Showing posts with label icelandic. Show all posts
Showing posts with label icelandic. Show all posts

23 February 2015

Hålogaland

Published, edited, images added & comments/annotations (in red) by Kenneth S. Doig)
Hålogaland
(From Wikipedia)


Coordinates: 67°N 14°E
Hålogaland around 1000 CE


Hålogaland was the northernmost of the Norwegian provinces in the mediaeval-Norse sagas. In the early Viking-Age, before Harald Fairhair, Hålogaland was a petty kingdom extending between the Namdalen valley in Nord-Trøndelag county and the Lyngen fjord in Troms county.

Etymology
Ancient Norwegians said that Hálogaland was named after a royal named Hölgi. The Norse form of the name was Hálogaland. The first element of the word is the genitive plural of háleygr, a 'person from Hålogaland'. 

The last element is land, as in 'land' or 'region'. The meaning of the demonym háleygr is unknown. Thorstein Vikingson's Saga, 1, describes it as a compound of Hial, "Hel" or "spirit," and "loge", "fire".

The Gothic historian Jordanes in his work De origine actibusque Getarum - a.k.a. Getica -, written in Constantinople in c. 551 AD, mentions a people "Adogit" living in the far North. 

This could be an old form of háleygir and a possible reference to the petty kingdom of Hålogaland, which based on some medieval accounts may have been inhabited by the Kven people during the first millennium, but also perhaps a long before. Jordanes' Vinoviloth is viewed by many historians as a reference to the Kvens of Northern Scandinavia and Fennoscandia:

And there are beyond these the Ostrogothae (Eastern Geats)(The géatas,{in OE, as in virtually all older and modern Gmc languages, languages, nationalities, weekday-names, holidays, months, are NOT capitalized, unless, at beginning of a sentence} Sw. götar [the Old-English, "OE",{West-Saxon}, diphthong /éa/corresponds to Swe., Nor.{bokmål}, Dan., /ö/, i.e., /ø/, western Old-Norse {Old-Icelandic, Old-Norw} /au/  are probably not the same tribe whom we today call Goths, Swedes goter. By comparing regular phonology [vocalism] between ancient & modern Gmc languages, géatas = Sw. goter) Raumarici (Romerike), Aeragnaricii (Ranrike), and the most gentle Finni (referring to either Sami or Finns), milder than all the inhabitants of Scandza (Scandinavia). Like them are the Vinoviloth (Kvens) also.

According to Emeritus Professor Kyösti Julku, in the modern-day Northern Norwegian county of Troms alone there are at least 12 prehistoric Kven place names. During Viking-Age, Troms formed the northernmost part of Hålogaland.



Alex Woolf links the name Hålogaland to the Aurora Borealis - the "Northern Lights" -, saying that Hålogaland meant the "Land of the High Fire", "loga" deriving from 'logi', which refers to fire.


In the medieval accounts of Ynglingatal and Skáldskaparmál, "Logi" is described as the personification of fire, a fire giant, and as a "son of Fornjót". In the medieval Orkneyinga saga and the account of Hversu Noregr byggðist ('How Norway was inhabited'), Fornjót is described as the King of Finland, Kvenland and Gotland. The royal lineages sprung from his children are discussed in these and other medieval accounts.


The beginning of the Þorsteins saga Víkingssonar ('Saga of Thorstein son of Víking') discusses King Logi who ruled the country north of Norway. Because Logi was larger and stronger than any other man in land, his name was lengthened from Logi to Hálogi, meaning 'High-Logi'. 
Derived from that name his country became called Hálogaland, meaning "Hálogi's land". Eventually the spelling of the name shaped to the modern-day Hålogaland.



The Hversu Noregr byggðist is an account of the origin of various legendary Norwegian lineages. It traces the descendants of the primeval Finnish ruler Fornjót (Fornjotr) down to Nór, who is here the eponym and first great king of Norway, who unites the Norwegian lands (petty kingdoms). 

The Hversu account then gives details of the descendants of Nór and of his brother Gór in the following section known as the Ættartölur ('Genealogies', a.k.a. Fundinn Noregr, 'Founding of Norway'). The Hversu account is closely paralleled by the opening of the Orkneyinga saga.


In 873 AD, according to the Egil's saga (written in c. 1240 AD) the Kvens and Norse cooperate in battling against the invading Karelians. The chapter XVII of Egil's saga describes how Thorolf Kveldulfsson (King of Norway's tax chief starting 872 AD) from Namdalen, located in the southernmost tip of the historic Hålogaland, goes to Kvenland again:


"That same winter Thorolf went up on the fell with a hundred men; he passed on at once eastwards to Kvenland and met King Faravid."
Based on medieval documents, the above meeting took place during the winter of 873-874 AD. Hålogaland's rather close vicinity to Kvenland is also demonstrated in c. 1157 AD in the geographical chronicle Leiðarvísir og borgarskipan by the Icelandic Abbot Níkulás Bergsson (Nikolaos), who provides descriptions of lands around Norway:


Closest to Denmark is little Svíþjóð (Sweden), there is Eyland (Öland); then is Gotland (Gotland); then Helsingaland (Hälsingland); then Vermaland (Värmland); then two Kvenlönd (Kvenlands), and they extend to north of Bjarmalandi (Bjarmia).


Modern usage
In modern times, the term Hålogaland is used in a variety of senses. For some purposes, all of Nord-Norge plus Svalbard and Jan Mayen are covered under the term Hålogaland. For other purposes the counties of Nordland and Troms constitute Hålogaland. Hålogaland or even Mid Hålogaland are frequent terms covering the smaller districts of Ofoten, Lofoten and Vesterålen, as well as the municipalities Bjarkøy, Gratangen, Harstad, Ibestad, Kvæfjord and Skånland of Troms county. The term has also been used in this last sense, minus the Lofoten archipelago.

27 November 2014

Why is English a Germanic language?

Image: Anglo-Saxon god Þunor (literally,
thunder), from Proto-Germanic,PGmc 
*Þunraz (*thunraz), Þórr in western
Old-Norse, Donar in OHG, cognate to
other Indo-European gods, seeming to be,
essentially the same personage. 

Published, edited, images added & annotations/comments (in red) by Kenneth S. Doig


Map/Karta/Mapa: Northwestern Europe, around the
Northsea. Whence the Scando-Germanic precursors to
the "Anglo-Saxons" (the English & many  eastern
Lowlanders (Scotland: Borders, Midlothian, etc.)
came. The semi-legendary year of the  "Anglo-Saxons'"
main invasion-arrival (Hengist & Horsa, brothers from the
Kimbric peninsula, modernday Denmark's Jutland, in
AD 449. At that time, they would have spoken various
 Common-Germanic dialects (Old-English 'OE', per se,
had yet to come into existence) came.


(websource: The Grammarphobia Blog}
(http://www.grammarphobia.com/)

Why is English a Germanic language?

Q: I’ve read that a majority of the words in English are derived from Latin or French? So why is English considered a Germanic (Gmc) language, not a Romance language?
A: Let’s begin with where linguists place English among the world’s languages.
English, Icelandic, Faroese, Norwegian, Swedish, Danish, Frisian, Flemish, Dutch, Afrikaans, German, and Yiddish are the living languages that are part of the Germanic family.
This family is divided into North-Germanic (NGmc) (Icelandic, Faroese, (or Faeroese), Norwegian, Swedish, Danish) and West-Germanic (WGmc) (English, Frisian, Flemish, Dutch, Afrikaans, German, Yiddish). The now defunct East-Germanic (EGmc) branch consisted of Gothic, which is extinct (also included; Vandalic, Lombardic/Langobardic. I believe that since, primarily, Gothic's very early attestation & equally importantly, scholars, antiquarians & archaeologists don't have the typical attributes/attestations of most ancient languages, in their graphogenesis. 

Many such ancient tongues have given us a scant few words, often scratched on bark, wood, carved in stone or crudely impressed into some malleable medium [claytablets, stylus, cuneiform]. The Gothic bishop, Ulfilas (or Wulfilas) of Thrace, but don't quote me on that, Thrace, a man of apparently mixed parentage, half-Gothic & half-Hellenic. Most of the EGmc may not be a separate branch from "Northwest-Germanic, i.e., NGmc, WGmc & Kimbro-Germanic[a] 

[a.k.a., as per prof. Gudmund Schütte, Univ. Hafniae, in his magnum opusOur Forefathers, the Gothonic Nations by Gudmund Schütte
Vol. 50, No. 2 (Feb., 1935), pp. 106-108 Published by: The Johns Hopkins University Press
states good reasons for not calling Germanic, "Germanic", first of all, it is an exonym, we are virtually positive they never called themselves by this appellation, given them by the 'Graeco-Roman' world. "German", itself, is of uncertain etymology. 

It resembles older Celtic words meaning "neighbor", or from Italic, i.e., from Latin, germanus, meaning brother; literally from the same germ, {womb}. Hence Ibero-Romance, Castilian, Galician, Portuguese, Catalan, etc., with forms like hermano/hermana, Castilian-Spanish brother/sister, Portuguese irmão , etc.
 The following in blueletter is from OED 

Ptolemy. German measles attested by 1856.german (adj.)

"of the same parents or grandparents," c.1300, from Old-French germain "closely related" (12c.), from Latin germanus "full, own (of brothers and sisters); one's own brother; genuine, real," related to germen (genitivegerminis) "sprout, bud," dissimilated from PIE *gen(e)-men-, from root *gene- "to give birth, beget" (see genus). Your cousin-german (also first cousin) is the son or daughter of an uncle or aunt; your children and your first cousin's are second cousins to one another; to you, your first cousin's children are first cousin once removed.
.
(Main article continues)
The other principal European language family is the Italic (popularly called Romance). This consists of the modern languages derived from Latin: Portuguese, Spanish, Catalan, Provençal, French, Italian, Rhaeto-Romance and Romanian.
These two families are branches of a single prehistoric language called Indo-European or Proto-Indo-European.
The language-group descended from Indo-European includes the Balto-Slavic, Albanian, Celtic, Italic, Greek, and Germanic families of languages.
It’s estimated that about half the earth’s population speaks a language from the Indo-European group, which is only one of several language groups that have been identified worldwide.
But back to English. Why do we call it a Germanic language?
As Calvert Watkins writes in The American Heritage Dictionary of Indo-European Roots, one of the dialects of Indo-European “became prehistoric Common-Germanic, which subdivided into dialects of which one was West-Germanic.”
This in turn, Watkins says, “broke up into further dialects, one of which emerged into documentary attestation as Old English. From Old English we can follow the development of the language directly, in texts, down to the present day.”
But while English is Germanic, it has acquired much of its vocabulary from other sources, notably Latin and French.
As Watkins explains: “Although English is a member of the Germanic branch of Indo-European and retains much of the basic structure of its origin, it has an exceptionally mixed lexicon. 

During the 1400 years of its documented history, it has borrowed extensively and systematically from its Germanic and Romance neighbors and from Latin and Greek, as well as more sporadically from other languages.”
Where exactly does our modern vocabulary come from? The website AskOxford cites a computerized analysis of the roughly 80,000 words in the old third edition of the Shorter Oxford Dictionary.

The study, published in 1973, offered this breakdown of sources: Latin, 28.34 percent; French, 28.3 percent; Old and Middle English, Old Norse, and Dutch, 25 percent; Greek 5.32 percent; no etymology given, 4.03 percent; derived from proper names, 3.28 percent; all other languages, less than 1 percent.

26 April 2014

Old Norse Pronunciation (alternative) : By Óskar Guðlaugsson

Published, edited, annotated (in red) & images added by Kenneth S. Doig

Old Norse Pronunciation (alternative)

By Óskar Guðlaugsson
‘Old Norse’ is a term very broad in both time and space. In Viking times, it is clear enough that people from all over Scandinavia, and from the peripheries of the Viking world, i.e., Norwegian settlements in the North Atlantic, Danish settlements in Britain, and Swedish settlements in the Baltic and south to the Ukraine, were mutually intelligible to each other in their speech.

30 January 2014

the germanic languages : a fairly detailed overview


"Proto-Germania Superior", c. 250 AD.
early protogermanic (or scando-teutonic or
gothonic) urheimat/proto-homelands
Published, edited, formatted, annotated (in red) & images added by Kenneth S. Doig

written by William G. Moulton & Anthony F. Buccini
(websource: Encyclopædia Britannica)

Germanic "Gmc" languages, branch of the Indo-European "IE" language family. Scholars often divide the Gmc languages into three groups: West-Germanic "WGmc, including English, German, and Netherlandic (Dutch); North-Germanic "NGmc", including Danish, Swedish, Icelandic, Norwegian, and Faroese; and East Germanic "EGmc, now extinct, comprising only Gothic and the languages of the Vandals, Burgundians, Rugians, Langobards (Lombards) and Heruli. (list, not exhaustive)
                    Idyllic Scanian farmhouse in the
                          very heart of Proto-Germania.
                         In modern-day southern Sweden's
                              province, Skåne
 
                 
In numbers of native speakers, English, with 450 million, clearly

28 September 2012

"On Breaking" (Vowel-breaking ['a' to 'éa' in WestSaxon, 'a' to 'iá/já' in Norse] in certain Germanic languages)



Published, formatted, heavily-edited, corrected, images added & annotations (in red) by Kenneth S. Doig Introductory Phonology By Hayes, Bruce (Google Affiliate Ad)


Preface
by K.S.Doig
Dr. Kortland is a superior linguist/philologist, but his typing & writing-skills (at least in English, I hope they´re better in his native Dutch. I spent hours just formatting the sentences from a PDF-file and much more time correcting spelling-errors, etc. I might have missed a few things.
by Frederik Kortlandt, PhD

Professor Emeritus, Leiden University, NL 


ON BREAKING
To the memory of Jörundur Hilmarsson

As H.F. Nielsen points out, for Old English 'it is fairly certain that breaking takes place prior to t-mutation, which itself precedes back-umlaut.2 [...]
Freemasonry Among the Anglo-Saxons (Google Affiliate Ad)

On the other hand, OE breaking must be later than OE fronting of α >  which is most likely to be an independent development' (1984:75, 80). This chronology suffices to show that the Old-English breaking cannot be identified with the Scandinavian breaking.History of the Scots Language (Google Affiliate Ad)

25 February 2012

OLD-NORTHGERMANIC


PUBLISHED, PAINSTAKINGLY EDITED, PROOFREAD, FORMATTED, ANNOTATIONS (IN RED) & IMAGES ADDED BY KENNETH S. DOIG



(Wikipedia)
Old Norse is a North Germanic language that was spoken by inhabitants of Scandinavia and inhabitants of their overseas settlements during the Viking-Age, until about 1300AD.

Proto-Norse (PN) developed into Old-Norse (ON) by the 8th century, and ON began to develop into the modern North-Germanic languages in the mid- to late-14th century, ending the language phase known as ON. These dates, however, are not absolute, since written ON is found well into the 15th century.

ON was divided into three dialects: Old-eastnorse (OEN), Old-westnorse (OWN), and Old-Gutnish (OG). OWN and OEN formed a dialect-continuum, with no clear geographical boundary between them. For example, OEN traits were found in eastern Norway, although Old-Norwegian (ONw) is classified as OWN, and OWN traits were found in western Sweden.


The most speakers spoke OEN in what are present-day Denmark and Sweden. Old-Gutnish, the more obscure dialectal branch, is sometimes included in the OEN dialect due to geographical associations. It developed its own unique features and shared in changes to both other branches.

25 August 2011

Case Alternations and the Icelandic Passive and Middle

Published by Kenneth S. Doig



Peter Svenonius, PhD, professor, University of Tromsö
Nov 11, 2006

1 Introduction
The main question to be addressed in this paper is, why is accusative case systematically
lost in passives (as illustrated in (1)), while dative case is retained, equally
systematically (as illustrated in (2))?
(1) a. Stormurinn
the.storm.NOM
blés
blew
strompinn
the.chimney.ACC
af
off
húsinu.
the.house
‘The storm blew the chimney off the house’
b. Strompurinn
the.chimney.NOM
var
was
blásinn
blown
af
off
húsinu.
the.house
‘The chimney was blown off the house’ (Zaenen and Maling 1984)
(2) a. Skipstjórinn
the.captain.NOM
sökkti
sank
skipinu.
the.ship.DAT
‘The captain sank the ship’
b. Skipinu
the.ship.DAT
var
was
sökkt
sunk
af
by
skipstjóranum.
the.captain
‘The ship was sunk by the captain’ (Zaenen and Maling 1984)
The question is particularly pertinent in Icelandic, which allows accusative subjects,
so that for example (1b) and (3) form a minimal pair.
(3) Strompinn
the.chimney.ACC
blés
blew
af
off
húsinu.
the.house
‘The chimney blew off the house’ (Zaenen and Maling 1984)
1
Examples like (3) are highly marked, and only possible for a finite list of verbs under
certain interpretive conditions. However, they represent a significant enough
feature of Icelandic syntax that the lack of accusative case in (1b) cannot be due
to a general requirement that accusative is dependent on nominative.
It is sometimes thought that the preservation of the dative in (2) is explained by
an assumption that dative case is ‘lexical’ or ‘inherent,’ but the middle construction
in Icelandic provides systematic alternations between dative and nominative
(Zaenen and Maling 1984); for example the dative object in (4) surfaces as nominative
in the middle form, as shown.
(4) a. Ég
I.NOM
týndi
lost
úrinu.
the.watch.DAT
‘I lost the watch’
b. Úrið
the.watch.NOM
týndist.
lost.MIDDLE
‘The watch got lost’ (Sigur􀀀sson 1989:269)
Descriptively, the accusative could be said to be more fragile than the dative, as
the dative is preserved in contexts where the accusative is lost (as in passive).
However, both are structural cases, and there are contexts in which both are lost
(as in the middle).
In this paper, I argue that the contrasts outlined here follow from two independently
motivated assumptions, cast in a decompositional model of lexical
structure. The first assumption is that dative case is determined lower in the decomposed
verbal structure than accusative case; this is motivated by the lexical
semantics which determine the distribution of dative case. The second assumption
is that the middle is lower in the structure than the passive, which is motivated
by the fact that the passive, but not the middle, implies the existence of an external
argument.
2 Object Case in Icelandic
Because case alternations are so central here, it is necessary to quickly review
some relevant facts about case assignment in Icelandic, before moving on to detail
the passive construction.
2
2.1 Basic facts
For most verbs, Icelandic shows a standard nominative-accusative pattern: the
subject gets nominative (in a finite clause), and an object, if there is one, gets
accusative.
(5) a. Hún
she.NOM
vekur
wakes
hana.
her.ACC
‘She wakes her’
b. Hún
she.NOM
hefur
has
litað
colored
hana.
her.ACC
‘She has colored it’
However, as noted, some verbs have dative, accusative, or genitive subjects, and
certain verbs have dative, nominative or genitive objects. In fact, dative is extremely
common on both subjects and objects in Icelandic, and there are many
hundreds of verbs which take the dative (cf. Bar􀀀dal 2001). Most dative subjects
are experiencers (see Jónsson 2003), and most dative objects can be characterized
as themes, especially themes of motion (Maling 2001). A few examples of dative
objects are given here to illustrate, as these are the objects whose case is preserved
in passives but lost in middles.
(6) a. Bandaríkin
the.USA
hafa
have
ekki
not
aflétt
lifted
viðskiptabanni
the.trade.embargo.DAT
á
on
Kúbu.
Cuba
‘The USA has not lifted the trade embargo on Cuba’
b. Sjómennirnir
the.sailors
reyndu
tried

to
bægja
drive
háhyrningunum
the.killer.whales.DAT
frá
from
netinu.
the.net
‘The sailors tried to drive the killer whales away from the net’
c. Þeir
they
flögguðu
flew
færeyska
Faroese
fánanum.
flag.DAT
‘They flew the Faroese flag’ (Maling and Thráinsson 1995)
Unlike German, where dative objects are oblique and behave syntactically much
like PPs (Vogel and Steinbach 1998), Icelandic dative case is structural, and dativemarked
objects pattern with ordinary accusative objects for various phenomena
such as control, binding, secondary predication, promotion under passive, and so
on (Maling 2001).
There are ditransitive verbs, which have nominative subjects and usually have
a dative argument followed by an argument that can be dative, accusative, or genitive,
as illustrated in (7a) (Yip et al. 1987 estimate that there are over a hundred
3
ditransitive verbs with a dative first object). There are also some ditransitives in
which the first internal argument is accusative, and the second is dative (25 verbs),
genitive (10 verbs), or accusative (one verb, kosta ‘cost’) (figures from Yip et al.
1987), as illustrated in (7b).
(7) a. Ég
I.NOM
hef
have
gefið
given
stráknum
the.boy.DAT
gjafir.
gifts.ACC
‘I have given the boy gifts’
b. Hann
he.NOM
svipti
deprived
konuna
the.woman.ACC
aleigu
asset.DAT
sinni.
her.DAT
‘He deprived the woman of all of her possessions’
In addition, there are numerous examples with dative subjects, usually experiencers
(Jónsson 1997-1998, Jónsson 2003). When such verbs have an object, it is
usually nominative.
(8) a. Mér
me.DAT
blæddi.
bled
‘I bled’
b. Fólkinu
the.people.DAT
sárnuðu
hurt
essi
these.NOM
ummæli.
words.NOM
‘The people were hurt by these words’ (Jónsson 2003:129–130
See Zaenen et al. (1985) for extensive evidence that such datives are subjects,
rather than topicalized internal arguments (see also Sigur􀀀sson 2004 for recent
discussion and references).
There is also a class of what can be called ‘quirky unaccusatives,’ verbs that
take a single internal (theme) argument in a non-nominative case (cf. (3) in §1),
but I postpone discussion of these special cases until §5, turning to an outline of
an account for the cases already discussed.
2.2 Bipartite case-assigners
There are reasons to think that accusative and dative objects get their licensing
from a combination of two heads in the projection of the verb. First, consider the
dative. As argued by Bar􀀀dal (1993; 2001), there are semantic correlations to the
distinction between dative and accusative objects. For example, themes of motion
often appear in the dative, as illustrated in (9), where the verb skjóta ‘shoot’ takes
the dative when the object is a projectile, but accusative when the object is an
affected patient.
4
(9) a. Þeir
they
skutu
shot
geimfaranum
the.astronaut.DAT
á
on
loft.
sky
‘They shot the astronaut into the sky’
b. Þeir
they
skutu
shot
geimfarann.
the.astronaut.ACC
‘They shot the astronaut’ (Maling and Thráinsson 1995)
I argued in Svenonius (2002) that dative appears on themes when the causing
subevent does not perfectly overlap with the subevent which describes the process
that the object undergoes; this is illustrated in the minimal pair in (10).
(10) a. Þeir
they
báru
carried
heyið
the.hay.ACC
upp
up
á
on
vagninn.
the.wagon
‘They carried the hay up onto the wagon’
b. Þeir
they
hentu
threw
heyinu
the.hay.DAT
upp
up
á
on
vagninn.
the.wagon
‘They threw the hay up onto the wagon’
The verb bera ‘carry’ describes accompanied motion, in which the external argument
participates in the event as long as the internal argument undergoes it.
The verb henda ‘throw,’ in contrast, involves a causing subevent followed by a
possibly non-overlapping movement subevent.
Thus, dative on theme objects in Icelandic correlates with a particular kind of
Aktionsart, one which is only possible with certain roots (henda but not bera, for
example, and one use of skjóta but not the other). At a first approximation, we can
say that verbs that take dative theme objects have a particular marked property,
call it DAT.
DAT is a necessary but not a sufficient condition for dative case on theme internal
arguments. Recall from (4) in §1 that dative case is lost in the middle. Dative
is also absent from the unaccusative variant of a typical causative-inchoative alternation.
(11) a. Skipstjórinn
the.captain.NOM
sökkti
sank
skipinu.
the.ship.DAT
‘The captain sank the ship’
b. Skipið
the.ship.NOM
sökk.
sank
‘The ship sank’ (Zaenen and Maling 1984)
5
This recalls Burzio’s Generalization (Burzio 1986), whereby the existence of an
external argument is necessary for the assignment of accusative case.
Assuming that the external argument is introduced by some sort of voice or
causative head, it appears that both that head and DAT are necessary for a theme
object to surface with dative case (cf. Watanabe 1993, in which structural case is
assigned by a combination of an Agr head and V for accusative or T for nominative).
Below I will assume that there are two heads associated with external
arguments, Voice and Init (for initiation, after Ramchand 2006). I will suggest
that dative themes are dependent on Voice.1
A similar argument can be made for accusative. First, note that accusative is
more ‘fragile’ than the dative in that it disappears under passive (cf. (1b) in §1),
where an external argument is implicitly present. The sensitivity of accusative to
passive voice, which affects the realization of the external argument, suggests that
accusative case has a licenser at or near the heads responsible for the expression
of the external argument (this is an implementation of the essence of Burzio’s
Generalization).
Now consider the facts about case assignment in ditransitives discussed by
Holmberg (2002). Holmberg points out that although Icelandic allows impersonal
passives, as in (12a), it does not allow impersonal passives of typical ditransitives,
as illustrated in (12b).
(12) a. Það
it
hafa
have
verið
been
skrifaðar
written.PL
rjár
three
bækur
books.NOM
um
about
etta.
that
‘There have been three books written about that’
b. *Það
it
hafa
have
verið
been
gefnar
given.PL
stráknum
the.boy.DAT
gjafir.
gifts.NOM
(Holmberg 2002)
This is all the more striking given that passives of ditransitives are perfectly acceptable,
if one or the other argument is promoted to subject position.
(13) a. Stráknum
the.boy.DAT
voru
were
gefnar
given.PL
gjafir.
gifts.NOM
‘The boy was given gifts’
b. Gjafirnar
the.gifts.NOM
voru
were
gefnar
given.PL
stráknum.
the.boy.DAT
‘The gifts were given to the boy’ (Holmberg 2002)
6
Holmberg argues that the unacceptability of (12b) is due to the fact that the dative
indirect object intervenes between the nominative and T (cf. 2000 “defective
intervention”). Movement of either argument to SpecTP resolves this and permits
the necessary Agree relation to obtain.
This means that an intervening dative DP in Icelandic blocks the formation
of an Agree relation with a more distant DP (see also Holmberg and Hróarsdóttir
2003). The dative in a typical ditransitive construction (such as (7a)) is clearly
between Init (the initiator head, also known as CAUSE or v) and the accusative; for
example the dative blocks Object Shift of the accusative, unless the dative itself
undergoes object shift (cf. e.g. Collins and Thráinsson 1996). So accusative case
cannot be assigned to an accusative direct object directly from Init under Agree,
in an example like (7a). Nor can there be a low source for accusative, or else there
would be no reason for the theme in (13) to be nominative.
The conclusion must be that two elements are necessary for the assignment of
accusative (just as with the dative): one which is relatively high up, near Init, and
interacts with passive; call this one AgrO. The other necessary element is relatively
low down, closer to to the accusative argument itself, perhaps even the head
introducing it. Schematically, this looks something like the following, assuming
a low applicative head R to introduce the two internal arguments (Pesetsky 1995,
Harley 1995, Ramchand 2006).
(14) AgrO
AgrOi Voice
Voice Init
D
Initj VDAT
VDATj R
D
[DAT] Ri D
[ACC]
AgrO and R are in a chain, as indicated by the subscripts; R can thereby check
case on the accusative theme. The heads Init and VDAT are also in a chain, and
7
check dative case on the goal.
If AgrO is absent, as it is in the passive, then R cannot assign accusative to the
object, and nominative case must be assigned instead. Suppose that nominative
case cannot be assigned through R, and that the nominative case assigner must
form a chain directly with the theme argument. This chain, formed as it is with a
nominal, is blocked by the dative, unlike the AgrO–V chain in (14).
Benefactive or Goal arguments with dative case in Icelandic are different from
dative themes in that their dative case is not affected by the middle.
(15) a. Pétur
Pétur
bauð
offered
mér
me.DAT
vinnu.
job.ACC
‘Peter offered me a job’
b. Mér
me.DAT
bauð-st
offered-MID
vinna.
job.NOM
‘I got the opportunity to get a job’ (Sigur􀀀sson 1989:260)
(16) a. Ég
I
hellti
spilled
mjólkinni
the.milk.DAT
niður.
down
‘I spilled the milk’
b. Mjólkin
the.milk.NOM
hellti-st
spilled-MID
niður.
down
‘The milk spilled’ (Sigur􀀀sson 1989:265)
As shown, the verb bjóða ‘offer, ask’ contrasts with the particle verb hella niður
‘spill’ in that the middle of the former preserves dative while the latter loses it. I
assume that the high licenser for the benefactive is Init, as indicated previously,
but I will argue below that the high licenser for the theme is Voice.
The assumptions about case made here can be schematized as follows: verbs
which take dative themes have a feature DAT; when VDAT enters into an Agree
relation with Voice, VDAT can check dative case on a DP.2 I will argue below that
Voice is present in passives, but absent from middles. The alternation can be
sketched as in (17) (if DAT is represented as a feature on V; alternatively it could
be projected as a head, as in Koopman 2005): in (17a), dative would be licensed
because both Voice and DAT are present, whereas in (17b), dative would not be
licensed (cf. (4), (16)).3
(17) a. Active (dative) b. Middle (no dative)
8
Voice
Voice Init
Init VDAT
VDAT D
[DAT]
Init
Init VDAT
VDAT D
Accusative case is checked by V when it enters into an Agree relation with AgrO,
a marked value of Agr which is inserted when there are two DPs in vP which have
unchecked case features. Setting aside some exceptional cases to be discussed in
§5 (the ‘quirky unaccusatives’), the accusative case-checking value of Agr, AgrO,
is inserted only when needed; namely whenever a verb phrase contains two noun
phrases whose case is not licensed. In such cases, AgrO is merged and enters into
Agree with V, which then checks accusative. In this way AgrO has an aspect of
‘economy’ (Chomsky 1993), and the advantages of competition models of case
are captured (cf. Yip et al. 1987, Marantz 1991, Haider 2000, Sigur􀀀sson 2003,
Woolford 2003). The unmarked value of Agr enters into an Agree relation not
with V but with a DP with unchecked case, as discussed below.
If AgrO is structurally higher than Voice, then the relatively greater fragility of
accusative than dative can be accounted for in terms of the height of the relevant
licensers in the verb phrase: an Agr is inserted too late to affect the licensing of
dative, but early enough to affect the licensing of accusative.
(18) a. Active (accusative) b. Passive (no accusative)
AgrO
AgrO Voice
Voice Init
Init V
V D
[ACC]
Voice
Voice Init
Init V
V D
Thus, a basic feature of the analysis is that there is an accusative licenser higher
than any dative licenser, and the passive affects verb structure which is higher than
9
the layer that the middle affects. Also, in general, marked options such as middle
or passive reduce options higher up, rather than lower down.
Notice that if the lower head R, like V, can bear a DAT feature, then the combination
of these two basic types of internal case leads to accusative-before-dative
ditransitives like that illustrated in (7b) above.
(19) AgrO
AgrOi Voice
Voice Init
D
Initj V
D
[ACC] Vi RDAT
RDATj D
[DAT]
In general, when I discuss datives in this paper, I refer to dative themes, not experiencers.
Details about the licensing of case on experiencers will not be relevant
in the discussion to follow, so I will not pursue them here. However, there is one
final point which is of importance. Icelandic generally allows nominative assignment
and number agreement across a VP boundary, even when there is an external
(experiencer) argument (see e.g. Sigur􀀀sson 2006 for recent discussion and references).
For example, in (20), the nominative remains in VP but controls number
agreement with the finite auxiliary (see also (13a)).
(20) Henni
her.DAT
höfðu
had.3PL
alltaf
always
leiðst
bored
strákarnir.
the.boys.NOM
‘She had always found the boys boring’ (Sigur􀀀sson 2006)
Assuming that this verb projects an InitP, some part of the verb phrase spells
out before T is merged (adopting Chomsky’s 2001 assumption that experiencer v
heads a strong phase). This means that something lower than T must agree with
the nominative. Given the assumptions made above about case, Agr can check
number and case inside VP before the verb phrase spells out, as long as this is not
10
blocked by a defective intervener; the experiencer dative (whose case licensing I
set aside here) does not block the assignment because it is extracted to become the
subject. This is sketched in (21), assuming an EPP feature to be inserted in Agr,
as a free option.4
(21) Agr
Di
Agr
EPP
Init
ti Init V
V D
I assume that a phase spells out when all uninterpretable features in it are checked
(“Impatient Spell-Out,” Svenonius 2001, Holmberg and Hróarsdóttir 2003). Suppose
this to be true of InitP in (21). Then [InitP leiðst strákarnir] spells out. After
that, some higher head, suppose it is T, checks the features on Agr. If T is finite and
the features are nominative, then the derivation continues, leading to agreement on
the finite auxiliary (see Svenonius 2004:275–279). Thus, the Agr–V combination
only checks accusative under certain conditions; in general, it can be said to only
check accusative via V, when there are two DPs with unchecked case features in
the vP. Otherwise, it checks nominative directly on a DP.
This is important for the understanding of impersonal constructions, in which
nominative subjects remain inside VP, but trigger agreement outside it. If there
were no phase boundary at the edge of the verb phrase, one might expect such
agreement to be in person as well as number. Agr belongs to the adjectival family
of agreement heads and as such cannot be specified for person features.
3 Subject Demotion in Icelandic
3.1 The Eventive Passive
The passive construction that is the focus of this study is the periphrastic passive,
illustrated in (22) and (23).5 Finite verb agreement is glossed here in order to
facilitate comparison with the participial agreement.
(22) a. Höskuldur
Höskuldur.NOM
sannfærði
convinced.3SG
hana.
her.ACC
11
‘Höskuldur convinced her’
b. Hún
she.NOM
var
was.3SG
sannfærð
convinced.F.SG.NOM
af
by
Höskuldi.
Höskuldur.DAT
‘She was convinced by Höskuldur’ (Zaenen and Maling 1984)
(23) a. Við
we.NOM
kusum
elected.1PL
á.
them.M.ACC
‘We elected them’
b. Þeir
they.M.NOM
voru
were.3PL
kosnir.
elected.M.PL.NOM
‘They were elected’ (Sigur􀀀sson 1989)
The participial suffix can be analyzed as having the base allomorphs -ð (for weak
verbs like sannfæra ‘persuade’ in (22)) and -(i)n (for strong verbs like kjósa
‘choose, elect’ in (23)), plus an adjectival agreement paradigm, agreeing with
the promoted nominative subject in gender, number, and case. The agreement
paradigm for strong participles deviates slightly from the strong paradigm for underived
adjectives, but is identical to the paradigm for the definite suffix, which
can also be analyzed as having a base -(i)n.6 I will return to this identity in §4.1
below.
Datives never control participle agreement, as illustrated in (24); the participle
in such cases is identical to the neuter singular nominative form.7
(24) a. Mér
me.DAT
var
was
boðið
invited
í
in
veisluna.
the.party
‘I was invited to the party’
b. *Mér
me.DAT
var
was
boðn-um
invited-M.SG.DAT
í
in
veisluna.
the.party
Accusatives, however, do control participle agreement if they are in the right configuration
(as noted in Andrews 1982, Sigur􀀀sson 1989:309, n. 42). To see an
accusative DP controlling participle agreement, a passive clause can be placed under
an ECM (Exceptional Case Marking) verb like telja ‘believe’; the promoted
internal argument then gets accusative from the ECM verb, as in (25a). For comparison,
the same configuration is shown for a dative example with bjóða ‘invite.’
As can be seen in (25b), ECM accusative does not overwrite dative, determined
lower down. Unlike the accusative example, the dative does not control agreement.
12
(25) a. Hann
he
telur
considers
hana
her.ACC
vera
be
sannfærð-a.
convinced-F.SG.ACC
‘He considers her to be convinced’
b. Hann
he
telur
considers
henni
her.DAT
vera
be
boðið/*boðin-ni.
invited/invited-F.SG.DAT
‘He considers her to be invited’
In sum, only nominative and accusative can be mediated or valued by the Agr
head, which will also be responsible for participial agreement. This pattern is systematic,
and in the examples below I generally don’t gloss the agreement, except
when it is under discussion.
3.2 Adjectival Passives
The usual diagnostics for adjectival versus verbal (or stative versus eventive) passives
show that Icelandic participles can be used as predicative adjectives.8
For example, (26a) is ambiguously eventive or adjectival, expressing either
that the breaking event occurred yesterday, or that the state of being broken held
yesterday. The example in (26b), however, is unambiguously eventive, because
an agentive ‘by’-phrase is not possible with an adjectival passive.9
(26) a. Rúðan
the.window
var
was
brotin
broken
í
in
gær.
yesterday
‘The window was broken yesterday’
b. Rúðan
the.window
var
was
brotin
broken
af
by
skrílnum
the.mob
í
in
gær.
yesterday
‘The window was broken by the mob yesterday’ (Thráinsson 1999)
Other diagnostics positively identify the adjectival use. For example, certain modifiers
such as mjög ‘much, very’ are not possible with verbal participles, so that
(27a) has only an adjectival reading; and the prefix ó- does not combine with the
eventive use of the participle, so that (27b) is also adjectival only.
(27) a. Rúðan
the.window
var
was
mjög
very
brotin
broken
egar
when
húsvörðurinn
the.landlord
kom.
came
‘The window was very much broken when the landlord arrived’
b. Rúðan
the.window
var
was
óbrotin
unbroken
í
in
gær.
yesterday
‘The window was unbroken yesterday’ (Thráinsson 1999)
13
With these diagnostics in hand, it can be seen that although dative case is preserved
in eventive passives (as noted in the introduction), it is lost in the adjectival
passive. Take the verb loka ‘close,’ which takes a dative object, as illustrated in
(28a). The eventive passive is given in (28b): as expected, dative case is preserved,
a by-phrase is possible, the interpretation is eventive, and there is no agreement
on the finite verb or the participle.
(28) a. Einhver
somebody
lokaði
closed
dyrunum
the.doors.DAT
klukkan
the.clock
sjö.
seven
‘Somebody closed the door at seven o’clock’
b. Dyrunum
the.doors.DAT
var
was
lokað
closed
(af
by
dyraverðinum)
the.porter
klukkan
the.clock
sjö.
seven
‘The door was closed (by the porter) at seven o’clock’
Now compare (29). The adjectival passive (as diagnosed here by the target state
interpretation and the impossibility of a by-phrase) does not preserve dative case;
the theme turns up nominative, and controls agreement both on the finite verb and
on the adjectival participle.
(29) Dyrnar
the.doors.NOM
voru
were
lokaðar
closed.F.PL
(*af
by
dyraverðinum)
the.porter
klukkan
the.clock
sjö.
seven
‘The door was closed at seven o’clock’
To show that animacy is not relevant, consider also (30), which shows the same
pattern.
(30) a. María
María
bauð
invited
honum.
him.DAT
‘María invited him’
b. Honum
him.DAT
var
was
boðið
invited
(af
by
Maríu).
María
‘He was invited (by María).’ (eventive reading)
c. Hann
he.NOM
var
was
(ó-)boðinn
un-invited.M.SG.NOM
(*af
by
Maríu).
María
‘He was (un-)invited’ (adjectival reading)
Another diagnostic which will become relevant later is the possibility of an
instrument phrase. In Icelandic, instrument phrases are typically introduced by
the preposition með ‘with, as illustrated in (31). The eventive passive allows the
instrument phrase, while the adjectival passive does not.
14
(31) a. Rúðan
the.window
var
was
brotin
broken
af
by
skrílnum
the.mob
með
with
hamri.
hammer
‘The window was broken by the mob with a hammer’
b. Rúðan
the.window
var
was
mjög
very
brotin
broken
(*með
with
hamri).
hammer
‘The window was very much broken (*with a hammer)’
c. Rúðan
the.window
var
was
óbrotin
unbroken
(*með
with
hamri).
hammer
‘The window was unbroken (*with a hammer)’
The same fact is illustrated for the dative case, below.
(32) a. Dyrunum
the.doors.DAT
var
was
lokað
closed
með
with
fjarstýringu.
remote.control
‘The door was closed with a remote control’
b. Dyrnar
the.doors.NOM
voru
were
lokaðar
closed.F.PL
(*með
with
fjarstýringu).
remote.control
‘The door was closed (*with a remote control)’
In the introduction I suggested that the contrast between the dative and the accusative
was that the accusative was more fragile than the dative, and is lost first.
The basic idea in the analysis developed below is that the largest structures (transitive
verbs) support accusative and dative, medium-sized structures (eventive passives)
support only dative, and the smallest structures (adjectives and middles) do
not support either internal case.
3.3 Middles
There are a great many verb forms in -st in Icelandic (cf. Sigur􀀀sson 1989:259–
262, Anderson 1990), covering much of the same range of meanings as reflexive
forms in Romance, Slavic, and other languages (e.g. anticausative, reflexive, passive;
cf. e.g. Alexiadou and Anagnostopoulou 2004). Setting aside a wealth of
data, I will concentrate on the type which is illustrated in (33), where a transitive
verb base in the -st form has a kind of anticausative, passive, or mediopassive
meaning. The verbs bjarga ‘rescue’ and kasta ‘throw’ take dative objects, but
their middles have nominative subjects.
(33) a. Jón
Jón
bjargaði-st
saved-MID
úr
out.of
eldinum.
the.fire
‘Jón got saved from the fire’
15
b. María
María
kastaði-st
throw-MID
út
out
úr
out.of
bílnum.
the.car
‘María got thrown from the car’
c. Bjórinn
the.beer
kláraði-st.
finished-MID
‘The beer got finished up’
The term ‘middle’ is used ambiguously in the literature on Icelandic to refer to a
morphological class (e.g. Ottósson 1992) or to a semantically distinct contruction
(e.g. Sigur􀀀sson 1989), and I will follow the latter tradition; the Icelandic middle,
as used here, is an agentless voice with the -st suffix. It is distinct from the
English middle, which has a characterizing semantics and typically requires adverbial
support (These cakes sell easily; Stroik 1992, Ackema and Schoorlemmer
1995, Lekakou 2005).
Sigur􀀀sson (1989) points out several ways in which the middle is distinct from
the passive. For example, he shows that it does not normally allow ‘by’ phrases.10
(34) a. Lögreglan
the.police
drap
killed
hundinn.
the.dog
‘The police killed the dog’
b. Hundurinn
the.dog
var
was
drepinn
killed
(af
by
lögreglunni).
the.police
‘The dog was killed (by the police)’
c. Hundurinn
the.dog
drap-st
killed-MID
(*af
by
lögreglunni).
the.police
‘The dog got killed’ (Sigur􀀀sson 1989:268)
Similarly, the middle, unlike the eventive passive, is incompatible with agentoriented
adverbials (Sigur􀀀sson 1989:268). The contrast can be seen in (35).
(35) a. Bíllinn
the.car
var
was
seldur
sold
(viljandi)
intentionally
(af
by
bílasala).
car.salesman
‘The car was sold (intentionally) (by a car salesman)’
b. Bíllinn
the.car
seldi-st
sold-MID
(*viljandi)
intentionally
(*af
by
bílasala).
car.salesman
‘The car got sold’ (Hrafnbjargarson 2005)
As already noted in the introduction, the middle causes not only accusative but
also dative case to be lost, just like the adjectival passive. This is illustrated again
below, for completeness. First, (36a) shows that læsa ‘lock’ takes a dative object.
16
Second, (36b) shows that the theme is nominative in the middle voice.
(36) a. Við
we
læsum
lock
dyrunum.
the.doors.DAT
‘We are locking the doors’
b. Dyrnar
the.doors.NOM
læsa-st!
lock-MID
‘The doors are locking!’ (Hrafnbjargarson 2005)
For comparison, (37) shows that the dative cannot be preserved in the middle, but
is preserved in the passive (note that the participle ending is -t after the s in the
root, creating a sequence st in (37b) which is nonetheless not an -st form).
(37) a. *Dyrunum
the.doors.DAT
læsa-st!
lock-MID
b. Dyrunum
the.doors.DAT
verður
become
læst!
locked
‘The doors are being locked!’ (Hrafnbjargarson 2005)
So far, then, the middle has the same properties as the adjectival passive: it does
not imply an agent, and it does not license dative or accusative case on its complement.
There is, however, another diagnostic which distinguishes the middle from the
adjectival passive. Alexiadou et al. (2006) show that certain PPs which presuppose
a causing event can be used to distinguish subclasses of agentless events. One such
PP is the instrumental, which was already shown in §3.2 to be possible with the
eventive passive and impossible with the adjectival passive.
(38) and (39) show that instrumentals are possible with the middle. Unaccusatives
are used to show the contrast.
(38) a. Drullukakan
the.mud.cake
bakaði-st
baked-MID
með
with
rafmagni.
electricity
‘The mud cake was baked using electricity’
b. Drullukakan
the.mud.cake
harðnaði
hardened
(*með
with
rafmagni).
electricity
‘The mud cake hardened (*using electricity)’
(39) a. Plakatið
the.poster
hengdi-st
hung-MID
upp
up
á
on
vegginn
the.wall
með
with
hamri.
hammer
‘The poster was hung up on the wall with a hammer’
17
b. Plakatið
the.poster
hékk
hung
upp
up
á
on
veggnum
the.wall
(*með
with
hamri).
hammer
‘The poster hung up on the wall (*with a hammer)’
As with by-phrases, this test must be applied carefully. For example, a PP translating
as ‘with tape’ (með límbandi) would be acceptable in (39b). The kind of
instrumental which distinguishes between adjectival passives and middles is the
kind which implies a causing event. Thus there is a sense in which the middle is
causative, like the eventive passive, but not so much so that the external argument
can be introduced explicitly in a by-phrase.
3.4 Nominalizations
I will not discuss nominalizations in this chapter, beyond what is stated in this
section. However, they provide an important backdrop for the analytic approach
taken here, because of the way in which different kinds of nominalizations show
the effects of embedding different sizes of verbal structure under a nominalizer
(Abney 1987, Alexiadou 2001).
Here it is sufficient to point out that dative and accusative case on theme objects
is systematically lost in nominalizations in Icelandic, as discussed by Maling
(2001) (judgments from Halldór Ármann Sigurðsson, personal communication).
(40) a. Sjómennirnir
the.sailors
björguðu
rescued
flóttamanninum.
the.refugee.DAT
‘The sailors rescued the refugee’
b. björgun
rescue.N
flóttamannsins
the.refugee.GEN
‘rescue of the refugee’
c. ?björgun
rescue.N
sjómannanna
the.sailors.GEN
‘the sailors’ rescue’
On the assumptions here, the nominalizer -un takes a smaller verb phrase than the
eventive passive, since not even dative case is licensed. The possibility of the external
argument here (in (40c)) might suggest that such nominalizations are larger
than the middle or the adjectival passive, but it is also possible that possessors have
such a wide range of interpretations that they do not clearly indicate the presence
of verbal argument structure (cf. Grimshaw 1990).
18
4 The Decomposition of the Verb Phrase
Given the preceding discussion, the different forms can be arranged in a hierarchy
according to the case properties (Nominative is not included in the table; it is
sensitive to finiteness but not to these features).
(41) Acc Dat
ACTIVE yes yes
EVENTIVE PASSIVE no yes
MIDDLE no no
ADJECTIVE no no
Still concentrating on transitive verbs, these same forms can also be arranged
around other properties such as implication of an agent or instrument.11
(42) (Implicit) Agt Instrument
ACTIVE yes yes
EVENTIVE PASSIVE yes yes
MIDDLE no yes
ADJECTIVE no no
I suggest that these two patterns can be unified and explained if case and thematic
roles are both linked to the introduction of subevents. Roughly, each verbal
head introduces a subevent, and each subevent can introduce one thematic role
(cf. Déchaine 1992); each case is checked by a chained pairing of a subevent head
with some other head, as suggested in §2.2.
The tables in (41) and (42) can be combined as in (43); the (implicit) agent and
the presence of dative do not give distinct results for these four categories, so they
are combined. The categories argued in §2.2 to be involved in the assignment
of VP-internal case are added to the column labels. A row is added for a verb
phrase type ‘Non-agentive’ which has Init, V, and R but not Voice, though there
is not space to discuss such structures here. In brief, they are transitive structures
with non-agentive verbs which cannot license dative themes or passivize, but they
have AgrO because they have two DP arguments one of which requires accusative
licensing.
19
(43) AgrO Voice Init V R
Case/Theta: Acc Dat Init- Under- Holder
iator goer
ACTIVE yes yes yes yes yes
EVENTIVE PASSIVE no yes yes yes yes
NON-AGENTIVE yes no yes yes yes
MIDDLE no no yes yes yes
UNACCUSATIVE no no no yes yes
ADJECTIVE no no no no yes
This can be compared to the analysis of English participial constructions in Embick
(2003; 2004), in which it is proposed that an adjectival participle, a resultant
state passive participle, and an eventive passive participle are derived by merging
a head ASP with successively larger structures.
The analysis draws more heavily on the analysis of verb structures in Ramchand
(2006), where the verb phrase is decomposed intomaximally three subevents:
init[iation], proc[ess], and res[ult]; it is from Ramchand that the thematic role labels
Initiator, Undergoer, and Result are taken. Ramchand, like Alexiadou et al.
(2006), uses adjuncts as probes for implicit subevents. I use the more traditional
label V for her proc and the shorter abbreviation R for her res.
The properties at issue here are arranged in a (partial) implicational hierarchy:
the availability of accusative (AgrO) implies the presence of two arguments, and
hence (normally) a causing event Init. However, AgrO does not strictly speaking
imply the presence of Voice, since non-agentive verbs can lack Voice but have
accusative case. Voice, however, implies the presence of a causing event, which
in turn implies the presence of a process event (V). The implications do not go
the other way; a process event may exist without a causing event (unaccusatives
and middles), and a causing event may exist without the availability of accusative
(passives). I include the Result head R in the table, but it seems that R may in fact
be absent from a normal transitive verb (see Ramchand 2006).
However much of the verb phrase projects, the possibility of combination of
the verb phrase with tense, aspect, and modality is essentially unchanged. That
is, whether a verb phrase is passive, unaccusative, middle, or active transitive, it
can combine with perfect aspect, past tense, or any modality or mood. I assume
that this is because the verb phrase spells out separately from the material in the
T-domain, as a phase (Chomsky 2000; 2001; 2005). For instance, if a maximal
clause consists of the categories in (44a), then a plausible assumption is that the
other licit clause types are at most those in (44b–d) (cf. Bresnan 1970, Stowell
20
1981).
(44) a. Force–Fin–T–Asp ... (finite main clause)
b. Fin–T–Asp ... (finite subordinate clause)
c. T–Asp ... (infinitive clause)
d. Asp ... (small clause)
Of course, if the hierarchy is richer (Cinque 1999), then there are more possibilities;
also, there may be subconstituents that cannot stand alone. Each category
can be assumed to have an unmarked value, a default interpretation which comes
about if no marked value is inserted (see Ramchand and Svenonius to appear).
In any case, applying the same logic to the verb phrase, if a maximal verb consists
of the elements in (45a), then every possible verb phrase consists of subsets
of this hierarchy as in (45b–d) (cf. Wurmbrand 2001).12
(45) a. Voice–Init–V–R (agentive verb)
b. Init–V–R (non-agentive transitive verb)
c. V–R (unaccusative verb)
d. R (stative verb)
For the purposes of this paper I will assume that all and only the sequences in
(45) can be embedded under Asp, and that the projection of the T-domain proceeds
independently of the size of the structure merged with Asp. I set aside the
question of what projections there are to host successive-cyclic A0 movement out
of vP (using the label vP now as a cover term for the verb phrase that spells out)
(cf. Belletti 2004 on topic and focus projections at the edge of vP).
I also assume that a phase spells out when all features in it are checked (Svenonius
2001), and that the edge is the space between the node that spells out and the
last head merged above it before it spells out (Svenonius 2004). Thus the edge
may consist of several heads, depending on what heads are in the hierarchy and
what features need to be checked. For example, if the last remaining unchecked
features in VP are checked when Asp is merged, then VP spells out and Asp,
Voice (if present), and Init (if present) constitute the edge.
I will illustrate how the system here works by stepping through some structures,
starting with actives and then moving to middles and passives.
21
4.1 Active Unaccusatives
A simple present active clause consists mainly of unmarked values for functional
heads. Being unmarked, they generally need not be filled with lexical material.
Take for example the sentence in (46).
(46) Það
it
dýpkar.
deepens
‘It deepens’
The root is dýp- ‘deep’ (cf. djúpur ‘deep,’ dýpi ‘depth’), and it has a verbal suffix
-ka which is used to form unaccusatives. For concreteness assume that -ka is
the spell-out of a V head which selects a resultative complement R (for Result,
cf. Ramchand and Svenonius 2002; Ramchand’s 2006 res), so that the lowest part
of the structure underlying (46) includes the root merged with R. R introduces an
argument position which must be saturated, so D is inserted.
This can be represented as in (47a), or as the equivalent (47b), representing
the heads R and D only once each, for simplicity, following Brody (2000).
(47) a. RP
R DP
D
b. R
D
I will adopt the latter and will add marked values for heads, when they appear, as
dependents with small capital terminals.
V is merged to R, and it introduces an argument position as well. Assume that
D is remerged, which I represent as movement leaving a coindexed trace. This
means that the DP moving is interpreted both as undergoing a process (by being
merged with V) and being affected or acquiring some property (by merging with
R) (see Ramchand and Svenonius 2002, Ramchand 2006 for discussion of the
conditions on remerging in theta positions).
(48) V
Di R
ti
22
I assume that V has uninterpretable temporal-aspectual features which must be
checked, and that D does as well. I also assume that the functional sequence
includes heads T, Agr, and Asp. Each has an unmarked value in the sentence in
(46). In general, marked values can be inserted to express meaning (e.g. past tense
in T) or to rescue a derivation (e.g. accusative case in Agr, only if needed).
(49) T
Di Agr
Asp
V
ti R
ti
Agr probes vP and enters into Agree with DP. Asp checks temporal features on
V. I suggested in §2.2 that a combination of two eventive heads could license a
case; suppose that in this case, Nominative case is checked on D by the chain
T–Agr. D is attracted by EPP features on T and spells out as Það, and its internal
structure is unavailable for further operations (though it can still be attracted for
A0 movement to a topic position, as long as it is not trapped within a phase).
Assuming obligatory V to T, the verb does not spell out before T is merged, and
then as dýp- with V -ka and T -r suffixed.
An unaccusative built from a dative-taking root has a marked value for V (recall
VDAT from §2.2), but cannot license dative in the absence of Voice (cf. (45a)).
The tree for (50) (cf. (11) in §2.2) is given in (51).
(50) Það
it.NOM
sekkur.
sinks
‘It sinks’
23
(51) T
Di Agr
Asp
V
ti DAT R
ti
If Init is not merged in to introduce an external argument, then the configuration
licensing dative does not arise. Nor does the configuration licensing accusative,
since the default value for Agr is inserted unless a non-default value is necessary
to distinguish two DPs, as detailed below. The default value of Agr probes for
a DP with unchecked case, finds DP, and coindexes with it, as with the previous
example.
In the perfect, there is a participial ending and an auxiliary hafa ‘have.’
(52) Það
it
hefur
has
dýpkað.
deepened
‘It has deepened’
The derivation proceeds as before up to the merge of V, but then a marked value
of the Asp head, corresponding to the participle, is merged. I label it ‘SPFC’
for ‘specific,’ as it picks out a contextually salient reference time, much like a
specificity marker (cf. Ramchand’s 2004 analysis of Russian perfectivity in terms
of definiteness).
(53) Asp
SPFC V
Di R
ti
The participial Asp also checks temporal features on V. I also assume, following
Ramchand and Svenonius (2004), that the participial projection is interpreted as
a predicate; predicate abstraction of the participle is achieved through choosing a
24
marked option in Agr mentioned previously, namely the EPP feature.
(54) Agr
Di
EPP Asp
SPFC V
ti R
ti
This means that there are no unchecked features remaining in AspP, and the predicate
spells out, in this case as dýpkað: the root as dýp-, V as -ka, and the participle
as -ð, which is the specificity marker; in the nominal domain it is interpreted as
definiteness, and in the verbal domain as reference to a specific result state.
Additional material must be inserted in order to support tense if this participle
is to appear as a predicate in a clause. I will assume that a verb corresponding to
‘be’ is inserted in T, and that T attracts DP.
(55) T
Di
BE Agr
ti
EPP ASP
dýpkað
Once T has checked the case on D, D can spell out. Once C is merged, D is
attracted to it, and the C–T–Agr complex spells out as hefur, as seen in (52) above
(cf. Kayne 1993 on the derivation of auxiliary ‘have’ from ‘be’ plus a nominal
element).
4.2 Active Transitives
Now consider a transitive example.
(56) Hann
he.NOM
litar
colors
hana.
her.ACC
25
‘He colors it’
The root is lit- ‘color,’ which is conventionally used as a noun (litur ‘color’) or
as a transitive resultative verb: as a verb (with -a), it implies that someone did
something to cause something else to change color. Thus, its lexical semantics
demand initiation, process, and result, which means that Init, V, and R must be
projected.
Starting with a D with a feminine feature F (because the object in (56) is
feminine), and merging this with R, then merging R with V and remerging D with
V, we get the following structure.
(57) V
Di
F
R
ti
When Init is merged, another argument is introduced, giving something like the
following, assuming that a masculine third person singular pronoun is chosen.
(58) Init
D
M
V
Di
F
R
ti
Now the pronouns have unchecked case features, and the verbal complex has
unchecked temporal-aspectual features, so no part of the structure can spell out
yet.
If the derivation were to proceed as with the unaccusative described above,
so that an unmarked Asp, Agr, and T were to merge with this structure, it would
eventually crash, as there would be no way to case-mark the object (assuming T
could case-mark only the subject).
Instead, a marked value of a projection is inserted which can license V to assign
accusative case, in the way discussed in §2.2. Following the spirit of Frampton
and Gutmann (2002), I assume that the insertion of AgrO is conditioned by
the presence of more than one DP with unchecked case features (in this way the
wrong value of Agr is not chosen, so later crash is avoided, but without lookahead).
I also assume that an agentive structure includes a Voice projection, as
26
shown. Finally, I assume that AgrO has an EPP feature and therefore attracts the
closest DP to its specifier.
(59) Agr
Dj
M
AGRO Asp
Voice
tj Init
tj V
Di
F
R
ti
The AgrO–V chain checks case on the internal argument. Recall that a V with
the DAT feature would check dative without a need for the AgrO head. At this
stage, VoiceP can spell out as a phase, but containing nothing but the object (since
Icelandic has verb movement).
Now consider the same structure but with the ‘specific’ (participial) value for
Asp.
(60) Hann
he.NOM
hefur
has
litað
colored
hana.
her.ACC
‘He has colored it’
27
(61) Agr
Dj
M
AGRO Asp
SPFC Voice
tj Init
tj V
Di
F
R
ti
The important difference here is that the participial Asp values temporal-aspectual
features on the verbal complex, unlike the default Asp. Once AgrO checks the
accusative case on the object and removes the subject, then, all features in Asp are
checked, and so the predicate can spell out as the participial litað hana (cf. (60)).
Agr–T–C spell out together as hefur, as before.
4.3 Passive
Passive in Icelandic suppresses the external argument. I assume the analysis of
English passives in Ramchand and Svenonius (2004) to be correct in outline for
Icelandic as well, namely that a crucial component of this type of passive is the
introduction of a marked value of Voice, PASS, which binds the external argument
introduced by Init; the only expression of the external argument which is compatible
with PASS is a null operator. Any other DP inserted in SpecInitP will lead to
a crash. So at a first approximation, the sentence in (62) contains the structure in
(63).
(62) Hún
she.NOM
er
is
lituð.
colored.F.SG.NOM
‘It is colored’13
28
(63) Voice
Opj
PASS Init
tj V
Di
F
R
ti
The passive in Icelandic, like the passive in English, requires the participial form.
This is a language-specific fact about PASS, since languages like Chichewa have
non-participial passives (Dubinsky and Simango 1996). This means that the participle
must merge outside PASS.
(64) Asp
SPFC Voice
Opj
PASS Init
tj V
Di
F
R
ti
The Asp and Voice heads together can be assumed to license a case, which the null
operator receives; this case is like the ergative case of ergative languages in being
assigned to an external argument. Recall that AgrO attracts a DP; the nearest DP is
the null operator. This presumably happens in case there is no internal argument,
namely in impersonal passives, such as those in (12a) above or (65) below.
(65) a. Páll
Páll
söng
sang
hátt.
loudly
‘Páll sang (loudly)’
b. Það
it
var
was
sungið
sung
hátt.
loudly
‘There was loud singing,’ ‘People sang loudly’ (Sigur􀀀sson 1989:308)
29
Instead, when there is an internal argument, some strategy must be found to allow
the internal argument to be targeted across the external argument (see Collins
2005). A salient distinction between the Icelandic perfect and the passive is the
presence in the latter of agreement. I hypothesize that this is related to the fact
that the passive allows one DP to undergo A-movement across another. The tree
in (64) can accordingly be expanded as in (66).
(66) Agr
Di
F
EPPi Asp
SPFC Voice
Opj
PASS Init
tj V
ti R
ti
Agr can either attract Di with the EPP option, for a typical promotion passive like
this one, or leave it below for an impersonal passive. Either way, it transmits a
core case; nominative here, because there is a finite clause, and there is no need to
distinguish the two DPs, since Op has ergative.
This means that the predicate no longer contains any uninterpretable features,
and can spell out, as lituð: the root lit-, the V -a, the participial suffix -ð, and a
feminine singular agreement morpheme which is realized here as a vowel change
in a !u (cf. M.SG.NOM gamall  F.SG.NOM gömul ‘old’).
(67) Agr
Di
F lituð
Note that since Agr is spelled out here as agreement, it cannot be used to form
‘have,’ so the auxiliary must spell out as ‘be’ (cf. Iatridou et al. 2001:222–223).
30
4.4 Adjectival Passive
Following Embick (2004), adjectival passives involve the same aspectual head as
eventive passives, but merged on top of a subset of the structure. Since adjectival
passives do not imply the occurrence of any process, I assume the participial Asp
can merge on R. I assume that R in these cases has its usual argument structure
(cf. Cinque 1990, Bennis 2004 on internal arguments in certain kinds of adjectives).
(68) Hún
she.NOM
er
is
lituð.
colored
‘It is colored’
(69) Asp
SPFC R
D
F
This structure merges with Agr, which ultimately spells out as adjectival agreement.
(70) Agr
Di
F
Asp
SPFC R
ti
The various properties of adjectival passives noted in §3.2 follow straightforwardly.
There will be no possibility of a by-phrase or of agent-oriented adverbials,
as there is no Init. There is no possibility of instrument phrases which presuppose
the occurrence of an event, either, and there is no temporal reference because there
is no process head V (following Ramchand 2006, R is stative). Importantly, dative
case will also be unavailable, regardless of the properties of R, because there is no
Init.
31
4.5 Middle
I will not be able to do justice to the Icelandic middle here. For extensive discussion
of the Icelandic middle and other -st forms, see Valfells (1970), Sigur􀀀sson
(1989:ch. 6), Anderson (1990); for recent discussion of anticausatives, see various
of the papers in Alexiadou et al. (2004).
A middle based on the verb lita ‘color’ is given in (71).
(71) Hún
she.NOM
lita-st.
color-MID
‘It turns color’
Recall that the middle has no implication of a causer, but does carry an implication
of a causing event (therefore allowing an instrumental phrase).
I take the middle to be a marked option at the level of Init, a specifier [ID MID]
(adopting Cardinaletti and Starke’s 1999 notation ID for the category of pronominal
clitics).
(72) Init
ID
MID
V
Di
F
R
ti
The meaning contribution is, roughly, that the event was caused in some way that
is dissociated from intent; it is not necessarily inadvertent, but is incompatible
with salient agency.
That the middle morpheme is a specifier of Init is suggested by several considerations.
First, it cannot be a V head because it combines with dative-licensing
roots, as shown above in (4), or below in (73). I have argued that dative-licensing
roots combine with a marked value for V. Assuming that there is a small number
of discrete marked values for V, and that these values cannot freely combine,
middle could not be a V head.
(73) a. Ásdís
Ásdís
kastaði
threw
spjótinu
the.javelin.DAT
53,15
53.15
metra.
meters
‘Ásdís threw the javelin 174.38 feet’
b. Skipverjinn
the.crewman.NOM
kastaðist
threw.MID
fyrir
for
borð.
board
32
‘The crewman was thrown overboard’
Second, it could not be an Init head because there is evidence that the highest verbal
head normally determines the conjugation class to which a verb belongs; for
example many Icelandic verbs which have strong unaccusative forms, but weak
transitive forms (cf. Zaenen and Maling 1984). This can be explained if there
is a causative Init which is weak conjugation and can combine with those verbs.
The middle attaches to strong verbs which remain strong (as with e.g. drepa ‘kill’
in (34c) or bjóða ‘ask, offer’ in (15b)) and to weak verbs which remain weak
(cf. (4b), (16b) or (73b)).
Despite the evidence that the middle is not an Init head, there is reason to think
that the middle merges in the Init projection. The verbs with which it combines
are often obligatorily transitive, and are expected to project Init; the instrument
phrases introduced in (38)–(39) in §3.3.
Third, the middle morpheme (-st) is enclitic, attaching outside finite tense and
agreement morphology.
(74) a. Maðurinn
the.man
kalla-ð-i-st
call-PAST-SG-MID
Jón
Jón
Glón.
Glón
‘The man was called Jón Glón’
b. Vanir
Vanir
köllu-ð-u-st
call-PAST-PL-MID
goðin
the.gods
í
in
norrænni
Norse
goðfræði.
mythology
‘The gods were called Vanir in Norse mythology’
If the middle is a specifier of Init, then it can combine either with V or VDAT. The
example in (73b) would have a structure something like the following (if there is
a result component to kasta ‘throw,’ a non-essential assumption).
(75) Init
ID
MID
V
Di DAT R
ti
The specifier is filled, and no external argument can be merged. The middle
cannot be combined with Voice; in this way it is like an unaccusative or nonagentive
causative, but unlike an obligatorily transitive verb such as (non-middle)
lita ‘color.’ This prevents dative case from being checked on a theme argument,
33
which gets nominative instead, after T-domain material is merged (with default
values in the case of (71)).
(76) T
Di
F
Agr
ti Asp
Init
ID
MID
V
ti R
ti
The middle can be combined with the participle to form a perfect, as noted by
Anderson (1990).
5 Quirky Unaccusatives
Ordinary unaccusatives lack Init (and Voice) not because of marked morphology
but because of the meaning of the root and the way in which the event is characterized.
As I have mentioned, normal unaccusatives in Icelandic do not assign
case to their sole internal argument, which surfaces as a nominative subject (or an
accusative, if placed under an ECM verb, for example, or PRO in a control structure,
etc.). Unaccusatives can form participles, though they cannot form middles
or passives, because of the absence of Init, and their participle forms can be used
as adjectives or can be embedded under the perfect, as might be expected.
Icelandic also has what can be called quirky unaccusatives; these are verbs
with a single theme argument that shows up in a non-nominative case, usually
dative or accusative. An example of this was given in §1; here are two causativeinchoative
pairs.
(77) a. Stormurinn
the.storm.NOM
rak
drove
bátinn
the.boat.ACC
á
on
land.
land
‘The storm drove the boat onto land’
b. Bátinn
the.boat.ACC
rak
drove
á
on
land.
land
34
‘The boat drifted onto land’ (Zaenen and Maling 1984)
(78) a. Jón
Jon.NOM
lauk
finished
sögunni.
the.story.DAT
‘Jon finished the story’
b. Sögunni
the.story.DAT
lauk.
finished
‘The story ended’ (Zaenen and Maling 1984)
Given the analysis of case I have presented here, each of these verbs must contain
a pair of verbal or aspectual heads which form a chain. Assuming that the lower
head in both cases is V (VDAT in the dative example), the question is what the
higher head is. I suggest that in both cases the verb is idiosyncratically combined
with a marked value of Init, one which implies causation by a natural force or by
an internal property of the theme. The Init appearing with the accusative verbs
can license accusative case by entering into Agree with V; in the case of the dative
verbs, Voice can merge as usual and license the dative.
As for the accusative, the conditions for insertion of AgrO are not met, as there
are not two DPs to be distinguished. I also assume that material outside vP cannot
be included in an idiomatic lexical entry (Svenonius 2005), so that AgrO cannot
simply be listed together with the verb root.
Thus, these verbs must contain Init; but ordinarily Init introduces an external
argument. Hence, this must be a marked value of Init, one that is listed along with
these verbs much in the way an idiom is listed. Is there any independent evidence
for this Init, apart from the needs of my analysis of case? There are in fact at least
two pieces of evidence.
The first is based on an observation by Zaenen and Maling (1984). They note
that although many causative-inchoative pairs in Icelandic have differing conjugation
classes, the quirky unaccusatives never do. On the assumption that the
conjugation class is determined by Init, when present, and that there is a weak
conjugation Init which is combined with other strong verbs, this suggests that the
quirky unaccusatives cannot combine with this causative head, something which
is most simply explained if they already have an Init projection.
Second, for many of these verbs there is a salient sense that something is
caused by natural forces or by internal properties. Sigur􀀀sson (to appear) divides
the accusative-taking quirky unaccusatives into two sets, the psych accusatives
(with essentially experiencer subjects) as in (79a) and the ‘fate’ unaccusatives
(with theme subjects) as in (79b).
35
(79) a. Mig
me.ACC
langar
longs
heim.
home
‘I yearn to go home’
b. Mig
me.ACC
tók
took
út.
out
‘I was swept overboard’
Eythórsson (2000) and Jónsson (2003) show that experiencer accusatives tend to
change to dative in modern Icelandic, while theme accusatives tend to change to
nominative, underscoring the difference between them. Sigur􀀀sson (to appear)
argues that the fate unaccusatives have a specific sense of uncontrolled causation,
especially by natural forces. I assume that this sense is contributed by a special
idiomatic Init which is listed together with these verbs, and which has the property
that it can bind V, thereby providing an internal case even in the absence of Voice
and AgrO. As before, I assume that accusative is the case which is licensed by V
when the binder of V is aspectually isomorphic to it. This Init is exceptional in
not introducing an external argument.
Quirky unaccusatives with dative themes can also be assumed to have an idiomatic
head, presumably Init, with the necessary aktionsartal properties to license
dative case. This makes these datives similar to the benefactive datives mentioned
in §2.2.
6 Conclusion
I have analyzed passives and middles in terms of a fine-grained decomposition of
the verb phrase, one which makes use of minimalist ideas of feature-checking and
cyclic spell-out. This account provides a framework in which to understand the
case alternations observed in Icelandic, which are complex but systematic. The
core observation is that case licensing in the Icelandic verb phrase is sensitive to
thematic and event-related (aktionsartal) semantics, but not so straightforwardly
that a given thematic role is reflected directly by a given case.
If the analysis is correct, it gives a probe on the material right at the juncture
between the vP phase and the T-domain, where voice-related phenomena are
encoded.
The model relies on a functional hierarchy which is presumably fixed for all
languages, one which distinguishes a vP phase and a T-domain. The vP consists
of some combination of the categories Voice, Init, V, and R; in general the
higher categories imply the presence of the lower ones. These categories allow
36
the characterization of different subevents and confer thematic interpretations on
DPs merged in their specifiers; in the details I largely follow Ramchand (2006).
I have maintained a careful separation of phonology and syntax, and a similar
separation of conceptual material from syntax, assuming late insertion of lexical
items which bear phonological and conceptual information. I have furthermore assumed
that the categories Voice, Init, V, and R have marked values which do have
syntactic consequences; Passive is a marked value of Voice and Middle a marked
value of Init. These marked values show some cross-linguistic variation and can
be thought of as functional heads, distinct from the phonological-conceptual lexical
items that are associated with them.
The same architectural assumptions apply to the T-domain, where there are at
least the heads Asp, Agr, and T, and doubtlessly many others (Cinque 1999). In
the T-domain, as in the vP, higher structure is dependent on the presence of lower
structure, so that every TP can be assumed to contain Asp but not vice versa. This
assumption constrains the range of combinatoric possibilities and allows a simple
treatment of category selection, for example one in terms of categorial grammar.
I assume that for each category, there is a default interpretation. Any marked
interpretation must be forced by the insertion of lexical material (though lexical
material need not necessarily have any phonological content, if it is learnable
indirectly). For example, if no value is specified for Asp, then reference time
is assumed to be identical to event time, in Reichenbachian terms. Any more

complex aspectual value must be specified.
This model has allowed me to link the conditions on case assignment in the
verb phrase in Icelandic to the assignment of thematic roles and the characterization
of event structure. Many details remain to be understood, for example the
exact properties of the mechanisms of feature-checking and the spelling out of
phases, but I hope to have shed at least a little light on the semantic contribution
of the passive and the middle in an exoskeletal model (Borer 2005a;b) with no
lexical or morphological component for rules that can affect argument structure.
Acknowledgments
I am greatly indebted to three people for detailed comments on a previous draft, and very helpful
discussion: Halldór Ármann Sigurðsson, Jóhannes Gísli Jónsson, and Gunnar Hrafn Hrafnbjargarson;
an anonymous reviewer also provided very useful feedback. Thanks also to Þorbjörg Hróarsdóttir
for additional assistance with the Icelandic. I have furthermore benefitted from discussions
with Gillian Ramchand, Michal Starke, Klaus Abels, Pavel Caha, and audiences in Reykjavík,
Vienna, Lund, and Tromsø in 2004-6. Finally, thanks to the editors for creating this volume and
including this paper in it.
37
Notes
1On the separation of Voice and causation, see Pylkkänen (2002); see also Doron (2003) on
the difference between causative and agentive verbs in Hebrew.
2The feature DAT could be formalized in terms of an uninterpretable feature Voice which must
be checked by interpretable Voice; a DP coindexed with V with checked Voice features spells out
as Dative. See Frampton and Gutmann (2000) for some discussion of the checking of uninterpretable
features by uninterpretable features, in a way that strives to avoid look-ahead.
3For the purposes of this analysis, the DAT feature could be lower than V, cf. the analyses of
ditransitives in Pesetsky (1995), Harley (1995), Pylkkänen (2002).
4 Cf. Alexiadou and Anagnostopoulou (1998) on the externalization of arguments from vP. As
noted in Svenonius (2001) and Holmberg (2002), Icelandic does not allow the order DP-Participle
in passives; thus EPP on Agr requires EPP on T, in Icelandic, perhaps via Agree.
5Much of the discussion to follow is based on the detailed and careful exposition in Sigur

sson
(1989).
6As pointed out to me by Tarald Taraldsen.
7Participles may have dative forms when used as attributive adjectives, e.g. peysa með smelltri
hettu ‘sweater with snapped.DAT hood.DAT.’
8See for example Wasow (1977), Dubinsky and Simango (1996), Marantz (2001). I will not
have occasion to make use of the distinction between eventive passives and what Kratzer (2000)
calls Resultant State passives (Embick 2004 calls them ‘resultative’). The adjectival passives here
can be equated with Kratzer’s Target State passives, or Embick’s stative passive.
9The expression í gær translates as ‘yesterday’; the preposition is not part of a verbal complex.
10Sigur

sson (1989:268) notes that the by-phrase af sjálfu sér ‘by itself,’ with the meaning that
something was not caused at all, has the opposite distribution: acceptable with middles, unacceptable
with passives (with passives, a literal, reflexive meaning is possible). In this respect middles
are like unaccusatives (cf. The door opened by itself ) (though see Chierchia 2004, in which the
distribution of such phrases is used as an argument that unaccusatives are underlyingly causative).
Gunnar Hrafn Hrafnbjargarson has pointed out to me a couple of examples of -st verbs which
allow af -phrases, including the following:
(i) Ísland
Iceland
byggðist
built.MID
af Norðmönnum
by Norwegians
‘Iceland was built by Norwegians’
I will assume that by-phrases can in certain cases be introduced without linking directly to the
external argument provided by the verb; for present purposes it is sufficient that the appearance of
by-phrases is much more restricted in middles than in eventive passives.
11This general way of approaching the problem has been explored at length by Michal Starke,
to whose Nanosyntax seminars in Tromsø in 2005-6 this analysis owes a great deal.
12Ramchand (2006) argues that verb can have process but lack result, for example ‘roll’ or
‘dance.’ The stricter hierarchy here would have to allow for a very bleached notion of “result,”
with a threat of vacuity.
13This string has a salient reading as an adjectival passive, like its English translation–cf. (68).
The eventive reading can be brought out by using the past tense, Hún var lituð, ‘She/It(FEM) was colored.’
38
References
Abney, Steven. 1987. The English Noun Phrase in its Sentential Aspect. Ph.D.
thesis, MIT, Cambridge, Ma.
Ackema, Peter and Maaike Schoorlemmer. 1995. Middles and non-movement.
Linguistic Inquiry 26: 173–197.
Alexiadou, Artemis. 2001. Functional Structure in Nominals: Nominalization
and Ergativity. John Benjamins, Amsterdam.
Alexiadou, Artemis and Elena Anagnostopoulou. 1998. Parametrizing AGR:
Word order, V-movement, and EPP-checking. Natural Language & Linguistic
Theory 16.3: 491–539.
Alexiadou, Artemis and Elena 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, edited by Artemis Alexiadou, Elena Anagnostopoulou, and
Martin Everaert, pp. 114–136. Oxford University Press, Oxford.
Alexiadou, Artemis, Elena Anagnostopoulou, and Martin Everaert (eds.) . 2004.
The Unaccusativity Puzzle. Oxford University Press, Oxford.
Alexiadou, Artemis, Elena Anagnostopoulou, and Florian Schäfer. 2006. The
properties of anticausatives crosslinguistically. In Phases of Interpretation,
edited by Mara Frascarelli, pp. 187–211. Mouton de Gruyter, Berlin.
Anderson, Stephen R. 1990. The grammar of Icelandic verbs in -st. In Modern
Icelandic Syntax, edited by Joan Maling and Annie Zaenen, vol. 24 of Syntax
and Semantics, pp. 235–273. Academic Press, San Diego, Ca.
Andrews, Avery D. 1982. The representation of case in Modern Icelandic. In The
Mental Representation of Grammatical Relations, edited by Joan Bresnan, pp.
427–503. MIT Press, Cambridge, Ma.
Bar􀀀dal, Jóhanna. 1993. Accusative and dative case of objects of some transitive
verbs in Icelandic and the semantic distinction between them. In Flyktförsök.
Kalasbok till Christer Platzack på femtiårsdagen 18 november 1993, från doktorander
och dylika, pp. 1–13. Lund University, Lund.
39
Bar􀀀dal, Jóhanna. 2001. Case in Icelandic: A Synchronic, Diachronic, and Comparative
Approach. Ph.D. thesis, Lund University.
Belletti, Adriana. 2004. Aspects of the low IP area. In The Structure of CP and
IP, edited by Luigi Rizzi, pp. 16–51. Oxford University Press, New York.
Bennis, Hans. 2004. Unergative adjectives and psych verbs. In The Unaccusativity
Puzzle, edited by Artemis Alexiadou, Elena Anagnostopoulou, and Martin
Everaert, pp. 84–113. Oxford University Press, Oxford.
Borer, Hagit. 2005a. In Name Only (Structuring Sense, vol. I). Oxford University
Press, Oxford.
Borer, Hagit. 2005b. The Normal Course of Events (Structuring Sense, vol. II).
Oxford University Press, Oxford.
Bresnan, Joan. 1970. On complementizers: Toward a syntactic theory of complement
types. Foundations of Language 6: 297–321.
Brody, Michael. 2000. Mirror theory: Syntactic representation in perfect syntax.
Linguistic Inquiry 31 1: 29–56.
Burzio, Luigi. 1986. Italian Syntax: A Government-Binding Approach. Reidel,
Dordrecht.
Cardinaletti, Anna and Michal Starke. 1999. The typology of structural deficiency:
A case study of three classes of pronouns. In Clitics in the Languages
of Europe, edited by Henk van Riemsdijk, pp. 145–233. Mouton, Berlin.
Chierchia, Gennaro. 2004. A semantics for unaccusatives and its syntactic consequences.
In The Unaccusativity Puzzle: Explorations of the Syntax-Lexicon
Interface, edited by Artemis Alexiadou, Elena Anagnostopoulou, and Martin
Everaert, pp. 22–59. Oxford University Press, Oxford. (first circulated in 1989).
Chomsky, Noam. 1993. A minimalist program for linguistic theory. In The View
from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, edited
by Kenneth Hale and Samuel Jay Keyser, pp. 1–52.MIT Press, Cambridge,Ma.
Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Step by Step:
Minimalist Essays in Honor of Howard Lasnik, edited by Roger Martin, David
Michaels, and Juan Uriagereka, pp. 89–155. MIT Press, Cambridge, Ma.
40
Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A Life in Language,
edited by Michael Kenstowicz, pp. 1–52. MIT Press, Cambridge, Ma.
Chomsky, Noam. 2005. On phases. Ms. MIT.
Cinque, Guglielmo. 1990. Ergative adjectives and the lexicalist hypothesis. Natural
Language and Linguistic Theory 8 1: 1–39.
Cinque, Guglielmo. 1999. Adverbs and Functional Heads: A Cross-Linguistic
Perspective. Oxford University Press, New York.
Collins, Chris. 2005. A smuggling approach to the passive in English. Syntax 8
2: 81–120.
Collins, Chris and Höskuldur Thráinsson. 1996. VP-internal structure and object
shift in Icelandic. Linguistic Inquiry 27 3: 391–444.
Déchaine, Rose-Marie. 1992. Predication and Category. Ph.D. thesis, University
of Massachusetts, Amherst.
Doron, Edit. 2003. Agency and voice: The semantics of the Semitic templates.
Natural Language Semantics 11: 1–67.
Dubinsky, Stan and Sylvester Ron Simango. 1996. Passive and stative in
Chichewa: Evidence for modular distinctions. Language 72 4: 749–781.
Embick, David. 2003. Locality, listedness, and morphological identity. Studia
Linguistica 57 3: 143–169.
Embick, David. 2004. On the structure of resultative participles in English. Linguistic
Inquiry 35 3: 355–392.
Eythórsson, Thórhallur. 2000. Dative vs. Nominative: Changes in quirky subjects
in Icelandic. Leeds Working Papers in Linguistics 8: 1–28.
Frampton, John and Sam Gutmann. 2000. Agreement is feature sharing.
Ms. Northeastern University.
Frampton, John and Sam Gutmann. 2002. Crash-proof syntax. In Derivation and
Explanation in the Minimalist Program, edited by Samuel David Epstein and
Daniel T. Seely, pp. 90–105. Blackwell, Oxford.
41
Grimshaw, Jane. 1990. Argument Structure. MIT Press, Cambridge.
Haider, Hubert. 2000. The license to license: Licensing of structural case plus
economy yields Burzio’s Generalization. In Arguments and Case, edited by
Eric Reuland, Linguistik Aktuell, pp. 31–55. John Benjamins, Amsterdam.
Harley, Heidi. 1995. Subjects, Events, and Licensing. Ph.D. thesis, MIT, Cambridge,
Ma.
Holmberg, Anders. 2002. Expletives and agreement in Scandinavian passives.
Journal of Comparative Germanic Linguistics 4: 85–128.
Holmberg, Anders and Thorbjörg Hróarsdóttir. 2003. Agreement and movement
in Icelandic raising constructions. Lingua 113: 997–1019.
Hrafnbjargarson, Gunnar Hrafn. 2005. The voice of -st. Ms. University of Tromsø.
Iatridou, Sabine, Elena Anagnostopoulou, and Roumyana Izvorski. 2001. Observations
about the form and meaning of the perfect. In Ken Hale: A Life
in Language, edited by Michael Kenstowicz, pp. 189–238. MIT Press, MIT,
Cambridge, Ma.
Jónsson, Jóhannes Gísli. 1997-1998. Sagnir me

aukafallsfrumlagi. Íslenskt mál
og almenn málfræ

i 19-20: 11–43.
Jónsson, Jóhannes Gísli. 2003. Not so quirky: On subject case in Icelandic. In
New Perspectives on Case Theory, edited by Ellen Brandner and Heike Zinsmeister,
no. 156 in CSLI Lecture Notes, pp. 127–163. CSLI, Stanford, Ca.
Kayne, Richard S. 1993. Toward a modular theory of auxiliary selection. Studia
Linguistica 47 1: 3–31.
Koopman, Hilda. 2005. Agreement configurations: In defense of “spec head”.
Ms. UCLA.
Kratzer, Angelika. 2000. Building statives. Paper presented at the 26th Berkeley
Llinguistic Society.
Lekakou, Marika. 2005. In the Middle, Somewhat Elevated: The Semantics of
Middles and its Crosslinguistic Realization. Ph.D. thesis, University College
London. Available at ling.auf.net/lingBuzz/000122.
42
Maling, Joan. 2001. Dative: The heterogeneity of mapping among morphological
case, grammatical functions, and thematic roles. Lingua 111: 419–464.
Maling, Joan and Höskuldur Thráinsson. 1995. Modern Icelandic verbs governing
dative objects. Ms. Brandeis and University of Iceland.
Marantz, Alec. 1991. Case and licensing. In Proceedings of the Eastern States
Conference on Linguistics, vol. 8, pp. 234–253. MIT.
Marantz, Alec. 2001. Words. Ms. MIT.
Ottósson, Kjartan G. 1992. The Icelandic Middle Voice: The Morphological and
Phonological Development. Ph.D. thesis, Lund University.
Pesetsky, David. 1995. Zero Syntax: Experiencers and Cascades. MIT Press,
Cambridge, Ma.
Pylkkänen, Liina. 2002. Introducing Arguments. Ph.D. thesis, MIT.
Ramchand, Gillian. 2004. Time and the event: The semantics of Russian prefixes.
In Nordlyd, TromsøWorking Papers on Language and Linguistics 32.2: Special
issue on Slavic prefixes, edited by Peter Svenonius, pp. 323–361. University of
Tromsø, Tromsø. Available at www.ub.uit.no/munin/nordlyd/.
Ramchand, Gillian. 2006. Verb meaning and the lexicon. Ms., University of
Tromsø; available at http://ling.auf.net/.
Ramchand, Gillian and Peter Svenonius. 2002. The lexical syntax and lexical
semantics of the verb-particle construction. In Proceedings of WCCFL 21,
edited by LineMikkelsen and Christopher Potts, pp. 387–400. Cascadilla Press,
Somerville, Ma.
Ramchand, Gillian and Peter Svenonius. 2004. Prepositions and external argument
demotion. In Pre-proceedings of Demoting the Agent: Passive and other
Voice-related Phenomena, edited by Torgrim Solstad, Benjamin Lyngfelt, and
Maria Filiouchkina Krave, pp. 93–99. University of Oslo, Oslo.
Ramchand, Gillian and Peter Svenonius. to appear. Mapping a parochial lexicon
onto a universal semantics. In The Limits of Syntactic Variation, edited by
Theresa Biberauer. John Benjamins, Amsterdam.
43
Sigur􀀀sson, Halldór Ármann. 1989. Verbal Syntax and Case in Icelandic: A Comparative
GB Approach. Ph.D. thesis, Lund University.
Sigur􀀀sson, Halldór Ármann. 2003. Case: Abstract vs. morphological. In New
Perspectives on Case Theory, edited by Ellen Brandner and Heike Zinsmeister,
pp. 223–267. CSLI, Stanford, Ca.
Sigur􀀀sson, Halldór Ármann. 2004. Icelandic non-nominative subjects: Facts and
implications. In Non-nominative Subjects, edited by Peri Bhaskararao and K.V.
Subbarao, vol. II, pp. 137–159. John Benjamins, Amsterdam.
Sigur􀀀sson, Halldór Ármann. 2006. The nominative puzzle and the low nominative
hypothesis. Linguistic Inquiry 37 2: 289–308.
Sigur􀀀sson, Halldór Ármann. to appear. Accusative and the Nom/Acc alternation
in Germanic. In Issues in Comparative Germanic Syntax. John Benjamins,
Amsterdam.
Stowell, Tim. 1981. Origins of Phrase Structure. Ph.D. thesis, MIT, Cambridge,
Ma.
Stroik, Thomas. 1992. Middles and movement. Linguistic Inquiry 23: 127–137.
Svenonius, Peter. 2001. Impersonal passives and the EPP: A phase-based analysis.
In Proceedings of the 18th Scandinavian Conference of Linguistics, edited by
Arthur Holmer, Jan-Olof Svantesson, and Åke Viberg, vol. 2, pp. 109–125.
Travaux de l’Institute de Linguistique de Lund, Lund.
Svenonius, Peter. 2002. Icelandic case and the structure of events. Journal of
Comparative Germanic Linguistics 5: 197–225.
Svenonius, Peter. 2004. On the edge. In Peripheries: Syntactic Edges and their
Effects, edited by David Adger, Cécile de Cat, and George Tsoulas, pp. 261–
287. Kluwer, Dordrecht.
Svenonius, Peter. 2005. Extending the extension condition to discontinuous idioms.
Linguistic Variation Yearbook 5: 227–263.
Thráinsson, Höskuldur. 1999. Íslensk Setningafræ

i. Málvísindastofnun Háskóla
íslands, Reykjavík.
44
Valfells, Sigrí􀀀ur. 1970. Middle voice in Icelandic. In The Nordic Languages
and Modern Linguistics, edited by H. Benediktsson, pp. 551–571. Reykjavík,
Reykjavík.
Vogel, Ralf and Markus Steinbach. 1998. The dative - an oblique case. Linguistische
Berichte 173: 65–90.
Wasow, Thomas. 1977. Transformations and the lexicon. In Formal Syntax, edited
by Peter Culicover, ThomasWasow, and Joan Bresnan, pp. 327–360. Academic
Press, New York.
Watanabe, Akira. 1993. Agr-based Case Theory and its Interaction with the A-bar
System. Ph.D. thesis, MIT, Cambridge, Ma.
Woolford, Ellen. 2003. Burzio’s Generalization, markedness, and locality constraints
on nominative objects. In New Perspectives on Case Theory, edited
by Ellen Brandner and Heike Zinsmeister, no. 156 in CSLI Lecture Notes, pp.
301–329. CSLI, Stanford, Ca.
Wurmbrand, Susanne. 2001. Infinitives: Restructuring and Clause Structure.
Mouton de Gruyter, Berlin.
Yip, Moira, Joan Maling, and Ray Jackendoff. 1987. Case in tiers. Language 63
2: 217–250.
Zaenen, Annie and Joan Maling. 1984. Unaccusative, passive, and quirky case. In
Proceedings of the Third West Coast Conference on Formal Linguistics, edited
by Mark Cobler, Susannah MacKaye, and Michael T. Wescoat, pp. 317–329.
Stanford Linguistics Association, Stanford, Ca.
Zaenen, Annie, Joan Maling, and Höskuldur Thráinsson. 1985. Case and grammatical
functions: The Icelandic passive. Natural Language and Linguistic
Theory 3: 441–483.
45

Good folks who follow this blog