¹ This has been the vulgate version since about 1870: it appears, for example, in Umpfenbach (1870), Dziatzko-Hauler (1898), Fleckeisen (Teubner, 1898²), Tyrrell (OCT, 1902), Laming (1902), Ashmore (1908), Sargeaunt (Loeb, 1912), Kauer-Lindsay (OCT, 1926), Marouzeau (Budé, 1956), Martin (1959), Rubio (1961), and Coury (1982); it is also the basis of the translations of Copley (1967), Radice (Penguin, 1967), and Due (1984). The only significant difference between these editions is that some (notably Kauer-Lindsay and Martin) prefer ‘exclude’ (∑) to ‘exculpe’ at 989, but this does not affect the argument of this article. Pratesi (1952) follows A except in 984–5.

² What follows is an amalgamation of the comments and stage-directions of the editors and translators listed in n. 1 above; see also Andrieu, J., Le Dialogue antique (Paris, 1954), pp. 223–4Google Scholar.

³ It is found, for example, in Muretus (1654), Camus (1687), Bentley (1727), Reinhardt (1827), Westerhoff(1831), Hickie (1854), Fleckeisen (18571), Parry (1857), Wagner (1869), and Sloman (1887), and it is the basis of Riley's translation (1898). Bond and Walpole (1879) give a mixture of A and ∑ readings.

⁴ See Grant, J. N., Studies in the Textual Tradition of Terence (Toronto, 1986)Google Scholar.

⁵ In general the Calliopian MSS of Terence, like most of their medieval Plautine counterparts, use three-letter abbreviations of the character-names to denote speakers, whereas the Bembinus uses an ‘algebraic’ system where the speakers are denoted by different Greek letters. But all these MSS presumably derive from earlier texts in which, as in early Greek papyrus texts, changes of speaker were marked by paragraphi and dicola but individual speakers were only occasionally identified. See Andrieu, op. cit., pp. 209–72, Lowe, J. C. B., BICS 9 (1962), 27–42Google Scholar, Jory, E. J., BICS 10 (1963), 65–78Google Scholar.

⁶ There are over 130 places in Terence where the OCT reports MSS disagreements over the speaker-assignation. Confusion tends to arise in scenes involving two characters of the same type: other good examples in Terence are Pho. IV.iii (Demipho and Chremes again), and the final scene of Eun. (involving the adulescentes Phaedria and Chaerea).

⁷ On the characterization of Demipho and Chremes, see Arnott, W. G., G&R 17 (1970), 45–52Google Scholar.

⁸ For the phrase rapere in ius, cf. Plaut, . Poen. 1336Google Scholar, Rud. 859, Hor. Sat. 1.9.77. The meaning of plain rapere has to be judged from the context. In similar scenes in comedy, sublimem rapere at Ter, . Andr. 861Google Scholar means ‘carry off inside’, and the same phrase at Plaut, . Men. 995Google Scholar means ‘carry off to the doctor's’, but rapere at Plaut, . Rud. 868ffGoogle Scholar. does mean ‘carry off to court’.

⁹ So Due, following the suggestion of Dziatzko-Hauler.

¹⁰ There is an interesting confusion in Martin, who in his text gives 982 (’adsequere…euoco’) to Chremes but then says in his notes (my italics) ‘addressed to Chremes, while Demipho turns to his own house to fetch reinforcements’; similarly in his text he gives 983 (‘enim nequeo…accurre’) to Demipho but then comments (on tecum 984) ‘sc. Demipho, who comes up in answer to Chremes' call for help…’.

¹¹ Thus there is no sign in the text of Ad. 155ff. that Parmeno enters with Aeschinus and Sannio until Aeschinus addresses instructions to him at 168ff.; and at Plaut, . Bacch. 799ffGoogle Scholar. there is no indication in the text which would pinpoint the departure of Artamo. On the lorarius as a silent character, see Prescott, H. W., CPh 31 (1936), 99–103Google Scholar; for other ‘silent’ exits in Terence, see Gilula, D., AJPh 100 (1979), 519–30Google Scholar.

¹² The presence of lorarii in this scene is assumed in Echard's 1689 English translation (as reproduced in Robert Graves's Belle Sauvage edition of 1962). This adds the stage direction ‘Enter a servant or two from Demipho's’ after line 984, and has Demipho uttering the commands of 985 (‘rape hunc’) and 988 (‘pugnos…ingere’) to the servants. I cannot forbear to quote also its stage directions at the end of the scene: ‘Enter Nausistrata; the Servants unhand Phormio; Chremes looks very sheepish.’

¹³ Though the speaker of 982 goes to fetch seruos (plural), the following singular imperatives (rape, etc.) imply that only one actually emerges. There is no real contradiction here.

¹⁴ The division of this utterance in ∑ between two speakers may go back to a misunderstanding of a dicolon after ‘opprime’; there are a number of cases in Greek papyrus texts where the dicolon marks a change of addressee rather than a change of speaker (e.g. Men, . Dysk. 177Google Scholar, Perik. 188–90: see Gomme-Sandbach ad locc.). Even if the utterance is left undivided (as in A), ‘inpurum’ is better taken as an anticipatory accusative with ‘uide’ than as agreeing with ‘os’.

¹⁵ The word ‘sequitur’ here reinforces the idea that ‘rape’ in 985 means ‘haul off’ and in particular ‘haul off to court’. If ‘rape’ just means ‘seize’, ‘sequitur’ has to be taken as equal to obsequitur (‘give in’).

¹⁶ This reading of the scene depends on the assumption that ‘dum ego hue seruos euoco’ in 982 can be interpreted as a stage direction ‘goes to house and calls out slaves’. This may be a bold assumption. The situation in the Ad. example quoted above (n. 11) is different in that the whole group is entering simultaneously from the wings; in the normal situation where the senex summons lorarii from indoors, either he goes to fetch them himself and returns with them (Plaut, . Bacch. 794–9Google Scholar, Men. 955/990, Most. 1038/1064) or he calls them out by name (Andr. 860, Plaut, . Capt. 657Google Scholar, Rud. 657). I am indebted to my colleague Elizabeth Duke for the alternative suggestion that lorarii do not in fact appear but that Terence is deliberately evoking the conventions of the lorarius scene in order to make the actions of the series seem even more feeble and ridiculous. This is a suggestion worth considering by those who do not accept the view advanced here.

¹⁷ See esp. Büchner, K., Das Theater des Terenz (Heidelberg, 1974), pp. 353–9, 479–81Google Scholar, Lefèvre, E., Der Phormio des Terenz und der Epidikazomenos des Apollodor von Karystos (Munich, 1978), pp. 34–7, 75–8Google Scholar.

¹⁸ On similar lines Büchner (op. cit., p. 321 n. 14) argues that ‘bei der Existenz eines Gynaikeion’ the Apollodoran Demipho could have more easily entered his house at Phorm. 311ff. without the risk of finding Sophrona and Phanium inside than his Terentian counterpart.

¹⁹ On Greek houses, see Walker, Susan, ‘Women and Housing in Classical Greece: The Archaeological Evidence’, in Cameron, A. and Kuhrt, A. (eds.), Images of Women in Antiquity (London, 1983), pp. 81–91Google Scholar.

²⁰ I can find only one example in Greek comedy and one in Roman.

²¹ Sommerstein, A. H.. ‘The Naming of Women in Greek and Roman Comedy’, QSt 6 (1980), 393–18Google Scholar.

²² There are calls to a slave, or slaves, or an unidentified ‘someone’ to open the door at Men, . Dysk. 459, 911Google Scholar, Epitr. 1076, Mis. 206; Plaut, . Amph. 1020Google Scholar, Bacch. 582, Capt. 830, Merc. 131, Mil. 1297, Most. 445, 900, 988, Rud. 413, 762, Stich. 309, Trin. 870, True. 663; Ter, . Ad. 634Google Scholar, Eun. 530, Phorm. 152.

²³ For Menander, see Frost, K. B., Exits and Entrances in Menander (Oxford, 1988), pp. 9–10Google Scholar. Frost cites three examples from Dyskolos of the ‘summoning by a shout’ of a freeborn character; but these all relate to people inside the shrine of Pan, which is rather a different matter from a private house, and none of them offers any real parallel to the Phormio situation: at 635 Simiche's ‘Gorgias! Where are you?’ is an exclamation of despair rather than a summons, at 667 Gorgias, who has just emerged from the shrine, calls back to Sostratos to come out and help, and at 889 Getas, having just emerged, calls back to Sikon. Demeas' call to Chrysis, at Sam. 730Google Scholar, also cited by Frost, belongs rather to the category of ‘instructions shouted inside’ than to that of ‘calling out on to the stage’.

²⁴ I am grateful to Geoffrey Arnott, Peter Brown, John Grant, Richard Hunter, Christopher Lowe, and Malcolm Willcock for helpful comments on a draft of this article; they should not be assumed to assent to its conclusions.