1 The Rise and Fall of Anticausative Constructions in Indo-European: The context of Latin and Germanic* Michela Cennamo, Thórhallur Eythórsson, Jóhanna Barðdal University of Naples, University of Iceland, University of Bergen Abstract The historical development of valence changes of verbs is generally an understudied phenomenon, and changes in lability of argument structure of verbs over time is even less studied. The aim of the present article is to contribute to a commencing discussion of the evolutionary paths of anticausativization in general and P-lability in particular. This is motivated by the fact that various genetically related languages, like the ancient and early Indo-European languages, have developed partly the same and partly different means for expressing anticausativization, including P-lability. At the synchronic level, different constructions compete with each other, and various semantic factors are here shown to influence the choice of constructions, like the inherent meaning of the verbal root and the degree of control of the P-subject. At the diachronic level, in contrast, the two languages we investigate, Latin and Old NorseIcelandic, display similarities in their development, suggesting that Latin represents an earlier stage than Old Norse-Icelandic in a common development of anticausativization where the Indo-European synthetic mediopassive is replaced with a reflexive in the anticausative function. Our data suggest, moreover, that the demise of the Indo-European synthetic mediopassive was caused by the rise and expansion of the reflexive construction, and not the other way around. The detriment of the IndoEuropean mediopassive construction, moreover, co-occurred with the loss of voice distinctions and the development of transitivity during the period from Latin to Romance. 1. Introduction In this article we investigate the encoding of the anticausative alternation in two early IE languages, Latin and Old Norse-Icelandic, in particular the role played by the interaction between the aspectual template of verbs, the verb’s inherent meaning, i.e. the root, and the nature of the P-subject, e.g. with regard to animacy and control, in determining the different morphosyntactic realizations of the anticausative alternation. Focusing on Latin and Old Norse-Icelandic, we present a first preliminary contribution to a comparative study of the formation, syntax and semantics of anticausatives in Indo-European languages in a diachronic perspective. * We thank Leonid Kulikov, Andrew Koontz-Garboden, Romano Lazzeroni, an anonymous reviewer and the audience at the workshop on lability in Thessaloniki in April 2009 for comments and discussions. We are also indebted to Ilja A. Serzant for help with the Baltic, Slavic and the Ancient Greek data in Section 4.1.5, and to Leonid Kulikov and Nikolaos Lavidas for welcoming this work into their guest-edited volume. We dedicate this article to our late friend, Kjartan G. Ottosson, who passed away in the prime of his life, and to whom anticausativization and valence reduction was not only an endearing and passionate topic, but constituted his life work. 2 A comparison of Latin and Old Norse-Icelandic reveals similarities and differences in the development of the anticausative alternation. In Early and Classical Latin the ancient Indo-European synthetic mediopassive construction functions as a general anticausative device, while this construction is already lost in the earliest attestations of both North- and West-Germanic, found only sporadically in Gothic. In both Latin and Old Norse-Icelandic a Reflexive Construction emerges, which gradually takes over the anticausative domain of the old Indo-European synthetic Mediopassive. In addition, there is also an Active Intransitive Construction, found in both languages, systematically alternating with a corresponding Transitive Construction. Moreover, Old Norse-Icelandic has a second Active Intransitive Construction, namely with oblique subjects where the case marking of the object of the Transitive has been maintained in the anticausative variant. Finally, Old NorseIcelandic has a fourth construction, characterized by the verbal suffix -na, which is well attested in Germanic but not found in Latin. In spite of these differences between Latin and Germanic, it seems, when considering the interplay between the Indo-European synthetic Mediopassive and the Reflexive, that Latin and Old Norse-Icelandic have followed a common evolutionary path in the development of anticausativization. During the Early and the Classical Latin period the Latin reflex of the Indo-European synthetic Mediopassive Construction was found with all kinds of verbs. The Reflexive Construction, in contrast, was much more limited, but gradually throughout the Latin period it expands and invades the functional domains of the earlier synthetic Mediopassive, including developing a passive reading. In Germanic, this development is more or less completed, as the synthetic mediopassive form is only found in Gothic, and is there confined to present tense, and is only attested with a passive reading, and not an anticausative reading. The Reflexive turns out to be the most pervasive anticausative device in Old Norse-Icelandic, suggesting that Old Norse-Icelandic represents a later point on the evolutionary path of anticausativity than Latin. Furthermore, our data suggest that the demise of the Indo-European Mediopassive was caused by the rise and expansion of the Reflexive, and not vice versa. The detriment of the IndoEuropean mediopassive construction, moreover, co-occurred with the loss of voice distinctions and the development of transitivity during the period from Latin to Romance. The discussion is organized as follows: Section 2 presents the concepts of anticausativity and lability, and their semantic constraints across languages. Sections 3 and 4, respectively, describe anticausativization in Latin and Old Norse-Icelandic, the constructions used to realize anticausativization, their synchronic distribution and their diachronic paths of evolution. Section 5 explores the relationship between changes in the morphosyntactic realization of anticausatives and the more general encoding of transitivity in the two languages. Section 6 explores whether the rise of the Reflexive Construction in anticausative function causes the Indo-European Mediopassive Construction to disappear or whether the demise of the Indo-European Mediopassive opened up for the rise and expansion of the Reflexive Construction. Finally, Section 7 provides a summary of the main content and the conclusions of this article. 3 2. The Anticausative Alternation and Lability The term anticausativization refers to the intransitive use of a transitive verb where the original inanimate object or the P-argument, the Undergoer, occurs as a subject. Languages may vary in the morphological devices used to encode this pattern. They may show no change in the verb form of the intransitive member of the alternation, i.e. lability proper, as in English (1a), or they have a dedicated morpheme signaling anticausative status. In Greek a non-active verbal morphology is used (1b) (Alexiadou & Anagnostopolu 2004: 116–117), while Italian uses the reflexive morpheme si obligatorily in such cases (Folli 2002, Schäfer 2008, Cennamo & Jezek 2009, among others): (1) a. b. c. Mark broke the vase. I supa kegete the soup burns.NACT ‘The soup is burning.’ Mario ruppe il vaso. Mario broke the vase ‘Mario broke the vase’ < < < The vase broke. O Janis ekapse ti supa the John burnt.ACT the soup ‘John burnt the soup.’ Il vaso si ruppe. the vase REFL broke ‘The vase broke.’ When there is formal identity between the transitive–intransitive use of a verbal form, the structure is referred to as lability, P-lability in the case of identity between the transitive and intransitive use of a verb, with the original object or the P argument of the transitive pattern occurring as subject of the corresponding derived intransitive pattern (cf. Dixon 1994: 6, Kulikov 2001: 887–888, 2003 and further references therein). Therefore the English example (1a) exemplifies the labile strategy used for anticausativization, while Italian, for instance, uses both P-lability and a dedicated verbal form, in this case the reflexive morpheme (see further discussion in Haspelmath 1987, Kulikov 2003, Schäfer 2008, Koontz-Garboden 2009). Two general semantic constraints on anticausativization have been recognized in the literature: (i) (ii) The spontaneous manifestation of an eventuality (Siewierska 1984: 77) The absence of agent-oriented meaning components or other ‘highly specific meaning components’ that debar the spontaneous interpretation of the verbal process (Haspelmath 1987: 15, 1993: 94) This means that only transitive causative verbs denoting events which may come about spontaneously, without the intervention of an external willful animate causer, may occur in the anticausative alternation (Siewierska 1984: 77, Haspelmath 1987: 15, Levin & Rappaport Hovav 1995: 102, Kulikov 2001, Lazzeroni 2004). In addition, verbs participating in this pattern must denote a nonspecific change of state (cf. Haspelmath 1987: 15, Koontz-Garboden 2009: 84 and references therein). Haspelmath (1993: 39) points out that actions implying specific instruments or methods are excluded, like for instance, bite, cut, dig, paint, etc. in English. In other words, verbs that lexicalize a manner component seem to be excluded from the anticausative alternation (Rappaport Hovav & Levin 2010). This lack of specification of the manner of a change has also been related to the possible restrictions on the subject of this type of change of state verbs. For instance, the subject of a verb like break which denotes a causative accomplishment, consisting 4 of two subevents, a causing eventuality and a resulting change of state (Van Valin 2005) can be an agent (2a), an instrument (2b), a natural force (2c), a stative eventuality (2d), as well as an event (2e) (cf., amongst others, Levin & Rappaport Hovav 1995: 85 and the recent discussion and references in Koontz-Garboden 2009: 84–85, from whom the examples below are adapted): (2) a. b. c. d. e. Mary broke the window. The hammer broke the window. The hurricane broke the window. The weight of the ball broke the window. The blow broke the window. We follow the Role and Reference Grammar classification of predicates, which adopts the Vendler/Dowty Aktionsart taxonomy comprising states, achievements, accomplishments, and activities. Each class has a corresponding causative counterpart. Thus, intransitive accomplishment and achievement verbs like break and shatter, denoting, respectively, a non-punctual and punctual, i.e., instantaneous, telic change of state, have corresponding transitive causative forms. These verbs differ from active accomplishments, the accomplishment use of activity verbs such as eat, drink, write, run, owing to the different relationship between their two subevents, involving causality in the case of causative accomplishments and achievements, and simultaneity in the case of active accomplishments (Van Valin 2005: 32–50). Therefore, with nonspecific change of state verbs such as break the participant in the causing subevent is thematically underspecified, i.e., it need not be an agent, bearing instead the role of Effector (Van Valin & Wilkins 1996, Koontz-Garboden 2009: 85). This property differentiates verbs like break from other change of state verbs like murder, kill, assassinate, which in spite of being characterizable as causative accomplishments, do not allow the anticausative alternation in various languages, including English. This morphosyntactic behavior, in fact, appears to reflect both the relational and the inherent nature of the participants of the event, with the subject of these non-alternating change of state verbs always bearing the thematic role of Agent (Koontz-Garboden 2009: 87), and the object being necessarily human (cf. Cennamo 1995: 91–92 for Italian). Recently, it has been suggested that the possibility of some verbs to participate in the anticausative alternation also stems from the semantic components lexicalized in the inherent meaning of the verbal root, rather than from the aspectual characteristics of the verb or predicate only, i.e. from the event structure template (Levin & Rappaport Hovav 2005). Verbs which lexicalize a manner component rather than a final, result or target state, i.e. verbs which denote a reversible state (Parson 1990), in fact, do not participate in the anticausative alternation, as illustrated in (3–4) for English (from Hale & Keyser 1997: 53, here cited from Levin & Rappaport Hovav 2005: 73): (3) a. b. We splashed mud on the wall. Mud splashed on the wall. (4) a. b. We smeared mud on the wall. *Mud smeared on the wall. 5 The non-alternating verb smear shares the event structure template with the nonalternating verb splash, since both are causative accomplishments. However, smear lexicalizes the means or manner in which the change of state comes about, rather than the result state, unlike splash, whose core meaning refers to the result state only (Levin & Rappaport Hovav 2005: 73, cf. also the discussion in Hale & Keyser 1997: 54). Current research on anticausativization and P-lability has been focused on the contribution of the idiosyncratic aspects, i.e. the root, and the structural components, i.e. the event structure template, of the verb meaning in determining the nature of the anticausative alternation. It has, however, been pointed out by Cennamo (2005) and Cennamo & Jezek (2009) that, although most typically anticausative constructions denote the spontaneous manifestation of a situation, as in (5a) below, anticausative constructions may also be found in contexts where such an interpretation is not available. This is evident with activity verbs (5b) and states (5c) in, for instance, Latin and Italian, and has also been documented crosslinguistically (Letuchiy 2009): (5) a. b. c. vulnus clauditur Latin wound closes.R ‘The wound heals’ bellum continuatur/continuat Latin war continues.R /continues ‘The war is still on’ una comunità omogenea si basa Italian a community homogeneous REFL bases anche su una mediocrità di fondo also on a mediocrity of background ‘A homogeneous community is based also on some sort of mediocrity’ Therefore, the widely accepted constraint of spontaneous manifestation of an eventuality, without the willful intervention of an external causer, only applies to the uses of the anticausative alternation with verbs denoting telic change of state, and marginally with verbs denoting change of location in some languages. In other words, these constraints only apply to verbs which lexicalize a final state, result or target state, i.e. causative achievements and accomplishments, which are well known as instantiating the core of the anticausative category in several languages (Cennamo & Jezek 2009; see also Koontz-Garboden 2009 for a criticism of the alleged suppression of the causer in the anticausative alternation). In our discussion in the following sections we address the issue of which factors determine the different morphosyntactic realization of the anticausative alternation in the diachrony of Latin (Section 3) and Old Norse-Icelandic (Section 4). In particular, we investigate the different contribution of the structural and idiosyncratic aspects of verb meaning in determining the distribution of the various strategies in the two languages investigated, both synchronically and diachronically. We focus our discussion on intransitive patterns derived from originally transitive ones, where the original transitive inanimate object occurs as an intransitive subject, this being optionally registered through different verb morphology or the case marking of the derived subject. 6 3. Anticausatives in Latin The following three constructions are employed to denote anticausativization in Latin, each exemplified in (6) below: (i) (ii) The Mediopassive Construction with the r-form The Reflexive Construction, i.e. the reflexive se together with a verb in the active voice The Active Intransitive Construction, i.e. P-lability (iii) (6) a. b. c. aperitur foris opens.R door.NOM ‘The door opens/is opened’ foris se aperuit door.NOM REFL opened ‘The door opened’ foris aperit door.NOM opens ‘The door opened’ (Plaut. Persa 300) Mediopassive Reflexive Active Intransitive The original mediopassive function of the verbal suffix -r is a long-standing matter of contention in Indo-European (cf. Kurzovà 1993: 157–171 and references therein). Functionally it covers the domains of middles, anticausatives, passives, and impersonals or subjectless clauses. In the synchronic grammar of Latin it has primarily a passive function (Flobert 1975, Baldi 1977). It may be regarded as a marker of the ‘non-active voice’, marking different points along a continuum of detransitivization, instantiated by patterns which depart from the prototypical transitive encoding of a situation whereby a highly agentive subject acts upon a patient participant, which in turn results in a change in the patient participant (Cennamo 1998: 78–81). Returning to the three anticausative constructions in Latin exemplified in (6) above, they are usually regarded as interchangeable in the literature, and differences in their distribution are viewed as reflecting different time spans in the history of Latin (Feltenius 1977, inter alia). The Mediopassive is the most commonly used anticausative construction in Early Latin, whilst the Active Intransitive Construction increases at later stages, alongside the Reflexive Construction, to the detriment of the Mediopassive Construction. This goes hand in hand with the gradual demise of the r-form, which in turn reflects the more general recasting of the voice system (Cennamo 1998, 2006, 2008). Contrary to current views, we argue that the distribution of the three strategies reflect aspectual and thematic differences, interacting, in turn, with changes in the encoding of voice and argument structure in the transition to Romance (Cennamo 1998, 2000b, 2006). 3.1 Synchronic Aspects A careful analysis of the morphological devices for marking anticausatives in Latin reveals that the use of the three constructions, the Mediopassive r-form, the Reflexive Construction and the Active Intransitive Construction, is not entirely equivalent, but reflects the interplay between the aspectual template of predicates, the meaning 7 components lexicalized in the verb, i.e. in the root, as well as inherent and relational properties of the subject, like animacy and degree of control. We will now discuss each of these constructions in turn. 3.1.1 The Mediopassive Construction The Mediopassive Construction with the r-form is found at all stages in the history of Latin with all verb classes which allow the anticausative alternation. These are causative achievements (7a), causative accomplishments (7b), gradual completion verbs (7), as well as activities (7d): (7) a. b. b. c. frangitur aestus breaks.R tide ‘When the rolling tide breaks’ tempora mutantur time changes.R ‘Time changes’ memoria minuitur memory decreases.R ‘Memory is impaired’ saxa volutantur stones roll.R ‘Stones are rolled’/Stones roll’ At times there is ambiguity between an anticausative and a passive interpretation, as in (7d), only resolved by the context. 3.1.2 The Reflexive Construction The Reflexive Construction in Early and Classical Latin is mainly found with causative accomplishments and achievements, like e.g., scindere ‘crack’ and rumpere ‘break’: (8) a. b. lutamenta scindunt se (Cat. Agr. 128) plaster.PL crack REFL ‘Plaster cracks’ brassica commutat sese sempre cum calore (Cat. Agr. 157,1) cabbage.NOM changes REFL always with heat.ABL ‘Cabbage constantly changes its nature with heat’ The reflexive is not attested with verbs of variable or reduced telicity, such as gradual completion verbs, as in (9). Examples of this type are not attested in Latin until the 1st century and onwards. (9) *memoria se minuit memory REFL decreases Intended meaning: ‘Memory decreases’ In Early and Classical Latin the reflexive is not attested in anticausative function with activities, like volutare ‘roll’ (10), for which only the r-form and the Active 8 Intransitive Construction are found in anticausative function (see also Cennamo 1998, 2000b and discussion in 3.2 below): (10) *saxa se volutant stones REFL roll Intended meaning: ‘Stones roll’ A provisional generalization that seems to emerge is that the Reflexive Construction is used in Early and Classical Latin as a strategy for anticausativization only with inherently telic predicates, i.e. with verbs lexically encoding a final, result or target state, as shown in (11) below: (11) a. b. valvae se ipsae aperuerunt (Cic. Div. 1, 34, 74) doors.NOM REFL themselves.NOM opened ‘The doors suddenly opened of their own accord’ brassica commutat sese semper cum calore (Cat, Agr. 157, 1) cabbage.NOM changes REFL always with heat.ABL ‘Cabbage constantly changes its nature with heat’ The Reflexive Construction seems to be preferred over the Mediopassive Construction when the subject, although inanimate, is personified, therefore showing some degree of control, as in (11a), which can be contrasted with (11b), where no personification is involved, and the presence of se simply marks the intransitive anticausative variant (cf. the discussion in Cennamo 1998 and Section 3.2). 3.1.3 The Active Intransitive Construction In Early and Classical Latin the Active Intransitive Construction is mainly instantiated by gradual completion verbs, as in (12a), and activities, as in (12b): (12) a. b. irae leniunt (Plaut. Mil. 583) anger.NOM.PL soothes.3PL ‘Anger soothes’ saxa volutant stones roll ‘Stones roll’ The Active Intransitive is not used in anticausative function with verbs lexically encoding a final state, i.e., causative accomplishments and achievements such as rumpere ‘to break’ and scindere ‘to crackle’ (13a–d), which may be regarded as the core of the category of verbs in Latin and in other languages which shows this type of transitive–intransitive alternation. One exception from this is found with rare examples like aperire ‘to open’ in Early Latin, e.g., Plautus, shown in (13e). This verb, however, denotes a reversible change of state, i.e., a target state, unlike scindere and rumpere, which denote a non-reversible change, i.e. a result state, and which therefore lexicalize a higher degree of telicity. (13) a. *foris rumpit door breaks Intended meaning: ‘The door breaks’ 9 b. c. d. e. *lutamenta scindunt plaster.PL crack Intended meaning: ‘Plaster cracks’ *vulnus claudit wound closes Intended meaning: ‘The wound heals’ *corrumpit iam cena spoils already dinner Intended meaning: ‘Dinner is spoiling already’ foris aperit door.NOM opens ‘The door opens’ (Plaut. Persa 300) The Active Intransitive Construction in anticausative function really flourishes in Late Latin, especially in 4th century non-fictive text, like veterinary texts. By this age verbs expressing causative accomplishments and achievements alternate freely between all three anticausative constructions, as illustrated with the verb rumpere ‘break’ in (14) (Pirson 1906, Feltenius 1977): (14) a. b. c. rumpuntur dentes (Chiron 776) break.R teeth dentes se rumpunt teeth REFL break rumpunt dentes break teeth ‘Its teeth break (sc. iumentum)’ Mediopassive Reflexive Active Intransitive 3.1.4 Interim Summary A close inspection of the distribution of the three different anticausative devices in Early and Classical Latin reveals that the choice between them reflects both the idiosyncratic, i.e. the root, and the structural aspect, i.e. the event structure template of verbal meaning. The Mediopassive is found with all types of verbs in Early and Classical Latin. The Reflexive Construction, in fact, can only occur with telic verbs, like scindere ‘crack, movere ‘move’ and aperire ‘open’ at this stage. The Active Intransitive, moreover, occurs most typically with verbs which do not lexicalize the attainment of a final state, i.e. the endpoint of the process. These factors, i.e. the idiosyncratic meaning of the root and the event structure template, interact, in the course of time, with changes in the voice system and the encoding of argument structure (Cennamo 1998, 2000b, forthcoming). 3.2 Anticausatives, P-Lability and Transitivity in Late Latin In Late Latin the distribution of the three anticausative constructions across verbs changes: the Mediopassive Construction, in fact, although still widely attested, is often instantiated by the same verbs as the Reflexive and the Active Intransitive, which have come to be used with different aspectual classes of verbs, with which they were not found in Early and Classical Latin. Therefore the choice of anticausative construction is no longer determined by the semantics of the predicate and the inherent and relational properties of the subject. 10 In particular, the Reflexive occurs also with verbs of reduced telicity, comprising gradual completion verbs like minuere ‘to decrease’ (15a), as well as with other types of accomplishments, for instance deadjectival and denominal verbs like sanare ‘to heal’ and cicatricare ‘to heal’ (15b) (Pirson 1906, Feltenius 1977): (15) a. minuente se morbo (Plin. Nat. 23, 50) decreasing.ABL REFL disease ‘When the disease is on the decline’ b. vulnera cum se cicatricaverint (Orib. Syn. 7, 10 Aa) wounds when REFL will-have-healed ‘When the wounds will have healed’ Late Latin The use of the mediopassive r-form in anticausative function in Late Latin, illustrated in (16), from a 6th century text, might also reflect the so-called deponentization (Flobert 1975), i.e. the widespread use of the passive morphology in active function with all kinds of verbs (Cennamo 1998, 2009 and references therein). (16) vulnera cicatricantur (Orib. Syn. 7, 3) wounds heal.R ‘The wounds heal’ Late Latin This has, however, been argued to be a part of the reorganization of voice distinctions and the consequent functional opacity of the voice morphology conveying them (Cennamo 1998, 2005, 2006, Herman 2002). The mediopassive r-form, in fact, may occur in active function, replacing active morphology, with both intransitive and transitive verbs, already in 4th century texts, as exemplified in (17a), and even more so at later stages, as in (17b), from the 9th century A.D. (see also Gianollo, this volume): (17) a. b. et sabbato non ieiunantur (Peregr. Aeth. 27, 1) and Saturday.ABL not fast.R.3PL ‘And they do not fast on Saturdays’ cum illo, qui eam ... dugatur uxorem with he.ABL who.NOM her.ACC take.R.3SG spouse.ACC ‘With that who will marry her’ or ‘With that who will take her as his spouse’ (Cod. Verc. cap. 192; Löfstedt 1977: 275) Towards the end of the imperial age, with attestations from the second half of the 4th century A.D., as in the Mulomedichina Chironis, and even more so in later texts, the Reflexive also occurs with non-inherently telic verbs, i.e. gradual completion verbs, as well as with activities (Cennamo 2001b, 2006: 333, fn. 8). This pattern, therefore, is also found with accomplishments like citare ‘cause’ and provocare ‘cause’, denoting the coming into being of an entity, and with activities like vexare ‘oppress’, servare ‘keep’ and excusare ‘justify, excuse’. At times ambiguity is found between an anticausative and a passive interpretation, i.e. between a spontaneous vs. an induced process reading, as shown in (18a) (Cennamo 1998, 2000b, 2006): 11 (18) a. b. mala … toto anno servare se possunt (Pall. 3, 25, 18) apples whole.ABL year.ABL keep REFL can ‘Apples ... can keep/be kept for the whole year’ stercora si se … provocaverint (Chron 230) excrements if REFL will-have-been-caused ‘Excrement, if ... it is induced’ The verb provocare ‘to cause, provoke’ in (18b), however, when occurring in the Reflexive Construction, can only denote an induced process, i.e. with the so-called passive reading. Such a reading is also the only possible reading found with activity verbs when occurring in the Reflexive, as shown in (19): (19) qui se vocat padule (Cod. Dipl. Bar. 5) that REFL calls marsh ‘(A place) that is called marsh’ Also the Active Intransitive increasingly occurs in anticausative function, alternating with the Reflexive, as in (20a–b), from the 4th century veterinary text Mulomedichina Chironis (Feltenius 1977: 82, Cennamo 2006: 317): (20) a. b. ut confirmet (sc. vulnus) (Chiron 670) Active Intransitive in-order-to heal (wound) ‘So as it (sc. the wound) heals’ donec cicatrix oculo se confirmet (Chiron 76) Reflexive till scar.NOM eye.DAT REFL heal ‘Until the scar in its eye heals’ In the same texts, illustrating the expansion of the Active Intransitive to include verbs with different aspectual templates, oblique anticausatives are also found, as shown in (21). In this construction the subject, or the S-argument, occurs in the accusative case, rather than in the canonical nominative case (cf. Cennamo 2009 and references therein):1 (21) multos languores sanantur many.ACC illnesses.ACC are-healed.3PL ‘Many illnesses heal’ This oblique encoding of the subject reflects more general changes in the encoding of the argument structure of the clause in Late Latin, leading to the use of the accusative in subject function, as a part of the loss of the case system and as a part of the general rise of head-marked patterns of active syntax in the transition to Romance (for a full discussion of this issue, cf. Cennamo 2009). To summarize, in Late Latin, therefore, the Reflexive spreads from inherently telic and punctual verbs denoting change of state and location, i.e., situations that can be brought about spontaneously, with verbs like mutare ‘change’, frangere ‘break’ and 1 We remain impervious here to the question of whether the accusative is an “inherent” or “structural” case, but refer the reader to Barðdal (2011a), where the generative distinction between structural and lexical case is contested on the basis of non-corroborated empirical predictions following from the dichotomy. 12 scindere ‘crack’, to non-inherently telic and atelic verbs. These are active accomplishments and activities, found either with the Reflexive in the passive function, or found in both the Active Intransitive and the Mediopassive in the anticausative function. It is only when the event clearly denotes an induced process that the Mediopassive r-form is preferred (Cennamo 1998, 2006). Following Haspelmath (1993: 93), verbs such as facere ‘do, make’, citare ‘cause’, vexare ‘oppress’ and provocare ‘cause, induce’ might be defined as having an Agentoriented meaning component, since they denote situations that most typically involve a human causer, which in turn makes the spontaneous interpretation highly unlikely (Cennamo 1998: 96). We believe that Haspelmath’s intuition can be made more precise if we restate it in aspectual terms, i.e. in terms of the aspectual template of verbs, as proposed above. Interestingly, in late texts, the reflexive pronoun and the r-form may co-occur in anticausative function, as in (22), at times with ambiguity between an anticausative and a passive reading, depending on the verb and on the syntactic context (see further discussion in Cennamo 1998): (22) a. b. si autem minutetur se medicamen if then pulverizes.R REFL drug ‘If then the drug pulverizes/gets pulverized’ (Orib. Eup. 4, 63; Svennung 1935: 463, n. 2) (VI A.D.) se festivitatis habebitur dies REFL festivity will-have.R day.NOM ‘It will be a day of festivity/The day of festivity will take place’ (Papyr. Mar. 8; Norberg 1943: 166) (VIII A.D.) The co-occurrence of the two morphological devices, the mediopassive r-form and the reflexive pronoun, suggests a functional equivalence of the two constructions, and is a clear sign of the restructuring of the voice system taking place in Late Latin, with the mediopassive r-form gradually demising in the spoken language and being replaced by the Reflexive Construction. This construction, in turn, spreads to all verb classes in Late Latin, with both animate and inanimate subjects. Hence, it is no longer confined to the truly reflexive and core anticausative domains, i.e., with verbs of telic change, as in Early and Classical Latin (Cennamo 1998; 2005; 2006). In particular, se together with the active form is no longer confined to the three functional domains in which it occurred in early and Classical Latin, namely: (i) reflexives, marking coreference between the A and P arguments (ii) inherent reflexives, a.k.a endoreflexives (iii) core anticausatives, i.e., verbs of telic change With inherent reflexives, i.e., with intransitive patterns where an original animate S argument occurs as subject, in Early and Classical Latin fluctuation is often found among voice forms. This may be taken as reflecting the degree of control of the subject over the verbal process, as illustrated in (23) (cf. Cennamo 1998: 86): (23) a. adplicor ignotis oris (Ov. epist. 7, 117) lean.R unknown.ABL shores.ABL ‘I land on unknown shores’ – Control 13 b. c. quo sese applicet (Pacuv. trag. 370) where REFL lean ‘Where he goes’ or ‘Where he leans’ quo applicem? (Enn. trag. 77) where lean ‘Where shall I go?’ + Control The r-form in (23a) marks the affectedness or inactive nature of the subject, whilst in (23b) the use of the reflexive signals volitionality of the subject. The active intransitive in (23c) is neutral as to the realization of this semantic feature. Although these alternations are usually explained in the literature as being due to stylistic reasons or personal choices on the part of the author, in many cases they convey a difference in control, as just illustrated in (23) above. The notion of control indeed plays an important role in the marking of transitivity in Latin, and probably in early Indo-European, an issue that we leave for further investigation. However, this notion is relevant for voice fluctuations with animate subjects (Cennamo 1998) and the ‘impersonal’ encoding of situations, as shown in (24), attested with activities, like delectare ‘please’ and states like poenitere ‘repent’, libere ‘like’ and others. Forms such as me fallit ‘I am wrong’, me delectat ‘I am delighted’, me libet ‘I am pleased’, registered in traditional Latin grammars as ‘impersonal’, may be regarded, instead, as crystallizing a usage that must have been very common in earlier stages of the language, with an oblique subject and an impersonal verb form which denotes lack of control of the subject over the verbal process. This is illustrated in (24) for the verb fallere ‘deceive’ (see full discussion in Cennamo 2010): (24) a. b. c. d. nisi memoria me fallit Active Transitive if-not memory.NOM me.ACC deceives ‘If memory does not deceive me’ nisi me forte fallo Reflexive if-not I.ACC strongly deceives/is-in-error ‘If I am not completely wrong’ nisi fallor Mediopassive if-not deceives/is-in-error ‘If I am not wrong’ quod me non fefellit Oblique Intransitive as-far-as-this I.ACC not deceives/is-in-error ‘I was not (I did not happen to be) wrong as far as this is concerned’ Interestingly, (24d) me non fefellit, is morphosyntactically identical with analogous Old Norse-Icelandic oblique subject patterns discussed in section 4.1.5. In Late Latin the notions of control and aspect no longer play a role in determining voice realizations and alternations in the reflexive-middle continuum and anticausatives, respectively. This, in turn, means that changes in the distribution of the strategies used to encode anticausatives reflect deep and wide-ranging changes taking place in the domains of voice and grammatical relations, as is partially illustrated above. We refer the interested reader to Cennamo (1998, 2000b, 2006, 2009) for further discussions on this topic. 14 3.3 Summary In Late Latin the Reflexive and the Active Intransitive come to be used in anticausative function with different aspectual classes of verbs. The Reflexive expands to verbs of variable and reduced telicity and to activities, while the Active Intransitive spreads to causative accomplishments and achievements. In late texts, therefore, all verbs may freely alternate between the three forms when used anticausatively, regardless of their structural or inherent features. This development can be regarded as part and parcel of wider and more pervasive changes in the encoding of argument structure in the passage to Romance, a result of which is the emergence of Oblique Intransitives, where the subject is in the accusative case rather than in the canonical nominative case. Both the event structure template of verbs and the meaning components lexicalized in the verb, i.e., in the root, in particular the type of change encoded as well as the degree of control of the P-subject, appear to affect the choice of construction, interacting, in the course of time, with changes in the encoding of transitivity. Although the progression of the change seems to be mainly aspectually driven, further investigation is needed, on a larger corpus, in order to give a more fine-grained description of the path of development. A more fine-grained analysis is also needed to detect, for instance, whether the anticausative Reflexive occurs earlier with gradual completion verbs, as seems to be the case, for instance with verbs such as minuere ‘decrease’, than with other types of non inherently telic predicates, such as active accomplishments like citare ‘cause’, and with atelic verbs, i.e. activities, such as vocare ‘call’. Moreover, the role played by inherent lexical properties of the verb, i.e. the root, and by the nature of the verbal argument, i.e., its animacy and control, in determining the interpretation of the construction as anticausative needs to be further explored, and in particular the interplay between these factors in cases where there is ambiguity between a spontaneous and an induced reading. We conclude, however, that in Late Latin the anticausative Reflexive Construction is no longer confined to verbs of inherent telicity, i.e. to verbs lexically encoding a final, result or target state. Conversely, the lack of such encoding and the presence of a manner or processual component (so-called ‘specific change’ in Haspelmath’s terms) in the verb’s inherent meaning appears no longer to be a requirement for the occurrence of the Active Intransitive in anticausative function, unlike in Early Latin. 4. Anticausatives in Old Norse-Icelandic Because of its status as the best-attested representative of North Germanic, we now investigate the constructions that are available in Old Norse-Icelandic for expressing anticausatives, their relationship with their transitive counterpart, and the interplay between the structural and lexical aspects of verb meaning in determining the selection of a particular construction. The main anticausativization strategies are the following (Ottosson 2009): 15 (i) (ii) (iii) (iv) Suffixed na-verbs The Reflexive construction, i.e. so-called ‘middle’ sk-verbs, originating from a cliticized reflexive pronoun The Nominative Active Intransitive Construction The Oblique Active Intransitive Construction We will now discuss each in turn. 4.1 Synchronic Aspects 4.1.1 The na-Verbs In Old Norse-Icelandic the verbal suffix -na has been regarded as a detransitivizer, marking the spontaneous manifestation of a situation and the lack of control or affectedness of the subject.2 The anticausative function of the suffix is very common and is attested with causative achievements, different types of accomplishments, and marginally with states. Some examples of causative–anticausative pairs are given in (24); the transitive verbs in question can be either strong (24a) or weak (24b): (24) a. b. brjóta – brotna ‘break’ bleikja ‘bleach’ (adj. bleikr ‘pale’) – blikna ‘grow pale’ Traditionally in the Icelandic literature, the -na suffix with the so-called na-verbs is considered as being inchoative, a term which refers to the inception or beginning of an event, signaling the transition to a state, since this suffix generally occurs with change of state verbs. A more accurate view is to regard the -na suffix as a general detranisitivizer, marking an inactive subject, having most typically, albeit not exclusively, a clear anticausative function (Ottosson 2009, Lazzeroni 2009). This analysis is corroborated by the fact that the suffix -na is, indeed, found alternating with both transitive and intransitive verbs, as in (25). (25) vekja ‘waken’ – vaka ‘be awake’ – vakna ‘wake up’ 4.1.2 The Reflexive The Old Norse-Icelandic Reflexive Construction with sk-verbs originates in cliticization of the reflexive pronoun to the verb (Ottósson 1992). The reflexive clitic pronoun became an affix, which is realized as -sk in all persons except for the first person singular, where it is -mk. The sk-form comprises a wide range of functions, notably reflexive, reciprocal and anticausative, and in rare cases in Old NorseIcelandic, passive (Ottosson 2009: 23), a function that later becomes dominant in Mainland Scandinavian (cf. Barðdal & Molnár 2003). This development is very similar to the one displayed by the Latin se. 2 See Lazzeroni (2009: 16) for a recent hypothesis that the detransitivizing -na suffix in Germanic, Baltic and Slavic is the same suffix as the homophonous transitivizing -na suffix in Sanskrit, Hittite, Greek, Celtic and Armenian. 16 The reflexive and reciprocal functions of the Reflexive Construction are shown in (26) below (cf. also Andersen 1990 and Barðdal & Molnar 2003: 253–254 for Modern Icelandic): (26) a. b. en þó lagðist hann niður og sofnaði but though laid-sk he down and fell-asleep ‘but despite this he laid down and fell asleep’ (Grænlendingaþáttur, Ch. 6) of drepask kváðu (Þjóðolfr ór Hvini Ynglingatal 11) of kill-sk said ‘they said they would kill each other’ The anticausative function of the Reflexive Construction is shown in (27) below. Verbs occurring in this construction are generally causative achievements and accomplishments, although some activities are also found, as in (28). (27) svo hristusk hringarnir then shook-sk rings-the.NOM ‘then the rings shook’ (28) a. b. b. c. jörð bifask earth.NOM shakes-sk ‘the earth shakes’ þar til er bogastrengrinn skarsk there till when bowstring-the.NOM cut-sk ‘until the bow-string got cut’ (Cleasby & Vigfusson 1874) þat mólsk á kverninni it ground-sk on handmill-the ‘it got ground on the handmill’ (Cleasby & Vigfusson 1874) þá mun brátt af þvásk öll sú sǿmð then will soon off wash-sk all that honor.NOM ‘then all the honor will soon get washed off’ (Cleasby & Vigfusson 1874) Finally, psychological verbs in the Reflexive Construction are usually states, as in (29a) below (cf. Barðdal 2001a), although of course not all psychological predicates are historically derived anticausatives (cf. Barðdal & Eythórsson 2009). (29) a. b. ek þess minnumk I.NOM this.GEN remember-sk ‘I remember that’ að minna konunginn á með nokkuru móti to remind king-the.ACC on with any way ‘to remind the king some way or another’ (Þormóðar þáttur [eftir Flateyjarbók], Ch. 2) Observe, that with these anticausatives an animate P-argument of the transitive becomes the S-argument of a corresponding anticausative. The example in (29b) gives the causative variant of the verb minna ‘remind, remember’ in Old Norse-Icelandic. 17 The main property of the anticausative Reflexive Construction seems to be to signal lack of external agency, whereas the -na suffix is mainly a marker of telicity, as well as of an inactive subject (Ottosson 2009: 18, Lazzeroni 2009: 16–17). 4.1.3 Interim Summary Our comparison between anticausative na-verbs and reflexive sk-verbs in Old NorseIcelandic reveals that they occur with different aspectual classes. Whilst the -na suffix occurs with achievements, accomplishments and marginally with states, the -sk suffix also occurs with activity verbs. Therefore there appears to be a clear distinction between the two devices: -na is incompatible with atelic, dynamic predicates, i.e. activities, whereas -sk is compatible with both telic and atelic predicates. Ottosson (2009: 24) observes that an important difference between the two constructions is that the one with na-verbs puts an emphasis on the effect of the situation denoted by the verb, i.e. it occurs with verbs which lexicalize a final, result or target state. The Reflexive, in contrast, conveys how this change comes about, occurring therefore also with verbs which lexicalize manner, i.e. with activities. It is also worth noticing that na-verbs and sk-verbs in Old Norse-Icelandic have in common that they are to a high degree lexicalized. That is, the causative and its anticausative variant are not necessarily examples of the ‘same’ verb, but have lexicalized in different ways. 4.1.4 The Nominative Active Intransitive (lability proper) The Active Intransitive strategy, i.e., P-lability, is apparently quite rare in Old NorseIcelandic (Ottosson 2009: 43). The issue, however, needs further investigation, with an in-depth study of the texts showing this pattern. Only very few examples, in fact, are given by Ottoson for the use of the Active Intransitive in anticausative function, with the derived subject occurring in the nominative case and agreeing with the verb, as in (30) below, of the gradual completion verb minnka ‘diminish, reduce’, where (30a) contains a transitive verb and (30b) its anticausative variant: (30) a. b. minnkuðu þessi orð mjök hans gleði diminished these words.NOM much his joy.ACC ‘these words greatly diminished his joy’ (Barlaams ok Josaphats saga 1981:207f) hans kraptr ok máttr minnkar his power.NOM and might.NOM diminishes ‘his power and might is reduced’ (Barlaams ok Josaphats saga 1981:20) This requires an investigation aiming at exploring the classes of verbs instantiating this construction, and at detecting whether the construction is also found with verbs which lexicalize a final, result or a target state, such as break, tear, crackle, open, or whether it is mainly confined to non-inherently telic verbs, as exemplified in (30b). 4.1.5 The Oblique Active Intransitive In contrast to the construction discussed in the previous section, a much more widely used type of lability in Old Norse-Icelandic involves verbs taking oblique subjects 18 (“impersonal detransitives” in Ottosson’s terminology). The nominative agentive subject of the transitive construction is “suppressed” in this anticausative variant, and the original object shows up as a subject preserving its original object case, accusative, dative, or more rarely genitive. In this pattern the verb is in the default third singular. The predicates occurring in this pattern most typically denote a telic change of state, i.e., they comprise causative achievements and accomplishments (Ottosson 2009: 45), as illustrated in (31–34): (31) (32) (33) a. blása hár af höfði e-m blow hair.ACC of head sby ‘blow hair off somebody’s head’ b. hafði blásit hauginn had blown mound-the.ACC ‘the mound had eroded’ a. þá hvelfðu þeir því sjalfir then capsized they.NOM it.DAT self ‘then they capsized it themselves’ b. í því hvelfdi skipinu skjótt in that capsized ship-the.DAT quickly ‘then the ship capsized quickly’ a. Þeir brutu skip sitt they.NOM broke ship.ACC their ‘they broke their ship/they were shipwrecked’ skipin braut ships-the.ACC broke ‘The ships were shipwrecked’ b. (34) a. b. ljúka einhverju finish something.DAT ‘finish something’ Einhverju lýkr something.DAT finishes ‘Something ends’ The nature of the subject is relevant as well; as with other verbs taking oblique subjects, it is never an agent (cf. Barðdal 2001b: 103, 2004). Ottosson (2009: 44) regards the Oblique Active Intransitive as a special Old Norse-Icelandic development. However, we have come across similar examples in Bavarian German, Modern Russian and Lithuanian, as illustrated in (35–37): (35) Es trieb den Kahn an den Strand. (Kainhofer 2002) it drove the.ACC boat to the beach ‘The boat drifted to the beach.’ Bavarian 19 (36) Sodą prinešė sniego. garden.ACC brought snow.GEN ‘The garden was filled with snow.’ (37) Lodku uneslo vniz po tečeniju. boat.ACC drifted-away down on stream ‘The boat drifted down the stream.’ Lithuanian Modern Russian Clearly there s a need to explore the relationship between the anticausative patterns illustrated in (31–34) in Old Norse-Icelandic, and the ones in (35–37) for some other Indo-European languages, and other constructions with oblique subject-like arguments, denoting lack of control of the S-argument over the verbal process. In all these structures, in fact, the situation expressed by the verb is encoded as affecting the nuclear argument, with no causer involved. This type of pattern is attested in Old Norse-Icelandic, as in (38b–39b), as well as in some other early Indo-European languages, like Old Russian, Ancient Greek and Old Saxon, shown in (40–43): (38) a. b. (39) a. b. fýsa einhvern einhvers urge someone.ACC something.GEN ‘urge someone to do something’ Einhvern fýsir til einhvers someone.ACC urges to something.GEN ‘Someone desires something’ gefa einhverjum eitthvat give somebody.DAT something.ACC ‘give somebody something’ og er þeim gaf byr and when they.DAT gave wind.ACC ‘and when they got wind’ The examples in (38–39) are like the examples in (31–37) in the sense that they convey a spontaneous event, happening by itself, with no external causer. However, they differ in that they are ditransitive and the S-argument is animate. Similar examples are also found in more early Indo-European languages: (40) fâhit im an sâlig thing Old Saxon brings him.DAT to holy thing ‘he is brought to holy things’ (Heliand, here cited from Dewey & Arnett 2010) (41) mē dē skotodinian iliggon te hymin empoiēsē Ancient Greek not already dizziness.ACC faintness.ACC and you.DAT cause.SUBJ ‘Let you be brought faintness and dizziness’ (Plat. Leg 892.e.7-a.1, here cited from Seržant 2010) (42) Jako g’nasta pout’ m’nog’, ti tako pristig’ša Old Russian so chased paths many thee.DAT such overtook ‘You have followed many paths, so you have been overwhelmed’ (Nest.Zhit.Theod.3, here cited from Smitherman 2010) 20 Observe that the Old Saxon example with fâhan ‘bring’ in (40) and the Ancient Greek example in (41) with empoeiō ‘bring’ are parallel to the Old Norse-Icelandic example with gefa ‘give’ in (39b) above, exhibiting exactly the same case pattern, Dat-Acc. The existence of Oblique Active Intransitive anticausatives and of other corresponding derived transitives with an oblique subject in other Germanic languages, both early and modern, as well as in Baltic, Slavic and Ancient Greek, suggests an inheritance, although perhaps only a West-Indo-European inheritance (see also discussion of similar lack of control patterns in Latin in section 3.2). The examples in (39b) and (40–41) are exactly like anticausatives in that they denote a spontaneous event without a willful external causer and that they are derived from corresponding causatives where the causer is the A-argument. They only differ from prototypical anticausatives in that they are derived from causative ditransitives and not causative transitives. The reason that examples like these have gone unnoticed in the literature on anticausatives is most likely because they are confined to the early or archaic Indo-European languages where nominal case marking is still intact, meaning that the dative and the accusative of the Dat-Acc case frame must be morphologically distinguishable in order for structures like these to be found. 4.1.6 Summary Old Norse-Icelandic presents a varied picture with respect to anticausative constructions. The formation with the na-suffix is quite common, while the Reflexive Construction is generally used to form anticausatives. In addition, a common anticausative strategy involves active intransitive verbs taking oblique subjects exhibiting the same object case as their transitive counterparts and a verb in the third person singular. In contrast, P-lability proper i.e., the strategy where the same verb form occurs both as a transitive and as an anticausative with a nominative subject, is only rarely attested. 4.2 Conjecturing the Diachronic Development of Anticausative Constructions in North Germanic As pointed out in Section 4.1.1, Germanic na-verbs ultimately derive from ProtoIndo-European nasal presents and already existed at the Proto-Germanic stage (Ringe 2006: 176–179, Suzuki 1989: 151–159, Ottoson 2009, Lazzeroni 2009: 176–179). In Old Norse-Icelandic anticausative na-verbs can generally be considered as having had a transitive counterpart, even though in many cases only a corresponding verbal adjective is attested. As discussed by Ottosson (2009), Gothic na-verbs also commonly enter into causative–anticausative pairs, although not as regularly as in Old Norse-Icelandic. In this light, it appears legitimate to conclude with Ottosson (2009) that the anticausative function of the -na suffix in Old Norse-Icelandic can be traced back to Proto-Germanic. The Reflexive -sk construction is presumably a North Germanic innovation, even though it is not attested in the early runic inscriptions. Owing to the limited scope of this corpus, the absence of the Reflexive Construction at this stage is not necessarily very significant. In fact, Ottosson (2008, 2009) argues that the Reflexive Construction must have originated in Proto-Nordic, because already in the earliest Old NorseIcelandic sources it exhibits a very extensive and systematic anticausative function. In 21 order for this development to have become so extensive, it must be assumed that it had already begun at an even earlier stage. However, the other West Germanic languages lacked reflexive pronouns, developing instead an anticausativization strategy involving formal identity between the transitive and intransitive use of the verb, i.e., P-lability. It needs be be investigated whether, and to which extent, the development of P-lability as the only anticausative strategy in these languages is related to the change of Oblique Active Intransitive anticausatives into Nominative Active Intransitive anticausatives, through the canonization of subject marking. Such a process is known under the label Nominative Substitution, and has been found in languages irrespective of the loss of morphological case (Seefranz-Montag 1983, 1984, Smith 1994, 1996, Barðdal & Eythórsson 2003, Eythórsson and Barðdal 2005). Clearly, however, when morphological case is lost, all potential Oblique Active Intransitive anticausatives automatically change into Nominative Active Intransitive anticausatives. The labile ‘proper’ alternation only occurs to a very limited degree in Old NorseIcelandic. Instead, this language has an anticausative construction with an oblique subject and default verb morphology. Similar structures also occur in Bavarian German and Old Saxon, and, outside Germanic, in Russian, Lithuanian and Ancient Greek, and may have been inherited from an earlier stage of the Indo-European languages. It should be noted that there is a sporadic tendency in the history of Icelandic for the oblique subjects of such verbs to be replaced by nominative subjects, through Nominative Substitution (Eythórsson 2000a–b, 2002, Barðdal & Eythórsson 2003, Barðdal 2011b), even though the case system of Icelandic has stayed intact throughout the Icelandic period, and no loss of case marking is detectable (Barðdal 2001b, 2008, 2009). Ottosson (2009) speculates that an important factor triggering these developments is the demise of the synthetic mediopassive of Proto-Germanic which would partly have had an anticausative function, i.e. the r-form. Within Old Germanic this passive category only survives in the present tense in Gothic. The loss of this passive would have contributed to the emergence of new, alternative anticausative formations, i.e., na- and sk-verbs. While it is undeniable that the encoding of anticausativization in Old Norse-Icelandic is to be seen within the wider context of the encoding of transitivity, there is no evidence for the demise of the synthetic mediopassive as being a possible trigger of the different anticausative strategies and their distribution in early West Germanic in general and Old Norse-Icelandic in particular. Moreover, contra Ottosson, we find it quite likely that Oblique Active Intransitives, which are robust in Old Norse-Icelandic, existed at the earliest stage of Germanic, suggesting that they are the result of a systematic earlier development. Comparative evidence from other languages suggests that these have a ProtoGermanic and even Indo-European origins (Barðdal & Eythórsson 2009). 5. Constraints on anticausativization in Latin and Old Norse-Icelandic: Similarities and differences The analysis of the morphosyntactic constructions expressing anticausatives in Latin and Old Norse-Icelandic has shown that they involve the aspectual templates of predicates, the meaning components lexicalized in the verb, and the degree of control or thematic (under)specification of the subject. 22 The relevant constructions in the two languages are generally of the same or similar kind, although the details differ according to language-specific possibilities. In particular, both languages make use of the reflexive pattern and the active intransitive. The mediopassive r-form in Latin and the -na suffix in Old Norse-Icelandic, although deriving from different forms, they both function as general detransitivizers, marking the affectedness or the inactive nature of the subject, and are used as anticausativization devices. The r-form is the most general anticausativization strategy in Early Latin, whereas the functional domains of the na-forms in Old NorseIcelandic are more limited, unlike the sk-forms which comprise a substantially wider range of verbs and functions. A comparison of the distribution of anticausatives in Latin and Old Norse-Icelandic across aspectual verb classes is presented in Table 1. As to the semantic development, in Early and Classical Latin the Active Intransitive mainly occurs with gradual completion and activity verbs and is only marginally attested with accomplishments encoding a target, i.e. a reversible state. In contrast, the Active Intransitive does not occur in anticausative function with verbs lexically encoding a final or a result state, i.e., with achievements. The Reflexive Construction, however, occurs in Early and Classical Latin as a strategy for anticausativization, mainly with inherently telic predicates, i.e. with verbs lexically encoding a final, result or target state, i.e. achievements and accomplishments. The Reflexive Construction is also preferred when the subject, although inanimate, is personified, involving some degree of control. Finally, the Mediopassive Construction is the most general device of the three, occurring at all stages of Latin with all verb classes which allow the anticausative alternation: achievements, accomplishments and activities. With the last two aspectual classes, however, ambiguity may occur between an anticausative and a passive interpretation, although this is resolved in each case by the immediate context. Table 1: Early/Classical Latin and Old Norse-Icelandic compared. Achievements Accomplishments Activities Early/Classical Latin Mediopassive r-form + + + 1 Reflexive Se+active + + – Active Intransitive – +/? + Old Norse-Icelandic na-verbs Reflexive sk-verbs Nom. Intransitive Oblique Intransitive 1 2 + + – + + + –/?2 + – + – + States + – –/? ? + – + Accomplishments comprise verbs with different degrees of telicity. The reflexive pattern does not occur with gradual completion verbs and active accomplishments. Active intransitives with nominative subjects are mainly attested with gradual completion verbs. In Old Norse-Icelandic the main anticausativization strategies are the na-suffix, the Reflexive Construction, i.e. sk-verbs originating in reflexives, and the Active Intransitive with Oblique subjects. The anticausative function of the detransitivizing suffix -na, marking the lack of control, affectedness or the inactive nature of the subject, is quite common and is attested with achievements and accomplishments and to a limited degree with states, but not with activities. That is, it is not attested with 23 atelic dynamic predicates. The Reflexive Construction, in contrast, the most general anticausativization device, is compatible with all these aspectual classes. In the Oblique Active Intransitive the subject retains the accusative, dative or genitive case of the object of the original transitive construction, whilst the verb reverts to the third person singular impersonal form. This pattern is attested with achievements and accomplishments and is much more common than the Nominative Active Intransitive form, the labile strategy in Old Norse-Icelandic, which seems mainly to be attested with non-inherently telic verbs, e.g., gradual completion verbs. The nature of the subject is relevant as well here; the subject is never an agent, neither in the Nominative nor in the Oblique Intransitive anticausative construction. An Oblique Active Intransitive form emerges in Late Latin, but the verb never occurs in the impersonal form and the oblique subject is always in the accusative case. More research is needed, however, in order to uncover the relationship between the Late Latin oblique anticausative pattern, the oblique case-preserving anticausative strategy attested in Old Norse-Icelandic and in other early Indo-European languages, and the other types of oblique subject constructions attested in Old Saxon, Old Russian, Ancient Greek and Latin, mentioned in sections 3.2 and 4.1.5. To conclude, a close inspection of the distribution of the different anticausative strategies in Latin and Old Norse-Icelandic reveals that their occurrence reflects the interplay and the degree of integration of both components of a verb meaning, the structural aspect, i.e., its event structure template(s), and the meaning components it lexicalizes, i.e., the root, as well as the inherent and relational properties of the derived P-subject. 6. The rise and fall of anticausative constructions: Relative chronology As mentioned in Section 4.2 above, it has been proposed in the literature that the original Indo-European Mediopassive construction, the so-called synthetic passive, disappeared and its demise may have triggered the development of a new construction with similar functions, in this case the Reflexive Construction. Ottosson (2009) suggests this for the development in Germanic, pointing out that one of the functions of the synthetic passive was anticausative. As we have discussed above, the Latin r-form, which is the Romance descendent of the Indo-European Mediopassive, also exhibits an anticausative function. The converse hypothesis, namely that the emergence of a new mediopassive construction may have caused the original synthetic passive to retreat, is also sustainable and we would like to entertain that idea here. On such a view, first a new construction arises which is functionally similar to the old construction, at least in part, causing the old construction to go down in use and frequency. The new construction gradually acquires more of the functions of the old construction, until it has acquired most if not all of them. At that point, the old construction is simply not needed anymore, and reduces in frequency even further until it either vanishes altogether or lives on as a relic. The question now arises as to whether the two possible scenarios, discussed above, make different empirical predictions, so that they can be compared and tested scientifically. We believe that they do. It seems to us that if a new construction gradually takes over the function of an existing construction, one should find a certain degree of usage overlap between the two constructions during the period when the new construction has emerged and before the old one disappears. In contrast, if the 24 old construction disappears first, one would not expect to find the two constructions being used equivalently at the same time, but rather that the new construction neatly replaces the old one. In other words, there is never an overlap in time where the two constructions are both used, but rather two subsequent time periods where the two constructions are used complementarily, the old construction during the earlier time period and the new construction during the later time period. In an extreme case, one might perhaps expect a time interval where neither of the two constructions is used. That would be the ultimate proof that first the old construction disappears and then the new construction expands and takes over the functional domains of the old construction, as a need for a new construction has clearly arisen. A second prediction is that if a new construction causes an old multifunctional construction to retreat, one might expect the new construction to gradually take over the different functions of the old construction, before the old one disappears, and hence that the new construction mirrors the old one functionally. In other words, there is a one-to-one mapping between the old and the new constructions, as both will be equally multifunctional and occupy the same functional domain. At that point, the two constructions may change morphologically, as representing the functional merger. In contrast, if a new construction arises after the old one has disappeared, then there is no reason to assume that a functional overlap across the functions of the old multifunctional construction will be found with the new construction, but rather that different new constructions will take over different functions of the old construction. This will yield a lack of a one-to-one correspondence between the old and the new construction, and rather a situation where several different constructions have taken over different functions of the old multifunctional construction. For a language family with many branches, this might manifest itself in different constructions being found across the branches as having taken over the different functions of the earlier common construction. However, one would perhaps first and foremost expect this last scenario in cases where the modern daughter languages have already branched off, before the old construction disappears. Let us now consider the development of the different anticausative constructions discussed here in order to reveal whether the picture conforms to the two specific predictions laid out above. Observe, first, that there is an overlap between Latin and Germanic in that both exhibit a Reflexive construction and a Nominative Active Intransitive. The oddball out in Germanic is the na-construction, which is not found in Latin. A close scrutiny of the na-construction, however, suggests that it is not a systematic device for valence reduction, as is otherwise true for anticausatives, since na-verbs are also found alternating with intransitive verbs. This is shown in (24) above, repeated here for convenience: (24) vekja ‘waken’ – vaka ‘be awake’ – vakna ‘wake up’ The vocalism found with the three verbs, i.e. umlauted e vs. a, suggests that vakna ‘wake up’ is derived from the intransitive vaka ‘be awake’, rather than from the transitive vekja ‘waken’. This category of na-verbs is therefore better regarded as a detransitivizing device, denoting change of state and an inactive subject, rather than as an anticausative construction. Turning to P-lability proper, both Latin and Germanic exhibit Active Anticausative Constructions, although the diachronic relation between the Nominative and the Oblique Constructions is far from clear. Comparative evidence from Baltic and Slavic suggests that the Oblique Active Intransitive may be inherited. Whether 25 the Nominative Active Intransitive has developed out of the Oblique Active Intransitive, through Nominative Substitution, or whether the two co-existed side by side, needs further investigation. However, if the Active Intransitive, the Nominative or the Oblique, is inherited, then this construction may have competed with the IndoEuropean synthetic Mediopassive Construction already at an early stage of the IndoEuropean languages. The difference between the two, however, is that the synthetic Mediopassive Construction had more functions than only an anticausative one, like for instance the passive function, while the Active Anticausative Construction is functionally confined to anticausativization. Finally, let us consider the relation between the Indo-European synthetic Mediopassive and the Reflexive Construction. It is a known fact that the IndoEuropean synthetic Mediopassive disappeared and is replaced with a Reflexive Construction, both in Germanic and Italic. The question is therefore whether the IndoEuropean synthetic Mediopassive disappeared first and its demise trigged the emergence of the Reflexive Construction or whether the synthetic Mediopassive retreated because of the expansion of the Reflexive Construction. As we have shown in Section 3 above, the Latin reflex of the Indo-European synthetic Mediopassive Construction, the r-form, is clearly the preferred anticausative construction in Early and Classical Latin, while the Reflexive Construction is more restricted. Gradually, the Reflexive Construction expands to new aspectual classes of verbs, living side by side with the r-construction, which also has a passive function. It is, thus, not until after the Reflexive Construction has expanded into the functional space of the rconstruction (or at least into parts of its functional space), that the r-construction starts regressing and disappears in Romance. This co-occurs with the merging of the voice distinction in Romance and a general development to transitive syntax. Not only does the Reflexive Construction in Romance gradually take over the middle and the anticausative domain of the Indo-European synthetic Mediopassiv, but also its passive domain. The same development as in Latin, and from Latin to its Italic daughter languages, may also be assumed to have taken place in Germanic. As discussed in Section 4 above, the Indo-European synthetic Mediopassive Construction is only found in the earliest Germanic language, Gothic, and only in one of its tenses, i.e. present tense. Also, its function is clearly passive and not anticausative, i.e. the process is an induced one and not a spontaneous one (cf. Braune & Heidermanns 2004:141, 148, 156). (43) a. b. ik daupjada (Greek: ego: baptizomai) I am-baptized.R ‘I am baptized’ (Mark 10.38) sa reiks … uswairpada ut (Greek: ho arkho:n … ekble:the:setai exo:) the ruler thrown-away.R out ‘the ruler ... is thrown out’ (John 12.31) In the other tenses, a periphrastic construction is used in Gothic to denote an induced process, i.e. in the passive reading. This is the same periphrastic passive construction that is found in all the other modern Germanic languages of today. (44) daupiþs was fram Iohanne (Greek: ekbaptisthe: hupo Ioannou) baptized.NOM was from John ‘He was baptized by John’ (Mark 1.9) 26 The Reflexive sk-construction, however, has already in the earliest North Germanic acquired an anticausative function (Ottosson 2009). A Reflexive Construction is also found in Old High German, whereas in Old English, the language where reflexive pronouns have disappeared, the third common anticausative construction, the Active Anticausative Construction, has become the dominant construction within the domain of anticausativization (Kitazume 1996, Gelderen 2010). Whether this is due to influence from Old French, in which P-lability proper is quite common (cf. Heidinger this volume), is difficult to know at this stage, although an investigation of anticausative usages in texts before and after the Norman invasion should be telling in this respect. So to conclude, Germanic shows the same development as Latin, and the situation in Germanic can be regarded as a later point in a development that we witness as starting in Latin. A comparison with the remaining Indo-European branches is needed to decide on whether the development found in Italic and Germanic is common to more IndoEuropean branches or not. We know that the Indo-European synthetic Mediopassive has also disappeared in Baltic and Slavic and a Reflexive Construction has arisen. So to concretize, how do the two predictions, discussed above, fare in relation to the data from Latin and Germanic? First, it is clear that both constructions, the IndoEuropean Mediopassive and the Latin Reflexive Construction, are found at the same time in Latin texts, suggesting that it may in fact have been the Reflexive Construction that brought the Mediopassive to bay, and not that the Mediopassive Construction first disappeared and hence gave space to the Reflexive Construction. The second prediction relates to the functional overlap of the old and the new constructions, and in this case, we have seen that the Reflexive Construction gradually takes over the functions of the Indo-European Mediopassive, as in Early Latin it could only occur with achievements and some accomplishment verbs in anticausative function, gradually occurring with more aspectual verb classes, and then finally it also acquired a passive function. At that point in the development, the Reflexive Construction has acquired all the functional domains of the Indo-European synthetic Mediopassive, i.e. both its middle and also its passive domain. In fact, the more we ponder this issue, the more conceptual problems we have with the general idea that an older construction first retreats, and then through this regress, a need for a new construction arises, to fill up the functional space that is left empty by the old construction. It is entirely unclear to us, and appears as highly illogical, why a perfectly functional construction should start backtracking without a reason. As long as there is no evident reason for why such a development would take place to begin with, and as long as such a scenario makes predictions which are not borne out, we regard it as vanishingly unlikely. To conclude, the two predictions resulting from the two possible scenarios both clearly show that it was the functional expansion of the Reflexive Construction that made the Mediopassive Construction retreat. This takes place in correlation with the loss of voice distinctions in Late Latin and the development of transitivity during the period from Latin to Romance. Hence, our investigation here has shown that there is no doubt that this particular language change, the demise of the Indo-European synthetic Mediopassive Construction, is caused by the rise and expansion of the Reflexive Construction and not vice versa. 27 7. Summary and Conclusions A comparison between Latin and Old Norse-Icelandic has revealed the following similarities and differences in the evolution of anticausativity: Three different constructions in Latin serve as anticausative, i.e. the Mediopassive Construction, the Reflexive Construction and the Active Anticausative Construction. It is generally claimed in the literature on Latin that the three constructions are interchangeable. However, we have shown here that there is a clear division of labor between the three constructions, based on aspectual properties. First, the Mediopassive Construction is the most common in Early and Classical Latin and it can be instantiated by all verbs which qualify for the anticausative construal. The Active Anticausative Construction, in contrast, occurs mainly with activity verbs, while the Reflexive Construction is only found with causative accomplishments and achievements. Gradually, through the course of history, the Mediopassive Construction has given way to the two other constructions. Hence, the Reflexive construction starts occurring with gradual completion verbs and active accomplishments, as well as with activities and marginally with states. The Active Anticausative Construction is extended to gradual completion verbs and other types of accomplishments. In Old Norse-Icelandic, in contrast, there are four different ways anticausativity may be manifested, of which two are common with Latin. These four ways are through the verbal -na suffix, the Middle Construction with the original reflexive -sk suffix, the Nominative Active Anticausative Construction and the Oblique Active Anticausative Construction. The Nominative Active Construction seems to be very rare in Old Norse Icelandic, while both the Oblique Active constructions and na-verbs are more common. Most common, though, is the Middle Construction. The verbs occurring with the -na suffix are mostly achievements and accomplishments, while the ones occurring in the Middle Construction with -sk suffix and Oblique Active Anticausative are not aspectually restricted. All three constructions, with na-verbs, oblique subjects and the -sk suffix, can be argued to be Proto-Germanic. The diachrony of anticausativization in the languages investigated shows an interplay of the strategies employed for the anticausative alternation with more general changes taking place in the domain of transitivity, such as the rise of the Reflexive Passive in Late Latin; a comparable change happens in the history of Mainland Scandinavian, but is marginally attested in Old Norse-Icelandic. Further phenomena involving the reorganization of voice and grammatical relations may be seen in the sporadic emergence of nominative subjects (Nominative Substitution) with oblique subject predicates in Old Norse-Icelandic. The data also show that the aspectual template of verbs and the meaning components lexicalized in the verb, e.g. the notion of reversible and non reversible change in Latin, as well as the nature of the subject, e.g., its degree of control, play an important role in determining the choice of strategy in the languages examined, both synchronically and diachronically. Finally, the investigation carried out in the present article suggests that the demise of the Indo-European Mediopassive Construction was caused by the rise and expansion of the Reflexive Construction, and not vice versa. This occurrence is related to the loss of voice distinctions and the development of transitivity during the period from Latin to Romance. We have, moreover documented this evolutionary path from Early and Classical Latin to Late Latin, and the earliest Germanic data are also compatible with this scenario. The Indo-European Mediopassive is only found in Gothic, where it is confined to the present tense, and only occurs with a passive 28 reading and not an anticausative reading. The Middle Construction originating in the reflexive pronoun is clearly the main anticausative device in Old Norse-Icelandic, showing that Old Norse-Icelandic represents a later point on this evolutionary path than Latin. References Alexiadou, A. & E. Anagnostopoulou. 2004. Voice morphology in the causative– inchoative alternation: Evidence for a non-unified structural analysis of unaccusatives. In The Unaccusativity Puzzle: Explorations of the Syntax–Lexicon Interface, 123–159. Eds. A. Alexiadou, E. Anagnostopoulou & M. Everaert. Oxford: Oxford University Press. Anderson, S. 1990. The Grammar of Icelandic Verbs in -st. In J. Maling & A. Zaenen (eds.). Syntax and Semantics 24: Icelandic Syntax, 253–273. NY: Academic Press. Baldi, P. 1977. Remarks on the Latin R-form verbs. Zeitschrift für vergleichende Sprachforschung 90: 222–257. Barðdal, J. 2001a. The Perplexity of Dat-Nom Verbs in Icelandic. Nordic Journal of Linguistics 24:47–70. Barðdal, J. 2001b. Case in Icelandic – A Synchronic, Diachronic and a Comparative Appraoch. Lundastudier i Nordisk språkvetenskap A 57. Lund: Department of Scandinavian Languages. Barðdal, J. 2004. The Semantics of the Impersonal Construction in Icelandic, German and Faroese: Beyond Thematic Roles. In Focus on Germanic Typology, 105–37. Ed. Werner Abraham. Berlin: Akademie Verlag. Barðdal. J. 2008. Productivity: Evidence from Case and Argument Structure in Icelandic. Amsterdam: John Benjamins. Barðdal, J. 2009. The Development of Case in Germanic. In The Role of Semantic, Pragmatic and Discourse Factors in the Development of Case, 123–159. Eds. J. Barðdal & S.L. Chelliah. Amsterdam: John Benjamins. Barðdal, J. 2011a. Lexical vs. Structural Case: A False Dichotomy. To appear in Morphology 21(1). Barðdal, J. 2011b. The Rise of Dative Substitution in the History of Icelandic: A Diachronic Construction Grammar Approach. Lingua 1201(1): 60–79. Barðdal, J. & Th. Eythórsson. 2003. The Change that Never Happened: The Story of Oblique Subjects. Journal of Linguistics 39(3): 439–472. Barðdal, J. & Th. Eythórsson. 2009. The Origin of the Oblique Subject Construction: An Indo-European Comparison. In Grammatical Change in Indo-European Languages, 179–193. Eds. Vit Bubenik, John Hewson & Sarah Rose. Amsterdam: John Benjamins. Barðdal, J. & V. Molnár. 2003. The Passive in Icelandic – Compared to Mainland Scandinavian. In Structures of Focus and Grammatical Relations, 231–260. Eds. Jorunn Hetland and Valéria Molnár. Linguistische Arbeiten 477. Max Niemeyer Verlag, Tübingen. Bertinetto, P. & M. Squartini. 1995. An attempt at defining the class of ‘gradual completion verbs’. In Temporal Reference, Aspect and Actionality. Vol. I: Semantic and Syntactic Perspectives, ed. by P. Bertinetto, V. Bianchi, J. Higginbotham & M. Squartini, 11–26. Torino: Rosenberg & Sellier. Braune, W. & F. Heidermanns. 2004. Gotische Grammatik. 20th ed. Tübingen: Max Niemeyer. 29 Cennamo, M. 1998. The loss of the voice dimension between Late Latin and Early Romance. In Historical Linguistics 1997, ed. by M. S. Schmidt, J.R. Austin & D. Stein, 77–107. Amsterdam: Benjamins. Cennamo, M. 2001a. Classi verbali e cambiamento sintattico: la reinterpretazione passiva del costrutto riflessive. In Semantica e Lessicologia Storiche, Z. Fábián & G. Salvi (eds.), 225–242. Rome: Bulzoni. Cennamo, M. 2001b. On the reorganization of voice distinctions and grammatical relations in Late Latin. In Actes du XemeColloque Internationale de Linguistique Latine, 51–65. Paris: Peeters. Cennamo, M. 2005. Passive auxiliaries in Late Latin. In S. Kiss & G. Salvi (eds). Festschrift Herman, Tubingen: Niemeyer, 179–196. Cennamo, M. 2006. The rise and grammaticalization pats of Latin fieri and facere as passive auxiliaries. In Passivization and Typology, ed. by W. Abraham & L. Leiso, 311–336. Amsterdam: Benjamins. Cennamo, M. 2008. The rise and development of analytic perfects in Italo-Romance. In Grammatical Change and Linguistic Theory: the Rosendal Papers, Th. Eythórsson (ed), 115–142. Amsterdam: Benjamins. Cennamo, M. 2009. Argument structure and alignment variations and changes in Late Latin. In The Role of Semantic, Pragmatic, and Discourse Factors in the Development of Case, J. Barddal & S.L. Chelliah (eds), 307–346. Amsterdam: Benjamins. Cennamo, M. 2010. Control and argument marking in Latin. A paper presented at the Workshop Subject and transitivity in Indo-European and beyond: a diachronic typological perspective. 43rd Annual Meeting of the Societas Linguistica Europaea, Vilnius, 2–3 September 2010. Cennamo, M. forthcoming. Impersonal constructions and accusative subjects in Late Latin. To appear in A. Malchukov & A. Siewierska (eds). Impersonal constructions: A crosslinguistic perspective. Amsterdam: Benjamins. Cennamo, M. & E. Jezek. 2009. The anticausative alternation in Italian: constraints and variation. A paper presented at the Workshop on Verb Typologies Revisited: A Cross-linguistic Reflection on Verbs and Verb Classes, Gent, 5–7 February 2009. Comrie, B. 1989. Language Universals and Linguistic Typology, 2nd ed. London: Blackwell. Dewey, Tonya Kim & Carlee Arnett. 2010. Oblique Subject Predicates in Old Saxon. Ms., University of California, Berkeley & University of California, Davis. Dixon, R.M.W. 1994. Ergativity. Cambridge: Cambridge University Press. Eythórsson, T., 2000a. Dative vs. nominative: changes in quirky subjects in Icelandic. Leeds Working Papers in Linguistics 8, 27–44. Eythórsson, T., 2000b. Fall á fallanda fæti [Unstable case marking]. Íslenskt mál og almenn málfræði 22: 185–204. Eythórsson, T., 2002. Changes in subject case marking in Icelandic. In: Lightfoot, D. (Ed.), Syntactic Effects of Morphological Change. Oxford University Press, Oxford, pp. 196–212. Eythórsson, Th. & J. Barðdal. 2005. Oblique Subjects: A Common Germanic Inheritance. Language 81(4): 824–881. Feltenius, L. 1977. Intransitivizations in Latin. Uppsala: Almqvist & Wiksell. Flobert, P. 1975. Les verbes déponents latins des Origines à Charlemagne. Paris: Les Belles Lettres. 30 Folli, R. 2002. Constructing Telicity in English and Italian, PhD thesis. Oxford: University of Oxford. Gelderen, Elly van. 2010. Valency Changes in the History of English. To appear in Journal of Historical Linguistics. Hale, K.L. & S.J. Keyser. 1997. The limits of argument structure. In A. Mendikoetexea & M. Uribe-Etxebarria (eds). Theoretical Issues at the Morphology-Syntax Interface. Euskal Herriko Universitaetea, Bilbao, 203–230. Haspelmath, M. 1987. Transitivity alternations of the anticausative type. Universität zu Köln (Arbeitspapiere N.F. 5). Haspelmath, M. 1993. More on the typology of inchoative/causative verb alternations. In Causatives and transitivity, ed. by B. Comrie & M. Polinsky, 87–120. Amsterdam: Benjamins. Hay, J., C. Kennedy & B. Levin. 1999. Scalar structure underlies telicity in “degree achievements.” In T. Matthews & D. Strolovitch (eds.), Proceedings of SALT IX, 127–144. Ithaca, NY: CLC Publications. Herman, J. 2002. La disparition du passif synthétique latin: nouvel essai sur l’écrit et le parlé en latin mérovingien. Esstudios Romanicos 24: 31–46. Kainhofer, J.M. 2002. Monadishce Ackusativ-Subjekt-konstruktionen im Isländischem. Magister Thesis, University of Salzbrug. Kitazume, Sachiko. 1996. Middles in English. Word 47(2): 161–183. Koontz-Garboden, A. 2009. Anticausativization, Natural Language & Linguistic Theory 27: 77–138. Kulikov, L. 2001. The vedic -ya Presents. Ph.D. Dissertation, Leiden University. Kulikov, L. 2003. The labile syntactic type in a diachronic perspective: the case of Vedic. SKY Journal of Linguistics 16: 93–112. Kurzová, H. 1993. From Indo-European to Latin: The Evolution of a Morphosyntactic Type. Amsterdam: John Benjamins. Lazzeroni, R. 2004. Inaccusatività indoeuropea e alternanza causativa vedica. Archivio Glottologico Italiano 89: 1–28. Lazzeroni, R. 2009. Causativi e transitivi indoeuropei: fra comparazione e tipologia. Studi e Saggi Linguistici 57: 7–23. Letuchiy, A. 2009. Towards a typology of labile verbs: Lability vs. derivation. In P. Epps & A. Arkhipov (eds.). New Challenges in Typology: Transcending the Borders and Refining the Distinctions. Berlin: Mouton de Gruyter, 247–268. Levin, B. & Rappaport Hovav, M. 1995. Unaccusativity. Cambridge: MIT Press. Levin, B. and Rappaport Hovav, M. 2005. Argument Realization. Cambridge: Cambridge University Press. Norberg, D. 1943. Syntaktische Forschungen: auf dem Gebiete des Spätlateins und des Frühen Mittellateins. Uppsala: Almqvist & Wiksell. Ottósson, K. 1992. The Icelandic Middle Voice: The Morphological and Phonological Development. Lund University Doctoral Dissertation. Lund: Department of Scandinavian Languages. Ottosson, K. 2008. The Old Nordic Middle Voice in the pre-literary period: Questions of grammaticalisation and cliticisation. In Josephson, Folke & Ingmar Söhrman (eds.), Interdependence of Diachronic and Synchronic Analyses, 185–219. Amsterdam: Benjamins. Ottosson, K. 2009. The anticausative and related categories in the Old Germanic languages. Ms. University of Oslo. Parsons, T. 1990. Events in the Semantics of English. Cambridge MA: MIT Press. 31 Pirson, J. 1906. Mulomedicina Chironis. La syntaxe du verbe. In Festschrift zum Allgemeinen Deutschen Neuphilologentage, Erlangen, 390–431. Rappaport Hovav, M. & B. Levin. 2010. Reflections on Manner/Result complementarity, in M. Rappaport Hovav, E. Doron & I. Sichel (eds), Syntax, Lexical Semantics and Event Structure, Oxford: Oxford University Press, 21–38. Ringe, Donald. 2006. From Proto-Indo-European to Proto-Germanic: A linguistic history of English 1. Oxford: Oxford University Press. Seefranz-Montag, A.v. 1983. Syntaktische Funktionen und Wortstellungsveränderung: die Entwicklung “subjektloser” Konstruktionen in einigen Sprachen. Munich: Wilhelm Fink Verlag. Seefranz-Montag, A.v 1984. ‘Subjectless’ constructions and syntactic change. In Fisiak, J. (ed.). Historical Syntax. Berlin: Mouton, 521–553. Seržant, Ilja. 2010. Dative Subject Predicate Candidates in Ancient Greek. Ms., University of Bergen. Siewierska, A. 1984. The Passive: A Comparative Linguistic Analysis. London: Croom Helm. Smith, H. 1994. Dative Sickness in Germanic. Natural Language & Linguistic Theory 12: 675–736. Smith, H. 1996. Restrictiveness in Case Theory. Cambridge: Cambridge University Press. Smitherman, Thomas. 2010. Dative Subject Predicates in Old Russian 2.1. Ms. University of Bergen. Suzuki, Seiichi. 1989. The Morphosyntax of Detransitive Suffixes -þ- and -n- in Gothic: A Synchronic and Diachronic Study. New York: Peter Lang. Svennung, J. 1935. Untersuchungen zu Palladius und zur lateinischen Fach-undVolksprache. Uppsala: Almqvist & Wiksell. Van Valin, R.D. 2005. Exploring the Syntax–Semantics Interface. Cambridge: Cambridge University Press. Van Valin. R.D & D.P. Wilkins 1996. The case for ‘effector’: case roles, agents and agency revisited. In M. Shibatani & S. Thompson (eds), Grammatical Constructions, 289–322, Oxford: Oxford University Press.