You are on page 1of 28

Recoverability of deletion in agrammatic production:

the omission of prepositions

Lidia Lonzi, Claudio Luzzatti, Francesca Vitolo

It is widely accepted that omission/substitution of function words is


direct evidence of an agrammatic brain damage. In this study a sequel
to Lonzi & Luzzatti (1995) we follow the view that the production of
agrammatic patients is subject to a constraint named Telegraph by Pesetsky
(1998) that is present in the pronunciation system of normal grammars:
Do not pronounce function words. We propose that this constraint, though
re-ranked higher in agrammatic production with respect to its position in
normal grammars, is still regularly subject to the condition of Recoverability
of deletion.
The results of two different tasks concerning prepositions a completion
and a grammaticality judgment task given to 4 stabilized agrammatic
patients comply with this view.*

1. Introduction

1.1. The Telegraph Hypothesis

Lonzi & Luzzatti (1995), henceforth L&L, have suggested that


the omission of function words observed in agrammatic production
obeys the Recoverability condition (see (1) below). They found that the
prepositions (Ps) omitted are those that introduce an argument that
is strictly subcategorized ([+ss]) by the verb, and argued that these Ps
can be considered recoverable inasmuch as they are obligatory.
The Recoverability condition was introduced in transformational
grammar by Chomsky (1964, 1965), and has remained a constant
reference in the subsequent development of the theory in connection
with deletion (see Chomsky 1981).1 The omission of function words,
on the other hand, was viewed as a grammatical constraint named
Telegraph (Tel) (Do not pronounce function words) by Pesetsky
(1998), in an Optimality-theoretic framework that allows us to
explain grammatical phenomena of non-pronunciation of function
words like, for instance, complementizers in different languages.
If grammatical processes may arise from tensions between
speaker-oriented desiderata and hearer oriented desiderata, and
Tel can satisfy the requirement on the part of the speaker (that is:
minimize effort vs. maximize parsability and easy interpretability;

Rivista di Linguistica, 18.2 (2006), p. 267-294 (ricevuto nellottobre 2005)


Lidia Lonzi et al.

Pesetsky 1998), it is reasonable to assume that the expressive


problems affecting agrammatic patients originate this reranking in
their production.
Of course, in Pesetskys proposal, Recoverability is among the
constraints that play a role in evaluating the best candidate solution:

(1) Recoverability (Pesetsky 1998:342):


A syntactic unit with semantic content must be pronounced unless it
has a sufficiently local antecedent.

Here, we take Recoverability as a general and intuitive


constraint that selectively permits zero forms in place of function
words here Ps in contexts where singular grammars require the
systematic implementation of their phonological realization. We think
that we can adopt Pesetskys proposal without appealing to specific
Optimality procedures, as far as the hierarchy interaction between Tel
and Recoverability appears to be respected in agrammatic production.
Our proposal, to be specified further, is given in (1):

(1) Recoverability of Ps:


In a telegraphic grammar, the lexical matrix of a verb provides the
required local antecedent for an unpronounced P, if P is selected by
the verb.

According to (1), Recoverability must primarily concern selected


Ps, that is Ps that, by definition, head [+ss] prepositional phrases
(PPs). These Ps must be distinguished from freely inserted lexical
Ps that introduce non-strictly subcategorized [-ss] PPs, although
we must take into account that also [+ss] PPs can be introduced by
freely inserted lexical Ps (particularly locative Ps, see (4) below and
relevant discussion). Our expectation is that Ps, whether selected or
not, heading [+ss] PPs, can be affected in agrammatic behavior.
In our view, the pathological functioning of the relevant process
in agrammatism could be conceived as follows:
the reranking of Tel in the hierarchy of conditions on function
words pronunciation, together with the resulting habit of non
pronouncing Ps that are recoverable, could bring the patient to
a generalized behavior of inattention and/or actual impairment
affecting both the semantically empty Ps and even though less
regularly the lexical, full ones, ultimately locative, in [+ss]
complements.
We must obviously emphasize that, although a constraint that
favors omission, Recoverability cannot have any influence by itself

268
Recoverability of deletion in agrammatic production

upon any substitution. Nonetheless, the results of L&L suggest


a correlation between omission and substitution in agrammatic
speech.2 It can be argued that substitution affects types of Ps that are
subject to omission at least in spontaneous speech. Accordingly, our
predictions are that [+ss] PPs can be affected, with omissions and/or
substitutions, and [-ss] PPs must be relatively unaffected.

1.2. Points of clarification: lexical Ps in [+ss] PPs

Verbs assign n -roles to their n arguments, either structurally


to their internal (object) and external (subject) arguments, or through
Ps to their indirect arguments and to the Location, Manner (very
few verbs of behavior), Cause, or other [+ss] complements.
We stipulate that indirect arguments receive their -roles
through Ps without a specific semantic content, like di of or a to,
selected by the verb, as in (2):

(2) Dubito della tua sincerit


(I) doubt of-the your sincerity

Other arguments, like the [+ss] complements mentioned above,


owe their -roles to Ps with a specific semantic content, as in (3):

(3) Le trote vivono nei ruscelli


trouts live in-the brooks

Encyclopaedic knowledge conceals the fact that in a locative


[+ss] PP the prepositional head is not univocally determined. The
semantic relevance of a P is associated with lexical choice, therefore,
in principle, with non-Recoverability. See (4):

(4) a - Ho messo una busta sul libro


(I) have put an envelope on-the book
b - Ho messo una busta nel libro
(I) have put an envelope inside-the book

From the -theoretic point of view (Chomsky 1981), locative [+ss]


Ps transmit the verb -role and, notably, contribute their specific
meaning: they are never empty. The relevant completion task, then,
requires the full interpretation of the sentence; more precisely the
lexical mastery of the verb and of both the direct object (if present)
and other complement nouns, whose relevant lexico-semantic features
play a role in the choice of the P. See (5) below, where (5b) and (5c)

269
Lidia Lonzi et al.

show that the alternative attested in (5a), as in (4) above, is not


always available:

(5) a - mettere le mani sulla/nella sabbia


put Det hands on / in the sand
b - mettere le mani sul/*nel pavimento
on / in the floor
c - mettere le mani *sulla/*nella/sotto la pioggia.
on / in/under the rain

In fact, locative Ps are underdetermined by V, hence not


recoverable (see (1)), but in the context of a [+ss] complement these
Ps become recoverable, mainly for extra-linguistic reasons. Indeed,
the contexts where the Tel hypothesis makes conflicting predictions
are [+ss] locative complements (see Table 1).

1.3. Previous Hypotheses

Since the work of Friederici (1982) we know that Ps are not


uniformly affected in agrammatism, therefore any explanation of
agrammatic behavior based on the fact that Ps are a homogeneous
linguistic class is not adequate. For this reason, neither is adequate
the phonologically based explanation (Kean 1980), nor the explanation
based on the datum that the privileged access to the closed class
items appears to be lost in Brocas patients (Bradley et al. 1980).3 Our
explanation must confront the explanations that admit the possibility
of a dissociated behavior according to different types of Ps.
Now, if we take into consideration the relevant aphasiological
literature (Friederici 1982; Bennis et al. 1983; Rizzi 1985; Grodzinsky
1989), we see that it is feature A and/or B that is deemed favorable to
the preservation of Ps in agrammatism:
A. presence vs. absence of semantic content of P (Friederici 1982,
1985; Bennis et al. 1983; Rizzi 1985). In particular, in Bennis et al.,
semantic content is associated with free lexical insertion, 4 in Rizzi,
with relevance for -theory;
B. presence vs. absence of syntactic independence (Grodzinsky
1989; Bennis et al. 1983). According to Grodzinsky, P is preserved if it
is not governed.
Bennis et al. (1983) show that A and B are two sides of the
same coin. The relevant hypothesis is that P is preserved when it is
a meaningful, autonomous item, untouched by an impairment that
is syntactic in nature, whether in a general (Bennis et al. 1983) or

270
Recoverability of deletion in agrammatic production

a specific, partial way (Grodzinsky 1989). For this distinction, see


Luzzatti et al. (2001).
Bennis et al. have drawn a theory-based distinction of Ps
(the relevant theory being Chomskys 1981) in three subclasses:
syntactic, subcategorized and lexical. The first subclass contains
pure case-assigning Ps, such as of between two nouns, and to in
English ditransitive verbs (a subclass of obligatory Ps that could
reasonably be grouped with our selected Ps, even if the Case Filter is
outside the pronunciation system; see Pesetsky 1998); the second one
contains Ps that in a vague intuitive sense form a lexical unit with
the verb, as in John was cared for (also these could reasonably be
grouped with our selected Ps); and the third one contains Ps that have
a lexical content of their own, instantiated by locative Ps (a subclass
that we propose to distinguish according to whether the relevant PPs
are [-ss] or [+ss]).
According to Bennis et al., the most impaired subclass in
agrammatic patients was predicted to be the syntactic one, while
subcategorized Ps were expected to behave like lexical Ps. Their
results failed under this respect (though reaching significance in
the differential error pattern between fluent vs. nonfluent patients):
no evidence was found for a differential loss of syntactic vs.
subcategorized Ps.
In our proposal, P is relatively spared when it is the head of an
optional PP, hence not recoverable. This point is very near to the
one made by Grodzinsky (1989), with a slight difference: we refer to
subcategorization by the verb, while Grodzinsky refers to syntactic
configuration. This slight difference yields a divergent prediction
in the case of locative Ps. While, under Grodzinskys analysis, Ps
heading [+ss] PPs are straightforwardly affected because governed,
independently from their lexical-semantic content, for our analysis
lexical-semantic content is crucial. The strict subcategorization of a PP
lends predictable results for Recoverability when it is matched with
the obligatoriness of the relevant P, but this is not the case for locative
Ps. The agrammatic answer can be determined by two alternative
criteria, and this is the reason for the (+/-) prediction in this case:
indeed, not an unfalsifiable prediction, for it is coupled with a (-) score
expectation for nonlocative Ps heading [+ss] PPs, on the one side, and
a (+) score expectation for locative Ps heading [-ss] PPs, on the other.
One proviso is, however, in order. The same conflict between
prediction of pronunciation vs. non-pronunciation of P (or preservation
vs. impairment) could extend to [-ss] locative PPs, if locative PPs
have indeed a special status in the linear order of adverbial PPs as

271
Lidia Lonzi et al.

suggested in Cinque (2002). Cinque suggests that there is a fixed


order of complement and adverbial PPs: Time, Location, PDat. For
reasons that are not relevant here, both in English and Italian this
order surfaces as PDat (that is, any indirect, or obligatory, argument),
Location, Time (not testable with our present methods based on
picturable events). If this were the case, both PDat and P in a
locative nonobligatory argument would be recoverable. The former by
hypothesis, the latter due to the primary position of Location in the
complement or adverbial serial order. For all other [-ss] PPs following
Location like Manner, Agent, Instrument, Purpose, etc. nothing
in the string could reveal the empty slot involved except the P itself,
which, therefore, cannot be omitted. We will revert to this alternative
hypothesis in the discussion.
In order to obtain comparable results and to evaluate the
different hypotheses, L&L established four types of Ps given in (6)
below, for which the expected pattern of performance according to all
the mentioned hypotheses is summarized in Table 1:

(6) A. locative Ps in [+ss] complements


(rich semantic content, absence of syntactic independence)
es. Il bambino sale su un tavolo the boy is climbing on a table
B. nonlocative Ps in [+ss] complements (no semantic content,
absence of syntactic independence)
es. Il bambino risponde alla maestra the boy is replying to-the
teacher
C. locative Ps in [-ss] complements
(rich semantic content, presence of syntactic independence)
es. La bambina mangia la torta in cucina the girl is eating the cake
in (the) kitchen
D. nonlocative Ps in [-ss] complements (poor semantic content,
presence of syntactic independence)
es. La ragazza legge la storia con interesse the girl is reading the
story with passion

In L&L, the relevant predictions were fulfilled: nonlocative


Ps in [+ss] complements were affected, with both omissions and
substitutions, nonlocative Ps in [-ss] complements5 were relatively
unaffected. As for locatives, [+ss] Ps were severely affected in
one patient (RO) and preserved in another one (AD), although,
notably, under a different experimental modality (see footnote 2).
This complies with the conflicting prediction of the Tel hypothesis
concerning Ps in locative complements.

272
Recoverability of deletion in agrammatic production

Table 1. Expected pattern of performance in the use of Prepositions according to


the different theoretical accounts.

Friederici Bennis et Grodzinsky Telegraph


1982, 1985; al. 1989 (Pesetsky 1998)
Rizzi 1985 1983 tested by L&L
(*) 1995
(a) loc [+ss] + + - +/-
(b) nonloc [+ss] - + - -
(c) loc [-ss] + + + +
(d) nonloc [-ss] +* +* + +
+ = relatively preserved
- = impaired (omitted, substituted, misjudged)
+/- = either preserved or impaired
* = not tested, only predictions

We must recall that, in L&Ls hypothesis, Recoverability is a


criterion for the omission of Ps. In (nonlocative) [+ss] complements,
as in example (2) above, P is univocally required hence recoverable,
whereas the contrary is true for Ps in [-ss] complements that are not
obligatory by definition, like Agent or Manner phrases (even if the
relevant P is not subject to lexical choice). Therefore, the relevant
opposition is between subclass (B), omitted, vs. (C)-(D), pronounced.
It is reasonable to assume that in addition there is a more vague
region occupied by (A), that is locative Ps in [+ss] complements, which
can be either preserved or not, depending on which one of their two
conflicting properties with respect to Recoverability does prevail:

1. their paradigmatic character: they are subject to lexical choice;


2. their relative obligatoriness: they introduce a [+ss] complement
and are pragmatically determined by encyclopaedic knowledge (if,
mutatis mutandis, this characterization should apply also to [-ss]
locative complements, no opposition between the two locative types
(A) vs. (C) could be predicted).

2. The study

In L&L, subjects were requested to judge whether anything


was missing in the sentences presented, and, if so, to provide the
appropriate word. In order to have a better assessment of the
question whether agrammatic omissions must be interpreted as direct
evidence of a specific impairment at the syntactic level, or, instead,

273
Lidia Lonzi et al.

as a manifestation of the Tel constraint, which is implicit in normal


grammar and re-ranked as a consequence of a different generalized
impairment, we devised a new experiment where the completion task
is kept distinct from the grammaticality judgment about correct vs.
incorrect/ missing Ps.
Based on our previous results, showing that both omission
and substitution can affect Ps introducing [+ss] complements, our
expectation was that an opposition would have emerged between:

[+ss] PPs, affected, with omissions and/or substitutions, vs.


[-ss] PPs, relatively unaffected.

As stated above, this opposition must concern nonlocative Ps


only. As for locative Ps heading [+ss] PPs, our expectation was
divided between impairment and preservation, subject to the specific
organization of the patients telegraphic grammar. That is, both
behaviors were deemed compatible with our hypothesis, contrary
to more common predictions (see Table 1). In contrast, locative Ps
heading [-ss] PPs were expected to be preserved.
The pattern shown by patient RO in L&L, with only 27% correct
completions in [+ss] locative complements (vs. 50% in [-ss]), and the
pattern shown by patient AD, who was not able to pronounce any P
but consistently required the expression of both types of locative Ps
(vs. nonlocative Ps heading [+ss] PPs), encouraged us in the direction
of a split prediction.
Passive da by was another focus of our study for which there
are different predictions in the literature. According to Grodzinskys
trace deletion hypothesis (for a review and an extensive discussion
see Grodzinsky 2000), passive is a benchmark for agrammatic patients,
given that their representation of a passive sentence presents no NP
trace in object position to allow the transmission of the theme -role to
the subject position: the relation between traces and their antecedents
is unavailable to them. As a consequence, both the subject, due to a
default strategy of agent -role assignment to the first constituent of
the clause, and the by-phrase, due to the regular assignment of this
role by the preposition, happen to bear an agent role, and the two
identical roles cause a chance performance. Under this view, while
comprehension performance of adjectival passives, also known as
lexical passives that is passives with no NP movement (ex.: The
woman was uninspired by the man) should score above chance,
comprehension of psychological passives, where the by-phrase bears the
experiencer -role (ex.: The girl was admired by the boy), should score

274
Recoverability of deletion in agrammatic production

even below chance, due to the puzzling competition between agent -


role of the first constituent and experiencer -role of the by-phrase.
On the contrary, according to the Tel hypothesis, we expected a
lower performance for the production of the obligatory, recoverable P
introducing the theme argument of the Psych-verb in the adjectival
passive form (Belletti & Rizzi 1989): here Argument Psych-da; and
a higher score for the production of Ps introducing the adjunct da
by-phrase of actional or even nonactional (psychological) passives:
here Passive da and Adjunct Psych-da, respectively. There is no
contradiction, in fact, between our and Grodzinskys predictions, that
is between good comprehension and impaired retrieval, although we
should expect specific problems with passives (see the extensive study
of Luzzatti et al. 2001).

3. Materials and methods

3.1. Subjects

3.1.1. Agrammatic patients


We tested 4 patients suffering from moderate Brocas aphasia
with agrammatism. Their spontaneous speech was nonfluent,
sentences were short and telegraphic, with omission of function
words, like prepositions, determiners, auxiliaries and pronouns
(in addition to licit null subject pro). Inflected forms were often
substituted, and almost always with less marked ones (e.g. verbs in
nonfinite form, adjectives and passive participles in the masculine
singular form). Two patients (AD and RO) were also involved in
L&Ls study.
AD is a 24-year-old woman who suffered from a brain
hemorrhage due to the rupture of a cerebral aneurism (May 1990).
A first language evaluation given about seven months after disease
onset revealed a severe Brocas aphasia with agrammatism and
very mild apraxia of speech. Sentences were short, with simplified
syntactic structure. Verbs and function words were often omitted.
Confrontation naming was severely impaired, repetition and reading
were only moderately impaired. She was treated for her language
impairment for the following 24 months. At the end of the treatment,
AD still suffered from mild agrammatic speech output and only mild
naming deficit. Repetition and reading aloud were only minimally
impaired. The spontaneous speech of 95 corresponds to the moment
of our study.

275
Lidia Lonzi et al.

In all the excerpts of spontaneous speech, names of persons and


places are fictitious to preserve patients privacy.

AD, March 1991

Ex: Tell me what happened to you

AD: Allora,..eh allora, venerd successo perch


Well,.. uh well,.... Friday happenedPstPrt because

..gui-guidato, guidato, guidato. Eh.la vena


...driven, driven, driven. Uh.the vein

si rotta e la macchina fuori di strada e..


broke Refl+Aux+PstPrt and the car out-of-street and..

eh perch, aspetta,.. in coma, eh.. io, allora,
uh because, wait,.. in coma, uh..I, then,

a Verona, unamica ehm Borgor-Borgoroma..


in Verona, a girlfriend ehm Borgoroma (hospital)..

E,.. son qua, eh.. allora, di qua, allora eh


And,.. am here, uh.. then, from this side, then uh

(a)spetta, a Verona..... eh di qua.... non (..)...recupero no?....
wait, in Verona.....uh from this side.... (there is) no (..)... recovery is
there?

AD, October 1995

Ex: Now describe me a day of yours: what do you do?

AD: Allora, pro- allora: al mattino, vabb,


Well, well: in the morning, OK,

mi alzo.. colazione, vabb.. poi.. eh ma


get up1stPSg.. breakfast, OK.. after.. uh but

a colazione con.. mangio con la radio.


at breakfast with.. eat1stPSg with the radio.

Poi niente, alle otto, otto e qualcosa,


Then nothing, at eight, eight and something,

276
Recoverability of deletion in agrammatic production

dipende, vabb.. mm.. cosa?.. eh..


(it) depends, OK.. mm.. what?.. uh..

incomincio a studiare.. studiare,


start1stPSg at studying.. studying,

(in)somma. eh ho, ci sono a casa


(in) sum. uh have1stPSg, there are at home

dei.. delle (e)ncicopedie..


someMsc..someFem encyclopediesFem..

che sono pi semplici.. quando


which are more simple.. when

eravamo da.. quando ero piccola,


were1stPPl from.. when was1stPSg a child,

RO was a 50-year-old white collar working in the administration


department of a trade company. In November 1987 he suffered from a
left hemisphere cerebral abscess, which determined a severe language
disorder. A first evaluation given about 4 months after disease
onset revealed global aphasia. Spontaneous speech was limited to
a few recurring utterances. The same degree of impairment also
emerged in all repetition, reading and naming tasks. Oral and written
comprehension were also severely impaired. The patient underwent
language rehabilitation, initially for his comprehension disorder and
his speech apraxia, and subsequently for his lexical and morpho-
syntactic deficits. The evolution of his impairment is documented by
several follow up evaluations.
Eight years after disease onset the patient still suffered from
severe nonfluent language deficits. Speech output was severely
agrammatic with two-to-three-word sentences and omission of
function words. In conversation, the patient used very few verbs and
only in nonfinite form (infinitives and past participles). There was a
mild-to-moderate confrontation naming and comprehension disorder,
a moderate repetition deficit and a mild-to-moderate reading deficit.

RO, September, 1991

Ex: Describe me how do you start your morning, after you wake up

RO: Sveglia, cinque, lavare, eh, sciacqua il viso..


Alarm, five, washInf, uh, rinces Det face..

277
Lidia Lonzi et al.

Barba, colonia, preparare caff; miele, crusca,


Beard, cologne, prepareInf coffee; honey, bran...

basta. Nina, v- svegliare, eh, svegliare, eh,


enough. Nina, wakeInf, uh, wake, uh,...

ostia (). Caff!, bacino, caff, tavola,


Interj. Coffee!, kiss, coffee, table,

eh, usti, Pavesini, mangiare, basta. ()


uh, Interj, Pavesini (biscuits), eatInf, enough. (...)

Eh, in (i) denti, lavare i denti.


Uh, Det teeth, washInf Det teeth.

(O)spedale, palole (parole), eh, stundiare (studiare),


Hospital, words, uh, studyInf,

basta. () Viaggio, chiuso. Riccardo.


enough. Transport, stop. Riccardo [logopedist].

Ex: And after you come home?

RO: Eh, bicchiere, limone, scitroppo (sciroppo),


Uh, glass, lemon, syrup,

spolverare larmadio, sala, eh, tante cose;


dustINF the wardrobe, living-room, uh, many things;

min-, preparare minnestra (minestra):


prepareInf soup:

pasta, riso; eh, tante cose, la mela, basta. Pranzo,


pasta, rice; uh, many things, the apple, enough. Lunch,

letto ona (una) ora. Poi, televisione, giornale,


bed one hour. Then, television, newspaper,

leggere, scrivere. Eh, giornale, fogli,


readInf, writeInf. Uh, newspaper, pages,

girare la pagine; eh, giornale.


turnInf theSg...pages; uh, newspaper.

278
Recoverability of deletion in agrammatic production

Caff (), grappa,


Coffee (...), grappa,

la eh, v-osti , limone, ostia,


theSgFem...uh,...Interj..., lemon, SgMsc Interj,

Ex: and then?

RO: Poi minestra: dadi, eh, po (poi)


Then soup: broth cubes, uh, then

la osti, () dadi, polpa, riso; eh,


theSgFem...Interj (...) cubesMsc, tomato pulp, rice; uh,

tante cose, pereparare (preparare) Nina, Serena, io


many things prepareInf Nina, Serena, I

mangiare minnestra (minestra).


eatInf soup.

Basta; televisione, amaro, amaro, cognas (cognac),


Enough; television, bitter, bitter, cognac,

a-maro, bottiglia, caff.. eh, basta.


bitter, bottle, coffee.. uh, enough.

PG is a 39-year-old right-handed man with a leading position


in a financial society. In February 1993 he suffered from a cerebral
vascular attack (thrombosis of the left internal carotid) causing right
hemiplegia and aphasia. A first language examination (May 1993)
revealed severe language impairment. His spontaneous speech was
limited to a few fragmentary clauses. There was only mild apraxia
of speech, but severe anomia. Sentences were very short (one or two
words) with systematic omission of verbs and function words. The
patient was treated for his language impairment for about two years.
The evolution of his disorder is documented by several follow up
examinations.
Two and a half years after disease onset his language is still
severely impaired. Speech output is agrammatic with frequent
omission of function words. Verbs are frequently omitted and
occasionally in nonfinite form. Articulation is only minimally
impaired. Confrontation naming and reading aloud are only mildly
impaired.

279
Lidia Lonzi et al.

PG, October 1995

Ex: Now, try telling me why you are here and what happened to you.

PG: Dunque, tanto tempo sono eh..


Well, much time have1stPSg uh..

sono.. stato a.. a casa con Silvia


have.. been at.. at home with Silvia

ho guardato ecco eh.. ho guardato


have1stPSg watched, right, uh.. have watched

il programma su canale cinque


the program on channel five

e mentre guardo eh sento qualche cosa..


and while watch1stPSg uh feel1stPSg something..

fastidio ecco.. febbre?


trouble, right.. fever?

e poi io dico Silvia sto male aiuta-mi.


and then I say Silvia feel1stPSg bad help me.

E poi.. caduto.. caduto per.. ospedale di Saronno.


And after.. fallen.. fallen but.. hospital of Saronno.

Dunque, TAC eh esami eh raggi tutti insieme, ecco.


Well, CT uh exams uh rays all together, right.

Ero muto e poi, piano piano..


was1stPSg mute and after, very slowly..

ora.. parlo, ora capisco.


now.. speak1stPSg, now understand1stPSg.

LZ was a 35 year-old social worker and university student. Few


days after the birth of her second daughter (August 95) she suffered
from right hemiplegia and aphasia due to a left hemisphere cerebral
hemorrhage. A first language evaluation given about 10 months after
disease onset revealed a Brocas aphasia with agrammatic output.
Spontaneous speech was scarce, made of one- or two-word sentences,
with frequent omission of function words. Verbs were also frequently
omitted or in nonfinite form. Reading and confrontation naming

280
Recoverability of deletion in agrammatic production

were severely damaged, while repetition was only mildly affected.


The patient underwent language rehabilitation for his lexical and
morpho-syntactic deficits. The evolution of her damage is documented
by several language evaluations.
In May 1997, her speech output is still severely agrammatic,
with constant omission of function words. Confrontation naming and
reading aloud had significantly improved.

LZ, November 1996

Ex: Now you should tell me the story of your disease

LZ: eh.. coma.


uh.. coma.

Ex: What happened exactly to you at the beginning, where were you?

LZ: ah. eh E-ster ed io ehm.. eh.. e bam-bi-ni.


ah. Uh Ester and I ehm.. uh....and children.

Ah.. eh madonna ehm Ester e io


Ah.. uh Interj ehm Ester and I

e bambi- eh Marco eh.. e Maria pan-cione.


and uh Marco uh.. and Maria big belly.

Ex: Who is Maria?

LZ: eh.. mi-a figlia!


uh..my daughter!

Ex: Try to explain me better.

LZ: Maria, in-cin-ta.


Maria, pregnant.

Ex: Maria is pregnant?

LZ: s, poi.
yes, after.

Ex: You were pregnant of Maria.

LZ: s s e poi eh ma-le alla te-sta eh ehm....


yes yes and then uh pain to the head uh ehm....

281
Lidia Lonzi et al.

lun- eh.. umpo-lan-za (ambulanza) eh.. ah no!


ambulance uh.. ah no!

cinque giorni Maria..... No?


five days Maria.....[five days after Ms birth] No?

E poi coma. Cinque giorni.


And then coma. Five days.

3.1.2. Control subjects


Nine control subjects of comparable age and education were
given the completion task. A further group of four control subjects
were given the grammaticality judgment task.

3.2. Materials

In the present study, Ps are still divided according to the four


major types given in (6) and summarized in Table 1, but several new
subtypes of (nonlocative) Ps are included in types B and D. Table
2 summarizes the types of Ps and of sentences entering both the
completion and the grammaticality judgment task.
In type B we included da introducing PPs that are subcategorized
by psychological verbs in the Aux (essere be) plus past participle form
(argument Psych- da). According to the classical analysis of Belletti
& Rizzi (1988), this form is an instance of adjectival passivization,
where da by introduces a theme argument rather than an agent
phrase (in the present analysis, an adjunct).5
In type D, besides da by, introducing agent da-phrases in
actional passives (passive da), we included a group of Ps introducing
manner (con), reason and purpose cause (per) [-ss] PPs;6 da by
introducing agent PPs in nonactional psychological passives (Adjunct
Psych-da); and con with, introducing PPs with an instrument -role.
By hypothesis, all Ps in type D assign their specific -role inherently.
This -role bears a univocal though not biunivocal - correspondence
with the relevant P.7
Verbs entering the sentences to be completed with the target
preposition were matched for word frequency (see Table 3).

282
Recoverability of deletion in agrammatic production

Table 2. Types of Prepositions used in the completion and the grammaticality


judgment tasks ([+ss] = strictly subcategorized; [-ss] = nonstrictly subcategorized).

Types of subtypes N examples


Prepositions
A. Locative [+ss] (24) Il ragazzo versa lacqua da un vaso
The boy pours the water from a vase
B. NonLocative Ps selected (23) La tenda appartiene a un bambino
[+ss] by V The tent belongs to a boy
argument (10) I passanti sono stati incuriositi dalla
Psych- da scena
Passers-by have been intrigued by the
scene
C. Locative [-ss] (14) Il bambino legge il libro su una sdraio
The boy reads a book on a deck-chair
D. NonLocative manner, (16) Il bambino legge il libro con attenzione
[-ss] reason, The boy reads the book with attention
and purpose
cause
passive da (13) Il cane addestrato da un bambino
The dog is trained by a boy
instrument con (13) La bambina lega il pacco con lo spago
The girl ties the parcel with a string
adjunct Psych- (10) Il cane amato da un bambino
da The dog is loved by a boy

Table 3. Word frequency of the verbs entering each sentence type.

loc nonloc loc nonloc


Comparison
[+ss] [+ss] [-ss] [-ss]
N= (Anova)
24 33 14 52
Verb frequency F(3,119)= 0.9;
307801 94183 158166 1291547
Mean st.dev p= ns

Oral word frequency values are taken from a corpus of 500.000 words (De Mauro et al. 1993).

283
Lidia Lonzi et al.

3.3. Experimental procedures

(1) Filling up the missing preposition


Each trial consisted of a 12 by 8 cm line drawing, taken or
adapted from various picture vocabulary tests, representing a certain
event (see Figure 1). The use of pictures was intended to favor the
comprehension of the sentences submitted and to avoid any possible
ambiguity concerning the P to be inserted.

Fig. 1. Example of a stimulus from the Completion task.

Il bambino si arrampica... un recinto [su]8


The boy climbs... a fence [on]

The drawing was coupled with a sentence describing the same


event verbally. Each sentence consisted of a NP subject, a finite V/
Aux + Past Participle (a NP object) and a PP in which, however, the
P was omitted. The examiner read aloud each sentence, reinforcing
the gap with a click. Patients and controls were asked to fill the gap
with the appropriate preposition. Latencies up to five seconds and
spontaneous repairs were accepted as correct responses. Longer
latencies were counted as errors.

284
Recoverability of deletion in agrammatic production

(2) Grammaticality judgment


In this condition, the patients were asked to judge the
grammaticality of sentences read aloud by the examiner. The
stimuli were the same sentences used in condition 1. All sentences
were presented twice, once in the correct form, and once in a form
where the relevant P was either omitted or substituted with an
inappropriate one (30 omissions vs. 93 substitutions). Correct and
incorrect sentences were randomized in a unique list for a total of
246 judgments. The grammaticality judgment task was given to the
patients a week after the completion task.

4. Results

Table 4 gives the raw results obtained by the four agrammatic


patients in the completion task.

Table 4. Results of the completion task (number of correct responses and %


score).

nonloc [+ss] non loc [-ss]


loc P M, Adjunct Total
Argument loc [-ss] passive
[+ss] selected R& Instrument Psych- correct
Psych- da da
by V PC da

AD 11 10 6 6 13 10 13 8 77
RO 14 8 9 4 14 10 13 8 80
PG 17 8 5 10 11 11 12 7 81
LZ 7 7 2 6 10 2 12 3 49
N= 24 23 10 14 16 13 13 10 123

%
AD 46 43 60 43 81 77 100 80 63
RO 58 35 90 29 87 77 100 80 65
PG 71 35 50 71 69 85 92 70 66
LZ 29 30 20 43 62 15 92 30 40
Average 51% 36% 55% 46% 75% 63% 96% 65% 58.5

M, R & P C = manner, reason, and purpose cause.

285
Lidia Lonzi et al.

We decided to keep the two Psych-da variables (argument and


adjunct) separate from their respective main categories, in order to
test an identical P with a different syntactic status in two analogous
morphosyntactic contexts (see footnote 5). Therefore, as a first step,
data relevant to all subtypes are presented separately (see Table 4).
In compliance with our predictions, the retrieval of Ps in
(nonlocative) [+ss] complements was significantly more impaired
than in (nonlocative) [-ss] complements (t(3) = 8.31; p<.002). Even
[+ss] argument Psych-da is significantly more impaired than [-ss]
adjunct da (that is da introducing agent da-phrases in both actional
and psychological passive): [paired t (3) = 3.13; p<.03]. A result that
is mainly due to patients AD and PG (RO has no problem with any of
these instances of da, while LZ has an impaired performance with all
of them).

Fig. 2. Results of the completion task, after aggregation in 4 major types of


Prepositions (% of correct responses).

FIGURE 2 shows these results with the scores grouped in the


four major categories. Nonlocative [+ss] PPs, with the inclusion of
argument Psych-da, improve slightly (from 36% to 41%). Nonetheless
the difference with respect to nonlocative [-ss] PPs is still significant
for all patients. (AD: 2(1) = 12.47; p<.001; RO: 2(1) = 12.5; p<.001;
PG: 2(1) = 13.6; p<.001; LZ: 2(1) = 5.02; p<.03).
Nonlocative [-ss] PPs with the inclusion of data concerning
instrument con and the two forms of passive da obtain the same

286
Recoverability of deletion in agrammatic production

mean score (75%). Furthermore, apart from the above-mentioned


behavior of RO and LZ with the three instances of da, for no variable
there is a significant difference between patients, with the exception
of locative Ps, where a difference between patients was predicted only
in our framework, given the split expectation for locative Ps in [+ss]
complements (see Table 1).
As for the grammaticality judgment task, the patients obtained
better scores than in the completion task across all the variables (t(3)
= 3.99; p<.015), with only minimal exceptions: in AD, the production
of Instrument con is better than the judgment on sentences of both
types; in RO, (i) the production of Instrument con is better than the
judgment on the wrong sentence; (ii) the production of argument
Psych-da is better than the judgment on the right sentence. Results
obtained by the four patients and the control sample are given in
FIGURE 3, grouped in the four major categories.

Fig. 3. Results of the grammaticality judgment task (% of correct responses).

If we take apart the wrong sentences from the correct ones,


the judgment on the wrong sentences is more impaired than on the
correct ones (t(8) 1.90; p, one sided <.05).
In fact, the opposition [+ss] vs. [-ss] inside nonlocative Ps is
confirmed though not always significant (AD: 2= 3.1, p 1-sided=.039;
RO: 2= 5.4, p 1-sided=.01; PG: 2= 6.7, p 1-sided <.005; LZ: 2= 0.8,
ns), whereas locative Ps ar more variable: they obtain an identical
mean score (93%) with the correct sentences and a different mean

287
Lidia Lonzi et al.

score (72% [+ss] vs. 84%[-ss], still not significant (Fishers Exact Test:
AD =.5; RO =.17; PG =.33; LZ =.35), with the incorrect ones. Only
the wrong sentences allowed the patients to evaluate omissions of
Ps. However, too few omissions per P type were tested to permit a
meaningful analysis.

In particular:
AD does not accept any omission in nonlocative [+ss]
complements (with only one exception). It is however interesting that
in the completion task she makes some omissions only within this
type of Ps, where 38% of her errors are omissions (5 out of 13, see
Table 3). In L&L, her behavior favoring locative vs. nonlocative Ps
complied with the semantically based hypothesis of Friederici (1982)
and Rizzi (1985); now, her low score with locative Ps requires a new
interpretation.
RO shows a trend (58% in [+ss] vs. 29% in [-ss] locative PPs)
with [+ss] > [-ss] that, however, does not reach the significance level
(2 = 3.14, p= ns; one-tailed p <.05). Notably, in the grammaticality
judgment task, this difference is not confirmed and the relation is
reverted. His difficulties with locative Ps, which he often replaces
with the polyvalent con with as a default P, seem generalized.
LZs completion is impaired in [+ss] PPs, whether nonlocative or
locative. Nonetheless, it is possible that also in this case the behavior
with locative Ps is somehow generalized, since the difference between
[+ss] and [-ss] PPs does not reach significance (Fishers Exact Test:
.88, ns).
PG, finally, meets difficulties only with nonlocative [+ss]
complements. He does not show any difference between the two
locative types (71% in both cases): a predictable result, however, when
[+ss] locative Ps are not affected.
Some examples of errors with the two types of locative Ps across
the two tasks are given below:

Completion task Grammaticality judgment task


(errors are given in italics) (errors are given in italics)
Correct P Wrong P (or )
AD
[+ss] La bambina si seduta ad uno sgabello su on: yes da from: no
the girl sat to a stool

[-ss] Il ragazzo beve una bibita da... la terrazza su on: yes a to: no
the boy is drinking from the terrace

288
Recoverability of deletion in agrammatic production

RO
[+ss] Il bambino si arrampica da un recinto su on: no : no
the boy is climbing from a fence

[-ss] I bambini aspettano la mamma sul giardino in in: no su on: yes


children are waiting their mother on the garden

LZ
[+ss] Il bambino pianta linsalata su lorto in in: yes : yes
the boy is planting the salad on the orchard

[-ss] Il cane mangia la carne con una ciotola in in: yes su on: yes
the dog eats the meat with a bowl

PG
[+ss] Il ragazzo si appoggia da un bastone a to: no in in: no
the boy is leaning from a stick

[-ss] La donna scrive con un blocco su on: yes : no


the woman is writing with a notebook

An unexpected consistent result is the homogeneous treatment of


both [+ss] and [-ss] locative Ps, in all the patients. In two patients (RO
and LZ), as noted above, the difference that can be observed between
the two types of locative Ps is not significant. Even the trend shown by
RO in the direction opposite to expectation does not reach significance.
Note that substitutions of locative Ps are intraspecific: locative Ps are
substituted with Ps still denoting location (with the exception of the
default use of con, observed in all patients)9. This indicates that the
representation of the relevant thematic grid is preserved.

5. Discussion

The aim of our study was to assess whether the error pattern of
agrammatic patients concerning prepositions allows us to maintain
that their behavior with respect to this class of function words is
compatible with the constraint of Recoverability of deletion (Telegraph
hypothesis). The relevant predictions overlap strikingly with other
predictions in the literature, with the exception of those concerning
locative prepositions. In Table 5 the overall predictions of Table 1
and the results of the present study are compared (note that there is
an implication relation between the [+/-] values for the two types of
locative Ps: if locative [+ss] impaired, locative [-ss] impaired etc.).

289
Lidia Lonzi et al.

Table 5. Expected pattern of performance in the use of Ps according to the


different theoretical accounts including the results of our present study.

Friederici Bennis et al. Grodzinsky Telegraph (Pesetsky 1998)
1982, 1985;
Rizzi 1985 (*) 1983 1989 L&L 1995 Present study

(A) loc [+ss] + + - +/- +/-


(B) nonloc [+ss] - + - - -
(C) loc [-ss] + + + + +/-
(D) nonloc [-ss] +* +* + + +

+ = relatively preserved
- = impaired (omitted, substituted, misjudged)
+/- = either preserved or impaired
* = not tested, only predictions.

It appears that the unexpected datum for Tel concerns Ps


introducing locative [-ss]complements. These Ps were predicted to
be spared, but patients having problems with Ps introducing locative
[+ss] complements (three patients out of four) had also problems with
the [-ss] ones. Before facing this question, we summarize our results,
based on the completion task:
(i) performance affected with [+ss] nonlocative complements (41%);
more precisely, with Ps selected by V;
(ii) performance relatively preserved with [-ss] nonlocative
complements (75%): agent da of actional and nonactional
passives (with the exception of one patient); con (instrument,
listed separately); and con (manner) together with per (reason
and purpose cause);
(iii) performance from affected to relatively preserved with both [+ss]
and [-ss] locative complements.
As for the grammaticality judgment task, even if a decision task
has by definition a higher rate of chance hits than a completion task
(two alternatives vs about 15), the relevant overall performance was
from slightly affected to well preserved, and the resulting trends
were entirely compatible with (i)-(iii) above. The difference between
production and judgment is consistent with the literature on the topic
(see Linebarger et al 1983) and with our starting hypothesis, based on
the consideration of general expressive difficulties in agrammatism.
The rest of our discussion will be focused on the completion task.
The observed variability between the patients for locative Ps
was specifically predicted, but only for Ps heading [+ss] PPs, while
RO, AD, and LZ had problems with locative Ps in general. The reason

290
Recoverability of deletion in agrammatic production

why Ps in [+ss] locatives could result impaired (their omission could


be licensed by the Recoverability condition as formulated in (1) on a
par with all other Ps heading nonlocative [+ss] PPs) does not easily
extend to locative Ps heading [-ss] PPs.
This result is problematic for the Tel hypothesis, but also (if not
more) for the remaining accounts (see Table 1). Even Grodzinsky (1989)
predicts problems with the [+ss] but not the [-ss] locative PPs. The
other accounts predict that all locative Ps should be relatively spared.
As noted in the Introduction, the same conflict between
pronunciation vs. non-pronunciation expectation could extend to Ps
heading nonstrictly subcategorized locative PPs, if these have indeed
a special status in the linear order of adverbial PPs, as suggested in
Cinque (2002). More precisely, both (locative) PDat - that is locative
P in a [+ss] PP and P in a locative nonobligatory argument that
is locative P in a [-ss] PP would be recoverable. The former by
hypothesis, the latter due to the primary position of Location in
the relevant serial order. Their lexical-semantic nature would then
determine the relevant unpredictability of agrammatic behavior.
Indeed, for reasons internal to the present experiment, our
nonstrictly subcategorized [-ss] locatives bear a semantic-pragmatic
relation with the predicate, and somewhat do form a homogeneous
group with the strictly subcategorized [+ss] ones (see Rizzi (1988),
where a tripartite structural analysis of locative PPs is suggested:
verb complement, VP adverbial, and sentence adverbial, or scene
setting locative according to Cinque (1999)). The results of our
experiment draw attention precisely on the principled homogeneity
between [+ss] and [-ss] locative PPs, and this specific result is a
welcome example of the contribution of aphasiology to linguistics.
We have proposed that Recoverability of complement PPs is the
source of the impairment of Ps in the agrammatic behavior. Locative
Ps, however, even in recoverable contexts, are not predictably
impaired, due to their rich semantic content. Recoverability of locative
Ps originates from different sources. In [+ss] PPs the source is the
obligatoriness of the complement PP (as for any other [+ss] PPs), in
[-ss] PPs the source is their prominent position (possibly a reflex of
the obligatoriness of location for any material action): in both cases,
however, the fact that they are subject to lexical choice yields the
observed variability in agrammatic patients.
A conclusion that is compatible with both the old and new
data is that locative Ps are indeed a category by itself, for which
Recoverability relies on supplementary licensing conditions, and, at
the same time, is not a sufficient condition for deletion.

291
Lidia Lonzi et al.

Address of the Authors:

Department of Psychology, University of Milano-Bicocca


<lidia.lonzi@fastwebnet.it>

Notes
*
A preliminary study, based on the same hypothesis expounded here, was
presented at the poster session of the Academy of Aphasia, San Diego, October
1995 (Lonzi & Luzzatti 1995). We thank the audience of the Conference, and in
particular Esterella De Roo for stimulating comments and criticisms. Thanks are
also due to physicians and speech pathologists of the Rho-Passirana Rehabilitation
Unit (G. Salvini General Hospital) and of the Veruno Medical Center Neurology
Department (S. Maugeri Foundation).
1
Chomsky (1981:275) suggests that Agr can be an antecedent for the null that
is, unpronounced subject.
2
We must emphasize that the correlation between omission and substitution
errors inside nonlocative [+ss] complements was established on the basis of two
different modalities in L&L: completion for RO, and judgment for AD. Within
these respective modalities, the behavior concerning locative [+ss] PPs was bad
only on the part of RO.
3
The fact that the retrieval system of the closed class items in Brocas patients
appears to be organized by frequency (Bradley et al. 1980) is potentially a basis for
a principled differentiation of Ps. However, a strict application of the hypothesis
of the word frequency account is implausible, given that high frequency Ps (like
a and di, respective frequency 12.001 and 19.915, present only in [+ss] PPs) are
the most impaired in our patients performance; while low frequency Ps (like con,
frequency 2.721, present only in [-ss] PPs) appear to be spared. Word frequencies,
normalized to a corpus of 500.000 lexical entries, were obtained from De Mauro et
al. 1993.
4
This is precisely the case for [+ss] locative Ps: although these Ps are never
omitted (or judged omissible) by our subjects in the experimental conditions, they
are often so in conversation: a datum that singles out these Ps and, at the same
time, lends an explanation for their substitutions. Interestingly, in conversation,
locative Ps can somehow license the null form of verbs that select them:
directional a to, for instance, can substitute a verb of directed motion. Like in
other cases of grammatical zero forms of V (Van Riemsdijks 2002), it appears that
V is selected by P, as if P would provide the required antecedent for the omitted
V (see (1) in the text). This phenomenon was observed in ADs spontaneous speech
(see our transcription above, March 1991).
5
Belletti & Rizzis (1988) analysis is far from being assessed (see discussion
in Pesetsky 1995). However, it sorts out a peculiar subclass of Italian Object-
Experiencer Psych-Verbs for which it is arguable that the da-NP is indeed an
argument.
6
With the term complement, we refer to any argument of V, independently
from its syntactic realization either as a sister or adjoined node and from its
[+strict]/ [-strict] subcategorization, respectively (or primary/ secondary status
in the thematic hierarchy). We adopt the view that there are both obligatory and
nonobligatory arguments, which may be realized both in a right (complements
proper) and in a left branching structure (VP adjuncts) (Phillips 2003). We

292
Recoverability of deletion in agrammatic production

stipulate that complements proper receive their -role from V, and adjuncts
inherently from Ps. Notice that, in Grodzinskys (1989) analysis of, for instance,
John plays tennis on a clay court, the locative PP is attached to a higher projection
than VP.
7
The present state of research in this field does not allow us to assess
sufficiently firm principles. This applies also to the analysis of agent da by: for
many authors it transmits the external -role of the passive suffix rather than
assigning it in an autonomous way, as assumed here (see the notion of argument/
adjunct in Grimshaw 1990; and of -roles matching in Lonzi 1997).
Like for agent da, also for instrument con, contrasting analyses have been
proposed. According to Kayne (1994), instrument con assigns case to its
complement, but not the respective -role (on the topic, see Pascual-Pou 1999).
Nonetheless, different linguistic analyses admit of identical predictions, as it
also appears from the present study. Rather than taking a stand on the specific
questions involved, we should perhaps try to make clear why a given linguistic
analysis is suitable to provide an explanation for a given aphasic behavior.
8
The P is obligatory here. The verb is unaccusative and does not allow the
transitive alternation.
9
In the completion task, RO produced 24 times the preposition con with as
a default solution (AD 9, PG 11, and LZ 26 times). Occasionally this P yields a
nonstandard form with some sense, as, for instance, in: raccontare una storia con
gli amici, to tell a story with the friends [instead of to the friends]; mostrare un
libro con i compagni, to show a book with the mates [instead of to the mates], or a
standard one that does not comply with the relevant picture.

Bibliographical References

Belletti Adriana & Luigi Rizzi 1988. Psych-Verbs and -Theory. Natural
Language and Linguistic Theory 6. 291-352.
B arbosa Pilar, Danny F ox , Paul H agstrom , Martha M c G innis & David
P esetsky (eds.) 1998. Is the Best Good Enough? Optimality and
competition in syntax. Cambridge Mass.: MIT Press and MITWPL.
B ennis Hans, Ronald P rins & Jan V ermeulen 1983. Lexical Semantic
versus Syntactic Disorders in Aphasia: the processing of Prepositions.
Universiteit van Amsterdam: Publikaties van het Instituut voor
Algemene Taalwetenschap 40.
Bradley Dianne C., Merrill F. Garrett & Edgar B. Zurif 1980. Syntactic
Deficits in Brocas Aphasia. In Caplan 1980. 269-286.
Caplan David (ed.) 1980. Biological Studies of Mental Processes. Cambridge
Mass.: MIT Press.
C homsky Noam 1964. Current Issues in Linguistic Theory. The Hague:
Mouton.
Chomsky Noam 1965. Aspects of the Theory of Syntax. Cambridge Mass.: MIT
Press.
Chomsky Noam 1981. Lectures on Government and Binding. Dordrecht: Foris.
Cinque Guglielmo 1999. Adverbs and functional Heads: A Cross-linguistic
Perspective. Oxford: University Press.
Cinque Guglielmo 2002. Complement and Adverbial PPs: Implications for
Clause Structure. GLOW Newsletter 48. 23-24.

293
Lidia Lonzi et al.

De Mauro Tullio, Federico Mancini, Massimo Vedovelli & Miriam Voghera


1993. Lessico di frequenza dellitaliano parlato - LIP. Milano: EtasLibri.
Friederici Angela D. 1982. Syntactic and Semantic Processes in Aphasic
Deficits: The Availability of Prepositions. Brain and Language 15. 249-
258.
F riederici Angela D. 1985. Levels of processing and vocabulary types:
evidence from on-line comprehension in normals and agrammatics.
Cognition 19. 133-166.
Grimshaw Jane 1990. Argument Structure. Cambridge Mass.: MIT Press.
Grodzinsky Yosef 1989. Syntactic Representations in Agrammatic Aphasia:
The Case of Prepositions. Language and Speech 31. 115-134.
Grodzinsky Yosef 2000. The neurology of syntax: Language use without
Brocas area. Behavioral and Brain Sciences 23. 1-71.
Kayne Richard S. 1994. The Antisymmetry of Syntax. Cambridge Mass.: MIT
Press.
K ean Mary Louise (ed.) 1980. Grammatical Representation and the
Description of Language Processing. In Caplan 1980. 239-268.
Linebarger Marcia C., Myrna Schwartz & Eleanor Saffran 1983. Sensitivity
to grammatical structure in so-called agrammatic aphasics. Cognition
13. 361-392.
Lonzi Lidia 1997. On Lasniks Subject Role. Lingua e Stile 32. 397-432.
L onzi Lidia & Claudio L uzzatti 1995. Omission of prepositions in
Agrammatism and Universal Grammar Constraint on Recoverability.
Brain and Language 51 (1). 129-132.
Luzzatti Claudio, Alessio Toraldo, Maria Teresa Guasti, Graziella Ghirardi,
Lorena L orenzi , Caterina G uarnaschelli 2001. Comprehension of
reversible active and passive sentences in agrammatism. Aphasiology
15. 419-441.
Pascual-Pou Montserrat 1999. The Instrumental Phrase: is it an adjunct,
an argument or a predicate? Master Thesis, Universitat Autnoma de
Barcelona. Catalan Working Papers in Linguistics.
Pesetsky David 1995. Zero Syntax. Cambridge Mass.: MIT Press.
Pesetsky David 1998. Some Optimality Principles of Sentence pronunciation.
In Barbosa et al. (eds.) 1998. 337-383.
Phillips, Colin 2003. Linear Order and Constituency. Linguistic Inquiry 34.
37-90.
Renzi Lorenzo (ed.) 1988. Grande Grammatica Italiana di Consultazione. Vol.
I. Bologna: Il Mulino.
Rizzi Luigi 1985. Two notes on the linguistic interpretation of Brocas aphasia.
In Kean, Mary Louise (ed.). Agrammatism. New York: Academic Press.
153-164.
Rizzi Luigi 1988. Il sintagma preposizionale. In Renzi (ed.) 1998. 507-531.
Van Riemsdijk Henk 2002. The Unbearable Lightness of Going. The projection
parameter as a pure parameter governing the distribution of elliptic
motion verbs in Germanic. Journal of Comparative Germanic Linguistics
5. 143-196.

294

You might also like