The Formation of Greek Place-nouns in-εών /-ών

This paper sets forth a new theory for the origin of the ancient Greek class of placenoun derivatives characterized by a stem formant that appears as either -ώνor -εώνdepending on the dialect. In the classical period and afterward, the stem formant acts as a simple productive suffix that derives place-nouns from noun bases. I propose that these place-nouns were originally formed as further derivatives of derived adjectival bases. Later, but still at a relatively early stage of Greek, the combination of the genitival suffix -ε(ι)οand the substantivizing ‘Strabon suffix’ -ōn̆was reanalyzed as monomorphemic and propagated as a productive unitary formant.


Forms of -(ε)ών Derivatives
Descriptively, there is a significant number of words in ancient Greek that end in nominative -ων, genitive -ωνος.But nouns with this descriptive ending are not all created equal; there are several distinct word-groups with the same surface stem final, but with different properties, meanings, and origins.One particular class of oxytone -ών-stem nouns, a fairly substantial group of what may generally be called 'derived place-nouns,' has eluded previous at-barth Indo-European Linguistics 4 (2016) 1-14 tempts at reconstruction.The present goal is to retrace this particular group to a Proto-Greek origin.
Before its phonological form can be reconstructed by comparison, first it may be helpful to differentiate several distinct groups of nouns in -ων, -ωνος.First, primary nouns and internal derivatives in -ών, -ῶνος and -μών, -μῶνος need no further comment other than to say that they share the accentual properties of derived place-nouns, but are of no further interest to the present discussion.Derivatives in the so-called 'Hoffmann suffix' (Hoffmann 1955) bear some semantic similarities to place-noun derivatives, but they often have different accentual behavior and show different dialectal forms.
Primary nouns/internal derivatives in -ών, -ῶνος and -μών, -μῶνος (1) ἀγών 'gathering, contest' Hom.+ (2) χειμών 'winter' Hom.+ (3) πρών 'headland, promontory' Hom.+ 'Hoffmann suffix' derivatives (noun + *-h x on-) (4) κέντρον 'goad' → κέντρων 'slave, i.e. one who bears marks of the goad' Soph.+ (5) ζυγόν 'yoke' → ζούγωνερ (pl.) 'working oxen, i.e. those who bear a yoke' Hesych.(Laconian) A third group of derivatives in -ων are formed with the substantivizing suffix -on commonly known as the 'Strabon suffix.'Accent is often retracted in these derivatives, though the precise rules are unclear (see footnote 4).Members of this final group can be distinguished from the others on phonological grounds.In this group alone, the shape of the suffix differs consistently between different dialects in the following way.In the classical period, such words end in long -ών, -ῶνος in Attic, but in Ionic the same words end in -εών, -εῶνος.For example, Herodotus uses the Ionic form of the word ἀνδρεών 'men's quarters' , but in the works of Aeschylus the same word is attested in the Attic form ἀνδρών.When such a cross-dialectal correspondence is encountered, the default assumption is that Attic long -ω-is the result of the characteristic Attic vowel contraction of -εω-to -ω-,1 while Ionic -εω-simply preserves the original uncontracted vowels in hiatus.
We have further evidence of synchronic variation between these forms from Phrynichus, the second century Attic prescriptivist grammarian who condemns the use of uncontracted forms ἱστεών meaning 'weaving shed,' καλαμεών 'reed bed' , and ἱππεών 'stable' in addition to ἀνδρεών itself.Of course, not every place-noun in Attic -ών or Ionic -εών is attested in both dialects, which complicates the question of their identification-if we find a solely Attic place-noun in -ών, other morphological sources cannot be ruled out definitively.On the other hand, Attic forms in -εών and Ionic forms in -ών can be discounted, as they do not show the correct forms, though with the caveat that by the postclassical period the original dialectally determined distribution had been obscured by borrowing between dialects in both directions, and the suffix variants became more or less interchangeable in practice.

Semantic Categories of -(ε)ών Derivatives
To sum up section 1.1, we can recognize nouns that exhibit predictable dialectal variation between -ών and -εών as comprising a formally homogenous group of derivatives.To this point, these derivatives have simply been referred to as 'place-nouns,' which is how they are traditionally described.But a closer examination of their semantic features is needed to hypothesize what underlying components comprise the complex stem formant of this derivative class.As the above examples demonstrate, most -(ε)ών derivatives may be subcategorized as either types of land or types of buildings.Most words for types of land name either groves of trees derived from tree names or cultivated plots of land derived from crop names, though other types of terrain derived from a variety of words are possible.The second group includes words for living quarters, other household rooms and areas, workrooms, shops, animal pens, and other miscellaneous building structures.It is perhaps worth noting that these place-nouns include concrete buildings as well as general areas within them.

1.3
The Base: Derivative Relationship The relationship between base and derivative for each of these place-nouns can be unified as semantically 'genitival,' meaning the kind of relationship that exists between a given x and its derivative, meaning 'belonging to x' or 'of x.' x → 'place of/for x's' (23) αἴξ 'goat' → αἰγών 'place for goats, i.e. goat stable' -ών Arist.+(24) κοίτη 'bed' → κοιτών 'place of beds, i.e. sleeping room' -ών Ar.+Despite the different shades of meaning that can be distinguished among these groups, the formation of all of these derivatives conforms to the following descriptive pathway: added to a noun x, the formant -(ε)ών makes a derivative, also a noun, with a degree of specialized function as a nomen loci characterized by the presence of one or more x(es).2In some cases, that pathway of formulation is straightforward: the base αἴξ 'goat' gives αἰγών meaning 'place for goats, that is, a goat stable.'In other cases, the derivative really refers to the immediate space comprised of the base noun: from μύλη 'mill' is made the derivative μυλών 'mill area, i.e. millhouse.'The base and derivative may even be more or less synonymous: ἐσχάρᾱ 'hearth' and its derivative ἐσχαρ(ε)ών 'hearth area' both mean virtually the same thing.
Having described the semantics of -(ε)ών derivatives, the next section addresses the phonological facts of the case.

2
Reconstructing the Proto-Form

Previous (Unsatisfying) Accounts of -(ε)ὠν Place-Nouns
Several previous attempts have been made to explain the history of this noun class, none of which is fully satisfying.Ehrlich (1905, p. 69 f.) traces -(ε)ών to the pie possessive *-u̯ en-(cf.Sanksrit -van-), proposing an original form *-ē-u̯ ōn.But there is no trace of -u̯ -in any derivatives of this type, even in dialects that preserve intervocalic digamma in writing.Even more importantly, deletion of barth Indo-European Linguistics 4 (2016) 1-14 -u̯ -is never followed by contraction of the resulting hiatus in Attic, as already mentioned in § 1.2.Wackernagel (1909, p. 334 f.) provides a solid descriptive account of the place-noun suffix with numerous examples.He suggests that variant suffix forms were in competition (-ών, -ειών, -εϝών), with different dialectal reflexes complicating their outcomes.Though his discussion provides valuable groundwork, it is not his immediate concern to determine which constituent morphemes might have originally combined to form the complex -ε(ι)ών formant.As such, he reaches no conclusions regarding the ultimate origin of the suffix.Petersen (1937) suggests that hypostasis of genitive plurals was ultimately responsible for the promulgation of place-nouns in -(ε)ών.For example, in such phrases as ὁ ἀνδρῶν (οἶκος), 'men's (household),' the substantive could be elided, leaving just the article and genitive plural modifier, reinterpreted as a nominative singular noun.The biggest problem for Petersen's theory is his explanation of the shift from a circumflex accent to an acute on -ών.He explains the change as the result of a uniquely close phrasal relationship between the genitive and the elided substantive, but this conjecture has no independent support.

2.2
Phonological Reconstruction The regular correspondence of Attic long -ō-and Ionic -eō-tells us to reconstruct an original sequence of a short or long e vowel + consonant + long ō, where the consonant is either *-s-or *-i̯ -.Hiatus like that of the Attic ending -εών arises in situations where, historically, an intervocalic consonant has been lost.The only consonants lost in this environment are *-s-(following lenition to -h-) and the semivowels -i̯ -and -u̯ -.While the Ionic form preserves the resulting hiatus, contraction of the adjacent vowels is a regular process in Attic.
The semivowel -u̯ -is another candidate for loss in the intervocalic position, but it can be eliminated straightaway, since even in Attic hiatus resulting from the loss of -u̯ -never results in the contraction of ĕō > ō.For example, in the genuinely Attic word κενεών meaning 'flank, i.e. hollow area between ribs and hip' (from the adjective base κενεός 'empty'), the vowels in hiatus do not contract after the relatively late loss of -u̯ -in the nominal base *keneu̯ ο-.It is probably best to analyze κενεών as a Strabon-type derivative.Since the stem had a -u̯ -to block contraction, even Attic authors (Xen.+)use κενεών rather than (an unattested) †κενών, so the -ε-of κενεών must be part of the base rather than part of a complex derivational -εών suffix.Indeed, any genuinely Attic words ending in uncontracted -εών must ultimately come from a source other than the placenoun formant.
Looking only at Attic/Ionic evidence, the short ĕ vowel in -εών is ambiguous, traceable either to an actual Proto-Greek short *ĕ or instead to Attic/Ionic ǣ (itself ultimately from Proto-Greek long *ā, whose predictable Attic/Ionic outcome is short ĕ before an o vowel).At this point, for purely phonological reasons we can hypothesize that the original place-noun formant of this type was, in its Proto-Greek form, either *-āsōn, *-esōn (with short or long e), *-āi̯ ōn, or *-eī̯ ōn (also with short or long e).
Evidence from other dialects now becomes crucial.The Homeric place-noun derivative χαλκεών 'forge or smithy' has a short -ε-in the ending.While a sole Homeric example may not be enough to establish a pattern,3 it does begin to suggest that long *ē, long *ā and long *ǣ are less likely reconstructions than simple *ĕ for the first vowel of the place-noun formant.
The -ε-in Elean ξενεονα '(acc.)guest chambers' supports the reconstruction of a genuine short *ĕ.And the -ιών-formant in the Laconian (gen.)place-name Κνακιῶνος is consistent with this, and also confirms once and for all that the lost consonant is not *u̯ , since an *-eu̯ o-sequence loses -u̯ -too late to undergo the normal Laconian change from εο > ιο.
(32) Chantraine (1963, pp. 15-17) o-di-do-si du-ru-to-mo a-mo-te-jo-na-de e-pi-pu-ta 50 a-ko-so-ne-qe 50 Pyl.Vn 10,2 'What the woodcutters give to the chariot-shop: 50 saplings and 50 axles.' As a final note on phonological reconstruction, since Indo-European and Proto-Greek stem formants are never -cvcv-in shape, it is clear from its shape alone that we are dealing with a complex stem formant, i.e. not a single morpheme, but a suffix historically composed of at least two distinct components.

Old Materials, New Combinations
At this point we are left with the question of what constituent morphemes could have produced a morpheme complex of the shape *-ei̯ ōn with the semantic properties discussed in section §1.An explanation falls into place when we consider the independently verifiable presence of two morphemes in Greek: a. genitival adjectives in -ε(ι)ο-, inherited from Proto-Greek b. the Strabon suffix -on-, which substantivizes thematic adjectives (see §1.1.3) The combination of these two morphemes to produce a complex ending -e(i̯ )oon, compositionally meaning 'that which is of x,' provides a fitting analysis, both phonologically and semantically, of the original formation of place-noun derivatives in -(ε)ών.The rest of this section will be dedicated to explaining the particulars of this proposed process.

3.2
Derived Adjectives in -ε(ι)ο-Taking point (a) first, Greek has a derivational formant -εο-(also appearing as -ειο-with metrical lengthening) that derives thematic adjectives from thematic nouns and consonant-stem nouns.In a few cases we also encounter derivatives from feminine ā-stems, likely by analogy to the more productive base types.These derived adjectives in -εο-, inherited from Proto-Greek and earlier *-ei̯ o-(to which may be compared Latin -eus and Vedic -áya-), have traditionally been perceived and classified as so-called 'material' adjectives, but a broader understanding of the -ε(ι)ο-adjective formant opens up new combinatory and explanatory possibilities.It is true that many adjectives formed with this suffix do indicate the material of which a thing is made, and in post-Homeric Greek they do so exclusively.Alongside (33) χάλκεος and (34) λίθεος in the table above, there are in particular a large number of adjectives derived from the names of metals.However, Nussbaum (2009) shows that the -ε(ι)ο-formant originally had a broader range of genitival functions in Homeric Greek, and that its later restriction to material adjectives was only a specialization of one possible use of genitivals.Adjectives derived from animal names tend to specialize in reference to the skin of that animal, and as such might be regarded as source adjectives rather than strictly material adjectives.The adjectives (36) κρόκεος 'saffroncolored' and (41) κύνειος 'dog-like' exemplify a different common semantic extension of genitivals, namely 'like x.'But the Homeric examples below, presented by Nussbaum (2009), show even clearer cases of -ε(ι)ο-adjectives with non-material (though still genitival) semantics.base, specialized as a nomen loci, with the same accentual properties observed in -(ε)ών place-nouns.The Strabon suffix alone does not regularly produce derivatives with the characteristic oxytone accentual pattern and dialectally conditioned short -ĕof place-nouns.But together with -ei̯ o-, a productive source of adjectival bases with the appropriate genitival semantics, we can identify a likely process for the creation of place-noun derivatives.This would make the homophonous ending of κενεών with place-noun derivatives only partially accidental: it shares the latter component, though not the former, of the originally bimorphemic -(ε)ών suffix.

3.4
-e(i̯ )o-Derivatives Substantivized with -on-Putting the previous two sections together, we can now analyze place-noun derivatives as formed through a two-step process.First, -ε(ι)ο-adjectives with the genitival semantics 'of x' are derived from noun bases of various types.Then, these adjectives formed the bases of Strabon-type substantives in -onthat were specialized in meaning and function as nomina loci.We then have a complex morpheme -e(i̯ )o-on-, compositionally meaning 'that which is of x.' At an early point, the combination of the two morphemes was reinterpreted as a monomorphemic stem formant and used to produce a new class of placenouns.
The extension and specialization from 'that which is of x' to 'place of x' is fairly straightforward.We have already seen this specialized extension in κενεών, an independent example of -on-substantivizing an adjective with 'place-noun' semantics and the oxytone accent that is characteristic of this noun class.4The combination has not only appropriate semantics at every step, but also the right phonological shape to produce the outcomes observed in 4 It should be noted that an explanation for the oxytone accent itself remains elusive.One possibility might be that the Strabon suffix is underlyingly accented, which has simply been obscured in many familiar cases by a vocative retraction of accent (cf. the recessive accent of the vocative of personal names in -ων, as Ἀγαμέμνων, voc.Ἀγάμεμνον (Smyth § 261, p. 63)).
In addition to proper names, Strabon derivatives are often epithets such as γλύκων 'sweet one' < γλυκύς 'sweet' which could share the same explanation.Some common nouns in the Strabon suffix do have oxytone accent, e.g.βαιών 'small fish' < βαιός 'small,' which could either be inherited from the adjective or due to an accented suffix.But in reality neither semantic category nor the accent of the adjective basis is a fully reliable predictor of accent in Strabon derivatives, somewhat undercutting this hypothesis.
Complicating the issue is a tendency for words in -ων from all sources to shift their accent rightward.Chantraine observes that ἀπατεών 'cheater' < ἀπάτη and λυμεών 'deceiver' < λύμη morphologically resemble place-nouns, which is certainly true in that they too have shifted These examples demonstrate derivatives of genitival adjectives being used with collective meaning, and naming a location characterized by a multitude of the base noun.Outside of Greek, we have an even closer parallel in Latin vīnea (Plaut.+)'vineyard,' a substantivization of vīneus 'made of/belonging to wine,' which is itself a so-called 'material' adjective derived from vīnum (Plaut., Cato+) 'wine-grapes.'

Conclusion
In conclusion, the origin of -(ε)ών place-noun derivatives has remained an unresolved issue because one crucial piece of the puzzle has always been missing, namely, an account of the -ε-in -εών.The long-standing erroneous doctrine that -ε(ι)ο-adjectives are and always were exclusively material adjectives rather than more generally genitival adjectives has prevented the formulation of this new theory, which holds that their earlier less restricted use was 'fossilized' in this specific compositional use with substantivizing -on-.This pathway of derivation for place-noun derivatives is a particularly appealing analysis because it does not require that we propose any new material or processes to account for it-neither a unique process of accent shift, nor any new morphemes with specific properties.Previous attempts to explain this derivative group have relied on proposing new, homophonous morphemes or rarely (if ever) occurring processes to make the reconstructions fit the forms we observe.But this paper provides an explanation that uses known derivational suffixes and processes that operate independently in Greek.In the end, a simple combination of morphemes is all that is needed to explain the heretofore elusive origin and properties of place-noun derivatives in -(ε)ών.