Marcelle Cole presents an interesting study of pronominal reference in Old English, nicely supplementing work available in the literature which shows, in brief, that in contexts with more than one possible referent, clause-initial nominative personal pronouns dominantly continue the topic (subject) of the previous clause, whereas clause-initial se-demonstratives dominantly switch the topic to a new referent of the previous clause. Cole adds to this with a study of the overall use of clause-initial pronouns in five Old English texts, which shows more variation than expected on the basis of this literature. Her conclusion is that se-forms by and large pick up discourse-new referents from the previous context. She further claims that her findings highlight how issues pertaining to style, such as the author–writer relationship, text type, subject matter and the conventionalism propagated by text tradition, influence anaphoric strategies in Old English. In this response article. I wish to counterbalance Cole's argument on two points, and make some suggestions for further research, based on recent psycholinguistic work.

First of all, Cole emphasizes two facts: the first reiterates the previous literature by observing that personal pronouns are considerably more free in their choice of referent. Secondly, she argues fairly strongly that so-called se-relatives are really Verb Second main clauses. This is an arguable choice, but the disambiguation of the main clause/relative clause distinction was certainly not as grammaticalized at this stage as it is in present-day Dutch or German. The presence of Verb Second primarily shows that se-relatives (which, like unambiguous se þe relatives, are always non-restrictive (Allen Reference Allen1980; van Kemenade Reference Kemenade1987)) do not have a relative conjunction þe, which would unambiguously mark them as subclauses, and possibly block Verb Second. This leaves us primarily with ambiguity between parataxis and hypotaxis, as argued by Los & van Kemenade (Reference Los, van Kemenade, Coniglio, Schlachter and Veenstra2017).

Cole has a number of observations that further qualify the observed tendencies. The primary refinement with respect to the previous literature is that, in contexts where there is no competing referent, se-forms dominantly pick up a discourse-new referent. This interesting result, which is of course entirely compatible with the topic switch function of demonstratives when there are competing referents, applies to clause-initial nominative as well as oblique se-forms. This observed distribution in Old English has not yet been investigated for present-day Dutch and German as far as I am aware.

My second point concerns the ‘crucial interaction’ between information structure-driven referent choices and matters pertaining to style and other textual factors, as postulated by Cole. The description of the texts in section 4.1.2 has a number of interesting observations. One of them is that some genres thrive on topic continuity. Not surprisingly, this includes recipes: clause-initial anaphors very dominantly pick up discourse-old referents, and even when they do pick up a new referent it is usually with a personal pronoun, but then there are usually no competing referents in this genre. The extract of a law text examined understandably shies away from ambiguity and repeats lexical items. Ælfric's Saints’ Lives and Homilies are very similar in their anaphoric strategies, and quite different from Wulfstan's homilies. Wulfstan's rhetorical homiletic (rather than narrative) style, addressed orally to a lay congregation, is overexplicit in its use of demonstratives, which even pick up discourse-old referents more than in any of the other texts. Cole attributes this to the spoken and admonitary character of Wulfstan's delivery. These are interesting observations, but the attempt to compare the avoidance of ambiguity of the medical recipes (topic continuation, dominant use of personal pronouns) to Wulfstan's need for unambiguous instructive moral clarity by means of a heightened use of demonstratives makes it clear that we are not yet looking at a crucial interaction between information-structural factors and matters pertaining to genre, text type and style.

My third point returns to the psycholinguistic motivations for differentiated use of personal pronouns and se-forms. Psycholinguistic work has so far concentrated on contexts with competing referents, and even there, Bosch et al.’s (Reference Bosch, Katz, Umbach, Schwarz-Friesel, Consten and Knees2007) account does not exhaust the factors relevant to pronoun choice. Kehler et al. (Reference Kehler, Rohde and Elman2008) show that the choice of topic continuity vs topic shift correlates strongly with specific discourse relations. Let's discuss this with some examples:

1. (1)

   • (a) John borrowed a bike from Mike. He ____

     (b) John passed the comic to Mike. He _____

     (c) John hugged Mike. He _____

     (d) John helped Mike. He _____

A speaker using topic continuity (reference to John in (1a)) is most likely explaining why John borrowed the bike, for instance because he was trying to get somewhere on time. A speaker using topic shift (reference to Mike in (1b)) is more likely to explain the result: what Mike was going to do with the comic. When the verb is an interpersonal one, there is generally a strong preference for the cause explanation, as observed by Majid et al. (Reference Majid, Sanford and Pickering2007). In other words, contexts like (1c) and (1d) elicit the cause explanation (why John hugs/helps Mike) more often than the result explanation (what Mike's response is).

Hypothesizing that the results of Kehler et al. might yield predictions for the use of clause-initial personal vs demonstrative pronouns in Dutch and German, and building on Kaiser (Reference Kaiser2011), Dimitrova & van Kemenade (Reference Dimitrova and van Kemenade2017) test this for present-day Dutch and German, specifically with regard to the choice of pronoun. The initial results for Dutch suggest a correlation between topic continuity, the use of a personal pronoun and cause continuation on the one hand, versus topic shift, the use of a demonstrative and result continuation on the other hand. These results need to be fleshed out further, but they suggest that the use of pronouns taps into the structuring of discourse in hitherto unexplored ways. It will be interesting to see whether this may shed further light on the details of pronominal usage in Old English, in relation to discourse structure.