¹ The few exceptions have been paraded tirelessly at least since Lucian Müller: Cat. 4.9 Propontida trucemue, 29.4 ultima Britannia, and indeed 4.18 impotentia freta (all of these in iambics), as well as Tib. 1.6.34 seruare frustra (Allen, W. S., Accent and Rhythm [Cambridge, 1973], pp. 140–1Google Scholar, and Hoenigswald, H. M., ‘A Note on Latin Prosody’, TAPA 80 [1949], 271–80Google Scholar, ‘Language, Meter, and Choice in Latin’, in Pieper, U. and Stickel, G. (eds.), Studia linguistica diachronica et synchronica [Berlin, New York and Amsterdam, 1985], pp. 377–83)Google Scholar, while the Ennian populea frus may well have to be struck off the list (Skutsch, O. [ed.], The Annals of Quintus Ennius [Oxford, 1985], p. 728).Google Scholar In what follows, ‘long’ and ‘short’, when applied to scansion (i.e. to syllables, not just to vowels), are interchangeable with ‘heavy’ and ‘light’, respectively.

² ‘…Plosive… + liquid… The evidence of early Latin verse…is quite clear that a syllable containing a short vowel followed by such a sequence was regularly light in quantity… At a later period, and under the influence of Greek practice, it became permissible to adopt for metrical purposes the alternative of treating syllables containing a short vowel followed by plosive + liquid as being of heavy quantity’ (Allen, W. S., op. cit. [n. 1], pp. 137–8).Google ScholarPace Timpanaro, S., ‘Alcune particolarità prosodiche nell' Anthologia Latina’, SCO 10 (1961), 156–60Google Scholar, and ‘Muta cum liquids in poesia latina e nel Latino volgare’, RCCM 7 (1965), 1075–1103Google Scholar, the long scansion is indeed far more probably a Greek importation than a Roman inheritance. Skutsch, O., op. cit. (n. 1), pp. 55–6Google Scholar, on Timpanaro and on Perini, G. Bernardi, Due problemi di fonetica latina (Rome, 1974), pp. 17–18, 70–7, 82–96, 108–9Google Scholar, states the case.

³ On Afranius see the end of the next footnote.

⁴ Kenney, E. J., CR n.s. 15 (1965), 188Google Scholar; Collinge, N. E., Collectanea linguistica (The Hague and Paris, 1970), p. 195Google Scholar; Allen, W. S., op. cit. (n. 1), pp. 140–1.Google Scholar Words like re-freno, re-fluit with both heavy and light scansion in their first syllables would seem relevant. Pit. Capt. 918Google Scholarcellas refregit omnis intus recclusitque armarium has an instructive coda; it was Lindsay who introduced … cc… into the text. In the case of muta cum liquida orthographic sanction in the form of gemination is of course not required from the post-scenic point of view (Kenney, loc. cit., is not wholly informative). Since ab–luit, ob–ruit have a long first syllable everywhere one would expect the converse for r–plet, re–primit, and also for re–flectit, the compounding seam always determining the ‘syllable boundary’. This, however, is not so, no doubt owing to the confusion – already Plautine – caused by syncopated reduplication in re-p(e)peri, re-t(e)tuli, red-(i)do and by the existence, however secondary etymologically, of the red- of red-eo etc. (see Sommer, F. and Pfister, R., Handbuch der lateinischen Laut und Formenlehre, v.l [Heidelberg, 1977], p. 158).Google Scholar Lucretius has heavy re–f.l… twice (3.502,4.442; both incidentally, in thesi) and light re–.fl… once, as well as six instances of heavy re–f.r… (four - 2.276,1121,4.440,6.531 - in the thesis of a spondee and two - 4.703,1085 - in arsi) against four instances of light re–.fr… On the other hand, the more lucid and also more artificial Lucretian compounds of the type largifluos, fluctifragus, ossifraga, siluifragus show a light …i–fl/r…–Vergil has nine cases of re–fi…, and one of re–fr…, all with light scansion (as well as bi–frontis, bi–frontem…) In Ovid re–fl… occurs six times and re–fr… five times; there are no heavy scansions. In Silius there are six light and three heavy (9.596, 15.738, 16.54, all in thesi) instances of re–fi… (there are none of re–fr…). In other words, aside from Lucretius (who, after all, also experiments not only with liquidus -… [hardly with i, i.e. with anomalous length by nature] but even with re-lig… - …, etc.), the f-clusters in these compounds actually tend to light scansion more than do the muta-cum-liquida ones. In this light Horace's cinif.lo becomes a little uncomfortable. Perhaps its compound nature was obscured. Perhaps, contrariwise, Horace, Sat. 1.2.98, with its tetrasyllable cadence and feminine caesura, comically suggests a Greek hexameter with a compound word occupying the same metrical position, as Nicander, , Alex. 605Google Scholar ⋯λβλαπτον Μελικέρτην∥, as the Editors have suggested to me. It is even remotely conceivable that the interior vowel is not just the ubiquitous compounding vowel (as in the less opaque largi-.fl…, etc.; … i-either generalized instead of the expected * …e-<* …o-; see Leumann, M., Lateinische Laut- und Formenlehre² [Munich, 1977], p. 390Google Scholar, or reflecting an etymological *…i…), and that we are instead dealing with a full-fledged *cinis-flo (cp. cinisculus) > cini(f)flo (cp., in a first syllable, dif-fluo); the MS. orthography of this virtual hapax – the word recurs once in Tertullian – hardly counts. The first syllable in Afer, Africa is long by nature, and so, presumably, is the middle syllable in Venafrum, whereas uafer, uafr…has ă. For Afranius the evidence, Hor. Ep. 2.1.57 || dicitur Afrani…, Lucan 4.4…castris Afranius illis||, 4.338…supplex Afranius armis||, is once again ambiguous.

⁵ Allen, W. S., Vox latino (Cambridge, 1965), p. 90Google Scholar; op. cit. (n. 1), p. 137.

⁶ This would not be entirely unnatural since f and s are both fricatives, though their privileges of occurrence differ as greatly as do their prehistoric antecedents. Still, there are ties: *sr, *bhr, *dhr all merge into fr/br, etc. See also Zirin, R. A., The Phonological Basis of Latin Prosody (The Hague and Paris, 1970), p. 41.Google Scholar

⁷ In Vergil (see below) and Ovid the avoidance is practically complete. Otherwise, in permitting himself exceptions, a given poet will use either the light (e.g. Hor. Sat. 1.10.72 saepe stilum uerlas…etc. [Satires only]; so also Lucilius, Lucretius, the Culex, Vergil [once only, Aen. 11.309, …ponite. spes…with ‘punctuation’], Propertius, Manilius; on Phaedrus, see Hoenigswald, H. M., op. cit. [n. 1], p. 381Google Scholar) or the heavy scansion (Cat. 17.24 pote stolidum etc.; also, Cicero, Tibullus, Grattius, Lucan, Statius, Silius, Juvenal, Martial, the Aetna; on Seneca, see Hoenigswald, H. M., op. cit. [n. 1], p. 381Google Scholar) but not both, except only for certain specific Homerisms like undo Scamandri at end of line, Cat. 64.357, etc. – an indication of the seriousness of the aesthetics involved. D. S. Raven calls the heavy scansions, together with their exceptions, ‘studied imitation[s] of Greek lines’ (Latin Metre [London, 1964], pp. 24–5Google Scholar). The details are in Hoenigswald, H. M., art. cit. (n. 1)Google Scholar; see also Postgate, J. P., Prosodia latino (Oxford, 1923), pp. 31–2Google Scholar; Collinge, N. E. (who is to be thanked for the reference to Pit. Rud. 198), op. cit. (n. 4), pp. 192–218Google Scholar, with important and somewhat different conclusions; Allen, W. S., op. cit. (n. 1), pp. 139–40.Google Scholar

⁸ Schulze, W., Zur Geschichte lateinischer Eigennamen (reprinted Berlin, Zurich and Dublin, 1966)Google Scholar, and Swanson, D. C., The Names in Roman Verse (Madison, Milwaukee and London, 1967)Google Scholar are on the whole useful guides, although Schulze, see above, p. 223, is awkward on the quantity in Martial 4.71.1.

⁹ i.e. to the exclusion of Phaedrus.

¹⁰ This reading of the metre avoids a split anapaest.

¹¹ So Frassinetti ad loc.

¹² Apparently, the f-clusters were perceived as close to the stop-and-liquid clusters with which they share their second, ‘liquid’ component, while qu did not fall in with the c-and-liquid clusters with which it could be said to share the first, or stop component. Instead, qu functioned as though it had no Greek model of any sort, and it was not allowed, except for Lucretius' occasional experimentation (n. 4), to make position (just as the su- of suadeo does not suffer avoidance as do the other word-initial s-groups; see n. 7). That there were, as there had to be, Greek transcriptions for qu, f, etc. is another matter; see Sommer, F., Handbuch der lateinischen Laut- und Formenlehre (Heidelberg, 1914), p. 284Google Scholar; Sommer, F. and Pfister, R., op. cit. (n. 4), p. 209.Google Scholar As Tarquinius is Ταρκύνιος, so Fabius is Φάβιος, and it is against that background that the MSS. spelling (two against one) Saphroni Martial 11.103.1, to which Professor M. D. Reeve has had the kindness to call my attention, must be seen in turn. This is also how Leumann, M., op. cit. (n. 4), p. 162Google Scholar, judges sporadic early renderings of ‘Φ’ by ‘f’ in genuine Greek words and names. Such spellings, especially in words not yet entrenched, become more frequent later (2nd century a.d. according to Allen, W. S., Vox graeca [Cambridge, 1987], pp. 22–6Google Scholar; 3rd/4th century a.d. according to M. Leumann, loc. cit.) when the Greek aspirates turn into fricatives. (Skutsch, O., op. cit. [n. 1], p. 703Google Scholar, may have in mind such things as the word-initial fr… of Fryg… Frug… Froeg… in the transmission of Apuleius, Met. 10.32, Flaccus, Siculus, agrim. p. 101Google Scholar, Turpilius 103 Rychlewska [frigus MSS.], Accius, , trag. 178Google Scholar [?], when he speaks, neutrally, of ‘alternation’ to throw doubt on the use of Livy's phalarica [ = fal…] as a sure argument for provenance from Fabius Pictor.) At the time, however, when the prosodic conventions were developed that accommodated Ruf.rae, uaf.ri, Saf.roni along with Ru.fria, Ru.fras, ua.fram, Sa.froni, there was no real symmetry: Greek Φ was, to put it crudely, closer to Latin f than any other Greek sound, but f was not the nearest available Latin rendering of Greek Φ, witness Aprodite [Praeneste], then Aprkodit…, and finally the familiar phl, phr and all the rest; see Leumann, M., op. cit. (n. 4), pp. 160, 162Google Scholar; W. S. Allen, see above (1987), p. 22.1 thank Professors W. S. Allen, G. N. Knauer, M. D. Reeve and O. Skutsch for reading versions of this note, and Professors J. A. Farrell and W. D. Smith for assistance with the resources of the Packard Humanities Institute, CD-ROM, at the Department of Classical Studies, University of Pennsylvania.