Agrammatic Aphasia as Weak Syntax: A
Case Study!
Deniz TatUniversity of [email protected]
February 1, 2011
Abstract
In this paper, I look into various aspects of the grammar of an individualwith agrammatic aphasia, and show that the Weak Syntax model proposed byAvrutin (2006) can correctly predict these. I maintain that a Phase-Theoreticapproach provides us with the right tools to analyze agrammatic speech. Itappears that interpretable features (such as argument structure and phi fea-tures) and uninterpretable features (such as Edge features and tense) haveindependent statuses, and it is crucially the uninterpretable ones that are un-derspecified and give rise to agrammatic speech. The outcome is thus a morepowerful machine that can generate a wider range of sentences with increasedgrammaticality.
Keywords: agrammatic aphasia, Weak Syntax, interpretable and uninter-pretable features, Phase Theory, Broca’s aphasia
1 Introduction
In this paper, I closely look at various aspects of the grammar of an individual with
agrammatic aphasia (AA), and show that the Weak Syntax hypothesis of Avrutin
!Special thanks to my language consultant E.D. without whom this study would be impossible.I am also grateful to University of Arizona Aphasia Research Lab directed by Pelagie Beeson. Manythanks to Tom Bever, Bill Badecker, Andrew Carnie, Heidi Harley, Simin Karimi and MassimoPiattelli-Palmarini for their comments and suggestions. Any errors in presentation and content areof course mine.
1
(2006) correctly predicts these. Avrutin (2006) claims that the grammar of agram-
matic aphasics is reduced to Narrow Syntax, which can generate acceptable utterances
as well as grammatical ones. I show that the subject in this study has a grammar
which has expanded to include acceptable utterances in production and comprehen-
sion as Weak Syntax (Avrutin, 2006) would predict. I propose that a Phase-Theoretic
Approach to understanding AA provides us with the right tools to make the right pre-
dictions. In particular, I propose that interpretable phase features, uninterpretable
phase features and Edge features (which are also uninterpretable) have independent
statuses in AA. It appears that the defining property of AA is the impairment of
uninterpretable phase features. Interpretable phase features and lexically specified
features, on the other hand, largely remain intact while the status of moving to the
Edges in a derivation depends on how costly it is to do so. The model proposed in
this paper, along the lines of Weak Syntax of Avrutin (2006), unifies several previ-
ously suggested accounts of AA including the Complexity Account of Thompson and
Shapiro (2005, 2007), tense and agreement dissociations shown by de Bleser and Luz-
zatti (1994); Nadeau and Gonzales-Rothi (1992); Friedmann and Grodzinsky (1997);
Wenzla! and Clahsen (2003), and the Derived-Order Problem Hypothesis of Bas-
tiaanse and van Zonneveld (2005) and Yarbay-Duman et al. (2008) while it poses
some problems to the Tree Pruning Hypothesis of Friedmann (2001); Friedmann and
Grodzinsky (1997).
This paper is organized as follows: Section 2 establishes the framework taken up
in this study. I briefly describe AA, present Avrutin’s (2006) Weak Syntax, and then
o!er my proposal. Section 3 presents the method. I provide the reader with infor-
mation about the subject, materials and procedures and a summary of the results.
Section 4 discusses these results in the light of Avrutin (2006) as well as other pre-
viously suggested hypotheses, and suggests that a unified theory of AA is not a very
unrealistic goal. Section 5 is a summary of my conlcusions.
2
2 Framework
2.1 What is agrammatic aphasia?
AA is generally described as a language disorder which manifests itself in the form of
shorter sentences that lack function words. It typically results from a stroke or some
external trauma. The term acquired agrammatism is also sometimes used, but it is
mostly redundant as most cases of agrammatism are acquired later in life, mostly in
adulthood. Several models have been proposed to account for the grammar of AA,
which can be roughly divided into two with respect to their assumptions about causes
of di"culty in the production and comprehension of sentences. On one hand, a group
of hypotheses makes claims about a particular area in the derivational process, for
example the TP, (e.g Wenzla! and Clahsen (2003) and Friedmann and Grodzinsky
(1997)) while others make claims about the movement of elements as being sources
of agrammatism and thus make distinctions between underlying forms and surface
forms (e.g. Bastiaanse et al. (2002), (Yarbay-Duman et al., 2008), Grodzinsky (1995),
Thompson and Shapiro (2005, 2007)).
When researchers use the term AA, most of the time, they refer to Broca’s apha-
sia. In this type of aphasia, speech is typically non-fluent, naming of nouns and verbs
is poor while comprehension is largely intact. Agrammatism resulting from damage
to Broca’s area might be language-specific. For example, Friedmann (2002) showed
that agrammatic speakers of languages that have auxiliary raising from T-to-C (e.g.
English, and Dutch (Ruigendijk et al., 2004)) fail to produce these questions signifi-
cantly when compared to those agrammatic speakers of languages that do not have
auxiliary movement.
2.2 Weak Syntax
Avrutin (2006) proposes that the distinction between an unacceptable and an un-
grammatical utterance is crucial in our understanding of AA. He maintains that an
utterance can be considered unacceptable because of a variety of reasons, and that
only some of these are syntactic in nature. He exemplifies this by cases as in an
utterance would be typically considered ungrammatical but still acceptable in certain
3
contexts. For instance, * John dance would normally be considered an ungrammati-
cal one because it violates the requirement that subjects in English must agree with
their corresponding verbs. However, given the right context, John dance???!!! Never!
would be an acceptable utterance. In fact, our daily conversations are full of these
regardless of the language we speak. In Dutch, it is perfectly acceptable to drop the
definite marker in a conversation if the speaker makes sure that the right presupposi-
tion is already established with the hearer (Avrutin, 2006). Even though this is quite
common, and in fact a natural characteristic of speech, we never hear anyone utter
something similar to the... read a.... So, even the acceptable yet ungrammatical ut-
terances are rule-goverened possibly by means of some Narrow Syntax requirements.
It is therefore important to make a clear distinction between what is grammatical
and what is acceptable.
Avrutin (2006, p. 52) assumes Narrow Syntax of Chomsky (1995), which he defines
as “a computational system that is isolated and encapsulated with respect to meaning;
that is, that such system conducts symbolic operations on lexical items, putting them
together in some specific order that is allowed in a given language.” So, the outcome
of this system must meet the requirements of Full Interpretation (Chomsky, 1995).
Avrutin (2006) proposes that Narrow Syntax, the meaning-interface, remains intact
in AA while the information structure, or the C-I interface, becomes costly. Since it
becomes costly, the speakers have to attend to another level, which is typically the
context.
Reuland (2001) argues that operations that take place at di!erent levels based
on an economy hierarchy, the syntactic ones being the cheapest and the discourse
ones being the most costly (for unimpaired speech). Avrutin (2006) proposes that
damage to Broca’s area would make syntactic operations more resource-consuming,
which would then make the economy hierarchy di!erent for the individual with AA:
syntactic operations become more expensive than, or at least equally expensive as
the discourse operations. In other words, syntax - the system required for conveying
and processing information - becomes “weak.”
Avrutin (2006) discusses some of the previously shown evidence in favor of his
Weak Syntax model. For example, Baauw et al. (2002) found a correlation between
omissions of tense and determiners in Dutch-speaking children and aphasics. They
4
showed that the more likely a sentence was to contain tense, the more likely it was
that a determiner would also be produced, suggesting that these are equally available
(or unavailable) at the same time to the speaker because they are part of the same
level of operation, syntax. In another study, Bastiaanse and van Zonneveld (1998)
showed that when asked to complete a sentence with a missing verb, individuals
with AA produced finite forms only 49% of the time while, in embedded clauses,
their performance increased radically to 86%. This was possibly because discourse
becomes more costly in the formation of embedded clauses, which then causes the
speaker to attend to syntax.
In this study, I aim to see whether Weak Syntax of Avrutin (2006) provides us
with the right framework in understanding AA. Since he is basing his claim on the
notion of Narrow Syntax proposed by Chomsky (1995), I establish the details of my
proposal on Chomsky’s Minimalist Program (MP). In the next subsection, I discuss
the Strong Minimalist Thesis (SMT) of Chomsky, which was based on earlier versions
of MP (Chomsky, 1995), and then developed in Chomsky (2001, 2005).
2.3 SMT
The SMT of Chomsky (2001, 2005)) is based on the earlier versions of the MP, which
derives from the principles of Economy of Derivation and Economy of Representation
(Chomsky, 1995). Economy of Derivation was proposed by Chomsky to account for
movement in languages. He introduced the notion of feature checking, according to
which uninterpretable features must be deleted by their interpretable counterparts.
For example, Number is an interpretable feature of a count noun, let’s say child as
opposed to children. This is an inherent property of the noun that it is singular, not
plural, and this information is relevant and indispensable for its semantic interpreta-
tion. If we have this feature as the subject of a sentence, Number must be checked
against the uninterpretable features of the T. The subject therefore moves to the
specifier of the TP, so that feature checking is possible within this minimal domain.
Hence, for example, we have The child is crying. If feature checking fails to occur,
then the derivation crashes as in The child are crying. This means that sentences
must always delete their uninterpretable features before they are sent to LF and PF
5
- the Principle of Full Interpretation.
The Principle of Economy of Representation1, on the other hand, requires that
every representation must have a purpose. For example, the TP is required to realize
checking of several features, one of which is the crucial feature of finiteness. Chomsky
(1995) dispensed with AgrSP and AgrOP on grounds that these are di!erent from
other functional projections like TP. Functional projections have interpretable fea-
tures which are visible at LF while AgrSP and AgrOP have uninterpretable features
which are not visible at LF. In other words, the interpretable features of CP, TP and
vP are relevant to interpretation while agreement features are not. Chomsky (1995)
states that the agreement projections were required only for theory-internal reasons,
and in a Minimalist account of grammar, we can easily manage without these. In
this framework, agreement is explained through an operation of Agree, where the in-
terpretable phi features of lexical heads like Number, Person and Gender are checked
against their uninterpretable correlates found in functional heads in a minimal domain2. If they are not in the same minimal domain, then the element bearing the inter-
pretable phi features (goal) attracts the element bearing the uninterpretable features
(probe) thus resulting in a movement operation. As a consequence, the agreement
features are no longer needed for interpretation since they have already done their
job: making everything interpretable at LF. Agree therefore lost its status in the the-
ory as a phrase but instead became a relation between two elements whereby feature
checking occurs. The status of Agree remained similar in further works of Chomsky
(2001, 2005) despite changes in the status of movement as last resort. To sum up,
AgrSP and AgrOP were dispensed with within the MP while CP, TP, vP and DP
continued to be treated as functional categories.
Interpretable phi features are inherent properties of lexical words and they there-
fore already bear these features in the Numeration. These features can undergo
multiple checking operations against the uninterpretable features of the same kind.
For example, in Romance, the gender feature of a noun can check this feature against
1The notion of representation tends to be replaced in Minimalism with the notion of derivation.Not everyone agrees fully with this radical replacement, and I will not enter into this issue here.What follows does not depend on this choice.
2The minimal domain X would be all the elements the projections of X contains and dominatesexcluding the projections themselves.
6
the uninterpretable gender features of multiple adjectives, which means that only the
interpretable phi features do not get deleted after Agree. What is crucial for AA
is that in a scenario where these features never check against their uninterpretable
counterparts, the former still remain interpretable at LF.
Another defining characteristic of early MP was a reinterpretation of the notion
of overt and covert movement. Evidence from quantifier scope, wh-in-situ languages
and V-to-T raising as a universal property of languages suggested that movement
could occur after spell-out (i.e. sending elements to PF). Although we do not have
any physical means to observe these movement operations in the word order of the
utterances, we know that they exist because, for instance, languages having no overt
movement of wh- words to form questions show similar island e!ects observed in
languages that overtly move the wh- words (e.g. Chinese, Huang (1982). Accordingly,
a Y-model of derivation (an upside-down Y where the branches go to LF and PF)
was proposed, where derivational content was spelled out - but yet continued to work
on the required syntactic operations before it arrived at LF.
In Derivation by Phase, Chomsky (2001) suggested that there are multiple spell-
outs within one clause. CP and vP (and possibly DP) must be phases: Each of these
has some propositional value, a predicative content. These must be sent to LF and
PF separately to eliminate redundant cycles and computational load. In this cyclic
model, once a chunk of representations is sent to LF, it becomes inaccessible for fur-
ther derivations. Chomsky calls this the Phase Impenetrability Condition (PIC). The
computational system must forget about these. However, not everything is sent to
LF; some elements must remain for further syntactic building. Chomsky suggested
that the complements of phases must be sent to LF; the phase head and its specifier
must remain to make possible, for instance, successive cyclic movement of wh- words
through their Edges. He called these features that also need to be checked Edge Fea-
tures. Edge Features also help us retain the traditional A and A’ positions, movement
sites for argument and non-arguments respectively. (1) demonstrates these phases:
(1)
7
In later work, Chomsky (2005) pointed out that T can only probe when it is
selected by a CP. TP would be able check both the Agree features and the Tense fea-
tures if these were its inherent properties. Since this is not possible, Chomsky (2005)
proposed that these features must be inherited from a CP. The resulting framework
has come to known as the SMT, “which holds that language is an optimal solution to
interface conditions that FL [Faculty of Language] must satisfy; that is, language is
an optimal way to link sound and meaning, where these notions are given technical
sense in terms of the interface systems that enter into the use and interpretation of
expressions generated by I-language” (Chomsky, 2005, p.3).
Let us now summarize the basic premises of SMT. We have already said that Full
Interpretation requires that all uninterpretable features must be checked before they
are sent to LF. Once they are checked, they become interpretable and must be spelled
out immediately to avoid look-back (Richards, 2007, p. 566). That is valuation of
the uninterpretable features and transfer of these to LF must happen at the same
time. The uninterpretable features of T actually belong to C and are transmitted to
T via Feature Inheritance as we already said, from which follows that uninterpretable
features must belong to phase heads. A second premise we have is that it must not be
the entire phase that is sent to LF because otherwise the derivation cannot continue.
8
This is why Chomsky (2005) suggested that the complement of the phase, a non-
phase must be sent to LF. It follows that the Edge of the CP and its complement TP
must be transfered separately.
2.4 Proposal
In the previous subsections, we have summarized the basic assumptions of Weak
Syntax (Avrutin, 2006) and the SMT (Chomsky, 2001, 2005). Based on these, let us
now discuss how we can characterize AA. Consider the following sentence:
(2) Whoi [ i kicked the rabbit]?
(2’)
In (2), the subject has interpretable features [+sg] and [+3]. We know that because
these are lexically specified in the word who. These features make this lexical item
have a semantic content that is visible at LF. This word also has an Edge feature [wh-
]. Another lexical item that has lexically specified features is the verb kick. Although
we have not discussed argument structure so far, we can say that it is also lexically
specified. That is, a verb like kick must have an agent DP and a patient DP. I
propose that individuals with AA do not have considerable di"culties with lexically
9
specified features. Since these are inherent properties of lexical items, they must be
indispensable parts of them. As long as the individual with AA can access these
words, these inherent properties must be intact. Furthermore, if these assumptions
about lexical specification are correct, then verbs that are lexically specified for tense
such as went must also be easier than those that need morphological tense marking
kicked. I therefore o!er the following hypotheses:
(3) the status of lexically specified features in aa
(a) Argument structure must be intact.
(b) Interpretable Agree features must remain intact.
(c) Suppletive past and participle forms of verbs must remain intact.
In (2), we also have a CP, which has the Edge feature [wh] that needs to be
valued. I hypothesize that this is the aspect of AA that is responsible for most of the
variability among and within individuals with AA. Valuation of the Edge features is
central to human language because it ensures that phrases can be built up by making
the Edges available to the next cycle but at the same time these are the most costly
movements because they involve long-distance dependencies. Since AA (and perhaps
all kinds of aphasia) is accompanied by several other processing di"culties like shorter
memory restrictions, longer times for lexical access and possibly several others, I am
convinced that these long-distance movements are di"cult for individuals with AA,
not necessarily because the Edge features are underspecified, but probably because,
as I said, due to the computational load of these operations.
(4) the status of Edge features of aa
(a) Their status varies depending on how costly it is to value them.
We also have the uninterpretable Agree features [sg] and [3] and other uninter-
pretable functional features like Case. These are inherited by the T head. We have
already said that the TP is central to AA, and there are several studies that show
that tense features - although it is assumed that these belong to the TP - are im-
paired even if agreement remains intact (e.g. de Bleser and Luzzatti (1994) for Italian,
10
Nadeau and Gonzales-Rothi (1992) for English, Friedmann and Grodzinsky (1997) for
Hebrew, Wenzla! and Clahsen (2004) for German). Since these are uninterpretable
features that need valuation in a non-phase, we might make a generalization and
claim that all uninterpretable features of T are impaired or underspecified. I would
like to use the latter term because the former may presuppose that TPs are com-
pletely deleted in AA. Note that I am not making the claim that the TP or the CP
disappears in AA.
(5) the status of uninterpretable features of CP
(a) They are underspecified.
So far, we have talked about the uninterpretable features of C but not mentioned
its interpretable features. First of all, we can say that there must be interpretable
features for the CP to be a visible representation both at LF and PF. I claim that
what makes CPs visible are at least two things: [+/-Realis] and [+/-Neg]. CPs are
propositions: they must have presuppositional semantics so that we can talk about
factives and non-factives in the world. Factives are propositions that must certainly
be true. Present and past events are factives: They are states or events that happened
in the past or are happening in the present. We are also capable of communicating
our wishes, hypotheses and expectations about the future or in some hypothetical
world. These are non-factives - things that have not yet happened or will possibly
never happen. We also have tools to negate both factives and non-factives. We can
talk about things that have not happened or will not happen. I claim that these
two are the interpretable features of C. Like other interpretable features, these must
remain intact in AA. I suggest that the relevant tense or aspect features of these are
uninterpretable features of the C and are thus subject to be underspecified in AA.
Table 1: Interpretable features of CP and their corresponding uninterpretable features+ Realis - Realis
+ Neg Past and present a!rmativefactives
Negative conditional and pur-pose (e.g. Unless..., lest...)
- Neg Past and present negative fac-tives
Future, conditionals, subjunc-tives
11
The above categorization of interpretable phase features has one thing in common:
In all cases, we can talk about a semantic truth-value. For example, in a sentence
like John went to the movies yesterday evening (+realis, -neg), we know that it is
true that at some point in the evening of the day preceding the time of the utterance,
John went to the movies. Similarly, a sentence like John wishes he were at the movies
now, we know that the truth-value of John being at he movies at the time of speaking
is False. There is another set of sentences, which do not have a truth-value similar
to the examples just given, but they are special in the sense that they provide a set
of possible truth-values, namely, questions. Yes/no questions are a set of exceptional
questions, where it is presupposed that a statement has a truth-value, in which we are
interested in learning whether it is True or False. In wh- questions, we are a bit more
flexible, but again limiting ourselves to a set of possible answers in a given world.
For example, the wh- element which has a special property of limiting the possible
truth conditions to a given context. Because of these properties of questions, we can
consider that a feature like +/- Q is also an interpretable feature.
(6) the status of interpretable features of CP
(a) They remain intact.
2.5 Predictions
In the previous subsection, I have proposed that interpretable, uninterpretable and
Edge features have independent statuses in AA, and suggested that it is crucially
the underspecification of uninterpretable phase features that give rise to the main
characterizations of AA. I have also suggested that, since Edge features require the
movement of elements in a derivation, they are computationally costly, and thus, their
status depends on how costly it is to derive them. Based on these hypotheses, and the
premises of Weak Syntax of Avrutin (2006), we can make the following predictions
about the grammar of the subject in this case study:
(7) He should be able to talk about factives and non-factives and negate these even
if the uninterpretable features associated with these are underspecified. He
should also be able to form both yes/no questions and wh- questions.
12
(8) Tense features and Case should be the main source of agrammatism. He should
therefore have regular past tense verbs consistently not marked for tense in the
matrix clause, and have auxiliaries underspecified in form possibly syncretized
for all. Tense features that normally lack in his production of simple clauses
should appear in embedded clauses as he would be relying on the syntactic tools
in embedding rather than discourse tools.
(9) Interpretable features of nouns as person, number and gender should be intact.
For example, he should not show any errors in comprehending or producing the
correct pronouns for female and male as well as number agreements between
subjects and their corresponding verbs.
(10) Argument structure should be intact. Because these are lexically specified in
the verbs themselves, as long as the subject has access to these verbs, he should
have access to their argument structure as well.
(11) Movement to the Edges must be selectively impaired. For example, shorter
movement operations like subject wh- movement should be easier than object
wh- movement.
(12) The errors in his production or comprehension should be grammatically accept-
able sentences in English, or maybe, in any given natural language as allowed
by Universal Grammar.
In the next section, I describe the method in this study including the subject,
materials and procedures, and the results. The data elicitation tasks aim at having
a wide range of sentence types to account for the predictions made above, and do
not necessarily target only one or two kinds of sentences. The sentence production
data is based on Thompson (n.d.)’s NAVS protocol along with sentence types like
unaccusatives, binding, control, ECM and raising constructions. Comprehension data
is based on grammaticality judgements of all of the sentence types. Having a wide
range of sentences has the advantage of coming across things we have not predicted
before although it has the expected disadvantage of having not quite direct statistics. I
would like to consider this data collection as analogous to doing field work. Assuming
13
that the subjects grammar has diverted from what it used to be, we might well
consider that his grammar has become a unique idiolect of English: It has English
words and structures but it is di!erent enough to be called a variety.
3 Method
In this section, I o!er a brief description of the subject, who took part in this study.
I then present the materials and procedures. These include the full NAVS proto-
col (Thompson, n.d.), supplementary materials and grammaticality judgements. I
provide the results to each of these data collection procedures as I present them in-
dividually to avoid any possible confusion. A summary of the results ispresented at
the end of this section.
3.1 The subject
The characteristic of the individual with AA looked at was as follows (Pelagie Beeson,
personal communication November, 2009): The subject is a 48 year-old, left-handed
man. He had a massive left hemisphere stroke during a surgical procedure, which
resulted in severe aphasia due to damage to the entire left perisylvian language regions
including Broca’s area, Wernicke’s area, the supramarginal gyrus and the insula. He
was first diagnosed as having Brocas aphasia at the University of Arizona Aphasia
Research Project two years post-onset. His speech lacked verbs completely and his
sentences were limited to two to four-word utterances. In time, his speech improved
in number of words and complexity, and he is now considered as a border-line aphasic.
3.2 Materials and procedures
In this section, I outline the materials and procedures: NAVS (verb naming, verb
comprehension, sentence formation, sentence production priming and sentence com-
prehension) supplementary materials (naming and recognition of unaccusative verbs,
supplementary sentence production priming and story telling), and grammaticality
judgements (of all the sentence types used in NAVS sentence production priming and
14
supplementary sentence production priming). I provide the results of each task as we
discuss each of them, but provide a summary of all at the end of this section.
The data presented in this study were collected in three sessions. In the first
two sittings, the full NAVS protocol was conducted. This was carried out in the
University of Arizona Aphasia Lab. The first half of the protocol was video-taped.
In the third sitting, the supplementary materials and the grammaticality judgement
tasks were used. This session was carried out in a quiet setting in a University of
Arizona classroom and was audio-taped.
3.2.1 NAVS
First, the full NAVS protocol of Thompson (n.d.) was followed. This was done in two
sessions. In the first session, the first three sections of the protocol were performed.
These include verb naming, verb comprehension and sentence formation with di!erent
argument structures. In the second session, sentence production priming and sentence
comprehension sections were implemented. I summarize the procedures and results
of these five tasks below:
Verb naming
In this task, the subject was presented with a set of 34 pictures that represent a
one-place predicate, a two-place predicate or what can be characterized, for the sake of
simplicity, as a three-place predicate. The one-place predicates involved unaccusatives
like dive, and unergatives like laugh. The two-place predicates included two categories:
transitive verbs that require an object like cut and transitive verbs that optionally
take an object like drive. And lastly, what can be considered three-place predicates
were those that take three arguments. These also may take an obligatory indirect
object like put or an optional indirect object like write.
The subject could name all the one-place predicates, most of two-place predicates
(62.5% of the ones with obligatory direct objects and 75% of the ones with optional
direct objects) but had di"culty with the three-place predicates. Although he could
name all the verbs that take an obligatory indirect object (there were only two of
them), he mostly failed to name verbs that optionally take indirect objects (37.5%
of the time correct). The average performance for two-place predicates and for the
15
three-place predicates is the same (68.75%).
Verb comprehension
In the verb comprehension, the subject was shown a set of four pictures and
was asked to point at the one he heard. The pictures involved the correct answer, a
distractor in the same verb category of the target verb, and two distractors in di!erent
verb categories. The subject did not make any errors in this task. Of the same words
used in the verb naming task, he pointed at the right picture at all times.
Sentence formation
In the sentence formation, he was provided with a set of 50 pictures that showed
an action. For each picture, the arguments as well as the verb itself were all provided
in print carefully placed around the picture, which did not always follow the order of
arguments from left to right to avoid any sentence formation facilitation. Figure 1 is
an example to illustrate this.
Figure 1: Example picture stimulus for sentence formation
The subject performed almost perfectly with argument structure. The results for
this section showed that he had one missing subject in an unergative construction
among 8, one missing internal argument in a transitive among 24, and 3 missing
optional indirect objects in a 3-place predicate. What is significant in this section
(yet not addressed in the evaluation criteria of the NAVS protocol) is the (non-)use
of tense. The subject did not produce tense in these sentences 46% of the time. For
16
those grammatical sentences, he chose the simple present instead of the progressive
for 5 of the sentences out of 27.
Below is a list of productions that involved an error in either tense or argument
structure, or both 3 . It also lists the grammatical sentences where he used the
present simple tense even if the practice items involved only the present progressive.
As illustrated, the subject has a significant tendency to leave out the auxiliary be and
either leave tense untreated or use the present tense marker -s for third person. In
one case, he used the past tense (2). In this sentence, he communicates a ditransitive
in two clauses that are coordinated.
1. Bark bark bark dog says.
2. Man sent a package so the woman got it.
3. Woman sitting.
4. Boy dives in...
5. Man wash clothes.
6. Man drive.
7. Baby crawling.
8. Man eating.
9. Postman delivers package.
10. Girl pulls boy.
11. Man sweeping lawn.
12. Woman throwing the ball.
13. Man dig the hole.
14. Man eat food.
15. Man laughing.
16. Dog its watching the television.
Some of the items in this task lack an optional direct or indirect object. Even
though these elements were not presented in the picture stimuli, the subject chose to
add them in his productions. Below, the items in parentheses were not in the picture
stimuli, but were produced by the subject. Preference for the unmarked argument
3The subject omitted the determiners in most cases, especially in the subject positions, but I didnot consider these as a source of ungrammaticality.
17
structure for these verbs indicates that argument structure is mainly intact in his
grammar:
1. Man sweeping (the lawn).
2. Man dig (the hole).
3. Dog its watching (the television).
4. Man swimming (on the...)
5. The cat is watching the dog (into the television).
6. Cat biting (the dog).
7. The father is reading (a book) (to the son).
Sentence production priming
In the second session, the subject was given the sentence production priming
task of NAVS. In this task, he was presented with a pair of pictures along with the
description of the first picture in: an active sentence, a yes/no question, a subject
wh-question, a subject relative, an object wh-question, a passive, or an object relative.
There were 5 sentences for each sentence type. The subject was then asked to describe
the second picture in the pair using the same structure. His utterance in the first
fifteen seconds is then evaluated. He was allowed to self-correct himself within this
period. He performed best in active sentences and subject wh- questions while poorest
in object relatives:
Table 2: Sentence production priming performances.Sentence type Performance Example
Object relative 0% There’s the squirrel who the rabbitis kicking.
Yes/no question 20% Is the rabbit kicking the squirrel?Subject relative 20% There’s the rabbit who is kicking the
squirrel.Passive 20% The squirrel is kicked by the rabbit.Object wh- question 80% Who is the rabbit kicking?Subject wh- question 100% Who is kicking the squirrel?Active sentence 100% The rabbit is kicking the squirrel.
It is important for the purposes of this study to qualitatively look at the errors the
18
subject made in the sentence production priming task. Table 3 is a list of the items
that the subject could not produce the target sentence within the expected amount
of time. For the object wh- question, he gave only one wrong answer. However, his
utterance involves a moved wh- item, the auxiliary and the subject. The missing verb
might be due to some di"culty in recalling the verb pulling.
Passives, subject and object relatives and yes/no questions on the other hand
seem to be explicit areas of di"culty. His utterances for passive sentences all include
the patient subject but have some di"culties specific to the complex nature of the
passive construction. In subject relatives, the expletive is missing in his responses that
were counted wrong in the NAVS protocol. In these sentences, he tends to produce
a mono-clausal version leaving out relativization. Object relatives, the relativized
object noun always precede the subject of the embedded clause. These sentences
however lack the wh- word. Yes/no questions are simply avoided by producing the
non-question versions of these sentences which lack the movement of the auxiliary
verb with the correct rising intonation for questions.
19
Table 3: Error analysis for sentence production priming.Stimulus Target Response
Object wh- questionWho is the boy pulling? Who is the girl pulling? Who is the girl... ?PassiveThe cat is chased by thedog.
The dog is chased by thedog.
1. The dog chasing... no2. The dog is chased... 3.The dog chased the cat.
The girl is pulled by theboy.
The boy is pulled by thegirl.
1. The boy is... 2. The boyis pulled from the girl.
The man is kissed by thewoman.
The woman is kissed by theman.
1. The woman is kissed tothe man. 2. The woman iskissed by the man.
The girl is shoved by theboy.
The boy is shoved by thegirl.
1. The boy is shoving. 2.The boy is shoved?
Subject relativeThere’s the rabbit who iskicking the squirrel.
There’s the squirrel who iskicking the rabbit.
1. One more time. 2. Thesquirrel is kicking the rab-bit.
There’s the boy who isshoving the girl.
There’s the girl who isshoving the boy.
The girl is shoving the boy.
There’s the boy who ispulling the girl.
There’s the girl who ispulling the boy.
The girl is pulling the boy.
There’s the dog who ischasing the cat.
There’s the cat who ischasing the dog.
The cat is chasing the dog.
Object relativeThere’s the girl who theboy is pulling.
There’s the boy who thegirl is pulling.
There’s the boy the girl ispushing.
There’s the squirrel whothe rabbit is kicking.
There’s the rabbit who thesquirrel is kicking.
1. There’s the rabbit kick-ing. 2. The rabbit is...
There’s the girl who theboy is shoving.
There’s the boy who thegirl is shoving.
The boy is shoved by thegirl.
There’s the cat who thedog is chasing.
There’s the dog who thecat is chasing.
The dog... the cat chasing.
There’s the man who thewoman is kissing.
There’s the woman whothe man is kissing.
1. The woman is kissing.2. The woman the mankissing.
Yes/no questionsIs the dog chasing the cat? Is the cat chasing the dog? The cat is chasing the dog?Is the rabbit kicking thesquirrel?
Is the squirrel kicking therabbit?
The squirrel is kicking therabbit?
Is the boy pulling the girl? Is the girl pulling the boy? The girl is pulling the boy?Is the woman kissing theman?
Is the man kissing thewoman?
The man is kissing thewoman?20
Sentence comprehension
The subject was lastly given the sentence comprehension section of NAVS. In this
task, he was presented a pair of pictures, given a sentence and then asked to point
at the picture which the presented sentence described. The sentences again involved
active sentences, yes/no questions, a subject wh- questions, subject relatives, object
wh- question, passives and object relatives. He pointed at the correct picture at all
times except in an object relative construction. For the sentence, There’s the girl who
the boy is pulling, he pointed at the picture where a girl was pulling a boy.
3.2.2 Supplementary data
The NAVS sessions elicited a crucial part of data in this study. But, the test is aimed
to test only a particular kinds of verbs and sentences. For example, verb naming
and comprehension had only a general category of one place-predicates, and did not
distinguish between unaccusative versus unergative verbs. The sentence structures
looked at did not have any constructions with unaccusatives, control and raising,
anaphors and ECM. The sentences in NAVS were also all based on third person
singular characters in the pictures, and thus did not yield any understanding as to
whether the subject distinguished between di!erent persons in agreement. To see
how the subject would perform in these constructions, a series of tasks has been
conducted.
For comprehension, he was given two tests: naming unaccusative verbs and gram-
maticality judgments. For production, he was given three tests: unaccusative verb
production, sentence production priming including structures not covered in NAVS
and a story telling. Prior to the session with the subject, all of these tests were ini-
tially conducted with a 34-year-old male native speaker of English with unimpaired
grammar to make sure all the pictures were good representations of the verbs and the
sentences, as well as to confirm the grammaticality judgments for sentences in the
judgement task.
Recognizing unaccusative verbs
21
The subject was tested to see if he had any di"culty comprehending verbs that
involved unaccusative verbs. Verbs break, arrive, burn, fly and melt were presented
in the same way pictures were presented in NAVS verb comprehension task. He was
shown a set of four pictures. One picture represented the target unaccusative verb;
another picture represented an unaccusative distractor. The other two pictures were
representing an unergative distractor and a transitive distractor. The subject did not
have any di"culty pointing at the right picture for all five verbs.
Unaccusative verb naming
The subject was presented some pictures and asked to name what is happening
in each. The target verbs were the same unaccusatives used in verb recognition test.
However, to avoid any potential short term memory e!ects, these two tasks were given
in di!erent halves of the session divided by a break. The subject was instructed that
he is expected to name the action, the verb, or what is happening in the picture, and
that he does not have to make a full sentence. Below is a list of verbs used in this
task along with the responses elicited from the subject.
Table 4: Naming unaccusative verbsStimulus picture Response
burn fire, smoke and firefly superman, go into outer space... ...
flyingarrive the groceries and the mama go into
the house... with... coke and... go-ing
melt ice is going, do the water, ice is...cold... ice getting warmer
break broken, plate broken into two... ...two, one plate and... breaks into...
For each of the five verbs, the subject was given as much as time he needed to
come up with verbs. He was encouraged at all times to further describe what is
happening in the picture. Among the five verbs, the subject came up with the right
verb only twice. For the first picture, he could only come up with nouns like fire and
22
smoke. For the verb fly he was presented a man flying like superman. The subject
came up with the verb fly only later after being asked to further tell us what is going
on in the picture. Arrive was a di"cult verb for the subject4. His responses for melt
and break were interesting. For the former, he gave us a very good description of
the change of state in the ice cube without using the verb melt at all. For break, his
initial response involved the participle form of break.
Supplementary sentence priming
To supplement NAVS sentence production priming, a similar test was given to the
subject. He was presented a set of pictures in pairs. First, he was given a sentence for
the first picture and then was asked to make a similar sentence for the second picture.
These sentences were designed to elicit subject agreement, unaccusatives, anaphoric
binding, control, raising and ECM. Five pairs of pictures were used for each category.
The subject provided correct sentences for all the agreement items. He had dif-
ficulty in 2 of the unaccusatives, 3 of the ECM constructions and 4 of the binding,
raising, and control constructions each. Table 5 is a list of items, where he provided
a wrong (but not necessarily ungrammatical) response.
4I am aware that the picture chosen for this verb may not be a good representation of the eventarrive.
23
Table 5: Error analysis for supplementary sentence production priming.Stimulus Target Response
UnaccusativesThe man arrived home. The woman arrived home. The woman arrive home.The ice-cream is melting. The ice is melting. The ice melts.BindingThe boy hurt himself. The girl hurt herself. The girl hurting... self... The
girl hurting... The girl hurts.The girl admires herself. The boy admires himself. The boy is admires himself.The man sees himself in themirror.
The woman sees herself in themirror.
The woman sees a man[laughs] ... The woman seesherself..
The man is enjoying himselfat the party.
The woman is enjoying her-self at the party.
The woman enjoys herself.
ControlThe man wants to go on va-cation.
The woman wants to go onvacation.
1. The woman and the manwent to vacation. 2. Thewoman is... also vacation.
The man wishes to get mar-ried.
The woman wishes to getmarried.
The woman married.
The woman hopes to go tocollege.
The man hopes to go the col-lege.
The boy hopes... to... highschool [laughs]... college too.
The woman wants to fly. The man wants to fly. 1. The man... the man...go... would... 2. The man...3. The man wants to fly.
RaisingThe girl is likely to crash intoa tree.
The boy is likely to crash intoa tree.
1. The boy is watching...the... the boy is actually...[laughs] the boy is... 2. Theboy is likely to crash and...burn.
The girl seems to be hard-working.
The boy seems to be hard-working.
The boy is hardworking.
The girl appears to be happy. The boy appears to be happy. The boy is appears to behappy.
The woman is likely to falldown.
The man is likely to fall down. The man is {slicely}... Theman {slightly} to... get down.
ECMI expect her to be successful. I expect him to be successful. 1. ... 2. I expect to... 3. I
expect him to [unintelligible]I expect her to crash into atree.
I expect him to crash into atree.
I expect him to crash andburn.
I believe him to be a nice per-son.
I believe her to be a nice per-son.
1. I believe her to... 2. I be-lieve her to do it... 3. I be-lieve her to... 4. I believe herto do the same.
24
The errors in sentences that had unaccusative verbs were either because they
lacked a tense marker or due to a preference for present tense. As I mentioned
earlier, the subject has a tendency to use present tense instead of the progressive.
And when he uses the progressive, he has a tendency to omit the auxiliary be. For
sentences that have anaphora, he did not have any problems with person agreement.
With the case of hurt oneself, he could not come up with the word herself and then
remembered that The girl hurt is a grammatical English sentence too (although it
does not have the exact same meaning). He tends to omit control constructions and
just provides a simple, finite clause, or he has di"culty incorporating the embedded
clause. In the case of The woman wants to fly, he provided the right answer after he
heard the initial structure three times suggesting that it is very likely that this is due
to short term memory e!ect. Raising is a case where he has a lot of di"culty too. He
shows a preference for simple finite clauses in this case too. In ECM constructions,
the significant errors have a pattern like I expect him but then the rest is problematic.
Story telling
The subject was also given a picture strip story for production. He was instructed
to look at the picture as much as he wants and then tell the story. The picture
used in this task was chosen because it had two di!erent characters, one male and
one female, which would yield gender agreement on pronouns. Also, the story line
consists of action that could help us elicit various kinds of verbs like an unaccusative
(e.g. arrive), unergative (e.g. walk), a transitive (e.g. hit the man), and possibly a
ditransitive (e.g. put the jacket in the closet). The subjects description was limited
to the following:
“Getting home to wife ... and wife um ... um ... looks ... marks and cigarette or
cigar and ... I dont know ... shirt o! or jacket o! and trousers o! and shirt and that
is enough [pounds on the table] mad, mad, mad her.”
25
Figure 2: Cartoon-strip for story telling
Piyale Madra, Radikal
The subjects utterances at 15 years post-onset in Aphasia Bank (MacWhinney,
2007) were looked at. The initial parts of the AphasiaBank protocol aims to elicit
narration of past events. Below is the transcript of the first parts of the session, where
he was asked to talk about (i) when and how he had a stroke as well as (ii) a story
that he would like to tell from his past.
Int: Do you remember when you had your stroke?
Part: um eighteen years ago.
Int: Can you tell me about it?
Part: um A . V . M then stroke. Doctor’s Hospital in San Francisco . stroke in .
hospital.
Int: What’s the first thing you remember after your stroke?
Part: Seventy three days . um got a call and that’s it.
Int: Tell me about your recovery. What kinds of things have you done to try to
get better since your stroke?
26
Part: um better in . um auditory . not better [touches wheelchair] in the chair
Int: Tell me about any kind of changes you’ve needed to make in your life.
Part: um after the hospital good things [thumb up] . I like it.
Int: Now, I’m gonna ask you to do a few more things where you need to talk.
Please talk as much as you can about each one cause we’re interested in finding out
about your language.
Part: Alright, I will.
Int: Thank you. Thinking back can you tell me a story about something important
that happened to you in your life? It could be happy or sad, um from any time, when
you were a kid or more recently.
Part: Me and Doug Smith um . and two [gesture two] bikes. and . [gesture up] .
um . go up [gesture up] and . um Dixon Avenue in Charlevoix, Michigan. And . I
guess the only thing is . um me and Doug Smith went crash.
The subject’s responses are typically reduced, short answers that lack verbs. The
two verbs that are in the past tense form are both irregular verbs: got and went.
3.2.3 Grammaticality judgements
For comprehension of sentences, the subject was also given a grammaticality judg-
ment test. He was simply instructed to say “good” for sentences that he thought
sounded grammatical and “bad” for those he considered to be ungrammatical. The
sentences consisted of grammatical and ungrammatical versions of the sentence types
looked at in NAVS (actives, passives, subject wh- questions, subject relatives, ob-
ject wh- questions, object relatives and yes/no questions) as well as sentence types
to test comprehension of structures that are not necessarily looked at in NAVS: un-
accusatives, agreement, anaphoric binding, control, raising and ECM. The subject
performed 100% correctly for all the grammatical sentences among the 64 sentences
(26 were grammatical). He called most of the ungrammatical sentences “bad” but
not all of them. The following is a list of ungrammatical sentences incorporated into
the full list, which was shu#ed to avoid short term memory e!ects as they were pre-
sented (for the whole list, see Appendix 2). The bold faced sentences are the ones
the subject considered good.
27
Table 6: List of ungrammatical sentences used for grammaticality judgements. Bold-faced sentences were considered to be “good” English sentences by the subject.
1. * The dog chasing the cat. 20. * This is the boy which the girl
is chasing.
2. * The woman is kissing to the man. 21. * The is the rabbit that the squirrel iskicking which.
3. * There’s tickling a boy. 22 * The airplane is arrived.
4. * Are the boy pulling the girl? 23. * There’s arriving a man.
5. * Mary wonders who is the boy
kissing.
24. * The woman is arrived home.
6. * Is who chasing the boy? 25. * The boy and the girl is reading.7. * Which the dog is chasing the cat? 26. * The man and the woman is
walking.
8. * This is the cat is chasing the dog. 27. * The boy are flying.9. * This is the squirrel is kicking the rab-bit.
28. * I saw me in the mirror.
10. * This is the boy is who shoving thegirl.
29. * The man admires myself.
11. * These are the girls are shoving theboys.
30. * You injured herself in the accident.
12. * Who the boy is kissing? 31. * The woman demanded the man
to buy a house.
13. * Which girl the boy is shoving? 32. * The boy hopes himself to go to col-lege.
14. * Which the dog is chasing the cat? 33. * It appears the boy to be sad.
15. The dog is barked by the cat. 34. * The girl seems it to be happy.
16. The dog has chased by the cat.. 35. * There seems a woman to fall down.17. * There’s chasing a dog the cat. 36. * I believe he to be a nice person.
18. * There’s kissed a woman by the man. 37. * I expect she to be sad.19. * This is the man the woman is kissingwho.
38. * I expect her fall down.
His judgement for 1 indicates that tense is not necessarily a production di"culty
but also a comprehension di"culty suggesting maybe this is a general property of his
grammar. Sentence 5, 6, 12 and 13 show that head-movement of auxiliaries is an area,
where he has di"culty most because he could not produce any sentences in the form
of yes/no questions in the NAVS sentence production priming either. In addition to
the present progressive auxiliary be and its movement for questions, he has di"culty
28
interpreting passive be too. He considered 16 grammatical, possibly interpreting has
as a passive auxiliary. It might also be the case that he interpreted this sentence as
having a perfective aspect and left the by phrase untreated. But this is not very likely
considering that he has little di"culty with argument structures.
Constructions with unaccusative verbs have quite interesting results. The par-
ticipant considered 22 and 24 grammatical, which may not be surprising from a
theoretical perspective at all. Since the arguments appearing in subject positions
are underlyingly the objects of the sentence, it may not be surprising that passive
forms of these were considered grammatical by the participant. He judged 23 as be-
ing grammatical as well as a grammatical form of this which is not in the list above
(There arrived a man at school). It may be well the case that he does not distinguish
between these two sentences is because the present auxiliary be in (23) is not inter-
preted. Similarly, he considered a sentence with an agreement error grammatical (25)
but another very similar sentence ungrammatical (26) showing once again that the
present verb be is an unpredictable area. We might say that he cannot distinguish
between the auxiliary be and have.
With Binding Condition A, he showed no di"culty in sentences with agreement
errors (e.g. 29 and 30), but considered 28 good, where the object must be bound.
31 is a case, where he has di"culty judging a control verb whose PRO is substituted
with an overt object. 33, 34 and 36 show the participant has inconsistent judgements
for raising and ECM constructions.
3.3 Summary of results
• Argument structure is mostly intact. In sentence formation task of NAVS, he
omitted one argument each in 5 sentences among 50. Three of these were op-
tional arguments in a three-place predicate. He used these NPs by conjunction.
(e.g. Man sent a package so the woman got it instead of The man is sending
a package to the woman). (His use of tense in these sentences were at chance
level.) He even added direct objects to verbs such as dig and sweep even if these
were not provided in the picture sets. In the grammaticality judgments, he had
one incorrect judgement for an item testing argument structure. (He considered
29
The woman demanded the man to buy a house as correct).
• The subject could name all the one-place predicates, 68.75% of the two-place
predicates (62.5% of the ones with obligatory direct objects and 75% of the ones
with optional direct objects) but had di"culty with the three-place predicates.
He could name all the three-place verbs that take an obligatory indirect object
(there were only two of them), he failed to name verbs that optionally take
indirect objects 37.5% of the time correct. The average performance for two-
place predicates and for the three-place predicates is the same (68.75%). He
had no di"culty comprehending these verbs.
• The subject came up with the right unaccusative verb 2/5 of the time in the
unaccusative verb naming task. Even if he was just asked to name the action
or the verb for these pictures, he relied on nouns or the participle forms of
these verbs in all items. For the verb melt, he gave us a very good description
of the change of state in the ice cube without using the verb melt at all. In
unaccusative sentence priming, he gave 3/5 of the sentences correct. 2/5 were
wrong due to ungrammatical or wrong tense morphology not due to argument
structure.
• He had no di"culty in active sentences in sentence production priming of NAVS.
His passives were on the other hand 1/5 of the time correct. However, the
ungrammaticality did not result from argument structure: In all sentences, he
started with the right NP but failed to give the correct auxiliary 3/5 of the
time. He had no di"culty in comprehending passives or actives.
• In the production of the subject relatives, he gave a simple version of these
sentences 4/5 of the time. For example, instead of Theres the cat who is chasing
the dog, he produced The cat is chasing the dog. In object relatives, all his 5
utterances varied in their structure but they all started with the object of the
main verb. His comprehension of subject and object relatives was 90% correct
with one mistake in object relative.
• He had significant di"culty in yes/no question formation. 4/5 of the time, he
produced the non-question versions of these with no tense problem correctly
30
producing the suppletive form of ‘to be’. In other words, the auxiliary verbs
were in their base position and not absent. He conveyed questions with rising
intonation at the end of the sentences.
• In sentence formation of NAVS, he used the suppletive form of be in progressive
sentences at a chance level (54%). In 5 sentences, he replaced the progressive
with present simple and marking the verb with the 3SG present morpheme -s.
When he was asked to narrate a story from past, he almost always left the verbs
and the relevant tense untreated except with the case of lexically specified verbs
like got and went.
• In control constructions, he failed to produce the correct sentence 4/5 of the
time. He either failed to produce the control verb or the embedded verb; he
did not produce both within the same sentence. In raising constructions, he
started the sentences with the right argument but then (i) failed to continue
the sentence, (ii) provided a mono-clausal version (i.e. The boy is hardworking
instead of The boy seems to be hardworking), or (iii) had problems in the tense
morphology. In ECM constructions, he had 3/5 incorrect responses. These all
started with the correct matrix clause (e.g. I expect him to...) but then failed
to continue with the embedded verb. All these sentences had the object of the
matrix verb marked with accusative case.
• In the sentence production priming of reflexives, he had only 1 sentence with
no reflexive pronoun and that sentence was still grammatical: The girl hurts.
Other ungrammatical sentences were due to either having no tense or shortening
the sentence by omitting the adjuncts (e.g. The woman sees herself instead of
The woman sees herself in the mirror). He had no di"culty in marking the
pronouns with the right gender neither in reflexive nor in personal pronouns (in
ECM or elsewhere).
• His normal speech (speech not elicited through sentence formation or sentence
production priming) has the characteristics of pro-drop (e.g. Getting home to
wife), short sentences that look like small clauses (e.g. Shirt o! ), lack of tense
31
marking in present progressive, preference for present simple verbs with 3sg
agreement, and frequent use of frozen forms (e.g. I dont know).
• His performance in identifying grammatical sentences as “good” was remark-
able. He considered all the grammatical sentences in the grammaticality judge-
ments as good sentences. However, he identified some of the ungrammatical
sentences as good too. These were ungrammatical because (i) they lacked T-
to-C raising in question formation (e.g. Who the boy is kissing? ), (ii) lacked an
auxiliary verb or had the wrong one (e.g. The dog has chased by the cat), (iii)
or lacked a wh- item where it was needed (e.g. This is the cat is chasing the
dog).
4 Discussion
I have started this paper with the hypothesis that AA results from underspecification
of uninterpretable features, while interpretable features remain intact. I have fol-
lowed Avrutin (2006), who, in his Weak Syntax framework, assumes Narrow Syntax
of Chomsky (1995). Narrow Syntax is a meaning system with symbolic operations,
but the relevant lexical items, when put together at the C-I level, must be syntac-
tically well-formed and thus interpretable. Within the Phase Theory of Chomsky
(2001, 2005), the tenets of early minimalism (Chomsky, 1995) are developed into a
more elaborate system, where Edge features, a type of uninterpretable features, are
incorporated into the theory. We can categorize syntactic features roughly as follows:
Table 7: Syntactic features+ interpretable - interpretable
Phasal +/- realis, +/- negation, +/-question
Tense features, auxiliaries(perfective, passive, etc.), wh-features
Lexical Argument selection, phi fea-tures, irregulars
wh- words, negative polarityitems
I have hypothesized that interpretable features must remain intact because those
32
are available at Narrow Syntax. Uninterpretable features, on the other hand, are
needed at the C-I level for agreement in order to “establish a functional relation”
(Miyagawa, 2010, p. 20). I have claimed that it is crucially the tense features that
are underspecified in AA, while successful checking of Edge features varies according
to how much it is costly to value each. In this section, I discuss the results of each of
these hypothesis with respect to the data provided in the previous section. The first
hypothesis is repeated below as (13):
(13) the status of lexically specified features in aa
(a) Argument structure must be intact.
(b) Interpretable Agree features must remain intact.
(c) Irregular past and participle forms of verbs must remain intact.
Argument Structure Complexity Hypothesis of Thompson and Shapiro (Thomp-
son and Shapiro, 2005, 2007; Thompson, 2003) predicts that verbs with more complex
argument structures are more di"cult to produce than verbs with simpler argument
structures (i.e. unergatives) in agrammatic speech. They show that treatment of
complex verbs results in improvement in the production of simpler structures, while
the opposite approach (i.e. treating simple structures first) do not provide any gen-
eralizations across structures. The findings in this study provide further support for
such a hierarchy of structures5. The participant in this study could name all of the
unergatives but 2/3 of the transitives and ditransitives. There was no significant
di!erence between transitives and ditransitives though. However, what concerns us
here is the production of argument structure when the participant has access to a
given verb either in his memory or by constrained tests where the verbs are present
in written form. The participant in this study showed that, whenever he has access to
a verb, the argument structure of that verb comes for free. In the sentence production
test of NAVS, he was shown several pictures (as in Fig. 1), where the required verb
and its arguments were provided in written form. When asked to make sentences
5For a comparison of L1 acquisition patterns and agrammatic patterns of nouns and verbs, seede Bleser and Kauschke (2003), who show that there is a parallelism between these two. In otherwords, noun and verb categories that are acquired in later stages of acquisition are lost first inagrammatic individuals - showing a clear distinction between transitives and intransitives.
33
using these cues, his performance was close to perfect with only one missing subject
in an unergative, one missing internal argument in a transitive, and three missing
optional indirect objects among 50 items total. He even added direct objects to those
verbs such as sweep and dig even though these direct objects were not provided in
the picture cues. His tendency to omit indirect objects that are optional in verbs
like read and to add direct objects to verbs like sweep indicate that he has a lexicon
of verbs with predetermined argument structures. This finding suggests that (13a)
correctly predicts that argument structure remains intact in AA provided that there
is access to the related verbs.
The supplementary sentence priming task included items that targeted at eliciting
agreement. For example, the participant was shown a picture of a girl reading a book,
and presented the sentence She is reading. He was asked to make a similar sentence
for the second picture which had several girls reading a book. The participant had no
di"culty producing the sentence They are reading. He consistently gave the correct
sentence across items, which included both singular to plural patterns and vice versa.
He also had no di"culty using the correct subject and object pronouns with the right
gender in these items as well as other items that involved reflexive pronouns. These
results suggest that agreement is not an area of much di"culty as predicted by (13b).
The participant’s use of the third person singular marking -s in present simple
was inconsistent. It was above chance but not perfect across the items in the sup-
plementary tasks (68.4%). Could this be because this marker is syncretized for tense
and person/number? Or, alternatively, was this because -s is the only morphological
marker for person/number in English, and is therefore perceptually salient? I do not
have a conclusive answer to either of these questions, but as far as the data pre-
sented in this study goes, agreement is intact for the most part except for the special
case of the su"x -s. This is in line with the hypothesis in (13b) that interpretable
phi features must remain intact in AA. This provides further evidence for tense and
agreement dissociations in AA reported by a number of scholars, who show that tense
is typically impaired while the agreement is largely intact (de Bleser and Luzzatti,
1994; Nadeau and Gonzales-Rothi, 1992; Friedmann and Grodzinsky, 1997; Wenzla!
and Clahsen, 2003, among others). I return to this issue shortly.
The last item in (13) claims that irregular past and participle forms must also
34
remain intact in AA mainly because these are words that have their past and participle
features lexically specified. In the AphasiaBank portion of the story telling task, the
participant was asked to talk about how his life changed since he had a stroke, and
also to tell a story that happened in the past. His response had few verbs and these
were all irregular (got, and went used twice). Also, in sentence production section
of NAVS, he produces sent and got once for each. This could have been a matter
of chance; nevertheless, it will su"ce for the purposes of the present discussion as
it directs to the conclusion that irregular forms of the verbs may indeed be lexically
specified for tense, and are thus potentially easier for individuals with AA.
Next, I have claimed that checking Edge features is dependent on how costly it is
do so:
(14) the status of Edge features of aa
(a) Their status varies depending on how costly it is to value them.
With respect to the checking of Edge features, I have predicted that shorter move-
ments with fewer landing sites must be easier for the subject compared to longer
movements with more landing sites. Sentence production priming performances of
the subject in NAVS indicate the following hierarchy which is in line with my predic-
tions:
(15)
subject wh questions
actives
!
object wh questions
!yes/no questions
subject relatives
passives
!
object relatives
The subject had no di"culty in the production of subject wh- questions and
actives. Among all items, these are the ones that are the least costly. For example,
the wh- item in its base-position, assuming that it has already moved to [spec, TP],
moves to two Edges respectively: C and [spec, CP] to check wh-features. On the other
hand, if we compare this to let’s say an object wh- question, we will see that there
35
is a longer movement operation (i.e. the internal argument is more deeply embedded
in the VP and thus travels a longer distance).
The hierarchy seen in (15) provides further support for Grodzinsky (1995), Grillo
(2008) andBastiaanse et al. (2002) among others. Grodzinsky’s (1995) Trace Deletion
Hypothesis is two-fold: deletion of traces (a moved element thus lacks a theta role) and
the compensatory strategy that necessitates the theta role assignment on the first NP6. Grodzinsky (1995) argues that individuals with Broca’s aphasia are unable to fully
interpret moved elements and traces because they rely on a non-syntactic strategy
taking the first DP in the sentence as the agent. This gives the correct interpretation
in subject relatives, subject gaps and subject clefts. However, when presented with
object relatives, object gaps and object clefts, this strategy does not help the individ-
uals with AA, who also have di"culty to interpret moved elements and traces. They
are therefore at a chance level to correctly comprehend these sentences. The hierar-
chy in (15) partially supports the Trace Deletion Hypothesis as the easiest sentences
for the subject in this study are agent-first sentences (subject wh- questions while
the hardest sentence type is object relatives, which are patient-first. And note that,
the results in (15) are from a production study, and the Trace Deletion Hypothesis is
largely based on comprehension data7.
As an alternative, Grillo (2008, p. 49) claims that “agrammatic aphasics can-
not represent the full array of morphosyntactic features associated with syntactic
categories,” and “underspecication selectively targets scope-discourse features, i.e.
features at the Edge of the nominal, verbal and clausal domain.” Grillo takes into
account the Relativized Minimality (Rizzi, 1990) e!ects of moving elements to the
Edges of vP, TP and CP, where movement of an element across another element of
the same type is problematic for individuals with AA. For instance, he shows that,
in comprehension of reversible sentences (i.e. sentences where thematic roles can be
6Grodzinsky bases his arguments on Jackendo" (1972)’s Thematic Hierarchy Condition, a uni-versal hierarchy that constrains the order of theta roles: agent ¿ experiencer ¿ location, source, goal¿ theme.
7In a recent study, Caramazza et al. (2005) challange the core predictions of Trace DeletionHypothesis. They look at the comprehension performance (sentence-to-picture tasks) of 38 Italianspeaking Brocas aphasics with verified damage to Brocas area in active and passive reversible sen-tences, and found out that only 15% of them were at chance level on passive sentences and betterthan chance on active sentences.
36
inverted), the deficit is selective: object relatives and clefts are at chance level, while
subject relatives and subject clefts are signicantly above chance. It is possible that
there was a similar e!ect in the sentence production priming task of NAVS in this
study as the arguments were reversible (e.g. boy vs. girl, and squirrel vs. rabbit) 8.
Bastiaanse et al. (2002) make the claim that derived forms are harder for aphasics
to produce than base forms. They assume that every language has a basic word
order. Dutch, being a verb second language, has a variety of word orders depending
on whether the verb is (non-)finite as well as a matrix or an embedded verb. They
show that individuals with agrammatism are much better at producing finite verbs
in their base positions (in embedded clauses) than finite verbs that move to verb
second position (in matrix clauses). Along the lines Bastiaanse et al. (2002), Yarbay-
Duman et al. (2008) show that Turkish speaking individuals with AA have equal
di"culties in producing subject and object relatives. Turkish is a head-final language
with an SOV basic word order. The DP heading the whole relative clause thus has a
head-final landing site for the relativized subject or object. The derived word order
for the subject relative is therefore OVS while for the object relative it is SVO. The
participants in Yarbay-Duman et als (2008) study show equal di"culties for producing
these word orders (correct word orders 32.8% of the time for subject relatives and
35% of the time for object relatives). These studies show that long-distance movement
is more costly than other kinds of movement, which is also predicted by the model
proposed in this study. In their analysis of the data, Yarbay-Duman et al. (2008,
p. 156) indicate that when the participants moved the relativized subject or object
to the correct position, they always produced the non-finite verb inflection. This
is crucial for the purposes of this study: We have suggested that uninterpretable
features of C are underspecified while interpretable features like [+/-Realis] are not.
Turkish relative clauses are actually nominals headed by a specific morpheme for the
subject relativization and another for object relativization. These are nominalized
clauses with no tense morphology but what Kornfilt (1997, 2003) calls as factives and
non-factives. Factives are used to refer to events that happen in the past or present.
The non-factives are used to refer to events that will happen in the future. Yarbay-
Duman at als (2008) study is based on picture description (e.g. picture of a woman
8For a similar claim in first language acquisition, see Friedmann et al. (2008)
37
sewing a jacket); consequently, the target relative clauses are all factives. Factives
and non-factives are only morphologically marked in embedded clauses; matrix clauses
have just tense or aspect morphology associated with the these. The factives and non-
factives are directly related to what we have already called [+Realis], an interpretable
feature of C, which remain intact in AA. Yarbay-Duman et al (2008) point out that
there should be a relationship between overt movement of the relativized noun and
its relevant verbal morphology but do not discuss this observation any further. These
facts as well can easily be accounted for with the model presented here.
I have further claimed that it is crucially the underspecification of uninterpretable
phase features that give rise to AA:
(16) the status of uninterpretable features of CP
(a) They are underspecified.
Similar claims have been made before. For example, the Tree Pruning Hypothesis
(Friedmann, 2001, 2006; Friedmann and Grodzinsky, 1997), as the name suggests,
argues that if the TP is a problem area for individuals with AA, then anything above
it must be inaccessible. This model can explain, for instance, why yes/no questions
are hard for English speakers but not Hebrew or Arabic speakers9. The former re-
quires the movement of the auxiliary to C while the latter languages produce yes/no
questions with intonation. An inaccessible CP explains this di!erentiation. However,
we observe in data in various studies that wh- movement - despite variation in the
type of wh- element we are talking about - is available even to individuals with severe
tense problems. The model presented here makes a clear distinction between Edge
features and uninterpretable features of C, and therefore can predict cases where indi-
viduals with AA fail to have tense features but perform much better in wh- question
formation. Wenzla! and Clahsen (2003), on the other hand, show that Tree-Pruning
Hypothesis cannot explain the data they presented from German-speaking individ-
uals with AA. Since agreement is above TP in German, Tree-Pruning Hypothesis
falsely predicts that agreement must be inaccessible in these individuals. Wenzla!
and Clahsen (2003), therefore, suggest that tense features of TP are underspecified
9For a similar study of Dutch question formation see Ruigendijk et al. (2004)
38
but [+/-Realis] features are not. Their analysis for German agrammatism is sup-
ported by facts that [+/-Past] features are secondary to [+/-Realis] typologically in
languages of the world as well as in language acquisition. I di!er from Wenzla! and
Clahsen (2004) only in the last point: the fact that [+/- Past] features are secondary
to [+/- Realis] is due to the fact that the former is an uninterpretable phase feature
related to the latter, which is a phase feature. In other words, the former is univer-
sal to languages, which may have di!erent manifestations of uninterpretable tense
features. The observation that these uninterpretable features are secondary typolog-
ically in world languages as well as in language acquisition is a natural outcome of
this di!erence in feature status, not the cause.
Because of its status as an uninterpretable feature, TP has drawn a great deal of
attention in agrammatism literature. The tendency has been to compare the status
of tense in AA as opposed to agreement (e.g. de Bleser and Luzzatti (1994) for
Italian, Nadeau and Gonzales-Rothi (1992) for English, Friedmann and Grodzinsky
(1997) for Hebrew, Wenzla! and Clahsen (2004) for German). The tense-agreement
dissociations are not surprising given the assumptions in this study. Along the lines of
Chomsky (1995), we have said that, TP is visible at LF, while agreement is required
for syntax-internal reasons to establish a functional relation, to keep a record of
movements in a derivation (Miyagawa, 2010). These two are thus di!erent things:
one is a meaningful content for the interpretation of a sentence, while the other is a
tool, a machine syntax needs to successfully operate.
Finally, I have made a claim about the interpretable phase features:
(17) the status of interpretable features of CP
(a) They remain intact.
I have said that these features are at least made of [+/-Neg] and [+/-Realis]. I
have also added that [+/-Question] is potentially a phase feature too. We have seen
evidence that the subject does not have di"culty marking sentences with negative
even if these lack tense auxiliaries. We have not seen any concrete evidence for Realis
because it is not overtly marked in English. As for questions, we have seen that he
subject can ask questions, but this needs to be further discussed.
39
The subject does not have di"culty with wh- words per se. In other words, he
does not have any di"culty accessing these words, or using them appropriately even
if the sentences may lack tense auxiliaries. In the sentence production priming task,
he produced all of the subject wh- questions correctly, and made only one mistake
in the object wh- question, and this was due to a di"culty in remembering the
verb, not because of the wh- word. In yes/no questions, however, we see that the
auxiliaries are present, but do not move to the CP (e.g. The cat is chasing the dog?
as opposed to Is the cat chasing the dog? ). This is noteworthy because the subject
has a great tendency to leave out auxiliaries. Why would they appear across all items
that have a yes/no question in sentence production priming? My answer would be as
follows: since question is a phase feature, and it is interpretable, it is a Narrow Syntax
requirement. Moving the auxiliary to the CP, on the other hand, is a requirement
of the C-I interface to delete the uninterpretable counterpart of this feature. So,
the individual has a choice between moving this element to a higher position, or
resorting to another interface, context, as proposed by Avrutin (2006). The subject
in this study prefers to turn to discourse because, for him, it is much less costly, and
therefore, produces an utterance like the following with the right intonation, which is
an acceptable English sentence both as a regular and an echo question:
(18) The cat is chasing the dog?
Interpretable features of other phases (DP and vP) are potentially intact too. Even
though I have not looked at these features closely, there is at least one instance in the
data that is noteworthy. In the production of unaccusatives, the subject described
the picture of a melting ice-cube as Ice is going, do the water, ice is... cold... ice
getting warmer. Even though the verb melt is not accessible to the subject at the
time of the elicitation, it appears that the subject ice is an undergoer of a change-of-
state marked by go, do and get. Can these be all heads of a vP, which dominate a
result state water or warmer? If the hypothesis in (17) is correct, then interpretable
features of vP such as “do,” “become” and “cause” must also be intact in AA. I do
not yet have a conclusive answer to this question, and thus leave this issue to future
research.
Avrutin’s (2006) Weak Syntax model helps us see particular aspects of naturally
40
occurring agrammatic speech as well. We have seen in story telling task that the
subject’s speech tends to drop subjects and consist of small clauses. I repeat his
description of the picture-strip story here as (19):
(19) Getting home to wife ... and wife um ... um ... looks ... marks and cigarette
or cigar and ... I dont know ... shirt o! or jacket o! and trousers o! and shirt
and that is enough[pounds on the table] mad, mad, mad her.
The ungrammatical sentences in (19) are all acceptable utterances in English.
For example, dropping the subject as in “getting home to wife” as well as small
clauses as “shirt o!” are all acceptable. The subject shows similar tendencies in
acceptability in comprehension too. In the grammaticality judgements, he identified
all the grammatical sentences as “good” English sentences, while he identified only
some of the ungrammatical ones as “bad”. And the ungrammatical ones he judged
as “good” English sentences are all acceptable utterances in English. I repeat an
example for each of the categories in Table 8.
Table 8: Examples for grammaticality judgementsGrammatical Ungrammatical Ungrammatical
Acceptable Acceptable Unacceptable
The girl is pulling the
boy.
The dog chasing the cat. The man is kissing to the
woman.
The data in Table 8 along with other types of data we have discussed so far all
indicate that Avrutin’s (2006) Weak Syntax model is making the correct predictions
about AA. The subject in this study seems to have a grammar that does not generate
unacceptable sentences. What appears is that, since the uninterpretable features are
underspecified, the generative machine becomes a “weaker” one. What was once a
powerful machine that would allow only a limited range of sentence types has now
become weaker in the sense that the range of grammaticality has expanded to cover
a much wider array of sentences.
41
5 Conclusion
In this paper, I have looked at various aspects of the grammar of an individual with
AA, and showed that they can largely be explained by the framework suggested by
Avrutin (2006), namely the Weak Syntax. I have shown that the Narrow Syntax
remained intact in the subject’s grammar while the syntactic operations at the C-I
level such as tense marking and moving elements to the Edges are impaired for the
most part. I have maintained that the Minimalist framework of Chomsky (1995,
2001, 2005) provides us with the right tools we have been looking for. In particular,
distinguishing between interpretable and uninterpretable features help us advance to-
ward a more unified theory of AA. We have seen that interpretable phase features
such as [+/-Question] and [+/-negation] as well as lexically specified features such as
argument structure and phi features remain intact. We have also seen that uninter-
pretable phase features such as tense as well as Edge features are, to a large extent,
main sources of agrammatism. I have aimed to provide an account of agrammatism
that can explain a wide range of observations in agrammatic speech, and showed that
a unified theory of agrammatism is in fact not a too far-fetched attempt. I am aware
that the findings in this study are very preliminary, and more research is needed for
more conclusive claims. In particular, aphasic speakers of languages that have richer
morphologies should be looked at especially those that overtly mark [+/-realis] and
other interpretable phase features. We can at least say that the findings in this study
point to a direction where agrammatism is a case of a less powerful grammar machine,
which can generate a wider range of sentences with increased grammaticality.
References
Avrutin, S. (2006). Weak syntax, in Y. Grodzinsky and K. Amunts (eds), Broca’s
region, Oxford University Press, pp. 49–62.
Baauw, S., de Roo, E. and Avrutin, S. (2002). Determiner omission in language
acquisition and language impairment: Syntactic and discourse factors, Proceedings
of the Boston University Conference on Language Development, Cascadilla Press,
Boston, pp. 23–35.
42
Bastiaanse, R., Hugen, J., Kos, M. and van Zonneveld, R. (2002). Lexical, morpho-
logical and syntactic aspects of verb production in agrammatic aphasics, Brain and
Language 80: 142–159.
Bastiaanse, R. and van Zonneveld, R. (1998). On the relation between verb inflection
and verb position in dutch agrammatic aphasics, Brain and Language 64: 165–181.
Bastiaanse, R. and van Zonneveld, R. (2005). Sentence production with verbs of
alternating transitivity in agrammatic broca’s aphasia, Journal of Neurolinguistics
18: 57–66.
Caramazza, A., Capasso, R., Capitani, E. and Miceli, G. (2005). Patterns of compre-
hension performance in agrammatic broca’s aphasia: a test of the trace deletion
hypothesis, Brain and Language 94: 43–53.
Chomsky, N. (1995). The Minimalist Program, The MIT Press, Cambridge, MA.
Chomsky, N. (2001). Derivation by phase, in M. Kenstowicz (ed.), Ken Hale: A life
in language, The MIT Press, Cambridge, MA, pp. 1–52.
Chomsky, N. (2005). On phases. Unpublished manuscript.
de Bleser, R. and Kauschke, C. (2003). Acquisition and loss of nouns and verbs:
parallel or divergent patterns?, Journal of Neurolinguistics 74: 1–25.
de Bleser, R. and Luzzatti, C. (1994). Morphological processing in italian agrammatic
speakers: syntactic implementation of inflectional morphology, Brain and Language
46: 21–40.
Friedmann, N. (2001). Agrammatism and the psychological reality of the syntactic
tree, Journal of Psycholongusitic Research 30: 71–90.
Friedmann, N. (2002). Question production in agrammatism: the tree pruning hy-
pothesis, Brain and Language 80: 160–187.
Friedmann, N. (2006). Speech production in broca’s aa: syntactic tree pruning,
in Y. Grodzinsky and K. Amunts (eds), Broca’s region, Oxford University Press,
Oxford.
43
Friedmann, N., Belletti, A. and Rizzi, L. (2008). Relativized relatives: types of
intervention in the acquisition of a-bar dependencies, Lingua 119: 67–88.
Friedmann, N. and Grodzinsky, Y. (1997). Tense and agreement in agrammatic
production: pruning the syntactic tree, Brain and Language 56: 397–435.
Grillo, A. (2008). Generalized minimality, syntactic underspecification in Broca’s
aphasia, PhD thesis, Utrecht University.
Grodzinsky, Y. (1995). Trace deletion, theta roles and cognitive strategies, Brain and
Language 51(3): 469–497.
Huang, C. T. J. (1982). Move wh in a language without wh movement, Linguistic
Review 1: 369–416.
Jackendo!, R. (1972). Semantic interpretation in generative grammar, The MIT
Press, Cambridge, MA.
Kornfilt, J. (1997). Turkish, Routledge, London.
Kornfilt, J. (2003). Subject case in turkish nominalized clauses., in U. Junghanns and
L. Szucsich (eds), Syntactic structures and morphological information, Mouton de
Gruyter, Berlin, pp. 129–200.
MacWhinney, B. (2007). The talkbank project, in J. C. Beal, K. P. Corrigan and
H. L. Moisi (eds), Creating and digitizing language corpora: synchronic databases,
Vol. 1, Palgrave-Macmillan, Houndsmills.
Miyagawa, S. (2010). Why agree? Why move? Unifying agreement-based and
discourse-configurational languages, The MIT Press, Cambridge, MA.
Nadeau, S. E. and Gonzales-Rothi, L. J. (1992). Morphologic agrammatism following
a right hemisphere stroke in a dextral patient, Brain and Language 43: 642–667.
Reuland, E. (2001). Primitives of binding, Linguistic Inquiry 32: 439–492.
Richards, E. (2007). Feature inheritence, Linguistic Inquiry .
44
Rizzi, L. (1990). Relativized minimality, number 16 in Linguistic Inquiry Monograph,
MIT Press.
Ruigendijk, E., Kouwenberg, M. and Friedmann, N. (2004). Question production in
dutch agrammatism, Brain and Language 91: 116–117.
Thompson, C. K. (2003). Unaccusative verb production in agrammatic aphasia: the
argument of structure complexity hypothesis, Journal of Neurolinguistics 16: 151–
167.
Thompson, C. K. (n.d.). The northwest assessment of verbs and sentences.
Manuscript in preparation.
Thompson, C. K. and Shapiro, L. P. (2005). Treating agrammatic aphasia within a
linguistic framework: treatment of underlying forms, Aphasiology 19: 1021–1036.
Thompson, C. K. and Shapiro, L. P. (2007). Complexity in treatment of syntactic
deficits, American Journal of Speach-Language Pathology 16: 30–42.
Wenzla!, M. and Clahsen, H. (2003). Tense and agreement in german agrammatism,
Brain and Language 89: 57–68.
Wenzla!, M. and Clahsen, H. (2004). Finiteness and verb-second in german agram-
matism, Brain and Language 92: 33–44.
Yarbay-Duman, T., Aygen, G. and Bastiaanse, R. (2008). The production of turkish
relative clauses in agrammatism: verb inflection and constituent order, Brain and
Language 105: 149–160.
45
Appendix
1. The participant’s production in WAB picture description task between 2 years post-onset
and 12 years post-onset
5/17/1994 - 2 years post-onset(Tell me what’s going on in the pictures and try to use sentences) I don’t know [whispering] I ... {ded} [pointing at his head and then his lips with his left index finger, shaking head] (Why don’t you start with what you can get?) tree ... I can {ded} [pointing at his head and lips, shaking head] (So, there’s these two people...) Yes (What are they doing?) [gesturing eating] (And this guy here?) [onomatopoeia and gesture for flying] (So, he’s flying a...) kite (And out here, what’s going on?) two ... sailors ... one [gesturing to wave] (Alright, and here?) one [gesturing to fish] (He’s digging or building with the...) [nodding] (He’s playing in...) sand
4/1/1998 - 6 years post-onset(Tell me what you see. Try to talk in sentences) I see um ... house and car and ... boat and four people ... and dog and car (Can you tell me any more?) Um ... trees ... and . flag ... ... and boy playing out there ... in the middle . of ... the lake (Anymore?). No. (Are you sure?) I don’t know . two people ... sailboat
4/2/1999 - 7 years post-onset(Tell me what you see and try to tell me in sentences) Tree ... house . car . boy in (Can you say it in sentences?) Oh ... (You are doing good. You’ve got those down.) [smiles] ... ... boy and girl ... eating and drinking ... boy and kite ... ... flag ... flying ... and girls on ... sailboat ... [gesturing to go] I know it ... I guess . this . is . it (Can you think of anything else?) I can’t
2/11/2000 - 8 years post-onset(Can you describe this picture for me? Tell me what you see. Try to talk in sentences if you can.) A lot of things . people ... places ... trees ... a house ... a car ... a flag pole ... um ... in this picture ... a dog and his owner ... um ... a boy and his kite ... and her ... drinking and smoking [laughs] sailboat ... two sailors and a sailboat ... [gesturing to fish] um ... he and his ... um I can’t ... um ... I can’t do it
3/20/2001 - 9 years post-onset(Tell me what you see. Try to talk in sentences.) Oh ... um ... sentences? (Just tell me what you can.) Jack and Jill ... um ... this is cool . Jack and Jill ... the ... um ... I can’t here [pointing at his head and mouth] (Tell us what you can.) tree . house . car . trees ... um ... kite ... somebody ... sun ... kite ... dog ... people ... {sand?}
4/25/2003 - 11 years post-onset(Tell me what you see in the picture. Please try to talk in sentences.) Um . here ... Mary and George ... um . sit in yard . um ... TV . um . radio ... I don’t ... I don’t . um ... (Trying to use sentences when describing it?) Yeah (Take a look over here. Just give it a shot.) Man um ... and ... kite ... um I don’t know ... two sailors sailing
46
2/11/2004 - 12 years post-onset(Tell me what you see and try to talk in sentences.) A tree in front of the house and car ... people ... um boy and girl ... um man and woman drinking a {toast} ... ... boy and kite . I don’t know ... boy the kite and the boy . I don’t ... dog is getting ... [sighs] ... sailboat and sailors in lake ... pole and ... I don’t know . alright (What’s in the pole?) flag . um little boy um ... and {lion} ... alright (You’re doing a good job.) I don’t ... I don’t ... (You don’t think so?) No
12/7/2005 -12 years post-onset(Tell me what you see and do your best to use sentences) Jeff and Jill reading and drinking . at . the . house ... on the lake ... good enough? (Tell me a little bit more) No, good enough for [left thumb up] (That was a very nice sentence, yes) Um ... flag waving ... and kite ... ... ... and dog ... and five people . two of them sailboat ... ... one of them dog ... and um fishing ... [gesturing OK?]
2. Full list of sentences used in grammaticality judgement
SENTENCE RESPONSE
1. The girl is pulling the boy. GOOD !! BAD !
2. * The dog chasing the cat. GOOD !! BAD !
3. * The man is kissing to the man. GOOD !! BAD !
4. * There"s tickling a boy. GOOD !! BAD !
5. The squirrel is kicking the rabbit. GOOD !! BAD !
6. Is the dog chasing the cat? GOOD !! BAD !
7. * Are the boy pulling the girl? GOOD !! BAD !
8. *Mary wonders who is the boy kissing. GOOD !! BAD !
9. Mary wonders who is chasing the dog. GOOD !! BAD !
10. Are the boys chasing the girls? GOOD !! BAD !
11. Who is kissing the man? GOOD !! BAD !
12. Who does Mary think is kissing the woman? GOOD !! BAD !
13. * Is who chasing the boy? GOOD !! BAD !
14. * Which the dog is chasing the cat? GOOD !! BAD !
15. Which woman is kissing the man? GOOD !! BAD !
47
SENTENCE RESPONSE
16. * This is the cat is chasing the dog. GOOD !! BAD !
17. * This is the squirrel is kicking the rabbit. GOOD !! BAD !
18. * This is the boy is who shoving the girl. GOOD !! BAD !
19. *These are the girls who are chasing the boys. GOOD !! BAD !
20. * These are the girls are shoving the boys. GOOD !! BAD !
21. Who is the rabbit kicking? GOOD !! BAD !
22. * Who the boy is kissing? GOOD !! BAD !
23. Which cat is the dog chasing? GOOD !! BAD !
24. Which girl the boy is shoving? GOOD !! BAD !
25. *Which the dog is chasing the cat? GOOD !! BAD !
26. * The dog is barked by the dog. GOOD !! BAD !
27. * The dog has chased by the cat. GOOD !! BAD !
28. The boy is pulled by the girl. GOOD !! BAD !
29. *There"s chasing a dog the cat. GOOD !! BAD !
30. *There"s kissed a woman by the man. GOOD !! BAD !
31. This is the cat which the dog is chasing. GOOD !! BAD !
32. * This is the man the woman is kissing who. GOOD !! BAD !
33. *This is the boy which the girl is chasing. GOOD !! BAD !
34. *This is the rabbit that the squirrel is kicking which. GOOD !! BAD !
35. This is the girl the cat is chasing. GOOD !! BAD !
36. The man is melting an ice cube. GOOD !! BAD !
37. * The airplane is arrived. GOOD !! BAD !
38. *There"s arriving a man. GOOD !! BAD !
39. *The woman is arrived home. GOOD !! BAD !
40. * There arrived a man at school. GOOD !! BAD !
48
SENTENCE RESPONSE
41. She is dancing. GOOD !! BAD !
42. *The boy and the girl is reading. GOOD !! BAD !
43. * The man and the woman is walking. GOOD !! BAD !
44. *The boy are flying. GOOD !! BAD !
45. The woman is shopping. GOOD !! BAD !
46. * I saw me in the mirror. GOOD !! BAD !
47. The boy hurt himself. GOOD !! BAD !
48. *The man admires myself. GOOD !! BAD !
49. * You injured myself in the accident. GOOD !! BAD !
50. The boys enjoy themselves at the party. GOOD !! BAD !
51. The man demands to go on vacation. GOOD !! BAD !
52. * The woman demanded the man to buy a house. GOOD !! BAD !
53. *The boy hopes himself to go to college. GOOD !! BAD !
54. The girl wishes to fly. GOOD !! BAD !
55. The boy hopes to be happy. GOOD !! BAD !
56. The girl is likely to crash into a tree. GOOD !! BAD !
57. The boy seems to be happy. GOOD !! BAD !
58. *It appears the boy to be sad. GOOD !! BAD !
59. *The girl seems it to be happy. GOOD !! BAD !
60. *There seems a woman to fall down. GOOD !! BAD !
61. I expect her to be successful. GOOD !! BAD !
62. * I believe he to be a nice person. GOOD !! BAD !
63. I expect him to crash into a tree. GOOD !! BAD !
64. * I expect she to be sad. GOOD !! BAD !
65. *I expect her fall down. GOOD !! BAD !
49
3. Pictures used for unaccusative verb naming
arrive melt burn break fly
4. Example picture pairs used in supplementary sentence production priming
AGREEMENT:
Stimulus: She is reading. Target: They are reading.
UNACCUSATIVES:
Stimulus: The plate broke. Target: The bottle broke.
50
BINDING:
Stimulus: The man sees himself in the mirror. Target: The woman sees
herself in the mirror.
CONTROL:
Stimulus: The man wishes to get married. Target: The woman wishes to
get married.
RAISING:
Stimulus: The girl is likely to crash into a tree. Target: The boy is likely
to crash into a tree.
ECM:
Stimulus: I expect her to fall down. Target: I expect him to fall down.
51