¹ Russell, D.A., Criticism in antiquity (London 1981) 4Google Scholar.

² Ibid. 6, 130; compare DK 82A1.

³ Stanford, W.B., Greek metaphor: studies in theory and practice (Oxford 1936) 18Google Scholar.

⁴ See Spengel, L., Rhetores graeci (Leipzig 1853–1856), II 254Google Scholar.

⁵ This will be obvious to anyone who browses, for instance, through the papers delivered at a Chicago symposium in 1978, and collected in Sacks, S. (ed.), On metaphor (Chicago and London 1979)Google Scholar. The presuppositions of views such as Stanford's are most explicitly challenged in Donald Davidson's ‘What metaphors mean’ (Sacks 29–45). Davidson's arguments are vigorously criticized by other contributors; but their essays will also help to undermine the illusion (if any still remains, sixty years after Stanford and twenty years after the Chicago symposium) that Stanford's simple formulations are adequate to their task. See particularly Ricoeur, Paul, ‘The metaphorical process as cognition, imagination and feeling’ (Sacks 141–57Google Scholar), Goodman, Nelson, ‘Metaphor as moonlighting’ (175–80)Google Scholar.

⁶ Particularly Rep. 392c–98b.

⁷ In making these statements I follow what I think is the usual interpretation of Aristotle; but I am not wholly convinced that it does him justice. It is true that his discussions of the effects of metaphors focus principally on their capacity to make ideas vivid, or to give pleasure to an audience, and that they do little to suggest that what a metaphor conveys differs in meaning from what is conveyed by a corresponding literal statement. But there are certain passages, notably Rhet. 3.2.13, which may indicate a stronger and more interesting view. That passage certainly deserves close scrutiny. For present purposes, however, we may put it aside, on the grounds that its implications (if indeed it has any) about the power of metaphor to affect meaning are both thoroughly obscure and uncharacteristic of Aristotle's overall approach; and it seems clear that they failed to make any impact on the dominant strand of the later tradition.

⁸ For the text, see Winnington-Ingram's Teubner edition (Leipzig 1963). There is a translation in my Greek musical writings, vol. 2 (Cambridge 1989)Google Scholar, which I believe to be preferable to that offered in Mathiesen, T.J., Aristides Quintilianus: On music in three books (New Haven and London 1983)Google Scholar. Mathiesen's introduction can usefully be consulted, however, on issues surrounding the question of the work's date and intellectual milieu.

⁹ I do not mean to argue that Aristides is working here with a fully developed distinction between ‘sense’ and ‘reference’, as articulated by Frege; but I do mean that any philosophically adequate analysis of his strategy would find it useful to deploy a distinction of just that sort. The theoretical paraphernalia on which (I believe) he consciously builds, and which I examine below, can fairly be said to have affinities with Frege's. But it is certainly not identical with it; and readers who persevere to the end of this paper may be inclined to judge that the Fregean analysis would have served Aristides' purposes better than did the Stoic apparatus on which, as I argue, he actually drew.

¹⁰ The claim that an ἔννοια is ‘never’ value-neutral may be a marginal – but only a marginal – exaggeration. In the chapters with which we are concerned, the word occurs seven times (65.23, 25; 67.16; 68.14; 69.1; 70.18; 72.7). The last two, taken in isolation, might not unreasonably be construed as referring to ‘conceptions’ without any implication that they are evaluatively loaded. None of the others can plausibly be interpreted in that way; neither can an earlier occurrence at 57.4. About the remaining three appearances of the word in other parts of Aristides' text (56.25, 76.29, 85.21) no clear decision is possible. Evaluative implications are quite plainly attached also to the two occurrences (68.19, 73.7) of the related word ἐννόημα, of which I shall say more below.

¹¹ Adv. math. 8.11–12. One might argue with some justice that Sextus is scarcely to be trusted on the delicacies of terminology. His testimony is worth recording none the less. It shows, if no more, that the identification between σημαινόμενον and λεϰτόν could be made, in a Stoic context, at this date; and since nothing in Sextus' critique hangs on the terminological point, there is no reason why it cannot have been made also, or even originated, within the Stoic camp.

¹² Alexander, In Ar. Top. 359.12–16; Stobaeus 1.136.21–137.6; Diogenes Laertius 7.60–1. See passages 30D, 30A and 30C respectively in Long, A. A. and Sedley, D.N., The Hellenistic philosophers (Cambridge 1987)Google Scholar.

¹³ But substantial help is given by Long and Sedley's notes, op. cit. I 181–3, II 181–5; see also the passages they cite on the subject of λεϰτά, and their discussion of the subject, I 195–202, II 196–204.

¹⁴ See for example Aetius 4.11.1–4. and other passages collected in Long and Sedley's Section 39. with their discussion.

¹⁵ It is used here, I think, in just the right manner and context for a Stoic discussion. See in particular Diogenes Laertius 10.33 (Long and Sedley 17E).

¹⁶ One consequence of the peculiar ontological status assigned by Stoics to ἐννοήματα. (and that assigned to λεϰτά too, though there the metaphysical issues are slightly less problematic) is that they are incapable of entering into causal relationships. As the PCPS reader pointed out to me, this entails that they cannot act upon anything, and hence – it would seem – cannot produce changes in anyone's soul. Hence no one and nothing can (apparently) be affected or altered by the meaning of anything that is said. Rather obviously, this is a conclusion that Stoics who sought to understand the power of language could not embrace without further discussion; the difficulty had to be circumvented, and the route around it was always likely to be complicated. But the fact that Aristides' account passes these problems by without the least acknowledgement is to be construed, I suggest, merely as a symptom of his lack of interest in refined ontological subtleties, rather than as a proof that the whole theory must after all be alien to Stoicism.

¹⁷ Aristides seems to have been a voracious reader, and to have absorbed (or partly absorbed) ideas from sources of many kinds. He compounds them in ways that do not always make consistent sense, and which certainly cannot always be understood as expressing the views of any particular ‘school’. I do not claim that his treatment of metaphor is either fully consistent (let alone fully adequate) or consistently Stoic in its inspiration; traces of Platonism and Aristotelianism, among other things, are not hard to find. What I am suggesting is that we can identify one significant, continuing thread in his treatment as conceptually and terminologically derived from the Stoic repertoire; and that within this Stoicizing material is embedded an approach to the subject that differs in very significant ways from all others that have survived from antiquity.

¹⁸ Earlier versions of this paper were presented at meetings of the Cambridge Philological Society in January 1991, and the West Midlands Classics Seminar in March 1997. I am grateful to members of both audiences for comments and suggestions made on those occasions. I owe a special debt to this journal's anonymous reader for the trenchant and far-reaching comments which he or she made on a previous draft. I do not expect my revisions to have convinced that reader; but I submit that there is a substantial case to answer.