Introduction to the special issue on variation in auxiliary selection

Abstract

The paper introduces the research field of the special issue: the use of the two types of auxiliaries be and have for expressing past events, and contextualizes research in this field in general.

Keywords:

• Auxiliary selection
• the Auxiliary Selection Hierarchy
• the Unaccusative Hypothesis

Background

In June 2016 a small symposium took place at Bernstorff Slot near Copenhagen, Denmark, organized by Anu Laanemets, who was then Marie Curie scholar at the University of Copenhagen.¹ The aim of the symposium was to bring together scholars with an interest in the perfect auxiliary systems in the languages of Europe in order to discuss recent developments within the field. The participants approached the topic from various theoretical, methodological and empirical perspectives which led to a number of interesting discussions. The papers in this special issue are all revised versions of presentations made during the symposium. Each paper has been reviewed by two or three anonymous reviewers, in accordance with the journal’s policy. We are grateful to the reviewers for their contributions and thank them cordially. Although some of the participants in the symposium chose not to submit papers for this issue, their presentations have provided inspiration for the articles in this issue.

The development of AUX as a problem and as a construction

Variation in the selection of auxiliaries has grown to be a topic of considerable interest for linguists of different persuasions in recent years. As McFadden (2007) in his concise review of research until then points out, the reason for the proliferation of research is that the early intuition of David Perlmutter on unaccusative verbs and unergative verbs (Perlmutter 1978) has led to a large number of descriptions in various languages, based on different types of data and various theoretical approaches. Perlmutter’s intuition concerned the status of subjects of various kinds of verbs: in English there seems to be a difference in the status of the subject of a transitive verb and that of different kinds of intransitive verbs. If we focus on the intransitive verbs, sentences such as (1) and (2) below typically assigns two different semantic interpretations to the entity which is the syntactic subject:

Table

(1)	John has worked	UNERGATIVE
(2)	John has arrived	UNACCUSATIVE

Subjects of unergative verbs behave like subjects for transitive verbs while subjects of unaccusative verbs behave more like objects. Perlmutter noted that this split fits the choice of the auxiliary in language which have a choice such that unergative verbs (and transitives) take have, while unaccusative verbs take be.

It is worth noting, as McFadden does, that this intuition, valuable as it has been, has not, until now, led to one theoretical explanation which covers all of the facts (McFadden 2007, 703). It is true that the main lines of research have benefited from Perlmutter’s original insights as well as from Antonella Sorace’s seminal paper on the Auxiliary Selection Hierarchy (ASH, cf. below) (Sorace 2000), but a number of recent refinements of the facts resist generalization (in addition to the papers below, see also the papers in Kailuweit and Rosemeyer (2015)).

Given that a lot of work has been done already, we may ask, as is often done in the Conversation Analysis tradition: Why this now? Why another collection of papers? We think that the distinctive characteristic of this collection is that it addresses variation in Auxiliary selection from different points of view and is based on many different types of linguistic data. Most of the approaches taken in the literature to date are present in this issue but are here applied to new facts. In this way, the contributions move the field forward toward better generalizations and eventually – hopefully – explanations.

First of all, as for the delimitation of the problem itself, the jury is still out. We all agree that we are concerned with Auxiliary selection, more particularly with explaining variation in the be and have Auxiliary selection but then agreement stops. Is this a fact about the verb stems (Sorace 2000; McFadden 2017), the participial structures that the AUX combines with (McFadden 2015), mood (Alexiadou 2015; McFadden and Alexiadou 2006), voice (Heltoft 2017; Nielsen 2017), the specific tense we are dealing with – the perfect (Thráinsson 2017; Fløgstad 2017), the various interpretations of the AUX chosen (Heltoft 2017; Nielsen 2017; Beliën 2017), the structure of the specific language we are dealing with (Heycock and Petersen 2017; Thráinsson 2017), not to mention the person and number issues found in some Italian dialects, recently discussed by D’Alessandro and Roberts (2010) (cf. also Drinka 2013, 626ff)?

Obviously, one’s theoretical stance will also lead to different delimitations of the linguistic facts to be addressed: Are we talking about standard languages or dialects, written language or spoken? For written language, are we dealing only with edited “correct” language or also with the unedited language of the social media (Beliën 2017)? What about language acquisition and learners’ language? For spoken language, are we looking at variation from a sociolinguistic point of view, i.e., taking note of the utterers’ age, gender, social background and place of origin (Laanemets 2017) or are we looking at a specific speech community? We believe that all of the above points are potentially relevant and that researchers need at least to be aware of these factors. We note with pleasure that all of the authors in this special issue have been quite explicit as to their delimitation of the problem and their use of data.²

Secondly, the position of the problem as a field of study within linguistics is at stake: Is it a syntactic issue, a semantic issue, or is it placed at the interface between semantics and syntax? Or is there no interface but only lexicogrammar (as in the Systemic Functional Linguistics tradition) and hence semantic aspects of syntax as well as syntactic aspects of semantics. Another version of this is the so-called semantics-syntax interface. An obvious case in point is Sorace (2015) who argues that it is feasible to integrate the Perlmutter syntactic generalization about unaccusatives and unergatives (which is dichotomous), and the ASH which by definition is gradient.

If we focus on the classification of the verb stems, as the ASH suggests, the further question arises whether this classification is universal and valid for all equivalents in all languages (and if so, why?) or whether the lexical semantic structures vary so much within this specific field that they are in fact hard to compare. As a point of reference here, Thráinsson (2017) fruitfully explores the use of adverbials to disambiguate various constructions. If verb stems were enough, adverbials would be superfluous as corroborating evidence.

Thirdly, there is the ever persistent question of the evolution of the attested patterns: What is the historical origin of variation in Auxiliary selection?

The historical perspective

As the very first speaker at the symposium, Bridget Drinka identified three important points for the Auxiliary selection patterns in the history of the languages studied here (Drinka 2013, 2017). Firstly, Drinka notes that the reconstructed Indo-European language did not have a perfect tense (and also did not have a verb of possession corresponding to have). Secondly, Drinka uses the available evidence on the status of the Greek language in the classical world, a language to be emulated by the Latin culture, to argue that Greek played a major role in facilitating the development of Latin perfect tenses. Thirdly, Drinka argues that the reign of Charlemagne (742–814) was instrumental in laying the foundations for a division of the European languages into central and peripheral. The Central languages German and French share with Italian and Albanian and ultimately with Romanian, Spanish, Dutch, Norwegian, and Danish the possibility of having have/be Alternation. This alternation, then, is to various degrees governed by the ASH. Outside this central area, have has been generalized to the West while be has been generalized to the East (note however that Thráinsson (2017) argues that a special kind of be-perfect has in fact developed in Icelandic). Furthermore, close to the border of the contiguous Charlemagne Sprachbund (van der Auwera 1998, 824), Drinka finds impressive variation. It is here that we find the interesting split between have and be being dependent on person, referred to above.³

Drinka (2017) is of course not alone in reconstructing the histories of the central European languages in this respect but the detailed argument she unfolds must be agenda setting from now on for this type of research. Seen from the perspective of this special issue, a particularly interesting point is Drinka’s emphasis on the role of language contact, as being historically founded sociolinguistics and as manifesting an integration of status perspectives and structural analysis. The analysis is meticulously documented with carefully glossed and analyzed, philologically precisely placed, examples. Drinka’s approach represents an alternative to the various strands of grammaticalization theory which builds upon universal tendencies for specific processes to occur. As such the contrast represents the age old (perhaps perennial) distinction between contextual approaches (language contact, sociolinguistics in general) and autonomous approaches (grammaticalization theory).

Grammaticalization theory is explicitly present in several of the contributions (Thráinsson, McFadden, Fløgstad, Nielsen and Heltoft) but the version of grammaticalization theory varies somewhat. Lars Heltoft, and his close collaborator Peter Juul Nielsen in a more synchronic analysis, both perform analyses of the synchronic oppositions which the structures at hand partake in before addressing diachronic issues. Fløgstad addresses the issue of functional explanations for the development. No language contact is involved in the accounts. There is also a lesson to be learnt by comparing Thomas McFadden’s contribution and Drinka’s strategy. McFadden builds on his earlier work with Alexiadou (McFadden and Alexiadou 2006, 2010) but here reveals the turning point when Present-Day English came into being ‘auxiliarywise’ by using data from the recently published corpus on English from 1707 to 1913. It turns out that the turning point was reached late in the nineteenth century. Incidentally, McFadden and Drinka both make use of the type of analysis which reveals not only how many examples have a be and how many have a have auxiliary but also points out how many verbs there are altogether such that we get a glimpse of the relative frequency of the periphrastic perfect.

Sorace’s ASH

As already mentioned, Sorace’s ASH has played an important role in the development of the field. We here summarize the defining properties of the hierarchy. In her seminal article, Sorace (2000) defines the ASH based on the aspectual and thematic characteristics of verbs, such as telicity and agentivity. According to Sorace, telicity – understood as achievement of an end/goal – is the most important factor for determining the choice of the auxiliary. Hence, as predicted by the ASH, verbs denoting telic change of location consistently select be, while verbs of non-motional controlled processes consistently select have. The verb classes between the top and the bottom of the hierarchy, i.e., verbs which are underspecified with respect to the defining semantic factors, exhibit variation (cf. Table 1, from Sorace 2000, 863). Yet, the variation in AUX selection is not random, but structured and orderly, and is largely in accordance with the lexical semantics of the verb.

Table 1. Auxiliary selection hierarchy Sorace (2000, 863).

CHANGE OF LOCATION	selects be (least variation)
CHANGE OF STATE
CONTINUATION OF A PRE-EXISTING STATE
EXISTENCE OF STATE
UNCONTROLLED PROCESS
CONTROLLED PROCESS (MOTIONAL)
CONTROLLED PROCESS (NONMOTIONAL)	selects have (least variation)

The ASH has been a very fruitful hypothesis leading to a number of insights from various (mostly European) languages. Just to mention one of the many brilliant papers springing from this fertile root: Melitta Gillmann (2015) contrasts German and Dutch and shows how telicity plays different roles in the two languages, being more important for Dutch than for German.⁴ This is a discussion which Beliën’s contribution to this issue latches directly on to by questioning the relevance of telicity as the explanatory principle for Dutch auxiliary choice.

Sorace has herself summarized the developments in her contribution to the important collection on Unaccusativity (Sorace 2004). The original formulation was based on grammaticality judgment data and could be used as predictions for actual usage. Since then Sorace has mustered neurolinguistic facts, attrition data and acquisition data for her hierarchy (Sorace 2015) but has not herself worked on actual usage. Other researchers, however, have used the ASH in e.g., historical research based on attested language (see e.g., Larsson 2015).

Sorace’s ASH is addressed in almost all the contributions to this issue but probably most directly by Heycock and Petersen, Beliën, and Laanemets. The latter two articles are based on spontaneously occurring data, written and spoken respectively.

The contributions

In his article Developing a new perfect: The rise of the Icelandic vera búinn að-perfect, Höskuldur Thráinsson casts light on the still on-going grammaticalization process whereby the phrase vera búinn “be finished,” combined with an infinitival phrase, is interpreted as a perfect. Thráinsson reviews earlier studies of this phenomenon and identifies certain key aspects, viz. the types of subjects and predicates which are possible at various stages as well as which readings are available. An extensive corpus study in the parsed historical Icelandic corpus shows how the construction is generalized. In addition, Thráinsson reports results from an acceptability study among more than 700 Icelandic speakers, divided into four age groups, which gives a clear indication of how the construction is gaining ground among younger speakers. Other groups which seem to readily adopt the new perfect are recent immigrants who are speakers of Icelandic as a second language. They even simplify the morphology to an uninflected marker buna. The way Thráinsson combines theoretical analysis, corpus studies, and large-scale informant studies is exemplary.

The aim of Caroline Heycock and Hjalmar Petersen’s article The have/be alternation in contemporary Faroese is to provide new data bearing on the use of be perfects and to add Faroese to the number of North Germanic languages studied for this particular phenomenon. Using data from two online questionnaires, inspired by Sorace’s ASH and McFadden and Alexiadou’s work on the have/be alternation in Earlier English (McFadden and Alexiadou 2006), they establish that be in Faroese perfects is not restricted to stative resultatives, as in Icelandic and Swedish. Rather the use of be as a perfect auxiliary resembles the pattern of use found in Danish (see Laanemets 2017). Determining whether or not this is a recent development, due to the close contacts between Danish and Faroese, requires more research on earlier stages of Faroese. Given that Danish was also used as the administration language in Iceland, this calls for a comparative study of the different effects that contact with a specific language may have on the grammatical systems of closely related languages.

Thomas McFadden’s article On the disappearance of the be perfect in Late Modern English is a detailed study of when perfect formation with be stopped being productive in English. Thanks to the availability of parsed historical corpora, McFadden is able to pinpoint this to the later part of the nineteenth century, but, more importantly, he also investigates which types of verb phrases are used in be perfects. He finds that until the end of the eighteenth century, be could combine with stative-resultative participles formed by both transitive and intransitive verbs. After around 1800, only participles formed by passivized transitive verbs are used productively in be perfects. McFadden links this restriction to the syntactic structure of the participial phrases.

In his article The function of supine auxiliaries in Swedish and Danish: Morphology and syntax in argument assignment, Peter Juul Nielsen carries out a detailed comparison of the role of the auxiliary and the supine in perfect constructions in Danish and Swedish and uncovers systematic differences. In Danish, where the supine lacks voice inflection, auxiliary selection is crucial to determining argument assignment and the auxiliary cannot be omitted. In Swedish, where the supine expresses voice and thereby determines argument assignment, the auxiliary only contributes primary (deictic) tense (present vs. past) and mood information and is in some syntactic contexts redundant. Nielsen’s analysis brings out relevant aspects of the interplay between morphology and syntax and is, like Heltoft’s, carried out within the framework of Danish Functional Linguistics. There is an interesting link with Beliën’s paper in that both focus on the construal or interpretation of the various constructions.

Guro Fløgstad’s article Revisiting Perfect/Preterit instability across Romance: On functional motivations for diverging paths addresses a different type of linguistic variation, viz. when analytic perfect forms, consisting of an auxiliary plus a participle, take on the same meanings as expressed by simple, synthetic, verb forms such as passé simple in French or Präteritum in German. After investigating a sample of more than 40 Romance varieties from Europe and South America (all employing the auxiliary have in perfect constructions), Fløgstad finds that the perfect expansion mentioned above is more common in Europe than in South America, where, in contrast, expansion of the preterite is more common. Given that the expansion of usage goes in both directions (from synthetic to analytic forms and vice versa), this may serve as a good illustration of competing motivations in functionalist explanations for language change. According to Fløgstad, target-based explanations fare better than source based for the developments in both Europe and South America. In addition, we need to consider the different perspectives of speakers and hearers with respect to processing related explanations.

Maaike Beliën’s article Auxiliary choice with particle verbs of motion in Dutch is a case study of three Dutch particle verbs showing variation in auxiliary selection. Several studies on Dutch auxiliaries have focused on telicity as the most relevant factor for predicting which auxiliary is used. Verbs denoting telic changes of location or state are used with zijn “be” as the auxiliary and atelic verbs use hebben “have”, in accordance with Sorace’s ASH. Beliën instead proposes that the relevant distinction is whether the event is construed, in the cognitive-linguistic sense, as a change in the subject or as a type of act, performed by the subject. Her data come to a large extent from informal writing on the Internet and support the hypothesis that the way the event is construed is more relevant than the Aktionsart (telic or atelic) of the verb phrase. Beliën does not explicitly relate her category “type of act” to what Sorace (2000) defines as “controlled process” in the ASH, among other things because she prefers to be agnostic as to the exact role of agentivity, volition and control. We agree with Belien that here is a point which needs more attention in future research.

The synchronic variation of the Danish auxiliary system is at the center of Anu Laanemets’ article The choice of the perfect auxiliary in contemporary spoken Danish. In contrast to almost all the other contributions to the field, Laanemets does not view variation as a witness to some historical process of contact or regrammation or any other structuralist vestige. In the variationist approach, structured variation is ubiquitous (Labov 1994). This means that variation may be latched on to by different groups of speakers, different generations or speakers who differ in any number of dimensions. This is not something we may know beforehand but something to be discovered by studying their actual usage when they speak the language. In her analysis, Laanemets uses the structured corpus of spoken Danish collected at the LANCHART Centre to show, through quantitative apparent and real time analysis, that indeed there is a difference between how the generations recorded during the period from 1987 till 2010 employ the auxiliaries. The subsequent micro-level analysis shows that the variation in auxiliary selection largely conforms to expectations (based on ASH). However, deviations are registered and even emerge with a categorical change of location verb such as komme “come”.

The title of Lars Heltoft’s contribution, The regrammation of paradigms: Auxiliaries from Gothic to Modern Danish, reveals his structuralist approach, as mentioned above. Based on the basic premise that auxiliary systems cannot be understood only from a syntagmatic perspective, Heltoft instead proposes a paradigmatic one. Taking his point of departure in the description of the system of auxiliaries in Modern Danish in the three volume reference grammar of Danish (Hansen and Heltoft 2011, 236ff), Heltoft analyses the earlier stages in Gothic and Old Danish. Heltoft places the predicate adjective on a par with the main verb. Thus, both predicative constructions and passives are part of the development of auxiliary systems leading to the bipartition of be and have perfects as well. The result is a description of the internal evolution of the system of auxiliaries in Danish from the earliest records to the present day. Though historical in its aim, the paper differs from Drinka’s approach by viewing systems as developing from structures already in evidence by consecutive and interlocking processes of reinterpretation. Here, Heltoft follows the influential Danish researcher Henning Andersen.

Notes on contributors

Elisabet Engdahl is a professor emerita of Swedish at the Department of Swedish, University of Gothenburg. She has worked on syntactic variation in the Scandinavian languages.

Anu Laanemets holds a PhD in Scandinavian Linguistics from the University of Tartu, Estonia. After two years as Marie Curie Research Fellow at the LANCHART Centre, University of Copenhagen, Denmark, she now works as an associate professor at the University of Oslo, Norway. Her research interests include Scandinavian languages, syntax, language use in spoken and written discourse, mainly combining comparative and corpus linguistic approaches.

Frans Gregersen is a professor of Danish language at the University of Copenhagen. From 2005–2015, he was directing the DNRF LANCHART Centre at the same university. He has written recently on style and intra-individual variation in general.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This work was supported by the Danish National Research Foundation [grant number 63]; and European Union FP 7 [Marie Curie Actions, grant number 630059].

Acknowledgments

We are also grateful to the LANCHART Centre, University of Copenhagen, for support. In addition the editors of the special issue are grateful to the Acta Linguistica Hafniensia for the opportunity to publish the papers from the symposium as a special issue and in particular to editor Hartmut Haberland for his untiring efforts to produce the best.

Notes

¹ The symposium was organized within the framework of the Marie Curie project AUX-Variation, which has received funding from the People Programme (Marie Curie Actions) of the European Union seventh Framework Programme (FP7/2007–2013).

² One approach which is missing from this special issue is Optimality Theory. OT has been applied to the study of the Italian dialects mentioned below which distinguish between persons by choice of auxiliary, by Geraldine Legendre in Legendre (2007, 2010).

³ This may be due to the fact that Italian is a Pro-drop language (cf. D’Alessandro and Roberts 2010).

⁴ Gillmann (2016) provides a very important historical analysis of the development of have and be perfects in German. One interesting fact is that this discussion is very old: Gillmann refers to a central paper by Paul (1905).