The Italic branch of Indo-European falls into two subgroups, Latin-Faliscan and Sabellic.Footnote ¹ Faliscan is so close to Latin that we could classify it as a Latin dialect, but the Sabellic group differs from Latin considerably. It consists of Oscan, Umbrian, South Picene and some poorly attested local dialects.

Within Sabellic, Oscan, spoken in the south of Italy, is understood best. It was written in four scripts: the Etruscan and Latin alphabets, the Oscan ʻnational alphabetʼ and the Greek alphabet.Footnote ² The Etruscan alphabet was derived from a West Greek source. Since Etruscan did not have voiced stops or a phoneme /o/, the letters <B, D, O> were not used outside alphabet tablets, just as Italian has a letter <k>, but does not use it in native words, where /k/ is written <c> before back vowels and <ch> before front vowels.

The Latin alphabet, though derived from the Etruscan, is a better fit for Oscan phonology, partly because Latin revived the graphemes <B, D, O> under Greek influence and created a letter <G> by adding a diacritic to <C>. The Etruscans had derived <C> from Greek <Γ>, but in Etruscan it represented /k/, not /g/. Later, Latin added <Y> and <Z> (= [y] and [z]) to its alphabet; their use was restricted to loans from Greek and only became necessary when the Romans began to appreciate Greek culture and to attempt to pronounce Greek words in Greek fashion. Latin itself did not have [z], as intervocalic /s/, originally pronounced [z], had changed to /r/ in the mid fourth century b.c., when Papisius Crassus renamed himself Papirius. Oscan did not undergo rhotacism; when Oscans used the Latin alphabet, <Z> represented the voiced allophone of /s/.

The ‘national alphabet’ was independently derived from the Etruscan alphabet. Like the Romans, the Oscans needed a set of new letters; the result was an alphabet and a number of spelling conventions that match the phonology of the language closely. The reformed alphabet (after 300 b.c.) had six letters matching the six vowel qualities: <i, í, e, a, ú, u> standing for /i, e, ε, a, o, u/. Oscan preserved the word-initial accent reconstructed for Proto-Italic. In initial syllables vowel length remained distinctive, while elsewhere quantitative distinctions were mostly lost; long vowels could optionally be written with a double letter.Footnote ³ There were fifteen consonantal phonemes: /p, t, k, b, d, g, f, s, h, m, n, r, l, w, j/. Each had a separate grapheme, except for /j/, written <i>. Geminate consonants were often written twice. <h> had consonantal value word-initially, but between vowels it merely showed that they belong to separate syllables instead of forming a diphthong. The phonemes /u, f, s, n/ had [y, v, z, ŋ] as allophones, but there were no special letters to represent these. [y], found after dental / alveolar and before /m/, could be rendered <iu>.Footnote ⁴ We only know about [v] from transliterations into other languages. Unlike in the Latin alphabet, [z] could not be written <z>; while there was such a letter, it instead stood for /ts/. We shall come back to [ŋ]. Perhaps /l/ also had an allophone [ɫ], attested for Latin by grammarians; in South Picene, the velar variant had a reflex <v>.

Z.’s excellent book examines how Oscan sounds were rendered in the Greek alphabet, and what such spellings tell us about Oscan phonology. As a side issue, he re-examines the intricate problem of dating inscriptions based on orthography, concluding that orthography is by and large unreliable as a dating criterion.

Vowel spellings are inconsistent. For /e/ we find <ι, ε, ει>, the last one mostly from the third century onwards. For /o/, <ο> and <ω> were used, but the latter not before 300. For /u/ we find <υ, ο, ου>, but υ is restricted to [y], and if the vowel was long, only <ου> could be used. For /εi/ there was <ει> as well as <ηι>. Tautosyllabic /ou/ was written <ου> or <οϝ>, and from 300 onwards also <ωυ> or <ωϝ>. For heterosyllabic /ou/ we find <οϝ, ουϝ, ωϝ>, the last one again not before 300.

This variability can be explained if we assume that Oscans were first and foremost literate in Greek and that they were simultaneously aware of two divergent phonological systems. Greek had five short vowels, /i, y, e, a, o/, written <ι, υ, ε, α, ο>, and seven long vowels, /ī, ȳ, ē, ǣ, ā, ō, ū/, written <ι, υ, ει, η, α, ω, ου>. The preference for rendering /εi/ as <ηι> rather than <ει> in endings can be explained from Greek spelling, where <ηι> is a common ending. But writing [y] as <ου> rather than <υ> makes sense only in Oscan, where the sound was an allophone of /u/.

Vowel spellings have repercussions for our understanding of Oscan morphology. The dative singular of o-stems goes back to *-ōi; the spelling is <-οι> or <-ωι>, and since <ω> is restricted to /o/, we know that the diphthong was shortened to -oi before *-ō- could be raised to *-ū-, as in other positions. The accusative singular of o-stems is spelled <-ομ> or <-ωμ>. In parts of the Oscan territory the pronunciation was /-um/, but <ω> indicates that the inherited /-om/ was preserved in the south. The genitive plural of o-stems goes back to *-ōm; it is sometimes assumed that this was shortened to *-ŏm in Proto-Italic. However, the spelling is <-ομ> or <-ουμ>, pointing to /-um/, which shows that there was no shortening in Proto-Italic, as *-ŏ- would not be raised, while *-ō- would become *-ū- and then -ŭ-.

The consonants are more straightforward. The shape of <h> is a good criterion for dating; <H> stopped being used around 200, while what looks like the left half of <H> is found at all periods. For /f/, we find <F>, <Θ> (fourth and third centuries), and <B>. The last two symbols stood for [t^h] and [b] in contemporary Greek, not yet for [θ] and [β] or [v]. We should not draw any conclusions about pronunciation from the Greek sound values, as <Θ> and <B> may just be distortions of <8>, the Etruscan symbol for /f/. In early Latin, the dative of Jupiter's name is DIOVEI, while southern Oscan has ζωϝηι, indicating that /dj/ had become [dz] and eventually [z]. The Oscan and Latin alphabets show that /j/ caused palatalisation and gemination of preceding /k, t, l, n/. The genitive of a personal name in the Greek script, ωυδδιηις, shows that the sound change also affected /d/.

Much space is dedicated to inscriptions from Messina and Lucania, where the Greek script was used, but where spelling conventions used in the ‘national alphabet’ can occasionally be seen. Thus, the genitive νιυμσδιηις exhibits a spelling of the name where [y] is rendered by a digraph, corresponding to <iu>, and where epenthetic [p] is not written between /m/ and /s/, while outside Lucania we find the nominative spelled νομψις.Footnote ⁵ Similarly, in personal names we find the digraphs <κσ> and <πσ>, corresponding to <ks> and <ps>, instead of the equivalent <ξ> and <ψ> normally used. Only the Greek-alphabet inscriptions from this area show <h> as a marker of hiatus, as in the ʻnational alphabetʼ, compare μαραhις ʻof Marasʼ. The occasional use of double vowels or double consonants to mark length could also show influence from the ʻnational alphabetʼ. By contrast, Z. connects the spelling <νγ> instead of the usual Greek-style <γγ> for [ŋg] with substandard Greek usages rather than what we find in the ʻnational alphabetʼ.

Z.’s book is methodologically sound and clearly structured, with a plethora of tables and useful indexes. The bibliography is up to date, and there is also a welcome lexical appendix discussing the etymology and phonology of individual words. I can recommend this book unreservedly.