THE NORWEGIAN PSEUDOPASSIVE IN LEXICAL THEORY

advertisement
Helge Lødrup:
Working Papers in Scandinavian Syntax 47 1991 118-29
THE N ORW EGIAN PSEUDOP ASSI VE IN LEXICAL THEOR Y 1
1. Bresnan 1982 argues that passive is lexical. She discusses and rejects
several arguments against this position. One of them concerns so-called
pseudopassives, i.e. passives like 1 whose subject corresponds to the
object of a preposition in the active:
1 The roof was walked on
These passives might seem to be impossible to handle in a lexical
theory. The reason is that lexical operations, like Bresnan´s lexical rule
of passive, are strictly local. They only affect the arguments of the verb,
and do not go "into" the internal structure of an argument.
Bresnan´s solution to this problem is "reanalysis", the verb and the
preposition are reanalyzed as a complex verb (e.g. walk on). This verb
takes an object that can become a passive subject with the ordinary
passive rule. Bresnan gives several arguments for reanalysis in English
pseudopassives. And reanalysis is well known from both traditional and
generative descriptions of English, cf. e.g. Jespersen 1969:138-39,
Hornstein and Weinberg 1981.
But what about other languages? Bresnan 1982 gives the impression
that a lexical passive presupposes reanalysis for pseudopassives. For a
lexical theory of universal grammar, this implies that all pseudopassives
in all languages involve reanalysis. But this cannot be correct.
Norwegian has a rather productive pseudopassive (as opposed to the
other Scandinavian languages). Some years ago several Norwegian
grammarians established that the arguments for reanalysis in English are
not applicable to Norwegian, and it was concluded that the Norwegian
pseudopassive cannot involve reanalysis. (Cf. Lødrup 1985, Christensen
1986, Hestvik 1986, Åfarli 1989a:170-73, 1989b.) The main point in
the argumentation was that the verb and the preposition never behave
like a syntactic unit in Norwegian pseudopassives. Cf. e.g. 2-5:
2 De må bli passet bedre på
´They must be looked better after´
For discussion and advice I would like to thank Kirsti Koch Christensen,
Helge Dyvik, Janne Bondi Johannessen and Victoria Rosén.
1
118
3 Hvorfor passes de ikke bedre på?
´Why look-PASSIVE they not better after´
(i.e. ´Why are they not looked after better´)
4 *Hvorfor passes på de ikke bedre?
´Why look-PASSIVE after they not better´
5 *en passet på hund
´a looked after dog´
As far as I know, the conclusion that the Norwegian pseudopassive does
not involve reanalysis has not been challenged in print. (But Holmberg
and Platzack 1991 assume reanalysis for "Mainland Scandinavian",
without discussing the problems involved.)
For a lexical theory like Lexical Functional Grammar (LFG), this
conclusion might seem to be rather damaging2. (Cf. Rosén 1988:25152.). But this is not necessarily so. In this paper I will try to show not
only that an LFG analysis of the Norwegian pseudopassive is possible,
but also that there is independent evidence for this analysis. The
analysis proposed is based on insights in Taraldsen 1979:49-50 and
Dyvik 1980:312-16.
2. Before we proceed, it is necessary to discuss briefly how the
description of pseudopassives should be divided between lexicon,
grammar and other components of linguistic theory.
It has often been pointed out that there are heavy restrictions on wellformed pseudopassives. Even if this is correct, pseudo-passives are
reasonably productive, so it would be misleading to list them as a kind of
A transformational theory like GB should be expected to handle
pseudopassives better. But they turn out to be very problematic for GB
theory. Without reanalysis pseudopassives must be derived by moving
the object of the preposition to subject position. This creates a problem
with the assignment of abstract case. The subject of the pseudopassive
will get nominative case from tense (or whatever assigns nominative
case). The problem is that its trace will get non-nominative case from its
governing preposition, which means that the chain consisting of the
subject and the trace of the prepositional object will get two different
cases. This is an impossible situation for GB theory. To avoid this
problem one would have to assume that passive morphology deprives
the preposition of the ability to assign case in pseudopassives. But this
would be entirely ad hoc. (It has been claimed, however, that case
conflicts are unavoidable in the description of Norwegian passives, cf.
Hestvik 1986, Nordgård and Åfarli 1990:138-39. To me this seems to
undermine GB theory.)
2
119
idioms3. Some pseudopassives
prepositions, as in 6-7:
involve
idioms
or
verb-selected
6 Per ble tatt rotta på
´Peter was taken the rat on´ (i.e. ´Peter was beaten´)
7 Per ble tenkt på
´Peter was thought of´
But most pseudopassives are not like that, cf. e.g. 8-11:
8 Sengen ble sovet i
´The bed was slept in´
9 Gaffelen er aldri blitt spist med
´The fork has never been eaten with´
10 Barna ble skiftet bleier/bukser/klær på
´The children were changed napkins/trousers/clothes on´
11 Talgen støptes lys av
[from Valdres - 900 aarsskrift, 1923]
´The tallow moulded-PASSIVE candles of´
And listing would not reveal the restrictions on the pseudopassive
anyway. Cf. e.g. 12-14:
12 Taket ble gått på av to barn
´The roof was walked on by two children´
13 ??Storgata ble gått på av to barn
´Main Street was walked on by two children´
14 Storgata er ikke blitt gått på i hele dag på grunn av asfaltering
´Main Street has not been walked on all day because of asphalting´
Examples like 12-14 indicate that there are pragmatic restrictions on
pseudopassives. In the literature on English, such restrictions have been
discussed, and they have been claimed to be exactly the same as on
passives in general. Bolinger 1977 proposes the following generalization:
15 "the subject in a passive construction is conceived to be a true
patient, i.e. to be genuinely affected by the action of the verb" (p. 67)
Riddle and Sheintuch 1983 criticize 15, and propose that
Bresnan 1982:50 claims that "there is a lexical relation between the
verb and its passivized subject: in each case the verb-preposition
combination expresses a lexicalized dyadic relation". On the other hand,
Bolinger 1977:59 claims that pseudopassives are "a completely open
set". Also Davison 1980, Riddle and Sheintuch 1983 emphasize the
productivity of English pseudopassives.
3
120
16 "the crucial condition on the occurrence of any NP as a passive
subject is that its referent play the most prominent role in the eyes of
the speaker within the situation described by the passive clause" (p.
559)
It is not clear to me to what extent these generalizations are adequate,
the point is that a condition like 15 or 16 is independently needed to
account for ordinary passives like 17-19:
17 Per ble forlatt av Kari og barna i 1968
´Per was left by Kari and the children in 1968´
18 ??Norge ble forlatt av Per i 1968
´Norway was left by Per in 1968´
19 Norge
rommet
blir
forlatt av
alle
levende
vesener i
filmen Flukt i
´Norway is left by all living creatures in the movie Escape in space´
It seems plausible that Per in 17 and Norge in 19, but not Norge in 18,
should "be genuinely affected" and "play the most prominent role".
The conclusion is that pseudopassives should be derived by a general
rule, and that overgeneration is unavoidable, because pragmatics must
be involved anyway.
3. What do we do without reanalysis?
The LFG treatment of passives has changed over the years. In current
LFG lexical rules in the traditional sense have been made superfluous by
"lexical mapping theory" (see Bresnan and Kanerva 1989, Bresnan
1990). Lexical mapping theory accounts for how the thematic roles of a
predicate are realized as grammatical relations by adding linking
information from different sources. This information comes from
classifications of specific roles (e.g. that theme will be assigned to
subject or object), from default classifications (e.g. that the role highest
on the role hierarchy will be assigned to the subject), from special rules
for groups of verbs, and from morphosyntactic operations like the
passive. In lexical mapping theory passive is an operation which makes
the verb´s highest role unavailable for linking. (The "agent phrase" is an
adjunct.)
It goes without saying that reanalysis can be implemented in this
framework (see Bresnan 1990:8). Reanalysis gives us verbs like walk on,
cf. 20:
20 walk on <agentive theme>
121
The thematic role assigned by the preposition (the theme) has become
a thematic role of the reanalyzed verb, and can be assigned to the
subject.
When we derive pseudopassives without reanalysis, we cannot "reach"
the role that the preposition assigns. This has the important
consequence that the subject of a pseudopassive cannot get a thematic
role from its verb, it must be treated as non-thematic. To be concrete,
let us look at the verbs gå ´walk´ and legge ´put´, as in 21-22:
21 gå <agentive locative>
22 legge <agentive theme locative>
Passive makes the highest role, agentive, unavailable for linking. Both
passive participles will have locative assigned to OBLlocative, and lagt will
have theme assigned to subject or object. This gives us 23-24
(disregarding lagt with a theme subject):
23 gått <OBLlocative>
24 lagt <OBJ OBLlocative>
These lexical entries are not well-formed at this stage of the derivation,
as a subject is required by the subject condition (Bresnan and Kanerva
1989:28). The subjects of 23-24 must be non-thematic, their entries
will then be as in 25-26 (where position outside the angle brackets
means non-thematic):
25 gått <OBLlocative> SUBJ
26 lagt <OBJ OBLlocative> SUBJ
One possibility is that they get a formal subject, we will then get the
lexical entries for sentences like 27-28:
27 Det ble gått på taket
´It was walked on the roof´
28 Det ble lagt duk på bordet
´It was put cloth on the table´
The other possibility is that they get a non-formal subject. I assume
(with Levin 1987) that non-thematic subjects and objects are inserted
"freely". But an NP with a governable grammatical relation must have a
thematic role from at least one predicate, except formal subjects and
objects. (This follows from the Semantic Coherence Condition in Levin
1987:25.) What we have to say is simply that the subject of the verb is
also the oblique´s object in functional structure. In other words, we
want the functional structure 29:
122
29 SUBJ taket
PRED gås
OBLlocative PRED på
OBJ
29 is simplified, and has the inflectional passive to make things simpler.
The curved line indicates that the subject of the verb is also the
oblique´s object.
Note that this analysis makes the pseudopassive very different from
other passives, since the subject has a "double" function. (In LFG, a
passive like The cake was eaten does not have the cake as an object of
eat in any sense at any level of representation.)
The pseudopassive in a way resembles the raising to subject
construction. In neither case does the subject get its role from its verb,
but from the head of an argument of its verb. In LFG raising to subject
verbs bear an equation saying that the subject of the verb is also the
subject of the verbal complement, as in 30:
30 (↑ SUBJ) = (↑ VCOMP SUBJ)
The corresponding equation for pseudopassives will say that the subject
of the verb is also the object of the oblique, as in 314:
31 (↑ SUBJ) = (↑ OBLΘ OBJ)
Pseudopassives represent a pattern which seems to be limited to
passive verbs. In general the passive uses patterns independently
established by active verbs. (See e.g. Anward 1989.) But in some cases
the passive extends these patterns, in the sense that a pattern that is
marginal in the active is more productive in the passive. An example is
raising to subject, which in the active is allowed only by a small number
of Norwegian verbs. In the passive, however, it is allowed by many verbs
which do not take raising to object in the active (e.g. påstås ´claimPASSIVE´). And there exist cases where the passive has a pattern of its
own. An example is passives of intransitive verbs in German (Gestern
wurde getanzt ´Yesterday was danced´), which are the only German
verbs that can appear without a constituent with a governable
grammatical relation. My only candidate for an active that is parallel to
the pseudopassive, in the sense that it needs the equation 31, is
Fødselsdagen din er ikke mer enn to uker til ´Your birthday is not more
than two weeks to´ (from Fretheim 1977:152).
4
123
Within the lexical rule framework of Bresnan 1982, the lexical rule for
pseudopassives (disregarding morphology and the agent phrase) will be
SUBJ -> Ø with the addition of 31.
Pseudopassives require that the oblique does not have a "competing"
object, cf. 32:
32 *Taket ble gått på bordet
´The roof was walked on the table´
In LFG this will follow automatically. In the functional structure of 32
both the subject and the oblique´s object will have two different values:
taket and bordet. This kind of situation is prohibited by a general wellformedness condition on functional structures which says that an
attribute cannot have more than one value. (This is called Functional
Uniqueness, cf. Kaplan and Bresnan 1982:181.)
The subject of a pseudopassive cannot be the object of any oblique, cf.
33-35:
33 Per tenkte på reisen til Drøbak
´Per thought of the trip to Drøbak´
34 Reisen til Drøbak ble tenkt på
´The trip to Drøbak was thought of´
35 *Drøbak ble tenkt på reisen til
´Drøbak was thought of the trip to´
35 is ungrammatical because the subject of a pseudopassive must be
locally available, in the sense that
a) the oblique whose object is subject of the pseudopassive, must itself
be an argument of the pseudopassive
b) the subject of the pseudopassive must be the "topmost" object in
the oblique
In LFG this is accounted for by a general restriction which prohibits
reference to more than two attribute names on either side of an
equation (disregarding unbounded dependencies). (This is called
Functional Locality, cf. Kaplan and Bresnan 1982:278, note 20.) This
means that 36, the equation needed for generating 35, would be too
"long", referring to three attribute names:
36 (↑ SUBJ) = (↑ OBLΘ OBLΘ OBJ)
It has often been claimed that pseudopassives cannot have objects of
adjuncts as subjects. (See e.g. Bresnan 1982 on English.) And there are
arguments that this is correct, e.g. the impossibility of 37-38:
124
37 *Åtte timer ble sovet i
´Eight hours were slept in´
38 *God samvittighet ble sovet med
´Good consciousness was slept with´
On the other hand, there are cases where this seems to be problematic,
e.g. 39-40:
39 Sengen ble sovet i
´The bed was slept in´
40 Gaffelen er aldri blitt spist med
´The fork has never been eaten with´
The distinction between obliques and adjuncts is a notorious problem in
linguistic theory, and cannot be discussed here. If we want to allow for
objects of adjuncts as subjects of pseudo-passives, 41 could be
introduced as an alternative to 31:
41 (↑ SUBJ) = (↑ ADJ OBJ)
(If obliques and adjuncts are a natural class, 31 and 41 could be
expressed as one rule.) 41 will overgenerate, and give us sentences like
37-38. But overgeneration is not really a problem if such sentences are
accounted for by general restrictions on subjects of passives. And they
are covered at least by the restriction of Bolinger 1977 (15 above)5.
There is, however, a technical problem involved, the equation 41
presupposes a treatment of adjuncts that is different from standard
LFG6.
5Note
that sentences like 37-38 cannot be excluded by general
conditions on preposition stranding. The reason is that Norwegian as the
main rule always allows preposition stranding in unbounded
dependencies, like the other Scandinavian languages (cf. Herslund 1984,
Maling and Zaenen 1985).
6LFG
treats adjuncts as sets. 41 then only works in sentences with one
adjunct, the reason is that "ADJ" in 41 is understood as "all members of
the set of ADJ". If we want to keep adjuncts as sets, a slight extension
of the formalism is required. The following alternative to 41 was
proposed to me by Helge Dyvik:
(↑ SUBJ) ∈ f [V g ∈ (↑ ADJ) [f = (g OBJ)]]
(f and g vary over functional structures. f means "the set of f such
that...", i.e. λ f.)
125
4. I have treated pseudopassives rather differently from other passives.
And there is independent evidence for this. Hovdhaugen 1977:29
pointed out that if there is an object in a pseudopassive, it must be
indefinite. Cf. 42:
42 Barna ble skiftet bleier/*bleiene på
´The children were changed napkins/the-napkins on´
(Exceptions involve parts of idioms, as in Per ble tatt rotta på ´Peter
was taken the rat on´, i.e. ´Peter was beaten´.) In other personal
passives there is of course no requirement that an object must be
indefinite, cf. 43-44:
43 Per ble tildelt medaljen
´Per was awarded the medal´
44 Medaljen ble tildelt Per
´The medal was awarded Per´
The requirement that an object must be indefinite is, however, well
known from the existential construction (active or passive)7. Cf. 45-46:
45 Det kom en mann/*mannen på veien
´It came a man/the-man on the road´
46 Det ble sett en bjørn/*bjørnen
´It was seen a bear/the-bear´
What is common to the existential construction and the pseudopassive
of a transitive verb? The generalization should be obvious, the
indefiniteness requirement holds of an unaccusative object whose verb
has a non-thematic subject. We see that the lexical analysis of
pseudopassives automatically gives us the indefiniteness effect8. (A
7A
difference between pseudopassives and the corresponding impersonal
passives is that an agent phrase is possible in pseudopassives, but
rather awkward in the corresponding impersonal passives. (Cf. Taket ble
gått på av to barn ´The roof was walked on by two children´, ??Det ble
gått på taket av to barn ´It was walked on the roof by two children´.)
This is observed by Hovdhaugen 1977:44, note 17, who also points out
that an agent phrase is not natural in passives of intransitives (p. 24).
But this seems to be a more general difference between passives with
an ordinary subject and passives with a formal subject, cf.
En bjørn ble sett av to barn ´A bear was seen by two children´
??Det ble sett en bjørn av to barn ´It was seen a bear by two children´
Pseudopassives of transitive verbs are more marginal in English than in
Norwegian, but not impossible, cf. Every time I sit down (...) I get
dumped confetti on (authentic example from Riddle and Sheintuch
8
126
similar analysis can be found in Taraldsen 1979:49-50 and Dyvik
1980:312-16, who proposed that pseudopassives are transformationally
derived from impersonal passives, while other passives are lexically
derived.) Note that the indefiniteness effect cannot follow in this way if
reanalysis is accepted, because then pseudopassives are just ordinary
passives9.
A problem for this account of the indefiniteness effect is that there are
stronger restrictions on the object in a pseudopassive than in an
existential sentence. This is pointed out by Taraldsen 1979:49, who
discusses the examples 47-49:
47 Brevet ble klistret frimerker på
´The letter was pasted stamps on´
48 ?Brevet ble klistret noen frimerker på
´The letter was pasted some stamps on´
49 ??Brevet ble klistret noen grønne frimerker på
´The letter was pasted some green stamps on´
There is no corresponding unacceptability in existential sentences, cf.
50:
50 Det ble klistret noen grønne frimerker på brevet
´It was pasted some green stamps on the letter´
But, as Taraldsen points out, 49 is much better than 51 with a definite
object:
51 *Brevet ble klistret de grønne frimerkene på
´The letter was pasted the green stamps on´
Besides, it must be considered that pseudopassives of transitive verbs
are somewhat marginal. The active object is of course generally the best
1983:556). The indefiniteness requirement is also operative in English,
as pointed out in Ziv and Sheintuch 1981, Riddle and Sheintuch
1983:556. This could be accounted for as in note 9 or 10. Another
possibility would be to try to account for the properties of the English
pseudopassive without reanalysis, and transfer the analysis of
Norwegian.
Taraldsen 1982:290 note 36 suggests that the indefiniteness
requirement is an argument for reanalysis. And it is of course difficult to
imagine uniquely referring NPs as parts of complex verbs. On the other
hand, this argument cannot be decisive, considering the arguments
against reanalysis (in Norwegian) in the literature.
9
127
candidate for a passive subject, so it should not be too prominent when
it gives up this privilege10. (It should not "play the most prominent role
in the eyes of the speaker within the situation described", in the words
of Riddle and Sheintuch 1983, cf. 16 above.)
5. I have tried to show that an LFG analysis of the Norwegian
pseudopassive is possible (which is not surprising, given the generative
capacity of this theory), and that there is independent evidence for this
analysis. The more general conclusion is that a lexical theory of universal
grammar can handle pseudopassives without reanalysis. This conclusion
is not only relevant for LFG, but also for related theories like HeadDriven Phrase Structure Grammar (HPSG).
LITERATURE
Anward, J. 1989 Constraints on passives in Swedish and in English. Working Papers in
Scandinavian Syntax 44, 15-30.
Bolinger, D. 1977 Transitivity and spatiality: The passive of prepositional verbs. In A.
Makkai et al. (eds) Linguistics at the crossroads. Padova: Liviana Editrice. Pp. 5778.
Bresnan, J. (ed.) 1982 The mental representation of grammatical relations. Cambridge,
Mass: MIT Press.
Bresnan, J. 1982 The passive in lexical theory. In Bresnan (ed.) 1982, pp. 3-86.
Bresnan, J. 1990 Levels of representation in locative inversion: A comparison of
English and Chichewa. Manuscript. Stanford University and Xerox Palo Alto
Research Center.
Bresnan, J. and J.M. Kanerva 1989 Locative inversion in Chichewa: A case study of
factorization in grammar. Linguistic Inquiry 20, 1-50.
Christensen, K.K. 1986 Complex passives, reanalysis and word formation. Nordic
Journal of Linguistics 9 135-62.
Davison, A. 1980 Peculiar passives. Language 56 42-66.
Dyvik, H.J.J. 1980 Grammatikk og empiri. Dissertation. University of Bergen.
(Reprinted as Skriftserie 24 and 25 1986, Department of linguistics and
phonetics, University of Bergen.)
Fretheim, T. 1977 Syntaktisk analogi - noe mer enn en billig nødløsning? In T. Fretheim
(ed.) Sentrale problemer i norsk syntaks. Oslo: Universitetsforlaget. Pp. 113170.
Herslund, M. 1984 Particles, prefixes and preposition stranding. Nydanske studier &
Almen kommunikationsteori 34-71.
Hestvik, A. 1986 Case theory and Norwegian impersonal constructions: Subject-object
alternations in active and passive verbs. Nordic Journal of Lingustics 9 181-97.
It could be possible to account for the indefiniteness effect by using
the idea that the object should not be too prominent when it gives up
the privilege of "becoming" the passive subject. But we have the formal
account anyway, we got it for nothing.
10
128
Holmberg, A. and C. Platzack 1991 On the role of inflection in Scandinavian syntax. In
W. Abraham et al. (eds.) Issues in Germanic syntax. Berlin: Mouton de Gruyter.
Pp. 93-118.
Hornstein, N. and A. Weinberg 1981 Case theory and preposition stranding. Linguistic
Inquiry 12, 55-91.
Hovdhaugen, E. 1977 Om og omkring passiv i norsk. In T. Fretheim (ed.) Sentrale
problemer i norsk syntaks. Oslo: Universitetsforlaget. Pp. 15-46.
Jespersen, O. 1969 (1937) Analytic syntax . New York: Holt, Rinehart and Winston.
Kaplan, R.M. and J. Bresnan 1982 Lexical-functional grammar: A formal system for
grammatical representation. In Bresnan (ed.) 1982, pp. 173-282.
Levin, L. 1987 Toward a linking theory of relation changing rules in LFG . Center for the
Study of Language and Information. Report No. 115.
Lødrup, H. 1985 Om reanalyse. Skriftserie 21, 97-110. (Department of linguistics and
phonetics, University of Bergen.)
Maling, J. and A. Zaenen 1985 Preposition-stranding and passive. Nordic Journal of
Lingustics 8 197-209.
Nordgård, T. and T.A. Åfarli 1990 Generativ syntaks: Ei innføring via norsk. Oslo:
Novus.
Riddle, E. and G. Sheintuch 1983 A functional analysis of pseudo-passives. Linguistics
and Philosophy 6 527-563.
Rosén, V. 1988 Norsk passiv: En LFG-analyse. Norsk lingvistisk tidsskrift 6 239-252.
Taraldsen, K.T. 1979 Remarks on some central problems of Norwegian syntax. Review
article of T. Fretheim (ed.) Sentrale problemer i norsk syntaks. Nordic Journal of
Lingustics 2 23-54.
Taraldsen, K.T. 1982 Remarks on government, thematic structure and the distribution
of empty categories. In R. May and J. Koster (eds.) Levels of syntacic
representation. Dordrecht: Foris. Pp 253-291.
Ziv, Y. and G. Sheintuch 1981 Passives of obliques over direct objects. Lingua 54 117.
Åfarli, T.A. 1989a The syntax of Norwegian passive constructions. University of
Trondheim Working Papers in Linguistics 9.
Åfarli, T.A. 1989b Passive in Norwegian and English. Linguistic Inquiry 20 101-8.
Helge Lødrup
University of Oslo
Department of linguistics and philosophy
Pb 1102, Blindern
N-0315 Oslo 3
NORWAY
129
Download