Research questions
Since Pollock 1989, much work has been done on the differences between languages in how finite verbs are placed with respect to sentence-medial adverbs and markers of negation. Thus in English an adverb such as often can immediately precede the finite verb but cannot intervene between a transitive verb and its object, while in French the equivalent adverb souvent shows the opposite behaviour; this is typically accounted for by postulating verb movement to a position on the left of the adverb in French (V-to-I) and its absence in English (rather than a difference in the syntax of the adverbials):
1. Louise (often) drinks (*often) red wine.
2. Louise (*souvent) boit (souvent) du vin rouge.
Pollock argues that this syntactic difference correlates with a difference in the extent of agreement morphology on finite verbs; this hypothesis is at the basis of a much more general, but controversial, claim that recurs in the syntactic literature: that syntactic variation is due to variation in inflectional morphology. Important data bearing on this claim have come from variation between the modern Scandinavian languages. The principal languages discussed occupy the two morphological extremes: Icelandic has an extensive agreement paradigm (and exhibits V-to-I), while the (standard) Mainland Scandinavian languages have lost all agreement on finite verbs (and lack V-to-I). These two extremes however are consistent with a number of different theories of the relation between agreement morphology and V-to-I; thus attention has turned to less categorical examples. Among living languages, the most crucial case is that of Faroese, which exhibits an agreement paradigm less rich than Icelandic but more extensive than Mainland Scandinavian, and which appears to have been losing V-to-I.
The main question that this project seeks to address is whether or not V-to-I is still part of the grammar of contemporary Faroese speakers. There is a degree of consensus that some speakers do not allow V-to-I; however there is considerable debate whether there are other speakers who have it as one option. We seek to determine whether there are such speakers; whether they are rarer in younger generations than older generations; whether their distribution correlates with an important dialect difference; and whether any such variation is between two grammatical options that are both completely acquired (even if one is preferred) or whether it is the reflection of a written norm imposed on a vernacular grammar without V-to-I, only the latter of which is truly acquired.
Aims and objectives
Our overall aim is to establish the status of verb-movement in contemporary Faroese.
To do this we aim to obtain reliable quantitative data concerning the acceptability of sentences that are only consistent with one or the other grammar. These data will include not only cases where the surface order of the finite verb and sentence-medial adverbs are at issue, but other phenomena which have been argued to depend on the presence or absence of V-to-I, thus giving insight into whether variable speakers are really accessing two fully grammatical options.
We aim to test subjects from a range of age-groups, in order to determine whether the situation is currently stable, or whether the acceptability of V-to-I decreases with younger speakers (a change in "apparent time"). We will include data from children who are not yet literate, as part of the investigation into whether variability is attributable to exposure to more conservative written norms.
Finally, we aim to obtain data from two dialect areas: from Streymoy, where the capital Torshavn is located, and from Suduroy, which has been argued to be the syntactically most "conservative" area in retaining V-to-I (Jonas 1996).
Research context
As indicated above, the correlation between V-to-I and "rich" agreement morphology has been taken as evidence that properties of morphological paradigms are the cause of syntactic variation between languages (e.g. Rohrbacher 1999). This synchronic pattern is strengthened by the diachronic observation that as Danish and Swedish lost rich agreement paradigms, so they lost V-to-I (e.g. Platzack 1988, Vikner 1997). However, there are some non-standard dialects of Swedish and Norwegian that share the lack of morphology with the standard languages but allow V-to I (possibly variably); it has also been noted that the loss of V-to-I in Danish appears to lag behind the loss of morphology by several generations (Bobaljik 2003). This has led to the proposal that there is only a one way implication between richness of agreement morphology and V-to-I, and that morphology follows syntax, rather than the converse (Bobaljik 2003, Bobaljik and Thrainsson 1998, Thrainsson 2003). The empirical difficulty here is that data from the past is necessarily impoverished in significant ways; the dialect data is limited and may be affected by dialect mixing with the standards. It is for these reasons that Faroese occupies such an important role in deciding this question, since it is currently spoken, and a standard language in its own right.
Research into contemporary Faroese can thus put us in a better position to "use the present to explain the past." In particular, we can inquire into speaker's judgments of acceptability, correlate the relative acceptability of phenomena that our theories predict should be related, and make use of constellations whose frequency in written data is rare - all opportunities denied to researchers working on historical records. Insights into the morphology-syntax relation gained here will be relevant not only to our understanding of developments of Scandinavian but also to our understanding of the history of English, which underwent a similar process (complicated in this case by the emergence of do-support), again with a problematic time-lag (see e.g. Roberts 1993, Kroch, Taylor and Ringe 1995, Bobaljik 2003).
Recent work in Faroese has begun to document the state of the language, but so far the data are fragmentary. Jonas (1996) reports that there are two dialects of Faroese, one without V-to-I, and the other with variable V-to-I. However, she gives no details about her pool of subjects, or exactly how their judgments were elicited, nor does she provide any quantitative measures. Petersen (2000) reports on a questionnaire study conducted with 18 Faroese school students around 20 years of age, and concludes that speakers of this generation "ha[ve] generally lost V-to-I movement" and that to the extent that these speakers accept sentences only consistent with a grammar with V-to-I this is because of the effect of the written language, and "is thus a typical instance of diglossia." Petersen's results are however reinterpreted by Thrainsson (2003), who adds data from a further questionnaire survey of 14 school students, and concludes that the degree of acceptance of V-to-I sentences in both studies is much higher than would be expected from speakers whose grammars lack this option altogether. Thrainsson has also examined texts from a range of authors born between 1819 and 1950 that show a decline in the frequency of unambiguous V-to-I orders but do not, he argues, suffice to demonstrate that these orders are excluded for contemporary speakers. Unfortunately, though, the crucial data that he reports are very sparse; one problem for corpus work in this area is that the relevant data are only found in a subset of subordinate clauses, which further must contain a sentence-medial adverb or negation.
Research methods
We plan to gather data concerning judgments of acceptability from native Faroese speakers through a series of experiments using the methodology of Magnitude Estimation (Bard et al 1996, Sorace and Keller 2005), which has been found to be particularly suitable for investigating gradient phenomena. Since it yields data on an interval scale, it is possible to subtract the estimate given to an unacceptable sentence from the estimate given to its corresponding acceptable counterpart: the relative magnitude of the number obtained is a direct indication of the speaker's ability to discriminate between acceptable and unacceptable sentences, and therefore a correlate of the strength of preference/dispreference for the acceptable/unacceptable sentence. A pilot study using this method was carried out in the Faroe Islands by Heycock in June 2003. The experiment was run with 24 subjects, 14 from the Northern dialect area and 10 from the Southern. The hypothesis that speakers still consider the Verb - Negation order a grammatical option was supported: there was a significant difference between judgements on this order and those on sentences with errors of case or verb conjugation. However the results also indicated that this order is significantly dispreferred, and that at least to some extent speakers are more likely to treat such cases as instances of "embedded main clauses" rather than syntactically subordinate clauses with V-to-I. The further hypothesis that southern speakers would be more willing to accept the Verb - Negation order was not supported, although no strong conclusion can be drawn given the small number of subjects who could be tested in the time.
We will expand the experimental design of this pilot and test subjects from three age groups using this methodology: young adult (15-25), middle-aged (40-55) and older (65+). If V-to-I is really a grammatical option, we expect judgments on the relative order of finite verb and adverbials to pattern together with other phenomena that are associated with this option, in particular the transitive expletive construction and the possibility of verb movement in certain infinitivals (Jonas 1996); on the other hand, if the verb - adverb order has been reanalysed as arising from (possibly generalized) embedded V2 or is simply recognised as characteristic of written texts without corresponding to any grammatical output for the speaker, we do not expect to find such a correlation.
In addition, we will test pre-school children (aged 4-5) using elicited production tasks (Thornton 1996) and child-specific acceptability judgment tasks (McDaniel and Cairns 1996). For a subset of children we will collect samples of parental language input in their everyday life. If the verb - adverb order is accepted by speakers only as a result of exposure to the written language and is not a grammatical output for them (and hence will not be in their output and in the data available to children), we do not expect preliterate children of this age to accept this order at all.
One significant advantage of our choice of research methodology is that it will give quantitative results that could be directly compared to results for the related Scandinavian languages. We will also be able to establish whether the clustering of judgments claimed in Jonas 1996 but not actually demonstrated does hold true of contemporary speakers.
