Bryn Mawr Classical Review 2007.05.10
Philomen Probert, Ancient Greek Accentuation. Synchronic Patterns, Frequency Effects, and Prehistory. Oxford: Oxford University Press, 2006. Pp. xxvi, 444. ISBN 978-0-19-927960-9. $150.00.
Reviewed by James Clackson, Jesus College, Cambridge (email@example.com)
Word count: 2822 words
Table of Contents
In his famously lugubrious preface to the second edition of his handbook on Greek accentuation,1 Chandler does not disguise his weariness with an 'insipid' subject in which he 'never took more than a languid interest.' However, Chandler's work remains one of the few Victorian compendia still used in modern Classical scholarship. The work under review is not a handbook (Probert has already published a short guide to Greek accentuation2), but it does much to counter Chandler's view of the tedium and inutility of the subject. Probert looks at how Greek word-accents actually work, how and why accent placements change through the history and prehistory of Greek, and she sets out to find answers to questions such as why the word for 'tree trunk', κορμός is accented on the final syllable, but ὅρμος, 'cord, chain' is recessive. P. combines extensive learning over a wide range of disciplines, with an engaging, often witty, style, and the book will repay the attention of textual critics and historical linguists as well as the usual suspects, comparative philologists and scholars of the Greek grammarians.
The work is divided into two parts. The first can be viewed as a base-camp, giving the reader the necessary knowledge of the terrain, the routes attempted by previous scholars, and instruction in using some of the relevant apparatus of modern linguistic theory, while in the second Probert takes up the challenge of explaining the differing accents of adjectives and nouns formed with five different suffixes: -ρο-, -το-, -νο-, -λο- and -μο-, and outlines how the work could be extended to other Greek suffixes. While Part II moves into completely new territory for ancient accentological research, Part I also contains much that is new. The opening chapter, which assesses the available evidence for the Greek accent through statements of ancient authors, grammarians, papyri, musical fragments, the manuscript tradition and medieval Greek, presents an overwhelming case for believing that the word-accents as reproduced in a good modern edition of Sophocles or Plato correspond to what the authors themselves would have said. Probert also discusses at some length the rather less conclusive evidence that the accentuation of the text of Homer was preserved through an oral tradition, and cautiously accepts this view. This chapter contains extensive evidence from grammarians and scholia (also well represented in later chapters), and brings out the significance of many of their observations that have often escaped earlier scholars.
In chapters 2 and 3, Probert moves on to the evidence for accentual variation in Greek across different dialects and within the history of the language and discusses the origin and workings of various restrictions in the location of the Greek accent, as, for example, the rule than an acute accent can only appear on the antepenultimate syllable if the final syllable does not contain a long vowel or end in a consonant cluster. Chapter 4 reviews the history of scholarship on the Greek accent, moving with equal surefootedness between Erasmus and generative phonology. In the course of these three chapters, Probert deftly includes comparative material from Sanskrit relevant to the placing of the Greek accents and discussion of the alternative ways of explaining the Greek law of limitation (of which the rule of acute accents on antepenultimate syllables given above is a part). Impressive also is her avoidance of some of the crevasses of the subject. Thus the date of the switch to a stress accent of the Modern Greek type is given as 'the early centuries AD' (p. 50) without further comment, and there is no discussion of the puzzling fact that the restriction of the circumflex accent on the penultimate syllable in some ways mirrors that of the acute on the antepenultimate, in that the final syllable cannot contain a long vowel, but in other ways it does not, since the final syllable can be closed by a consonant cluster, as in the Greek words κλῖμαξ and κῆρυξ. The traditional accentuation of the latter word is especially troubling, since in other cases, such as the genitive κήρυκος, the vowel υ in this word is long.
In Part II, Probert attempts to find some sort of reason behind the irregularity of Greek accents on adjectives and nouns. Her working hypothesis, arrived at from a consideration of various pieces of evidence, such as the accentuation of loanwords into Greek and the comparison of accented words with exact cognates in Sanskrit and Germanic, is that there is an inherent tension in Greek between a default lexical accent which is recessive, i.e. in which the accent is placed as near to the beginning of the word as the law of limitation allows, and various morphological suffixes which inherently determine non-recessive accentuation. Probert considers various explanations for why recessive or another accentuation wins out in individual cases. It could be, for example, that words formed earlier in the history of Greek favour one accent or another, or that Greek has two or more suffixes of identical shape, which differ only with regard to whether they are accented or not, or that various different semantic classes of words share particular accentuation patterns. Closer scrutiny, however, reveals that these explanations, if they have any validity at all, only work for a portion of all the words in the set. Thus, among the nouns ending in -νο-, it is possible to identify a set of neuters denoting tools and instruments ending in -ανον or -ανα, and these are all recessive, but this does not help to explain why στέφανος 'crown' is accented on the first syllable but οὐρανός 'heaven', and δελκανός kind of fish on the last.
In the absence of any explanation which will give a neat account of why some words with the suffixes -ρο-, -νο- etc. are recessive and others not, Probert investigates how the accentuation of words with these suffixes is distributed against their frequency of use. Although there is nothing in Ancient Greek comparable to the corpora of modern spoken languages now available, it is at least possible, owing to the Perseus Digital Library, prepared by Gregory Crane and a team from Tufts University (Perseus Project), to find word frequencies in a corpus of Greek texts of several million words. Probert interprets the results of her frequency analysis as follows. The suffixes -ρο-, -το-, -νο- and -λο- were all originally inherently accented, and this accentuation is generally retained in adjectives formed with these suffixes, since in these cases the meanings of the suffixes are largely predictable and most remain productive within the history of Greek (in the case of -λο- and -μο-, where the simple suffixes are not productive as adjectives, there is a corresponding decline in the regularity of final accentuation). However, nouns formed with these suffixes, many of which originate in substantivised use of early adjectives, in general tend towards recessive accentuation, since, in most cases, the predictability of the meaning of the noun is weaker; hence the synchronic identity of the suffix is lost, and the accent falls in with the default. As a corollary to this, she adduces the case of the suffix -μο-, which also seems to have been inherently accented. This suffix is found on very few inherited adjectives in Greek, but it is used with a high degree of semantic predictability to form abstract nouns (usually translatable by English gerunds ending in -ing). These nouns retain the final accentuation, except in cases where the meaning has greatly altered or the derivation has become opaque such that the identity of the suffix is lost. Hence, ὅρμος, 'cord, chain', is no longer recognised as connected with εἴρω 'tie, join' and so has the default recessive accent, but κορμός 'tree trunk' could still be seen as a step away from κείρω 'cut' (and was so derived in ancient etymologies).
Exceptions to this general pattern are found for very high frequency or very low frequency nouns, in the former case since the words are so common that speakers retain the original accentuation, in the latter, because words are so uncommon that they are not stored in the individual mental lexicons of speakers, but are recreated every time they are used. Thus we have some means of explaining the divergent accents of στέφανος, οὐρανός, and δελκανός given above. The word for heaven is attested 488 times in Probert's count of the Perseus Digital Library, and the fish name not at all. These two words are at either end of the frequency scale, and so retain their original accentuation, but στέφανος, with 361 hits, comes in between. As an analogous process Probert compares the accentuation of English verbs ending in the suffix -ate, which originally were stressed on the penultimate syllable, but for which the stress has changed in line with the default accentuation of English verbs, to final stress for disyllables, antepenultimate stress for trisyllabic and longer words. This change is observed in more frequently used words, such as dictáte and cóncentrate, but not for rarer vocabulary such as erúctate and fórmate, where the suffix -ate is still analysed as present by speakers.
Probert's argumentation throughout these chapters is always lucid and rigorous, and her theory of a 'leftward' drift in the accentuation of Greek nouns has much to recommend it (although I am not sure I would want to argue for a default recessive accentuation on Greek adjectives). Probert is always careful in her use of data and statistics, but I wondered whether her figures for frequency counts were really such a strong support for the overall argument as it may appear. Our corpus of surviving Greek texts is in no way comparable to the corpora of contemporary spoken (or spoken and written) English such as the American National Corpus or the British National Corpus. We have no record of spoken Greek, and the size of our ancient Greek corpus is small compared to the modern corpora (the American National Corpus is projected to include 100 million words, comparable in size to the British National Corpus, thirty times the size of the Perseus text-base at the date when Probert worked on it). This in itself would probably not be a problem, but unlike the modern corpora, the Greek corpus consists of literary texts, which span around a thousand years, from Homer to the second century AD. One wonders whether work based on this corpus can really be called synchronic in any meaningful sense.
Furthermore, in the Greek corpus it is possible that a single author's vocabulary preference may skew the results. Probert notes, for example, on page 170, that the word ἕταρος, which is the eighth most frequent noun with a suffix -ρο- in her sample, occurs almost exclusively in Homer, but otherwise she only gives a shorthand indication of the occurrence of specific items in the appendix, and no word is excluded from the count on grounds of its distribution. Thus, to take just a single example of many, one of the three most frequent nouns ending in -το- is βροτός 'mortal man (/woman)', a word which occurs principally in Homer and later poetry, and is avoided in prose. The effect of single author preferences is most pronounced in the case of words of low attestation. As an example, consider the words κύσσαρος 'anus' and κύτταρος 'hollow, cavity', which at first sight appear to be simple dialectal variants of one another, although the etymological dictionaries (followed by Probert) opt for their existence as separate words. Of these two words, κύσσαρος, which appears in Hippocrates and Galen, is allocated a frequency count of zero in Probert's table on p. 170, since Hippocrates and Galen are not in the Perseus digital library, but κύτταρος has a count of 3 since this word turns up three times in the text of Aristophanes (as well as in Aristotle and Theophrastus, who are not included in the Perseus texts). This might not seem to matter much in the larger scheme of things, but κύτταρος ends up included in the words of medium frequency, between 1 and 100 occurrences in the corpus, alongside words such as γαμβρός, attested 62 times in the corpus, while κύσσαρος is considered a word of very low frequency. With cases such as this in mind, I wondered what would happen if one were to alter the boundaries for how words with low attestation were counted. If we were to count attestations of five and less together with the words of zero attestation, what happens to Probert's figures? For nouns with the suffixes -ρο-, -το- and -νο-, the effect of this change is to level out the differences which Probert finds between nouns of very low frequency and those of neither low nor high frequency. The middle group no longer stands out with a predilection for a shift to recessive accentuation, there is rather a tendency, if anything, for decrease in final accentuation as nouns become less frequent: thus for nouns with a suffix -νο-, 60% of those with a frequency of over a 100 are finally accented, 46% of those with a frequency of between 6-100, and 32% of those with a frequency of 5 and lower; the figures for nouns with a suffix -ρο- are respectively 45%, 31% and 27%, for nouns with a suffix -το- (excluding certain types of formations following Probert) respectively 86%, 76% and 79%.
A second concern I had for the frequency counts that Probert gives is the selection of the material. In keeping with her aim to find out about the diachronic development of accentuation from Proto-Indo-European to Greek, Probert selects only material where she can be confident that the lexical item was at one stage in the history or prehistory of Greek formed with one of the suffixes she is interested in. Thus many common Greek words which do not have reasonably secure etymologies, or which were originally loanwords, or which can be segmented differently, are left out of the count. And conversely, words that we now know from comparative evidence to have one of these suffixes may not have been so analysed by the speakers. Following this reasoning, the following common words (amongst others) are excluded from the analysis of nouns with a -ρο- suffix: βλέφαρον 'eye-lid', μέγαρον 'large hall' and σωρός 'heap'. However, for each of these words there are associated forms within Greek which are at least as semantically close as many of the 'genuine' suffixed forms included in Probert's list (for example, αγρός 'field' and άγω 'lead, carry' (p. 334)): βλέπω 'look', μέγας 'big' and σῶμα 'heap'.3 Can we be sure that the Greeks themselves, who of course did not have access to their linguistic prehistory, did not analyse these words as containing a suffix? One way to check would be to look at ancient etymological works, as Probert herself does in the discussion of nouns in -μο- (pp. 253f.). Of course the defence for the exclusion of the etymologically uncertain material is that we have no way of knowing whether the location of the accent in, say σωρός, is due to the original inherent accentuation of the -ρο- suffix or for some other reason. But if the Greeks themselves could analyse the word as containing a suffix, does it actually matter whether the word can be identified as originally containing a suffix or not? In any case the comparative evidence for the location of the accent on a words such as αγρός (Sanskrit ájra-) is not exactly straightforward. This line of criticism may seem pedantic, especially since Probert herself stresses that the focus of the book is not etymology (p. 8), but it is relevant, since some of the conclusions made on the base of frequency counts depend on a very small number of lexical items. Again, for the sake of argument, consider what happens if we include the three nouns βλέφαρον, μέγαρον and σωρός in the frequency counts of nouns in -ρο- (remembering that it is impossible to repeat the same search that Probert made on Perseus in 1999). The recessive nouns βλέφαρον and μέγαρον both appear in the very frequent category, each with over a hundred attestations, but the finally accented σωρός, with 23 hits, enters into the middle category. With these nouns included, the percentage of finally accented nouns of very frequent occurrence drops from 45% to 38%, while the percentage with between 6-100 occurrences climbs to 36%. The results of the frequency analysis consequently seem to my mind inconclusive.
Despite my reservations about the frequency counts, there is no doubt that this is a book of the highest scholarly standards (there are remarkably few misprints), which opens up new avenues for research and which will set the field for all future discussion of the Greek accent. I only hope that, when she completed this detailed and painstaking research, Probert had the presence of mind not to follow Chandler's example, and that she refrained from consigning all her notes and collections of material relevant to Greek accentuation to the flames.
1. H.W. Chandler (1881), A Practical Introduction to Greek Accentuation, 2nd edition. Oxford: Clarendon Press.
2. P. Probert (2003), A New Short Guide to the Accentuation of Ancient Greek. London: Bristol Classical Press.
3. See the entries for βλέφαρον, μέγαρον and σωρός in P. Chantraine (1968-80), Dictionnaire étymologique de la langue grecque: Histoire des mots. Paris: Klincksieck.