Transcript PSYCHOLOGY 506b Cognitive Neuroscience Core Course (LouAnn

Netherlands Graduate School of Linguistics
LOT Summer School 2006
Issues in the biology and evolution of
language
Massimo Piattelli-Palmarini
University of Arizona
Session 3 (June 14)
Loss of speech and two
interesting hypotheses
Some ancient history



Along many centuries, physicians and common
people had observed cases in which a person
had suffered some loss of speech.
The most ancient (indirect) testimony is to be
found in the Bible (in the Ecclesiastes)
“And if I am found to have forgotten thy name,
oh Sacred Jerusalem, may my right arm remain
limping along my body and my tongue remain
glued to the palate”.
LOT Summer 2006
Loss of speech
2
This association cannot have been
invented




In fact, severe damage to the left hemisphere,
indeed, typically produces the paralysis of the
right arm, and loss of speech.
This is called, in modern terminology, “global
aphasia”
It is the most severe form of loss of speech.
We will see in a moment less severe, more
specific syndromes.
LOT Summer 2006
Loss of speech
3
Further evidence






The cognitive neuropsychologist John
Marshall of Oxford has collected a large
sample of historical documents referring to
specific “modular” cognitive deficits.
Personal letters (incidentally) describing
specific symptoms
Medical records
Military chronicles
Novels, poems, plays etc.
From the ancient Roman times to the 19th.
Century
LOT Summer 2006
Loss of speech
4
An interesting consideration






Ample evidence of subtle and specific
cognitive deficits has been commonly
available for centuries
YET
The very idea of the modularity of mind is
barely 25 years old
And it still meets strong resistance.
No one was “ready” to see it.
Speech impairments have been a very
important domain
LOT Summer 2006
Loss of speech
5
A revealing synthesis
(Russell A. Poldrack, UCLA, 2006)
 749 published papers, 3222 comparisons (from
the BrainMap database) TICS Vol10 (2)
February 2006, pp.61-62
Language study








Activated
Not activated
166
703
Not language study
199
2154
Jeffreys & Poldrack Bayesian factor f:
posterior odds/prior odds
1 < f < 3 weak evidence
3 < f <10 moderate evidence
10 < f
strong evidence
Here, for the reverse inference, it is 2.3
LOT Summer 2006
Loss of speech
6
The standard “textbook”
picture of aphasia
Before, approximately, 1970
For a concise review, see Edgar B. Zurif Brain regions of relevance to
syntactic processing. (Chapter 13) in Volume 2 of An Invitation to
Cognitive Science Second Edition (D. N. Osherson, Editor) MIT Press
1995
Paul Broca (1824 -1880)

In a series of papers published between
1861 and 1866 employing the clinicopathological correlation technique to analyze
a loss of speech (aphémie), Broca
persuaded a majority of his colleagues that
there was a relatively circumscribed center,
located in the posterior and inferior
convolutions of the left frontal lobe, that was
responsible for speech (langage articulé).
LOT Summer 2006
Loss of speech
8
In a memorable meeting in 1862
he demonstrated the brain lesion
of his first patient (Tan) who had
suffered from aphémie (renamed
aphasia later by Armand
Trousseau (1801-1867)). Broca
concluded that the integrity of the
left frontal convolution was
responsible and necessary for
articular speech. David Ferrin
(1843-1928) is responsible for
naming this region "Broca’s
convolution- the motor speech
area."
LOT Summer 2006
Loss of speech
9
Brain of patient with motor aphasia (the famous “Tan”) studied by
Broca is preserved uncut. The damaged area receives its blood from the
left middle cerebral artery.
LOT Summer 2006
Loss of speech
10
Notice that it extends beyond the (now canonical) Broca’s area
showing lesions of the underlying insula and the white matter
(as it is often the case in such patients) (See Nina Dronker’s comment
in BBS 2000)
LOT Summer 2006
Loss of speech
11
Broca’s famous patient (“Tan” -1861), after a stroke, had
completely lost the ability to speak. The patient could apparently
understand language, but the only syllable he could produce was
"tan", over and over again. Broca referred to this patient as “Tan”.
After Tan's death, Broca performed an autopsy and determined the
site of the stroke.
LOT Summer 2006
Loss of speech
12
The “foot” of the 3rd gyral convolution of the left frontal lobe
LOT Summer 2006
Loss of speech
13
Early polemics



Pierre Flourens (1794-1867), Broca’s
distiguished colleague (whom we “met”
yesterday as a discoverer of the cerebellar
functions) defied Broca’s conclusions
Damage to other regions (allegedly) had similar
effects on speech
(Strangely) he maintained that the superior
functions cannot be “localized” in one precise
cerebral region.
LOT Summer 2006
Loss of speech
14
Classical description of Broca's aphasia

A person that suffers a lesion to that area
understands language, communicates
nonverbally, and writes, if not also paralyzed,
but cannot articulate or speak fluently. (A
sudden drop in fluency may, in fact, signal an
impending stroke).
LOT Summer 2006
Loss of speech
15
Broca's aphasia: The clinical textbook description




This is a form of aphasia in which speech
output is severely reduced and is limited
mainly to short utterances, of less than four
words.
Vocabulary access is limited in persons with
Broca's aphasia, and their formation of
sounds is often laborious and clumsy.
The person may understand speech relatively
well and be able to read, but be limited in
writing.
Broca's aphasia is often referred to as a 'nonfluent aphasia' because of the halting and
effortful quality of speech.
LOT Summer 2006
Loss of speech
16
LOT Summer 2006
Loss of speech
17
Brodmann areas
Parietal
Frontal
Occipital
Temporal
LOT Summer 2006
Loss of speech
18
Brodmann areas 44 & 45: Broca’s area
Parietal
Frontal
Occipital
Temporal
Spoken language production (motor component)
Brodmann area 45 is called the “triangular portion”.
LOT Summer 2006
Loss of speech
19
Brodmann areas 22 (Wernicke’s area) and 41 & 42: Herschl’s gyri
Parietal
Frontal
Occipital
Temporal
Spoken language comprehension
LOT Summer 2006
Loss of speech
20
Brodmann areas 39 and 37
Parietal
Frontal
Visual assoc
areas
Occipital
Temporal
Sylvian fissure
or “lateral sulcus”
Written language perception
All in all, language is said to reside in “peri-Sylvian areas”
LOT Summer 2006
Loss of speech
21
Wernicke’s Aphasia


LOT Summer 2006
It is characterized not by apparent
loss of fluency but by absence of
meaning in what the person says.
The words don't add up to
informative sentences; or the
person may have problems naming
familiar objects, and call a cup an
ashtray, for instance, or be unable
to name a loved one.
Loss of speech
22
Wernicke’s Aphasics




These patients could not understand language,
spoken or written, but could apparently produce
copious flowing speech.
Their speech, however, made absolutely no
sense.
Subjects and verbs would be strung together in
a seemingly grammatical order, like fragments of
real sentences, but the sentences seemed to
bear little relation to what the patient was trying
to express.
Presumably the patients could not understand
what came out of their own mouths.
LOT Summer 2006
Loss of speech
23
Conduction aphasia



Broca's and Wernicke's speech areas
intercommunicate via a thick arching bundle
(called the arcuate fasciculus or bundle).
When damage to this pathway disconnects
the two speech areas, language fluency and
comprehension are not affected;
However, the sufferer cannot repeat newly
presented phrases.
LOT Summer 2006
Loss of speech
24
Perisylvian language networks of the
human brain (Catani, et al., 2005)


Next slides adapted from those of Katrina
Nichols (University of Arizona)
Damage to arcuate fasciculus (white matter
tract) connecting Broca’s and Wernicke’s areas
results in conduction aphasia (intact
comprehension and production but inability to
repeat what heard)
Catani, M., Jones, D. K., and Ffytche, D. H. (2005).
Perisylvian language networks of the human brain
Annals of Neurology, 57, 8 - 16.
LOT Summer 2006
Loss of speech
25
Perisylvian language networks of the
human brain (Catani, et al., 2005)



Conduction aphasics (temporoparietal lesions) are a
heterogeneous group in terms of impairment (some
more Broca-like, some Wernicke-like) suggesting
arcuate fasciculus may not be the only underlying
structure
Authors were interested searching for other possible
perisylvian (temporoparietal areas) language networks
to explain data
Used in vivo diffusion tensor magnetic resonance
imaging tractography (to extrapolate white matter
trajectories)
LOT Summer 2006
Loss of speech
26
Perisylvian language networks of the
human brain (Catani, et al., 2005)

Newly discovered but evolutionarily older
structure in addition to the classical arcuate
fasciculus



parallel and lateral to classical arcuate fasciculus
Connects “Geschwind’s territory” (inferior parietal cortex) to
classical language areas (Broca’s and Wernicke’s areas)
Rudimentary form exists in the brain of other primates
(macaque)
LOT Summer 2006
Loss of speech
27
Perisylvian language networks of the
human brain (Catani, et al., 2005)
Arcuate fasciculus (long segment; medial)
and “new” network (anterior segment and posterior segment)
LOT Summer 2006
Loss of speech
28
Perisylvian language networks of the
human brain (Catani, et al., 2005)

Corroborates neuropsychological evidence for
different subtypes of conduction aphasia
conduction aphasia – long segment
lesion; failure in automatic repetition
 Transcortical aphasia – anterior segment lesion;
failure to vocalize semantic content
 Sensory aphasia – posterior segment lesion;
failure of auditory semantic comprehension
 Classical
LOT Summer 2006
Loss of speech
29
Perisylvian language networks of the
human brain (Catani, et al., 2005)

“The fact that these pathways [ the ones including the
anterior and posterior segments] appear to exist – in
more rudimentary forms – in the brains of monkeys may
also have bearing on the search for the evolutionary
origins of language. ‘These data suggest that language
evolved, in part, from changes in pre-existing networks,
not through the appearance of new brain structures,’
said Catani.”
 American

Neurological Association press
release, 2006
This is in line with non-drastic, piecemeal
changes occurring in evolution, not complicated
overhauls of structures
LOT Summer 2006
Loss of speech
30
Global aphasia




This is the most severe form of aphasia, and is
applied to patients who can produce few
recognizable words and understand little or no
spoken language.
Global aphasics can neither read nor write.
Global aphasia may often be seen immediately
after the patient has suffered a stroke and it
may rapidly improve if the damage has not
been too extensive.
However, with greater brain damage, severe
and lasting disability may result.
LOT Summer 2006
Loss of speech
31
Mixed non-fluent aphasia


This term is applied to patients who have sparse
and effortful speech, resembling severe Broca's
aphasia.
However, unlike persons with Broca's aphasia,
they remain limited in their comprehension of
speech and do not read or write beyond an
elementary level.
LOT Summer 2006
Loss of speech
32
Anomic aphasia





This term is applied to persons who are left with a
persistent inability to supply the words for the very
things they want to talk about - particularly the
significant nouns and verbs.
As a result, their speech, while fluent in
grammatical form and output, is full of vague
circumlocutions and expressions of frustration.
They understand speech well, and in most cases,
read adequately.
Difficulty finding words is as evident in writing as
in speech.
Mostly associated with damage to the angular
gyrus
LOT Summer 2006
Loss of speech
33
Fisher, S. E., & Marcus, G. F.
(2006). The eloquent ape:
genes, brains and the evolution
of language.
Nature Reviews Genetics,
Vol. 7 (January), pp. 9-20.
LOT Summer 2006
Loss of speech
34
The study of aphasia comes of age






Edgar B. Zurif and Alfonso Caramazza, since
the mid-Seventies, characterize agrammatic
aphasia
Gabriele Miceli and Alfonso Caramazza draw
close comparisons between agrammatic
aphasics in English and in Italian
David Caplan (MGH and Harvard Med. Sch.)
and collaborators explore in detail “disorders of
syntactic comprehension”
The traditional classification collapses:
Wernicke aphasics also show syntactic
impairments, while Broca aphasics also show
semantic impairments.
The “single case” approach is proposed, raising
serious objections
Yoseph Grodzinski (ever since 1984) proposes
the TDH (trace deletion hypothesis) (see infra)
LOT Summer 2006
Loss of speech
35
The single-case method





Specific impairments ascertained in one patient
are explained by certain functional disconnections
Normal performance in the same patient on other
cognitive tasks is explained by other connections
being intact
Flow charts are then constructed
The legitimacy (or the non-legitimacy) of
postulating processing units and connections is
put to test
The most cogent test is one of the “necessity” of a
certain node, or connection, for a given cognitive
task
LOT Summer 2006
Loss of speech
36
Typical dissociations







Some patients show an impairment with nouns
(they use thing, stuff, not, say, cake, chair, etc.),
but not with verbs
Other patients show an impairment with verbs
(they omit them altogether, or pause, or just use
do, get), but not with nouns
Word-endings can be affected
Or word-beginnings
Only in writing, or only in speech (modalityspecific deficits)
And many more kinds of dissociations
Different brain areas are correspondingly
affected (verbs  frontal and fronto-parietal,
nouns  temporal and fronto-temporal)
LOT Summer 2006
Loss of speech
37
Accurate and complete oral description, but the verb is omitted in
writing. Inverse symmetric deficits have also been reported.
From Rapp & Caramazza, 1998
A modality-specific deficit like this one cannot be “semantic”
LOT Summer 2006
Loss of speech
38
Agrammatism (left frontal lesions)




Morphosyntactic processing is impaired in production
(inflection and agreement) regardless of the lexical
class of the items (verbs, nouns, adjectives), functional
elements are omitted in zero-morphology languages,
inter-substituted in non-zero-morphology languages.
Goodglass and Berko, 1960; Goodglass, 1968; Berndt
and Caramazza, 1980.
The differentiation between the morphological, lexical,
syntactic and semantic components, and the
corresponding brain lesions, is often debated (see
Caramazza and Shapiro, in Jenkins’s volume)
In later years, (refined)-EEG, PET, fMRI, MEG and
rTMS (repeated Transcranial Magnetic Stimulation) add
new data. The correspondence with brain lesions is not
always neat, but approximately satisfactory.
LOT Summer 2006
Loss of speech
39
Lesions versus imaging




Participation of a brain region in a task is not
the same as the localization of that task in that
region.
In a nutshell: Lesions show whether a given
brain region is “necessary” for that task
Imaging shows that the region is (strictly, and
selectively, in some cases) recruited for that
task.
rTMS (is rather a suppression than a
stimulation) is closer to lesions, because it
generates a “temporary virtual lesion”
LOT Summer 2006
Loss of speech
40
Zooming onto Broca’s area lesions









Caramazza and Zurif (1976) challenge the
(then) prevailing view that comprehension is
unaffected in Broca’s aphasics
Semantically non-reversible sentences
The apple that the boy is eating is red.
Versus semantically reversible sentences
The boy that the girl is chasing is tall.
Good to perfect performance by Broca’s
aphasics on the first, poor on the second.
Hypothesis: A common mechanism is affected
in understanding and in production
Agrammatic production goes with “asyntactic
comprehension”
But patients not obeying this generalization
were soon discovered.
LOT Summer 2006
Loss of speech
41
Rectifying the claim:





Production and comprehension may not exploit
a common mechanism
Clinical categories (such as Broca’s aphasia)
do “not uniquely determine the nature of the
underlying deficit in the patient included in
those categories.” (Caramazza et al. 2001)
Notice: The reversible/irreversible character of
the test sentences was lexical-semantic
But it could also be reinterpreted as “canonical”
agent-theme order versus an inverse order.
This is what Grodzinsky emphasized, coming
up with a new syntactic hypothesis: The TDH
(Trace Deletion Hypothesis)
LOT Summer 2006
Loss of speech
42
The Trace Deletion Hypothesis
Let’s see it in the “pure” form,
before we discuss the problems
Broca’s aphasics have no problem
understanding:







Basic syntactic trees and violations of basic
phrase-structure rules
Lexical meanings and their interface with
syntax (violations of sub-categorizations are
easily detected by them)
Argument structure (when movement and
traces are not involved)
(Full interpretation is unaffected)
Basic inter-sentential dependencies (relatives,
subordinates etc. When no movement and
traces are involved)
Case assignment (esp. in languages that have
overt case, such as Serbo-Croat)
Binding and anaphoric relations (with some
exceptions)
LOT Summer 2006
Loss of speech
44
Zooming onto Broca’s area lesions






Grodzinsky’s TDH1 (Trace Deletion Hypothesis
first version): Broca’s aphasics have an
impairment in comprehension only when
confronted with sentences that present
movement and traces, but not otherwise.
A “particularly influential hypothesis” (said by an
opponent: Alfonso Caramazza)
Methods: Picture selection task upon
presentation of sentences, truth judgments, and
detection of non-grammaticality
Reminder: They cannot “produce” such
sentences (but see the tree-pruning hypothesis)
Above-chance (often 100%) correct
performance signals absence of a
comprehension impairment
Chance performance signals a comprehension
impairment
LOT Summer 2006
Loss of speech
45
What’s being ruled out:







Generic “working memory” impairment
Generic “linking” impairment
Generic impairment with “inversions” (extending
to non-linguistic domains, see Grodzinsky’s
2000 BBS paper)
Generic impairment with “complex” syntactic
constructions of all kinds
A problem confined to lexical knowledge
And/or to “encyclopedic” knowledge
The spoon ate the table is understood perfectly,
in spite of its semantic implausibility
LOT Summer 2006
Loss of speech
46
Typical data: Minimal pairs
(AC=above chance, C=chance)
 Active versus passive
 The woman is chasing the man
AC
The man is chased by the woman C
 Subject relative versus object relative
 The woman who is chasing the man is tall AC
The man that the woman is chasing is tall C
 Subject-gap versus object-gap
 Show me the woman who is chasing the man AC
Show me the man who the woman is chasing C
 Subject cleft versus object cleft
 It is the woman that is chasing the man
AC
It is the man that the woman is chasing C
LOT Summer 2006
Loss of speech
47
Other data















No impairment with head-movement as such.
Grammaticality judgments are perfect for:
Could they have left town?
*Have they could leave town?
John did not sit.
*John sat not.
OK with full interpretation
Who did John see?
*Who did John see Joe?
*Mary ate the bread that I baked a cake.
OK with selectional restrictions on transitive
complements and object deletion
The children sang.
*The children sang the ball over the fence.
*The children threw.
The children threw the ball over the fence.
LOT Summer 2006
Loss of speech
48
Refining the hypothesis





Broca’s area as “neural home to receptive
mechanisms involved in the computation of the
relation between transformationally moved
phrasal constituents and their extraction sites”
(Grodzinsky, 2000, BBS)
But those patients perform successfully in
constructions that involve movement of the VPinternal subject to [Spec, IP] (Hickok, 1992)
A better hypothesis: -conflict
Somehow, the patient is receiving thematic
information that both NPs in the sentence have
the same -role, and that, therefore, any one of
them can be matched to the agent and the
patient argument in the sentence.
The performance is, thus, 50% (chance)
LOT Summer 2006
Loss of speech
49
Minimal pairs







Which man t touched Mary? AC
Which man did Mary touch t? C
The man who t is touching Mary is tall. AC
The man who Mary is touching t is tall. C
A non-grammatical strategy assigns the role of
agent to the first NP, and that of patient (or
theme) to the second (in English)
When this matches the grammatical (impaired)
processing, all is OK
When it does not, a conflict arises, and
performance is at chance level
LOT Summer 2006
Loss of speech
50
A specific hypothesis





The TDH hypothesis does not bear upon the
mere syntactic “complexity” of the sentence
Nor upon the “first” versus “second” position of
the NPs and the traces
Nor upon the “left” versus “right” position
It bears upon the standard position of -roles
and arguments in the patient’s language,
whatever that is.
In fact, in Chinese (an otherwise SVO language
like English) the heads of the relative clauses
follow the relative, contrary to English
LOT Summer 2006
Loss of speech
51
Minimal pairs in Chinese and English








The h index indicates the head of the relative
clause
[t zhouei gou] de mauh hen da
C
chased dog that cat very big
[mau zhouei t] de gouh hen xiao AC
cat chased that dog very small
The inverse is the case in English
The cath that [t chased the dog] was very big AC
The dogh that [the cat chased t] was very big C
Mirror-image results: Agent/agent conflict in
English, theme/theme conflict in Chinese
Similar results in Hebrew, Spanish, Korean and
German (see Grodzinsky’s papers for biblio)
LOT Summer 2006
Loss of speech
52
Scrambling in Japanese: Minimal pairs






2 possible configurations, with different results
in Broca’s aphasics (Hagiwara and Caplan,
1990)
(a) Taro-ga Hanako-o nagutta
AC
Taro hit Hanako
Subject Object Verb
(b) Hanako-o Taro-ga t nagutta
C
Object Subject t Verb
SOV is the basic order, while OStV is the
“scrambled” (secondary) order (Hale, 1983;
Saito, 1985; Miyagawa, 1997)
In (b) Hanako must c-command the VP, so it
must have moved to adjoin a higher projection
than that of the subject (Taro)
(See many details in the BBS paper)
LOT Summer 2006
Loss of speech
53
Scrambling in Japanese aphasics







Basic (unscrambled) actives and indirect
passives are above chance
Derived (scrambled) actives and direct
passives are at chance
This shows that it is not the passive
construction as such that is affected
(This has exasperated Caramazza)
The real discriminator is movement (and traces)
Not insensitivity to overt morphological “cues”
of the passive construction (-en and the
preposition by in English)
Nor semantic “plausibility” (as we saw)
LOT Summer 2006
Loss of speech
54
The -conflict hypothesis TDH2 (2000)









Full interpretation is preserved (see previous data): A
thematic role must be assigned to every NP in the
sentence.
A non-syntactic linear order strategy is also applied
(see Bever 1970, Bever and Townsend 2002, Jaeggli
1986)
<NP1=agent; NP2=theme> (in English)
Whenever you identify one -role for one NP, then you
assign the other -role to the other NP
When constituents are moved, the interpretation of
traces becomes crucial
But the Broca’s aphasic cannot interpret traces, nor
moved constituents. His/her syntactic representation is
“traceless”
So he/she resorts to the linear order strategy
When the linear order matches the syntactic
assignment, all is OK
When not, a conflict arises, and comprehension is at
chance level
LOT Summer 2006
Loss of speech
55
The -conflict hypothesis











We have a partial thematic representation and a nonsyntactic cognitive strategy (à la Bever et al.) that tries to
compensate for the impotence of a “traceless” syntax
When the “theme” role is correctly assigned to the second
NP in the sentence, then the agent role is correctly
assigned to the first NP
Which man t touched Mary? AC
................................. Mary =theme (no mediation of the
trace)
Therefore
Which man = agent (regardless of the trace)
But
Which man did Mary touch t? C
Mary =agent of the relative clause (no trace mediation)
Which man = agent in virtue of the linear strategy
Therefore conflict, and 50%-50% resolution of the conflict
LOT Summer 2006
Loss of speech
56
The -conflict hypothesis











There should also be cases of systematic inversion (that is,
below chance performance)
In fact there are (Grodzinsky, 1995, 2000)
with psych verbs (admire, love, adore, fear, etc.) (Belletti
and Rizzi, 1988; Pesetsky, 1995)
The syntactc subject is not really an “agent”, it’s rather an
“experiencer”, and the “object” is really a “theme”
Normal assignment in yellow, Broca’s aphasic’s
assignment in blue:
Theme
Experiencer
[The girl]i was t’j admired ti by [the boy]j
Agent
Experiencer
No conflict here. The inversion is systematic
No problem with the active counterpart
The boy admired the girl
LOT Summer 2006
Loss of speech
57
Additional evidence for TDH: fMRI




With important caveats, it should be the case
that normal subjects asked to perform on
sentences with inverted traces show an
activation of Broca’s area.
One has to tear this apart from other, generic,
estimates of sentence complexity
Maintain sentence length, number of words,
embeddings, functional to lexical categories
ratio etc. constant
Let only the relevant parameters (number of
arguments and movement, and the position of
the traces) vary
LOT Summer 2006
Loss of speech
58
Minimal pairs






M. Ben-Shachar’s minimal pairs in Hebrew
Object relative clauses versus main verbs that
take CP complements
‘azarti la-yalda [Se-Rina pagSa t ba-gina]
helped-I to-the-girl that-Rina met t in-the-garden
I helped the girl [that Rina met t in the garden]
‘amarti le-Rina [Se-ha-yalda yaSna ba-gina]
told-I to-Rina that-the-girl slept in-the-garden
I told Rina [that the girl slept in the garden]
Switch the verb (meet exchanged for sleep)
and you have an ungrammatical counterpart for
each sentence
The subjects were asked to make
grammaticality judgments for each minimal pair
LOT Summer 2006
Loss of speech
59
Results:



More intense signal in the left inferior frontal
gyrus (Broca’s area) for the sentences that
involve movement and trace
In agreement with the data on Broca’s aphasics
Also Hershl’s gyri (Brodmann area 22) are
activated bi-laterally (temporal lobes)
LOT Summer 2006
Loss of speech
60
R
L
From Ben-Schachar et al. (in Grodzinsky’s 2000 paper)
LOT Summer 2006
Loss of speech
61
Herschl’s gyrus
Parietal
Frontal
Occipital
Temporal
Areas 22 are also activated bi-laterally. Syntax is not all in the left
hemisphere
LOT Summer 2006
Loss of speech
62
Similar fMRI data for scrambling



Activation of the same areas (Broca’s left, and
Herschl’s bilaterally) is observed for scrambled
embedded double-object verbs in German
(Röder et al, 2002)
Jetzt wird der Astronaut dem Forscher den
Mond beschreiben
Now will the astranaut [to] the scientist the
moon describe
Jetzt wird dem Forscher den Mond der
Astronaut t t beschreiben
LOT Summer 2006
Loss of speech
63
And for overlapping regions being
activated by:


Embedded wh-questions versus yes/no
questions
Object topicalized versus non-topicalized main
clauses
LOT Summer 2006
Loss of speech
64
Additional data from psycholinguistics







Priming and gap-filling (see J. D. Fodor, D.
Swinney et al, and Zurif’s chapter in Osherson)
Sentences heard over earphones
The man liked the tailori with the British accent
who (t)i claimed to know the queen.
Lexical probe (clothes and boat) flashed on a
screen, alternatively in position 1 (pre-gap) and
2 (at the gap). Lexical decision time is attributed
to re-activation of the tailor at the gap
The man liked the tailori with the British accent1
who2 (t)i claimed to know the queen.
Broca’s aphasics manifest no difference at all
Normal subjects and Wernicke’s aphasics show
a robust gap-filling effect.
LOT Summer 2006
Loss of speech
65
A virtuoso extension (see the paper)








Can fMRI data “decide” between competing
syntactic hypotheses?
Data from location and intensity are needed
Location tells about same versus different
computations
Intensity tells about effort (primary versus
additional computations)
Double object constructions (Larson, 1988;
versus Aoun and Li, 1989)
Data from fMRI on Hebrew stimuli (BenSchachar and Grodzinsky, 2002)
What type of movement (if any) is involved?
Which complement order (dative or double
object) is derived, and which base-generated?
LOT Summer 2006
Loss of speech
66
Data on dative/double object
Dative-shift contrast versus topicalization
contrast on minimal pairs in Hebrew
 Activation of different brain regions would signal
a difference in processing
 Different relative intensity would signal greater
effort (greater mental computation) for the
construction that is not base-generated
 There is a difference of activation for
topicalization versus dative shift
 The latter activates right frontal brain regions,
different from those activated by topicalization
 In these regions, activation is significantly
higher for the double object than for the dative
shift
 In Hebrew, double object constructions appear,
therefore, to be derived and not basegenerated.
LOT Summer 2006
Loss of speech
67

Comments to Grodzinsky’s BBS
target article









Other hypotheses can explain all the data
The data are mostly (all) wrong
Only parts of the hypothesis are correct
Issues with definitions, categorizations,
anatomical localizations
Compatible with GB, but not minimalism,
therefore minimalism is wrong
Better compatible with minimalism, rather than
GB, therefore minimalism is right
The implications of the hypothesis are quite
different
Checks with “better data” are suggested
Better confirming data are actually offered
LOT Summer 2006
Loss of speech
68
Comments to Grodzinsky’s BBS
target article






Other conclusions should be derived from the
data
Other kinds of data should be explored
The basic assumptions are flawed (modularity,
specificity of language, the value of generative
grammar, etc.)
This work has merits, in spite of specific
problems
This work is basically flawed, in spite of its
limited success
Accusations of neglect of other (falsifying) data
LOT Summer 2006
Loss of speech
69
Not everything that shines is gold
Caramazza, Capitani, Rey and Berndt (2001)
Brain and Language, Vol. 76, pp.158-184
 Brandishing Popper’s criterion of falsifiability,
 Produce additional data, and re-examine
Grodzinsky’s own data, concluding that
 There is a mixed bag of symptoms, in Broca’s
aphasics, including instances of TDH, but a lot
more,
 And cases in which there is no evidence of
DTH: “a number of distinct patterns in different
patients”.
 Their specific target is active vs passive and
subject vs object relatives and clefts
 A battle on quantifiers: “all” patients versus
“most” patients, “necessarily”
versus “possibly” 70
LOT Summer 2006
Loss of speech

Conflicting statistical analyses





Berndt et al. (1996) and Grodzinsky et al.
(1999) reach different conclusions from the
same data bank of patients (over 40 cases), on
tests on passive versus active sentences.
Only 1/3rd of the patients confirm the pattern
predicted by Grodzinsky’s TDH
Is the grouping of patients under the category
“Broca’s aphasics” consistent?
Excruciating (to us) details follow as to the
proper classification of patients, the proper
methods of testing, and the right (as opposed to
“creative”) kind of statistical analyses.
Over 83% correct is labelled “above chance”,
55% or lower is labelled “chance”
LOT Summer 2006
Loss of speech
71
The Caramazza et al.’s calculations
(in essence)




If a population of patients (as it were) flips a coin to
understand a “hard” sentence, we get a single-mode
binomial distribution (a Gaussian) distributed around
the mean 50%
But, if the population is not uniform for this task, and
rather shows great individual variance for this task, then
we have a multi-modal distribution, with important “tails”
at the extremes. These must not be merged with the
rest in the statistical analysis
They show that the distribution is, indeed, multi-modal
Broca’s aphasics do not constitute a homogenous
population for the active/passive comprehension test.
LOT Summer 2006
Loss of speech
72
Caramazza et al.’s conclusions






Grodzinsky’s hypothesis is falsified
Even when one concedes the “corrections”
suggested by Grodzinsky
The suggestion to re-classify bona fide Broca’s
aphasics is an objectionable move,
(At least until new objective and rigorous
criteria are provided)
Making the hypothesis un-falsifiable, and
therefore un-scientific
Popper’s famous case of black swans is cited
LOT Summer 2006
Loss of speech
73
Grodzinsky’s reply:





Data on passives are indeed not neat (the case
of Dutch). There is individual variation
Caramazza et al.’s claims would indeed be
worrysome, if correct.
Critical re-analysis of the replication failures
(test-design, patient diagnosis, test
administration and statistical analysis)
In spite of inevitable (minor) individual
differences in the lesions, the diagnosis of
Broca’s aphasia is sufficiently consistent across
languages, laboratories and groups.
A very large data set for the statistics has been
assembled
LOT Summer 2006
Loss of speech
74
Statistical criteria






77 patients tested on 2 groups of constructions
(1) ac-TDH Above chance (actives, subject
relatives, subject questions, subject clefts)
(2) c-TDH chance (passives, scrambled actives
- in German, Spanish, Hebrew and Korean-,
object relatives, object questions, object clefts)
Number of patients showing c (40-70% correct)
and, respectively, ac (over 80% correct) is
plotted against percentage correct
The range of individual variation on c is much
greater that than on ac
Mean around 50% for c, quasi-asymptote to
100% for ac
LOT Summer 2006
Loss of speech
75
From Y. Grodzinsky (in the Jenkins volume)
The solid curve is indistinguishable from chance
LOT Summer 2006
Loss of speech
76
Is this one group of subjects?





The wide distribution of the c curve has
suggested to Caramazza et al that we are
dealing with a heterogeneous group of subjects
Grodzinsky and colleagues deny that this is the
case
Essentially, they compute the numerical
advantage represented by a bi-modal
Gaussian, versus a uni-modal Gaussian
They conclude that there is no advantage (no
better approximation to the data) to the bimodal
And conclude that we are indeed dealing with
one homogeneous group of patients (one
syndrome)
LOT Summer 2006
Loss of speech
77
A new development



Dan Drai and Yosef Grodzinsky A new empirical angle on the
variability debate: Quantitative neurosyntactic analyses of a
large data set from Broca’s Aphasia Brain and Language,
Volume 96, Issue 2, February 2006, Pages 117-128
David Caplan, Gayle DeDe and Hiram Brownell Effects of
syntactic features on sentence–picture matching in Broca’s
aphasics: A reply to Drai and Grodzinksy (2005) •
DISCUSSION Brain and Language, Volume 96, Issue 2,
February 2006, Pages 129-134
The punch line: “Performance variation within a group of
patients in itself does not preclude the existence of structure
in their deficit. Thus in aphasia the data may present interpatient variability, but the challenge for us is to try and
discover commonalities at the group level in the face of this
variability”.
LOT Summer 2006
Loss of speech
78
A new development


We have compiled a large data set of raw data of
performance scores, and adopted a statistical methodology
which gives a precise quantitative meaning to the question of
the existence of a significant difference in performance —
analyzed at the group level— between two types of sentences.
Note that the approach is general and can be applied to any
subject group, and to any categorization of the sentences. We
did apply our method to performances of Broca’s aphasics,
for whom we have different existing categorizations of
sentences.
This method reveals a highly significant performance
difference when the data are categorized by a syntactic
principle (whether or not sentences contain a
Transformational Movement relation); more traditional ones
(Mood [active/passive], Sentence Complexity [high/low])
yield no contrast.
LOT Summer 2006
Loss of speech
79
What seems to be uncontroversial:






Language processing is modular
Different brain regions (mostly, but not
exclusively, in the left cerebral hemisphere) are
involved in different linguistic computations
Some regions are involved specifically in
syntactic processing
Most Broca’s aphasics match the picture given
by the Trace Deletion Hypothesis:
They have a specific impairment with
infrasentential dependencies involving traces,
when constituents have been extracted from
the object (or theme) position
Data from lesions and data from imaging are
comparable and approximately converging
LOT Summer 2006
Loss of speech
80
A new book that says it all:



Yosef Grodzinsky and Katrin Amunts Broca’s
Region Oxford University Press (2006)
Proceedings of a conference, much re-worked
and updated
With a new translation of Broca’s original paper
and with other historical landmark papers.
LOT Summer 2006
Loss of speech
81
Broca’s Area Revisited



A few cytoarchitectonic borders, e.g., between
the somatosensory cortex and the primary
motor cortex and between the primary and
secondary visual areas, are obvious and can be
clearly defined by pure visual inspection.
For these borders, discrepancies between
different observers are marginal.
However, the cytoarchitecture of the vast
majority of cytoarchitectonic areas does not
differ this distinctly.
Amunts, K., Schleicher, A., Buerger, U., Mohlberg, H., Uylings, H. B. M., & Zilles, K.
(1999). Broca's region revisited: Cytoarchitecture and intersubject variability.
The Journal of Comparative Neurology, 412, 319-341.
LOT Summer 2006
Loss of speech
82
Broca’s Area Revisited



The observer-independent procedure
introduced in this paper localizes the precise
position of the border even in such cases
because the definition of borders is based on
single and significant peaks in the multivariate
distance function.
This procedure allows one to identify a
transitional area as a distinct cytoarchitectonic
unit with reproducibly definable borders.
A similar statistical approach is widely used in
functional imaging studies because clusters of
activation are considered meaningful only if
there are significant differences between signal
and ‘‘noise’’.
Amunts, K., Schleicher, A., Buerger, U., Mohlberg, H., Uylings, H. B. M., & Zilles, K. (1999).
Broca's region revisited: Cytoarchitecture and intersubject variability. The Journal of
Comparative Neurology, 412, 319-341.
LOT Summer 2006
Loss of speech
83
Broca’s Area Revisited






The laminar distribution of cell densities was measured with an
automated microscopic scanning procedure using an image
analysis system (Schleicher and Zilles, 1990).
The resulting gray level index (GLI) is a reliable measure of cell
packing density in the cortex (Wree et al., 1982).
Statistically significant changes in the laminar distribution of the
GLI can be detected at the transition between two cytoarchitectonic
areas but not within an homogeneous area (Schleicher et al., 1995,
1998).
A procedure based on the detection of such changes in GLI has
been shown to be sensitive for interhemispheric, ontogenetic, and
areal differences (Zilles et al., 1986b; Schlaug et al., 1995b;
Amunts et al., 1996, 1997; Geyer et al., 1996, 1997).
Thus, this procedure avoids observer-dependent influences in
defining areal borders and makes possible the statistical testing for
significant changes in cytoarchitectonic organization.
An analysis of 10 human brains
Amunts, K., Schleicher, A., Buerger, U., Mohlberg, H., Uylings, H. B. M., & Zilles, K. (1999). Broca's region
revisited: Cytoarchitecture and intersubject variability. The Journal of Comparative Neurology, 412, 319-341.
LOT Summer 2006
Loss of speech
84
Mean cell density profiles
Ordinates: Gray Level Index (GLI) in %
(Amunts et al. 1999)
0% depth = boundary between layers I and II;
100% = boundary with the white matter
LOT Summer 2006
Loss of speech
Averages of 10
individual profiles
85
Vertical lines of penetration
Location of the image
in A
LOT Summer 2006
Loss of speech
86
The results, in synthesis:






Area 44 was identified on the basis of conspicuously
large pyramidal cells in deep layer III and in layer V and
by a barely recognizable dysgranular layer IV, which
was invaded to different degrees by layer III and V
pyramidal cells.
Area 45 differed essentially from area 44 by the
presence of a clearly visible layer IV.
Due to the more pronounced layer IV, the horizontal
layering of area 45 also appeared more conspicuous
than did that of area 44.
Layer IV of area 45, however, was less distinct than that
in the rostrally adjoining prefrontal cortex, e.g., in areas
10 or 46
The cytoarchitectonic borders did not coincide with
reliably identifiable macroscopic features, e.g., with a
fundus of a sulcus. Thus, macroscopic anatomy and
areal borders differ independently.
Large intersubject variability in the
cytoarchitecture of each cortical area
Loss of speech
LOT Summer 2006
87
The results, in synthesis:






All five male brains showed a left-larger-than-right
asymmetry in area 44.
By contrast, the Asymmetry Coefficients for females in
area 44 and for both sexes in area 45 were more
evenly distributed, i.e., no interhemispheric asymmetry
could be detected in these cases.
Intersubject variability in cytoarchitecture was accompanied by
intersubject variability in the positions of areas 44 and 45 relative to
sulci and gyri.
The spatial variability of borders relative to macroscopic features
therefore is a relevant factor for the architectonic interpretation of
functional imaging studies with a spatial resolution of a few
millimeters.
The classic maps do not provide information concerning the
position of borders in the depths of a sulcus. Kononova (1938)
found areal borders of areas 44 and 45 that did not coincide with
the fundus of a sulcus.
In some sections, area 45 does not reach the inferior frontal
sulcus; in other sections, area 45 occupies parts of the dorsal wall
of the middle frontal gyrus
LOT Summer 2006
Loss of speech
88
Inter-hemispheric asymmetries








Although the volumes of area 44 differed across subjects by up to a factor
of 10, area 44 but not area 45 was left-over-right asymmetrical in all
brains.
The interhemispheric differences in the volume of area 44 appear
to be greater than those in cytoarchitecture.
A left-over-right asymmetry was detected in all 10 cases.
This finding supports that of a previous study in which a
significantly larger area 44 was found on the left than on the right
(Galaburda, 1980). Nine of the 10 subjects of that study had a
larger area 44 on the left side.
If the present sample is combined with that in the study by
Galaburda, the incidence of leftward asymmetry (19 of 20 cases,
95%) matches the frequency of left-sided speech dominance in the
general population and may structurally reflect the functional
lateralization
No consistent left–right difference in the volume of area 45 was
found.
Area 45 is not more symmetric than area 44; it is more heterogeneous
pertaining to the direction of asymmetry.
The data on volumes of areas 44 and 45 also clearly showed that the
variability of this parameter and of the total brain weights (and, thus, brain
volumes) are considerable.
LOT Summer 2006
Loss of speech
89
Individual
Statistically defined maps versus visible anatomy
Individual
LOT Summer 2006
Loss of speech
90
Brodmann 1909
Variation in the
anatomical definition
of Areas 44 and 45
(Broca’s area)
Economo &
Koskinas 1925
(Amunts et al. 1999)
Sarkisov et al. 1949
LOT Summer 2006
Loss of speech
91
Broca’s aphasia revisited (Friedmann 2006)






Who did the cat chase? AC
Which dog did the cat chase? C
WHO is a pure operator, while WHICH is
discourse-dependent and therefore
computationally “more costly” (Hickok and
Avrutin 1995; Tait 1995; Avrutin 2006)
Strong asymmetry between tense and subjectverb agreement inflection (tested in many
languages)
Subject-verb agr. 65% correct or more
Tense is at chance level
LOT Summer 2006
Loss of speech
92
Broca’s aphasia revisited (Friedmann 2006)









A = affected NA = not affected
A Subject pronouns
NA Object pronouns
A Relatives
NA Reduced relatives
A Wh-questions
NA yes/no questions
A Subordination conjunctions
NA Coordination conjunctions
Language variability:
A yes/no questions in Dutch, English and
German
NA yes/no questions in Hebrew and Arabic
LOT Summer 2006
Loss of speech
93
The Tree Pruning Hypothesis (TPH)
(Friedmann, 1998, 1999, 2006; Friedmann and Grodzinsky,
1997, 2000)

In essence: The highest functional nodes in the
syntactic tree are selectively affected
CP
Wh-question
milder
C’
C
Complementizer
TP
severe
T’
T
Tense
NegP
AgrP
Agr’
LOT Summer 2006
vP
AgreementLoss of speech
94
The Tree Pruning Hypothesis (TPH)
(Friedmann, 1998, 1999, 2006; Frriedmann and Grodzinsky,
1997, 2000)

Recovery of S.B. a 20 years-old Hebrew speaker with
traumatic brain injury (Friedmann, 2005, 2006)
CP
Wh-question
15 months
C’
C
Complementizer
TP
6.5 months
T’
T
Tense
NegP
4.5 months
AgrP
Agr’
LOT Summer 2006
vP
AgreementLoss of speech
95