…but there are some who surprise. Remembering that this data-mining thing is a sledgehammer and not a delicate chisel, among the findings in the first quick and dirty round of datamining tests for status in Greek tragedy is that the servant (therapon) in Euripides’ Heracleidae does not talk like other servants and non-elite individuals. Much legwork remains in following all the markers through to separate the meaningful from the meaningless, but at the macro-level this makes some sense. In the scene with Iolaus (630ff), the servant takes something of a commanding tone with the aged hero and the interaction does not seem quite the same as that between either protagonist and messenger or protagonist and servants elsewhere. Does the servant’s less servant-ish speech help shape the contrast with old Iolaus’ attempt to arm himself? (This is, not coincidentally, a scene often suspected of being a specimen of Euripidean comedy.) That this particular servant is specifically marked at his appearance as a penestes (“serf”? cf. Aristophanes Wasps 1237), one wonders whether there might be something to his exceptionality. There is also the question of whether the messenger speech later on in the play (where we hear that old Iolaus has been rejuvenated as superman Iolaus) is in fact spoken by this same servant of Hyllus. For the data-mining runs we are using modified versions of Perseus project texts, so these two characters are currently separated out in the speaking parts and there seems to be no problem with fitting the messenger in with the rest of the non-elite individuals. Depending on how the servant’s speech is working at more detailed levels (and there are various filters for this, depending on how the search is shaped, e.g. in terms of surface features, lemmatized matches, or pairs of lemmas and pairs of words), this sort of difference between the therapon and the (possibly therapon) messenger speech might shed some light on the line attribution question as well.