The vulgate text at Prop. 3.10.17–18 reads as follows:Footnote ¹

According to Stephen Heyworth,Footnote ²polles is ‘a securely correct reading that T did not derive from any extant earlier MS … Such a conjecture would have been within the compass of the age of humanism; yet pellere is not absurd here (cf. Cic. Cael. 36) and pellas would be a more obvious correction’. If Heyworth is right, Prop. 3.10.17 provides strong evidence in favour of T's independence from the Petrarchan tradition (FLP).

Four further arguments support the hypothesis of an authentic polles. First, a similar confusion appears in Plin. HN 20.101, where pellere must be corrected to pollere (TLL 10.1.1009.46 s.v. pello). Secondly, the relative clause qua polles surely conforms to classical syntax, as shown by a parallel example in Valerius Maximus (TLL 10.1.2539.15–18 s.v. polleo):

Thirdly, the collocation of polles and ablative qua referring to forma is semantically equivalent to the noun phrase forma potens that occurs at 2.5.28 (Cynthia, forma potens; Cynthia uera, leuis) and 3.20.7 (est tibi forma potens, sunt castae Palladis artes).Footnote ³ Fourthly, polles in verse 17 may provide an etymological wordplay with pollice in verse 14 (et nitidas presso pollice finge comas).Footnote ⁴

Yet, another story remains possible. The first attestation of qua (or quo) polles dates from the very beginning of the fifth century:

and the following sample suggests that this relative clause came to function as a stock expression in Late, Medieval and Renaissance Latin:

Such evidence makes a humanistic correction plausible, all the more so since polleo does not occur elsewhere in Catullus, Virgil, Propertius, Tibullus or the Tibullan corpus; Horace uses it on one occasion only (Ars P. 242), and Ovid six times (Fast. 1.269, 2.425, 6.54; Met. 5.508, 7.196; Tr. 3.10.55) but always in the present participle form. It follows that pollice (verse 14), far from supporting polles, may have favoured the interpolation.

The alternative possibility of reading a form of pello in 3.10.17 is supported by two passages that deal with the attraction exerted by feminine beauty, that is by a true forma potens:Footnote ¹⁴

Yet, if we opt for the verb pello, further adaptations prove necessary. Neither future pelles nor subjunctive pellas capture the fact that Propertius is dealing with the actual powers of the girl's beauty – a point Heyworth seems to overlook when he claims that ‘pellas would be a more obvious correction [than polles]’. Indicative pellis, while appropriate to the context and recalling qua primum oculos cepisti ueste Properti | indue in verses 15–16, creates a metrical oddity, viz. an unusual lengthening of short -is at the penthemimeral caesura. Book 2 contains two pentameters with the same anomaly:

But even if we do not emend them away, we should observe that, in both cases, the metrical deviation at hand involves a lexical repetition supplemented with rhyme, and that the initial aspiration of haec increases the plausibility of indifference to quantity in 2.8.8.Footnote ¹⁵ Thus the mere presence of leonine rhyme between pellis and perennis need not be seen as sufficient for the verse et pete, qua pellis, ut sit tibi forma perennis to be metrically acceptable.

The solution consists in assuming that ut stems from the corruption of a second occurrence of imperative pete:Footnote ¹⁶

Confusions between p and u abound in manuscripts,Footnote ¹⁷ so that pete > uete could have been read as uti and (subsequently or simultaneously) corrected into ut for the sake of metre; indeed, classical poetry rules out pyrrhic uti (see for example 1.15.24, 2.22.35, 2.23.9, where the second syllable is scanned long). Once pellis was followed by ut, scribal correction to pelles easily eliminated the metrical anomaly.

Ovid has a quite similar repetition of pete … petes preceding sit tibi:

We may surmise that he was struck by the morphophonological and grammatical pattern of his Propertian model, where leonine rhyme, alliteration in pe-, and the distribution of vowels (eight occurrences of short e; five of short i) contribute to isolating qua … forma. In fact, his line exhibits a still more complex network of sound parallelisms: dumque pe- nearly copies tu pete; (sit) tibi tac- and (tac)ta manus are mirror images of tu pete and dumque pe(tes). From a syntactical point of view, the two verses differ radically: in Ars am. 578, pete and petes take cibum as their direct object and the absence of ut before sit tibi in Ovid, far from being a grammatically optional feature as in Propertius, follows from the fact that the coordinate structure of the sentence links the subjunctive form to imperative pete. But such morphophonological echoes frequently involve syntactically heterogeneous materials: for instance, nec tibi crediderim uotis contraria uota (Lygdamus = [Tib.] 3.4.83) imitates both tunc ego crediderim uobis et sidera et amnes | posse … ducere (Prop. 1.1.23–4) and quod si forte tuis non est contraria uotis (Prop. 1.5.9).

Notice, to conclude, that examples of pello with an implicit direct object appear at Ter. Eun. 1080, Ov. Met. 10.486, Tac. Ann. 14.31.3 and perhaps Verg. Aen. 11.901 (where the variant reading poscunt may be preferred on the basis of 4.614).Footnote ¹⁸