Several reasons have been proposed for the decline of infinitival complementation in Ancient Greek: the fact that the infinitive became morphologically restricted, the inherent redundancy of the Classical complementation system, and language contact. In this article, I explore yet another reason for the decline of the infinitive: I argue that the system of infinitival complementation became fundamentally ambiguous in its expression in later Greek. As has been noted previously, the loss of the future and perfect tense had a serious impact on the use of infinitival complementation. However, rather than there being an 'omission' of temporal distinctions, as previous studies have claimed, I argue that the present and aorist infinitive became polyfunctional, being used for anterior, simultaneous, and posterior events. Next to temporal ambiguity, a second type of ambiguity occurred: 'modal' ambiguity or ambiguity with regard to the speech function of the complement clause. Already in Classical times, the present and aorist infinitive could be used after certain verb classes to encode both 'propositions' and 'proposals' (offers/commands), an ambiguity which continues to be found in later Greek. The study is based on a corpus of documentary texts from the Roman and Byzantine periods (I-VIII AD). 1

In the present article, I apply Fauconnier's Mental Spaces Theory to the diachronic analysis of the Ancient Greek periphrastic perfect. I argue that the periphrastic construction started out as a 'resultative' perfect, with FOCUS and EVENT located in the same mental space. I show that, contrary to what is sometimes believed, the construction was not limited to a purely stative meaning, but underwent the cross-linguistically attested semantic shift from resultative to anterior, whereby an additional non-FOCUS EVENT-space was constructed. In fourth-century Classical Greek, we witness the further extension of the periphrastic construction with regard to semantics, morphology and discourse context. I close the article with some remarks on the possible aoristicization of the periphrastic perfect.

It has been claimed that Archaic and Classical Greek had two main types of headed relative clauses: (i) postnominal externally headed relative clauses; and (ii) internally headed relative clauses (Perna 2013a;2013b;Fauconnier 2014;Probert 2015). In this article, we take a closer look at the semantic and syntactic properties of the second category in Post-classical and Early Byzantine Greek (I-VIII AD). Analysing a corpus of documentary texts, we show that a good deal of the examples in this period do not correspond to the established properties of internally headed relative clauses in the history of Greek. This leads us to propose that at least some examples that are apparently internally headed should be revised as a third relative clause type, namely prenominal externally headed relative clauses. We hypothesise that such examples came into existence through form-function reanalysis of internally headed relative clauses, a process which we suggest took place already in the Classical period (V-IV BC). In the last part of our article, we investigate the motivation for the choice of internally headed and prenominal externally headed relative clauses over the postnominal ones: we show that such examples occur strikingly frequently in formal texts such as contracts, petitions and formal letters. We propose that in such texts, internally headed and prenominal externally headed relative clauses, which are syntactically more complex, function as 'transparent signifiers' (Hodge & Kress 1988), serving as a marker of a higher social level. . 6 In this article, we refer to the (nominal) constituent that is semantically shared by both the relative clause and the matrix clause as 'head noun' or simply as 'head'. A number of alternative labels have been chosen in the literature to refer to the same (syntactic) object as 'head (noun)', among which are 'antecedent', 'pivot', or 'domain nominal'. For the purposes of this article, nothing hinges on these terminological differences. 7 With HEAD standing for 'head noun', rel pro for 'relative pronoun', i for 'anaphoric relationship', V for 'verb', and Ø for the gap/presumably deleted constituent at the relativisation site. 8 Compare Fiorentino (2007: 275-6) for this notation. The head noun assumes a grammatical function both within the relative clause, precisely at the relativization site, and outside of it. Very much in line with Fiorentino (2007: 275-6), we assume that this is established with a double anaphoric relationship in the relative clause construction: (i) between the head noun and the relative pronoun; and (ii) between the relative pronoun and a 'deleted' constituent. 530

Classical Greek (V -IV BC) is known for the complexity of its complementation system, involving infinitival, participial and finite verb forms. In Post-classical Greek (III BC -VI AD), a simplification of this system takes place, whereby finite complementation becomes much more frequent, and ὅτι is used as a 'generic' complementiser. This article analyses to what extent complementation patterns other than ὅτι with a finite verb form and the accusative with infinitive are still used in the Post-classical period (I -VI AD), focusing on documentary sources (that is, letters and petitions). I show that various 'minor' complementation patterns are (still) attested; some of them are known from Classical Greek, while others are entirely new formations. I furthermore argue that 'factivity' and 'formality' are two key factors in explaining the distribution of these patterns.

While the Greek documentary papyri constitute a rich resource for linguistic innovation, they cannot be considered a linguistically homogeneous corpus. In this article, I show that even in a single archive, the “ katochoi of the Sarapeion” archive, considerable variation exists in terms of phonology/orthography, morphology, and syntax. I argue that the different types of documents contained in this archive—that is, dreams, letters and petitions—can be situated on a linguistic continuum or “register” continuum. To account for the linguistic differences between these documents, the so-called “tenor”-vector plays a particularly important role.