Unaccusatives as lexical argument reduction: evidence from aphasia

1
Unaccusatives as lexical argument reduction
evidence from aphasia

• Karen Froud
• Teachers College, Columbia University
• BBSQ6111 - Current issues in speech-language
  pathology Linguistic theory, language
  development and language breakdown
• November 21st, 2005

2
Split intransitivity

• The Unaccusativity Hypothesis (Perlmutter, 1976)
  there are (at least) two types of intransitive
  verbs
• Agentive (unergative) The cat jumped
• The girl danced
• Non-agentive (unaccusative) The flowers wilted
• The sun rose
• Baker 1988 - Uniform Theta Assignment Hypothesis
  (UTAH) / Williams 1994 Theta Related Argument
  Configuration (TRAC)
• Theta roles are uniformly assigned to particular
  structural positions theta structure represents
  pre-movement structure
• UTAH / TRAC Principles of UG
• By UTAH / TRAC, a subject which has thematic
  properties of an object (i.e. is non-agentive)
  must underlyingly be an object

3
Split intransitivity English evidence

• Expletive subject constructions There arrived
  three wise men from the East
• There telephoned three students from Buffalo
• Agreement in there-expletive constructions There
  have arisen several problems
• There has arisen several problems
• Cannot passivize an unaccusative predicate
• The package was accumulated on by dust.
• The bridge was existed under by trolls.
• The oven was melted in by the ice cube.
• The woods were vanished in by Little Red
  Riding Hood.
• Early Modern English unaccusatives selected be
  as perfective aux He is fallen into a vat of
  wine
• He is jumped into a vat of wine

4
Split intransitivity cross-linguistic evidence

• In languages that use both have and be as
  auxiliaries (e.g. for perfect / past tense),
  unaccusatives commonly take be and unergatives
  take have
• Italian
• Giovanni è arrivato. John arrived (be)
  Giovanni ha telefonato John telephoned (have)
• Dutch
• Jan is gevallen. John fell (be)
• Jan heeft gelachen. John laughed (have)
• Some languages have different morphology for
  unaccusative vs unergative versions of the same
  verb e.g. Korean
• Nwun-i nok-ass-ta The snow was caused to melt
• Nwun-i noka-ci-ess-ta The snow became melted
  (i.e. melted by itself)

5
Split intransitivity cross-linguistic evidence

• Ne-cliticization a classic argument from Italian
• Direct-object bare quantifiers require ne
• Gianni inviterà molti studenti.
• Gianni inviterà molti / Gianni ne inviterà
  molti.
• Preverbal subject bare quantifiers do not allow
  ne
• Molti inviteranno Gianni / Molti ne
  inviteranno Gianni.
• Molti affondarono / Molti ne affondarono.
• Post-verbal nominative bare quantifiers with
  passive and unaccusative verbs REQUIRE ne
• Ne saranno invitati molti (passive) / Ne
  affondarono molti (unaccusative)
• Suggests that these pre-verbal subjects originate
  in direct object position
• Burzios Generalization (1986) If a verb
  licenses accusative case, it has an external
  argument. (Unaccusative verbs do not license
  accusative case, therefore they do not have an
  external argument.)

6
Split intransitivity evidence from first
language acquisition

• In a study of Italian children, more overt
  subjects were produced to unaccusative than
  unergative predicates and often unaccusative
  subjects were produced post-verbally (Lorusso et
  al 2004)
• Even in English, children will sometimes produce
  post-verbal subjects to unaccusative predicates,
  but the same construction with unergatives is not
  attested
• Go truck, Come Mommy, Fall the cradle

7
Split intransitivity

• All this suggests that something odd is going on
  with the syntax of unaccusative verbs
• Things we can all agree on
• The subject starts off in object position (where
  it gets its theta role)
• Then it Moves to subject position

John
8
Split intransitivity evidence from second
language acquisition

• There may be a semantic gradient of
  unaccusatives, with some verbs being unaccusative
  in many languages (arrive) and some displaying
  variable cross-linguistic behavior (run, decay)
  (Sorace 2000)
• Interference errors are common in L2 acquisition
• Intermediate and quite advanced L2 learners have
  persistent problems with unaccusative verbs in
  English, Japanese, Italian, French and Chinese
  (e.g. Montrul, 2005)
• Some errors attested in English
• passive unaccusatives
• causativized (transitive) unaccusatives
• avoidance of S-V order with unaccusatives

9
Semantics of unaccusatives

• Things we do not agree on
• Unaccusatives can be viewed as being derived from
  transitive counterparts
• Such a derivation would involve suppression or
  elimination of one of the theta roles assigned by
  the verb
• Reinhart (1997, 2000) - Argument reduction both
  theta roles refer to the same thing (also called
  lexical binding)
• For X verbs Y
• If ?1 ?2, then X verbs X REFLEXIVE
• If ?2 ?1, then Y verbs Y UNACCUSATIVE

10
Argument reduction

• REFLEXIVES internal argument reduction (the
  internal argument is the one thats lost)
• Zach washed Zach Zach washed himself Zach
  washed
• UNACCUSATIVES external argument reduction (the
  external argument is the one thats lost)
• The stone rolled the stone some property of the
  stone itself caused it to roll, with no external
  aid The stone rolled
• Problems this account doesnt capture the
  differences between reflexives and unaccusatives
  e.g. why is the reduced argument retrievable
  after internal argument reduction (as an
  anaphor), but not after external argument
  reduction? Seems like a qualitatively different
  process
• Reinhart Siloni 2004 maybe theres a
  difference between argument reduction which
  applies in the lexicon (? reduction in the number
  of theta roles the V must assign, making
  unergatives from transitives) and argument
  reduction which applies in the syntax (?
  reduction in the number of case features
  available, making reflexives from transitives)
• Unaccusativity is theta role elimination, so on
  this view it must be a lexical operation

11
Zero morphemes

• More things we dont agree on
• Pesetsky 1995 there are a number of morphemes
  with null phonological realizations, which can
  have various effects on the argument structure of
  verbs
• E.g. CAUS a causativization morpheme
• For Pestesky, unaccusatives and reflexives are
  the underlying forms affixes like CAUS add theta
  roles to these representations
• So The vase broke (unacc) CAUS ?
• John broke the vase (i.e. John caused the vase
  to break)

12
Some other properties of unaccusatives

• More observations than arguments
• Unaccusatives can be associated with overt or
  covert morphology
• Semantically, unaccusatives are the loci of
  parametric variation ? unaccusative mismatches
  between languages
• E.g. unaccusatives are stative in Italian, telic
  in Dutch (Sorace 1995)
• Or, unaccusatives result from projection of
  Aspectual heads in the syntax, which affect
  interpretive properties and licensing constraints
  (Borer 2005)
• All of this combines to sound very much like
  properties of functional heads

13
MC

• 74 years old at start of testing, 10 years post
  CVA (LHS embolic stroke secondary to MCI),
  previously right-handed
• Multiple aphasic deficits somewhat similar to a
  Boston nonfluent-type profile, but marked
  differences (greater articulatory agility,
  greater range of grammatical structures in
  spontaneous speech
• More impaired at naming actions than objects
• Non-significant effects of imageability and
  frequency on his word recognition
• A highly skilled communicator ? need to stress
  his linguistic skills to get at the real nature
  of his deficit
• Single word reading tasks

14
MCs language production
Narrative production
Reading connected text

• Once upon a time..There was a prince and
  princess. And they are going a baby. Erm... the..
  baby is arrived and it is the christening to...
  erm...erm...the all the gentry were present...5
  sec and in the gallery was lots of fairies
  which, which dancing, and erm each one took to
  point to the baby and erm... and... like... like
  to give a present. Erm, but I cant do the fairy
  individually. Okay, erm... It is sixteen. What
  happened with the fairies? Oh no, the other
  thing is the wicked fairy, the curse, which is a
  pin, or a - whats the thing called...a bodkin?
  Erm.. which will die.

Once upon a time there was a prince and princess
and they are going to a baby. The baby duly
arrives. The gentry are present. In the gallery
there are lots of fairies and... each one of them
to baby and to a present. The wicked fairy ...
erm... oh, I cant think of that word... cursed
the spindle and said that would be baby should be
touched to die. Well, sort of...oh dear.
Written text

• Once upon a time there was a prince and princess,
  and they were going to have a baby. When the baby
  arrived there was a christening at which all the
  gentry were present. In the gallery were lots of
  fairies dancing, and each one took a turn to
  point to the baby and give a present. But the
  wicked fairy cursed a spindle and said that the
  baby would die when she touched it.

15
(No Transcript)
16
Function word recognition reading

• Written function word comprehension was good, BUT
• Reading written function words in isolation was
  very poor
• Suggests that he can access information about the
  semantics of FCs, but not their phonological
  forms
• Characteristic reading errors limited to the
  supercategory of FCs

17

• Out of 684 function words
• 83 correct (12.13)
• 546 function word substitution errors (79.83)
• 55 other errors / failures to respond
• Out of 784 substantives
• 661 correct (84.31)
• 0 function word substitution errors
• mainly derivational / inflectional errors (stems
  were correct)

18
Experiment reading unaccusatives

• If unaccusatives are associated with a functional
  head, MC is expected to make function word
  substitution errors when reading them
• Unlike other verbs associated with FCs,
  unaccusatives are not reliant on a sentential
  context for their unique argument-structure
  realization they are lexically frozen
• Reading list
• 25 non-alternating unaccusatives
• 50 alternating unaccusatives
• 25 unergatives
• 25 low frequency transitives
• Rated and controlled for imageability and
  frequency
• Presented on five separate occasions, three
  months apart

19
Results

• Things to notice
• Mostly correct most errors in all classes were
  substitutions and morphological errors
• FC substitution errors to unaccusative verbs
  only and not all the time

20
Discussion

• At some level MC could identify and respond to
  unaccusativity
• Generating the structure associated with the
  syntactic reflex of unaccusativity is problematic
• This can be captured in the same terms as the
  rest of MCs deficit FCs can be identified in a
  superordinate kind of way, but not in a
  specific way
• This results in function word substitution errors
• How was MC able to read any of the unaccusatives
  correctly? morphological affixation errors
• In the same way as he sometimes reads prove as
  proven, or earn as earnings,
• He can also misidentify VUNACC as a base form
  (transitive)
• Support for this view sometimes MC does appear
  to substitute another argument-structure-changing
  zero morpheme for the UNACC morpheme
• Examination of MCs spontaeous speech corpus
  revealed the following errors
• Also appeared the father (unaccusative subject
  remaining in base position)
• They appeared the witches (transitive or
  reflexive realization of unaccusative)
• A little shoe was appeared (illegal
  passivization of an unaccusative)

21
Converging evidence

• Kegl 1995 theoretical perspective proposing that
  unaccusatives should be as difficult as passives
  for agrammatic aphasics to interpret, because
  they involve movement of an argument and changes
  to theta properties of the predicate
• Thompson 2003 group study reports more problems
  with unaccusatives than unergatives - because
  unaccusatives are more complex (the Argument
  Structure Complexity Hypothesis)
• Finnochiaro 2003 single case study reporting
  better performance on agentive than
• -agentive verbs - because the patient had a
  semantic deficit restricted to -agentive verbs

22
Conclusions

• The true nature of MCs deficits were manifest at
  the single word level, due to his communicative
  competence masking his difficulties at higher
  levels
• MCs deficits were mainly morphological in nature
• His affixation errors on substantives appeared to
  be very different to his function word
  substitution errors on reading functional
  categories
• But a close examination within a theoretical
  framework reveals that the deficits are the same
  morphology is the locus of MCs problem in
  linguistic representation
• Clinical data obtained from work with MC provide
  support for a particular theoretical formulation
  of unaccusativity as a morphological operation
  affecting argument structure
• So the relationship between linguistic theory and
  clinical investigation is bi-directional

23
Acknowledgements

• MC (in memoriam)
• Neil Smith
• Judit Druks
• Diane Blakemore