Power point

Title Page

Does the Thematic Hierarchy hold in people with aphasia and across the lifespan? Evidence from the Event Task.

by

Sophia Norvilas

Submitted to the Graduate Faculty of the

Dietrich School of Arts and Sciences and

Fredrick Honors College in partial fulfillment

of the requirements for the degree of

Bachelor of Philosophy

University of Pittsburgh

2022

ii

Commit tee Membership Page

UNIVERSITY OF PITTSBURGH

DIETRICH SCHOOL OF ARTS AND SCIENCES

This thesis was presented

by

Sophia Norvilas

It was defended on

November 11, 2022

and approved by

Dr. Tessa Warren, Professor, Psychology

Dr. Julie Fiez, Professor, Psychology

Dr. Jamie Reilly, Professor, Communication Science and Disorders

Thesis Advisor: Dr. Michael Walsh Dickey, Professor, Communication Science and Disorders

iii

Copyright © by Sophia Norvilas

2022

iv

Abstract

Does the Thematic Hierarchy hold in people with aphasia and across the lifespan? Evidence from the Event Task.

Sophia Norvilas

University of Pittsburgh, 2022

Aphasia is a neurological disorder that can disrupt language production and

comprehension, impairing both written and spoken language (Dresang et al., 2019). This condition

is typically brought on by brain damage following a stroke. Research indicates that people with

aphasia sometimes rely on event knowledge to compensate for their language impairment

(Caramazza & Zurif, 1976). However, we know little about event processing in people with

aphasia (PWA), even though event knowledge supports a multitude of crucial cognitive processes,

including language comprehension, language production, memory, and perception. One type of

event knowledge that has been studied thoroughly in linguistics is the entities, objects, and

locations (event roles) that are involved in events. Linguists have developed a hypothesis, known

as the Thematic Hierarchy, that some of these event roles are more cognitively salient than others.

The research I present here uses evidence from a new assessment that measures event knowledge

(the Event Task) to evaluate whether this thematic knowledge is maintained in older adults and

people with aphasia, while also examining whether the performance of PWA on the Event Task is

aligned with the Thematic Hierarchy. PWA (N = 26) and neurologically healthy adults (N = 182)

completed the Event Task, which instructed participants to identify whether a depicted event was

plausible or implausible. Analyses showed that the Thematic Hierarchy did not appear to guide

the performance of PWA or neurologically healthy adults across the lifespan. However, PWA and

neurologically healthy controls alike displayed the same patterns of both accuracy and reaction

v

time performance. In neurologically healthy adults, increased age was correlated with decreased

accuracy and increased reaction times. In addition, neurologically healthy adults with 12+ years

of education were found to have increased accuracy and decreased reaction times. The current

findings could be the foundation for future research regarding aphasia and event knowledge.

vi

Table of Contents

Preface ............................................................................................................................................ x

1.0 Introduction ............................................................................................................................. 1

1.1 Semantic memory and event knowledge ...................................................................... 1

1.2 The Thematic Hierarchy ................................................................................................ 2

1.3 The Event Task ............................................................................................................... 5

1.4 The Current Study ......................................................................................................... 7

2.0 Methods .................................................................................................................................... 8

2.1 Participants ..................................................................................................................... 8

2.2 Event Task Stimuli and Procedure ............................................................................... 8

2.3 Data Compilation ............................................................................................................ 9

2.4 Event Taxonomy Questionnaire .................................................................................. 10

2.5 Data Analysis ................................................................................................................ 11

3.0 Results .................................................................................................................................... 12

3.1 PWA and Matched Controls Analyses ....................................................................... 12

3.2 PWA Analyses ............................................................................................................... 14

3.3 Lifespan Analyses ......................................................................................................... 17

3.3.1 Age Analyses .......................................................................................................18

3.3.2 Mini-Mental State Examination Analyses .......................................................20

3.3.3 Years of Education Analyses .............................................................................22

4.0 Discussion............................................................................................................................... 24

4.1 PWA and Matched Controls ....................................................................................... 25

vii

4.2 Neurologically Healthy Adults Across the Lifespan .................................................. 27

4.3 Clinical Application ...................................................................................................... 30

4.4 Limitations .................................................................................................................... 34

5.0 Conclusion ............................................................................................................................. 35

Bibliography ................................................................................................................................ 37

viii

List of Tables

Table 1: Canonical and Exclusive Event Role Violations ....................................................... 11

ix

List of Figures

Figure 1: PWA and matched control participants' accuracy performance. ......................... 13

Figure 2: PWA and matched control participants’ reaction time performance. .................. 14

Figure 3: Individual PWA accuracy performance per event role violation type. ................. 16

Figure 4: Individual PWA reaction time performance per event role violation type .......... 17

Figure 5: Healthy participants’ accuracy performance in relation to age. ........................... 18

Figure 6: Healthy participants’ reaction time performance in relation to age. .................... 19

Figure 7: Healthy participants’ accuracy performance in relation to MMSE Score. .......... 20

Figure 8: Healthy participants’ reaction time performance in relation to MMSE Score. ... 21

Figure 9: Healthy participants’ accuracy performance in relation to years of education. .. 22

Figure 10: Healthy participants’ reaction time performance in relation to years of education.

........................................................................................................................................... 23

x

Preface

First and foremost, I would like to express my gratitude to Dr. Michael Walsh Dickey, my

mentor and Bachelor of Philosophy advisor. Thank you for the years of guidance, encouragement,

and insight. I will never forget the invaluable lessons he taught me and the countless hours he

dedicated to my project. Joining the Language and Brain Lab and learning under his guidance truly

changed the trajectory of my career, and I cannot express the full extent of my appreciation. In

addition, I am so lucky to have studied under Dr. Tessa Warren, my incredible Honors in

Psychology advisor. Her mentorship has been imperative in honing my writing skills and finding

my voice in the scientific community. Her dedication and expertise have been instrumental in

helping me become the researcher I am today.

I am very grateful to Dr. Julie Fiez and Dr. Jamie Reilly, for their enthusiasm and

willingness to serve on my thesis committee. I would like to acknowledge the University of

Pittsburgh Honors College for providing this wonderful opportunity and supporting my work with

resources such as the Brackenridge Fellowship. I am extremely appreciative of Emily Goldberg,

for assisting with data compilation efforts and cheering me on since my very first lab presentation

over two years ago. Finally, I would like to thank my friends and family, for providing unwavering

support and encouragement as I pursued this project.

1

1.0 Introduction

1.1 Semantic memory and event knowledge

Semantic memory is an individual’s long-lasting memory for general knowledge about the

world (Dresang et al., 2019). From birth, we begin collecting knowledge about the world. Semantic

memory is critical when trying to produce or understand language. For example, when you hear a

word or sentence, you consult your knowledge about objects, people, actions, and things to

understand the meaning of that sentence. Semantic memory is often impaired in PWA, and an

accurate measure of semantic memory impairment is crucial for treatment (Antonucci & Reilly,

2008). Existing picture-based assessments of object-related semantic memory include Pyramids

and Palm Trees (PPT) (Howard et al., 1992) and the Camels and Cactus Test (CCT) (Bozeat et al.,

2000). There is also a picture-based assessment of action-related semantic memory called the

Kissing and Dancing Test (KDT) (Bak & Hodges, 2003). However, there are few tests to measure

semantic memory for events in PWA (Dresang et al., 2019).

Events can be defined as collections of people, objects, actions, places, and the context in

which they occur (Dresang et al, 2019). Event roles (or thematic roles) are the people, places, and

things that play a role in an event. The agent is the initiator of an action (Ünal et al., 2021). The

patient is the entity undergoing the effect of some action, often undergoing some change in state

(Ünal et al., 2021). The instrument is the tool used to perform an action (Ünal et al., 2021). The

location is the place in which something is situated (Saeed et. al, 1996). The goal is the entity

towards which something moves, either literally or metaphorically (Ünal et al., 2021). An example

of this would be the following sentence: “The boy hit the baseball with his bat on the field.” The

2

agent would be “the boy” as he is the one doing the act. The action is hit. The patient is the baseball

because it is moved by the action. The instrument is the bat because the boy uses the bat as a tool

to hit the ball. The location is the baseball field on which he is playing the game.

Event role knowledge is activated during language processing. Some evidence for this

comes from a study in which adult participants were given line drawings of two-participant

events—for example, a man throwing a football to a woman—and were asked to describe the

events (Griffin & Bock, 2000). Eye-tracking data showed that the participants’ eye movements

were guided by event features prior to sentence formulation. Participants gazed at the agent prior

to speech, then looked to the patient as they began speaking, and finally returned their gaze to the

agent. The results showed that participants “rapidly extract roles of event participants” while

describing the line drawing (Griffin & Bock, 2000). Event knowledge provides context that helps

individuals understand novel events by providing prototypical events for comparison (McRae &

Matsuki, 2009). Language processing is reliant on event knowledge, as it supports verb retrieval

and thematic-role processing (Dresang et al., 2019).

1.2 The Thematic Hierarchy

The Thematic Hierarchy is the idea that certain event roles are more prominent than others

(Ünal et al., 2021). An event role near the top of the hierarchy would have more prominence,

meaning that it should be identified earlier and processed more quickly than a non-prominent event

role. Agents are at the top of the hierarchy and are most salient or prominent. Next are Patients,

followed by Goals. Instruments are second-to-lowest on the thematic hierarchy. Finally, locations

are the least prominent role. Linguistic research indicates that agents are most prominent because

3

they are linguistically encoded as subjects (Baker, 1997; Jackendoff, 1990). Patients are next

because they are direct objects of the verb. Goals, instruments, and locations are the least likely to

be in these syntactic positions, leading to a lower prominence on the thematic hierarchy.

Instruments are rarely selected as verb arguments and even referred to as “secondary roles” (Baker,

1997). This hierarchy naturally predicts that event roles towards the top of the Thematic Hierarchy

will be mentioned more frequently or identified more easily than event roles from the bottom of

the Thematic Hierarchy.

I will now outline the experimental evidence supporting the Thematic Hierarchy. Agents

and patients are expected to be prioritized over other event roles because of their positions towards

the top of the Thematic Hierarchy. Consistent with this, Hafri et al (2018) found that information

about agents and patients is often extracted even when it is not directly encouraged by a task or

while the participant is preoccupied with another task (Hafri et al., 2018). Participants viewed an

array of stimuli images depicting a simple interaction between a male and a female and were

instructed to answer which character, the male or the female, was on the left or right half of the

screen. When presented with subsequent images in which the event role of the target character

differed from one trial to the next, the participants displayed an event role switch cost, a delay in

reaction time caused by the shift in event role. Hafri et al.’s study found that although task

instructions did not explicitly refer to event roles, participants’ reaction times were affected by the

event role change, indicating that event role information is spontaneously encoded even when not

directly related to the task at hand.

Research suggests that instruments are not prioritized unless the instrument is deemed

particularly relevant to the story being told. Studies show that English speaking adults consistently

omit instruments in retelling stories involving instrument events (Lockridge & Brennan, 2002).

4

The instrument is even more likely to be omitted from the retelling if it is a common or typical

instrument. Lockridge and Brennan (2002) found that during story retellings, atypical instruments

were most often explicitly disclosed to the partner early in the story, whereas an instrument that is

inferable from the story was not mentioned as frequently. These findings suggest that although

instruments can provide crucial information about an event, they are not attended to with the same

focus that an agent or patient would receive, which is consistent with their low role on the Thematic

Hierarchy.

Locations are positioned at the bottom of the Thematic Hierarchy. Ferretti et al (2001) used

single word priming to investigate how verbs activate event role knowledge (Ferretti et al., 2001).

Results showed that verbs primed all event roles except for locations (Ferretti et al., 2001). These

findings are consistent with the idea that locations are the least salient of all event roles, resulting

in location’s place at the bottom of the Thematic Hierarchy.

From a cognitive perspective, the Thematic Hierarchy would predict that entities in more

prominent event roles would be remembered better and identified more easily than entities in less

prominent roles. This claim is supported by a study by Ünal et al. (2021) that tested young learners

of English and Turkish in order to determine whether asymmetries in event role prominence

generalize to other languages. The study had two parts: a linguistic description task and a change-

blindness task. In the linguistic description task, participants were asked to describe event stimuli

images. Event role prioritization was measured by how frequently participants mentioned each

event role when describing a stimuli image. For the change-blindness task, four new stimuli images

were created from each original stimuli image by changing the color of one event role. In the task,

the original stimuli image was displayed, followed by a grey screen. Next, the new stimuli image

featuring the color change was shown, followed by another grey screen. The cycle would then

5

repeat. Participants were instructed to respond as soon as they detected the changing object. In this

task, event role prioritization was measured by how quickly the participant responded to the

changed event role.

Findings suggested that agents were prioritized in both language groups. Patients and goals

tended to be similarly prioritized over instruments in both languages as well. The relative salience

of event roles was similar across young learners of both English and Turkish (Ünal et al., 2021).

This is particularly interesting because even though there is variation in the way these event roles

are encoded across the two languages, event roles were prioritized in similar ways, which aligns

with the Thematic Hierarchy. My thesis explores the theory that people should be faster and more

accurate to detect incongruency when it appears in more prominent roles than in less prominent

roles.

1.3 The Event Task

The Event Task is a picture-based assessment of semantic memory. The preliminary

version of the assessment was developed in the Language and Brain Lab at the University of

Pittsburgh. The task’s goal is to measure how functional or impaired an individual’s ability to

access their semantic memory for events is. In the task, the participant is given photographs of

events that are either plausible or implausible. Plausible events would be those that would

commonly occur in everyday life, such as a person doing work at a desk or boxing in a gym.

Implausible events are impossible or unlikely, and do not match our everyday experience or our

world knowledge based in semantic memory. Images like a man playing the violin underwater and

a woman with her head in a bag are clear examples of implausible events taken directly from the

6

Event stimuli. In each implausible image, at least one event role is incongruent with the event

depicted. Participants were shown one image at a time and were asked to indicate if this scene was

something that may normally happen.

A study conducted by Dresang et al. (2019) assessed the preliminary validity of the Event

Task. The study’s first aim was to characterize typical performance and distinguish stimulus

characteristics of the Event Task. To that end, it gathered data from 90 neurologically healthy

adults across the lifespan and identified both average performance at multiple age-ranges and

variation in processing time across the plausible and implausible images (Dresang et al., 2019).

The second aim was to establish the Event Task’s sensitivity to neurological impairment. Findings

concluded that the Event Task was successful in detecting those participants afflicted with aphasia,

indicated by distinctions between neurologically healthy age-matched controls and PWA (Dresang

et al., 2019). The third and fourth aims were to demonstrate that the Event Task shows preliminary

evidence of convergent validity with measures of language processing and assessments of action

and object knowledge. Poor Event Task performance was correlated with greater deficits in verb

retrieval (specifically, verb production) and thematic-role processing (specifically, producing

agent, patient, and other thematic roles in sentences), and Event Task performance was predicted

by KDT (action based semantic memory assessment) performance and PPT (object based semantic

memory assessment) performance (Dresang et al., 2019). These findings suggest that Event Task

performance is indicative of verb retrieval abilities and thematic-role processing abilities, which

are crucial for language processing. The current study expands upon the foundational work of

Dresang et al. by examining whether the Thematic Hierarchy’s predictions regarding the salience

of various event roles hold for Dresang et al.’s Event Task data.

7

1.4 The Current Study

My research investigated performance on the Event Task in relation to the Thematic

Hierarchy. My research tested whether there was a correlation between the relative prominence of

event roles according to the Thematic Hierarchy (agents > patients > instruments > locations) and

accuracy and reaction time for implausible Event Task pictures which contain violations of our

expectations about those event roles. For example, the stimulus image of a man playing the violin

in the ocean would be a location violation, as the location (the ocean) is not a typical place for that

event to occur. I hypothesized that there would be a positive correlation between Thematic

Hierarchy position and accuracy, with higher accuracy for more prominent event roles. I

hypothesized that there would be a negative correlation between Thematic Hierarchy position and

reaction time, with lower (or faster) reaction times for more prominent event roles.

If the event roles identified as incongruent most accurately and quickly were those event

roles near the top of the Thematic Hierarchy, that would suggest the Thematic Hierarchy is a good

indicator of event role prominence. If the Thematic Hierarchy proves to be a good indicator of

which event roles elicit the fastest or most accurate responses in the Event Task, then it might

suggest that aphasia treatment could be adjusted to focus on improving comprehension of those

event role categories that PWA respond to inaccurately or slowly. If the predicted relationship

between the Thematic Hierarchy and Event Task performance is not found, that would suggest

that the Thematic Hierarchy does not guide event roles in relation to PWA.

8

2.0 Methods

2.1 Participants

As reported in Dresang et al. (2019), Event Task data was collected from 208 participants,

including 26 PWA and 182 neurologically healthy participants across the lifespan. These healthy

participants were divided into groups of 15 in the following age ranges: 20-29, 30-39, 40-49, 50-

59, 60-69, 70-79. The healthy control groups allowed for Event Task performance analysis across

the lifespan. All healthy participants were required to have no history of language, speech, or

neurological impairments. All participants attained a score of >24 on the Mini-Mental State

Examination, an assessment of cognitive function (MMSE) (Folstein et al., 1975). All participants

were monolingual native English speakers with normal or corrected-to-normal hearing and vision.

All participants provided informed consent and received compensation for their participation in

the study (Dresang et al., 2019). 4 participants were found to have incomplete Event Task data,

resulting in exclusion from the analysis.

2.2 Event Task Stimuli and Procedure

The Event Task is composed of 260 colored images, including 256 experimental images

and 4 practice images (Dresang et al., 2019). The photographs depict people engaging in basic

actions in a variety of environments. Half of these images portray plausible events, such as a person

boxing at a gym, and the other half portray implausible events, such as a man playing the violin

9

underwater. Implausible events are impossible or unlikely, and do not match our everyday

experience or our world knowledge based in semantic memory. Implausible event images were

characterized as stimuli that “violates so-called ‘world knowledge’ about typical human actions in

ecological environments” (Proverbio & Riva, 2009). In each implausible image, at least one event

role is incongruent with the event depicted. Incongruent event roles conflict with an individual’s

semantic memory for typical events of this type. Participants were shown one image at a time and

were asked to indicate if this scene was something that may normally happen by pressing a button

on the keyboard—the key labeled “1” indicated that the event “might normally happen” (a

plausible event) and the key labeled “5” indicated that the event “might not normally happen” (an

implausible event). Participants were instructed to respond as accurately and as quickly as possible

(Dresang et al., 2019). 

2.3 Data Compilation

I imported the Event Task data from E-Prime to Excel, and then consolidated the data into

one Excel file. The final Event Task Dataset included information such as participant code, trial

number, correct answer, accuracy, and reaction time. All participant demographic information was

manually transferred from printed files into an Excel file, including date of birth, education, Mini-

Mental State Examination score, and participant code. Each PWA was matched to a neurologically

healthy participant with the same age and years of education. The dataset consisting of PWA and

matched healthy participants was used to compare event role accuracy and reaction time between

the two groups.

10

2.4 Event Taxonomy Questionnaire

In order to evaluate which event role may have caused the incongruence in each

implausible image, referred to as an event role violation, we used data from an Event Taxonomy

Questionnaire. In the questionnaire, 37 volunteer participants were given event role definitions

and were asked to indicate which event roles created the incongruency in the 130 implausible

stimuli images from the Event Task. For example, in the image of a nail being hammered with a

shoe, the implausible event role would be the instrument. 5 practice images were given with

explanations as to which event role was violated to ensure comprehension of the instructions. The

data consisted of 33 participants’ responses, as 4 participants were excluded due to incomplete

questionnaires.

The individual Event Taxonomy Questionnaire responses were compiled into one Excel

file. The number of ratings for each event role dimension per stimuli image was totaled. The mean

number of rated violations for each event role dimension across all incongruent stimuli was

calculated. To be considered a canonical event role violation (for any given stimuli image), the

number of ratings had to be 2 SD above the mean for that dimension. For example, if a stimuli

image’s number of agent violation ratings was 2 SD above the mean number of agent violation

ratings across all incongruent stimuli, that stimuli image would be classified as having a canonical

agent violation. To also be considered an exclusive event role violation, the stimuli image must

have canonical ratings for only one event role dimension. The data was then converted into binary

data (“1” for an event role violation and “0” for lack thereof) which was imported to the Event

Task Dataset. This data indicated which event role was violated per stimuli image per participant

trial, allowing thematic hierarchy data analysis to occur.

11

2.5 Data Analysis

The Event Taxonomy Questionnaire data was summarized via pivot table, presenting the

number of canonical and exclusive event role violations in the incongruent stimuli set.

Table 1: Canonical and Exclusive Event Role Violations

Agent Patient Instrument Location

Canonical Event Role Violations 25 23 18 29

Exclusive Event Role Violations 16 19 13 23

Exclusive event role violations were used to generate the ordered factor Thematic

Hierarchy variable. Using exclusive violations ensured that stimuli images each have one clear

event role violation, which made it possible to map that violation to a single position on the

Thematic Hierarchy. Data analysis was conducted via R using a regression analysis including both

linear mixed effects models and generalized linear mixed effects models.

12

3.0 Results

3.1 PWA and Matched Controls Analyses

An initial generalized linear mixed-effects model compared accuracy at detecting

violations of four different event roles across PWA and healthy control participants who were

matched to the PWA by age and years of education. All analyses were conducted using only

incongruent stimuli. Model glmer(EventAccuracy ~ Group*ThemHierarchy + (1 +

ThemHierarchy|Subject) + (1|Stimulus)) compared accuracy performance among event role

violation types across both groups, while accounting for random effects across subjects and items.

Agent violations were coded as 1 (i.e., the reference level) due to agent holding the highest position

on the Thematic Hierarchy. Because agents are at the top of the Thematic Hierarchy, we expected

them to yield the highest accuracy and lowest reaction time. Both groups’ average accuracy

performance per event role violation type is depicted in Figure 1. Mean accuracy was relatively

high for the agent and patient violations, at around 86%, and not much lower for the location

violations. Instrument violations had the lowest accuracy. However, none of these differences were

reliable (Agent vs. Instrument: Estimate = -0.75798, SE = 0.44675, p value = -0.0898), (Agent vs.

Patient: Estimate = 0.01634, SE = 0.40747, p value = 0.9680), and (Agent vs. Location: Estimate

= -0.21022, SE = 0.39464, p value = 0.5943). There was no reliable effect of Group (Estimate =

0.24334, SE = 0.34004, p value = 0.4742) and no reliable interaction between Group and patient

(Estimate = 0.14145, SE = 0.30828, p value = 0.6463), instrument (Estimate = 0.53204, SE =

0.34735, p value = 0.1256), or location (Estimate = 0.28329, SE = 0.30130, p value = 0.3471).

13

Figure 1: PWA and matched control participants' accuracy performance.

A parallel analysis of reaction time included only trials on which participants answered

correctly. We analyzed the time it took for participants to detect event role violations using the

following model: lmer(RT ~ Group*ThemHierarchy + (1 + ThemHierarchy|Subject) +

(1|Stimulus)). Figure 2 displays both groups’ average reaction time performance per event role

violation type. Patient violations yielded the highest reaction times in both groups. In PWA, patient

violation reaction times were about 200 to 300 ms slower than the average for agent, instrument,

and location violations, which ranged from about 1950 ms to 2050 ms. In the healthy control group,

patient violation reaction times were about 100 to 200 ms slower than the average for agent,

instrument, and location violations, which ranged from about 1720 to 1820 ms. However, none of

these differences were reliable (Agent vs. patient: Estimate = 126.01, SE = 104.99, t value =1.200),

(Agent vs. instrument: Estimate = 29.25, SE = 114.58, t value = 0.255), and (Agent vs. location:

Estimate = -172.57, SE = 100.63, t value = -1.715). Group had a reliable effect reaction time

(Estimate = -388.502, SE = 158.553, t value = -2.450). No reliable interaction was found between

0.6

0.65

0.7

0.75

0.8

0.85

0.9

Agent Patient Instrument Location

A cc

ur ac

y

Event Role Violations

PWA and Matched Healthy Controls: Accuracy

PWA

HC

14

Group and patient (Estimate = -34.384, SE = 74.984, t value = -0.459), instrument (Estimate =

33.311, SE = 77.879, p value = 0.428), or location (Estimate = -8.582, SE = 69.416, p value =

-0.124).

Figure 2: PWA and matched control participants’ reaction time performance.

3.2 PWA Analyses

The next set of analyses were identical to the ones reported in the previous section, except

that they only included data from PWA. These analyses aim to investigate whether the Thematic

Hierarchy guides PWA accuracy and reaction time performance on the Event Task, examining

individual PWA performance in more depth than the previous analysis. Model

glmer(EventAccuracy ~ ThemHierarchy + (1 + ThemHierarchy|Subject) + (1|Stimulus))

compared accuracy performance among event role violation types in the PWA sample. As depicted

500

700

900

1100

1300

1500

1700

1900

2100

2300

2500

Agent Patient Instrument Location

R ea

ct io

n T

im e

(M S)

Event Role Violations

PWA and Matched Healthy Controls: Reaction Time

PWA

HC

15

in Figure 1, mean accuracy was approximately 85% for agent and patient violations with location

violations following close behind, averaging 83%. Instrument violations had the lowest accuracy,

averaging 76%. None of these differences were reliable (Agent vs Instrument: Estimate = -

0.75384, SE = 0.40886, p value = 0.0652), (Agent vs. Patient: Estimate = 0.01364, SE = 0.37801,

p value =0.9712), and (Agent vs. Location: Estimate = -0.16820, SE = 0.36538, p value = 0.6453).

In contrast to the mean accuracy across all PWA per event role violation type reported in

Figure 1, Figure 3 shows an in-depth description of accuracy performance per individual PWA.

As depicted in Figure 3, 19 of 26 PWA displayed the lowest accuracy performance on instrument

violations. This unexpected result was identified in the majority of PWA’s accuracy performance,

which is particularly interesting considering the higher accuracy average of location violations,

which are ranked lower on the Thematic Hierarchy than instruments and therefore were expected

to display the lowest accuracy performance. One PWA displayed the expected pattern: agent

violations being most accurate, followed by patient violations, instrument violations, and, finally,

location violations. The six PWA who were furthest from the expected pattern displayed unique

accuracy performance patterns which did not align with the expected pattern or common

performance across all PWA. For example, participant 218 had the highest accuracy for instrument

violations and the lowest accuracy for agent violations, which is far from both the expected pattern

and the performance of most PWA in this sample. The PWA that were furthest from the expected

pattern exemplify some of the variability seen across PWA accuracy performance.

16

Figure 3: Individual PWA accuracy performance per event role violation type.

Model lmer(RT ~ ThemHierarchy + (1 + ThemHierarchy|Subject) + (1|Stimulus))

compared reaction time performance among event role violation type in the PWA sample. Mean

reaction time was relatively stable at around 2000 ms across agent, instrument, and location

violations. Patient violations had the highest reaction time, averaging 2245 ms. However, none of

these differences were reliable (Agent vs. Patient: Estimate = 198.706, SE = 123.046, t value =

1.615), (Agent vs. Instrument: Estimate = -24.131, SE = 133.104, t value = -0.181), and (Agent

vs. Location: Estimate = -33.977, SE = 115.503, t value = -0.294).

In contrast to the mean reaction times for PWA per event role violation type provided in

Figure 2, Figure 4 provides reaction time performance for individual PWA across all event role

violation types. 18 of 26 PWA displayed the slowest reaction time for patient violations. None of

the PWA exhibited the expected pattern of reaction time performance. One PWA was identified

as being furthest from the expected pattern: participant 222 displayed the highest reaction time for

agent violations, followed by instrument, patient, then location violations. This performance

17

pattern did not align with the expected pattern or common performance across all PWA. Although

participant 222 exhibits a particularly unique pattern, reaction time variability is seen across all

PWA.

Figure 4: Individual PWA reaction time performance per event role violation type

3.3 Lifespan Analyses

The following analyses were conducted using Event Task data from 170 neurologically

healthy adults across the lifespan, utilizing the incongruent Event Task trials to assess accuracy

and reaction time in relation to the Thematic Hierarchy. These analyses investigate whether the

Thematic Hierarchy guides Event Task performance on neurologically healthy adults across the

lifespan. Additionally, these analyses include demographic information including age, MMSE

score, and years of education to examine what other factors impact Event Task accuracy and

reaction time performance.

500

1000

1500

2000

2500

3000

3500

4000

4500

201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 227

R ea

ct io

n T

im e

(M S

)

PWA

PWA Individual Reaction Time Scores

Agent Patient

Instrument Location

18

3.3.1 Age Analyses

A generalized linear mixed effects model investigated the relationship between accuracy,

age, and the Thematic Hierarchy. The Thematic Hierarchy variable was coded with the R function

as.integer, meaning that each event role violation type corresponds with an ascending number in

accordance with the Thematic Hierarchy rankings. Agent, patient, instrument, and location

violations are coded as 1, 2, 3, and 4 respectively. Model glmer(EventAccuracy ~

Age*ThemHierarchy + (1 + ThemHierarchy|Subject) + (1|Stimulus)) indicated that age has a

reliable effect on accuracy (Estimate = -0.015105, SE = 0.007156, p value = 0.0348). Figure 5

provides average accuracy per event role violation type across the lifespan. As age increased,

accuracy decreased on all event role violation types. The Thematic Hierarchy did not have a

reliable effect on accuracy (Estimate = 0.028055, SE = 0.146167, p value = 0.8478). There was no

reliable interaction found between age and Thematic Hierarchy (Estimate = -0.002082, SE =

0.001851, p value = -0.2608).

Figure 5: Healthy participants’ accuracy performance in relation to age.

0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

15 25 35 45 55 65 75 85

A cc

ur ac

y

Age in Years

Neurologically Healthy Participants: Accuracy and Age

Agent Patient Instrument Location Linear (Agent) Linear (Patient) Linear (Instrument) Linear (Location)

19

A linear mixed effects model assessed age and the Thematic Hierarchy in relation to

reaction time. Model lmer((RT ~ Age*ThemHierarchy + (1 + ThemHierarchy|Subject) +

(1|Stimulus)) indicated that age had a reliable effect on reaction time (Estimate = 7.161, SE =

1.932, t value = 3.707). Figure 6 displays average reaction time performance per event role

violation type across the lifespan. As age increases, reaction time increases. Thematic Hierarchy

did not have a reliable effect on reaction time (Estimate = 29.83, SE = 29.09, t value = 1.026).

There was no reliable interaction found between age and Thematic Hierarchy (Estimate = 0.013,

SE = 0.2705, t value = 0.048).

Figure 6: Healthy participants’ reaction time performance in relation to age.

500

1000

1500

2000

2500

3000

3500

4000

15 25 35 45 55 65 75 85

R ea

ct io

n T

im e

(M S

)

Age in Years

Neurologically Healthy Participants: Reaction Time and Age

Agent Patient Instrument Location

Linear (Agent) Linear (Patient) Linear (Instrument) Linear (Location)

20

3.3.2 Mini-Mental State Examination Analyses

Model glmer(EventAccuracy ~ MMSE*ThemHierarchy + (1 + ThemHierarchy|Subject) +

(1|Stimulus)) assessed the relationship between MMSE Score and the Thematic Hierarchy in

relation to accuracy. Figure 7 provides average accuracy performance per event role violation type

in relation to MMSE score. MMSE Score did not have a reliable effect on accuracy performance

(Estimate = -0.0001468, SE = 0.1550574, p value = 0.999). Similarly, the Thematic Hierarchy did

not have a reliable effect on accuracy (Estimate = -1.9259424, SE = 1.2819478, p value = 0.133).

There was not a reliable interaction found between MMSE score and the Thematic Hierarchy

(Estimate = 0.0617778, SE = 0.0433683, p value = 0.154).

Figure 7: Healthy participants’ accuracy performance in relation to MMSE Score.

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

25 26 27 28 29 30

A cc

ur ac

y

MMSE Score

Neurologically Healthy Participants: Accuracy and MMSE Score

Agent

Patient

Instrument

Location

21

A linear mixed effects model examined MMSE score and Thematic Hierarchy in relation

to reaction time. Model lmer(RT ~ MMSE*ThemHierarchy + (1 + ThemHierarchy|Subject) +

(1|Stimulus)) revealed that participants with an MMSE score of 25 had higher reaction times than

participants with MMSE scores of 26-28. Figure 8 provides average reaction time performance

per event role violation type by MMSE score. MMSE score had a reliable effect on reaction time

(Estimate = -88.520, SE = 43.632, t value = -2.029). Thematic Hierarchy did not have a reliable

effect on reaction time (Estimate = -12.306, SE = 186.204, t value = -0.066). There was not a

reliable interaction between MMSE and Thematic Hierarchy reaction time (Estimate = 1.543, SE

= 6.245, t value = 0.247).

Figure 8: Healthy participants’ reaction time performance in relation to MMSE Score.

600

800

1000

1200

1400

1600

1800

2000

2200

2400

25 26 27 28 29 29.5 30

R ea

ct io

n T

im e

(M S

)

MMSE Score

Neurologically Healthy Participants: Reaction Time and MMSE Score

Agent

Patient

Instrument

Location

22

3.3.3 Years of Education Analyses

A generalized linear mixed effects model examined years of education and Thematic

Hierarchy in relation to accuracy. Figure 9 depicts average accuracy performance per event role

violation type in relation to years of education. Model glmer(EventAccuracy ~

YoE*ThemHierarchy + (1 + ThemHierarchy|Subject) + (1|Stimulus)) indicated that participants

with 11 years of education were less accurate on all event role violation types than participants

with 12-22 years of education. Neither years of education (Estimate = 0.023906, SE = 0.059237,

p value = 0.68653) nor Thematic Hierarchy (Estimate = -0.009782, SE = 0.267561, p value =

0.97083) had a reliable effect on accuracy. Additionally, there was no interaction between years

of education and Thematic Hierarchy (Estimate = -0.006824, SE = 0.015047, p value = 0.65016).

Figure 9: Healthy participants’ accuracy performance in relation to years of education.

0

0.2

0.4

0.6

0.8

1

1.2

11 12 13 14 15 16 17 18 19 20 21 22

A cc

ur ac

y

Years of Education

Neurologically Healthy Participants: Accuracy and Years of Education

Agent

Patient

Instrument

Location

23

A linear mixed effects model analyzed how years of education and the Thematic Hierarchy

predict reaction time. Figure 10 depicts average reaction time performance per event role violation

type in relation to years of education. Model lmer(RT ~ YoE*ThemHierarchy + (1 +

ThemHierarchy|Subject) + (1|Stimulus)) found that participants with 11 years of education had

higher reaction times than participants with 12-22 years of education. However, years of education

did not have a reliable effect on reaction time (Estimate = -3.1859, SE = 17.4183, t value = -0.183).

The Thematic Hierarchy did not have a reliable effect on reaction time (Estimate = 42.8134, SE =

46.1222, t value = 0.928). Years of education and Thematic Hierarchy did not have a reliable

interaction (Estimate = -0.5739, SE = 2.3879, t value = -0.240).

Figure 10: Healthy participants’ reaction time performance in relation to years of education.

600

1100

1600

2100

2600

3100

11 12 13 14 15 16 17 18 19 20 21 22

R ea

ct io

n T

im e

(M S

)

Years of Education

Neurologically Healthy Participants: Reaction Time and Years of Education

Agent

Patient

Instrument

Location

24

4.0 Discussion

This paper reports new analyses examining accuracy and reaction time performance of

PWA and healthy control participants on incongruent trials of the Event Task. In the task,

participants saw photographs of events that are either plausible or implausible and were asked to

indicate if the scene depicted is something that may normally happen. Each incongruent stimuli

image had at least one event role violation, and accuracy and reaction time performance were

analyzed per event role violation type.

I predicted violations that involved event roles from higher on the Thematic Hierarchy

would be detected more accurately and more quickly. For example, agent violations were expected

to be detected faster and more accurately, whereas location violations were expected to be detected

less accurately and more slowly. This prediction was based on experimental evidence from the

literature indicating that information about highly prominent event roles, such as agent and patient,

is spontaneously encoded even when it is not directly relevant to the task (Hafri et al., 2018). In

contrast, event roles at the bottom of the Thematic Hierarchy receive less attention: common

instruments are often omitted from story retellings (Lockridge and Brennan, 2002) and locations

were the only event role not primed in a single-word priming task (Ferretti et al., 2001).

25

4.1 PWA and Matched Controls

Analyses of the current dataset did not yield any reliable differences between PWA and

healthy controls; however, an interesting numerical pattern of responses was revealed. Both PWA

and the healthy control participants displayed the same pattern: patient violations yielded the

highest accuracy, followed by agents, locations, and instruments. This is inconsistent with the

original hypothesis that agent violations would be most accurate, followed by patient violations,

instrument violations, and location violations. On average across both groups, instrument

violations were less accurate than other event role violations. In the healthy control group, patient,

agent, and location violations were on average between 87% to 88% accurate while instrument

violations averaged only 82% accuracy. Similarly, PWA were approximately 83% to 85% accurate

on agent, patient, and location violations, while instrument violations were only 76% accurate on

average. The majority of PWA were the least accurate in identifying instrument violations in

comparison to other event role violation types on the Event Task incongruent trials.

Both PWA and matched healthy participants reacted fastest to location violations, followed

by instrument violations, then agent violations, and finally, patient violations. This finding is

inconsistent with the hypothesis that both groups would have the lowest reaction time for agent

violations, followed by patient violations, instrument violations, then locations violations. These

findings indicate that the Thematic Hierarchy does not guide either group’s accuracy or reaction

time performance on the Event Task. This finding leads to yet another question: if not the Thematic

Hierarchy, what is guiding their performance?

Both groups displayed the same pattern of accuracy and reaction time performance, most

notably the low instrument event role violation accuracy. One possibility is that the size of the

event role in the stimuli image affects accuracy and reaction time performance on the Event Task.

26

Instruments are often smaller than other event roles, such as agents and locations. Perhaps event

roles which take up more pixels of the image, like locations, are easier to identify accurately and

quickly in comparison to event roles which are often smaller and take up fewer pixels, such as

instruments.

This finding of relatively poor detection of atypical instruments seems to potentially

contrast with Lockridge and Brennan’s (2002) finding that atypical instruments were often

explicitly mentioned early in story retellings, whereas typical instruments were often omitted,

suggesting that incongruent or atypical instruments receive more attention than inferable

instruments. . However, their study compared typical instruments to atypical instruments, focusing

solely on one event role category; in the current study, atypical instruments were compared to

atypical agents, patients, and locations. While the Lockridge and Brennan experiment may speak

to the prominence of incongruent instruments in comparison to common instruments, our study

found that instrument violation accuracy was lower than all other event role violation types,

indicating that instruments are not as prominent as agents, patients, or locations.

Additionally, Lockridge and Brennan (2002) examined instrument prominence in story

retelling, an act reliant on language comprehension and production. Participants were told the story

and then asked to retell it, allowing ample time for story comprehension and the language

production necessary for retelling. However, when making plausibility judgements in the Event

Task, the participant relies on semantic memory rather than language processing. It might be the

case that atypical instruments are prominent enough to receive more attention in language

processing, aligning with their position above locations on the Thematic Hierarchy, but perhaps

are not prominent enough for quick encoding of information during split-second judgements, such

as the plausibility judgement in the Event Task.

27

4.2 Neurologically Healthy Adults Across the Lifespan

Across the neurologically healthy adult lifespan analyses, Thematic Hierarchy did not have

a significant effect on accuracy or reaction time. This finding is similar to the findings for PWA

and matched controls in showing that the Thematic Hierarchy does not guide Event Task

performance on incongruent stimuli items. Age had a significant effect on accuracy in

neurologically healthy adults; increased age was correlated with lower accuracy on the Event Task.

Additionally, age had a significant effect on reaction time, indicating that as an individual increases

in age, their reaction time gradually slows.

One potential mechanism behind this age-related decline in performance on the Event Task

may be inefficiencies in older adults’ semantic networks. Semantic networks are the structural

organization of mental representations which portray semantic relations in a fashion similar to that

of a map. An efficient semantic network would be flexible with strong connectivity and little space

between concepts on the map. A study conducted by Cosgrove et al. (2021) examined the effects

of aging on semantic networks. They predicted that older adults would display semantic networks

with less flexibility, indicating weak connectivity and larger distance between nodes. In their

study, neurologically healthy older and younger adults participate in a categorical verbal fluency

task. This task measures the ability to recall semantic knowledge from long-term memory, and

results from this type of task are often utilized for mapping semantic networks. This data was

examined using a percolation analysis to create visualized semantic networks. Findings indicated

that older adults’ semantic networks were less efficient than those of younger adults (Cosgrove et

al., 2021).

In theory, older adults’ inefficient semantic networks may hinder performance on a variety

of semantic memory tasks, including plausibility judgements such as the Event Task trials. I

28

believe that an efficient semantic network would lead to accurate, fast responses. For example, a

younger person with efficient semantic networks may see the Event Task stimulus image of a man

playing the violin underwater and quickly identify the incongruent location. Some concepts in a

semantic network for the violinist could be a concert hall, orchestra, cello, or musician. However,

the concepts of ocean and violinist would not be close to each other on a semantic network,

informing the participant that the location must be incongruent. I predict that an older individual

with an inefficient semantic network could take longer to conclude that the stimulus is incongruent

because their semantic network has more distance between concepts, making the judgement a

longer and more strenuous process. I think it is possible that as age increases, performance on

semantic plausibility judgements worsens due to inefficient semantic networks.

I also investigated potential relationships between performance on the Mini State Mental

Exam (MMSE), an assessment of cognitive function (Folstein et al., 1975), and performance on

detecting violations in the Event Task. The MMSE assesses skills relevant to orientation (i.e.,

knowledge of what year it is), attention, recall, and language. It seemed likely that a person with a

lower MMSE score might struggle to attend to the Event Task trials, resulting in low accuracy or

high reaction time. In my analyses, MMSE score was not related to accuracy or reaction time

performance. However, participants with a MMSE score of 25 had an average reaction time of

2097 MS. Participants with a MMSE score ranging from 26-30 had an average reaction time of

1570 MS. In summary, participants with a MMSE score of 25 were approximately 527 MS slower

to respond than participants with a MMSE score ranging from 26-30. MMSE scores classified as

“normal” range from 25-30 (Folstein et al., 1975). The slower reaction time of participants with a

25 MMSE score could be indicative of their lower cognitive function, as their score is at the bottom

of the normal range.

29

Years of education also did not have a significant effect on accuracy or reaction time

performance. However, participants with 11 years of education were around 60% accurate for most

event role violation types, while participants with 12-22 years of education were approximately

90% accurate on most event role violation types. It seems that having 12 years of education

increases accuracy performance by nearly 30% in comparison to those with 11 years of education.

This finding indicates that having a high school degree is correlated with increased accuracy on

the Event Task. Although years of education did not significantly affect reaction time, participants

with only 11 years of education displayed remarkably higher reaction time performance.

Participants with 11 years of education reacted in 2924 milliseconds (ms) on average. Participants

with 12 years of education had a slightly faster reaction time of 1811 ms. Participants with 13-22

years of education displayed a consistently faster reaction time around 1496 ms. Participants with

11 years of education were approximately 1428 ms slower to respond to Event Task trials than

participants with 13-22 years of education. This finding reinforces the interpretation that attaining

a high school education level may improve performance on the Event Task in relation to both

accuracy and reaction time.

One potential explanation for this distinction between 11 years of education and 12+ years

of education is that high school coursework may impact the richness of one’s semantic memory

representations. High school students are exposed to a wide variety of coursework which can

provide new insights about the world, adding to their ever-growing semantic memory. It’s

plausible that high school classes introduce information which adds to their semantic memory and

assists in making plausibility judgements. Completing a high school education may be critical in

adding important general knowledge to one’s semantic memory during a person’s formative years.

For example, exposure to broader subject matter through various courses could create more

30

efficient semantic networks featuring a wide range of concepts, leading to more accurate and faster

responses to plausibility judgements such as the Event Task trials.

Additionally, high school coursework is designed to improve critical thinking and

analytical skills. High school assignments, such as analyzing passages in a literature class or

solving complex open-response problems in math class, allow students to practice critical thinking

on a daily basis, ultimately resulting in improved analytical skills. The Event Task plausibility task

is similar to the theoretical questions high school students are often asked to answer in classes; a

question is posed, and students use problem solving skills and general knowledge to come to the

best possible answer. It’s possible that the completion of a high school education leads to improved

analytical skills, resulting in better performance on the Event Task in comparison to individuals

without a completed high school education.

4.3 Clinical Application

With this research comes the opportunity for future clinical advancements. Verb Network

Strengthening Treatment (VNeST) is an aphasia treatment focused on ameliorating generalized

word retrieval (both noun and verb retrieval) in PWA (Edmonds et al., 2014). I believe that

implementing the Event Task into VNeST protocol could lead a better understanding of how

aphasia treatment mitigates event knowledge deficits. Clinicians could utilize the Event Task

before and after VNeST to evaluate potential accuracy and reaction time performance

improvement, acting as a measure of VNeST’s success rates. Additionally, initial Event Task

performance could help clinicians identify areas of difficulty (such as specific event role deficits)

which could be focused on in treatment. Implementing event role analysis into VNeST could

31

provide useful insight into the ways that PWA process and store event role information, leading to

novel clinical applications of VNeST. In the VNeST protocol, the clinician cues the PWA to

generate agents and patients as a way of activating their semantic networks, followed by

encouraging the expansion of schemas (i.e., where, when, and why is the agent doing that action?).

PWA then attempt to independently produce a verb relevant to the event. Finally, the protocol is

repeated with the omission of the clinician cues to promote independence. An additional step

includes asking the PWA to produce a subject-verb-object sentence describing an event image.

Future research could entail adding eye tracking to this step to examine the order in which

PWA look at each event role and the duration of each glance. I would be curious to see if PWA

would display a similar pattern of results as to those found by Griffin and Bock (2000). Their eye-

tracking data indicated that neurologically healthy adults gazed at the agent before speaking, then

glanced at the patient as they began speaking, and concluded by looking to the agent once again.

Adding an eye tracking step to VNeST treatment would allow us to examine how/if the order of

event role gazing or the duration of gaze changes over the course of treatment. This would expand

upon the previous experiment by Griffin and Bock (2000) by tracking PWAs’ gazing patterns in

relation to all four event role types, whereas the previous study only examined neurologically

healthy adults’ eye movements for simplistic images featuring only agents and patients. I would

predict that, in the beginning phases of treatment, PWA may look between the event roles multiple

times prior to speech, spending more time gazing at any event role with which they are struggling

with word retrieval. I think that PWA would spend the most time gazing at agents and patients.

Additionally, it’s possible that PWA may spend less time glancing at instruments, as the current

study’s results indicate that instruments may not be particularly salient. However, I would expect

32

that over the course of treatment, PWA may begin to show the same pattern as the neurologically

healthy participants in the Griffin and Bock (2000) experiment.

Another future direction could entail examining what factors influence generalized lexical

retrieval, specifically in relation to event roles and semantic relations. Semantic memory can be

divided into two distinct sectors: taxonomic and thematic knowledge. Taxonomic knowledge

consists of the relation between two entities founded in similarities such as common features (i.e.,

dog and cat), whereas thematic knowledge refers to the relationship between two entities which

commonly occur in the same events (i.e., dog and leash) (Mirman et al., 2017). Taxonomic

organization would separate entities by categories (such as event roles) while thematic

organization would distinguish those same entities by events in which they often co-occur (Mirman

et al., 2017).

I would expect that PWA would exhibit improved generalized lexical retrieval in response

to untrained words that are thematically related to trained words. For example, I predict that a

trained word such as “dog” might elicit generalized retrieval for untrained words such as “leash”

or “fur” which are thematically related. I hypothesize that generalized lexical retrieval may rely

heavily on the richness of one’s semantic networks. An efficient, dense semantic network would

have less distance between related concepts, which I predict would result in a higher success rate

of generalized word retrieval in PWA.

The distinction between thematic and taxonomic relations arises yet another question:

could lexical retrieval improvement of trained words result in generalized lexical retrieval across

untrained words of the same event role category? For example, perhaps a person trained on the

word “hammer” could easily retrieve the untrained word “baseball bat,” not because they are

thematically related, but because they fall into the same taxonomic category of “instrument” and

33

share the purpose of helping the agent to complete an action. I hypothesize that the success of

generalized lexical retrieval across event role types would be dependent on the strength of the

participant’s semantic networks, more specifically the distance between concepts which are

taxonomically related instead of thematically related. Efficient, taxonomically organized semantic

networks could result in generalized lexical retrieval across event role types.

Future research could examine semantic network efficiency and organization to gain a

better understanding of how semantic relations are represented in the mind, answering questions

such as: are there taxonomic semantic networks solely detailing the categorization of entities into

event roles? Opposingly, do semantic networks display both thematic and taxonomic relations on

the same semantic network? If so, I would expect that thematically related concepts would be

closer together on the map, whereas taxonomically related concepts would have more distance

between nodes. These questions can be addressed in future research delving into how event

representations are stored in an individual’s semantic memory.

To summarize, examining the gazing patterns of PWA during VNeST treatment could

provide useful insight into the ways that PWA process event role information before and during

speech. Considering taxonomic relations, thematic relations, and semantic network efficiency

could lead to novel findings describing the mechanisms behind generalized lexical retrieval in

PWA. In the future, the semantic memory research conducted in relation to the Thematic Hierarchy

and event knowledge could lead to the implementation of new and improved treatment methods

for those PWA.

34

4.4 Limitations

A limitation of this research is the variation across stimuli images. The Event Task dataset

consists of 256 stimuli images, including both congruent and incongruent stimuli. Among the 128

incongruent stimuli, there are examples of each event role violation type; however, the number of

stimuli portraying each event role violation type varies. For example, there were 23 location-

exclusive violation stimuli, whereas there were only 13 instrument-exclusive violation stimuli.

This severely limited the power of my analyses to detect effects. Additionally, the stimuli images

vary in how many event role violations are present. Some Event Task stimuli images include

multiple event role violations—for example, a single stimuli image may have both agent and

patient violations. Future studies could include a new stimuli set which would ensure that all event

role violation types are featured equally across stimuli images and each stimuli image contains

only one event role violation. Creating a specialized stimuli set featuring carefully selected event

role violations could lead to statistically stronger results in future work.

In addition to the variable number of items per event role violation type, the current study

featured only a small number of items. The low number of incongruent stimuli items led to a lack

of statistical power in our analyses, resulting in reduced ability to detect meaningful differences

between event role violation types. Creating a new stimuli set including equal and larger numbers

of items per event role violation type would help to address this limitation in future work.

35

5.0 Conclusion

The current study examined accuracy and reaction time performance on the Event Task in

relation to the Thematic Hierarchy. In summary, the Thematic Hierarchy does not seem to guide

Event Task performance in PWA or neurologically healthy adults across the lifespan. However,

PWA and neurologically healthy controls alike displayed the same patterns of both accuracy and

reaction time performance, indicating that an unknown force may be guiding their performance.

Most notably, participants displayed low accuracy performance on instrument violation

incongruent trials. This finding could be driven by event role size or lack of instrument prominence

when viewing the Event Task stimuli.

Although no results were found to be statistically reliable, interesting numerical patterns

emerged from the current study. In neurologically healthy adults, age had a reliable effect on

accuracy and reaction time performance, which may be influenced by the deterioration of semantic

networks associated with aging (Cosgrove et al., 2021). Additionally, neurologically healthy adults

with 12+ years of education were found to have better accuracy and reaction time performance

than adults with only 11 years of education, which may be indicative of how high school education

completion may impact critical thinking skills and semantic memory richness.

The current study provided novel findings which may be the foundation for future research.

Next steps would include creating an improved version of the Event Task featuring stimuli images

specifically designed to assess performance on each event role type. Future studies could examine

VNeST in relation to the event roles with the hopes of getting a better understanding of how PWA

process event roles while simultaneously improving treatment methods. Finally, next steps could

include studying the role of semantic networks in regard to aphasia treatment and generalized

36

lexical retrieval. Future research examining PWA and event knowledge is imperative for both the

scientific community and the clinical population.

37

Bibliography

Antonucci, S., & Reilly, J. (2008). Semantic Memory and Language Processing: A Primer. Seminars in Speech and Language, 29(1), 005–017. https://doi.org/10.1055/s-2008- 1061621

Bak, T. H., & Hodges, J. R. (2003). Kissing and dancing—A test to distinguish the lexical and conceptual contributions to noun/verb and action/object dissociation. Preliminary results in patients with frontotemporal dementia. Journal of Neurolinguistics, 16(2), 169–181. https://doi.org/10.1016/S0911-6044(02)00011-8

Baker, M.C. (1997). Thematic roles and syntactic structure. In L. Haegeman (Ed.) Handbook of generative syntax (pp. 73–137). Dordrecht: Kluwer.

Bozeat, S., Lambon Ralph, M. A., Patterson, K., Garrard, P., & Hodges, J. R. (2000). Non-verbal semantic impairment in semantic dementia. Neuropsychologia, 38(9), 1207–1215. https://doi.org/10.1016/S0028-3932(00)00034-8

Caramazza, A., & Zurif, E. B. (1976). Dissociation of algorithmic and heuristic processes in language comprehension: Evidence from aphasia. Brain and Language, 3(4), 572–582. https://doi.org/10.1016/0093-934x(76)90048-1

Cosgrove, A. L., Kenett, Y. N., Beaty, R. E., & Diaz, M. T. (2021). Quantifying flexibility in thought: The resiliency of semantic networks differs across the lifespan. Cognition, 211, 104631. https://doi.org/10.1016/j.cognition.2021.104631

Dresang, H. C., Dickey, M. W., & Warren, T. C. (2019). Semantic memory for objects, actions, and events: A novel test of event-related conceptual semantic knowledge. Cognitive Neuropsychology, 36(7–8), 313–335. https://doi.org/10.1080/02643294.2019.1656604

Edmonds, L., Mammino, K., & Ojeda, J. (2014). Effect of Verb Network Strengthening Treatment (VNeST) in Persons With Aphasia: Extension and Replication of Previous Findings. American Journal of Speech-Language Pathology / American Speech-Language-Hearing Association, 23. https://doi.org/10.1044/2014_AJSLP-13-0098

Ferretti, T. R., McRae, K., & Hatherell, A. (2001). Integrating Verbs, Situation Schemas, and Thematic Role Concepts. Journal of Memory and Language, 44(4), 516–547. https://doi.org/10.1006/jmla.2000.2728

Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). “Mini-mental state”: A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research, 12(3), 189–198. https://doi.org/10.1016/0022-3956(75)90026-6

38

Griffin, Z. M., & Bock, K. (2000). What the Eyes Say about Speaking. Psychological Science, 11(4), 274–279. https://doi.org/10.1111/1467-9280.00255

Hafri, A., Trueswell, J. C., & Strickland, B. (2018). Encoding of event roles from visual scenes is rapid, spontaneous, and interacts with higher-level visual processing. Cognition, 175, 36– 52. https://doi.org/10.1016/j.cognition.2018.02.011

Howard, D., Patterson, K., & Thames Valley Test Company. (1992). The pyramids and palm trees test: A test of semantic access from words and pictures. Thames Valley Test Company.

Jackendoff, R. (1990). Semantic structures. Cambridge, MA: MIT Press.

Lockridge, C. B., & Brennan, S. E. (2002). Addressees’ needs influence speakers’ early syntactic choices. Psychonomic Bulletin & Review, 9(3), 550–557. https://doi.org/10.3758/bf03196312

McRae, K., & Matsuki, K. (2009). People Use their Knowledge of Common Events to Understand Language, and Do So as Quickly as Possible. Language and Linguistics Compass, 3(6), 1417–1429. https://doi.org/10.1111/j.1749-818X.2009.00174.x

Mirman, D., Landrigan, J.-F., & Britt, A. E. (2017). Taxonomic and thematic semantic systems. Psychological Bulletin, 143(5), 499–520. https://doi.org/10.1037/bul0000092

Proverbio, A. M., & Riva, F. (2009). RP and N400 ERP components reflect semantic violations in visual processing of human actions. Neuroscience Letters, 459(3), 142–146. https://doi.org/10.1016/j.neulet.2009.05.012

Ünal, E., Richards, C., Trueswell, J. C., & Papafragou, A. (2021). Representing agents, patients, goals and instruments in causative events: A cross-linguistic investigation of early language and cognition. Developmental Science, 24(6), e13116. https://doi.org/10.1111/desc.13116