonable degree of education, and b) spelling conventions always disguise the writer's vernacular to some extent.

These informal texts are often carelessly spelled and written, and it is sometimes hard to be sure how much weight to attach to spellings which appear to anticipate later developments (since some of the spellings mistakes are clearly random). However, by around 400 b.c.e. at the latest there is sufficient evidence pointing to non-elite dialect in Attica, which in most (but not all) respects anticipated the later development of the elite standard; this variety (in reality a range of varieties) reflected some of the developments that were taking place in neighbouring Boeotia. It is worth noting that we are unlikely to be dealing with a simple High → Low continuum; in most speech communities the non-elite dialects are the most innovative varieties, and there are (therefore) a number of competing variants at this level (a model of social dialects will be pyramid-shaped). A characteristic of elite language is that it is conservative: innovative features are typically suppressed.

The spelling of informal texts suggests the following:Footnote ²⁵

VI

We have considered some of the evidence for non-elite or ‘substandard’ varieties of Attic; speakers belonging to the political and literary elite probably maintained a conservative pronunciation—at least in formal registers—until considerably later (see §VII.1 below). The difference between urban and rural environments must also have been an important factor, and will have intersected with other variables such as the degree of education; even after the end of formal education, there will have been ongoing differences in exposure to literary texts and performances, and to new philosophical and sophistic theories about language. A distinction which cuts across these categories is articulated in a fragment of Aristophanes (PCG 706), preserved in Sextus Empiricus:

For an idea of this ‘effeminate’ urban speech style, we can turn to Aristophanes’ Knights (1375–81), where τὰ μειράκια … τἀν τῶι μύρωι (‘the lads in the perfume shops’) are mocked for an affected Ionicizing style, in particular the use of adjectives in -ικός. The comic poets seem careful to avoid mocking ‘uneducated, rustic talk’; the nearest Aristophanes comes to this is in the linguistic characterization of some of his ‘anti-hero’ old men, whose speech may be marked by parataxis (Strepsiades, Philocleon, Euelpides). When the linguistic defects of politicians are attacked, these are characterized as ‘barbarian’.Footnote ³⁴ In a play called Cleophon, Plato Comicus portrayed the politician's mother speaking broken Greek to him (βαρβαρίζουσαν πρὸς αὐτόν) and referred to her as Thrassa ‘Thracian woman’ (61 PCG); the chorus at Aristophanes’ Frogs (679–84) makes the same connection between Cleophon's diction and alleged barbarian roots.Footnote ³⁵

A sour comment on Athenian speech-habits in the classical period survives in the Constitution of the Athenians of the ‘Old Oligarch’. The writer implies that Athenian naval supremacy has been responsible for a range of decadent behaviours, and in particular (Xen. [Ath. pol.] 2.7–8):

This comment points to a variety of Attic which we would want to posit on other grounds: an expanded variety, influenced by Ionic and associated particularly with mercantile activity and the navy. This ‘Piraeus Attic’ will have been a non-elite variety, perhaps marked by many of the phonological changes outlined above; it must also have been a precursor of the spoken koinē in the Hellenistic period. (The Greek of Xenophon, an upper-class Athenian who spent much of his life in a military context and outside of Attica, is marked by a number of lexical items that are alien to Attic.)

Metics and slaves will also have had an input into non-elite vernacular. Athens was, by ancient Greek standards, a relatively open city to foreign workers and residents, and slaves (many of whom will have originated from outside Attica) formed a large proportion of the population. The linguistic influence of slaves should not be underestimated, since they played an important role in the upbringing of the young citizen (nurses, paidagōgoi and others). In the last decades of the fifth century the rural population of Attica was often, owing to the Peloponnesian War, crowded inside the walls of Athens. This is likely to have played a role in the spread of non-conservative variants; and it explains why it would have been invidious for comic playwrights, in the context of domestic hardship and foreign enemies, to make non-elite or non-urban variants the subject of jokes.

VII

Gender is one of the central variables in most, perhaps all, linguistic communities.Footnote ³⁶ There are often marked differences in phonology, lexicon and idiom; differences in morphology and syntax have also been noted. It is very difficult to assess the effect of gender on language in ancient Greek, for the simple reason that we have almost no written sources by women, and very little reference to female speech in male writers.

Comedy, our principal source for the representation of everyday language, gives no suggestion of phonological differences between male and female characters. There is, however, a range of distinctive oaths, expletives and forms of address that are put into the mouths of women; certain lexical items also seem disproportionately frequent in female characters, and this includes a genitive of the personal pronoun which could be classified as an innovative morphological feature.Footnote ³⁷ Forms of address in male-female interactions, as opposed to same-sex interactions, tell us more about social relations between the sexes in Athens than about language stricto sensu, and the same is true of the observation that women avoid obscene language in the presence of men. However, all of this information contributes to a wider picture of the ethnography of language in a community. Analysis of oaths and expletives reveals some lexical differences between stage characters. A study by Sommerstein of Aristophanes and Menander shows that the adjectives τάλας, δύσμορος, γλυκύς, φίλος in vocatives are exclusively or principally used by women;Footnote ³⁸ these data are likely to be skewed by a male view that women's language was more emotional, or less serious, than men's language (in this connection Sommerstein points out that φίλος as an address is common between men in Plato's dialogues). As Euripides has Megara say in Euripides’ Herakles (536):

Many lament expletives in tragedy (οἲ ’γώ and others) are exclusively female; as McClure notes, ‘many of the expressions associated with women signify a state of heightened pathos’.Footnote ³⁹ Specialized lament vocabulary also recalls a specific social practice (perhaps the only one) in which women's voices were allowed, namely ritualized mourning for the dead. For the most part our sources are explicit that women were expected to remain silent.

Two further points in Megara's utterance are worth noting. First, she has prefaced the line with an apology: despite the fact that the entire family is about to be slaughtered, she apologizes to her father-in-law for speaking before him, which ‘is his right’. Second, the line contains the softening particle πως. Analysis of the particle usage of female characters provides some evidence of ‘hedging’, a feature that linguists have identified in women's speech in many modern communities in which women are disempowered. It is described by Lakoff as follows: ‘These sentence types provide a means whereby a speaker can avoid committing himself, and thereby avoid coming into conflict with the addressee … The more particles in a sentence that reinforce the notion that it is a request, rather than an order, the politer the result.’Footnote ⁴⁰ Although our data are meagre, Sommerstein has identified a tendency to use concessive particles (ἀλλ’ οὖν, καίτοι and clusters with γε), and Duhoux suggests a higher use of τοι (collaborative) and εἰ than for male characters.Footnote ⁴¹

In a much-discussed passage in Plato's Cratylus (418b–e), Socrates comments on changes in Greek phonology:

Some commentators have quoted alleged parallels from modern cultures, since in the early days of sociolinguistics some studies concluded that women tend to use a more conservative linguistic variety than men (closer to the prestige variety).Footnote ⁴² These studies tended to explain sex differentiation in language in terms of socio-economic class: women, feeling less secure than men for social and political reasons, associate themselves linguistically with a higher social group. Modern sociolinguistic work has questioned this implicit ranking of class over gender: in many contexts women are likely to create prestige forms rather than merely adopting them.Footnote ⁴³ Given the social networks in which women in Athens participated, there is no reason to predict that they would have favoured conservative speech styles, and some reason for thinking the opposite. They had minimal access to education, perhaps the most important agent in linguistic conservatism.Footnote ⁴⁴ Even if they were not legally barred from the theatre, it seems unlikely that large numbers of citizen women will have been in attendance; they were certainly excluded from the political and legal institutions where power derived from the manipulation of language. The day-to-day interactions of aristocratic women would have been largely with slaves, and the household may have been a locus of a sub-elite innovative variety (at a point in time which is hard to define, boys in particular would have been exposed to a weak diglossia—home language vs the public standard—which grew over time as the standard fossilized into the koinē). Non-elite women were less likely to have been confined indoors, since many needed to work; in Aristophanes working-class women take their place alongside men in the market and are not to be messed with.Footnote ⁴⁵ There is clearly no reason to think that they would have used anything other than an innovative variety similar to the one spoken by their male counterparts.

If in the Cratylus passage cited above the implied subject of the sentence starting ‘But now they change’ (νῦν δὲ … μεταστρέφουσιν) is ‘women’, the best explanation is that of Sommerstein, who argues that, if women were indeed saying [i:] for [ε:] and [d] for [zd], then they spoke an innovative rather than a conservative variety: these features were shared with Boeotian.Footnote ⁴⁶ Willi has also argued that innovative features can be detected in the language of female characters on the comic stage: for example, in the use of the genitives μου / σου to indicate possession, and in the use of lexical items such as καθάπερ ‘like’ and ὑπάρχω ‘be’.Footnote ⁴⁷ If, as seems more natural, the verb ‘they change’ refers generally to contemporary speakers, Plato is alleging an unexpected change of [i(:)] to [e(:)] and [ε:], exactly the reverse of what we reconstruct. I interpret this as a straightforward reference to hypercorrection, whereby a speaker is unsure of, or cannot hear, the distinction between a prestigious and a non-prestigious variant: the speaker therefore overcompensates by using the prestige form where it is historically incorrect (the joke in Catullus 84: Arius says hinsidias instead of insidias). Plato argues, or pretends to, that ἡμέρα [hε:mera:] ‘day’ is a modern hypercorrect substitution for the ancient ‘authentic’ ἱμέρα [hi:mera:] or εἱμέρα [he:mera:]; this could only happen in the case of a speaker for whom [ε:] and [e:] had merged with [i(:)]. This enables him to make an etymological play with the verb ἱμείρω ‘desire’. The argument is that women, being conservative, still say ἱμέρα [hi:mera:]; since this is, in fact, an innovative pronunciation, it would still be in line with Sommerstein's argument.

VIII

The obsession with mapping letters (eta, etc.) onto sounds presupposes, and flows from, two fifth-century spelling reforms: the use of the digraphs (‘spurious diphthongs’) ΕΙ, ΟΥ for [e:], [o:] and the adoption of the Ionic alphabet (official in 403 b.c.e.) by which Η was renamed eta and reassigned from [h] to inherited open [ε:]. Before the second half of the fifth century both long e vowels were written Ε, and it would have been difficult for Plato to frame his argument in the Cratylus. The script reforms in late fifth-century Athens, in the context of Sophistic interest in defining and classifying language (orthoepeia, associated particularly with Protagoras and Prodicus), fed a phonographic revolution in Greek thought about language. Words were reimagined as a sequential ordering of discrete minimal elements (stoicheia), each of which contributed its correct pronunciation. This opened the way for a reading pronunciation of words, which was more correct as it captured an ‘original’ state of the word and its relationship with reality. This is a powerful prod to a linguistic culture to start reimagining language variants as correct or incorrect (rather than as merely vulgar or noble) and is a necessary ingredient in diglossia: the phonographic articulation of the prestige variety contrasts with the ‘organic’ production of the vernacular. This is how a classicizing pronunciation of high-prestige or written Greek continued until, presumably, the time of Libanius (many of the grammatical categories of Attic would have been indistinguishable in the fifth century c.e. in the phonology of contemporary vernacular); it may even have been maintained as a primary speech style by some.

In any case, if hypercorrection is the right explanation of the Cratylus passage, it indicates that there was an awareness of sound change by the early / mid fourth century in Athens, and a desire to undo it.Footnote ⁴⁸ This points to a new self-consciousness in the linguistic culture of Athens: Attic now had value as a cultural and epistemic property, partly connected with its emergent status as the prevailing written form of Greek. If women or market-traders said [ni:ki:] for NIKH, this was to be explained by their lack of ownership in the shared cultural heritage of the Greek world, in which Athens was the natural leader. At around the same time Isocrates was making an explicit connection between paideia and Greek identity, and saw the Attic dialect as playing a central role in this.Footnote ⁴⁹

IX

Languages generally move in the direction of a ‘substandard’; this is merely because languages change continuously, and the standard is a conservative variety which suppresses innovation, and is generally based on the written language and (usually) a literary canon. One cannot predict which of the competing non-standard elements will prevail to become part of the new standard; some will disappear, and others will remain in the non-standard varieties. But at a given point in time, all are part of the grammar of the linguistic community.

At the start of the fifth century in Athens there was no clear concept of grammaticality, since writing in prose was in its infancy. Over the next two centuries literacy increased and the book trade grew; language became an object of philosophical enquiry and rhetorical training, and prose as a genre became culturally central. By the end of the fourth century an Attic-based Panhellenic standard had emerged, and in Athens at least it became possible to compare ways of speaking with this standard; in this context a polar opposition between correct and improper language established itself in the social consciousness of Greek speakers (and then Latin speakers). Linguistic heterogeneity is reduced to a simple dichotomy, and this is a familiar strand in European thought about language. Popular or colloquial language is given the same status as standard language or regional dialect; but they are abstract terms covering networks of registers and isoglosses.