The extent to which morphology determines syntactic structure has been in focus for much syntactic research over the past two decades. Icelandic has often played a central role in the argumentation because of its rich inflectional system, which distinguishes it from the mainland Scandinavian languages. In this dissertation, Ásgrímur Angantýsson investigates how certain morpho-syntactic properties of Icelandic may affect the word order in embedded clauses and what consequences this has for syntactic theory. He identifies four relevant constructions where modern Icelandic differs from the mainland Scandinavian languages and also to some extent from modern Faroese. The relevant constructions involve the order of sentential adverbs and the finite verb in various kinds of subordinate clauses, Topicalisation in embedded clauses, Stylistic Fronting, with and without Expletive Insertion, and the so- called Transitive Expletive Construction. Each of these constructions has received attention from linguists in recent years, but so far no one has looked in detail at possible correlations between acceptability judgements for all of these constructions. Angantýsson is able to provide this thanks to judgement data from approximately 1600 Icelandic speakers of various ages. He has also compared these with judgements from speakers in the Faroe islands, in West Jutland and in Älvdalen in north Dalecarlia, where he has carried out fieldwork within the ScanDiaSyn project.Footnote ¹ In some cases he has supplemented the questionnaire data with corpus studies. The result is a comprehensive overview of the actual use of these constructions in Modern Icelandic.

In the background chapter, Angantýsson gives an overview of previous research on clause structure, in particular on so-called root phenomena. He also discusses advantages and disadvantages of various methods for gathering data and eliciting judgements on relevant linguistic examples. Using a written questionnaire is often the preferred method when it comes to gathering a large amount of data in relatively short time, but the method is also unreliable since it is difficult to check that the subjects have understood the task, that they are attending to it and that they actually interpret the example sentences in the way intended by the experimenters. These drawbacks can be countered to some extent by the so-called ‘oral elicitation’ method (see Cornips & Poletto Reference Cornips and Poletto2005) whereby the researcher asks the speakers, either individually or in a group, to judge the sentences and discuss acceptability as well as when such sentences would be used. In addition the researcher can control for intonational effects such as contrastive stress and destressing. Angantýsson reports on an interesting triangulation of methods (written questionnaires, filling in blanks and responses to audio-recorded stimuli) which shows that the written questionnaires give reliable results on the whole, but that there are constructions where the availability of a recorded stimulus made a difference (pp. 50f.).

The Icelandic data used in the dissertation was collected mainly through three large overview questionnaires in the Icelandic Dialect Syntax project (2005–2007) (Thráinsson forthcoming) involving some 1600 speakers of various ages. The data on Faroese, West Jutlandic and Övdalian was collected by the author during field work sessions organised within the ScanDiaSyn project.

In the theoretical overview, Angantýsson discusses the two main approaches to the constructions under investigation, the morpho-syntactic approach, according to which differences in the morphology lead to different structures (see Holmberg & Platzack Reference Holmberg and Platzack1995), and the so-called cartographic approach, according to which semantic or discourse-related categories play a role in the syntax. He summarises this in terms of hypotheses concerning the number of functional projections and the setting of IP-parameters (pp. 52 ff.). Angantýsson follows Bobaljik & Thráinsson (Reference Bobaljik and Thráinsson1998) in assuming that if a language has rich agreement, then it will have V-to-I-movement. Icelandic has the required tense/person agreement and V-to-I-movement is thus predicted to apply in all embedded clauses. Faroese and Övdalian have somewhat poorer agreement than Icelandic, but more agreement than the mainland Scandinavian languages, and hence provide interesting material for comparisons.

Chapter 3 is entitled ‘Verb/adverb placement in subject initial embedded clauses’. Angantýsson investigates the acceptability of both Vfin–Adv order and Adv–Vfin order in að-clauses (see (1)), indirect questions, relative clauses and adverbial clauses.

1. (1)

If the finite verb always moves into the I-domain, it should precede adverbials that are adjoined to VP. The data analysis shows that Vfin–Adv is strongly preferred in all types of subordinate clauses in Icelandic, whereas the Adv–Vfin order is preferred in Faroese and West Jutlandic; Övdalian seems to have a mixed system (see the comprehensive overview in Table 27, p. 102). Among the Icelandic informants, the youngest speakers are more willing to accept the Adv–Vfin order in að-clauses than the oldest speakers, whereas the opposite pattern is found in relative clauses. In this clause type, Angantýsson shows that the type of subject also matters. If the subject of the relative clause is a pronoun, more people accept the Adv–Vfin order, illustrated in (2a), than if it is a lexical NP, as in (2b).

1. (2)

Examples like (2c), where the subject position is empty, were not included in the study, for the stated reason that they would be instances of Stylistic Fronting (p. 69, n. 47) (see below). To me this seems like an unfortunate oversight, since it would be interesting to know how such sentences are judged by Icelandic speakers, regardless whether or not they are analysed as Stylistic Fronting by the linguist. Angantýsson cites one spontaneous example of this type from the ÍS-TAL corpus of spoken Icelandic, see (3), which also shows the increasing tendency among Icelanders to insert að ‘that’ after the complementiser in relative clauses, indirect questions and adverbial clauses (pp. 127ff.).

1. (3)

Although the Vfin–Adv order is clearly preferred in all types of subordinate clauses, the Adv–Vfin order is judged as OK by 20–30% of the informants in some clause types. How can this be if the verb always moves out of the VP in Icelandic? Angantýsson follows Bobaljik & Thráinsson (Reference Bobaljik and Thráinsson1998) here and assumes that the verb has indeed moved into the IP domain, but that the adverb has adjoined higher up in the IP domain, to TP. This proposal is supported by the fact that such higher adjoined adverbs often are prosodically prominent, as in the example in (4).

1. (4)

Here the negation receives contrastive stress, highlighting the fact that what he did not say was more important than what he did say. If the higher Adv position is marked, one might expect that it would not be used when the adverb is unaccented, as is presumably the case when the negation associates with a narrow focus further on in the clause, as in (5).

1. (5)

Chapter 4 deals with Embedded Topicalisation (ET) as in the example in (6)

1. (6)

Angantýsson finds that there is a lot of variation in the judgements and that the younger speakers in general are more reluctant to accept such examples than the older speakers. ET is judged more acceptable in complements of factive verbs like segja ‘say’, halda ‘think, believe’ and uppgötva ‘discover’, which fits well with Hooper & Thompson's (Reference Hooper and Thompson1973) classification. He supplements the findings from the questionnaires with an overview of ET examples found in written and spoken corpora. What is noticeable in the spontaneously produced examples is that the topicalised phrases are often very short and consist of a personal or demonstrative pronoun such as það ‘it’ or þetta ‘this’, whose reference is provided in the preceding discourse. Context sentences were sometimes provided in the questionnaires, but all the ET examples involved fronted lexical phrases. One way of testing more naturally sounding ET examples in the future might be to insert the test sentences in short conversations.

Chapter 5 is devoted to Stylistic Fronting (SF), which involves the fronting of a past participle, an infinitive or an adverb to a preverbal position in clauses without subjects, first discussed in Maling (Reference Maling1980). Angantýsson surveys the extensive literature on the topic, including the question whether SF only applies to heads, as exemplified in (7a), or also covers XP-fronting as in (7b).

1. (7)

Angantýsson establishes that although younger speakers are more restrictive vis à vis SF than older speakers, both age groups prefer SF over a subject gap. In some cases younger speakers prefer expletive insertion over SF, which is not the case with older speakers. A search for SF in the corpora provided a fairly large number of instances and revealed that the type of fronted item varies with clause type. Past participles are most often fronted in relative clauses, whereas fronted adverbs are most common in all other clause types. The corpus investigation also revealed that for certain expressions, SF is not just a stylistic option but actually the unmarked, or only, option (pp. 158 ff.). The question arises whether these obligatory SF phrases might act as triggers for the learner. Given that younger speakers were fairly negative towards SF, it would be interesting to look at the frequency of such expressions in school text books and literature for children.

The Faroese speakers generally seem to prefer expletive insertion to SF, with some variation between the clause types. The West Jutlandic and Övdalian speakers largely rejected all cases of SF, leading Angantýsson to conclude ‘that the possibility of SF does not seem to be available at all in the Mainland Scandinavian languages’ (p. 183). I do not think this is entirely correct. In Swedish, just as in Icelandic, there are a number of more or less frozen SF expressions, as in (8a). Furthermore anaphoric temporal and locative adverbs are often fronted, as in (8b). They may be contrastively stressed, but not necessarily so.Footnote ²

1. (8)

In addition PPs are sometimes fronted in order to prevent an unintended attachment as in (9), from a column in the newspaper Dagens Nyheter, 3 March, 2011:

1. (9)

Fronting ‘in Israel’ makes it clear that the writer is talking about ‘the silence that prevails in Israel’. If ‘in Israel’ had appeared in the usual place for locative adjuncts at the end of the VP, then it would most naturally have been interpreted as modifying ‘Palestinian areas’. In view of the existence of examples like (8) and (9), I believe that it would be interesting to conduct comparable corpus searches in Swedish to establish the extent to which the position before the finite verb in subjectless clauses is actually occupied.

In Chapter 6, Angantýsson investigates the correlation in acceptability judgements for the investigated constructions, a type of investigation that has not been carried out before in any Scandinavian language on such a large scale. He finds that the type of embedded clause has a major impact on the speakers’ judgements. For instance, ET receives the highest acceptance rates in að-complements of bridge verbs, but this is precisely the context where the Adv–Vfin order receives the lowest score. He also finds that the speakers who were more willing to accept the Adv–Vfin order were also more willing to accept ET and SF. This type of correlation cannot be explained on a strict morpho-syntactic account and Angantýsson concludes that what is needed is a more fine-grained structure which reflects semantic/pragmatic differences among the types of embedded clauses. In Chapter 7, he proposes some explicit structures for the CP domain in the clause types under investigation, discussing also the variation found in regional Northern Norwegian (see Julien Reference Julien2007, Bentzen Reference Bentzen2009 and Wiklund, Bentzen, Hrafnbjargarson & Hróarsdóttir Reference Wiklund, Bentzen, Hrafnbjargarson and Hróarsdóttir2009). He makes use of semantic features such as +/– Assertive and +/– Factive. The chapter is rather short and exactly how the semantic features determine the internal structure of the CP and project the relevant information to SubP, the maximal projection of the embedded clause, does not become entirely clear. More research is obviously needed here.

All in all, this dissertation represents a successful combination of theoretical exploration and thorough empirical work. It makes a valuable contribution to our knowledge of Scandinavian clause structure, in particular through the analysis of the extensive data collection reported here and presented in very useful overview tables. Angantýsson establishes some highly interesting correlations between acceptability judgements on the order between adverbs and finite verbs, embedded topicalisation, stylistic fronting and expletive insertion in various types of embedded clauses. As he points out, the emerging patterns have clear theoretical consequences for the way different embedded clause types should be distinguished. In addition differences in the acceptability patterns between the Icelandic age groups will most likely contribute to the understanding of ongoing language change.