In Vowel length from Latin to Romance, Michele Loporcaro takes the reader on a fascinating linguistic journey through time and space, focusing particularly on Northern Italy. The main objective of this well-documented journey, which combines detailed historical evidence and extensive experimental data from contemporary Romance varieties, is to unravel the history of vowel length. In Classical Latin, long and short vowels occurred in free distribution and were contrastive, as in the minimal pairs in (1).

1. (1)

   

Long and short vowels occurred in stressed and unstressed, open and closed and final and non-final syllables. According to Loporcaro, this contrastive vowel length was lost in the transition from Latin to the modern Romance languages, as the result of the lengthening of stressed vowels in open syllables (Open Syllable Lengthening; OSL) and the shortening of unstressed vowels. Its effects are still observable in Standard Italian, in Tuscan dialects and in the majority of central-southern Italo-Romance dialects. In these varieties, stressed vowels in open non-final syllables are lengthened, and occur in complementary distribution with short stressed vowels before geminates, as illustrated in (2) (p. 62).

1. (2)

   

In Northern Italo-Romance varieties such as Milanese, Cremonese and Friulian, a different situation obtains. In Milanese and Friulian, contrastive vowel length occurs in oxytones, e.g. Friulian [ˈliːs] ‘worn out’ vs. [ˈlis] ‘smooth’, [ˈdiː] ‘to say’ vs. [ˈdi] ‘day’ (p. 98). In Cremonese, contrastive vowel length is not limited to oxytones, but occurs also in paroxytones, e.g. [ˈroːda] ‘wheel’ vs. [ˈrota] ‘broken (fem.sg)’, [ˈpeːza] ‘scales’ vs. [ˈpesa] ‘rag, cloth’ (p. 85). The central questions that Loporcaro addresses are where vowel length in these varieties came from, and why it was later lost again in some of them. In my opinion, this book more than succeeds in its aim: presenting a reasonably detailed and reliable inventory of the facts required to provide a sensible diachronic account of the changes which affected vowel length in the development from Latin to the Romance languages.

Chapters 1 and 2 provide the background for the further discussion of vowel length in Romance. On the basis of a detailed examination of available evidence, Loporcaro convincingly argues that contrastive vowel length is a feature of spoken Classical Latin, contrary to claims of its early loss in the 3rd century BC (Pulgram Reference Pulgram1975). A quotation from Cicero about audiences protesting when a single quantity mistake is made (p. 8) clearly shows that contrastive vowel length (CVL) was a feature of spoken Latin during the Republic/Early Empire. Iasucthan's hexameters (pp. 46–51) show OSL in early 3rd century AD in Africa, as demonstrated by the fact that feet like mīlītēs and c≥pīt≥ are well-formed hexameters, realised as mīlītĕs and cāpīt≥ as a result of the lengthening of short stressed vowels and the shortening of long unstressed vowels. Distinct regional varieties of spoken Latin differed in displaying either CVL, a conservative feature, or OSL, which Loporcaro considers to be its diachronic successor.

An additional argument in support of late CVL that Loporcaro could have used is the principled absence of words like c≥pīt≥ in classical poetry. For example, antepenultimate stressed words cŭpīdītas ‘desire’ and m≥rītīmus ‘at sea’ are found neither in Ovid's Ars amatoria or Vergil's Aeneid. Three light syllables in a row simply don't fit in a hexameter with six feet (either (HLL) or (HH)). Similarly, m≥nīpŭlum ‘fistful, battalion’ and pĕrīcŭlum ‘danger’, which both have short antepenultimate vowels in Modern Italian manipolo and pericolo, differed crucially in Latin. Given a short stressed i in manipulum but a long one in periculum, the former can only occur in syncopated form in poetry, and the latter both as periculum and periclum.

The first two chapters succeed in establishing a solid base for the assessment of competing views on the further development of vowel length in the following chapters. One piece of evidence that Loporcaro might have considered to further pinpoint the co-occurrence of the merger of CVL and the rise of OSL are disyllabic words with hiatus, e.g. piam ‘pious’, viam ‘road’, meam ‘mine’ and deum ‘god’. All of these words had a short initial vowel in Classical Latin, but for the evolution of Latin to French we have to assume a long i for piam (the vowel survived as [i]), but a short i for viam (the vowel became [e], and after diphthongisation ended up as [wa] in Modern French voie). Italian di, Portuguese dia and Spanish dia also indicate a long i in words like diem ‘day’. Similarly, a long e has to be assumed for meam (the vowel became [e] and, just as in viam, [wa] in Old French meie, moie), but a short e for deum (the vowel became [ε] and, after diphthongisation, [jε] (dieu)). Lengthening of these stressed vowels in open syllables must clearly have occurred before the CVL merger.

Chapters 3 and 4 deal in detail with the development of vowel length in Romance and its analysis in Northern Romance. Chapter 3 presents a detailed overview of the evolution of the various Romance languages, and shows that, alongside many varieties that lack contrastive vowel length (Standard Italian, Tuscan dialects, central-southern Italo-Romance, Ibero-Romance, Daco-Romance), several other varieties (Northern Italo-Romance, Gallo-Romance) do have vowel-length contrasts which did not originate from a preservation of the Classical Latin contrasts, but arose later. In some of these varieties, this secondary CVL has been lost, but its historical traces can still be found today. An important aspect to mention in the evolution of the discussed varieties is that not only has vowel length arisen anew, but also, after degemination, consonantal length, as illustrated for Hauteville Franco-Provencal in (3) (p. 184).

1. (3)

   

Long stressed vowels (3a) are followed by a single consonant, while short stressed vowels from original closed syllable geminates either preserve the geminate or create a new one (3b).

The same allophonic distribution of consonant length occurs in the dialect of Semigallia (p. 185), as well as in Southern Emilian Lizzano, as illustrated in (4) for proparoxytones (p. 189).

1. (4)

   

The voiced geminates in (4a) show that gemination must have followed lenition, while the forms in (4b) show that it took place before syncope. Loporcaro (p. 185) observes: ‘the same situation – with systematic degemination only before stress and at least traces of gemination in proparoxytones – is documented for varieties like 12th–13th-century Romagnolo … or 13th–14th-century Veronese’, and mentions that in Carrara gemination affects both proparoxytones and paroxytones, as in [sal’vatːiko] ‘wild’ and [pu’tanːa] ‘whore’.

Chapters 4 and 5 offer a detailed analysis of vowel length in Northern Romance. After wandering up and down the slopes of the Apennines, tracking through the Alps and looking into France and Switzerland, this fascinating linguistic sightseeing tour ends with the panoramic and synoptic view in (5) (p. 207).

1. (5)

   

The Romance varieties discussed in the preceding chapters offer detailed evidence for the typology in (5). Dialects differ in four ways. Those in column (iv), such as modern Standard French, most dialects of Piedmont and eastern Lombardy, are varieties that no longer display CVL, but had it in earlier stages. The dialects in (iii), such as Milanese, now show only CVL in the final syllable. In (ii), e.g. Cremonese, Emilian and Ligurian, CVL is limited to the final and penultimate syllables, whereas the dialects in (i), such as the Rhaeto-Romance dialect of the Tavetsch valley, show the phonologised outcomes of CVL in any of the last three syllables, as illustrated in (6) (p. 205).

1. (6)

   

Prior to the loss of final vowels other than [a], stressed [e] in open syllables lengthened by OSL in both paroxytones and proparoxytones (a) before diphthongising to [ei̯] in open syllables, whereas stressed [e] diphthongised to [ja] in closed syllables (b).

The typological overview in (5) also reflects the way in which the rise and fall of vowel length took place historically. That is, the dialects in (ii) must once have been like Tavetsch in (i), and Ligurian and Emilian dialects such as Genoese and Bolognese indeed still show traces of vowel length in proparoxytones, e.g. Genoese [ˈpaːgwa] ‘sheep’ < [ˈpeːguÓa] < pecora and [ˈlaːgrima] ‘tear’ < lacrima (p. 206). Cremonese too still has contrastive vowel length, as in [ˈtaːvula] ‘table’ vs. [ˈfabula] ‘tale’ (p. 87). Similarly, type (iii) dialects must once have been like type (ii) dialects, and Old Milanese indeed shows CVL in paroxytones (pp. 170–171). Thus CVL arose anew, and was lost to a varying extent again in these Romance varieties, first in proparoxytones, then paroxytones and finally oxytones.

What these contexts have in common cannot be explained by apocope (deletion of final unstressed vowels) or by compensatory lengthening, i.e. lengthening of a vowel to compensate for the deletion of a consonant, but has to be linked back in time to OSL. According to Loporcaro, the drawback of previous accounts of vowel length, such as Montreuil (Reference Montreuil, Wanner and Kibbee1991), Repetti (Reference Repetti1992) and Prieto (Reference Prieto and Mazzola1994, Reference Prieto and Repetti2000), is precisely that they attempt to derive vowel length synchronically, instead of assuming an account in which vowel length is underlying and contrastive. Loporcaro does not claim that vowel length cannot arise due to either of these two processes – on the contrary, he provides examples of this from Gallo-Romance (p. 104, following Morin Reference Morin, Gess and Arteaga2006) and Friulian (p. 236). His claim is essentially that CVL in Northern Italo-Romance arose as a result of OSL. More precisely, when degemination occurred in the phonological history of these varieties, the lengthened output of OSL was phonologised as a result of degemination. A strong point in favour of the link between degemination and CVL is that it allows for a unified account of all the cases where contrastive vowel length is found. If true, it also shows that Kiparsky's (Reference Kiparsky and Goldsmith1995: 656) claim that CVL can only arise in systems that already have an independent vowel-length contrast is in need of revision.

The question that remains is why contrastive vowel length, after having resurfaced, disappeared again rather quickly. The answer is provided in §5.4.1. Due to a general phonetic constraint, which is termed rhythmical compensation by Loporcaro, citing, among others, D'Imperio & Rosenthall (Reference D'Imperio and Rosenthall1999), the length of a stressed vowel decreases as the number of syllables to the right of the vowel increases. This is reflected by the arrow in (5); the further away from the right word boundary, the less natural it is to have vowel length. Rhythmical compensation also explains the direction of retreating CVL in the varieties discussed: first in proparoxytones, then paroxytones and finally oxytones. Why, then, is CVL lost from the final syllable, if rhythmical compensation suggests that this is the most natural environment for vowel length? The answer is provided by the markedness scale in (7) (p. 215, from Mester Reference Mester1994 and D'Imperio & Rosenthall Reference D'Imperio and Rosenthall1999).

1. (7)

   

Loporcaro claims that this hierarchy is purely phonological, and does not follow from any substantive constraints. According to the hierarchy, allowing final heavy stressed syllables is the most marked option, whereas the most unmarked option for the final syllable is to be the weak branch of a foot, or to be extrametrical, as in Classical Latin. The hierarchy explains why OSL in Standard Italian does not affect the word-final syllable, but does occur in paroxytones and in proparoxytones, as in [ˈʤoːvane] ‘young man’, [ˈaːvido] ‘greedy’ and [ˈaːrabo] ‘Arab’. According to Loporcaro, the major factor explaining the retreat of CVL in Northern Romance was the general phonetic constraint mentioned above, ‘the increasing difficulty of maintaining length as the number of unstressed syllables following the stressed one increased’ (p. 234), and its loss from the final syllable can be considered the result of the scale in (7).

Let us consider contemporary Central Friulian in more detail. This variety does not have long vowels in etymologically CVCV words which did not undergo apocope, as illustrated by [ˈkrude] ‘raw (fem.sg)’ < crudam, [ˈlove] ‘wolf (fem.sg)’ < lupam, but still has long vowels in oxytones, as in [ˈkruːt] ‘raw (masc.sg)’ < crudum, [ˈloːf] ‘wolf (masc.sg)’ < lupum (p. 236). The shortening of short stressed vowels in paroxytones is ‘confidently ascribed to rhythmical compensation’ but does not ‘imply that no [vowel-length] contrast at all is allowed to occur in the penult’, where, as a matter of fact, it does occur as the result of compensatory lengthening, as in [ˈspaːla] ‘shoulder’ < spatulam (p. 236). Rhythmical compensation was thus strong enough to shorten penultimate long vowels resulting from the first occurrence of lengthening (OSL), but later compensatory lengthening reintroduced CVL in the penult. Loporcaro rightly argues that the proposals mentioned above that try to derive vowel length synchronically as a result of devoicing of a following obstruent (Prieto Reference Prieto1992) or the loss of a following vowel (Repetti Reference Repetti1992) are unable to analyse long vowels in words like [ˈspaːla] and [ˈloːf] as ‘surface manifestations of one and the same series of underlying entities’. Conversely, in the analysis proposed by Loporcaro, ‘CVL, first arising via degemination, has become a property of underlying representation which may be retained or lost in different contexts [final vs. penult]’ (p. 237).

Contrary to what Loporcaro claims, however, the markedness scale in (7) is in fact not purely phonological in nature, but can be understood as following from the difficulty of realising a pitch accent in the final syllable in the close vicinity of a boundary tone (Gussenhoven & Jacobs Reference Gussenhoven and Jacobsforthcoming). Languages tend to resolve this crowding of tonal realisations on the final syllable either by lengthening the final syllable or by avoiding stress on the final syllable; this is also the reason why iambic languages often have initial rather than final stress in disyllabic words.

Rhythmical compensation, or rhythmical compression, as it is more commonly termed (Lindblom & Rapp Reference Lindblom and Rapp1973, D'Imperio & Rosenthall Reference D'Imperio and Rosenthall1999), accounts for duration differences as being due to vowel shortening as a consequence of word compression or position in the word. The greater the number of syllables in a word, the shorter the individual duration. There has been a great deal of discussion (e.g. Fischer-Jørgensen Reference Fischer-Jørgensen, Channon and Shockey1987) as to whether this is word-based (striving for more or less constant duration of the word) or foot-based (striving for more or less constant distance between stressed syllables), but in any case compression seems not to be limited to vowel length. Fischer-Jørgensen (Reference Fischer-Jørgensen, Channon and Shockey1987: 54) reports on a study in which it was found that the [il] in dealer is shorter than in deal, and that both the [i] and the [∫] in fish are longer than the corresponding sounds in fishing, noting that Lehiste (Reference Lehiste, Hammerich, Jakobson and Zwirner1971) found that all sounds, not only vowels, are shortened (see White & Turk Reference White and Turk2010 for a more nuanced account of polysyllabic shortening). This makes one wonder whether rhythmical compression is in fact the explanatory factor for the rapid retreat of CVL in Northern Romance. For one thing, there seems to be a conflict with the apparent ease with which new geminates arise, as illustrated in the forms in (4) above. Why is consonant length apparently more easily tolerated in proparoxytones than vowel length? It might also be possible to understand the rapid retreat of vowel length as the result of the difficulty of maintaining lexical vowel length in a system where it is not truly contrastive. In other words, vowel length, arising from OSL, became lexicalised and underlying after degemination, but failed to become a distinctive, contrastive property in the same sense as, say, the difference between voiced and voiceless, and was therefore soon lost again. In Dutch, for instance, vowel length is not underlyingly contrastive. However, long vowels do occur, but only in loanwords, mainly from English and French, and can form near-minimal pairs (crème ‘cream’ [krεːm] vs. tram [trεm] ‘tram’, serre [sεːrə] ‘bay window’ vs. verre [vεrə] ‘distant’, zoomen [zuːmə] ‘to zoom’ vs. zoemen [zumə] ‘to buzz’). Vowel length is certainly a lexical feature of Dutch, but at the same time it is not a contrastive property of Dutch phonology.

There is really just one overall point of criticism that I would like to raise. Throughout the book the author voices his objections to ‘studies in theoretical phonology and/or experimental phonetics which apply their synchronic machinery to the explanation of sound change without a clear notion of the historical change to be thereby ‘explained’, as though knowledge of the data to be accounted for were not a precondition for the soundness of the analysis’ (p. 16). Crosswhite (Reference Crosswhite2001) is taken as an example of such a study; Loporcaro claims that ‘the author, notwithstanding the merit of having included in her database linguistic evidence from a domain in which she is not a specialist, nevertheless cannot be credited with checking appropriately the extensive literature which has established the historical facts she mentions’ (p. 16). An explanation for the alleged absence of Standard Italian vowel lengthening in proparoxytones proposed by Prince (Reference Prince1990) and by Hayes (Reference Hayes1995) (cf. [ˈʤoːvane, ˈaːvido, ˈaːrabo] above) is ‘patently mistaken’ (p. 230), and can be put forward only ‘at the cost of [Calabrese (Reference Calabrese1984–85)] misrepresenting the empirical data’. On page 143 he observes that lengthening of the stressed vowel does happen in Cremonese and is ‘thus flying in the face of Prieto's categorical ban and calling for a subtler explanation’. Furthermore, alternative analyses of vowel length achieve poor descriptive accuracy, as ‘such theories keep on squeezing sound change into the synchronic phonology’, camouflaging the English Great Vowel Shift as a synchronic phonological rule of modern English. This is ‘not a thing of the past, since this type of abstract analysis is still proposed by practitioners of generative phonology (under the Kiparskian label ‘opacity’: see e.g. McCarthy Reference McCarthy and de Lacy2007: 108; Baković Reference Baković, Goldsmith, Riggle and Yu2011)’ (p. 143). This reads as if opacity cannot arise in synchronic phonology at all, and is thus not a real problem for phonological theory. Opacity can arise synchronically even as the result of postlexical allophony in modern Romance varieties, as in the opaque interaction of voicing and spirantisation across word boundaries in Corsican or Gran Canarian Spanish (Jacobs Reference Jacobs2009), or as it did in Latin, due to the regular application of syncope (Jacobs Reference Jacobs2008).

Loporcaro pays particular attention in Chapter 6 to the metrical foot, which is deemed not to be a relevant concept, as ‘current assumptions in metrical stress theory … prove disadvantageous, in my opinion, for a satisfactory account of the data analysed here’ (p. 227). He  illustrates this point by criticising, rightly, Mester's (Reference Mester1994) account of Latin syncope. The arguments put forward on pages 228 and 229 against the moraic trochee-based account of syncope as resolving ‘trapped’ syllables are well taken, and indeed are put forward by Lahiri et al. (Reference Lahiri, Riad, Jacobs and van der Hulst1999), who are cited on the following page, not for criticising Mester on rather similar grounds, but for defending a quantity-insensitive initial foot preceding main stress, given that ‘several of those examples show syncope under stress’ (p. 229). Loporcaro writes: ‘Mester (Reference Mester1994: 38) is misled by his sources here, as he quotes Lindsay (Reference Lindsay1894: 173) for the (incorrect) generalization that ‘post-tonic’ syncope under the penultimate accent law …’ (p. 229). Ironically, Loporcaro himself is here ‘misled’ by the same source as Mester: Lindsay (Reference Lindsay1894: 170–177) is very precise in pointing out that syncope always, throughout the history of Latin, affected an unstressed vowel, but that the stress system of Preliterary, Early or Plautinian Latin and Classical Latin differed, so what may seem to be syncope of a stressed vowel from a Classical Latin perspective (balineum > balneum) was in fact syncope of an unstressed vowel in Plautinian Latin, at which time quadrisyllabic words with the first three syllables light were still stressed on the first syllable: ′balineum (cf. Jacobs Reference Jacobs, Fikkert and Jacobs2003 for a more detailed account).

These final critical remarks certainly do not detract from the overwhelming appreciation I have for Loporcaro's book, which most definitely should be read by anyone interested in Romance and Latin phonology. All in all, the sightseeing tour provides lots of empirical data, often new, and is extremely enjoyable, whether or not the reader shares the theoretical approach of the tour guide.