This paper presents new empirical evidence for a process description of quantity in Swedish. Like most other Germanic languages, Swedish possesses a rich array of stress-governing derivational suffixes which cause stress and length alternations in the stems to which they are attached. Such alternations bear crucially on the choice between a unit-oriented or process-oriented approach to Swedish phonology. Inasmuch as 18-vowel-phoneme solutions presuppose lexically inherent vocalic length, they result in a multitude of morphophonemic alternations between long and short vowels. In process solutions, such quantitative alternations follow predictably from independently motivated rules. These results have important implications for the description of the phonemic system, phonotactics, rule component, and morphological structure of the language. The conclusions are directly valid also for Norwegian.

This paper presents a syntactic account of the distribution of subject clitics in the Oslo dialect of Norwegian. In addition to providing a unified explanation for some seemingly unrelated facts about subject clitics, this account will be shown to have implications for the proper definition of the notions ‘variable’ and ‘anaphor’.

In Norwegian, these two types of constructions expressing an NP with a (prenominal) possessor exist:

(i) Pers bil ‘Peter's car’

(ii) Per sin bil Peter his, refl. car ‘Peter's car’

I argue that the s in (i) is not a case affix, but rather a variant of sin. Furthermore, I discuss how [Per] in (ii) assigned Case, and the fact that Per and sin apparently share a theta-role. I suggest that they constitute a chain in (ii), to which Case and a theta-role is assigned.

I shall compare the definition of “chain” in Chomsky 1981 to the modified definition of Kayne 1981. The modified definition, together with other principles suggested by Kayne 1981/1983, not only permits our analysis of (ii), but also provides interesting predictions about this construction, and indirectly about the lack of it in a related language, Icelandic.

This paper presents a unified account of the distribution of Danish der within the framework of Government and Binding, analyzing der as a subject filler. The existence of empty subjects in cases of extraction out of embedded questions is explained by assuming proper government by AGR, not due to verbal inflection as in pro-drop languages, but rather due to the absorption by AGR of the features of an empty co-superscripted postverbal position.

In Norwegian, the verb få ‘get, receive’ differs from all other verbs with respect to its use in imperative sentences and with respect to control properties in infinitival clauses. In the imperative, the deleted subject may be first person, and in clauses embedded after ‘ask’, få has unambiguous subject control. These syntactic facts are related to the lexical semantic properties of the verb få, and it is shown that this syntactic behavior is not due to some idiosyncratic peculiarity of this particular verb. The semantic properties of this and other Norwegian verbs account for the uniqueness of få in this respect.

This paper reviews various approaches to describe the verb second phenomenon of Germanic languages within generative transformational grammar. The solution to the descriptive problem seems to be to assume that the finite verb in main clauses has the same position as the complementizer in subordinate clauses. Various ways to explain the presence of this word order in Germanic languages are presented in the final part of the paper.

Certain word order facts in Icelandic have been claimed, by several linguists, to provide evidence that Icelandic has no VP constituent (in s-structure) in sentences where the main verb is finite. If correct, this entails that Icelandic phrase structure is not always binary branching, contradicting the hypothesis that all phrase structure is binary branching. It is shown to be incorrect. A new theory of Icelandic sentence structure is presented, based on Platzack (in press), but strictly observing binary branching. The theory presented is shown to be empirically more adequate than the less constrainedĀ-nary branching theories.

In this article I argue that Taraldsen's (1983) analysis of verb particle constructions cannot be maintained because it predicts a considerable subset of the grammatical particle constructions to be ungrammatical, as well as some ungrammatical particle constructions to be grammatical. In my analysis of verb particle constructions as causative constructions, I borrow the device of reanalysis from Taraldsen. This reanalysis is crucially taken as a licensing condition for the rule of particle movement, but it does not constitute a sufficient condition for the possibility of particle movement, since I will also argue that this rule is subject to essentially non-syntactic restrictions so that its output cannot be wholly deduced from GB syntax subtheories.

Various linguists working within the theory of Government and Binding (e.g. Hornstein & Weinberg (1981), Kayne (1981)) have attempted to provide a unified account of preposition-stranding. This article uses evidence from Icelandic to show that preposition-stranding is not a unified phenomenon. Although Icelandic freely allows preposition-stranding in wh-movement constructions, it lacks prepositional passives in which the prepositional object of an active verb corresponds to the grammatical subject of a passive verb. Various syntactic tests which distinguish between grammatical subjects and topicalized NPs are used to demonstrate this. Our conclusion is that while lexical reanalysis is needed to account for prepositional passives, no such reanalysis is warranted for preposition-stranding due to wh-movement; hence, an adequate theory of preposition-stranding must allow for two separate parameters.

Two models are presented of how the listener detects words in utterances. The first model assumes that the listener takes advantage of phonetic word boundary signals (WBSs), non-phonetic information not being necessary for word detection. The second model assumes that word detection relies on the use of non-phonetic knowledge of the language, words being detected through the recognition of the preceding word. Thus WBSs may not be necessary for word detection. The WBSs suggested for Finnish are evaluated against this background. The phonotactic WBSs are found unreliable or useless, the others limited in applicability or experimentally unsupported. The models and the results of the survey will direct future investigations of word detection and recognition in Finnish.