1 Tradition and Originality in Roman Poetry (Oxford, 1968), pp. 743ff.
3 Williams' discussion is briefly criticised by Perret J. for being ‘beaucoup trop radicale’ (Latomus 28 , 718). For a favourable appraisal, see Benediktson D. T., ‘Vocabulary Analysis and the Generic Classification of Literature’, Phoenix 31 (1977), 341ff.
4 Williams was not the first to offer this criticism of Axelson: cf. Ernout A. in his review of Unpoetische Wörter, Rev. Ph. 21 (1947), 68, on the subject of diminutives: ‘un simple index, une pure statistique, même comparative, sans examen des conditions d'emploi, ne peuvent, à eux seuls, fournir les éléments suffisants d'un jugement sur le style, ou la langue’.
5 On diminutives in Augustan poetry, see Gow A. S. F., CQ 26 (1932), 150ff. The comments on Aen. 4.328 of Austin R. G., P. Vergili Maronis Aeneidos Liber Quartus (Oxford, 1955), echoed by Williams R. D., The Aeneid of Virgil, Books 1–6 (London, 1972), demonstrate that followers of Axelson have not been slow to proceed to this second stage.
6 This assumption, which underlies both Axelson's and Williams' discussions, is not always valid, as I hope to demonstrate.
7 Cicero, for instance, has uxor 84 times in his speeches; coniunx appears 34 times, but mainly in stock phrases such as coniuges liberique.
8 The Massilians' impassioned appeal to Caesar for neutrality reaches its climax with a statement that they do not fear to suffer ‘pro libertate’ the fate of Saguntum: ‘pectoribus rapti matrum frustraque trahentes I ubera sicca fame medios mittentur in ignes | uxor et a caro poscet sibi fata marito, | volnera miscebunt fratres bellumque coacti | hoc potius civile gerent’ (3.351ff.).
9 Prop. 2.28.22, 32.57, Ov. Am. 1.9.36, A.A. 2.362, 3.20, 110, R.A. 775, Fast. 2.627, E.P. 3.1.121, Tr. 1.6.19, 2.405, 5.14.37.
10 A.A. 3.685f.: ‘nec cito credideris: quantum cito credere laedat | exemplum vobis non leve Procris exit’.
11 So also the examples at Ov. A.A. 1.556, 686.
13 Cf. Fraenkel E., Horace (Oxford, 1957), p. 193: ‘when I have reached Aphrodite's serene speech at the conclusion of the poem [where uxor is found] I seem to have forgotten all about the girl Galatea’.
14 I.e. dat. sing., nom./voc./acc. pl. The gen. pl. would be possible before a vowel, but is not in fact so used.
15 Here it is acc. sing.; virginem (before a vowel) is not used by Virgil but does occur at Ov. Met. 6.524.
16 In Comedy, virgo is used 55 times by Plautus and 68 by Terence; puella 26 by Plaut., 6 by Ter., normally referring to babies (e.g. Plaut. Cist. 124, 135, Ter. H.T. 627) or, of older girls, to meretrices or ancillae. Among prose writers, puella is not used by Caesar, Cicero in his Speeches, or Sallust.
17 I have discussed this word-pair in greater detail elsewhere; see Glotta 61 (1983), 119–43.
18 A possible exception is the reference to Syrinx at 1.712.
19 But contrast 762: ‘virgineas ne crede manus’. The alliterative effect was also, no doubt, an important consideration in the choice of puella here.
20 E.g. Prop. 1.1.15, 3.19.26, Ov. A.A. 1.109, 125, 458, 2.187, 3.631.
21 The exception is Horace's Odes, where virgo appears twice with reference to unmarried women in an erotic milieu (1.6.17, 3.15.5: in the latter case the age factor is prominent, the young virgines being contrasted with the ageing ‘uxor Ibyci’).
22 E.g. Cat. 64.86, 97, Lucr. 3.1008, Livy 3.44ff., Sil. 10.494, Plin. Ep. 1.14.8.
23 For a similar distinction based on genre, compare Hor. Ep. 2.1.132, where the phrase ‘castis cum pueris ignara puella mariti’, is employed of the same persons who, in the more elevated opening lines of the Carmen Saeculare, are referred to as ‘virgines lectas puerosque castos’ (line 6).
24 Many occurrences of minister-tra must be discounted, where it is not a synonym for famulus/servus. Often it has a specialised sense of waiters at a banquet (e.g. Virg. Aen. 1.705, Tib. 1.5.34, 9.25, Ov. Met. 11.119) or assistants at a religious ceremony (e.g. Ov. Met. 2.717, 9.233, Prop. 4.4.44); on many occasions it is used in the general sense of attendant (not of servile status), e.g. Virg. Aen. 2.100, Ov. Met. 2.837, 9.90.
25 Cf. Westermann W. L., The Slave Systems of Greek and Roman Antiquity (Philadelphia, 1955), p. 2; Finley M. I., The World of Odysseus (London, 1956), p. 62; Calhoun G. M. in A Companion to Homer, ed. Wace A. J. B. and Stubbings F. H. (London, 1963), p. 442.
26 Servus is used to the exclusion of famulus by Caesar, and by Cicero in his Letters, Speeches and Rhetorical works.
27 Serva occurs twice in the Aeneid (5.284, 9.546) but may be regarded as an archaism.
28 Ernout/Meillet s.v. suggest that famulus denoted a servant in some specific capacity (such as valet), whereas servus refers to the legal status of a slave.
29 See R. D. Williams in his commentary on Book 3 (Oxford, 1962), ad loc.
30 E.g. with reference to Achillas, Ptolemy 's praefectus regius (10.522), or Pothinus (8.538, 10.100, 341), a eunuch acting as regent for the young king (cf. Caes. B.C. 3.108, 112).
31 See Fordyce C. J., Catullus (Oxford, 1961), pp. 262–3.
32 On the stylistic difference between Am. 2.7 and 2.8, see further my article ‘Ovid Amores 2, 7 and 8: The Disingenuous Defence’ in WS 96 (1983) 91–103.
33 Propertius uses dapes in a mythological context at 3.22.30.
34 For Virgil's ‘concern to invest commonplace notions with a distinction consonant with their place in a heroic narrative’, see Fordyce C. J.'s comments on Aen. 7.110ff., with the references to ancient rhetorical theory there quoted (P. Vergili Maronis Aeneidos, Libri VII–VIII, Glasgow, 1977).
35 I cannot accept this premise, since the second half of the Aeneid is closer to contemporary Rome only in location, not chronology, and the conventional features of the epic world (e.g. divine machinery) are retained.
36 The following is a complete list (each word occurs once unless otherwise specified): Books 1–6: apud (4), atque before consonants (8), eos, etsi, horum, huius (3), idcirco, ideo, illius, (4), iucundus, lassus, metuo — forms from perfect stem (2), mi (2), minime, nemo (3), obitus, obtrunco (3), omnino, onustus, palmula, parvulus, praeclarus, praeficio (2), pravus, propter (2), puella (2), quocirca, rogito, trucido, vero (not in formulae); Books 7–12: adficio, apud (2), atque (27), autem (not in formulae), cadaver, capillus, condicio, etenim, etsi, fortasse, gladius (5), horum (7), huius (2), illius, illorum, imperito, intersum, istius, lassus, metuo, mulier, muliebris, multo, mortem obeo, neco, nemo, obitus, obtrunco (3), occīdo (3), omnino, opera, prae (2), praeclarus (3), praesidium, praeter, profecto, proinde (2), propter, quivis, rogito, sagulum, sane, trucido.
37 Another factor that seriously affects the statistics is the preponderance of atque before a consonant in the second half. In fact, if this usage is discounted, along with words connected with war, the difference between the two halves is virtually cancelled out: we are left with 26 words used 41 times in the first 6 books and 31 used 43 times in the second.
38 Cf. Marouzeau J., Traité de Stylistique latine (Paris, 1935), p. 134.
40 He uses it there 64 times.
41 Brutus does occur once (1.34.9), but in a different sense. Demens is found at 1.37.7 and 3.19.23 and insanus at 1.16.15, 3.21.3 and 3.7.6; none of these five examples, however, refer to persons, the normal application of stultus.
42 For a full discussion see Monteil P., Beau et Laid en Latin (Paris, 1964), pp. 71–109, and with particular reference to Axelson on this word-pair, see A. Ernout, op. cit., pp. 64–7.
43 Cf. Cat. 86.5–6; ‘Lesbia formosa est, quae cum pulcherrima tots est, | tum omnibus una omnis surripuit Veneres’; here formosa incorporates the notion of physical beauty implicit in pulchra, but with the additional ingredient of sexual attractiveness. On the semantic development of formosus see Monteil, op. cit., pp. 23–60.
44 Instructive here is a comparison with Virg. Aen. 7.430f., where Turnus is urged to drive the Trojans from his native Tiber (‘Phrygios qui flumine pulchro | consedere duces pictasque exure carinas’); here the sentiment underlying pulcher is one of patriotism rather than mere delight in the river's physical appearance.
45 It is thus employed by Tibullus 9 times, Propertius 27 and Ovid in the Am. and A.A. 31. It may also refer to parts of the body, which are described as formosus in that they belong to a formosus person (e.g. Ov. Am. 1.18.12, of Corinna's feet).
46 In the Odes, formosa is used once (appropriately) of an anus who wants to appear formosa (i.e. sexually attractive to men). Pulcher is employed either of inanimate objects, often in the sense ‘well-omened’ (e.g. 1.36.10, 4.4.65), or as an objective term for physical beauty, referring to women in whom the poet himself has no sexual interest (e.g. 1.16.1, 2.8.7, 4.13.8). The only erotic context where formosa might have been substituted is 3.9.21.
47 Pulchra when applied to Dido reads like a stock Homeric epithet. The difference between Virgil's and Ovid's usage in respect to this word-pair can be attributed to the latter's more ‘elegiac’ attitude to love.
48 In some cases, the words may have other uses: e.g. pinus meaning pine-tree rather than ship, patres for senators, parentes for parents/ancestors. Accordingly, the figures include only those examples where there is an exact correspondence in meaning.
49 The passages alluded to are as follows: Ov. Am. 1.2.4, 5.25, 13.13, 14.22; 2.13.2; 3.5.1, 7.80, 11.8,13, Hor. Od. 2.6.7; Prop. 1.3.45; 2.13.28, 15.46, 26A.2, 33.26, 34.75; 3.7.69 (for lassus); Ov. Am. 1.15.10; 2.6.44, 8.20, 11.12, 39, 13.10; 3.2.34, 12.38, A.A. 1.388, 2.671, 3.94, 259, 311 (for mare) and Ov. Am. 2.11.50 (navis).
50 Fessus: 1.2.2, 3.88, 8.68, 10.42; 2.5.45; 3.10.10, 17.2; ratis: 1.3.40, 4.46, 7.20, 9.10; 2.3.40; 3.3.10, 5.24, puppis: 1.4.45. Two exceptions are ratis at 2.5.40 and puppis at 2.5.45, both in a narration.
51 Op. cit., chapter 5 (‘Tibull Analogetiker?’), pp. 114–33.
52 At Prop. 4.3.34, gladios is read by some editors (e.g. Fedeli, Richmond, Schuster, Rothstein), but the passage is much disputed (R. Hanslik in the latest Teubner  edition has radios; others who do not accept gladios include Barber, Helm and D'Arbela).
53 Prop. 1.21.7; 2.8.22; 3.14.11; 4.10.46, Hor. Od. 1.16.10, 37.23; 3.1.17; 4.15.19.
54 Cf. Ross D. O. Jr, Style and Tradition in Catullus (Cambridge, Mass., 1969), pp. 18–19.
55 In the case of this word-pair, the figures for Horace, and even more for Tibullus, are so small as to render them of dubious value.
56 Possible exceptions are 3.6.40 (eulogy of early Rome in the last ‘Roman’ Ode) and 4.6.20 (mythological allusion in a hymn to Apollo).
58 I have omitted three words included by Axelson: asellus, capella and masculus. The first two are the normal words in poetry for ass and nanny-goat respectively (cf. Gow, op. cit., p. 153 n. 2), while the diminutive status of masculus is dubious.
59 Cf. Gow, op. cit., p. 154.
60 Catullus (68.148) uses lapis in the specialised sense of a pebble used to mark lucky and unlucky days; this is elsewhere designated by lapillus (e.g. Mart. 9.53.5; cf. Plin. Ep. 6.11.3 calculus).
61 His second criterion — the degree to which the poet's personality enters the work — was capable of being dismissed more briefly (see above).
62 Cf. my statistics in Table I and the discussion of Diminutives.
63 I would like to thank Professor R. G. M. Nisbet for helpful comments on an earlier draft of this paper.