2. I am referring here, in obviously general terms, to overall strategies, and I am not trying to deny the important influence of Hesiod on the De rerum natura. For more detailed observations on this matter cf. MD 19 (1987) 29ffGoogle ScholarPubMed.

3. For a specific reference to didactic poetry cf. Poetics 47b.

4. There are, of course, small-scale plots to be found in some narrative sections of the poets, such as – to name but one – the Iphigenia episode. These, however, do not structure the poem as a whole.

5. The interpretation of the ‘story of Memmius’ is by no means univocal. In a stimulating paper (in Mega nepios. Il destinatario nell'epos didascalico, Schiesaro, A., Mitsis, P. and Strauss Clay, J., eds. (Pisa 1993)Google Scholar [= MD 31]), P. Mitsis has recently argued, for instance, that the ‘ideal reader’ of the poem is expected to distance his or her experience from those of Memmius, who, according to Mitsis, is portrayed as a nepios who reacts incorrectly to the revelations of Epicurus. Others incline to interpret the ‘fading of Memmius’ in the books presumably composed at a later stage (3, 4 and 6) as evidence of the fact that Lucretius must have lost any hope of converting his patron to Epicureanism (thus Townend, G. B., ‘The fading of Memmius’, CQ ns. 28 (1978) 267ff.CrossRefGoogle Scholar; cf. also, by the same author, ‘The original plan of Lucretius' de rerum natura’, CQ n.s. 29 (1979) 101ff.)CrossRefGoogle Scholar.

6. It is hardly necessary to add that I will dispense with a number of so-called traditional issues of Lucretian criticism: whether the poem is ultimately optimistic or pessimistic, for one, but, more importantly, whether what we read was Lucretius' final decision or whether he would have gone back to a previous arrangement of the books which some scholars feel they are able to reconstruct.

7. Friedländer, Paul, ‘The pattern of sound and atomistic theory in Lucretius’, AJPh 62 (1941) 16–33Google Scholar. Cf. also the recent book by Dionigi, Ivano, Lucrezio. Le parole e le cose (Bologna 1988)Google Scholar.

8. 67 A 6 DK = Arist., Metaph. 985b13ffGoogle Scholar.

9. Pol. 278d; Theaet. 202bff.

10. L.L. 6.39.

11. 1.22–4.

12. 2.93; Cf. Plut. De Pyth. orac. 399e (= p. 362, 6–9 Usener).

13. Dionigi (n.7) 11ff.

14. Marouzeau, J., ‘La leçon par l'exemple’, REL 14 (1936) 58–64Google Scholar.

15. Hinds, S., ‘Language at the breaking point: Lucretius 1.452’, CQ n.s. 37 (1987) 450–3CrossRefGoogle Scholar.

16. I am not sure, however, that I can agree completely with Dionigi's claim that the circular arrangement of physical and linguistic reality prevents an identification of the primary model and the determination of which of the two systems precedes the other (p. 19, cf. also p. 33). Notwithstanding Dionigi's correct references to Plato and Aristotle (p. 35), there exists a difference, indeed an important one, between a hermeneutic model and the reality it tries to account for. While the model can indeed be traced to the observation of linguistic reality, it should be equally stressed that in the reductionist Epicurean atomism the absolute priority of physical reality is never questioned, irrespective of what interpretive model is chosen to account for it. And Dionigi's incidental remark that the alphabetical model is ‘more than an analogy’ (p. 35) should be interpreted, as I will try to do, in a direction exactly opposite to the one he privileges: since everything is made of atoms and void, letters, too, have a material dimension. Hence my radical disagreement with Dionigi's final line: uerba tene, res sequetur (p. 37).

17. I have attempted a more thorough analysis in Simulacrum et imago. Gli argomenti analogici nel De rerum natura (Pisa 1990)Google Scholar.

18. Simulacrum et imago (n. 17) 28ff.

19. There exist in the poem, however, analogies which illustrate or explain certain phenomena and are not ontologically related to the explanandum: these are ‘illustrative’ examples which help clarify an explanation. Cf. Simulacrum et imago (n. 17) 30ff.

20. Simulacrum et imago (n. 17) 30ff.

21. As I anticipated earlier, I speak of the ‘reality’ of letters in a sense which is very different from Dionigi's, who ultimately inclines to think of letters and words as ‘more real’ than the reality they represent. I maintain that – while there is no doubt that atoms and void are the fundamental components of the Epicurean cosmos – letters and words have a specific material aspect to them.

22. By this process of mental creation we can also think of non-existing things: cf. Kleve, K., ‘Wie kann man an das Nicht-existierende denken? Ein Problem der epikurischen Psychologie’, SO 37 (1961) 45ffCrossRefGoogle Scholar.

23. I refer to the fuller version of this paper (see n. 1) for a more extensive discussion of the modalities of perception presupposed by this interpretation.

24. I will use this expression again for convenience's sake, while the correct definition, as I indicated above, is that the poem contains the instructions for the creation of a set of material bodies in our mind.

25. A comprehensive analysis of these issues is offered by Schrijvers, P. H., Horror ac divina voluptas. Etude sur la poétique et la poésie de Lucrèce (Amsterdam 1970) 87–147Google Scholar, to which I am indebted throughout this section and to which I refer for further details.

26. 4.533–4: haud igitur dubiumst quin uoces uerbaque constent | corporeis e principiis.

27. I accept Schrijvers' interpretation of sub uerbo as ‘immediately after a word’, ((n. 25) 95–7); I also print the text of line 784 as he suggests at p. 98 n. 4. The reader for PCPS rightly points out that the word-play between simulacra and simul ac in lines 781–2 embodies the phenomenon vividly expressed by sub uerbo: anne uoluntatem nostram simulacra tuentur | et simul ac uolumus nobis occurrit imago.

28. Schrijvers (n.25) 104–6, and Long, A. A., ‘Aisthesis, prolepsis and linguistic theory in Epicurus’, BICS 18 (1971) 114–33Google Scholar, both with further bibliograpy. Cf. Epicurus, Ad Hdt. 38.

29. Simulacrum et imago (n. 17) 149ff.

30. Cf., in the proem, words such as genetrix (line 1), uoluptas (1), alma (2), frugiferentis (3), concelebras (4), concipitur (5), suauis (7).

31. The pattern of growth and decay is treated by Minadeo, Richard, The lyre of science: form and meaning in Lucretius' De rerum natura (Detroit 1969)Google Scholar.

32. The transposition of 1247–51 after 1286 was first proposed by Bockemüller in 1873, and has been recently defended with excellent arguments by Fowler, Peta G., ‘A commentary on part of Book Six of Lucretius' De Rerum Natura’, diss. Oxford 1983Google Scholar, and ‘Lucretian conclusions’, in D. P. Fowler and D. Roberts, forthcoming.

33. Cf. Fowler (n. 32).

34. 2.1128–32: nam certe fluere atque recedere corpora rebus | multa manus dandum est; sed plura accedere debent, | donec alescendi summum tetigere cacumen. | inde minutatim uiris et robor adultum | frangit et in partem peiorem liquitur aetas.

35. It is reasonable to consider the final praise of Epicurus at the beginning of book 6 as a delayed example of the ‘peak’ of civilization reached by mankind.

36. It is worth remembering in this context Peta Fowler's important observation ((n. 32) 30) that the last few lines of De rerum natura show traces of a process of decomposition which thematizes the poem's coming to an end. So, for instance, consanguineos at line 1283 breaks down into cum sanguine at 1285, morbus (1250) becomes mors (1251), and temptaret (1251) yields tempore (same line).

37. nam quacumque prius de parti corpora desse | constitues, haec rebus erit pars ianua leti, | hac se turba foras dabit omnis materiai. | haec sic pernosces parua perductus opella; | namque alid ex alio clarescet nec tibi caeca | nox iter eripiet quin ultima naturai | peruideas: ita res accendent lumina rebus (1.1111–17).

38. The repetition of this passage, with slight changes, at 1.921ff. and here is one of the most vexed questions of Lucretian philology. True to my initial promise I will dispense here with a detailed analysis of the problem.

39. Interesting observations on repetition in Dionigi (n. 7) 75ff., whose conclusions, however, I cannot wholly endorse (cf. below n. 41). Cf. also MD 24 (1990) 47–70Google ScholarPubMed, with further bibliographical indications.

40. A fine analysis of the importance of memory and meditation in Lucretius is offered by Clay, D., Lucretius and Epicurus (Ithaca–London 1983) 176ffGoogle Scholar.

41. … (transl. Loeb).

42. (transl. Loeb).

43. The same principles could be seen at work at the level of metric configuration of the line, and the particular kind of connection between lines.

44. Excellent general remarks – if somewhat exaggerated – are still to be found in Ferrero, L., Poetica nuova in Lucrezio (Florence 1949)Google Scholar.

45. There is a clear difference, however, between the number of the types of atoms and that of letters. Both are finite, but the number of atoms is ‘inconceivably’ (aperileptos) high, while the number of letters is small and perfectly well known. But the analogy is not fundamentally flawed.

46. Dionigi (n. 7) offers a very interesting analysis of this phenomenon in his third chapter. He maintains that the poem's insistence on reduplication, through, for instance, anaphora and repetition of near-synonyms, gives it a ‘fisionomia simmetrica’ which he relates in turn to the fundamental law of isonomia: ‘il bilanciamento di principi ed eventi opposti [è] affidato all'equilibrio e alla specularità delle forme linguistiche e stilistiche’ (p. 83). This suggestion is fascinating, and I cannot do it full justice here; yet the isonomia between creative and destructive forces would require, at the stylistic level as well, stylistic devices that ‘destroy’ some of their counterparts. This is why I am inclined to believe that isonomia works better as a general organizing principle at the level of the overall architecture of the poem.

47. On this notoriously difficult passage cf. at least Furley, D., ‘Nothing to us?’ in Schofield, M. and Striker, G., eds., The norms of nature: studies in Hellenistic ethics (Cambridge–Paris 1986) 75–91Google Scholar; Sorabji, R., Time, creation and the continuum (Ithaca 1983) 176–9Google Scholar.

48. De civitate Dei 22.28: genethliaci quidam scripserunt esse in renascendis hominibus quam appellant Graeci; hanc scripserunt confici in annis numero quadringentis quadraginta, ut idem corpus et eadem anima, quae fuerint coniuncta in homine aliquando, eandem rursus redeant in coniunctionem.

49. I offer a fuller discussion of the philosophical implications of this passage in ‘La “palingenesi” nel De rerum natura (3.847–869)’, to appear in the proceedings of the Congresso Internazionale di Studi Epicurei held at Naples in May 1993.

50. On ordo, Dionigi (n. 7) passim.

51. On ‘proemi al mezzo’ see Conte, G. B., Virgilio. Il genere e i confini (ed. 2, Milan 1984)Google Scholar. Conte traces the notion of such programmatic proems in the middle to Hellenistic models, while I stress the function that the proem to book IV has in the structure of De rerum natura as a whole.

52. I refer again to Peta Fowler's chapter in Fowler and Roberts (n. 32).

53. Again, according to Bockemüller's transposition.