The article deals with Stephen R. Anderson's Icelandic Vowel Shift Rule as an example of the general concept of exchange rule and involving marked vs. unmarked rule orde˚. The evidence adduced from Modern as well as Old Icelandic is reexamined in detail and shown to be in fact nonexistent. The paper ends with a few concluding remarks on the general notions involved.

An interesting restriction on Norwegian identity-of-sense anaphoric reference when the anaphor occurs as second object is discussed. Fretheim (1974) proposed to account for this by a global rule constraining pronominalization in such contexts, and claimed that the necessity of this constraint renders any interpretive account of identity-of-sense anaphoric reference untenable. The present paper purports to demonstrate the contrary, that Fretheim's observations in fact provide support for an interpretive theory of identify-of-sense anaphora as far as Norwegian is concerned.

The purpose of this paper is to present an analysis of french schea deletion where, in agreemant with some basic principles of natural Generative Phonology, all rules are obligatory and not extrinsically ordered. This analysis is at variance with Dell's (1973) treatmant. It is contended that even if, as suggested by Rudes (1976), marking each phonological rule for applying only within a certain tempo can eliminate from the phonological component true optional rules, a certain amount of optionality must still be allowed within a rule. Furthermore, the nature of French schwa deletion forces us to reject rudes' model where the output of a tempo is the input to the next faster one.

This article is a review of Thorstein Fretheim (ed.), Sentrale problemer i norsk syntaks, Universitetsforlaget 1997. The review groups together the papers by Even Hovdhaugen, Jan Engh and Thorstein Fretheim. It argues that these three papers largely fail to support the authors' claims because post-Aspects developments in transformational, generrative grammer are ignored. Also, specific analyses based on TG are proposed as alternatives to some analyses based on analogy. Finally, lars Hellan's contribution is examined in a rather detailed manner, and some counterproposals are sketched.

122 Norwegian speech errors occurring in spontaneous speech have been analysed in the light of models of speech performance. Methods and problems in connection with data collection are discussed.

Systematic contrastive analysis (CA) and error analysis (EA) of Swedish are just a little more than a decade old, but have rapidly grown into a flourishing field of study. However, the contributions to this area of research are scattered over numerous and often rather diverse publications, and a compilation of the relevant literature should therefore be useful to researchers in the field as well as textbook writers and teachers who seek to apply the results of Swedish CA and EA in instructional materials and class-rooms.