The thesis consists of four interconnected studies of various topics in the fields of Latvian dialectology and historical linguistics: (1) apocope and shortening; (2) loss of the present 2^nd singular endings *–i and *–ɨ; (3) the development of the participle desinences *–ājis *–ējis; (4) the ā– and ē–stem locative singular endings. A recurrent theme is a potential loss of *j in the phonological sequences *–āji(s) *–ēji(s), which were subsequently contracted to *–āj(s) *–ēj(s). In this regard, the first study provides the necessary East Baltic context, while studies 2, 3 and 4 investigate the relevant evidence in the Latvian dialects. At the same time, however, each of the topics is also studied on its own terms.

In the first study, an improved account of apocope and shortening is formulated. It is argued that the primary apocope affected all unaccented short vowels, including *u, but that it took place before the accent retraction. Endings that were only affected in immobile paradigms were restored on the model of their counterparts in mobile paradigms. The secondary apocope in second posttonic syllables, which traditionally accounts for the occasional loss of inherited long vowels and diphthongs, is dated after accent retraction but restricted to *i. This implies that a series of endings such as d.sg. –am, d.pl. –Vms, ill.sg. –Vn, 1.pl. –Vm a.o. contained short vowels when the primary apocope set in. Unless Leskien’s Law operated in Latvian, the evidence suggests that the final vowels in these endings were short. If Lithuanian excludes Leskien’s Law in a given case, then the Latvian evidence is compelling, e.g., PEB d.sg.m. *–amu or *–ami rather than *–amọ̄(i̯).

The second study investigates present 2^nd singular forms with and without a distinct ending in the Latvian dialects. The most archaic distribution seems to be the one found in Blīdene₁₀₆, where only the derived presents are endingless in the indicative. Studies 2, 3 and 4 also give a comprehensive overview of the development of word-final *–āj(s) *–ēj(s) and *–āji(s) *–ēji(s). Both *–āj *–ēj and *–āji *–ēji yielded –ā –ē in Low Latvian, while High Latvian might have a contrast between accented –āi̯ –ēi̯ and unaccented –ā –ē. High Latvian reflects *–ājs *–ējs and *–ājis *–ējis as –ājs –ējs, but the Low Latvian developments are unclear. Accented *–ājs *–ējs yield –āš –ēš but in unaccented position there may be up to three phonologically regular reflexes: –āš –ēš, –aiš –eiš and –ais –eis. In Vidzeme and Zemgale, *–ājis *–ējis must have merged with *–ājs *–ējs, but in Courland *–ājis *–ējis were possibly apocopated only after *–ājs *–ējs had become –āš –ēš.

Accordingly, only Low Latvian –â –ê can be reflexes of the old inessive. Low Latvian –ai –ei cannot be cognate with –â –ê but might continue the old adessive. If Low Latvian –ã² –ẽ² reflect the illative, then the three pairs, along with dialectal adverbs in –uop, represent all four East Baltic local cases, which therefore must have existed at least as syntactic constructions in Proto–East Baltic. The enigmatic High Latvian ā–stem locative singular ending –â is identified with Low Latvian –ai rather than –â, which explains its remarkable lack of labialisation but implies a development *–ˌaî > –ˌâ.

The utility of the proposed phonetic loss of *j is limited to disyllabic forms like l.sg. *tâji ‘that’ prs.2.sg. *smeji ‘laugh’ ptc. *gãjis ‘go,’ all of which might also be analogical, and it seems best to reject it.