The thesis investigates the development of the group-stress system in French from the earliest textual records to 1500. Empirical work is based on a corpus compiled especially for the study, which comprises 87 extracts from medieval French texts totalling over 250,000 words, composed mainly of verse texts to make use of the extra phonological information provided by the form. A unique metrical and syntactic annotation is used in the corpus to permit studies of phonological phrasing and stress placement in lines of verse. Much octosyllabic narrative verse, in particular texts associated with oral performance, is shown to have an iambic rhythmic tendency in the pre-1250 period, which is particularly strong in the earliest texts. No such effect is found in lyric texts or plays, or in narrative from after 1250. Additionally, a phonological phrase boundary is commonly found in the middle of the line. Iambic rhythmic organization is argued to be incompatible with group stress and associated ‘stress deafness’ effects observed in modern French. From this data, group stress is argued to have developed between the mid-12th and mid- 13th centuries. Work on modern French (e.g. Post, 2000) suggests that the stress group is the phonological phrase. Through reconstruction of the phonological phrasing of medieval French, the thesis demonstrates that regular word-final stress, the phonological phrase internal process of stress clash resolution, and the frequency of monosyllabic words combine to favour reanalysis of the French stress system in the pre-1250 period. Finally, the hypothesis that prosodic change affected verb-second word order in medieval French is reconsidered. It is argued that light clause-initial constituents which do not form their own phonological phrase (i.e. short adverbs, subject pronouns) become unstressed, a development which triggers syntactic changes that lead to the introduction of non verb-second word orders.