This article presents a quantitative study of the referential status of PPs in clause-initial position in the history of English. Earlier work (Los 2009; Dreschler 2015) proposed that main-clause-initial PPs in Old English primarily function as ‘local anchors’, linking a new clause to the immediately preceding discourse. As this function was an integral part of the verb-second (V2) constraint, the decline of local anchors was attributed to the loss of V2 in the fifteenth century, so that only the contrasting and frame-setting functions of these PPs remain in PDE. This article tests these hypotheses in the syntactically parsed corpora of OE, ME, EModE and LModE texts, using the Pentaset-categories (New, Inert, Assumed, Inferred or Identity; Komen 2011), based on Prince's categories (Prince 1981). The finding is that Identity clause-initial PPs decline steeply from early ME onwards, which means the decline pre-dates the loss of V2. A likely trigger is the loss of the OE paradigm of demonstrative, which functioned as standalone demonstrative pronouns as well as demonstrative determiners, and the loss of gender marking more generally. From EModE onwards, main-clause-initial PPs that still link to the preceding discourse do so much more indirectly, by an Inferred link.

The spelling conventions for dental fricatives in Anglic languages (Scots and English) have a rich and complex history. However, the various – often competing – graphemic representations (<þ>, <ð>, <y> and <th>, among others) eventually settled on one digraph, <th>, for all contemporary varieties, irrespective of the phonemic distinction between /ð/ and /θ/. This single representation is odd among the languages’ fricatives, which tend to use contrasting graphemes (cf. <f> vs <v> and <s> vs <z>) to represent contrastive voicing, a sound pattern that emerged nearly a millennium ago. Close examinations of the scribal practices for English in the late medieval period, however, have shown that northern texts had begun to develop precisely this type of distinction for dental fricatives as well. Here /ð/ was predominantly represented by <y> and /θ/ by <th> (Jordan 1925; Benskin 1982). In the late fifteenth and early sixteenth centuries, this ‘Northern System’ collapsed, due to the northward spread of a London-based convention using exclusively <th> (Stenroos 2004). This article uses a rich body of corpus evidence for fifteenth-century Scots to show that, north of the North, the phonemic distinction was more clearly mirrored by spelling conventions than in any contemporary variety of English. Indeed, our data for Older Scots local documents (1375–1500) show a pattern where <y> progressively spreads into voiced contexts, while <th> recedes into voiceless ones. This system is traced back to the Old English positional preferences for <þ> and <ð> via subsequent changes in phonology, graphemic repertoire and letter shapes. An independent medieval Scots spelling norm is seen to emerge as part of a developing, proto-standard orthographic system, only to be cut short in the sixteenth century by top-down anglicisation processes.