This paper focusses upon the semantics of cojunctions and their referential properties in texts.

It is show that the function of conjunctions can be descrined in terms of deictic theory of didcourse reference. The main argument is that there is in principle neituer a functional nor an original historical difference berween pronominal/deictic expressions on the one hand, and conjunctions on the other.

Some general aspects of the historical evolution of conjunctional expressions in Germanic languages are presented which are necessary in order to reach an explanatory level in linguistic description. Thus an analysis of the processes of univerbation, deletion, morphological differentiation, and shifts of morphological classes is given, based on data from the different branches of Germanic.

Recent papers by Cathey and Demers (1975, 1976, 1977, ms) have incorporated an analysis of Old Icelandic phonology in which purely phonetically conditioned rules of i-Umlaut and u-Umlaut are crucial not only to the analysis itself, but to the theoretical implications theauthors draw from that analysis. Synchronically, however, the non exceptional application of i-Umlaut is restricted to just four morpho-syntactic classes: athematic and j-thematic verbs in the preterite subjunctive, j-thematic verbs in the non-preterite, adjectives with a consonant-initial comparative suffix, and athematic feminine or plural r-stem nouns. Historical u-Umlaut continues to round a to Q before u in the next syllable even in literary Old Icelandic, but its application is also morphologically induced, viz. a → Q in all feminines, plurals, and u-stems not followed by a vocalic suffix. All intraparadigmatic umlaut alternations are covered by these morphologizations, which in turn obviate the abstractions proposed by Cathey and Demers.

In some dialects of Old Norwegian unstressed non-low vowels have a distribution which has traditionally been defined as vowel harmony. However, such vowels after low stressed vowels have a distributional pattern which cannot be fully accounted for in terms of vowel harmony. On the basis of data from one vowel harmony dialect of Old Norwegian (Trøndelag) the present note suggests that the distribution of unstressed non-low vowels should be regarded as the result of four separate processes rather than one, including both vowel harmony and vowel reduction.

This article is an attempt at combining sentence level and text level by similar means of analysis, based on a DEEP CASE MODEL. Tentative lists of DEEP CASES and TEXTUAL RELATIONS are given and discussed.

The theoretical apparatus has been developed through empirical analyses of important notion of LINK FAMILY is introduced. The ultimate goal of this type of analysis is to find the INFORMATION STRUCTURE of texts, but its connections wiht e.g. “pure” TEXTUAL COHESION are also treated.