Semantic Reversal Errors in Agrammatic Aphasics

Semantic Reversal Errors in Sentence
Comprehension
Zachary Ekves
SREBCS, University of Pittsburgh
Mentors: Julius Fridriksson, Paul Fillmore
August 1, 2013
Outline
 Semantic reversal
 Approaches to agrammatism
 Language comprehension background
 Methods
 Results/Discussion
Semantic Reversal
 “The boy paints the girl.”
 Arguments:
 the boy (subject/agent)
 the girl (object/theme/patient)
 Canonical vs non-canonical
Agrammatism Overview
 Looking at chronic stroke patients
 Lesion data
 Areas required for a function
Trace Deletion Hypothesis
 Traces are deleted from representations
 Grodzinsky (1995)
 No traces leads to agent first processing
 Normal syntactic processing for second
NP
 But, no preference is shown for double
agent representations
(Beretta and Munn, 1998)
Resource Based Theories
 Comprehension difficulties arise from a
deficit of a resource, perhaps working
memory
(Caplan & Waters, 1999)
 Syntactic features of specific sentence
types are not the cause of error
 All comprehension is affected to
different extents
Eyetracking Research
 Representations are possibly the same
as normal populations
(Dickey & Thompson, 2009)
 Errors in comprehension arise from
lexical retrieval/processing
(Meyer, Mack, & Thompson, 2012)
Temporoparietal Area
 Associated with thematic role
assignment errors
(Thothathiri, Kimberg, & Schwartz, 2012)
 Thematic semantics
(Mirman & Graziano, 2012)
 Angular Gyrus: semantic structure
recognition
(Humphries, Binder, Medler, & Liebenthal, 2006)
Other Posterior Regions
 Lexical access to verbs and verb
argument knowledge, and conceptual
verb argument knowledge
(Den Ouden, Fix, Parrish, & Thompson,
2009; Wu, Waller, and Chatterjee, 2007)
 Canonical Sentence Processing
(Magnusdottir et al.,
2012)
 Retrieval of lexical and semantic
information
(Hickok & Poeppel, 2007)
Anterior and Middle Temporal Lobe
 Linguistic knowledge of verb arguments
(Wu, Waller, and Chatterjee, 2007)
 Non-canonical sentence processing
(Magnusdottir et al., 2012)
 Activation for syntax manipulations,
not semantic
(Liebenthal, 2006)
Participant Information
 33 chronic stroke patients analyzed
 Mean age of 67.70 (SD = 9.74)
 Tested 3.93 years later, on average
Behavioral Testing
“It is the girl who tickles the boy.”
 9 Sentence Types (1,4,8 : canonical; 2,5,9 : non-canonical)
 1. Active sentences with a main verb and overt
case marking
 “The girl paints the boy.”
 2. Passive Sentences with overt case marking
 “The boy is painted by the girl.”
 3. Truncated passive sentences with overt case
marking
 “The boy is painted.”
 4. Subject cleft sentences with overt case marking
 “It is the boy that paints the girl.”
 5. Object cleft sentences with overt case marking
 “It is the girl that the boy paints.”
 6. Sentences with topicalized object and a main
verb
 “The girl, the boy paints.”
 7. Sentences with topicalized object and an
auxiliary verb
 “The girl, the boy is painting.”
 8. Referential which clauses (subjects) with a
main verb
 “Which boy paints the girl?”
 9. Referential which clauses (objects) with a
main verb
 “Which boy is the girl painting?”
Taken from Magnusdottir et al. (2012)
Behavioral Results
Type
Mean
Standard
Deviation
Semantic Error
(overall)
7.55
6.94
1
0.33
0.60
2
1.03
1.42
4
0.30
0.77
8
1.82
1.26
9
0.70
1.19
Lex Error (overall)
0.97
1.65
Neuroimaging Data
 VLSM
False Discovery Rate of P < .05
Semantic Errors Overall
0.05 = -2.59
Lexical Errors Overall
0.05 = -3.53
Type 1: Active
Type 2: Passive
.05 = -3.54
.05 = -2.57
Type 4: Subject cleft sentence
0.05 = -2.9
Type 8: Referential which clauses (subjects)
.05 = -2.58
Type 9: Referential which clauses (objects)
.05 = -2.92
Discussion
 Lexical errors
 Indicative of verb retrieval
 Supports previous research
 Active sentence errors
 Thematic role assignment in simple sentences
 Supports and extends previous research
 Passive sentence errors
 Temporal regions: thematic role deficits/lexical deficiencies
 Temporal pole: syntactic integration
 Direct comparison between active and passive
 Subject cleft sentence errors
 Frontal association in a type without movement
 Referential which clause (subjects)
 Angular gyrus
 Referential which clause (objects)
 Not as posterior
Overall Semantic Errors
 Temporal Pole
 Syntactic integration
 Middle Temporal Gyrus
 Linguistic verb argument knowledge
 Lexical processing
 More involvement with more complex syntax, in general
Future Directions
 Temporal pole
 Morphosyntactic processing?
(Dronkers et al., 1994)
 Teasing apart functions of the temporal
pole and middle portions of the
temporal lobe
Acknowledgements
 Julius Fridriksson
 Paul Fillmore
 Everyone involved with data collection
 SREBCS