2.1 Different explanations about -ι⟨o⟩ς, -ι⟨o⟩ν#/-ί⟨o⟩ν#: phonetic vs. non-phonetic origin

Although the question has been debated for over a century, complete agreement on the origin and characteristics of the phenomenon reflected in the spellings -ι⟨o⟩ς, -ι⟨o⟩ν /-ί⟨o⟩ν has not been reached. Different theories have been proposed, yet some authors limit themselves to only attesting to the appearance of these variants without explaining them (Dickey 2009: 157), or they vaguely refer to possible analogical influences while recognizing that “the explanation still remains uncertain” (Browning 1983: 38). Those who have addressed the analysis in greater detail have proposed either morphological or phonetic explanations, or a combination of both. Initially, using a morphological approach,¹⁰ different possible influences were proposed to explain the phenomenon’s origin: that of the abbreviated Latin terms of the type Clodis vs. Clodius, Cornelis vs. Cornelius (an old proposal, found in Ritschl 1861; Foy 1879: 127; Hatzidakis 1892: 112), or the influence of the tendency to create hypocorisms “by violently shortening proper names” (Jannaris 1897: 293). Both hypotheses have been rejected, although they are sometimes considered to be secondary influences for the spread of the variation (Mayser 1938: 16, fn. 1). On the other hand, Gignac (1981: 28) stated that “the nouns so affected are either diminutives or other nouns, including names, which lend themselves easily to shortened forms (…) This suggests that the forms in -ις, -ιν (later -ι) are shortened forms of -ιος, -ιον rather than strictly phonetic simplifications”.

Nevertheless, other authors have proposed a primarily phonetic explanation, although considering that at some point there could have been a morphological influence. Based on the parallel change -ιος → -ις in the genitive singular of nouns of the third declension, e.g. πόλι⟨o⟩ς (← πόλεως) or πράσι⟨o⟩ς, Brixhe (1994: 222) ruled out a morphological origin. Mayser & Schmoll (1970: 130–131) had pointed out the similarity with the genitives in -ιος of Egyptian names of the third declension, such as Ὀννώφρι⟨o⟩ς. Like Schwyzer (1939: 472), they hypothesized a contraction ι + ο > ι (rejected by Gignac 1981: 28, fn. 4). However, they also considered as alternatives that these Egyptian genitives might simply show a use of an indeclinable nominative form (so also Gignac 1981: 75) or a mere confusion of case endings.

Georgacas (1948: 257–259) put forward a different interpretation proposing a process [ion]> [iən]> [ii̯n]> [in], with a change from -o- to -ə- in an unstressed position before -n, with subsequent assimilation -ii̯- and contraction in -i-. He considered the possibility of the -ι⟨o⟩ς variant being analogical, a result of the influence of the accusative singular.

More recently, Horrocks (2010: 175) pointed out that “from as early as the 3rd century BC the o-vowel in certain categories of 2nd-declension noun, namely masculine personal names in -ιος [-ios] (the only such subclass) and neuters in -ιον [-ion]/ -ίον [-ˈion], started to be lost in popular speech, either through syncope of an unstressed vowel in a period before the synizesis of antevocalic [i] had begun (…), or through the regular change of prevocalic [i] to [j], followed by loss of the o-vowel and revocalization of [j] > [i] (through a process known as samprasarana)”. This second alternative agrees with Méndez Dosuna’s interpretation (2000: 283). Regardless of the details of the explanation, it should be noted that Brixhe (1994: 235), Méndez Dosuna and Horrocks deliberated on what could have been the chronological sequence between the loss of -ι- due to synizesis and the omission of -o-, which are usually analysed separately. Gignac (1981: 28–29) had already drawn attention to the appearance in the 2nd declension of the spellings -ι⟨o⟩ς, -ι⟨o⟩ν/-ί⟨o⟩ν together with the spellings that would reflect a synizesis in -⟨ί⟩ου, -⟨ί⟩ῳ, although both types of variants were interpreted as results of different processes, one morphological and the other phonetic.

2.2 Synizesis and omission of ⟨ι⟩

There seems to be general agreement in the analysis of spellings omitting ⟨ι⟩ in the genitive and the dative, e.g. ἀργυρ⟨ί⟩ου or ἐκφορ⟨ί⟩ωι. The majority view is that they reflect a synizesis of /i/, usually explained as a regular/popular allegro pronunciation [iV] > [jV] with loss of syllabicity and consonantization of the glide leading to its omission in writing (Mayser & Schmoll 1970: 126, Gignac 1976: 302, Teodorsson 1977: 237, Horrocks 2010: 169). This general explanation is however not without objections. Thus, Brixhe (1994: 235, fn. 42) drew attention to the lack of influence of [j] on the preceding consonant, especially in groups that have a clear evolution in ancient dialects, such as *[pj]> [pt]. Brixhe pointed out that this did not pose problems following alveolars, such as [r], [n], [l], but was less evident after [k], [m], [p]. He concluded that “cependant, en l’absence de toute perturbation graphique autre que l’absence du iota, ne faut-il pas accorder, provisoirement au moins, un préjugé favorable à cette solution?” Méndez Dosuna (2000: 281–282) also raised objections of this type. He recognized that the omission of prevocalic ⟨ι⟩, interpreted as a popular and/or allegro pronunciation [j], would be admissible after ρ in forms such as κυρά (for κυρία) “possibly [kyˈra] (<[kyˈrja])” or τριακόσα (for τριακóσια) “possibly [trjaˈkosa]”, in light of parallels in Modern Greek, such as νοικοκυρά or τρακόσα. However, in other contexts “we must be dealing with simple mistakes” and he specifically quotes the cases of omission of ⟨ι⟩ before a vowel and following a λ or a nasal, as κογχύλ⟨ι⟩ον.

2.3 Possible influence of the stress and the alteration of vowel length

It is generally accepted that, starting in Hellenistic times, important changes took place in the phonetic and phonological system, such as the disappearance of the differences in vowel length and the change of accentual type. Greek papyri document the evolution of pitch accent to stress accent (Gignac 1976: 325). This process continued in different places and at varying linguistic levels of the Koine during several centuries. Georgacas and Horrocks specifically mentioned the unstressed character of the omitted vowel. This fact, together with the associated loss of distinctiveness in vowel length, may be relevant to the disappearance of ⟨o⟩ and ⟨ι⟩ in popular/allegro pronunciation contexts.

On the other hand, a change from paroxytone to oxytone accentuation has been deduced when tonic -ί- disappeared before a back vowel. Together with [ˈi]> [j] there would have been a shift of stress to the final syllable [ˈion]> [jˈon] (Gignac 1976: 302, fn. 3; Teodorsson 1977: 237, fn. 270). It would be an accentual change equivalent to καρδιά, with [i] > [j], in Modern Greek. Although Gignac considered that the omission of ⟨o⟩ was not a phonetic process (whereas the omission of ⟨ι⟩ was), he acknowledged the influence of the stress: “Phonetic patterns, however, conditioned the change. The -o- is lost in the nom. and acc. when the accent is proparoxytone; the -i- is lost in the gen. and dat. when the accent is paroxytone, corresponding to a general shift of prevocalic stressed /i/ to /j/ with a concomitant shift of accent to the final vowel”. It is very interesting that in Greek loanwords in Coptic, the disappearance of ⟨o⟩ is also observed in both the -ιον and -ιος endings, which Girgis (1965–1966: 83) attributes to the weakened pronunciation of Greek unstressed -o-.

2.4 Morphological distribution

An interesting point is the restriction mentioned frequently of subclasses of words that have spellings -ι⟨o⟩ς, -ι⟨o⟩ν. Their presence is often considered to be limited to neuter nouns in -ιον/-ίον and to names in -ιος (Mayser & Schmoll 1970: 130, Browning 1983: 38, Horrocks 2010: 175). Gignac (1981: 25), however, noted its appearance in Roman and Byzantine times in names, forms of address, titles, occupational designations and diminutives. Sometimes its appearance in adjectives is mentioned (Mayser 1938: 15), or noted that “there is little sign of a similar change in common nouns and adjectives in -ιος” (Browning l.c.). But Georgacas (1948: 243, fn. 2) categorically refuted this, arguing that the supposed adjectives would have become nouns before being affected by the change. Gignac (1981: 115) and Brixhe (1994: 240) refer to the analogical influence of other adjectives in -ος, -α, -ον. Recently, Holton et al. (2019: 84) have stated that “it must be noted that the restriction of the phenomenon to nouns (though certainly not all nouns, e.g. ἥλιος) and the exclusion of the phonetically identical adjectives in -ιος, -ιον from this development (e.g. ἄξιος, τίμιος, πλούσιος) indicates that the causes of the phenomenon cannot have been purely phonetic and that morphological leveling of some sort must have played a part”.

Sometimes mention is made of a new type of late declension of these forms: nom. -ις (masc.), -ιν (neut.), acc. -ιν, gen. -ίου, dat. -ίῳ (Mayser 1938: 15; Gignac 1981: 25; Dickey 2009: 157); or, taking into account the forms with synizesis: -ις / -ιν, -ιν, -οῦ, -ῷ, type κῦρις, κῦριν, κυροῦ, κυρῷ (Gignac 1981: 28).

4.1 -Cι⟨o⟩ς

When reviewing the 863 regularizations of ⟨o⟩ in words ending in -ιoς, we find another noteworthy fact. Between III BC and II AD they do not appear mainly in the nominative singular of the thematic declension, as might be expected from previous analyses, but rather in the genitive of names of the third declension with nominative in -ις. They are adaptations of Egyptian names to said Greek inflection, type Πετoσῖρις, genitive Πετοσίρι⟨ο⟩ς.²⁸ There are 645 regularizations (74.74 % of the total of -ι⟨o⟩ς). The TmI lists only 3 regularizations (0.35 %) of -ιος in genitive singular of the third declension in common nouns (2 ἄμι⟨ο⟩ς, 1 κόνι⟨ο⟩ς). Omissions in the genitive of Egyptian names are documented from III BC (4 items in TmI) to VIII AD (1 item), although the largest number of cases (225), as well as the largest FI (70.74 in 318,080 words) occurs in I AD. In I BC, for example, we have 27.5 regularizations in 102,525 words (FI 26.82), and in II AD there are 204.5 regularizations in 971,843 words (FI 21.04).²⁹

If we now examine the regularizations proposed for the nominative of names of the second declension, type Ἀπολλώνι⟨ο⟩ς, we count 155 (17.96 %). Chronologically, the TmI lists omissions gathered from I AD to VII AD. However, as can be seen from the examples cited in previous works and in searches carried out in the DDbDP, there are clear examples from Ptolemaic times, such as Ἀπολλώνι⟨ο⟩ς (P.Ryl. 2.249.1, II BC; SB 1.4281.4, II BC) or Νουμήνι⟨o⟩ς (P.Tebt. 4.1108.84, II BC).³⁰ They are forms that were not regularized in the corresponding editions, so they do not appear in TmI. As pointed out in section 3, while the use of these lists is useful for an extensive examination of the phonetic context in which the change took place and for the general analysis, it is not a valid source for accurate dating. According to the TmI lists, the largest number of regularizations in modern editions appears on papyri from II AD (66.25 items, with a FI 6.82 in the aforementioned subcorpus of 971,843 words), although the largest FI (8.41) occurs in I AD (with 26.75 examples in a subcorpus of 566,834 words).

View Full Size

Figure 2

Nº of regularizations and FI of genitive -ι⟨o⟩ς in 3rd decl. names

Citation: Journal of Greek Linguistics 23, 1 (2023) ; 10.1163/15699846-02301003

• Download Figure
• Download figure as PowerPoint slide

Regarding the nominative singular of nouns and adjectives, type κύρι⟨o⟩ς, there are 60 regularizations (6.95 %). The number is lower than that of names, but not insignificant. There are 13.5 regularizations and a higher FI (4.24) in I AD, and 24.5 regularizations and a FI 2.52 in II AD.

Therefore, there is no total restriction, but there is a relevant disproportion between some morphological types and others. The greater number of items in the genitive of Egyptian names could be due to the greater presence of this morphological type³¹ in the documentation. To this we may add Mayser & Schmoll’s (1970: 130) hypothesis, followed by Gignac (1981: 75), that the spellings in -ις of this ending can be nominative forms in non-inflectional use. However, given their high number and the corroboration of the same phenomenon in Greek names, it is difficult not to recognize an equivalent phonetic omission in the graphic omission of ⟨o⟩ in the genitive of Egyptian names. One could ask whether a lower morphological pressure to maintain graphic regularity in the genitive of Egyptian names compared to the nominative of Greek names, could have caused the greater presence of the phenomenon in them. Moreover, Vierros (2012: 174–175) has pointed out that the confusion of the nominative ending -ις and genitive ending -ιος shows that the unstressed vowels were likely to be dropped (and be added hypercorrectly), but that sometimes, however, the names were intentionally in the nominative. She considers several possible conditioning factors (phrase-initial inflection, the dominance of the nominative case, context influence, etc.). Therefore, in the future, it will be necessary to carefully study the attestations to determine to what extent there could have been a confluence of different internal and external factors in the development of the phenomenon.

View Full Size

Figure 3

Nº of regularizations and FI of -ι⟨o⟩ς in 2nd decl. names

Citation: Journal of Greek Linguistics 23, 1 (2023) ; 10.1163/15699846-02301003

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Figure 4

Nº of regularizations and FI of -ι⟨o⟩ς in common nouns and adjectives

Citation: Journal of Greek Linguistics 23, 1 (2023) ; 10.1163/15699846-02301003

• Download Figure
• Download figure as PowerPoint slide

Regarding the phonetic context, accentually all the forms are proparoxytone. Furthermore, of the 645 regularized ⟨o⟩ omissions in the genitive of Egyptian names, only in 3 is the grapheme preceding the sequence -ι⟨o⟩ς a vowel. In the remaining 642 cases, we find 13 different consonantal graphemes, but over half of the items appear after three specific graphemes: σ (133), ρ (123) and ν (81).

Regarding the names of the thematic declension, of the total of 155 regularizations in the nominative, the preceding consonants are mostly ν (54), λ (48) and ρ (21). Omission occurs in 12 items after σ. Therefore, the omission of ⟨o⟩ is especially frequent after coronal-alveolar consonants.

In the case of nouns and adjectives ending in -ι⟨ο⟩ς, the omission of ⟨o⟩ is more frequent after ρ (35) and δ (20, all corresponding to the term γέρδι⟨o⟩ς). Again, the most frequent context is after a coronal-alveolar or dental consonant. In contrast, there are no items after λ and very few, comparatively, after ν (only 2), as can be seen in table 2.

4.2 -C⟨ι⟩oς

If we now go on to examine the 48 examples of the other graphic variant, -⟨ι⟩ος, with omission of ⟨ι⟩, we find 38 regularizations in the genitives of Egyptian names, type Nεχθμίν⟨ι⟩oς,³³ appearing more frequently after ν (15), ρ (6), λ (5) and φ (5, 4 of them being in the name Πετετρίφ⟨ι⟩ος). There are 10 regularizations of the nominative in names in -ιος of the second declension, type Δημήτρ⟨ι⟩ος, appearing more frequently after ρ (4), λ (2) and ν (2). That is, the consonantal contexts in which ⟨ι⟩ is omitted are mostly the same as the contexts in which ⟨o⟩ is omitted. In Egyptian names the preceding syllable has a tonic -ί- in 27 items, in a type -ίC⟨ι⟩ος, such as Πετοσίρ⟨ι⟩ος, or a tonic grapheme to note the pronunciation of [i] or a closed vowel: ύ (2), η (3), οι (1), ει (1), ου (3). Therefore, dissimilation is a possibility.

This conditioning factor is not so clearly appreciated in the few examples that exist in the names of the second declension with omission of ⟨ι⟩: Δημήτρ⟨ι⟩ος (2, III and II BC), Πομπών⟨ι⟩ος (I BC), Μάρκ⟨ι⟩ος (II AD), Πωσιδών⟨ι⟩ος (II/III AD), Αἴλ⟨ι⟩ος, [Οὐαλέ]ρ⟨ι⟩ος (III AD), Γερόντ⟨ι⟩ος (III/IV AD), Ἀνατόλ⟨ι⟩ος (VI AD), Μακάρ⟨ι⟩ος (VI/VII AD).

4.3 -Cι⟨o⟩ν# / -Cί⟨o⟩ν#

There are 342 regularizations collected in TmI regarding the omission of ⟨o⟩ in -Cι⟨o⟩ν.³⁴ Only 36 (10.53 %) correspond to names of the second declension, which begin to be documented in the TmI lists from I AD (13.5), the period when their highest FI occurs (4.24 in a subcorpus of 318,080 words). In the case of -ι⟨o⟩ν, the highest number of regularizations (306 = 89.47 %) corresponds to nouns and adjectives. Regarding chronology, TmI contains data from III BC (ἀργύρι⟨o⟩ν in O.Wilcken 2.329.5), which increase, together with the FI, until II AD (98.66 items, FI 10.15 in a subcorpus of 971,843), although the largest FI occurs in V AD (15.5 items and a FI of 12.44 in a subcorpus of 124,582 words) and VI AD (46 items and a FI of 13.58 in a subcorpus of 338,765 words) respectively.

Of the 36 regularizations in names, 9 are neuter female names (7 in -άρι⟨ο⟩ν) and 27 are accusative male names (including an adj. Ἀντιοχήσιον) in -ι⟨ο⟩ν. They appear mainly after coronal-alveolar ρ and ν.

View Full Size

Figure 5

Nº of regularizations and FI of -ι⟨o⟩ν in nouns and adjectives

Citation: Journal of Greek Linguistics 23, 1 (2023) ; 10.1163/15699846-02301003

• Download Figure
• Download figure as PowerPoint slide

In nouns and adjectives ending in -ι⟨o⟩ν, the omission occurs after a large number of consonants, but again a higher percentage is observed after coronal-alveolar or dental consonants. Especially after ρ, ν, δ, τ, λ and σ. There is only one different context in which its frequency stands out: after the bilabial μ.

It is noteworthy that, although the accentuation is mostly proparoxytone, there are also omissions of ⟨o⟩ in paroxytone words in -ίον> -ίν (17), and even in some oxytone word in -ιόν> -ίν (δεξι⟨ó⟩ν). However, attention should be drawn to the fluctuations in the accent notation of some of these forms in papyrus editions and lexicons, a question which, due to space limitations, I cannot address here.

4.4 -C⟨ι⟩oν# / -C⟨ί⟩oν#

We also find some omissions of ⟨ι⟩, albeit very few, in -ιον/-ίον endings. Only 4 regularizations are documented at the end of names (Δημήτρ⟨ι⟩ον, Οὐαλέρ⟨ι⟩ον, Πατερμούθ⟨ι⟩ον, Γεώργ⟨ι⟩ον) between I/II AD and VI AD, while in nouns and adjectives there are 12 regularizations from the Ptolemaic period (παιδάρ⟨ι⟩ον, ἱμάτ⟨ι⟩ον III BC, ἐργαστήρ⟨ι⟩oν II BC) up to VI AD. In only two forms is ⟨ί⟩ tonic (βιβλ⟨ί⟩oν II AD and γραμματ⟨εῖ⟩ον IV AD, at a time when the itacistic pronunciation of ει would already have consolidated). Like in the rest of the lists, it is possible to find more forms not regularized in their editions. For example, ἀρν⟨ί⟩ον already in III BC, in P.Gurob 22.3, 15, 34 (instead of the supposed irregular nominative ἄρνον [sic] instead of ἀρήν, as interpreted, for example in the Greek-English Lexikon of Liddell & Scott s.v.).³⁵ Obviously, these examples are few in number and can be interpreted as mere errors with no phonetic cause. However, we cannot rule out the alternative that they portray minority variants arising in the process of resolving the vowel hiatus. The highest number of omissions (5) appears after ρ: δηνάρ⟨ι⟩ον (2), ἑπτακελλάρ⟨ι⟩ον, ἐργαστήρ⟨ι⟩ον, παιδάρ⟨ι⟩ον.

4.5 -C⟨ί⟩ου#

In TmI there are 98 cases of graphic omission of -ι before -ου. Of these, 41 are in names, all before the ending of the genitive singular of the second declension except for 1 case in the beginning of the word (⟨Ἰ⟩ουστίνου in P.Oxy. 54.3774.9, IV AD). There are cases from III BC to VIII AD, and a few from the Ptolemaic period such as Ποσειδων⟨ί⟩ου (III BC), Δημητρ⟨ί⟩ου (II BC) or Ἀπολλων⟨ί⟩ου (II BC). Again, synizesis occurs mostly after coronal-alveolar consonants (like the omission of ⟨o⟩ in the endings -ι⟨o⟩ς, -ι⟨o⟩ν), especially after ν (13) and ρ (6).

There are 57 omissions in nouns and adjectives before -ου. Inside the word there are 2 items (ῥᾳδ⟨ι⟩ουργίας BGU 1.226.14, I AD, πρ⟨ι⟩ουάτου P.Flor.3.320.3, IV AD) and 1 of an accusative plural κυρ⟨ί⟩ους (P.Herm. 46.2, IV AD). The 54 corrections in genitives are dated from III BC to VIII AD. In Ptolemaic times we find ἀργυρ⟨ί⟩ου (III BC), ὑπο]βρυχ⟨ί⟩ου (II BC) and κυρ⟨ί⟩ου (ΙΙ BC). Again, the highest number of cases is after coronal-alveolar, especially ρ (46 items), with 21 quotes of the genitive κυρ⟨ί⟩ου.

4.6 -C⟨ί⟩ῳ# / -C⟨ί⟩ων#

In TmI there are 61 cases of omission of ι before ω.³⁶ Of these, 28 correspond to dative singular endings of the second declension: 7 in names and 21 in nouns and adjectives, especially κυρ⟨ί⟩ῳ or its equivalent spelling κυρ⟨ί⟩ο. 14 items are dated between VI AD and VIII AD. The oldest datives with ι omission are from II AD, e.g., Διδυμαρ⟨ί⟩ῳ or ὑπτ⟨ί⟩ῳ.

However, in other morphological forms it is possible to find omissions already in Ptolemaic times, specifically the nominative Σαραπ⟨ί⟩ων (UPZ 2.180a and b.3, II BC). In fact, the omissions of ⟨ι⟩ before -ω found in TmI include 33 items in contexts other than dative singular endings. Of these, 9 are omissions in genitives plural such as δεηνα[ρ]⟨ί⟩ων (P.Yadin 1.11.7, II AD) οr κλαν⟨ί⟩ων (P.Hamb. 1.10.46, II AD), and 10 in nominatives ending in -ί⟨ω⟩ν, while in the rest of the regularizations (14) the omission appears inside the word. Of all the 61 regularizations of -⟨ι⟩ω, 23 occur after ρ, 6 after π and 4 after λ.

4.7 Other cases of -ι⟨V⟩(C)#, -⟨ι⟩V(C)#

This review has been limited to the sequences -Cιος, -Cιον, -Cίον, -Cίου, -Cίῳ at the end of the word. Only occasionally have we referred to sequences like -Cίους οr -Cίων. However, one should remember that, although in smaller numbers, there also are omissions in sequences of -ιoC# preceded by a vowel (-Vιος, -Vιον),³⁷ such as Πτολεμαῖ⟨o⟩ς (CPR 28.1, III BC; BGU 16.2674.199, I BC), πλοῖ⟨o⟩ν (BGU 16.2661.11, I BC), etc., as well as in forms with a different V₂ vowel, such as σιτομετρί⟨α⟩ν (UPZ 2.158.5, col. A, b and c, III BC), τρισχιλί⟨α⟩ς (P.Lips. 1.7.19, II BC), κεράμι⟨α⟩ (P.Tebt. 1.122.2, I BC), or κυρ⟨ί⟩αν (P.Diosk. 18.48, II BC), νεκρ⟨ί⟩αν (UPZ 1.18.14, II BC), παιδάρ⟨ι⟩α (SB 14.11960.121, II AD), etc. Due to space limitations here, I plan to analyse them elsewhere.

5.1 Hiatus resolution strategies and their graphic expression

Examining the editorial regularizations mentioned previously, we are able to review the hypothesis mentioned in section 2. To this end, it is useful to also take into account data resulting from the typological comparison. As has been observed in recent decades, many languages, through different procedures, avoid heterosyllabic contiguous vowel sequences [V₁.V₂]. This can lead to a variability of results and said variability can be verified within the same language. Among the different possible solutions, we can cite (Casali 2011: 1434; Holton et al. 2019: 79): the elimination of V₁ or V₂; the loss of syllabicity and conversion into glide of one of the vowels ([V̯₁V₂]/[V₁V̯₂]) with the ensuing consonantization of the glide; the diphthongization of the sequence;³⁸ the coalescence or contraction of the two vowels in a third V₃; or the development of an epenthetic consonant between the vowels in hiatus [V₁.V₂]> [V₁^CV₂] / [V₁CV₂]. Often, the hiatus sequence [V₁.V₂] coexists with other phonetic variants that resolve it. These variants would depend on developments such as a faster pronunciation, more careless or more colloquial. Ancient Greek lacked phonological rising diphthongs of type /iV/. In Greek dialects of the first millennium, before the Koine, this sequence was usually heterosyllabic [i.V]. However, occasionally a consonantic element was placed between the vowels [i^jV], or a tautosyllabic pronunciation replaced the hiatus by a diphthong of increasing sonority [iV] / [i̯V] (Lejeune 1972: 163, 245). From what we have seen, the documentary papyri show that the tendency to avoid the vowel hiatus in [i.V(C)#] increased in the Koine.

It is also important to bear in mind that the same spelling could be used to represent different phonetic realizations. Thus, in pre-Koine Greek the same normative spelling -ιV- was used for different pronunciations [i.V], [i^jV], [i̯V]/[iV]. Therefore, we can infer that the same spelling in the documentary papyri could have been the result of more than one phonetic realization. This would depend on whether the orthographic norm predominated or phonetic perception imposed itself against the norm, as well as which were the features that were perceived the most. Faced with possible allophonic variants, the listener and/or the scribe would use one spelling or another.

5.2 Variability and constraining factors in the resolution processes

The variability of results may be conditioned by different causes and restrictions in each language (Casali 2011; Garrido 2013). One can observe internal factors in the phonetic, phonological and morphosyntactic systems as well as external factors. This article can only focus on factors of the first group, without going into possible external factors such as the frequency of use of the words affected or sociolinguistic influences.³⁹ For this reason, attention has been paid to accent distribution and the consonantal context as well as to the possible relationship with the morphological subtypes (names of Egyptian origin, names of Greek origin, nouns and adjectives), given the importance granted to these conditioning factors in previous works. In this regard, we can confirm a reduced presence of changes in adjectives (section 2.4). It is however not limited to nominalized adjectives as Georgacas argued. There are clear examples of adjectives used as adjectives: e.g. ἁλιεὺς ποτάμι⟨ο⟩ς P.Teb. 2.316.90 (I AD), γύης δημóσι⟨o⟩ς P.Mich. 5.272.4, 5 (I AD).

5.3 Phonetic origin and contexts of appearance

As we have seen, the omission of ⟨o⟩ appears both in the nominative and accusative singular of the second declension, as well as in the genitive singular of the third, confirming Brixhe’s opinion (section 2.1) about its phonetic origin. But there is an important remark to be made: the more relevant data is provided by the Egyptian names, not by the Greek nouns. The analysis of the editorial regularizations points to this noteworthy fact that has gone unnoticed: the number of quoted words and FI is much higher in the genitives in -ι⟨o⟩ς of the Egyptian names, e.g. Πετοσίρι⟨o⟩ς. In addition to the frequency of this subtype, this apparent non-inflectional use of the nominative (sections 2.1 and 4.1) could be caused by the fact that the elision of ⟨o⟩ was subjected to less morphological pressure in order to maintain the standard pronunciation than in other subtypes.⁴⁰ Therefore, there would have been less restriction for the resolution of this hiatus. In the case of -ι⟨o⟩ν, on the other hand, its greater presence in common nouns is confirmed, although its number and FI are lower than in the genitives in -ι⟨o⟩ς of Egyptian names.

The quantity and early date of the cases point to the same phonetic origin for -ι⟨o⟩ς and -ι⟨o⟩ν, as opposed to Georgacas’s interpretation (section 2.1). Georgacas advocates a phonetic origin for -ι⟨o⟩ν and an analogical extension for -ι⟨o⟩ς, a hypothesis that must, however, be rejected especially due to the important evidence of the genitives in -ι⟨o⟩ς.

There are also a smaller number of spelling variants omitting ⟨ι⟩ instead of ⟨o⟩ in -⟨ι⟩ος, -⟨ι⟩ον. Such omissions seem at first glance to be equivalent to those of -⟨ί⟩ου and -⟨ί⟩ῳ endings. These spellings are lower in number and FI than the omissions of ⟨o⟩ in -ι⟨o⟩ς, -ι⟨o⟩ν, although chronologically simultaneous.

Regarding the context, the consonants preceding the omitted ⟨ι⟩ and ⟨o⟩ are of different types, but there is a clear majority of cases after coronals, especially coronal-alveolar and coronal-dental, such as ρ, ν, λ and σ (mostly in Egyptian names), δ, τ. This can be seen in both the omission of ⟨o⟩ and of ⟨ι⟩. Therefore, vowel hiatus resolution occurs preferentially after coronal consonants, but this does not determine which, V₁ or V₂, is omitted.

The omission of ⟨o⟩ occurs mostly in proparoxytone words. All the forms regularized in -ι⟨o⟩ς are proparoxytone (except ὑι⟨ό⟩ς), while the forms in -ι⟨o⟩ν account for only 4.97 % (17) with oxyton ending -ί⟨o⟩ν. Therefore, the data seem to support mostly the link pointed out by Georgacas and Gignac (section 2.3) between stress accent and ⟨o⟩ omission contrasting with other opinions that reject the influence of accent (e.g. Brixhe 1994: 223). In analyzing other languages, the influence of the prosodic structure and the distance with respect to the stressed syllable has also been pointed out. A frequent solution is the diphthongization of the unstressed vowel strings of increasing sonority [i.V]> [iV].⁴¹ In the case of Ptolemaic Greek, Georgacas (section 2.1) also pointed to a change [io]> [iə] in an unstressed position. Again, Gignac (1976: 325) mentioned a change [o]> [ə] “in unaccented syll. esp. before ς”. We must remember too that “schwa is often the result of vowel reduction. Its short duration and its consequent tendency to co-articulate make schwa a likely candidate for the vocalism of stressless domains (…) schwa is especially susceptible to deletion, and thus may alternate with zero under varying conditions” (Silverman 2011: 641). Given the characteristics of the schwa, a diphthong [iə] in postonic position could have tended to a pronunciation [i(ː)]. In fact, in English we have a parallel: [iə] in words like year or here: “many Scottish and American speakers (…) will pronounce hear with a pure vowel followed by a consonantal r” (Clark et al. 2007: 35). A proposal has been made of a previous consonantization of the glide [i]> [j] in the resolution of the sequences [i.o], followed by samprasarana [j]> [i], as reported in Latin Florentiae > MedIt. Fiorenze> MnIt. Firenze (Méndez Dosuna 2000: 283), or in Osc. pakis <*-yo-s vs. Lat. Pacius, Umbr. safinim ⟨ *safnyom vs. Lat. Samnium (Tikkanen 2011: 17). However, it is also possible that the diphthongization did not imply the consonantization of the first vowel. It has been found that in some languages a string of two vowels in hiatus has a longer duration than that same sequence pronounced tautosyllabically in a diphthong,⁴² and that the two elements of the diphthong tend to modify and have different constriction, duration and prominence, amplitude or sonority. So that it would be possible for these parameters to be different and change during the hiatus resolution process, depending on the circumstances. We could find the process [i.o]⟩ [io]> [iə]> [i^ə]> [i(ː)] in unstressed endings.

In the case of -ίoυ, -ίῳ endings, except for a few examples⁴³ of V₂ loss, ⟨ί⟩ is usually omitted. Obviously, maintaining the morphological distinctiveness of the declensions must have played an important role. Furthermore, simultaneous to the omission of ⟨ί⟩, the displacement of the stress to the final syllable has been inferred. This shift, noted by Gignac and others, points to a trend found in Modern Greek and other languages. Heterosyllabic sequences [ˈi.V] acquire a tautosyllabic pronunciation through a shift in the stress [ˈi.V] → [i̯ˈV].⁴⁴ This displacement implies increased sonority, duration and intensity of V₂ compared to V₁ which turns into a glide.⁴⁵ Therefore, the process in stressed endings evolved in different circumstances, a development such as [ˈiV]> [iˈV]> [i̯ˈV]> [^i̯ˈV]> [ˈV(ː)] being possible. The reduction of these characteristics in V₁ would imply its disappearance -ίoυ → -οῦ, -ίῳ → -ῷ, favored at a time when, together with the change to a stress accent, the differences in vowel length disappeared. In the case of -ίον endings, the analogical influence of the unstressed ending -ιον could favor a result -ίον → -ί⟨o⟩ν instead of -⟨ι⟩όν, the tonic -ί- remaining. There could perhaps be a few exceptions where the analogical pressure had no influence: ἀρνόν = ἀρν⟨ί⟩oν, βιβλόν = βιβλ⟨ί⟩ον, etc.

5.4 Gradience of the process

However, although few in number, there are also a number of spellings with omission of ⟨ι⟩ in unstressed endings -⟨ι⟩oς, -⟨ι⟩oν. How can they be interpreted? Can we look for alternatives to just seeing graphical errors without a phonetic trigger? In the case of Egyptian names like Πετοσίρ⟨ι⟩oς we have observed a probable dissimilatory influence of the vowel in the preceding syllable, mostly -ι-. When the hiatus was in the process of being resolved, this influence would have conditioned the omission of ⟨ι⟩ instead of ⟨o⟩. But the nominative and accusative singular forms of the second declension such as Πομπών⟨ι⟩ος, Ἀπολλόν⟨ι⟩oν, παιδάρ⟨ι⟩ον οr ἱμάτ⟨ι⟩ον do not correspond to that context. They are either mere spelling errors or they are part of variants with a small number of cases. If the second alternative is followed, it could perhaps be explained by considering the hiatus resolution process as a gradient non-categorical process. Casali (2011: 1455) mentions that “hiatus resolution patterns (glide formation and/or vowel elision) that had previously been described as categorical in two languages, Modern Greek and Igbo, respectively, actually involve gradient and highly variable timing adjustments”. Casali bases himself on earlier works by Baltazani (2006) and Zsiga (1993; 1997), dealing with the external or postlexical hiatus, to which other later works can be added: e.g. Pelekanou & Arvaniti (2002), Kainada (2012), and others. The idea of gradience in these processes—and it is tempting to apply it to other languages and to the internal hiatus—could suggest the existence of other phonetic variants. Although graphically expressed by keeping -o- and omitting ⟨ι⟩, they could actually represent allophones other than the cardinal or extreme vowels usually represented by said spellings.⁴⁶ The pronunciation of the sequence [V₁V₂] would be perceived precisely as a monophthong with distinctive features closer to [o] than to [i], which then influenced how it was written.⁴⁷

The gradient nature of the process also raises a doubt regarding the categorical change [i.V]> [jV] in Koine Greek, with a consonant glide equivalent to that which has been identified for earlier and later stages in the evolution of Greek. We must remember the objections raised by Brixhe and Méndez Dosuna (section 2.2) regarding the apparent lack of effect in the consonant preceding [j]. The large number of cases casts a doubt on considering them mere graphic errors without any phonetic grounds. There are authors who have differentiated between different types of glides (Padgett 2008; Levi 2011). Padgett differentiated between vowels, semi-vocalic glides and consonantal glides [i, i̯, j] depending on dynamics and stricture. So that “within a language a glide might be realized as either semivocalic or consonantal”. He considered that “it is very plausible, even likely, that this variability between semivocalic and consonantal realizations is gradient”. Therefore, if, taking stricture into consideration, a gradient [iV] → [i̯V] → [jV] → [^jV] can be established, it is possible that in the language of documentary papyri only one stage was reached [iV] → [i̯V] → [^i̯V], with the disappearance of the semi-vocalic glide and without any palatalizing effects on the preceding consonant, which a consonant glide [jV] → [^jV] would have had.