Toshiko Yamaguchi CDU 803.959:801.56 National University of Singapore IMPERSONAL PASSIVES IN ICELANDIC 1 This paper argues that /celandic impersonal passives have a constructional property that express­es generic events. After having demonstrated their basic characteristics, it is shown that the notion of 'activity' is inadequate for the explanation ofwhy the construction is licensed. The behaviour ofthe adverbial ser moovitandi is diagnosed to show indefiniteness tied up with the construction. l. Introduction Impersonal passives are highly productive grammatical constructions iri modem Ice­landic. They typically appear with a past participle form of intransitive as well as transi­tive verbs, often containing an expletive pao 'it, there' at the initial position ofa sentence. 2 (1) l>a6 var synt. it was swum 'There was swimming' (2) l>a6 var skrifa6. it was written 'People wrote' This construction is called impersonal because, syntactically speaking, it does not have a nominal govemed by the base verb (see Klaiman 1991: 6, 8). The construction has received much attention in the literature of syntax and semantics particularly on Germanic languages such as Dutch and German (e.g. Perlmutter 1978; Kathol 1994; Pollard 1994; Zaenen 1993). In Icelandic too some research, though less extensively, has been put forward by severa! scholars (Zaenen and Maling 1990; Smith 1993; Van Valin 1991). A common idea in the previous discussion is that the applicability of im­personal passives is accounted for by means of the lexical property of a base verb. Following Perlmutter's original distinction between unergative and unaccusative, it is stated that impersonal passives are formed by unergative verbs, those verbs that are lexically characterized as having an extemal argument and, hence, an agentive subject. After it became obvious that the 'Unaccusative Hypothesis' does not hold up well in many languages (e.g. Levin and Rappaport 1989), researchers paid close attention 1 would like to thank Magmis Petursson who provided me with the data and took the tirne to discuss tricky areas ofthe Icelandic impersonal passive. Without his help and generosity this paper would not ha ve taken this form. 1 thank Peter Sells who read an earlier version of this paper and made helpful suggestions. Errors are exclusively mine. In this paper we mainly deal with intransitive verbs and treat transitive verbs only marginally. The exact treat­ment ofthis construction with transitives is beyond the scope ofthis paper (see footnote 5). to impersonal passives on the basis of different semantic criteria. Dowty (1991 :607ff) appeals to the idea ofbifurcation of intransitive verbs by considering the semantic ele­ment 'volition' as a salient component of agentivity; the presence or absence of voli­ti on for the triggering of the action determines the classification of intransitive verbs. Preserving the notions, agentive verbs come to be considered unergative, while non­agentive verbs are considered unaccusative. Thus, the ungrammaticality of the sen­tence in (3) is explained by the fact that sokkva 'sink' is an unaccusative verb in that volition is definitely not a relevant factor for an entity such as a ship to sink. (3) *l>ao var sokkio. it was sunk 'Sinking took place' In contrast to (3), the reason that (1) is grammatical is that the activity of swimming necessarily involves volition or, more precisely, the swimmer has a will to do it; hence, synda is classified as unergative. In effect, Perlmutter's original proposal finds expres­sion in Dowty in that verbs which undergo impersonal passivisation are felicitous with an agentive or volitional interpretation. As shown in (4), the impersonal passive in Dutch arises from an unaccusative verb stinken 'stink' only when the agent's volition is emphasised in such a way that the woman is interpreted as intentionally exuding the bad odours (example cited from Zaenen 1993: 139, (37); cited as well in Levin and Rappaport Hovav 1987). (4) Er werd door de krengen gestonken. 'There is a stink from the nasty woman' Although Zaenen (1993: 136) admits the presence ofthe component ofvolition in Dutch, she further argues (ibid 138) that the whole construction of impersonal passive encodes atelicity that is not determined by the lexical property of the verb. According to her, the acceptability of the impersonal passive ultimately depends on the aspect of the sentence as a whole but not purely on the lexical property of the base verb. As exemplified by the contrast in (5) and (6), the Dutch verb /open 'run' is only compat­ible with impersonal passive when it is atelic, whereby the element of volition appar­ently plays no relevant role (examples cited from Zaenen 1993: 138, (32) and (33)). (5) Er werd gelopen. 'There is running' (6) *Er werd naar huis gelopen. 'There is running home' In this paper, I will present an analysis of Icelandic impersonal passives that di­verges in many respects from Dutch. Although for some researchers the unerga­tive/unaccusative distinction is taken for granted in Icelandic (see Smith 1993, for example ), 3 Icelandic does not provide sufficient syntactic grounds for this statement (Yamaguchi in press). It does not, as will be discussed below, exhibit semantic unac­cusativity either. Neither does aspect play a relevant role. It will be shown that Icelandic impersonal passives are sensitive not merely to the lexical semantics of the base verb but also to the existence of the impersonal passive construction that expresses a 'generic event', an event that people or the members of a given circumstance generally take part in. 4 lmpersonal passives in Icelandic do not allow an adverbial expression ser meovitandi 'consciously, be aware of', for this ad­verbial assigns the component of definiteness to the sentence. This mismatch is self­explanatory in that generic events do not describe specific or individual, hence, defi­nite episodes with which the meaning ofthis adverbial might be felicitous. The reason volition or aspect does not play a part in Icelandic is that these components are, in effect, insensitive to the notion of (in)definiteness. The organisation of the paper is as follows. In section 2, 1 demonstrate some basic facts relevant to our ongoing discussion. In section 3, 1 review critically Van Valin's ( 1991) analysis that approaches Icelandic impersonal passives from a different seman­tic perspective. 1 argue that his proposal that the semantic primitive 'activity' is a deci­sive factor in the formati on of lcelandic impersonal passives is not on the right track. In section 3, 1 show how my proposal can explicitly account for the applicability ofimper­sonal passives in Icelandic. Finally in section 4, while summing up the findings, 1 refer to remaining problems which cannot be fully understood by the present proposal. 2. Some Basic Facts 2.1. Expletive pao As briefly referred to at the outset of this paper, lcelandic impersonal passives are formed by an expletive or a placeholder pao followed by an auxiliary vera 'be' and a past participle of the verb. Note that this past participle is always in a neuter and sin­gular form. As illustrated in (7)b and (7)c, pao does not appear in the sentence, there­by the past participle remains neuter and singular and there is no change in meaning among these variants. 5 3 Smith (1993: 480-481) states that verbs such as koma 'come',fara upp 'go up',fara ofan 'go down' lioa 'pass' are unaccusative verbs with a theme subject just like sčikkva 'sink', hence not forming impersonal passives. However, he is wrong in this statement because these verbs are uncontroversially good in impersonal passives; the themehood of subject cannot thus be a criterion for unaccusativity. 4 I follow the general idea presented in the versions of Construction Grammar that grammatical constructions represent pairings of form and meaning that are present independently ofthe base verbs occuring in them ( e.g. Goldberg 1995, Kay and Fillmore 1999). However, the detailed theoretical criticism of the nature of gram­matical constructions is beyond the scope of the paper. 5 In Icelandic some transitive verbs appear in the expletive construction. There are two types. The first type is shown in (i) where the passive form ofthe verb skrifa 'write' agrees with the noun brekur 'books' in gender (7) a. I>a6 er hlaupi6. it is run (NEU.SING) 'People are running' b. i porpinu er hlaupi6. in village.the is run (NEU.SING) 'In the village people are running' C. i grer var hlaupi6. in yesterday was run (NEU.SING) 'People ran yesterday' 2.2. A.f-phrase In Icelandic the agent is expressed by the phrase headed by a preposition af'by' in both passive constructions. In the personal passive construction the active agent is defocused and can appear in the afphrase, as in (8). 6 By contrast, impersonal passives do not usually express the agent in the afphrase, as in (9). As indicated by (10), how­ever, impersonal passives can allow the afphrase only when the agent refers to people in general or a collective number of people or things. In many studies in the past the subject of the active sentence is considered to be suppressed, defocused, or demoted structurally to give rise to the passive altemant (cf. Comrie 1977; Shibatani 1985). Examples below represent a case that impersonal passivisation cannot be explained fully by reference to this type of theoretical consideration, but it is shown to involve much more subtle aspects which are, in effect, semantic in nature. In descriptive terms, Icelandic impersonal passives function as 'impersonalisation' of a participant in the and number. The agent can appear in the afphrase, as in personal passives in (ii), but it does not necessarily do so in a particular context in which the speaker puts more emphasis on the generic nature of the activity of writing, which cannot hold for personal passives. (i) l>a3 eru skrifaessar brekur eru skrifaa3 er skrifa3 (*af Joni). it is written (NEU.SING) by John 'Someone wrote' It appears that the expletive constructions with transitive verbs ofthe first type might be said to lie somewhere between personal passives and impersonal passives in morphosyntactic and semantic terms, and the second type is almost identical with impersonal passives with intransitives. 6 The term 'defocused' here refers to the mention of an agent in a non-prominent way in the sentence. designated event. Because of this semantics, only the agents who are not specified in that event are able to appear in the passive, as indicated by (1O)7. (8) Dansinn var dansaour (afMariu). dance was danced (by Maria) 'The dance was danced by Maria' (9) l>ao var dansao (*afMariu). it was danced (by Maria) 'There was dancing' (10) l>ao er dokao vio (*afJ6ni/afpeim/aff61kinu). it is stayed (by John/by them/by people) 'People stayed' 2.3. Aspect We noted that Dutch impersonal passives are sensitive to atelic aspect (section 1). However, 1 will show here that Icelandic impersonal passives are insensitive to aspect; they are available with or without the directional phrase. 8 ( 11) a. l>ao er flutt. it 1s moved 'People are moving' b. l>ao er flutt til islands. it is moved to Iceland 'People are moving to Iceland' (12) a. l>ao er komio. it 1s come 'People are coming' b. l>ao er komio til Mssins. it 1s come to house.the 'People are coming home' 7 As 1 will mention in section 4, there are verbs which do not accept an overt expression of an a.f-phrase at ali, for which 1 do not have an explanation at presen!. 8 German behaves, in some aspects, more similarly to Dutch. The verb ankommen 'arrive' is not compatible with impersonal passives when it co-occurs with a directional phrase, as exemplified by (i) and (ii) . (i) Es wurde angekommen. (attested) it was arrived 'There was an arrival' (ii) *Es wurde angekommen nach Hause. it was arrived to house 'There was an arrival to the house' In Yamaguchi (2000), it is claimed that the availability of the auxiliary vera 'be' is sensitive to the component of directionality if it exhibits a perfective reading, i.e. verbs such asjlakka 'wander' take vera only when a direction is overtly expressed. However, impersonal passives differ significantly from perfective auxiliary constructions, as seen in (13), in that the selection ofvera is not affected by directionality, as seen in (14). The different forms ofpast participles in (13) and (14) are due to the fact that they agree with subject in gender and number when they express perfectivity ( section 2.1 ). (13) Perfective: a. *Hans er flakka6ur. Hans is wandered b. Hans er flakka6ur til stMvarinnar. Hans is wandered (MASC.SING) to station.the 'Hans (is) wandered to the station (and he may be now at the station)' (14) Impersonal Passive: a. l>a6 er flakka6. it is wandered (NEU.SING) 'People are wandering' b. l>a6 er flakka6 til stMvarinnar. it is wandered to station.the 'People are wandering to the station' 2.4. Volition Volition can be expressed linguistically by adverbials such as, viljandi 'intentional­ly, knowingly', afasettu raoi 'intentionally', or i peim tilgangi 'on purpose' in Ice­landic.9 Although it is stated in terms of Dutch that verbs which co-occur with opzet­telijk 'on purpose' are good in impersonal passives (Zaenen 1993: 133), Icelandic verbs cannot be subcategorised by this criterion. Consider verbs such as synda 'swim' and rulla 'roll' which behave in exactly the same manner with respect to these adver­bials but they differ in terms of impersonal passivisation; the former permits it, while the latter does not. (15) a. l>a3 var synt. it was swum 'People swam' 9 It is in fact difficult to give exact English glosses which clearly distinguish between viljandi and afasettu raoi. Generally speaking, afasettu radi is more frequently used and expresses a stronger and established intention. A verb likefremja sjalftmoro 'commit suicide' allows only the latter, because ifone commits suicide, he/she has an established intention, i.e. that what he/she does is expected to lead him/her to death. b. Hans syndir viljandi. af asettu rMi i peim tilgangi (16) a. *I>a5 it var was rulla5. rolled b. Hans n.llla5i viljandi. af asettu rMi i peim tilgangi Verbs which behave analogously to synda and mila are listed below. (17) synda fara 'go' ganga 'walk' h/aupa 'run' hr6pa 'cry' kenna 'teach' koma 'come' vaka 'be awake' ritlla birtast 'aooear' bogra 'crawl' deyfa 'make dim' hverfa 'disappear' lykta 'smell' vakna 'become awake' ve/ta 'tumble' The irrelevance of volitionality for the formation of impersonal passives in Icelandic is also obvious with respect to the verb lykta 'smell' whose behaviour is not influenced by the pragmatic context, as we noted with regard to example (4) above. Lykta does not form an impersonal passive under any circumstances. (18) *I>a5 er illa lykta5. it is badly smelt 'There is a stink' 3. Van Valin (1991) Van Valin ( 1991) in his investigation of Icelandic verbs with Role and Reference Grammar (henceforth, RRG) framework claims that impersonal passives are available when verbs encode an activity predicate in their logical structure (ibid 190). He takes an adverb kroftuglega 'vigorously, energetically', which expresses an action, as a test to show the existence of activity as well as the applicability of impersonal passives in Icelandic. In this section 1 shall try to show by providing ample a3 var skolfi3. it was shivered 'There was shivering' However, one problem arises when we look at a near-synonymous verb titra 'shiver' which is compatible with kroftuglega and behaves almost similarly to skjalfa in active sentences, as in (20a), although it is bad in impersonal passive, as in (20b ). We come to recognise that this contrasting behaviour of these two verbs can hardly back up Van Valin's claim. (20) a. Bami3 titra3i kroftuglega/?viljandi/*af asettu ra3i/*i peim tilgangi. child shivered vigorously/intentionally 'The child shivered vigorously' b. *I>a3 var titra3. it was shivered 'There was shivering' One salient difference between skjalfa and titra is semantic. The activity of shiver­ing encoded in skdlfa is caused by natura! phenomena like coldness due to a low tem­perature, for instance, whereas the shivering encoded in litra is caused by interna! human conditions such as fear or excitement. The contrast in (21) illustrates the case in point; titra does not allow kulda 'coldness' asa causal element. (21) a. Bami3 skalf vegna kulda. child shivered because of coldness 'The child shivered because of the coldness' b. Bami3 titra3i vegna *kulda/af resingu. child shivered because of coldness/by excitement 'The child shivered/shook because of the coldness/excitement Van Valin gives a further example for the legitimacy of the semantic relevance of activity. For instance, an expression vera heima 'stay/remain at home' permits an impersonal passive, as shown in (22), and this expression isto be an activity predicate in spite of the presence of the copula vera 'be', which signals, as a rule, the stativity of a predicate. He provides, for instance, a sentence like (23) as a diagnostic test for the validity of the activity component. (22) l>a6 var rniki6 veri6 beirna a kvoldin. it was rnucb been at.borne in the.evening 'People rernained borne a lot in the evening' (Van Valin 1991: 187, (58b)) (23) F6lk var ekki beirna af asettu rMi a kvoldin. people was not at borne intentionally in tbe evening 'People intentionally did not rernain at borne in tbe evening' (see Van Valin 1991: 188 (59a); sligbtly rnodified by tbe autbor) Note, however, that (23) co-occurs with afcisettu rcioi, being a signal for the pres­ence of volition, but not for that of activity. It is striking that, as in (24), kroftuglega, being a signal for the presence of activity, is infelicitous and this empirical fact con­tradicts his own argument, i.e. that the compatibility ofafcisettu rcioi does indicate that impersonal passivisation is possible even though verbs are not necessarily classified as activity predicates. (24) *F6lk var ekki beirna kroftuglega a kvoldin. people was not at horne vigorously in the evening In fact, it appears that it is almost impossible to single out the activity component with respect to the availability of impersonal passives. First, there are weather verbs which disallow volitional components as modification but only allow an action com­ponent and, interestingly, they do not permit an impersonal passive at ali. (26) lists other kinds of weather verbs which bebave in exactly the same way. (25) a. l>a6 rignir kroftuglega/*viljandi. it rains vigorously/intentionally 'It is raining bard' b. *l>a6 er rignt. it is rained (26) blasa 'blow', frj6sa 'freeze', skina 'sbine', snj6a 'snow', ... One might argue that the inapplicability of impersonal passives is brought about by the inanimacy linked to the weather verbs. For instance, Van Valin claims (1991: 186) that the Icelandic impersonal construction "is restricted to intransitive verbs which take animate actors". However, there are passivisable verbs such asfalla 'fall',fara 'go', gr6a 'grow', koma 'come', sigla 'sail' which assign what he calls undergoer to their single argument. This is exemplified by the examples (27) and (28), where falla and sigla are compatible with animate and inanimate subjects and absolutely good in im­personal passive. That is to say, the sentence Pao var fallio in (27)b, for instance, can refer either to animate ( e.g. an athlete) or inanimate entities ( e.g. signposts on the street) which can fall. In this respect, it might not be correct to make a generalisation that impersonal passives in Icelandic are restricted to intransitives only with animate actors. (27) a. Skiltio/Ipr6ttamaourinn fellur. signpost/athlete fell 'The signpost/athlete fell' b. l>ao var fallio. it was fallen 'Falling took place' (28) a. Maourinn/Skipio sigldi. Man/ship sailed 'The man/ship sailed' b. l>ao var siglt. it was sailed 'There was sailing' There are verbs apart from weather verbs that permit krofluglega but do not form an impersonal passive. One such verb is detta 'fall'. Although it is a near synonym to falla, it does not permit an inanimate entity as subject. The reason is that the meaning of detta involves human motivation, i.e. actions conducted by humans, not necessari­ly intentional, which brings a fall about (e.g. a mistake by an athlete). (29) a. Ipr6ttamaourinn dettur kroftuglega. athlete fell vigorously 'The athlete fell vigorously' b. *l>ao var dottio. it was fallen There are also verbs such as hvilast 'rest' which bebave in a reversed manner; they permit an impersonal passive although they are not compatible with kroftuglega. (30) a. *Maria Maria hvildist rested kroftuglega. vigorously b. l>ao var hvilst. it was rested 'People rested' It follows that the isolation of the component 'activity' is not a relevant criterion for the availability of impersonal passives contra Van Valin. A serious problem asso­ciated with his analysis might be that he deals with restricted ao var vitao. it was known 'People knew' (35) a. l>rostur pekkti *kroftuglega/*viljandi. l>rostur knew vigorously/intentionally b. *l>rostur var ao pekkja. l>rostur was at know 'l>rostur was knowing' C. I>ao var pekkt. it was known 'People knew' Ifthe stativity, which is treated as another primitive in RRG, should play a role, as Van Valin claims, the behaviour of vita and jJekkja would be highly controversial. However, the ao var almennt/*pers6nulega flogio it was generally/personally flown C. l>ao var flogio (*af Hans/ af f6lkinu). it was flown (by Hans/by people) Verbs forming impersonal passives are, in fact, large in number in lcelandic. The following two lists give a bird's eye view ofthe distribution ofverbs that allow or dis­allow an impersonal passive. Although it is beyond the scope of this paper to answer the question ofwhy verbs in (38) disallow the formati on ofimpersonal passives, a ten­tative suggestion might be that events designated by these verbs are considered not to happen to people in general; but it suffices to say for the purpose ofthis study that the conventional meanings ofthese verbs might at best concem events that happen to spe­cific individuals or entities (cf. section 4.2). (37) Verbs forming impersonal passives: aka 'drive', berjast 'fight', dvelja 'stay', elska 'love',fal/a 'fall',jljuga 'fly',jlytja 'move', fremja sjalfsmoro 'commit suicide', dansa 'dance',fara 'go, ganga 'walk', geispa 'yawn', giftast 'marry', gr6a 'grow', ha/da 'think, consider', hja/pa 'help', hlaupa 'run'' harfa a'look at'' hlusta (a) 'listen to'' hr6pa 'cry'' hugsa 'think'' hvilast 'rest', kenna 'teach', koma 'come, arrive', komafram 'appear, come into being', /rera 'leam', minnast 'recall', 6ska 'wish', sakna 'miss', sigla 'sail', sitja 'sit', sja 'see', skjalfa 'shiver', snokta 'sob', synda 'swim', syngja 'sing', vaka 'be awake', vita 'know', yfirvega 'think over', pekkja 'know', pvo 'wash', ... (38) Verbs not forming impersonal passives: birtast 'appear', blatna 'become wet', bagra 'stoop, crawl', brotna 'break', detla 'fall', deyja 'die', deyfa 'make dim', dofna 'become weak', dropa 'drop, leak', heyra 'hear', hverfa 'disappear', kafna 'suffocate', teka 'leak', lika 'like', lykta 'smell', neita 'refuse', ril/la 'roll', sofna 'fall asleep', sokkva 'sink', litra 'shiver', vakna 'become awake', ve/ta 'tumble', ... 4.2. Indefiniteness In the previous section 1 claimed that Icelandic impersonal passives express gener­ic events. This section provides further evidence for its validity by referring to the adverbial expression ser meovitandi 'consciously, knowingly' which is claimed to ex­press definiteness. IO 1O The expression ser meovitandi is a compound expression ofa dative reflexive pronoun ser for third person and an adjective meovitandi, a derived form of the noun meovitund 'consciousness', arising originally from the verb vita 'know' to which the preposition meo 'with' is concatenated. Let US first consider the behaviour ofthis adverbial expression. The reflexive ser usu­ally means 'for the sake of' or 'for the advantage of' and is used to supplement this semantic element to verbs that appear as transitive. 11 This is the reason why (39)b is ungrammatical. Consider (39)c in which the appearance of ser stresses Magga's inten­tional action. Therefore, (39)e is ungrammatical because an inanimate entity such as a ship cannot have an intention. Thus, (39)a differs from (39)c in that Magga in (39)a sank independently of her will. Although ser has something to do with intention or will, this specific meaning fades away when it occurs in the combination ofmeovitandi, as in (39)f. (39) a. Magga sokk. Magga sank 'Magga sank' b. *Magga sokk ser. Magga sank self C. Magga sokkti ser. Magga sank self(DAT) 'Magga sank' d. Skipi6 sokk. ship sank 'The ship sank' e. *Skipi6 sokkti ser. ship sank self f. Magga sokkti ser me6vitandi. Magga sank consciously 'Magga was aware of sinking' Ser meovitandi can also co-occur with verbs, normally not taking ser on their own. As demonstrated by the verb detta 'fall' in (40), the element 'for the sake of' is not required by detta. Together with (39)f, ser in ser meovitandi is not exactly the same thing as a normal reflexive ser; hence, it might be fair to say that this complex expres­sion is semantically a unified construct that independently expresses the semantics of consciousness or awareness. (40) a. *Hans datt ser. Hans fell self b. Hans datt ser me6vitandi. Hans fell consciously 'Hans was aware offalling' 11 Certain verbs such as sleppa 'escape, release', sokkva 'sink', stokkva 'jump', sme/la 'snap, bang', ve/ta 'roli' have two distinct past tense forms. As seen in (39), the verb sokkva bas sokk for intransitive and sokkti for tran­sitive verbs. My contention is that the behaviour ofser meovitandi accounts far the applicabili­ty oflcelandic impersonal passives that express genericity. It is a well-known fact that the meaning of genericity is often conveyed by indefinite expressions ( e.g. Krifka et al 1995). Far instance, bare noun phrases in Icelandic are often responsible far the expression of generic sentences. Bj6rar byggja stiflur. beavers build dams 'Beavers build dams' Consider the contrasting examples in ( 42) and ( 43) where verbs that form an imper­sonal passive are not compatible with ser meovitandi, while this adverbial is felicitous when the verb does not form an impersonal passive. This is somewhat surprising at first sight since both hlaupa 'run' and mila 'roll' are semantically close to each other, in the sense that they encode a motion component. (42) a. I>ao er hlaupio. it is run 'People are running' b. *Magnus Magnus hleypur ser meovitandi. run s consciously (43) a. *I>ao er It is 'People are rolling' rullao. rolled b. Magnus Magnus rullar rolls ser meovitandi. consciously In comparison, as illustrated in (44) and (45), these verbs uniformly accept adver­bials such as viljandi 'intentionally' or afasettu raoi 'intentionally' (see also section 2.4), indicating that the component of voliti on does not provide an explanation ofwhy (42) and (43) bebave differently. (44) Magnus hleypur viljandi/af asettu rao. Magnus run s intentionally (45) Magnus rullar viljandi/af asettu rao. Magnus rolls intentionally Consider the examples in ( 46) and ( 47) where definite and indefinite interpretations neatly distinguish ser meovitandi from viljandi and afasettu raoi; the former co-occurs with the definite noun phrase, while the latter can appear irrespective of the definite­ness distinction ((46)a and (47)a)). An interesting fact that further confirms our obser­vation is that relative clauses serve to impart a definite property, that is, ser meovitan­di is acceptable whenf6lk 'people' is a head noun modified by the relative clause head­ed by sem 'who', as shown in ( 4 7)b. (46) Indefinite: a. *F6lk syngur ser meovitandi. People sing consciously b. F6lk syngur viljandi/ afasettu raoi People sing intentionally 'People sing intentionally' (47) Definite: a. F6lkio 'People.the syngur ser meovitandi. sing consciously' b. F6lk, sem er rauoklrett, People who are dressed in red 'People who are dressed in red sing cons syngur sing ciously' ser meovitandi. consciously c. F6lkio syngur viljandi/ af asettpeople.the sing intentionally 'People sing intentionally' u raoi The distinction with respect to the definite or indefinite properties goes essentially along with our characterisation of the impersonal passive construction. More precise­ly, the reason ser meovitandi is infelicitous with verbs that form an impersonal passive, as shown in (42), is clearly that the definiteness encoded in this adverbial expression does not match the semantics of the base verb which does not count as definite. Following this, the fact that the verb mila 'roll' does not license an impersonal passive (43) is due to the property of definiteness associated with this type of verbs, and, of course, this semantic component does not meet the generic characterisation of the impersonal passive. A question arises. How can we distinguish between definite and indefinite mean­ings encoded in verbs? One solution might be to regard this distinction as purely lin­guistic. In other words, the assignment of definite and indefinite meanings to events denoted by verb forms is to be taken as 'arbitrary' (Saussure 1916 [1983]). It goes without saying that, it is indeed not easy, on cognitive grounds, to explain why the event expressed by hlaupa 'run' is indefinite, while that ofrulla 'roll' is definite, inso­far as we can say that in both cases one can, in principle, be aware of doing a desig­nated activity. It might suffice to say, however, that, due to arbitrariness of language, this given distinction is made possible when native speakers of lcelandic are in agree­ment with it when they make an utterance. Thus, what is relevant in our ongoing discussion is to identify the fact that the con­structional meaning ofgenericity is affecting, or imposing constraints upon, the bebav­iour of verbs in Icelandic. Accordingly, the present discussion not only justifies our intuition that there is a linguistic construct that might adequately be dubbed 'the Icelandic impersonal passive construction' but also explains why semantic factors pro­posed on the basis of other languages such as Dutch do not hold for Icelandic. 5. Summary and Remaining Problems The discussion in the preceding sections indicates that there are fine-grained semantic factors that systematically take part in the formati on of lcelandic impersonal passives. The existence ofthese factors bas also been shown, if not in any direct man­ner, to be unmotivated by the unergative/unaccusative distinction. We then claimed that one crucial factor that licenses the construction is genericity encoded in it. By showing that the adverbial expression ser meovitandi 'intentionally, knowingly', which picks out the definiteness component, is infelicitous with passivisable verbs, we provided good evidence that impersonal passives in Icelandic express generic events that characterise or summarise what people generally do. This observation led us to the fact that expressions such as viljandi or afasettu raoi, which are taken as expressing volition, are insensitive to (in)definiteness, and this explains why Icelandic differs from languages such as Dutch (Perlmutter 1978; Zaenen 1993) where volition is con­sidered to be a relevant factor. In this respect, it might be correct to say that fine­grained semantics underlying the formation of Icelandic impersonal passives is, as far as we can surmise, largely language-specific. We have also drawn attention to the fact that classifications such as activity or stative predicates can hardly count as semantic primi ti ves, as opposed to Van Valin ( 1991); our discussion made it clear that the noti on of activity is obviously stili coarse-grained and ambiguous. The inappropriateness of his analysis lies crucially in his failure to observe the very fact that the Icelandic imper­sonal passive encodes genericity. Researchers working within the Construction Grammar framework state that there are grarnmatical constructions that exist independently of verbs which instantiate them. This idea appeals to our Icelandic data, while we stili feel that the description of constructions in terms of the argument structure of a predicate such as X CA USES Y TO RECEIVE Z, along the lines proposed in Goldberg (1995, 1998), might not count as an appropriate representation to our present finding. In other words, constructional meanings are, in our terms, built more on our interaction with extralinguistic compo­nents, and these components clearly extend beyond the number of arguments and type of predicates encoded. In this regard, the proposal in Kay and Fillmore (1999) might provide us a sound testing ground for further research on the nature of grammatical constructions. Although I trust that the present proposal explains a great deal ofthe relevant facts ofthe lcelandic impersonal passive, there are stili facts, as given below, that might not be explained purely semantically nor along the lines proposed above. I believe that if we sol ve, or gain insight into, these problems, a unified account ofimpersonal passives in Icelandic will certainly be arrived at, and, concurrently, we will make a contribution to our real understanding of the nature of natural language. (48) (1) Ali verbs which take a reflexive sig 'self' (e.g. baoa sig 'take a bath') do not form impersonal passives. (II) There are verbs such as deyja 'die', spretta 'grow', vaxa 'grow' which are, when appearing with vera 'be', ambiguous and open to generic and resultative interpre­tations depending on the given context. (III} There are a small set ofverbs such asflyija 'move', boroa 'eat', whose behaviour is not consistent with the present proposal; they are compatible with ser meovi­tandi 'consciously, knowingly', while forming an impersonal passive. (IV) Verbs such as geispa 'yawn', siija 'sit', skilja 'understand', snokta 'sob', vinka 'wave', when they form impersonal passives, do not allow an overt expression ofan a/phrase with a generic NP (e.g. aff6lkinu 'by people'). References COMRJE, Bemd. 1977. In Defense ofSpontaneous Demotion: The Impersonal Passive. In Cole, P, J. M. Sadock (eds.), Syntax and Semantics, vol. 8. Grammatical Relations. San Diego: Academic Press. DOWTY, David. 1991. Thematic Proto-Roles and Argument Selection. Language, 67 (3), 547-619. GOLDBERG, Adele. E. 1995. A Construction Grammar Approach to Argument Structure. Chicago: The University of Chicago Press. GOLDBERG, Adele E. 1998. Pattems of Experience in Pattems of Language. In Tomasello, M. (ed.), The New Psychology ojLanguage, 203-220. New Jersey/London: Lawrence Erlbaum Associates, Publishers. KATHOL, Andreas. 1994. Passives without Lexical Rules. In Nerbonne, J., K. Netter, and C. Pollard, (eds.), German In Head-Driven Phrase Structure Grammar, 237-272. Stanford, Calif.: CSLI Publications. KAY, Paul and Charles FILLMORE. 1999. Grammatical Constructions and Linguistic Generalizations: The What's X doing Y? Construction. Language, 75(1), 1-33. KLAIMAN, M. H. 1991. Grammatical Voice. Cambridge: Cambridge University Press. KRIFKA, Manfred, Francis J. PELLETIER, Gregory N. CARLSON, Alice TER MEULEN, Godehard LINK, and Gennaro CHIERCHIA. 1995. Genericity: An Introduction. In Carlson, Gregory N. and Francis J. Pelletier ( eds. ), The Generic Book, 1-124. Chicago and London: The University ofChicago Press. LEVIN, Beth. and M. RAPPAPORT. 1989. An Approach to Unaccusative Mismatches. Proceedings ojNELS, 314­ 329. Amherst: University ofMassachusetts. PERLMUTTER, David. 1978. Impersonal Passives and the Unaccusative Hypothesis. Jaeger, J. J. et al. (eds.), Proceedings ojthe 4th Annual Meeting ojthe Berkeley Linguistics Society, 157-189. Califomia: University ofCalifomia, Berkeley. POLLARD, Carl. 1994. Toward a Unified Account ofPassive in German. In Nerbonne, J., K. Netter, and C. Pollard, (eds.), German In Head-Driven Phrase Structure Grammar, 273-296. Stanford, Calif.: CSLI Publications. SAUSSURE, Ferdinand de. 1916 [1983]. Course in General Linguistics. ed. by Charles Bally and Albert Sechehaye in collaboration with Albert Reidlinger. Translated from the French by Wade Baskin. London: Duck-worth. SHIBATANI, Masayoshi. 1985. Passives and Related Constructions: A Prototype Analysis. Language, 61 ( 4 ): 821-848. SMITH, Henry. 1993. Linking Changes in lcelandic. Historical Linguistics. Papers jrom the 9th International Conjerence on Historical Linguistics, 467-484. Amsterdam/Philadelphia: John Benjamins Publish­ing Company. VAN VALIN, Robert. Jr. 1991. Another Look at lcelandic Case Marking and Grammatical Relations. Natura/ Language and Linguistic Theory 9: 145-194. YAMAGUCHI, Toshiko. (2000). lcelandic Auxiliaries Hafa and Vera. A Semantic Account. In Alan K. Melby and Arle R. Lommel (eds.), LACUS Forum XXVI, 305-320. Fullerton, Calif.: The Linguistic Association of Canada and United States. ZAENEN, Annie. 1993. Unaccusativity in Dutch: Integrating Syntax and Lexical Semantics. In Pustejovsky, J. (ed.), Semantics and the Lexicon, 129-161. Dordrecht: Kluwer. ZAENEN, Annie. and Joan MALING. 1990. Unaccusative, Passive and Quirky Case. In Maling, J. and A. Zaenen ( eds. ), Syntax and Semantics. Modern Icelandic Syntax, vol. 24, 137-152. San Diego: Academic Press. Povzetek BREZOSEBNI TRPNIK V ISLANDŠČINI V prispevku se utemeljuje misel, da islandski brezosebni trpnik izraža generična glagolska deja­nja, v katerih delujejo ljudje. Dokaz so diagnostični testi, ki kažejo, da so brezosebni trpni stavki združljivi samo z izrazi, ki se ne nanašajo na posamična glagolska dejanja. Tako so testi neuspešni s prislovnim določilom ser meovitandi 'zavestno, zavedajoč se', o katerem se navaja, da izraža do­ločnost. V nadaljevanju avtorica odkloni pojem dejavnosti kot činitelja, ki da sproža islandski brez­osebni trpnik (Van Vanlin 1991), ker je ta pojem preohlapen, da bi se mogle z njim izvesti kake po­splošitve. Avtorica vidi v generičnosti lastnost stavkov, ti pa so idiomatksa povezava med obliko in pomenom. To misel šteje za obetavno, vendar prepušča prihodnjim raziskavam, kako bodo povezale obravnavane pojave z zgradbenostjo (prim. Goldberg 1995, Kay in Fillmore 1999).