UNIVERSITY OF CALIFORNIA Los Angeles Measuring Linguistic

UNIVERSITY OF CALIFORNIA

Los Angeles

Measuring Linguistic Empathy:

An Experimental Approach to Connecting

Linguistic and Social Psychological Notions of Empathy

A dissertation submitted in partial satisfaction of the

requirements for the degree of Doctor of Philosophy

in Applied Linguistics

Trevor Kann

2017

Trevor Kann

2017

ABSTRACT OF THE DISSERTATION

Measuring Linguistic Empathy:

An Experimental Approach to Connecting

Linguistic and Social Psychological Notions of Empathy

Trevor Kann

Doctor of Philosophy in Applied Linguistics

University of California, Los Angeles, 2017

Professor Olga Tsuneko Yokoyama, Chair

This dissertation investigated the relationship between Linguistic Empathy and

Psychological Empathy by implementing a psycholinguistic experiment that measured a

person’s acceptability ratings of sentences with violations of Linguistic Empathy and

correlating them with a measure of the person’s Psychological Empathy. Linguistic

Empathy demonstrates a speaker’s attitude and identification with a person or event in an

utterance (Kuno & Kaburaki, 1977; Kuno, 1987; Silverstein, 1976; Yokoyama, 1986). This

identification is represented in the utterance through a speaker’s unconscious/automatic

selection from grammatically valid options that convey pragmatically different attitudes.

On the other hand, Psychological Empathy is a social psychological notion that allows a

person to understand and experience the emotional reality of others. The capacity for

Psychological Empathy is known to differ among individuals, and assessment of this

capacity is often implemented in a clinical setting to indicate those at risk of conditions

iii

with deficits of Empathy, such as Autism Spectrum Disorder. This study measured

Psychological Empathy with the Empathy Quotient test, which was designed for both

clinical and non-clinical settings (Baron-Cohen & Wheelwright, 2004).

This study extended the notion of Linguistic Empathy from a linguistic phenomenon

that is represented in speech to a measurable trait in an individual. The measure of

Linguistic Empathy specifies the individual’s capacity to notice and rate sentences that are

grammatically valid but contain unnatural violations of Linguistic Empathy phenomena

(e.g.,

I met Nancy

versus

Nancy met me

). The results of the experiment showed that

Linguistic Empathy is a measurable and systematic trait, and that it has a significant

positive correlation with Psychological Empathy. This correlation suggests that despite

their disparate theoretical origins, Linguistic and Psychological Empathy share a common

information processing component. One important clinical application of the results is to

use a test of Linguistic Empathy as an unbiased screen for individuals at risk for a deficit of

Psychological Empathy.

The dissertation of Trevor Kann is approved.

Eran Zaidel

Jesse Aron Harris

Olga Tsuneko Yokoyama, Committee Chair

University of California, Los Angeles

2017

To Jenni,

for challenging and inspiring me

desde Simon Dice

TABLE OF CONTENTS

CHAPTER 1

EMPATHY: AN INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.0 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 EMPATHY IN SOCIETY AND LANGUAGE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 GOALS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.2.1 MEASURING LINGUISTIC EMPATHY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.2.2 CORRELATIONS OF LINGUISTIC EMPATHY

AND PSYCHOLOGICAL EMPATHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.2.3 METHODOLOGICAL IMPACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.2.4 CONFIRMATIONS AND EXTENSIONS OF EMPATHY. . . . . . . . . . . . . . . . . . 7

1.3 SUMMARIES OF CHAPTERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

CHAPTER 2

PSYCHOLOGICAL AND LINGUISTIC NOTIONS OF EMPATHY . . . . . . . . . . . . . . . 10

2.0 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2.1 PSYCHOLOGICAL EMPATHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2.1.1 AFFECTIVE AND COGNITIVE EMPATHY . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.1.2 PSYCHOLOGICAL EMPATHY MEASURES . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.2 LINGUISTIC EMPATHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.2.1 EMPATHY PERSPECTIVE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.2.2 EMPATHY HIERARCHIES AND

THEIR PRAGMASEMANTIC CONNECTIONS . . . . . . . . . . . . . . . . . . . . . . . . 22

2.2.3 POINT OF VIEW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2.2.4 SENTENTIAL TOPICS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

2.2.5 PASSIVE CONSTRUCTIONS AND RECIPROCAL VERBS . . . . . . . . . . . . . . 37

2.2.6 UNIVERSAL AND INDIVIDUAL NOTIONS OF EMPATHY. . . . . . . . . . . . . . 40

2.3 EMPATHY IN THIS STUDY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

vii

CHAPTER 3

EXPERIMENTAL APPROACH TO EMPATHY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3.0 INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3.1 RATIONALE FOR EXPERIMENTATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3.1.1 MEASURE OF LINGUISTIC EMPATHY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3.1.2 MEASURES OF PSYCHOLOGICAL EMPATHY. . . . . . . . . . . . . . . . . . . . . . . . 51

3.1.3 BENEFITS OF LINGUISTIC EXPERIMENTS . . . . . . . . . . . . . . . . . . . . . . . . . 54

3.2 LINGUISTIC MANIPULATION OF THE STIMULI. . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3.2.1 OPERATIONALIZATION OF EMPATHY HIERARCHIES . . . . . . . . . . . . . . . 56

3.2.2 TARGET SENTENCES AND CONTEXT SENTENCES . . . . . . . . . . . . .. . . . . 58

3.2.3 CONDITIONS OF THE STIMULI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

3.3 PREDICTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

3.3.1 STIMULUS TYPES AND EMPATHY HIERARCHIES . . . . . . . . . . . . . . . . . . . 66

3.3.2 EXPERIMENT DESIGN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

3.3.3 SENSITIVITY TO LINGUISTIC EMPATHY . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

3.3.4 CORRELATIONS BETWEEN LINGUISTIC AND

PSYCHOLOGICAL EMPATHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

3.4 METHODOLOGY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

3.4.1 PARTICIPANTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

3.4.2 STIMULI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

3.4.3 PROCEDURE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

CHAPTER 4

EXPERIMENT: CORRELATIONS BETWEEN

LINGUISTIC AND PSYCHOLOGICAL EMPATHY. . . . . . . . . . . . . . . . . . . . . . . . . . . 78

4.0 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

4.1 RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

4.1.1 RECIPROCAL ITEMS AND EXPERIMENT VALIDATION . . . . . . . . . . . . . . . 78

viii

4.1.1.1 RECIPROCAL DESIGN

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

4.1.1.2 DEFINING CONDITIONS

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

4.1.1.3 LINGUISTIC EMPATHY SENSITIVITY AND

CORRELATIONS WITH PSYCHOLOGICAL EMPATHY (EQ)

. . . . . . 81

4.1.2 ACTIVE/PASSIVE ITEMS AND EXPERIMENT VALIDATION . . . . . . . . . . . 84

4.1.2.1 ACTIVE/PASSIVE DESIGN

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

4.1.2.2 DEFINING CONDITIONS

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

4.1.2.3 LINGUISTIC EMPATHY SENSITIVITY AND

CORRELATIONS WITH PSYCHOLOGICAL EMPATHY (EQ)

. . . . . . 87

4.1.3 COMPARISON ACROSS SENTENCE TYPES . . . . . . . . . . . . . . . . . . . . . . . . . 89

4.2 DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

4.2.1 DEFINING CONDITIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

4.2.2 DIFFERENCE MEASURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

4.2.3 LINGUISTIC EMPATHY MEASURE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

4.2.4 IMPLICATIONS FOR THE LINGUISTIC EMPATHY HIERARCHIES . . . . . 94

4.3 LIMITATIONS AND EXTENSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

CHAPTER 5

CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

5.0 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

5.1 METHODOLOGICAL CONTRIBUTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

5.2 MEASURE OF LINGUISTIC EMPATHY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

5.3 CONNECTION BETWEEN LINGUISTIC EMPATHY

AND PSYCHOLOGICAL EMPATHY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

5.4 ELECTROPHYSIOLOGICAL EXTENSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

5.5 UNDERLYING EMPATHY PROCESSING COMPONENT . . . . . . . . . . . . . . . . . . . . . . 106

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

LIST OF TABLES AND FIGURES

Figure 2.1: Kuno & Kaburaki’s “Empathy and Camera Angles” . . . . . . . . . . . . . . . 18

Table 3.1: Example Trials, Violations, and Conditions . . . . . . . . . . . . . . . . . . . . . . 66

Table 3.2: Reciprocal Stimulus Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

Table 3.3: Active/Passive Stimulus Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

Figure 3.1: Sample Trial of the Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

Figure 4.1: Mean Ratings, Reciprocal Items . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

Table 4.1: Sample Reciprocal Sentences for the Defining Conditions . . . . . . . . . . . 80

Table 4.2: Mean Ratings and Correlations, Reciprocal Items . . . . . . . . . . . . . . . . . 80

Table 4.3: Mean Ratings and Correlations, Reciprocal Difference Measures . . . . . 81

Figure 4.2: Correlation of EQ and Linguistic Empathy, Reciprocal Sentences . . . . 83

Figure 4.3: Mean Ratings, Active/Passive Items . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

Table 4.4: Sample Active/Passive Sentences for Defining Conditions . . . . . . . . . . . 85

Table 4.5: Mean Ratings and Correlations, Active/Passive Items . . . . . . . . . . . . . . 86

Table 4.6: Mean Ratings and Correlations, Active/Passive Difference Measures . . 86

Figure 4.4: Correlation of EQ and Linguistic Empathy, Active/Passive Sentences. . 88

LIST OF ACRONYMS AND SPECIAL TERMS

Language/Empathy Terms:

NP Noun Phrase

ASD Autism Spectrum Disorder

EQ Empathy Quotient

HES Hogan Empathy Scale

IRI Interpersonal Reactivity Index

QMEE Questionnaire Measure of Emotional Empathy

Rasch Analysis A method for measuring latent traits like attitude or ability

Statistics Terms:

ANOVA Analysis of Variance: statistical method for testing between

and among groups

SD Standard Deviation

M Mean

SE Standard Error: standard deviation of the sampling

distribution of a statistic

Mean of the within group variances

MSE

Mean Squared Error: risk function corresponding to expected

value

Probability value of obtaining a result (statistical significance)

Partial Eta Squared: default measure of effect size in ANOVAs

r Correlation Coefficient: measures how closely 2 variables are

t-test Assessment of whether 2 groups are statistically different

Fisher Z Test Converts correlations into a normally distributed measure

z Comparison of a test population with a normal population

∑ Sigma: the sum of the values that follow

ACKNOWLEDGMENTS

Although my dissertation research has focused on Empathy, what I have learned

about this topic was exponentially eclipsed by the insight I have gained from my loved ones

and from my committee. The Empathy that the people mentioned below have shown me is

the central reason that I have been able to complete this dissertation.

I must first and foremost thank those in my personal life for their love, support, and

patience throughout academic career. Jenni, the passion you exhibit toward everything you

undertake, no matter how daunting the task, is both humbling and inspiring. Since we

were thirteen, the intellectual and emotional fortitude with which you carry yourself have

challenged me to strive in ways I could have never imagined. Mom, you demonstrate more

Empathy than anyone I know, and I live my life hoping to show others the type of kindness

that comes so naturally to you. I owe my love for teaching and entertaining to you. Dad,

without the love you have shown and the sacrifices you have made, I would not have had

the opportunities that I have enjoyed. Also, I know for a fact that others are super grateful

to you for making me as sarcastic as I am. Dylan, without the high academic standards you

set for us, I would have never pursued academia. I could not imagine a funnier, more

thoughtful, and more loving brother (despite always making me turn out the light even

though I was already asleep). Jee, your love and counsel that began in my youth continue to

guide me through the challenges I face. Thank you for being the first to intrigue me with

and expose me to higher educational pursuits. Ardelle, I cannot thank you enough for your

tireless generosity throughout this process; your selflessness astounds me. Finally, I must

acknowledge Mike, who instilled in me a love for teaching and linguistics at an early age,

and always smothered his tutelage and our friendship in a healthy layer of wry humor. I

regret that I could not finish this dissertation in time for him to see its completion.

xii

I will forever be indebted for the guidance, compassion, and understanding that my

committee have exhibited. I am truly humbled by the unbelievable support they continue to

demonstrate through the difficult circumstances I have faced in this dissertation process.

Olga, you have continually exceeded the role of a mentor by providing me with layers upon

layers of support that I never knew I needed. Thanks to your mentorship, you have

motivated me to apply myself and to believe in myself, and you continue to model a

staggering level of dedication to both professional pursuits and personal convictions. I have

always felt encouraged and understood, and with this has come trust and respect for you as

an academic, an advocate, and a friend. Eran, I admire your passion for your work and for

your family, and I appreciate your ability to provoke your colleagues into meaningful

discussion and action. You personify an enviable combination of kindness and brilliance in

an advisor. Jesse, your sharp insight and thoughtful support are astounding, and your

adeptness as a younger academic is inspiring. Thank you also for providing a roof for the

experiments when we were suddenly without one. Finally, John, since I began my graduate

studies, you have embodied what I had hoped was true: that seriousness and levity are not

mutually exclusive in academia. To all of my advisors, I can only aspire to one day provide

mentorship for others with a fraction of effectiveness as you have shown me.

I would also like to acknowledge the generous contributions from the Dr. Ursula

Mandel Endowed Fellowship, which helps to fund research that endeavors to contribute to

the medical field, and to the UCLA Grad Division for the Dissertation Year Fellowship.

Additionally, I would like to thank the Department of Applied Linguistics for the Teaching

Assistance positions and other funding opportunities that have helped me become a more

effective educator and researcher.

Finally, I would also like to acknowledge my colleagues in the research team who

have contributed significantly to the development of this project. Portions of this

xiii

dissertation (primarily segments in chapters 3 and 4) overlap with a co-authored

manuscript that is currently in preparation (cited below). Olga Yokoyama and Eran Zaidel

are responsible for conceiving of and initiating the study, for assembling the wonderful

research team, and for contributing to the co-authored paper below. Additionally, I must

thank Michael Cohen for his brilliance in generating and analyzing the statistics in chapter

4. Despite the contributions from my co-authors, any mistakes in this dissertation are

entirely my own.

Kann, T., Yokoyama, O.T., Cohen, M.S., Gülser, M., Goldknopf, E., Berman, S., Erlikhman,

G., Zaidel, E. (2017).

Linguistic Empathy: Behavioral measures, physiological

correlates and correlation with Psychological Empathy.

Manuscript in preparation.

xiv

VITA

2004 B. A. Linguistics, Portuguese Minor

University of California, Los Angeles

2010 M. A. Applied Linguistics & TESOL

University of California, Los Angeles

PROFESSIONAL EMPLOYMENT

2014 - 2016 Graduate Research Assistant, Zaidel Psychology Lab

University of California, Los Angeles

Department of Psychology

Designed, oversaw, and conducted behavioral and neurophysiological

experiments in psycholinguistics.

2009 - 2016 Teaching Assistant/Teaching Associate/Teaching Fellow/Instructor

University of California, Los Angeles

Department of Applied Linguistics

Courses Taught

o Appling 10W: Language in Action

o Appling 30W: Language and Social Interaction

o Appling 40W: Language & Gender

o Appling 101W: Language Learning and Teaching

o Appling 102W: The Nature of Language

o Appling 144M: Fundamentals of Translation

o Appling 161W: Talk and the Body

Department of Linguistics

Course taught:

o Linguistics

1: The Study of Language

Department of Writing Programs

Courses taught:

o ESL 32: Academic Interaction

o ESL 33C: Academic Writing

o ESL 34: Public Speaking

o ESL 38B: Pronunciation

UCLA Anderson School of Business

International Student Summer Program

Speaking/Listening for Business English

2012 - 2013 Instructor of English as a Second Language

Santa Monica Community College

Department of English as a Second Language

Courses taught:

o 11B: Basic English II

o 21B: English Fundamentals II

AWARDS AND HONORS

2015-2016 Dissertation Year Fellowship, UCLA

Awarded on the basis of a thorough dissertation research proposal,

organization and initiative for preliminary findings, and prospective

contributions to the field.

2015-2016 Dr. Ursula Mandel Scholarship, Endowed Fellowship

Awarded to dissertation fellows in scientific fields whose research is of

value to the medical field.

2013 Celce-Murcia Outstanding Teaching Award, UCLA

Received by one student annually. Nominated and awarded by

professors in Applied Linguistics and the Social Sciences for excellence

in teaching performance, mastery of materials, student satisfaction,

and original contributions.

2010-2013 Scholarship, Applied Linguistics, UCLA

PRESENTATIONS

Yokoyama, O. T. & Kann, T. (2014).

EEG methods in pragmatics research

. Paper

presented at the Second Annual American Pragmatics Conference, UCLA, Los

Angeles, CA.

Kann, T. (May 2012).

The prosodical son: The influence of music and other modalities

on evolution of language.

Paper presented at the

11th Annual Conceptual

Structure, Discourse, and Language Conference University of British Columbia,

Vancouver, BC, Canada.

CHAPTER 1

EMPATHY: AN INTRODUCTION

Social psychology and linguistics share a term,

Empathy

, to designate two distinct

concepts:

Psychological Empathy

and

Linguistic Empathy

. Psychological Empathy is

described as the ability to “understand emotionally another’s feelings and experience”

(Ehrlich & Ornstein, 2010, p. 4) or as “the ability to see the world through others’ eyes so as

to sense their hurt and pain and to perceive the source of their feelings in the same way as

they do” (Watson, 2002, p. 446). In other words, Psychological Empathy registers a person’s

emotional reactions to others and the emotional understanding involved in these

experiences. Alternatively, Linguistic Empathy is a notion based on perspectives in

linguistic structure that capture a speaker’s attitude toward people and things that are

referenced in an utterance. As defined by Kuno (1987), Linguistic Empathy is “the speaker’s

identification, which may vary in degree, with a person/thing that participates in the event”

(p. 206). Linguistic Empathy is used to explain utterances that differ with respect to the

speaker’s perspective, yet defy standard grammatical explanation. For instance, the

utterances

I met someone last summer

and

Someone met me last summer

are logically and

structurally equivalent; however, the latter seems to be markedly less acceptable than the

former. Similarly, Linguistic Empathy is used to explain lexically correct options at a

speaker’s disposal (e.g., a mother asking her child

Did daddy call?

rather than

Did Terrance

call?

). Psychological Empathy and Linguistic Empathy both involve the understanding and

application of perspective taking; however, there is no consensus of how the two types of

Empathy are related. This study is designed to shed light on this question.

1.1 Empathy in Society and Language

The popular use of the term Empathy is often synonymous with compassion. When

people strive to show Empathy, they are attempting to understand the world based on

another’s circumstances and to convey this understanding through a compassionate

response. Although there may be individual and cultural nuances, the implementation of a

version of Empathy exists across cultures. Empathy is universally acknowledged as a

personal and social asset: It can serve as a window into morality (Baier, 1958; Hogan,

1969), it can benefit cross-cultural interaction, and it is a quality that parents endeavor to

impart to their children. Barack Obama has even stressed the personal importance of

learning about Empathy from his mother by using her “simple principle — ‘How would that

make you feel?’ — as a guidepost for [his] politics.” He has also argued that the United

States is “suffering from an empathy deficit” (Obama, 2006, p. 66-67). Ciarrochi et al. (2016)

cite the extensive prosocial benefits of Empathy, from having positive effects on

relationships and communication, to promoting conflict management and emotional

learning.

Although the arguments against Empathy as a prosocial behavior are few, the

promotion of Empathy does convey some inherent drawbacks. First, because Empathy is

highly valued socially and culturally, those who experience a deficit of Empathy on a

clinical level (e.g., Autism Spectrum Disorder, psychopathy, schizoid personality disorder)

can become disadvantaged and disenfranchised. Their lack of Empathy is considered

antisocial and becomes stigmatized. However, for them, a lack of Empathy is not a result of

complacence, selfishness, or cold calculation; it is not a choice. It is a condition that is

ingrained in their neurology, which can be addressed through medical and therapeutic

channels, if desired.

Bloom (2016) describes another drawback with respect to Empathy. He argues that

a paradoxical pitfall of Empathy involves a phenomenon where Empathy is more likely to

be triggered when exposed to the harrowing plight of an individual as opposed to the notion

of widespread human tragedy. Logically, the suffering of many should outweigh the

suffering of a single person. Nevertheless, providing personal detail about an individual’s

struggle provides a face to the tragedy, whereas a mass of people loses out on specificity and

poignancy. Bloom argues that Empathy is potentially a cultural downfall for this reason,

and instead, if humans attempted to exercise compassion with more impartiality and logic,

there would be less suffering. This argument suggests that when the experience of

another’s suffering is described as specific, real, and relatable, it becomes a more relevant

target of Empathy than suffering that may be more unjust and tragic. If the suffering

affects a larger population, it becomes more difficult to describe with precision an

experience that affects many. When this specificity is lost, the struggle becomes less

relevant and relatable.

Unfortunate circumstances can feel much more significant for a person when these

details are relevant to this person’s life. Yokoyama (1986) cites the example from the 1980s

of Americans coping with increased gasoline prices, versus the concurrent tragedy of food

shortages experienced by the Polish (p. 148). For most Americans, the issue of gasoline

prices was far more relevant and palpable than the distant issue of Poland’s food shortage,

despite the fact that coping with devastating health issues like starvation and malnutrition

exceeds the struggle of confronting more expensive gas. Relatedly, Yokoyama (1986, pp. 28

ff.) describes the linguistic notion of a

relevance requirement

, in which a message must

contain information that is relevant for the hearer. Yokoyama provides the examples of a

stranger who approaches someone at a train station and utters

you dropped your ticket

This utterance is considered relevant and valid to the hearer considering the importance of

a ticket in a train station. However, if the stranger approaches and utters

I’m cooking fish

tonight

, this would be irrelevant and invalid from the hearer’s perspective (Yokoyama,

1986, p. 28). When the

relevance requirement

is applied on to Empathy, it suggests that

when a struggle like starvation is distant (e.g., in Poland) or less specific (e.g., a mass of

people), then it is considered less relevant to those who learn of these issues. Linguistic

messages, like descriptions of suffering, become more relatable when the principle of

relevance is observed.

The term Empathy also remains problematic for linguistics. Because Linguistic

Empathy is a little known term outside of the discipline, the notion of Psychological

Empathy is the default for both the layman and academics outside of linguistics. In fact,

Linguistic Empathy is often avoided within the field of linguistics and overshadowed by

traditional notions of syntax. Linguistic Empathy takes a functional view of language

analysis in which grammar and pragmatics are interrelated, and the language participants

(e.g., speaker, author, hearer, addressee) are inseparable from the speech event. This is an

alternative approach to traditional syntax, which endeavors to establish and describe rules

of language by investigating linguistic structure. However, grammar on its own is not the

quintessence of language. Yokoyama (1986) argues that “grammar and pragmatics are

inseparable; they … verbalize both universally human (“phylogenic”) experience, as well as

narrowly personal (“ontogenic”) experience” (p. 186). Linguistic Empathy is a linguistic

phenomenon that is firmly planted in both grammar (e.g., subject/object alternation in

met someone last summer

versus

Someone met me last summer

) and pragmatics (e.g.,

speaker/hearer relationship that justifies the utterance

Did daddy call?

). By extension, it

considers both universal human tendencies of language, as well as situational

idiosyncrasies when applied to discourse.

1.2 Goals

1.2.1 Measuring Linguistic Empathy. As mentioned, Linguistic Empathy considers

both linguistic structure and individual circumstances when analyzing language behavior

in order to describe a speaker’s identification and point of view with respect to entities

within a sentence. When this linguistic phenomenon is quantified (e.g.,

Empathy or

less

Empathy), it refers to the extent to which a speaker shares the perspective with an

entity in a sentence. For instance, the utterance

Joey was hit by Katie

expresses more

Empathy with

Joey

than the utterance

Katie hit Joey

Katie hit her neighbor

(in which

her neighbor

and

Joey

are co-referential). Similarly, when Psychological Empathy is

quantified (e.g.,

Empathy or

less

Empathy), it can refer to a situation in which an

experiential perspective is shared more or less with others. For example, two people might

have different emotional responses upon learning that someone has broken an arm while

riding a unicycle: One might experience

Empathy by feeling upset with the person

and by expressing concern, while another might experience

less

Empathy and instead feel

that this person deserved to have an accident for riding a unicycle.

Unlike Linguistic Empathy, Psychological Empathy is also conceived of in terms of

an individual’s capacity to experience more or less Empathy. Various medical conditions,

such as Autism Spectrum Disorder, are related to deficits of Psychological Empathy.

However, Linguistic Empathy is not currently measured as an individual trait in the same

way that Psychological Empathy can be measured as a screening tool for deficits of

Empathy. Although difficulty in language acquisition and processing is one of the most

prominent symptoms for those who experience a Psychological Empathy deficit, Linguistic

Empathy is not currently conceived of as a quantifiable notion. A primary goal of this study

is to extend the theory of Linguistic Empathy from a notion that describes a speaker’s

identification within a speech event to include a measure of a person’s capacity for

experiencing Linguistic Empathy.

1.2.2 Correlations of Linguistic Empathy and Psychological Empathy. A driving

force in this study is to explore the correlations between Linguistic Empathy and

Psychological Empathy. This is examined by issuing a psycholinguistic experiment

designed to probe both notions of Empathy. In this experiment, a person’s capacity to

experience Psychological Empathy is determined through a prevalent measure of

Psychological Empathy, i.e. Empathy Quotient (Baron-Cohen & Wheelwright, 2004). Then,

experimental participants are exposed to sentences that vary in felicitousness, and they are

asked to provide acceptability ratings for these stimuli that manipulate phenomena within

Linguistic Empathy. If a measure of Linguistic Empathy is found (as posited in section

1.2.1), and if this measure correlates significantly with the measure of Psychological

Empathy, then it follows that Linguistic Empathy and Psychological Empathy share more

than a common term, Empathy, and are in fact interrelated phenomena. Uncovering the

extent that this overlap exists is a central goal of this study.

1.2.3 Methodological impact. This dissertation takes a transdisciplinary approach to

assess correlations between Psychological Empathy and Linguistic Empathy. The goals of

this approach, as stated above, are to yield a measure of Linguistic Empathy as well as

establish correlations between Psychological and Linguistic Empathy. The aforementioned

psycholinguistic experiment produces quantifiable results that can apply to both the

linguistic phenomena that are manipulated in the experiment and to the individuals who

participate in the experiment. When the results are analyzed with respect to the linguistic

phenomena, this serves to solidify current understandings of Linguistic Empathy, and to

nuance the understanding of the linguistic phenomena that are represented in the

experiment. This methodology is relatively standard in experimental linguistics. However,

when the results are analyzed by applying these results to the participants themselves, this

methodology investigates individual difference factors that can account for variation in

acceptability ratings. The success of the goals in sections 1.2.1 (Linguistic Empathy

measure) and 1.2.2 (correlations of Linguistic and Psychological Empathy) are predicated

on the analysis of the results reflecting individual differences in Empathy processing and

issuing language acceptability ratings. The goal of this methodological approach is to

challenge linguists to examine acceptability judgments in language not only through

grammatical rules that reflect linguistic structure, but through idiosyncratic and pragmatic

accounts of language that reflect individual differences in linguistic capacity.

1.2.4 Confirmations and Extensions of Empathy. In addition to the individual

differences that are investigated in this dissertation, confirmation and extension to existing

Linguistic Empathy theory will be discussed. The experiment provides a quantitative

means to solidify linguistic phenomena within Linguistic Empathy, which can serve as

evidence for a united theory of Linguistic Empathy. Additionally, if Psychological Empathy

is demonstrated as a correlate of Linguistic Empathy, then the extended view of Linguistic

Empathy must accommodate this united view of Empathy.

1.3 Summaries of Chapters

Chapter 2

Chapter 2 of this dissertation examines the notions of Psychological Empathy and

Linguistic Empathy more deeply. One goal of this chapter is to present the disparate

applications of these notions while also demonstrating an overlap in the fundamentals of

perspective taking. The chapter investigates the evolution of Psychological Empathy

measures and evaluates strengths and weaknesses of the various approaches, including the

Empathy Quotient test (Baron-Cohen & Wheelwright, 2004). Despite its shortcomings,

Empathy Quotient is presented as an eminent measure of Psychological Empathy, and it

serves as the measure of Psychological Empathy for the subsequent experiment in this

dissertation. Then, the notion of Linguistic Empathy is reviewed, beginning with Empathy

Hierarchies that are posited by Kuno & Kaburaki (1977) and Kuno (1987). The specifics of

three of these hierarchies, the Surface Structure Empathy Hierarchy, the Person Empathy

Hierarchy, and the Topic Empathy Hierarchy, are examined in depth because these

hierarchies were central to creating the stimuli for the experiment in the study.

Additionally, this chapter situates Kuno’s approach with other notions of viewpoint in

linguistics and related fields, and consolidates it with notions of Linguistic Empathy in

other approaches (e.g., Silverstein, 1976; Yokoyama, 1986; Deane, 1992) for a unified

Linguistic Empathy model.

Chapter 3

The purpose of chapter 3 is to bridge the theoretical notions of Empathy discussed in

chapter 2 with a psycholinguistic experimental approach, and to illustrate the methodology

for this approach. This chapter first provides a rationale for the experimental approach.

Next, the specific features of the Linguistic Empathy Hierarchies can be isolated and

operationalized as variables within an experimental setting, so the chapter demonstrates

the process of linguistic manipulation of the stimuli in the experiment. The chapter then

submits specific predictions based on the manipulation of the stimuli, and it concludes by

providing the methodology of the experiment.

Chapter 4

The results and discussion of the experiment are presented in chapter 4. The

purpose of this chapter is to present the quantitative evidence that supports the pursuit of

a valid individual measure of Linguistic Empathy as well as a correlation between

Psychological and Linguistic Empathy. The goals of the results are to confirm the

predictions of Linguistic Empathy theory, an individual measure of Linguistic Empathy,

and an Empathy processing component that is common to both Psychological and Linguistic

Empathy. Additionally, the limitations and potential extensions of this experiment are

considered.

Chapter 5

Chapter 5 expands upon the implications of the experiment that were introduced in

the previous chapter. This chapter extends the application of these results into four areas.

First, the implications of a measure of Linguistic Empathy sensitivity are introduced.

Second, the correlations between Psychological Empathy and Linguistic Empathy are

reviewed, and the influence that a test of Linguistic Empathy can exert on Psychological

Empathy measures is discussed. Third, the methodological contributions of experimental

approach challenge linguists to rethink the methods and implications of acquiring native

speaker grammatical judgments. Finally, extensions are proposed for further Linguistic

Empathy and Psychological Empathy experimentation.

CHAPTER 2

PSYCHOLOGICAL AND LINGUISTIC NOTIONS OF EMPATHY

This chapter describes varying notions of Psychological and Linguistic Empathies,

including where they overlap and where they diverge. After establishing current

psychological understandings of Empathy, the chapter discusses the means of measuring

an individual’s capacity for Psychological Empathy. The chapter then considers Linguistic

Empathy theory, beginning with the Empathy Hierarchies proposed by Kuno and Kaburaki

(1977) and Kuno (1987), and continuing with notions related to linguistic point of view,

such as animacy (Silverstein, 1976) and entrenchment (Deane, 1992). Next, the chapter

examines the role that passive constructions, reciprocal verbs, and preceding context can

have in Linguistic Empathy. Based on these discussions, the approach to Empathy taken in

this dissertation is then illustrated.

2.1 Psychological Empathy

The term Psychological Empathy in this study is used to refer to the social

psychological notions of Empathy in order to distinguish Psychological Empathy from

Linguistic Empathy. Psychological Empathy refers to the ability of a person to understand

the thoughts and emotions of others, to share in this emotional experience, and to respond

to these situations appropriately. Psychological Empathy allows a person to navigate

productively the social and emotional landscape in which the interaction between the

speaker and others takes place. In the field of psychology, this term is typically discussed in

terms of Affective Empathy and Cognitive Empathy.

2.1.1 Affective and Cognitive Empathy. Affective Empathy reflects the emotional

and prosocial component that is applied to seeing the world from another’s perspective. As

defined by Bryant (1982, p. 414), Affective Empathy is “the vicarious emotional response to

the perceived emotional experiences of others.” The emotional responses that comprise

Affective Empathy are typically categorized as either parallel (i.e., experiencing the same

emotion) or reactive (i.e., sympathy, pity, or compassion beyond the experiencer’s emotion)

(Davis, 1980; Lawrence et al., 2004). Crucially, the emotional response must be socially

appropriate to be considered as part of Affective Empathy. Taking delight in another’s

misfortune, for instance, is self-oriented, and therefore not considered a prosocial response

to another person (Baron-Cohen & Wheelwright, 2004; Lawrence et al., 2004).

Sympathy and Affective Empathy are neither equivalent nor distinct from one

another; they overlap. Davis (1994) cites that the origins of sympathy and Empathy have

different theoretical roots. The concept of sympathy originated from 18

century moral

philosophy to describe the emotional response that occurs when a person observes someone

else experiencing a significant emotion. This means that sympathy originates from focusing

on another’s situation, but maintaining one’s own point of view. Contrarily, the concept of

Empathy originates from German aesthetics to describe the phenomenon of a person

inserting her/himself into another’s situation. Thus, a contrast between Empathy and

sympathy is that sympathy maintains a person’s perspective, and Empathy involves

leaving one’s own point of view in order to take another’s (Davis, 1994). Affective Empathy

thereby demands the recognition of another’s emotions and requires the ability to craft an

appropriate emotionally congruent response, which sometimes includes sympathy, and

sometimes includes other emotional responses.

Cognitive Empathy, on the other hand, involves parsing out the emotions and

reactions of Affective Empathy from the unemotional and rational aspects of Empathy. The

focus is placed on the ability to understand others’ emotions (Kohler, 1929) as well as

processes like role taking and perspective switching (Mead, 1934). As defined by Hogan

(1969, p. 313), Cognitive Empathy is the “intellectual or imaginative apprehension of

another’s condition or state of mind.” This approach is not necessarily prosocial. If this

person were to exploit or manipulate others with the knowledge that accompanies

emotional understanding, he/she might display a lack of Affective Empathy but a wealth of

Cognitive Empathy. Similarly, Smith (2006) contends that an imbalance of Cognitive and

Affective Empathy is associated with Empathy deficits such as schizoid personality disorder

and antisocial personality disorder, and Jolliffe & Farrington (2006, p. 592) suggest that the

apparent social charm of psychopaths can be attributed to this imbalance. Jolliffe &

Farrington go on to argue that, as a means of identifying this imbalance, it would useful to

have Empathy measures that accurately reflect different facets of Empathy. However,

Baron-Cohen & Wheelwright (2004) maintain that despite the theoretical existence of

multiple facets that Empathy embodies, they are so inherently tied that, empirically, it is

exceedingly difficult to tease them apart.

2.1.2 Psychological Empathy measures. Hogan (1969) argues for the importance of

developing an Empathy measurement in order to quantify concepts directly and indirectly

related to morality, and thus developed a test for measuring these concepts. Developing this

scale, he argues, provides a necessary step toward shifting conceptions of Empathy to a

quantifiable construct. However, Jolliffe & Farrington (2006) argue that a shortcoming of

the Hogan Empathy Scale (HES) is that it does not measure Empathy; rather, it identifies

individuals belonging to high and low Empathy groups based on other conditions. The HES

is no longer considered a valid measure of Cognitive Empathy, but it remains an accurate

predictor of behavior that is indicative of criminal or delinquent behavior (Jolliffe &

Farrington, 2006, p. 591).

Other common Empathy measures include the Questionnaire Measure of Emotional

Empathy (QMEE) (Mehrabian & Epstein, 1972), which measures Emotional (i.e., Affective)

Empathy, and the Interpersonal Reactivity Index (IRI) (Davis, 1980), which measures both

Affective and Cognitive Empathy. Both tests are self-reports in which respondents agree or

disagree with statements that are designed to reflect Empathy. However, Jolliffe &

Farrington (2006) argue that some statements falsely equate sympathy and Empathy (e.g.,

“I get very angry when I see someone being ill-treated” [Mehrabian & Epstein, 1972], “other

people’s misfortunes do not usually disturb me a great deal” [Davis, 1980]). While these

statements reflect instances of sympathy (or the lack thereof), Jolliffe & Farrington

consider only the parallel responses of Affective Empathy (i.e., experiencing the same

emotions) and not the reactive responses of Affective Empathy (feelings beyond the

experiencer’s emotions) under their definition of Affective Empathy (2006, pp. 591 ff.). As

such, they argue that questions relating to sympathy (i.e., reactive Affective Empathy)

should be disqualified from Affective Empathy measures. Nevertheless, there appears to be

a consensus that Affective Empathy includes any appropriate parallel or reactive emotional

response (Davis, 1994; Baron-Cohen & Wheelwright, 2004; Lawrence et al., 2004).

Additionally, despite the attempts of the QMEE to measure solely Affective

Empathy, the creators of the QMEE (Mehrabian & Epstein, 1972) cite that its items were

likely to tap into a single empathic construct that includes both cognitive aspects and

emotional arousal (Baron-Cohen & Wheelwright, 2004). Other critiques of the IRI state

that it does not attempt to measure Cognitive Empathy; rather, they attempt to measure

more general perspective taking ability. In other words, instead of measuring a person’s

ability to understand the emotions of others, the IRI measures the person’s ability to take

another’s perspective without necessarily understanding the emotions involved (Jolliffe &

Farrington, 2006).

Baron-Cohen and Wheelwright (2004) justify the need for a unifying test for

Psychological Empathy, citing previous shortcomings of the IRI and QMEE (Baron-Cohen

& Wheelwright, 2004). While previous Empathy measures may accurately express factors

within Empathy (e.g., Cognitive Empathy, empathic concern, etc.), they assert that

Affective and Cognitive Empathy cannot be parsed easily, and one should not be analyzed

without the other. Consequently, Baron-Cohen and Wheelwright sought to develop a

measure of Empathy, Empathy Quotient (EQ), to investigate the relationship between

Empathy and Autism. Baron-Cohen et al. (2001) had already developed a test called the

Autism Spectrum Quotient, but this quotient did not focus on Empathy. They argued that a

lack of Empathy, both affectively (i.e., appropriate emotional responses) and cognitively

(i.e., theory of mind), requires a unifying measure that does not attempt to parse and

measure individual aspects of Empathy.

Initial tests of reliability and replicability for EQ reinforce the validity of the

measure. Lawrence et al. (2004) demonstrate that the EQ test yields strong inter-rater

reliability, test-retest reliability, and correlation with key aspects of the IRI. However, the

study also indicates that social desirability (i.e., not wanting to self-report unfavorably)

may influence participants in some questions. Additionally, despite Baron-Cohen and

Wheelwright’s (2004) claim that facets of Empathy (e.g., Affective versus Cognitive) cannot

be parsed, Lawrence et al. (2004) employed a Principal Components Analysis to determine

that the EQ questions could be accurately categorized into three subtypes of Empathy that

measure: Cognitive Empathy, emotional reactivity, and social skills separately. Other

evaluations of EQ (e.g., Muncer & Ling 2006, Berthoz et al. 2008) again confirm that the

measure of these three separate types of Empathy is valid. Furthermore, Allison et al.

(2011) support dividing the test into three categories of Empathy by applying a Rasch

analysis, to measure latent personal traits, to evidence the view of EQ as a unidimensional

measure of Empathy. Crucially, through all this, the validity of the EQ remains, and

considering EQ as multi-factor measure of Empathy or a unidimensional measure of

Empathy are valid approaches and applications of EQ.

Additional criticisms of EQ cite that participants with Autism Spectrum Disorder

(ASD) often underreport their own Empathy compared to control groups (Baron-Cohen &

Wheelwright, 2004; Lawrence et al., 2004; Allison et al. 2011). With EQ questions and ASD

status both relying on self-reporting, the opportunity for variance arises between those with

ASD who report their condition and those who do not (Lawrence et al., 2004). Despite these

criticisms, Lawrence et al. (2004) report that EQ accurately distinguishes between control

groups (i.e., participants without ASD) and clinical groups (i.e., participants with ASD)

because of the robust sample size. This particular shortcoming targets EQ in

clinical/diagnostic settings; however, this does not convey that EQ is a false indicator of a

person’s Empathy. Additionally, if someone is hoping to get a baseline measure of

Psychological Empathy without heeding the ASD implications, this test is described as

“suitable for use as a casual measure of temperamental Empathy by and for the general

population” (Baron-Cohen & Wheelwright, 2004).

To measure a socio-cognitive construct like Empathy with reliability is no simple

task. Certain tests have proven successful at reliably targeting specific aspects of Empathy

or Empathy related behavior (e.g., HES - offending behavior, EQ - singular Empathy

construct, QMEE - emotional arousal, IRI - perspective taking). As Psychology continues to

improve upon previous versions of Empathy tests, the most prominent tests all seem to

follow a self-reporting style, which appears to be an inherent conflict to the impartiality of

what it attempts to measure. Many of the questions on these exams are quite direct and

transparent toward the goal of assessing social awkwardness or emotional confusion. This

creates a motive for someone who may want to “correctly” answer a question to avoid being

labeled at risk for a deficiency of Empathy.

2.2 Linguistic Empathy

The purpose of this section is to provide an overview of Linguistic Empathy, which is

the central principle under investigation in the psycholinguistic experiments in this paper.

As mentioned, Linguistic Empathy and Psychological Empathy derive from different

theoretical constructs, which happen to share the term

Empathy

. However, as will be

demonstrated in this section, many of the theoretical foundations, like perspective-taking

and appropriate responses, overlap between the two concepts. The extent to which these

two distinct concepts overlap is a central research question to be explored.

2.2.1 Empathy perspective. Linguistic Empathy refers to the vantage point from

which an event is encoded in a speaker’s language. Empathy is the perspective with respect

to certain participants (e.g. the agent or the patient of the action in an utterance) or events

in a sentence. In expressing his/her point of view, the speaker identifies with entities in a

speech event to varying degrees. As defined by Kuno (1987):

(2.1) Empathy: Empathy is the speaker’s identification, which may vary in degree,

with a person/thing that participates in the event or state that he describes in a

sentence (Kuno, 1987, p. 206).

(2.2) Degree of Empathy: The degree of the speaker’s Empathy with x, E(x), ranges

from 0 to 1, with E(x) = 1 signifying his total identification with x, and E(x) = 0 a

total lack of identification (Kuno, 1987, p. 206).

One facet of Linguistic Empathy is that it explains certain linguistic data that defy

standard syntactic explanation. It manifests in the speech events produced in the form of a

speaker’s momentary and subconscious choice of expression among grammatically or

lexically correct options at his/her disposal. Linguistic Empathy has been used to explain

how the subconscious selection is reached. For example, when a mother asks her five-year-

old son

Did daddy call?

rather than

Did Joey call?

Did my husband call?

, the mother is

sharing the perspective of the child, who thinks of his father as

daddy

instead of

Joey

(his mother’s) husband.

The mother, however, more likely considers this same person as

Joey

my husband

when expressing her own perspective. Besides playing a role in the

choice of referential expressions (

daddy

versus

Joey

), Linguistic Empathy has been

proposed as the force behind other lexical choices (

come

versus

), as well as behind the

choice of the grammatical subject of reciprocal verbs and some active/passive options (see

more on this below), reflexive forms (

you

versus

yourself

), and other phenomena observed

in several typologically unrelated languages (Silverstein, 1976; DeLancey, 1981; Kuno,

1987; Yokoyama, 1999; Oshima, 2007a).

Since language speakers are typically heavily reliant on their visual sense, this

vantage point is often conceived of through visual representation, as in

Figure 2.1

below.

This depicts Kuno and Kaburaki’s (1977, p. 628) reference to “camera angles” as a manner

of experiencing language, where the sentences listed in (A), (B), and (C) correspond to the

metaphorical vantage point of (A), (B), and (C). The notion of “camera angle” corresponds to

where a camera would be placed to capture the scene depicted in the utterances. The

utterances are all logically equivalent, and they vary only in grammatical structure and

referential expressions. The utterances in (A) provide a relatively neutral depiction of the

scene, whereas the utterances in (B) and (C) convey

Teddy’s

and

Lucy’s

perspective,

respectively.

Kuno’s initial description of Linguistic Empathy suggests that a sentence can be

conceived of in terms of camera placement as if the scene were filmed. Although this

embodies much of how Linguistic Empathy exists in language, this notion oversimplifies

Linguistic Empathy and falls short of representing its crux. The camera placement aligns

nicely with

viewpoint

, in which we literally

see

the scene described in the sentence from the

appropriate angle. However, this does not account for sentences that are

experienced

and

cannot be

viewed

, in literal terms. Like the theory of perspective discussed by Uspensky

(1972), Linguistic Empathy goes beyond the visual experience implied through the use of a

camera angle and instead represents the entire visceral experience of an event. The extent

to which a speech event portrays this experience reflects Linguistic Empathy. For instance,

a sentence like

her stomach turned

is not something for which “camera placement” is

appropriate. This sentence is not about the film-able signals of such an event (e.g., a

wincing facial expression that indicates someone’s stomach might be turning); rather, this

sentence is about the internal experience of one’s stomach turning. The version of this

(A) Teddy hugged Lucy.

Lucy hugged Teddy.

Teddy and Lucy hugged.

(B) Teddy hugged his sister.

brother.

Adapted from Kuno & Kaburaki (1977, p. 628).

Figure 2.1: Kuno & Kaburaki’s “Empathy and Camera Angles”

sentence that would be film-able (e.g.,

[it seemed that] her stomach turned

) would shift the

perspective from internal to external.

For a classic illustration of a shift in Linguistic Empathy that defies syntactic

explanation, we turn to Kuno and Kaburaki (1977) and some examples with the Japanese

verbs

yaru

and

kureru

, which share the same logical meaning,

to give.

The difference in

these verbs is entirely empathic:

yaru

is used when the speech event is described from the

sentential subject’s (i.e., the giver’s) perspective, and

kureru

is used when the speech event

is described from the sentential dative’s (i.e., the receiver’s) perspective, as seen in the

following two sentences, both of which have the core meaning “Taroo gave the money to

Hanako.”

(2.3) a. Taroo wa Hanako ni okane o

yatta.

(Subject-Centered)

Taroo-NOM Hanako-DAT money-ACC gave

‘Taroo gave the money to Hanako.’

b. Taroo wa Hanako ni okane o

kureta.

(Dative-Centered)

Taroo-NOM Hanako-DAT money-ACC gave

‘Taroo gave the money to Hanako.’

(Adapted from Kuno & Kaburaki 1977, p. 630)

The syntactic structure and case marking of these two examples are identical, and

the logical truth value for each sentence is the same (i.e.,

Taroo

was the giver of money to

Hanako

). However, despite the fact that both (2.3a) and (2.3b) are grammatical and

acceptable, there is not free alternation of these verbs and their corresponding perspectives.

By uttering (2.3a), the speaker conveys the perspective of

Taroo

, thereby unequivocally

signaling an identification with him. It would most naturally be uttered by someone who is

part of

Taroo’s

in-group (e.g.,

Taroo’s

friend or relative). Alternatively, by uttering (2.3b),

the speaker conveys the perspective of

Hanako

, thereby unequivocally signaling an

identification with her. It would most naturally be uttered by someone who is part of

Hanako’s

in-group (e.g.,

Hanako’s

friend or relative). For a speaker who is part of an in-

group to utter a sentence that corresponds with the perspective of the other entity would

sound jarring to the hearer and cause confusion as to the speaker’s allegiance. A speaker

who is not part of either in-group or who is part of both in-groups choosing between these

two verbs must therefore indicate whether to view and convey the logical truth of the

utterance from the perspective of the giver (

Taroo

) or from the perspective of the receiver

(

Hanako

) by means of verb choice

It is possible for the identification with one of the entities in the sentence to be made

lexically explicit when certain referential expressions are used. For instance, uttering the

Japanese first person male pronoun

boku

identifies the speaker unequivocally with the

speaker’s own perspective (i.e., the speaker must identify with himself), whereas uttering

the name

Taroo

does not explicitly identify the speaker’s perspective. Sentences are

unacceptable when the identification with one of the sentential entities is made lexically

explicit, and the perspective conveyed by the verb conflicts with the explicit perspective of

that sentential entity, as in examples (4) and (5) below.

(2.4) a.

Boku

wa Taroo ni okane o

yatta.

(Subject-Centered Verb)

I - ACC Taroo-DAT money-ACC gave

‘I gave money to Taroo.’

A truly neutral speaker, e.g. a judge in a court case, would avoid either verb and resort to

a formal Sino-Japanese high register word like

zouyo-suru

‘gift’ or an archaism high

ataeru

‘give’, neither of which would be usable in normal speech.

b. * Taroo wa

boku

ni okane o

yatta.

(Subject-Centered Verb)

Taroo-ACC I-DAT money-ACC gave

‘Taroo gave me money.’

(2.5) a. *

Boku

wa Taroo ni okane o

kureta.

(Dative-Centered Verb)

I - ACC Taroo-DAT money-ACC gave

‘I gave money to Taroo.’

b. Taroo wa

boku

ni okane o

kureta.

(Dative-Centered Verb)

Taroo-ACC I-DAT money-ACC gave

‘Taroo gave me money.’

(Adapted from Kuno & Kaburaki 1977, p. 631)

In all examples above, the perspective is explicit by means of the first person

pronoun

boku

. In other words, when

boku

is uttered, the speaker conveys his/her own

perspective, and the sentence should be experienced from this vantage point. When

boku

used as the subject of the sentences in (2.4a) and (2.5a), the subject-centered verb

yaru

(2.4a) correctly corresponds with the perspective of the speaker (

boku

), whereas the dative-

centered verb

kureru

in (2.5a) clashes with the perspective that is already established by

boku

. Similarly, when

boku

is used as the indirect object of the sentences in (2.4b) and

(2.5b), the dative-centered verb

kureru

in (2.5b) correctly corresponds with the perspective

of the speaker (

boku ni

), whereas the subject-centered verb

yaru

in (2.4b) clashes with the

perspective that is already established.

Examples (2.4) and (2.5) illustrate that a shift in Empathy perspective to an

unnatural or unexpected Empathy locus can hinder the acceptability of an utterance. Kuno

and Kaburaki, (1977) and Kuno (1987) use violations of Linguistic Empathy like those in

(2.4b) and (2.5a) to inform and establish rules of acceptable perspective taking. Like

Silverstein’s (1976) predictions of the distribution of split-ergative case markings, these

rules form Empathy Hierarchies that can predict violations and acceptability of sentences

where syntax fails to do so.

2.2.2 Empathy Hierarchies and their pragmasemantic connections. Kuno (1987)

posits that the Degree of Empathy is managed by a series of rules in the form of Empathy

Hierarchies (EHs) that determine preferences and constraints for the perspectives shared

in utterances. A speaker establishes a vantage point, or Empathy locus, through different

discourse channels (e.g., referential expressions, structure, semantic relations, context) that

manifest within these hierarchies (Kuno & Kaburaki, 1977; Kuno, 1987), three of which are

described below.

(2.6) Surface Structure Empathy Hierarchy: It is easier for the speaker to empathize

with the referent of the subject than with the referents of other noun phrases

(NPs) in the sentence.

E(subject) > E(other NPs), (Kuno, 1987, p. 211)

Observed:

Juli met a strange man at the party.

Violated:

(?) A strange man met Juli at the party.

This Surface Structure EH formalizes the notion that perspective is typically established by

the subject of the sentence. In (2.6a), the sentential subject

Juli

is the Empathy locus.

However, in (2.6b), the Surface Structure EH selects the Empathy perspective of the subject

A strange man

. The sentence becomes awkward once the name

Juli

is uttered since a

proper name like

Juli

is typically more relevant than an indefinite NP like

a strange man

In addition to the structural account of establishing Empathy perspective, a speaker

typically express identity with her/himself over others by using the first person pronouns

and

, which is the notion asserted in the Person EH below.

(2.7) Person Empathy Hierarchy (also known as the Speech-Act Empathy Hierarchy):

The speaker cannot empathize with someone else more than with himself.

E(speaker) > E(others), (Kuno, 1987, p. 212)

Observed:

I like Gary.

Violated:

(?) Gary is liked by me.

The Person EH formalizes the notion that a speaker identifies with her/his own point of

view over all others. The logical truth value in (2.7a) and (2.7b) above is identical; however,

when the Empathy locus shifts from the speaker (

) to a third person entity (

Gary

), the

acceptability of the sentence suffers. The effect that passive voice exerts on sentences like

(2.7b) will be discussed in section 2.2.5. The Topic EH, below, is the final EH to be discussed

here, and it addresses the role of context in perspective taking.

(2.8) Topic Empathy Hierarchy: Given an event or state that involves A and B such

that A is coreferential with the topic of the present discourse and B is not, it is

easier for the speaker to empathize with A than with B.

E(discourse topic) ≥

E(nontopic), (Kuno, 1987, p. 210)

a. Observed:

Did you hear what happened to John? He got into a car accident

with Anna.

b. Violated:

Did you hear what happened to John? (?) Anna got into a car

accident with him.

The Topic EH formalizes the notion that entities that are topically relevant are typically

the focus of Empathy. Examples (2.8a) and (2.8b) again convey the same logical content;

It is not clear what motivated Kuno (1987) to use the ≥ instead of the > in this EH. This

paper will continue the use of the ≥ within this EH, but it does not take a strict stance as to

which sign is more appropriate.

however, (2.8b) is less acceptable than (2.8a). This is due to the first sentence in each

example, in which the referent

John

is mentioned. The question

Did you hear what

happened to John?

establishes

John

as a relevant topic for the following discourse, and the

Topic EH suggests that Empathy with a discourse topic is more natural than with a non-

topic.

When the Empathy relations across these hierarchies align, then the sentence’s

Empathy perspective is consistent. However, discrepancies in Empathy locus can occur

within one utterance, which can often cause marginality in the acceptability of a sentence.

Kuno postulates a

Ban on Conflicting Empathy Foci

, which posits that “a single sentence

cannot contain logical conflicts in Empathy relations” (1987, p. 207). According to this

hypothesis, the Empathy relations must remain logically consistent across these

hierarchies. Following Kuno’s terms, when the inequalities align, then there is no Empathy

violation. However, when the inequalities conflict, this causes an Empathy violation, which

affects sentence acceptability. It should be mentioned that Uspensky (1972), Yokoyama

(1979, 2000), and Christensen (1994) have all provided counterexamples in Russian and

Polish that refute Kuno’s ban. In these examples, multiple distinct perspectives occur in the

same sentence. Yokoyama and Christensen have argued against the ban on these grounds.

Instead, the Ban on Conflicting Empathy Foci can be considered a guideline. Let us

consider some examples of the sentences that are in violation, and analyze them using

Kuno’s ban.

(2.9)

I met a strange man at the party.

a. Surface Structure EH: E(

) > E(

a strange man

)

b. Person EH: E(

) > E(

a strange man

)

c. Topic EH: N/A

(2.10)

A strange man met me at the party.

a. Surface Structure EH: E(

a strange man

) > E(

)

b. Person EH: E(

) > E(

a strange man

)

c. Topic EH: N/A

(2.11)

I like Gary.

a. Surface Structure EH: E(

) > E(

Gary

)

b. Person EH: E(

) > E(

Gary

)

c. Topic EH: N/A

(2.12)

Gary is liked by me.

a. Surface Structure EH: E(

Gary

) > E(

)

b. Person EH: E(

) > E(

Gary

)

c. Topic EH: N/A

(2.13)

Did you hear what happened to John? He got into a car accident with

Anna.

a. Surface Structure EH: E(

John

) > E(

Anna

)

b. Person EH: N/A

c. Topic EH: E(

John

) > E(

Anna

)

(2.14)

Did you hear what happened to John? Anna got into a car accident with

him.

a. Surface Structure EH: E(

Anna

) > E(

John

)

b. Person EH: N/A

c. Topic EH: E(

John

) > E(

Anna

)

Examples (2.9), (2.11), and (2.13) all explicitly invoke two of the three EHs discussed

above, and the Empathy relationships for these hierarchies do not conflict. Contrarily,

examples (2.10), (2.12), and (2.14) also invoke two of the three EHs, but the Empathy

relationships for these hierarchies are in conflict. For instance, examples (2.11) and (2.12)

investigate the sentences

I like Gary

and

Gary is liked by me

, respectively, in terms of the

Person EH and the Surface Structure EH. Since the Surface Structure EH places the

Empathy perspective on the subject of the sentence, and the Person EH places the Empathy

perspective with the speaker (

I/me

) above other NPs,

is valid as the Empathy locus by

means of both EHs in example (2.11), but the Empathy perspective is conflicting in example

(2.12). This conflict, according to Kuno, causes marginality in sentences. Since there is no

explicit topic in this example (e.g., no preceding context), the Topic EH was not considered

in the examples above unless there was a context sentence given. Kuno does not make

explicit how a sentential entity becomes a “topic of the present discourse,” in cases when no

preceding context includes a topic, so for now, it is assumed that the sentential entity must

be referenced previously. The role of Topic EH when there is not explicit entity that is

referenced previously will be discussed in section 2.2.4.

Other linguistic hierarchies have been proposed that are closely related to Kuno’s

EHs. Silverstein (1976, 1981) and Deane (1992) propose hierarchies that largely overlap

with Kuno’s and describe closely related phenomena. The Silverstein Hierarchy was

originally proposed to explain variability in split-ergative languages. Ergativity allows

perspectives and identities to be assigned without specific lexical items or structural

changes. Ergative-absolutive languages assign the same case marking to the subject of an

intransitive verb as the object of a transitive verb, whereas nominative-accusative

languages, such as English, assign the same case marking to subjects of transitive and

intransitive verbs. Split-ergative languages contain sentences that can vary between

sentences with ergative-absolutive case markings and sentences with nominative-

accusative case markings. Silverstein establishes a hierarchy based on animacy (e.g.,

human > nonhuman

, and

person > 2

person > 3

person

, etc.) that predicts which

sentential NPs prefer a given case based on the lexical properties of this NP — that the

“‘split’ of case markings is not random” (Silverstein, 1976, p. 113). In other words, NPs that

are higher on the Silverstein Hierarchy more naturally take nominative-accusative case

markings, and the NPs that are lower on the hierarchy take ergative-absolutive case

markings. This parallels the hierarchies that Kuno proposed for Linguistic Empathy in

terms of the ease with which a speaker can identify/empathize with an entity in a speech

event. Kuno argues that in order to share an entity’s perspective, that entity must be

capable of having a perspective in the first place. That excludes inanimate entities and non-

human animate entities, which must be personified in order for the speaker to assume their

perspective. Importantly, Silverstein focuses on the lexical properties of the varying NPs

when establishing order to the assignment of case markings. DeLancey (1981)

acknowledges the Silverstein Hierarchy and builds upon the lexical properties of NPs to

include the use of aspect in determining viewpoint in split-ergative languages. Verb aspect

(e.g., perfect vs. imperfect) in these languages can shift viewpoint from internal to external

without a change in case markings or lexical items. Other languages, such as English, must

default to changes in structure, like passivization, in order to accomplish a similar effect.

Deane (1992) applies the notions of animacy, agency, and salience to the NPs of an

utterance in order to extend the Silverstein Hierarchy. Deane argues that the extent to

which a sentential entity embodies these qualities correlates with how naturally a

speaker/hearer should share its perspective. Animacy, agency, and salience unify Deane’s

hierarchy of entrenchment with Kuno’s EHs, and these qualities can map onto Kuno’s EHs.

The first factor of these factors, animacy, is the conceptual relation that an entity has in

common with the speaker. That is, the more in common an entity has physiologically with

the speaker, the more natural it is to empathize with this entity. As mentioned above, the

notion of animacy is central to the Silverstein Hierarchy (Silverstein, 1976). Animacy

hierarchies rank humans over non-humans, which is typically further delineated in the

following order:

Humans > Animals > Plants > Natural Forces > Concrete Objects >

Abstract Objects

. The

Humans

portion of this hierarchy most broadly unravels into the

following order:

First Person > Second Person > Third Person

(Yamamoto, 1999). Animacy

and humanness map directly onto Kuno’s Person EH and provide the means to investigate

greater granularity to the concepts of

speaker

and

other

. A crucial difference between

Deane’s entrenchment hierarchy and Kuno’s Person EH is that the granularity introduced

in the Deane’s hierarchy is based on tendencies of Empathy relations, but not strict rules;

violating these tendencies does not necessarily result in an invalid utterance. However, the

Person EH is based on the binary notion of

speaker

versus

others

, and it claims that

utterances that violate this preference, i.e.,

others

over

speaker

, are infelicitous.

The agency of a NP is another factor that diminishes sequentially along Deane’s

entrenchment hierarchy and the Silverstein Hierarchy. With humans at the top, and non-

humans toward the bottom, the relevance for establishing agency also diminishes as one

moves down the hierarchy. In other words, semantically speaking, humans should be more

capable of serving as agents in narrated events, which is a notion that aligns with the

animacy hierarchies. Hopper & Thompson (1980) consider agency interrelated with the

transitivity and salience of a narrated event. They argue that in discourse, story lines are

moved along by performing actions, and that agency is linked with entities performing

these actions (Hopper & Thompson, 1980, p. 286). Then, using the Silverstein Hierarchy

(1976), they demonstrate the correlation of agency with grounding/salience. Hopper and

Thompson posit agency as a factor in determining the degree of transitivity by

demonstrating that a sentence with greater agency is predicted to occur in discourse with

greater salience. Agency maps onto Kuno’s Person EH as those higher on human hierarchy

are more likely to be sentential agents.

The salience of a narrated event is another factor that is related to determining the

recipients of Empathy perspective. Salience refers to the degree to which an entity or action

is likely to enter the mental state of a language user. The salience increases when concepts

are common and stable, or if they are contextually activated in speech events (Schmid,

2007, pp. 119 ff.). An item is considered salient if “it has been loaded . . . into current

working memory and has thus become part of a person’s center of attention” (Schmid, 2007,

p. 119). Similarly, Yokoyama (1986, pp. 31 ff.) argues that topics that are in the center of

current concern are more relevant as discourse topics. In other words, with respect to

Empathy, the more real and tangible an entity is, the more natural it becomes to share its

perspective. Relatedly, Hopper & Thompson (1980) discuss the Transitivity Hypothesis in

which the relative level of transitivity in a narrated event is highly correlated with its

salience. They argue that transitivity can be predicted based on linguistic parameters like

agency, kinesis (i.e., action transfer), and punctuality; and that these parameters are

heavily involved with the foregrounding (higher transitivity) and backgrounding (lower

transitivity) of the speech event. The terms “foregrounding” and “backgrounding” indicate

in binary terms higher and lower degrees of salience, respectively, of an utterance (Cramer,

2011, p. 87), as opposed to a sliding scale such as Fillmore’s Saliency Hierarchy (Fillmore,

1977, p. 78). Hopper & Thompson qualitatively cite universal manifestations of

morphosyntactic features that support the Transitivity Hypothesis across unrelated

languages, and then establish a quantitative correlation of linguistic parameters with

salience that are found within the narrative. They argue that although grammatically valid

structures exist that flout the tendency for correlation between salience and transitivity,

these manifestations are rare and marked. This argument parallels the grammatically

relevant structures of Linguistic Empathy for which violations can be grammatically valid

but are marked.

Salience maps onto Kuno’s Person EH and the Topic EH. Regarding the Person EH,

the notions of humanness that are central to this EH are more real and tangible to a

speaker than abstract concepts. Additionally, the concept of a self-referential first person is

more salient than a different human, especially one who is grammatically indefinite, plural,

or unfamiliar to the speaker. It is highly implausible for a speaker to enter the mind of

someone she/he does not know or cannot identify. Animacy is a strong indicator of

transitivity, and entities higher on the animacy hierarchy are shown to exist primarily in

prominent positions of salience (Hopper & Thompson, 1980). Regarding the Topic EH,

discourse in which an entity is mentioned previously increases the salience of that entity

with respect to the current concern of the interlocutors. Hopper & Thompson argue that

topicality (i.e., definiteness and referentiality) “follow[s] naturally from the fact that

foregrounded [i.e., salient] clauses typically continue talking about the same participant

within one episode, rather than introducing a new participant” (1980, p. 286). Since

background precedes foreground, and background sentences can suggest or predict the

topics of the following foreground sentences, Hopper & Thompson argue that topicality,

salience, and agency are all closely correlated. Similarly, topicality, salience, and agency

have all been discussed as factors that influence the Empathy perspective of utterances and

map onto the Empathy Hierarchies as defined by Kuno (1987).

2.2.3 Point of view. Similar to the interdisciplinary notions of Empathy, point of

view is a construct that can take shape across disciplines. Point of view can refer to a visual

point of view, a dogmatic stance, a cognitive standpoint, or, as in linguistics, these vantages

as reflected in language. It should be noted that many scholars use

point of view

viewpoint

and

perspective

, interchangeably. If there is a difference among these terms, it is not

crucial here; instead, the current discussion focuses on understanding point of

view/viewpoint/perspective, and its overlapping traits with Linguistic Empathy.

Uspensky (1972) delineates how internal and external perspectives are invoked in

artistic expression and shared with the audience, arguing that the “behaviour of a human

being can be described either externally (in terms of his objective acts with reference to an

external observer), or with reference to his internal state, which is, generally speaking, not

accessible to outside observation” (p. 8). In literature, these perspectives can include the

thoughts, feelings, and experiences of individual characters, and in fine art, they can

include the vantage point of a painting. In order to successfully convey these perspectives,

the creator of the piece of work must portray perspective faithfully and accurately with

respect to the individual and any cultural expectations. It is the choice of the artist or the

author to convey perspectives internally or externally. With respect to literature, Uspensky

denotes four categories by which an external or internal perspective can be identified:

psychological characterization (i.e., the thoughts, feelings, and experience of a person),

verbal characterization (e.g., the dialect or the referential expressions that reflect a belief or

perspective), spatial and temporal characterization (i.e., spatiality: akin to Kuno’s visual

“camera angle;” temporality: whether the perspective is told in the present, which is

typically internal, or past, which is typically external), and ideological evaluation (i.e., the

beliefs and stances as reflected in the discourse).

Uspensky does not create a hierarchy of rules with respect to perspective in the arts;

instead, he describes and exemplifies how the conveying of perspective is realized using the

four categories above. These verbal means reflect perspectives that convey pragmatic

meaning beyond the mere truth value of an event. To diverge from the type of expression

that a character might use would miss these pragmatic considerations or even seem

unnatural, not unlike a violation of a Linguistic Empathy Hierarchy. This description of

perspective in the arts utilizes many of the principles that govern Linguistic Empathy, and,

significantly, Uspensky’s theory of perspective extends into Psychological Empathy as well.

More specifically, the narrator must place her/himself into the world that has been created,

she/he must understand what the character must be experiencing (Cognitive Empathy),

and she/he must convey these experiences with appropriate expression (Linguistic

Empathy) so that characters (and the audience) can have authentic emotional responses

(Affective Empathy).

The notion of point of view has additional applications within linguistics. Kuroda

(1973) reports that point of view is traditionally considered either reportive, where an

outsider describes her/his perception of another’s point of view (e.g.,

[I think that] Grace is

tired

) or nonreportive, where the point of view is reported directly from the source (e.g.,

Grace: “I am tired”

). Kuroda’s assessment corresponds with Uspensky’s notion of internal

versus external perspective. DeLancey (1981) argues that viewpoint manifests in language

through the implications of words (e.g.,

bring

from the viewpoint of the destination,

take

from the viewpoint of the object being taken) and structures (e.g., active versus passive

constructions), and that viewpoint manifests semantically in natural attention flow.

Natural attention flow is a focal point of attention that is given to the prototypical or

natural unfolding of the predicate, and it most naturally begins with the agent of a

sentence and flows to the patient. This semantically-defined focal point differs from

Yokoyama’s (1986) and Schmid’s (2007) centers of current concern/attention, which are

defined more broadly by the discourse situation, including inter-speaker pragmatic history

and context. DeLancey argues that contradictions between viewpoint and the natural

attention flow result in unacceptable sentences. This, he argues, explains why

The bear ate

the salmon

is more natural than

The salmon was eaten by the bear

: the flow of attention

naturally begins with the

eater

and finishes with the

eaten

Oshima (2007b) identifies how perspective takes shape in natural language through

construction alternations, anaphora, deixis, semantics-pragmatics interface, narrative

styles, and discourse. Oshima discusses that perspective is a term that is often used across

linguistics in order to describe multiple linguistic constructs (i.e., deixis, logophoricity,

Empathy, etc.), and that these constructs are often wrongly equated with one another.

Oshima (2007b) argues that although there is significant overlap among deixis,

logiphoricity, and Empathy, it is necessary to distinguish these constructs so as to avoid

confounding them. For example, Oshima teases apart the complexities of

deictic center

from

the

Empathy locus

, which other scholars (Culy, 1997; Sells, 1987) label as equivalent.

Deictic center

refers to a reference point that is “the object relative to which the meaning

(content) of a deictic expression is determined” (Oshima, 2007b, p. 23), whereas

Empathy

locus

refers to the physical or conceptual center from which perspective is experienced by

the entities in a sentence. Although these often coincide and the meaning is typically

derived from the perspective that the speaker identifies with, Oshima exemplifies and

summarizes that despite these significant correlations, they operate under different

guidelines and reflect different notions. For a discussion that parses these concepts through

examples from Japanese, see Oshima 2007b (ff p. 22).

Oshima’s approach is to tease apart related phenomena that are associated with

perspective while establishing a correlation between them. This helps to establish, predict,

and utilize individual differences toward understanding and applying Empathy, which is

not unlike the approach in this dissertation to attempt to correlate Linguistic and

Psychological Empathy. Although the definitions of Linguistic and Psychological Empathy

vary, the core principles of the varying points of view can be distinguished through

explication and example. Point of view is an essential component of Linguistic and

Psychological Empathy, whether discussed in a social context, as a literary device, or

analyzed in linguistic terms.

2.2.4 Sentential topics. The Empathy perspective of an utterance can be influenced

through contextual priming, in which entities mentioned in previous discourse become

relevant for future discourse. Kuno (1987) uses the term “topicalization” to refer to this

inter

-sentential phenomenon. However, “topicalization” traditionally refers to the

phenomenon of

intra

-sentential movement of an argument to the topic of a sentence. In

English, this corresponds to the leftward movement of an expression, as in a sentence like

You still have much to learn

transforming into

Much to learn, you still have.

This

dissertation adopts Kuno’s use of this term to refer to inter-sentential contextualization.

Previous discourse can contribute to competition among sentential entities to be

inserted as the topic of current discourse. For the hearer, this can prime an expectation for

certain entities to be the topic of the following sentence. In English, this occurs as

competition for subjecthood. As reviewed in section 2.2.2, Kuno establishes an Empathy

Hierarchy (EH) to manage the effect of context on Empathy perspective by claiming the

following:

(2.8) Topic EH: Given an event or state that involves A and B such that A is

coreferential with the topic of the present discourse and B is not, it is easier for

the speaker to empathize with A than with B.

E(discourse topic) ≥ E(nontopic), (Kuno, 1987, p. 210)

a. Observed:

Did you hear what happened to John? He got into a car accident

with Anna.

b. Violated:

Did you hear what happened to John? (?) Anna got into a car

accident with him.

In this definition of the EH, Kuno suggests that making an entity relevant in previous

discourse increases its right to be empathized with. However, Kuno does not elaborate on

whether “topic of the present discourse” refers to explicit mentioning of a topic or making it

relevant by some other means. A strict interpretation of “topic of present discourse” could

refer only to entities that have been explicitly mentioned previously, as in (2.8a) and (2.8b).

In these examples, the speaker mentions

John

in the first sentence (i.e., context sentence),

which explicitly makes

John

a relevant discourse topic.

In addition to an explicitly referenced topic, it is possible for a topic to be made

relevant through other entities in the sentence, as in

I took the bus to Long Beach.

Unfortunately, the driver didn’t have change

. This example derives from Prince (1981, pp.

233 ff.), in which she proposes

assumed familiarity

of knowledge and connotations between

interlocutors. In this example, the phrase

took the bus

suggests that other entities and

actions were involved in this event. Clark (1975) describes this phenomenon as

bridging

, in

which the relevant but implicit information derives from a series of implicatures. For

example, a bus, a bus driver, paying fare, and other passengers are primed as candidates

for being involved in this event. Similarly, in Fillmore’s (1976) notion of frame semantics,

bus/driver/money are all interrelated and derive meaning from one another in the same

semantic frame. This knowledge is culturally shared by the interlocutors, hence they are

reasonable impositions. Thus, without explicitly mentioning these other entities, the first

sentence makes other entities, such as

the driver

, valid as the topic of the second sentence.

Along with topicalization through previous explicit reference and through semantic

priming, an entity can be “coreferential with the topic of present discourse” if it is of mutual

current concern between the speaker and the hearer. Yokoyama (1986) argues that the set

of notions {

you, me, here, now

} is the minimum set of items that is required to be mutually

acknowledged between interlocutors. As a result, the entities

you

(hearer) and

I/me

(speaker) are, by default, relevant “topic of the present discourse” in a typical exchange

between interlocutors. Compare the following sentences in (2.15a) and (2.15b), in which

David

is the only entity mentioned in the previous discourse.

(2.15)

David definitely likes to exercise.

You first met him at a gym, after all.

Surface Structure EH: E(you) > E(David)

Person EH: E(you) > E(David) or N/A

Topic EH: E(David) ≥ E(you)

He first met you at a gym, after all.

Surface Structure EH: E(David) > E(you)

Person EH: E(you) > E(David) or N/A

Topic EH: E(David) ≥ E(you)

Examples (2.15a) and (2.15b) follow the same preceding context sentence in which

David

is the only sentential entity that is explicit. Therefore, according to the strict reading

of the Topic EH,

David

should be easier to empathize with than

you

. Additionally, the

Surface Structure EH dictates that the entity in subject position is the preferred recipient

of the Empathy perspective. If these two factor together, then (2.15b) should be preferred to

(2.15a). However, according to a Yokoyama’s notion of mutually acknowledged context,

you

is an entity that is of mutual concern, and therefore, it is a relevant “topic of the present

discourse.”

This updated understanding of the Topic EH, where “present discourse” includes

{

you, me, here, now

}, along with the further delineation of the Person EH

, would suggest

that both (2.15a) and (2.15b) should have complications resulting from conflicts in EHs.

Indeed, the acceptability of (2.15a) and (2.15b) is hardly distinguishable. Thus, candidates

for topics that are “coreferential with the topic of the present discourse” should include

entities that are explicitly referenced, entities that are suggested frame-semantically, and

entities that are part of mutual current concern. In all of these cases, competition could

arise among multiple allowable topics of following discourse. Depending on the strength of

the topicalization, certain contexts could create expectations for certain entities to be

preferred over others.

2.2.5 Passive constructions and reciprocal verbs. Passive constructions and

reciprocal verbs are distinctive in English in that they can alter the surface structure of a

sentence while maintaining the logical truth value. The active/passive alternation in

sentences like

Danny likes Alison

and

Alison is liked by Danny

does not alter the logical

Kuno’s (1987) formulation of the Person EH, as represented in (2.9), recognizes only the

speaker (first person) versus other (non-speaker). In this formulation,

you

would not be

higher on the Person EH than David; they would both be considered “other”. However,

other formulations of the human hierarchies (e.g., Deane’s entrenchment hierarchy [1992],

Silverstein’s animacy hierarchy [1976], explicitly suggest that second person you should be

higher than proper names like David. Kuno and Kaburaki (1977) also make this distinction

in the Person EH by positing that speaker > 2nd person > 3rd person. Additionally,

Yokoyama’s (1986) notion that {

you, me, here now

} are necessarily in mutual current

concern to start a conversation suggests that

you

should be higher on the Person EH than

an unreferenced third person entity.

equivalence of these sentences; similarly, the alternation of subject/object in sentences with

reciprocal verbs like

Joey married Jessica

and

Jessica married Joey

does not alter the

logical equivalence of these sentences. Reciprocal verbs usually refer to verb forms that

convey “each other” in the meaning (e.g.,

Joey and Jessica married each other

). In some

languages, the reciprocal pronoun can be obligatory, as in the Spanish version

Joey y

Jessica se casaron.

In this study, reciprocal verbs are discussed without the explicit use of

“each other,” following the implementation of “reciprocal verbs as Empathy verbs” from

Kuno (1987, pp. 209 ff.). As previously discussed, Kuno’s Surface Structure EH argues that

a speaker expresses identity with the subject of a sentence, and the sentential subject is the

Empathy locus. When there are possible alternations in the structure of a sentence, as in

passive constructions and reciprocal verbs, then the speaker chooses the subject of the

sentence based on which entity is the speaker Empathizes with. Since a sentence’s

structure is a common means of reflecting a speaker’s Empathy perspective, this section

establishes that usage of passive constructions and alternations in reciprocal verbs provides

speakers a structural choice in encoding Empathy perspective.

Passive constructions are often taught as a stylistically less preferred option in

English because they can be less direct or more difficult to follow (Strunk & White, 1999

[1958]; Ferreira, 1994). However, the notion that passive constructions should be avoided is

not universally true. Let us revisit examples (2.7a) and (2.7b) above, which contain an

active sentence that is transformed into a passive sentence:

(2.7) a.

I like Gary.

(?) Gary is liked by me.

I argue that the passive voice in (2.7b) is not

per se

the primary cause of its marginal

acceptability. In these examples, and indeed in all passive constructions, one of motivations

for passivization is the Empathy locus on the patient rather than the agent of the predicate.

Note that when (2.7b) is compared with another sentence in the passive voice like

Gary is

liked by his peers

, the latter is more acceptable. The difference between

Gary is liked by me

and

Gary is liked by his peers

is that the former violates Linguistic Empathy through a

conflict between the Surface Structure EH (E(

Gary

) > E(

)) and the Person EH (E(

) >

Gary

)), and the latter does not. Furthermore, many consider the passive

I was just

mugged (by someone)!

to sound more natural than the active form,

Someone just mugged

me!

. When the first person pronoun is the subject (

I was just mugged!

), passive is more

acceptable than when the indefinite pronoun is the subject (

Someone just mugged me!

)

despite the passive construction because the speaker is the Empathy locus in the passive

form. Thus, a speaker uses a passive construction when he/she identifies with a sentential

entity that is not the agent of the proposition.

Similarly, reciprocal verbs also allow for speaker choice when establishing Empathy

perspective. Kuno argues that sentences with reciprocal verbs “share one characteristic

with passive sentences: the subject of these sentences has been chosen as subject by the

speaker’s design” (1987, p. 211). Reciprocal verbs, which include verbs like

meet

date

, and

fight

, are verbs whose subject and object reflect equal participation in the action. As a

result, reciprocal verbs can often have multiple constructions which all share the same

logical truth value. For instance,

Bob is dating Nancy

Nancy is dating Bob

, and

Nancy and

Bob are dating (each other)

all describe the same situation. However, Kuno’s Surface

Structure EH argues that the subject of the sentence is determined by the identification of

the speaker with sentential entities. In other words, the Empathy perspective determines

the subject of the sentence. So, while the sentences are all identical semantically, the

pragmatic motivation differs.

Sentences with reciprocal verbs and sentences with passive constructions differ in

that passives have the additional markedness of a change of grammatical voice (e.g.,

likes

becomes

is liked

). Additionally, when a transitive verb is converted into a passive

construction, the agent of the sentence is no longer mandatory but may be added after the

predicate as a

-agentive. The example above

I was just mugged

does not require the

-agentive, and in fact some I argue that adding the indefinite and anonymous pronoun as

the agent, as in

I was just mugged by someone

, sounds less acceptable than simply

I was

mugged

. Despite these differences, passive constructions and reciprocal verbs share the

property of speaker choice as the determining factor for the sentence’s final structure, and

that this choice directly encodes the Empathy perspective of the utterance.

2.2.6 Universal and individual notions of Empathy. Yokoyama (1986, pp. 167 ff.)

depicts Linguistic Empathy into a dichotomy of phenomena: Anthropological Empathy and

Personal Empathy. Anthropological Empathy is an objective phenomenon that aligns with

the notions of Empathy established by Kuno in which the Empathy locus is based on

grammatical (e.g., subjects of passive predicates) or semantic (e.g., human vs. non-human)

categories. Like the hierarchies discussed above (e.g., Silverstein, 1976; Kuno, 1987; Deane,

1992), Anthropological Empathy is governed by the “general human tendency to identify

with certain referential and semantic categories over others” (Yokoyama, 1986; p. 167).

Yokoyama continues by establishing the expression of Anthropological Empathy as a

universal phenomenon across languages that does not vary depending on prior knowledge

or the relationship between interlocutors.

Personal Empathy, on the other hand, is a subjective phenomenon that is distinctly

individualized to reflect both the relationship between interlocutors as well as the attitudes

and emotions toward sentential referents. This is manifested by speakers’ contributing an

“egocentric display of personal factors” that may conflict idiosyncratically with the

universal hierarchies of Anthropological Empathy (Yokoyama, 1986; p. 167). An important

distinction is necessary between egocentricity as it relates to animacy hierarchies (i.e.,

first

person

second person

, etc.), and egocentricity as it relates to personal factors. These

personal factors can help to determine appropriate identification with NPs that are

otherwise equal on the Silverstein Hierarchy with a speaker’s personal relationship with

these entities (e.g., two proper names, one of whom is in the speaker’s in-group), as in (2.16)

below. These factors can also place the Empathy locus on an item that should otherwise

violate the EHs, as in (2.17) below. Consider the following examples:

(2.16) Context: A mother (B) conversing about her son, Dylan, and his wife,

Carolyn, with a friend (A) familiar with everyone involved (i.e., {A and B’s

knowledge and A and B’s current concern include: +Dylan, +Carolyn,

+marriage})

a. A:

So how did they meet?

They met (each other) while volunteering for Americorps.

b. A:

So how did they meet?

Dylan met Carolyn while volunteering for Americorps.

c. A:

So how did they meet?

(?)

Carolyn met Dylan while volunteering for Americorps.

(2.17) Context: A father speaking to his young daughter:

Daddy wants you to chew with your mouth closed

For sequence (2.16a-c), the friend (A) knows that her interlocutor’s son,

Dylan

, is

married to

Carolyn

. Because of the mutual knowledge of B’s greater referential proximity to

her son (i.e.,

Dylan

Carolyn

), it is only appropriate for

Dylan

to be the subject of the

sentence, be it individually as in (2.16b) or as a compound as in (2.16a). Placing

Carolyn

the subject position in (2.16c) is grammatically acceptable and valid according to the

Silverstein Hierarchy and Kuno’s EHs; however, it violates the speaker’s Personal Empathy

when the speaker places the Empathy locus away from her son,

Dylan

. Furthermore, in

example (2.17), the use of

Daddy

as self-referential for the speaker should also be a

violation of the Silverstein Hierarchy and Kuno’s Person EH. In this example,

Daddy

presumably the referential expression that the hearer (the speaker’s young daughter)

would use, which suggests that the speaker is adopting the perspective of the hearer over

his own. This is a phenomenon that was previously expressed by Yokoyama (1994, p. 4).

Personal Empathy considers the relationship between the father and his young daughter,

which reflects the father’s preference to share his daughter’s perspective. Personal

Empathy accounts for this sort of variation where other EHs do not. Similarly, the regular

violations of the EHs that occur in Russian and Polish, as introduced by Uspensky (1972,

pp. 20-21), Yokoyama (1994), and Christensen (1994, pp. 39-41), can be accounted for by the

empathic associations posited by Personal Empathy.

Additionally, Personal Empathy and discourse topic mutually influence one another.

Yokoyama (1986, pp. 59 ff., 255 ff.) points to the phenomenon of imposition, which is the

result of a speaker unilaterally deciding that a referent or proposition of personal current

concern is part of mutual concern. Yokoyama argues that the speaker and the hearer need

not experience a discourse situation identically. In the case of imposition, the state of the

addressee’s mind as it is and as it is assumed by the speaker are not identical, and the

speaker commits an “assessment error” (pp. 31 ff.). An imposition, nevertheless, does not

disrupt the exchange as long as the hearer accepts this imposition as reasonable. For there

to be an imposition between interlocutors, their relationship and the mood of the discourse

must permit such an imposition; otherwise, the addressee can refuse to accept it and the

transaction fails, exposing the speaker’s violation of a sociolinguistic (or Psychological

Empathic) norm. In interactions where there are multiple eligible discourse topics (e.g.,

multiple items relevant frame-semantically or multiple items mentioned in previous

discourse), and the item that the speaker chooses as topic is not of current concern for the

hearer, these items are reasonable candidates for being accepted impositions. When there is

less relevance, the imposition becomes more jarring to the hearer. Imposition and Personal

Empathy coincide when a topic of discourse that is of personal concern, be it a referent

and/or a proposition, is hoisted into the conversation as relevant and appropriate. In these

cases, Personal Empathy and imposition can help to validate violations of Kuno’s EHs. For

an example, let us consider the following:

(2.18) Context: The speaker (A) has waited quite a long time to get married to her

boyfriend (C) and, at long last, they have gotten engaged. The hearer (B) has

knowledge that includes {+A dating C, +A waiting to get married, +C has a

fear of commitment, - A and C have gotten engaged.}. As indicated by the

sentence

Guess what?

, A’s dating life is not part of B’s current concern prior

to the following exchange.

Guess what? He’s finally marrying me!

Wow, congratulations! That’s so exciting!

The utterance

He’s finally marrying me

, when preceded by a null context like

Guess

what?

, is marginal or unacceptable. This utterance is in violation of Kuno’s Person EH.

However, Personal Empathy and imposition help to validate the utterance. The imposition

of referential knowledge of the pronoun

in this utterance, despite being impositional and

not objectively appropriate, is retroactively validated by B’s utterance. The

that A utters

can only refer to A’s romantic partner (C) that exists in the knowledge sets of both A and B.

Secondly, there is an imposition of the disjunction

he will marry me

he will not marry

, which is made relevant by the use of the word

finally

and the accentuation of the verb

marrying

. This accentuation manifests an additional imposition of the information that the

question of marriage was up in the air until now, while the word

finally

shows that the

decision to get married was delayed because of the referent

. Typically, Linguistic

Empathy would deem that the speaker should be locus in a sentence with a reciprocal verb

as uttered by A, were it not for the Personal Empathy assumed to be shared between

interlocutors by means of speaker imposition. In example (2.18), the speaker commits an

assessment error in that she identifies

in the current concern of both A and B; however,

is not in the current concern of B. Thus, this is an example of a topic that is imposed

upon the hearer, and the influence of B’s Personal Empathy with A makes it possible to

ignore the violation of Anthropological Empathy, (i.e.,

He’s finally marrying me

instead of

I’m finally marrying him

). If the propositional knowledge that A was waiting to get married

was not reasonably imposable into B’s set of knowledge, then A’s utterance would likely be

considered unnatural.

2.3 Empathy in this Study

As discussed, Linguistic Empathy is a linguistic notion contained in narrated events,

as opposed to a cognitive notion like Cognitive Empathy or an emotional notion based on

the actions and feelings of individuals like Affective Empathy. A person’s ability to feel

Psychological Empathy influences his/her emotions, speech, actions, and understanding. By

contrast, Linguistic Empathy has thus far been considered a property of language that

reflects the extent to which a speaker identifies with people or entities in a speech event.

Linguistic Empathy has not been considered as a personal trait that can be measured. A

person with greater Psychological Empathy is considered more adept at managing and

processing Psychological Empathy stimuli; however, a person is not currently considered to

have more or less adeptness for Linguistic Empathy. In contrast, what I argue is that a

language user’s ability to recognize and manage phenomena within Linguistic Empathy can

be considered a quality of an individual. This quality exists as a speaker’s ability to convey

his/her identification with sentential entities and as a hearer’s ability to notice violations of

Linguistic Empathy. The varying degrees to which Linguistic Empathy is experienced by

speakers and hearers is never discussed explicitly, and this gap in the discussion provides

an opportunity to investigate the extent to which individual language users experience

Linguistic Empathy.

This dissertation’s approach aligns with the uncontroversial stance that Linguistic

Empathy is encoded in the linguistic structure by the speaker. However, I argue that in

addition to being a feature of linguistic structure, it is also a

measurable

trait in language

users. The measure of Linguistic Empathy in speakers would reflect the ability to convey

Empathy perspective naturally (e.g., the ability to adhere to rules like Kuno’s EHs or to

navigate notions like Yokoyama’s Personal Empathy) as much as it reflects the addressee’s

ability to evaluate sentences in terms of their adherence to it. In basic terms, this means

that a speaker and a hearer with a greater ability to manage Linguistic Empathy would

prioritize Empathy perspective when generating or processing an utterance, whereas a

speaker or hearer with less awareness of Linguistic Empathy would more easily neglect

Empathy relations and instead prioritize syntactic and semantic validity (e.g., she/he would

rate

I like Gary

and

Gary is liked by me

as equivalent). In this dissertation’s experiment,

only the hearer’s measure of Linguistic Empathy is under investigation; the speaker’s

measure of Linguistic Empathy will remain untested.

The extent to which a hearer’s ability to manage Linguistic Empathy is a

measurable and systematic phenomenon is under investigation in this dissertation. The

measure of Linguistic Empathy in hearer would reflect the ability to notice nuances in

perspective taking (e.g., imposition) and violations in Linguistic Empathy (e.g., conflicting

Empathy foci). For instance, if an addressee finds an utterance

Gary is liked by me

to sound

as natural as

I like Gary

, then I argue that this addressee understands the semantics of the

two utterances but not the significance in alternating perspective between the speaker

and the third person referent

Gary

. This person would have a lower sensitivity to violations

of Linguistic Empathy. Alternately, a person who finds the utterance

Gary is liked by me

less acceptable would register the negative effect of the speaker identifying with another

entity over her/himself. This person would have a higher sensitivity to violations of

Linguistic Empathy. Thus, if the measure of Linguistic Empathy in an addressee involves

understanding the Empathy perspective of the speaker and the pragmatic implications of

this perspective, then evaluating the ability to recognize when a speaker violates Linguistic

Empathy amounts to a valid test for this measure.

By conceiving of Linguistic Empathy as a measurable trait in individuals, the

overlap of Empathy across linguistic and social psychological notions can be compared. The

experimental approach this dissertation takes is to measure an addressee’s ability to

recognize violations in Linguistic Empathy. The experiment implemented in this

dissertation focuses on violations of Kuno’s Linguistic Empathy Hierarchies. Baron-Cohen

and Wheelwright (2004) support their measure of Psychological Empathy (Empathy

Quotient) as meaningful by comparing the results of the EQ test between participants with

Autism Spectrum Disorder (a related Psychological Empathy deficit) and participants

without this condition. This resulted in a significant correlation between lower EQ scores

and participants with high-functioning Autism. Similarly, this dissertation’s measure of

Linguistic Empathy can be shown as meaningful if the results correlate with a related

phenomenon. Psychological Empathy and Linguistic Empathy both incorporate the capacity

for adopting viewpoints and understanding others’ perspectives, and thus both involve

overlapping processes. More specifically, if a person has difficulty with shifting perspective

when language is not involved (i.e., low measure of Psychological Empathy), then this

difficulty will manifest in managing Empathy perspective in language. I hypothesized that

there exists a significant correlation between Linguistic Empathy sensitivity and EQ, and

that this correlation would validate the measure of Linguistic Empathy.

As presented, Empathy is sometimes conceived of as a social and philosophical

notion (e.g., Davis, 1994), and other times conceived of very mechanistically with a rigid set

of linguistic guidelines (e.g., Kuno, 1987). Beyond the mechanistic view, Linguistic

Empathy has been presented as having universal elements, cross-linguistic variability,

personal/idiosyncratic influence (Yokoyama, 1986, 2000), as well as suggested to have

cultural roots (Oshima, 2007b). Similarly, the notion of Psychological Empathy has been

approached more mechanistically in experimental settings and controlled settings (e.g.,

Baron-Cohen & Wheelwright, 2004). These notions arose independently to account for

phenomena that involve perspective taking within language and psychology. One goal of

this study is to operationalize Kuno’s mechanistic rules of Linguistic Empathy, which lend

themselves to an experimental setting, and then to investigate the correlations and

applications that these results have with respect to Psychological Empathy. The

experimental section of this dissertation examines the extent to which Psychological

Empathy and Linguistic Empathy influence one another.

CHAPTER 3

EXPERIMENTAL APPROACH TO EMPATHY

The goal of this chapter is to elucidate the rationale, predictions, and methodology for

the experiment in this dissertation. The chapter begins by illuminating extensions of

current Linguistic Empathy theory and gaps in previous approaches to Psychological

Empathy. The chapter argues in favor of experimentation that examines Linguistic

Empathy and Psychological Empathy together in order to maintain objectivity and to

investigate a potential correlation between the two forms of Empathy. Subsequently, the

chapter describes the process by which the linguistic stimuli were manipulated in order to

target the specific Linguistic Empathy Hierarchies. Based on these stimuli, the chapter

then submits predictions for the experiment, and concludes with the details of the

experimental methodology.

3.1 Rationale for Experimentation

In attempting to uncover new insight into Linguistic Empathy, one could either

expand upon existing theory using similar methodologies, or one could approach the theory

using an alternate methodology. In researching Linguistic Empathy Kuno and Kaburaki

(1977) and Silverstein (1976) follow a prominent and tested trajectory in linguistic

investigation. The first step is typically to highlight a range of linguistic phenomena in

sentences that have either been generated by the linguist or that derive from attested

discourse. Next, the linguist demonstrates that these phenomena are not be accounted for

in current theoretical understandings. Finally, the linguist accounts for these phenomena

in a nuanced or novel theory. However, since the current study pursues a different line of

investigation (i.e., a correlations and a measure of Linguistic Empathy as opposed to a

novel/nuanced theory of Linguistic Empathy), I have chosen a fundamentally different

experimental methodology.

When analyzing linguistic phenomena, it is standard for the linguist to make

acceptability judgments or for the linguist to solicit acceptability judgments from native

speaker informants. There is often variability in acceptability ratings of marginal

sentences; however, this variability is not often researched. A hypothesis of the current

study is that variance in acceptability does not only apply to the ambiguity of the linguistic

phenomena, but also applies to individual differences of those who rate the sentences. To

evaluate this hypothesis, this study diverged from the traditional approach to Linguistic

Empathy by creating and issuing experiments. This involved first generating language

samples based on existing parameters of Linguistic Empathy (i.e., not highlighting new

phenomena), and then conducting experiments that yield quantitative feedback regarding

these sentences. Instead of this data only applying to the theory of Linguistic Empathy, the

data also provides information about the individuals who participated in the experiment.

Once quantitative and qualitative patterns emerge with respect to the individuals, then the

results provide the measure of a person’s sensitivity to these Linguistic Empathy

phenomena.

The goal of Linguistic Empathy thus far has been to establish Empathy as a

linguistic phenomenon and to show that it operates in utterances generated by a speaker.

This study built upon this goal and explored the extent to which a viable measure of

Linguistic Empathy in individuals is possible. The present study attempted to conceive of

acquiring Empathy measures through a novel methodological approach, to be described

below. The study also determined how measurements of Linguistic Empathy correlate with

traditional measures of Psychological Empathy.

3.1.1 Measure of Linguistic Empathy.

As discussed, the goal of linguists who

research Empathy has been to establish and detail the theory of Linguistic Empathy using

their own language judgments or judgments from native speaker informants. These

judgments are then used in linguistic argumentation for establishing Linguistic Empathy.

The tacit assumption for linguists who follow this paradigm is that acceptability judgments

from native informants are valued equally, and that judgments can vary. It is assumed that

these differences reflect individual idiolects; however, differences in inter-subject

acceptability ratings are not typically investigated. Alternatively, the current experimental

approach evaluates whether inter-subject acceptability ratings are the result of a

systematic difference in language competence with respect to Linguistic Empathy

phenomena.

Linguists have not attempted to measure person’s capacity for noticing and adhering

to the principles of Linguistic Empathy, nor have they hypothesized that this capacity may

co-vary with measures of Psychological Empathy. Variation in acceptability judgments of

sentences with Linguistic Empathy violations are often attributed to “the strength that the

Empathy principles that have been violated have in each person’s idiolect” or “mitigating

circumstances can be readily found that would make up for the violation” (Kuno, 1987, p.

296). The individual differences in acceptability ratings have been acknowledged, but they

have not been researched. The experiment in this dissertation examines these individual

differences in acceptability ratings in order to provide insight into the strength of different

types of Linguistic Empathy violations, as well as to provide evidence for an approximate

measure of Linguistic Empathy within an individual. By determining patterns of the

ratings within an individual participant, this study approximated a person’s sensitivity to

violations of Linguistic Empathy in sentences that are otherwise grammatically correct. It

was hypothesized here that this measure of sensitivity to Linguistic Empathy correlates

with a person’s measurement of Psychological Empathy.

3.1.2 Measures of Psychological Empathy.

The Psychological Empathy measures

that were reviewed in the previous chapter have all faced criticism. Criticism of these

various measures typically focuses on the theoretical approaches of the tests, such as

separating sympathy and or only measuring certain aspects within Empathy. Nevertheless,

even if a test was argued to provide only portions of an Empathy measure, these criticisms

did not take issue with the accuracy of these measures. For example, despite their critique

of the Interpersonal Reactivity Index (IRI) as a measure of only Affective Empathy, Jolliffe

& Farrington (2006) claim that IRI results are meaningful, and they use correlations with

the results of IRI measures to validate another Empathy test. With respect to EQ,

Lawrence et al. (2004) verify that despite its flaws, “EQ provides a reliable and valid way of

measuring Empathy via self-report in both healthy individuals and clinical populations” (p.

919). Because of its widespread implementation, its validity in clinical and healthy

populations, and its accessibility, EQ was utilized as a measure of Psychological Empathy

for this dissertation.

One consistency across all of these assessment approaches is that they all employ a

self-report questionnaire. Chronologically, each measure attempted to some extent to

improve upon previous tests’ criticisms by creating fixes to these issues instead of taking an

alternative approach. Although these tests approach Empathy with varying definitions and

theoretical vantage points, they attempt to improve upon previous iterations while

maintaining this consistent format. Unfortunately, attempting to create a measurement

system using this standard format has created a culture for Empathy measures with

consistent limitations that are intrinsic to the format. More specifically, each test solicits

responses to questions that directly reflect Empathy and empathic behavior, and inherent

shortcomings of self-report questionnaires remain unaddressed. I argue here that

questionnaires attempting to accurately reflect Empathy are inherently flawed for three

primary reasons. Firstly, basing the measurement of a prosocial trait on direct, self-report

questions creates a conflict of interest when there is a disconnect between a person’s desire

to be portrayed positively and a person’s true feelings and behavior. Secondly, issuing a

questionnaire that measures an emotionally-founded trait like Psychological Empathy is

subject to fluctuate based on the emotional state of the participant. Finally, with regard to

test-retest reliability, the nature of a questionnaire familiarizes the participant with the

subject matter, thus creating difficulties for reissuing the questionnaire without a sizable

lapse of time.

With respect to the first issue, Empathy is a prosocial behavior, which causes an

inherent conflict for a person who does not want to self-incriminate; subjects might

recognize prosocial criteria in self-diagnostic questions, and they may not respond

truthfully if this answer conveys antisocial tendencies. Even if subjects can be trusted to

avoid this, a level of self-incrimination is built into these measures, which leaves open the

possibility of subconscious bias. Some examples of statements from the EQ test for

participants to agree or disagree with are: “I am very blunt, which some people take to be

rudeness, even though this is unintentional”, “other people tell me I am good at

understanding how they are feeling and what they are thinking”, and “friendships and

relationships are just too difficult, so I tend not to bother with them” (Baron-Cohen &

Wheelwright, 2004, pp. 172-173). People might imagine themselves differently than others

perceive their behavior, or people might simply hope or want to act differently than they

typically feel.

Some of these concerns were in fact reported in by Lawrence et al. (2004), who found

a correlation between EQ and social desirability (i.e., a disconnect between how people hope

to appear and how they actually feel/behave). Participants’ responses to the EQ test were

measured against their responses to the Social Desirability Scale (SDS), which is another

questionnaire designed to measure a person’s tendency to answer questions with socially

preferable responses (Crowne & Marlowe, 1960). The correlation was found in five of the 40

items on the EQ test (three of which were significant). Although a trend in a few

individuals who might answer questions with bias could raise questions, the fact that this

trend correlates across the entire population of the study raises concern for the impartiality

of the questionnaire.

It can be argued that the statistical significance of the test’s validity despite these

confounds proves the reliability of the EQ. Nevertheless, aside from Lawrence et al. (2004),

it is not standard practice to issue an SDS concurrently with the EQ test, or with any

Empathy measure. So, if an individual or group beyond that study were to answer

questions inaccurately caused by the bias of social desirability, this trend would likely go

unnoticed and unreported. Furthermore, while the statistics overwhelmingly suggest that

EQ is a reliable measure of Empathy across large number of subjects, Empathy

measurements are typically used for diagnostics in individuals with an Empathy deficit. If

an individual so desires to avoid a diagnosis of Autism Spectrum Disorder or psychopathy

in order to adhere to social norms, the bottom line is that issuing a test with overt and

direct questions creates a more explicit opportunity to do so. Rather than intending to

improve upon surveys that directly question a person’s Empathy, the potential for a test

that avoids the potential for conscious/unconscious social desirability should be proposed.

The second issue with respect to Empathy questionnaires is that since Psychological

Empathy is an inherently emotional quality, the state of a person’s emotions could

influence the results of the questionnaire. A person in an unusually happy or affectionate

state, for instance, could provide more prosocial responses than that if that same person

were in an unusually agitated state. Of course, a person’s emotional state is a variable in

any experiment; however, a questionnaire that asks direct questions of a person’s socio-

emotional tendencies is in direct conflict with emotional objectivity. Even if a questionnaire

attempts to recognize behavioral patterns as opposed to emotional states, surely there

would be potential for significant emotional impact.

The final issue with Empathy questionnaires deals with test-retest reliability.

Lawrence et al. (2004, p. 912) cite the efficacy of test-retest reliability of the EQ test when

measured one year apart. While this supports the reliability of the EQ test, it leaves room

for logistical difficulties. For instance, if a person wants to verify an EQ score, it hardly

seems viable to wait an entire year before retaking the test. There is only one version of

typical Empathy questionnaires, so other than shifting to a different type of Empathy test,

new questions could not be issued. These restrictions highlight the need for a test that

contains multiple possible versions with equivalent stimuli.

3.1.3 Benefits of linguistic experiments.

Linguistic experiments can be designed to

explore the issue of individual differences with respect to sentence acceptability ratings.

The experiment in this dissertation considers acceptability as a numerical gradient.

Standard treatment of sentence acceptability by linguists considers sentences as

unacceptable (designated with *), acceptable (no designation or √), or marginally acceptable

(designated by a single question mark, ?, or any number of them, e.g., ???). The current

study applies a similar gradient scale onto numerical values. Participants are instructed to

rate the sentences numerically on a scale from 1 (low acceptability) to 4 (high acceptability)

for all stimuli. Since the experiment design is quantitatively rich, ratings can be averaged

for individual sentences as well as across categories of sentences. Most importantly, this

design yields acceptability judgments that will be used to develop an individual measure for

Linguistic Empathy sensitivity.

Additionally, if Linguistic Empathy is demonstrated to correlate with Psychological

Empathy, then a Linguistic Empathy test could be developed to acquire measurements that

would approximate an individual’s Psychological Empathy. Psychological Empathy

measures are often used in clinical settings as a diagnostic tool for individuals at risk of

Empathy deficits (Lawrence et al., 2004; Rogers et al., 2006). Issuing a linguistic test to

assess a person’s Empathy would be a radically new approach to Psychological Empathy,

and it would avoid the three issues described above in section 3.1.2 that arise from the

questionnaire-based measurements. Firstly, with respect to social desirability, there is no

issue of self-incrimination of antisocial feelings and behavior when it comes to linguistic

acceptability judgments. As such, rating a sentence does not factor into social desirability.

Indeed, Rogers et al. (2006) conclude that a measure of Empathy with more objectivity and

less intrinsic conflict is needed. Secondly, linguistic ratings are an unemotional endeavor.

While linguistic experiments are prone to the same emotional variance as any experiment,

the subject matter is not inherently emotional, and therefore avoids the conflict created by

Psychological Empathy self-report questionnaires. Finally, retaking a test that involves

language acceptability is logistically more flexible than a fixed questionnaire. If it is

deemed invalid to participate in an experiment with the same linguistic stimuli, then

another set of stimuli that manipulate Linguistic Empathy can be issued. The ability to

adjust an experiment based on linguistic variables is more flexible than the closed set of

questions that ask about a person’s Psychological Empathy behavior.

3.2 Linguistic Manipulation of the Stimuli

By implementing psycholinguistic experimentation, this study attempted to account

for individual variability in acceptability judgments. The experiment applied the argument

that a person’s acceptability ratings can measure her/his capacity for recognizing particular

linguistic phenomena. The present study focused on phenomena within Linguistic

Empathy, in which sentences often sounded marginal or awkward to most, but were

grammatically correct. Patterns that emerge from these acceptability ratings would

demonstrate that a person’s capability for recognizing violations of Linguistic Empathy

phenomena is actually a measure of Linguistic Empathy sensitivity. This experiment also

investigated inter-subject variability and correlation with the Psychological Empathy

measure of Empathy Quotient.

3.2.1 Operationalization of Empathy Hierarchies. The experiment measured the

effect of manipulating two of the Empathy Hierarchies (EHs) proposed in Kuno and

Kaburaki (1977) and Kuno (1987): Person EH and Topic EH. These EHs are replicated

below:

(3.1) Person EH: The speaker cannot empathize with someone else more than with

himself.

E(speaker) > E(others), (Kuno, 1987, p. 212)

(3.2) Topic EH: Given an event or state that involves A and B such that A is

coreferential with the topic of the present discourse and B is not, it is easier for

the speaker to empathize with A than with B.

E(discourse topic) ≥ E(non-topic), (Kuno, 1987, p. 210)

The first hierarchy, the Person EH, argues that it is unnatural for a speaker to share the

Empathy perspective of another entity (e.g.,

you

s/he

) over the self (

). The Person EH

was manipulated through the use of target sentences with structures that permit the

reversal of the subject and object, i.e., reciprocal verbs and active/passive constructions,

e.g.,

I like Gary

versus

Gary is liked by me

. The second hierarchy, the Topic EH, argues

that when the topic of a sentence has been mentioned in or made relevant by previous

discourse, it is easier for the speaker to empathize with this entity than an entity that was

not mentioned nor relevant. The Topic EH was manipulated by including preceding context

sentences before the target sentences that either referenced a sentential entity (e.g.,

Let me

tell you about Gary

references

Gary

), or referenced no entity and were conceived of as

occurring ‘out of the blue’ (e.g.,

Guess what?

A third hierarchy, the Surface Structure EH, was also relevant in creating the

stimuli; however, it was not manipulated as a variable in the experiment. Kuno formulates

the Surface Structure EH as follows:

(3.3) Surface Structure EH: It is easier for the speaker to empathize with the referent

of the subject than with the referents of other noun phrases (NPs) in the

sentence.

E(subject) > E(other NPs), (Kuno, 1987, p. 211)

The Surface Structure EH posits that the subject of a sentence is the most natural

Empathy locus. Instead of being manipulated as an experimental variable, this EH served

as a reference point for conflicts of observation/violation of the Person EH or the Topic EH.

When there was a violation of the Person EH or the Topic EH in this experiment, these

violations conflicted with the Surface Structure EH. For example, in a sentence like

Gary is

liked by me

, there is a conflict between the Person EH and the Surface Structure EH. The

Person EH identifies the speaker

as the Empathy locus (E(

) > E(

Gary

)), and the

Surface Structure EH identifies

Gary

as the Empathy locus (E(

Gary

) > E(

)). However, in

the sentence

I like Gary

, there is no such conflict: The Empathy relation is E(

) > E(

Gary

)

for both the Person EH and the Surface Structure EH.

Across these examples, Surface Structure EH is not manipulated because of the

strength that subjecthood has in conveying Empathy perspective in English. However, the

relationship between subjecthood and Linguistic Empathy cannot be trusted universally.

The strength of subjecthood is varies in other languages, especially those with

morphological case marking and freer word order, like Slavic languages (Yokoyama, 1986),

Native American languages like Navajo, Cree, or Jinghpaw (DeLancey, 1981; Oshima,

2007a & 2007b), or Japanese (Kuno, 1987; Oshima, 2007a & 2007b). In these languages,

while Linguistic Empathy is a major factor affecting the linguistic form of the sentence, it

does not necessarily depend on the choice of the grammatical subject the way it does in

English. Since English is an SVO language, the speaker typically chooses a sentence

structure in which the referent the speaker empathizes with is the grammatical subject.

Therefore, the Surface Structure EH is automatically observed, and when there is a conflict

between the Surface Structure EH and either the Person EH (e.g.

I like Gary

versus

Gary is

liked by me

) or the Topic EH (e.g.,

Let me tell you about Gary. I like him

), the Surface

Structure EH is observed, and the other EH is violated.

3.2.2 Target sentences and context sentences. The stimuli used in this experiment

consisted of two types of sentences: target sentences and context sentences. Target

sentences were designed to manipulate the Person EH. These were the sentences that the

participants rated, and they consisted of two structurally different groups: a Reciprocal

group and an Active/Passive group. Stimuli were presented to the participants as if an

external speaker were uttering all sentences in the stimuli, as opposed to the participants

imagining the utterances as self-generated. The Reciprocals contained sentences with

reciprocal verbs like

meet

date

fight

, in which the truth value of the sentence did not

depend on which of the two participants in the sentence was chosen as the grammatical

subject; if

I dated Terry

is true, so too is

Terry dated me

. (Reciprocal verbs also allow plural

subjects such as in

Terry and I dated [each other]

, but this option was not considered.) In

the Active/Passive group, the alternation between active and passive constructions was

used to maintain the truth value across sentences; if

I like Sherry

is true, so is

Sherry is

liked by me

. Stimuli in both the Reciprocal and Active/Passive groups were limited to

sentences in which the first person

I/me

was one of the sentential entities, and a third

person proper name was the other. In the Active/Passives, the grammatical subject of the

active version was always

Context sentences preceded target sentences and provided background for the

participants to interpret the target sentence. Designed to manipulate the Topic EH, context

sentences were divided into two categories: null context and priming context. A null context

corresponds with the beginning of an interaction or with utterances introduced by phrases

Guess what, did you hear the news?

A null context merely establishes the mutual

willingness of the interlocutors

and

you

ready to engage at a given time and place without

designating a specific topic. Yokoyama (1986, pp. 31 ff.) argues that at a minimum, the

topics {

you

here

now

} are relevant to the discourse because both interlocutors must have

them in their current concern in order to communicate. Thus, the topics

and

you

should be

included in the set of things that are “coreferential with the topic of present discourse”

(Kuno, 1987, p. 210), as the Topic EH stipulates, even if they are not explicitly referenced.

This means that when

is the subject of a sentence that follows a null context, the sentence

observes the Topic EH (e.g.,

Guess what? I like Gary

). When a third person entity is the

subject of a sentence that follows a null context, it violates the Topic EH because this entity

does not have a legitimate reason to be part of mutual current concern (e.g.,

Guess what?

Gary is liked by me

In priming contexts, third person entities were explicitly referenced in order to

create the expectation that these entities could be used as a topic in future discourse. For

example, the relevance of

Gary

is elevated by the content of the sentence

Let me tell you

about Gary

. Thus, when

Gary

is mentioned, it becomes a candidate for topic of a sentence

that follows; the mutual current concern is revised and now contains both

and

Gary

Which of the two is chosen by the speaker as the focus of Empathy thus becomes contested.

Let us consider the following sentences:

(3.4)

Guess what, did you hear the news? Gary is liked by me.

a. Surface Structure EH: E(

Gary

) > E(

)

b. Person EH: E(

) > E(

Gary

)

c. Topic EH: E(

) ≥ E(

Gary

)

(3.5)

Let me tell you about my friend Gary. He is liked by me.

a. Surface Structure EH: E(

) > E(

)

b. Person EH: E(

) > E(

)

c. Topic EH: E(

) ≥ E(

)

In example (3.4) there is no mention of any entity in the context sentence

Guess

what, did you hear the news?

, so the relevant entities for topic in the following discourse

derive from the NPs of mutual current concern, i.e., {

I, you

}. Thus, the Empathy relation of

the Topic EH is in conflict with the Surface Structure EH in (3.4). A sequence with a

priming context like (3.5)

sounds better than the equivalent sequence with a null context

like (3.4) because there is no conflict between the Topic EH and the Surface Structure EH

in the priming context (though it is still marginal because the Person EH is in conflict). In

(3.5),

Gary

is mentioned in the context sentence, which enters

Gary

into mutual current

concern. Thus,

Gary

is a relevant entity to be the topic of the following discourse.

A sequence similar to (3.5) would sound even more natural if there were further

contextualization of the third person referent

Gary

. For example, let us consider the

following:

(3.6)

Let me tell you about my friend Gary. He’s liked by his peers, he’s liked by

his friends, and he is liked by me.

The final phrase in (3.6),

he is liked by me

sounds better in this example than in (3.4) or

(3.5) because of the extended context of which

Gary

is clearly the topic. Furthermore, this

extended context arguably makes

Gary

more competitive than

to be the topic, as in the

following:

(3.7)

Let me tell you about my friend Gary. He’s liked by his peers, he’s liked by

his friends, andI like him.

If the target sentence

I like him

in (3.7) were judged as marginal, this would confirm that

long contextual strings with a heavily emphasized topic can prioritize

him/Gary

over

for a

clear violation of the Topic EH. However, the practical considerations for this experiment

prevented the design from including stimuli with this type of extreme contextual priming.

Instead, the strength of the contextual priming was tested in sequences that resemble those

like (3.5) in which the expectation for the explicitly mentioned topic,

Gary

, creates

competition between

Gary

and the unverbalized topic of mutual concern,

. I stipulate that

when the explicitly referenced entity is topic of the following discourse, the Topic EH is

observed, like in example (3.5). Contrarily, when the topic of the following discourse is

this competition triggers a violation of the Topic EH, as in (3.8) below:

(3.8)

Let me tell you about my friend Gary. I like him.

A violation to the Topic EH occurs in two conditions. Let us consider the following

sentence quartet in order to contrast these conditions using the following examples of

stimuli from the experiment:

(3.9)

Matt can be such a bully. This morning, I fought Matt.

(3.10)

Matt can be such a bully. This morning, Matt fought me.

(3.11)

Guess what happened this morning, I fought Matt.

(3.12)

Guess what happened this morning, Matt fought me.

The first type of Topic EH violation corresponds with the previous explanation of (3.5) and

(3.6), and applies to (3.9). This violation occurs when a sequence with a priming context is

followed by a target sentence with

as the subject. In (3.9) and (3.10),

Matt

is referenced in

the context sentence. Once this entity is introduced into the discourse, it creates

competition with the other relevant entities {

I, you

}. This competition creates a conflict

between the expectation of what the topic will be and what is uttered. Sentences like (3.9),

for this reason, are considered in violation of the Topic EH according to this experiment

design.

Sentences like (3.10) observe the Topic EH because

Matt

is the subject of the target

sentence. The second type of Topic EH violation occurs when a sequence with a null context

is followed by a target sentence with a third person referent. The contexts in (3.11) and

(3.12) contain no competition for topic; of the entities in the target sentence, the topic

Like with all reciprocal verbs, some readers might be tempted to apply an agentive

reading to the subject of

fight

, so that the utterance means something like

Matt picked a

fight with me

Matt assaulted me

. However, I argue that

fight

entails reciprocal

confrontation between the two sentential actants, as evidenced by the semantic oddness of

the utterance

Matt fought me, but I did not fight Matt.

If both participants do not engage in

the

fighting,

then the attempt to

fight

is unsuccessful.

the only viable option for the subject of the sentence. Thus, (3.11) observes the Topic EH

since

is the subject of the target sentence, and (3.12) violates it since

Matt

is the subject of

the target sentence. The two types of violations can be contrasted between examples (3.9)

and (3.12). In (3.9), there is competition for topic of the context sentence (i.e.,

Matt

explicitly referenced,

is not referenced but is relevant by default;

is used as topic), and in

(3.12), there is no competition (i.e.,

Matt

is not referenced and not relevant,

is not

referenced but is relevant by default;

Matt

is topicalized).

Target sentences within the same item group have obvious differences caused by

reciprocal or active/passive alternation; for example, the difference between the target

sentences in (3.9)

I fought Matt

and (3.10)

Matt fought me

is the alternation of

Matt

and

I/me

. Aside from this difference, it was necessary to design the target sentences to be as

homogenous as possible in order to isolate these experimental variables. As a result, the

effort toward minimizing differences across sentences resulted in some stimuli with

constructions that were less than ideal. Let us consider the alternation between (3.9)

fought Matt

and (3.10)

Matt fought me

. The use of the name

Matt

in both of these sentence

constructions was necessary to compare these alternations. However, when the sentence

was preceded by the priming context, as in (3.10)

Matt can be such a bully. This morning,

(Matt fought me)

, the mentioning of the name

Matt

in the context sentences creates a

possibility (and a preference) for a pronoun to be used in the target sentence. However,

instead of using the pronoun

, the name

Matt

is repeated. Although the repetition of the

proper name

Matt

can affect the acceptability of the stimulus, it was necessary in order to

keep sentences homogenous within item groups. Indeed, Gordon, Grosz, and Gilliom (1993)

demonstrate the

repeated-name penalty

, in which proper names that are the center of focus

and are repeated present more difficulty in processing than their pronominal counterpart.

This issue was also dealt with in some stimuli with use of backward

pronominalization, in which a pronoun appears before a coreferential proper name instead

of the proper name appearing first and the pronoun appearing subsequently. In these

stimuli, the entity was referenced in the priming context using a pronoun, and the target

sentence contained the proper name. For example, a stimulus with backward

pronminalization resembles the following:

He can be such a bully. This morning, Matt

fought me.

Cole (1974) discusses how sentences with backward pronominalization are often

considered marginal, and he argues that sentence acceptability in these sentences can

depend on syntactic and semantic constraints, as well as intonational variation and

individual preferences. The use of backward pronominalization eliminated the issue of

repetitive reference to the same person using a name in subsequent sentences; however, it

introduced its own potentially marginalizing effects.

There was another type of context sentence that used parallel grammatical subjects,

which is exemplified in (7) and (8) from

Table 3.1

below. This type of context does not have

the drawbacks of repeating the proper name or of backward pronominalization. Parallel

grammatical subjects occur in sentences such as

Jack fell down and Jill came tumbling

after

with parallel grammatical subjects (

Jack

and

Jill

) belong to a previously established

topic set (Yokoyama 1986, 314 ff.). Coming after

Jack fell down

Jill

in the coordinated

parallel sentence (

and Jill came tumbling after

) is recognized as a member of a topic set

{

Jack, Jill

}. In the same way, coming after

Dan and Sue have had quite a busy morning,

Dan got a haircut,

in the coordinated parallel sentence (

and Sue was massaged by me

)

should be recognized as a member of a topic set {

Dan, Sue

}. Both the explicit mention of

Sue

and the implication of the topic set affect interference between the two entities now high on

the Topic EH (

and

Sue

3.2.3 Conditions of the stimuli. The sentence quartets in the experiment gave rise to

four stimulus types, which are defined as follows: the observation of both EHs (condition A),

the violation of the Topic EH and the observation of the Person EH (condition B), the

observation of the Topic EH and the violation of the Person EH (condition C), and the

violation of both EHs (condition D). The four stimulus types are abbreviated to the

condition labels of {A, B, C, D} so that they can be compared across Reciprocal and

Active/passive groups, as well as placed into graphs and analyses in the experimental

results. Participants were asked to rate the acceptability of the target sentences given the

previous context sentence.

Table 3.1

, stimulus examples are organized by the type of context, target,

violation (+ or -), and condition (A, B, C, D). The examples of a null context for a reciprocal

verb

fight

are presented as (1) and (2), and examples of a priming context for the same

reciprocal verb are (3) and (4). Examples of a null context for an active/passive pair with

verb

massage

are (5) and (6), and examples of the priming context for the same

active/passive verb are (7) and (8).

3.3 Predictions

3.3.1 Stimulus types and Empathy Hierarchies. In the Reciprocal group, the Person

EH was manipulated by creating stimuli in which the grammatical subjects and objects

alternated between first and third persons. In the Active/Passive group, the Person EH was

similarly manipulated by creating stimuli in which the object of an active sentence was

used as the subject of the passive version, and the subject of an active sentence was used as

the

-agentive in the passive version. I predicted that in null contexts, in which

established as a default topic, target sentences with first person grammatical subject (e.g.,

fought Matt

I massaged Sue

) should be more acceptable than targets with third person

Context

Target sentence

Violations

T P

Condition

Label

Guess what happened this morning?

I fought Matt.

+ +

Guess what happened this morning?

Matt fought me.

- -

Matt can be such a bully. This

morning,

I fought Matt.

- + B

Matt can be such a bully. This

morning,

Matt fought me.

+ -

Good to see you. Can I tell you about

what happened last night?

I massaged Sue.

+ +

Good to see you. Can I tell you about

what happened last night?

Sue was

massaged by me.

- -

Sue and Dan have had quite a busy

day. Dan got a haircut, and

I massaged Sue.

- + B

Sue and Dan have had quite a busy

day. Dan got a haircut, and

Sue was

massaged by me.

+ - C

Table 3.1: Example Trials, Violations, and Conditions

Table 3.1:

Example trials for all types of stimuli (1-4 Reciprocal, 5-8 Active/Passive) with

columns for contexts, targets, Topic and Person EH observation (+) or violation (-), and

condition label.

grammatical subject (e.g.,

Matt fought me

Sue was massaged by me

). This is

demonstrated in

Table 3.1

in examples (1)

Guess what happened this morning? I fought

Matt.

and (5)

Good to see you. Can I tell you about what happened last night? I massaged

Sue

, where the Topic (T) and Person (P) EHs are both satisfied (+), and these stimuli are

defined as (T+, P+), condition A. The Topic EH and Person EH are both violated (-) in

examples (2)

Guess what happened this morning? Matt fought me

and (6)

Good to see you.

Can I tell you about what happened last night? Sue was massaged by me

, and these stimuli

are defined as (T- P-), condition D.

The priming context introduced a third person entity (

Matt

Sue

) into the

discourse beforehand, thereby escalating it on the Topic Empathy scale. I predict that the

interference between the two entities high on the Topic EH (

and

Matt

, or

and

Sue

) would

be in competition with one another, which would negatively affect the acceptability of the

target sentences when the Person EH was satisfied. In these cases, I treat the Topic EH as

violated when the context suggests

Matt

Sue

as the topic, but the subject of the following

target sentence is

, as in examples (3)

Matt can be such a bully. This morning, I fought

Matt

and (7)

Sue and Dan have had quite a busy day. Dan got a haircut, and I massaged

Sue

. These sentences were defined as

(T- P+), condition B. Alternatively, the Topic EH is

treated as satisfied when the context suggests

Matt

Sue

as the topic, and the subject of

the following target sentence is

Matt

Sue

, as in examples (4)

Matt can be such a bully.

This morning, Matt fought me

and (8)

Sue and Dan have had quite a busy day. Dan got a

haircut, and Sue was massaged by me.

Meanwhile, the Person EH is violated in these

examples since the third person is the subject, and the first person is the object (reciprocals)

by-

agentive (active/passives). These sentences were defined as (T+ P-), condition C.

Let us examine in some detail the two relationships between a context sentence and

a topic sentence that can cause a violation of Topic EH (T-). The first relationship can be

found in examples (2)

Guess what happened this morning? Matt fought me

and (6)

Good to

see you. Can I tell you about what happened last night? Sue was massaged by me

, in which

the null context is followed by a target sentence with a third person subject. In these

examples, the null context does not mention a third person entity, which means that the

only relevant topics are {

you

(hearer/participant),

(speaker of experiment text),

here

now

}. Thus, the preferred topic of the following sentence should default to the first person.

In these examples, there is a violation of the Topic EH because the third person subject

(

Matt

Sue

)

is not an established discourse topic, yet it is in the subject position. Thus, I

predict that examples with a T+ such as (8)

Sue and Dan have had quite a busy day. Dan

got a haircut, and Sue was massaged by me

should sound better than (6)

Good to see you.

Can I tell you what happened last night? Sue was massaged by me

despite having the same

target sentence because (6) violates the Topic EH and (8) does not.

The second relationship between the context and the target sentence in which T-

occurs can be found in examples (3)

Matt can be such a bully. This morning, I fought Matt

and (7)

Sue and Dan have had quite a busy day. Dan got a haircut, and I massaged Sue

, in

which the priming context is followed by a target sentence with a first person subject. In

these examples, the priming context mentions third person entities,

Matt

in example (3)

and

Sue and Dan

in example (7), which places these entities into the set of mutual current

concern). In these instances, T- signifies that there is competition between possible topics to

be the subject of the context sentence. I predict that examples like (5)

Good to see you. Can

I tell you what happened last night? I massaged Sue

will be rated higher than examples

like (7)

Sue and Dan have had quite a busy day. Dan got a haircut, and I massaged Sue

despite having the same target sentence

I argue for this prediction because even though

the target sentence

I massaged Sue

is identical in each, there is no competition for the

topic/subject entity in example (5), and there is competition for the possible topic/subject in

example (7). Thus, by referring to entities in the context sentences, an increased

expectation for this entity to be the topic of the following (target) sentence is created for the

addressee/participant. When this expectation is not realized, it is considered a violation of

the Topic EH (T-).

3.3.2 Experiment design. Overall, the experimental design is considered successful

when it produces a statistically significant interaction between the two manipulations of

the Topic EH, designated as “Context (Null vs. Priming)” and the two manipulations of the

Person EH, designated as “Subject (First vs. Third)” for the Reciprocal Group and

“Structure (Active vs. Passive)” for the Active/Passive group. This interaction has four

defining conditions: Null First/Active (condition A), Null Third/Passive (condition D),

Priming First/Active (condition B), Priming Third/Passive (condition C). A significant

interaction of Context x Subject and Context x Structure would demonstrate that each of

the four conditions {A (T+ P+), B (T- P+), C (T+ P-), D (T- P-)} yield different and predictable

acceptability measures relative to the other conditions. The predictions followed that when

a Linguistic EH is violated, the acceptability ratings are lower than when it is observed.

Condition A contains no violations, and it should be the highest rated condition. Condition

D contains two violations, and it should be the lowest rated condition. Since conditions B

and C each contain one violation, I predict that the ratings of both B and C to fall between

A and D. Thus, if the measures of the four conditions in the interaction of Context x Subject

or Context x Structure are significant, then the follow-up paired-sample t-tests will show

that A > {B, C, D} and {A, B, C} > D.

There is no prediction regarding the relative difference between conditions B and C;

however, this difference is important for investigating the relative strength of the two

hierarchies. B (T- P+) represents the violation of Topic EH while the Person EH is observed.

Given this assumption, B represents the independent contribution of the Topic EH.

Similarly, C (T+ P-) represents the independent contribution of the Person EH.

Consequently, B > C would imply that the violations of the Person EH are stronger (i.e.,

result in lower ratings); conversely, C > B would imply that violations of the extended Topic

EH are stronger. Therefore, B-C is a first step toward a measure of the asymmetry between

the two hierarchies. However, this asymmetry is systematically influenced by instances of

imposition, which itself contains subjective factors (e.g., speaker’s whim, egocentric

preoccupation, the cooperative mindset or lack thereof on the part of the addressee, the

cultural conventions, etc.). Thus, this asymmetry is predicted to be approximate, at best.

To summarize, the subjects were expected to find sentences more or less acceptable

depending on the conditions given below in

Table 3.2

(Reciprocal group) and

Table 3.3

(Active/Passive group).

Context

Null

Topicalizing 3

Person

Sentence Subject

First Person

Third Person

First Person

Third Person

Topic Empathy

Hierarchy

Observed Violated Violated Observed

Person Empathy

Hierarchy

Observed Violated Observed Violated

Violations (T, P)

(T+ P+)

(T- P-)

(T- P+)

(T+ P-)

Condition

Acceptability Rating

High

Low

Mid

Example

Context

Guess what

happened this

morning?

Guess what

happened this

morning?

Matt can be

such a bully.

This morning,

Matt can be

such a bully.

This morning,

Example

Target Sentence

I fought Matt.

Matt fought

me.

I fought Matt

Matt fought

me.

Context

Null

Topicalizing 3

Person

Structure

Active

Passive

Active

Passive

Topic Empathy

Hierarchy

Observed

Violated

Observed

Person Empathy

Hierarchy

Observed Violated Observed Violated

Violations (T, P)

(T+ P+)

(T- P-)

(T- P+)

(T+ P-)

Condition

Acceptability Rating

High

Low

Mid

Example

Context

Good to see

you. Can I tell

you about

what

happened last

night?

Good to see

you. Can I tell

you about

what

happened last

night?

Sue and Dan

have had quite

a busy day.

Dan got a

haircut, and

Sue and Dan

have had quite

a busy day.

Dan got a

haircut, and

Example

Target Sentence

I massaged

Sue.

I massaged

Sue.

Sue was

massaged by

me.

Sue Was

massaged by

me.

Tables 3.2

and

3.3

: The matrices demonstrate the EH violations for Context (Topic EH)

and Subject (Person EH) for Reciprocal verbs (

Table

3.2

) and for Active/Passive

constructions (

Table 3.3

). The matrices predict acceptability ratings of sentences relative

to other sentence types based on these violations. In descending order, these ratings are

High (no violation), Mid (one violation of either EH), and Low (violation of both EHs).

Table 3.3: Active/Passive Stimulus Matrix

Table 3.2: Reciprocal Stimulus Matrix

3.3.3 Sensitivity to Linguistic Empathy. The four types of stimuli, A, B, C and D,

provide several different representations of Linguistic Empathy where one or both of the

EHs in question were violated. As mentioned above, in this experiment, the stimuli that did

not violate any Linguistic EH (i.e., T+ P+, condition A) are predicted to be rated as with the

highest acceptability across participants. The stimuli that violated both Linguistic EHs

(i.e., T- P-, condition D) are predicted to be rated with the lowest acceptability across

participants. Thus, the difference between the ratings for stimuli of condition A and ratings

for stimuli of condition D provided the crudest reflection of how violations of the Topic EH

and the Person EH with respect to the Surface Structure EH are perceived. The difference

between condition A and condition D therefore represents a person’s tendency to notice

violations of these two Linguistic Empathy Hierarchies relative to sentences with no

violations of these hierarchies. The prediction follows that the greater the distance is

between A and D for an individual participant, the more heightened that person’s

awareness is to violations of Linguistic Empathy. This difference, thus, is the best

representation of an individual’s sensitivity to Linguistic Empathy in this experiment

design.

3.3.4 Correlations between Linguistic and Psychological Empathy. In addition to the

manipulation of Empathy Hierarchies and the potential measures of Linguistic Empathy

sensitivity, this experiment measured the EQ for each participant. It was hypothesized that

if there is a correlation between a person’s Linguistic Empathy (i.e., a person’s sensitivity to

Linguistic Empathy violations) and Psychological Empathy (i.e., EQ), this would signify

that the two distinct forms of Empathy are somehow overlapping or related. This study

predicted that the acceptability measure of Linguistic Empathy is related to Psychological

Empathy (EQ): if a person demonstrated a heightened awareness to violations of Linguistic

Empathy, then that person should also demonstrate an increased measure of Psychological

Empathy. This study also predicted that Linguistic Empathy and Psychological Empathy

correlate positively with each other. I argue that the correlation would be most significant

between a participant’s EQ score and the difference between points A and D. As a person’s

EQ score increases, so does the distance in ratings between (Null, First/Active) stimuli (T+

P+, condition A) and (Null, Third/Passive) stimuli (T- P-, condition D). Moreover, in this

view, a person’s capacity for sensitivity to Linguistic Empathy is measurable within

individuals, and there are individual differences in Linguistic Empathy just as in

Psychological Empathy.

3.4 Methodology

3.4.1 Participants. Thirty-four UCLA undergraduate students (19 females, 15 males)

participated in the experiment in exchange for course credit. One additional participant

was run, but did not complete the experiment due to equipment failure. All participants

were classified as right-handed based on self-report and observation of right hand

dominance for writing. All participants were also native speakers of English between 18 -

22 years of age.

3.4.2 Stimuli. The experimental session contained a mixture of trials intended to

manipulate a number of different variables related to Linguistic Empathy. The set of 208

items presented to each participant was composed of twelve different item groups, where an

item group includes all possible sentence variants presented with a particular verb. These

twelve item groups consisted of three types of items: reciprocal verbs, active/passive verb

structures, and sentences with genitive phrases. The items with genitives manipulated

different variables that do not map simply onto the variables of the other two groups, and

thus, they are not discussed in this study. Additionally, the reciprocal and active/passive

sets each included one item that manipulated separate variables from the other items in

that group, on an exploratory basis. Those items are also excluded from the analysis. The

data discussed here thus includes two sub-experiments: the Reciprocal group consists of

three item groups (48 trials) using reciprocal verbs, and the Active/Passive group consists of

three item groups (48 trials) in which active/passive construction was contrasted.

The reciprocal sentences included item groups with the following three verbs:

fought

(finally) met,

and

dated.

Eight trials in each item group had a first person pronoun as the

grammatical subject and a single-syllable proper name as the object, e.g.,

I fought Matt

The subject-object choice was reversed for the other eight trials, with the third person

grammatical subject and a first person object (e.g.,

Matt fought me

The active/passive constructions included item groups with the following three

verbs:

coached

stabbed

, and

massaged

. For those item groups, eight trials included an

active construction with a first person pronoun as the grammatical subject and a single-

syllable proper name as the object (e.g.,

I massaged Sue

), while eight trials included a

passive construction with the third person subject and first person

-agentive (e.g.,

Sue

was massaged by me

In each of the Reciprocal and Active/Passive groups, half of the trials were preceded

by a null context sentence, while half were preceded by a priming context sentence, as

Met

is intended to mean “made the acquaintance of,” which is why the adverbial

finally

was included with these stimuli. Kuno (1987) used the verb

met

in his examples as “to run

into;” however, I argue that “made the acquaintance of” carries an equal or greater

reciprocal meaning.

described above. Context length was controlled across sentences within each set within five

total characters. Trial order was randomized for each participant, and this avoided possible

biases of neighborhood effects from surrounding trials that could have shared or differed in

experimental conditions.

Additionally, this experiment was designed to explore potential hemispheric

differences in processing Empathy. Language processing predominantly occurs in the left

hemisphere (Taylor & Taylor, 1990); however, there is reason to believe that the right

hemisphere is selectively involved in social aspects of human interaction in general, and in

the pragmatics of natural discourse in particular (Brownell & Joanette, 1993).

Furthermore, there is evidence that the right hemisphere is selectively involved in

Psychological Empathy (Shamay-Tsoory, Tomer, Berger, & Aharon-Peretz, 2003). Given the

hypothesis in this dissertation that Linguistic Empathy is correlated with Psychological

Empathy, this leads to the conjecture that the right hemisphere is selectively involved in

Linguistic Empathy as well. The experiment was devised to be able to examine this issue by

lateralizing the target word to one visual hemifield and requiring the participant to rate the

stimulus with the ipsilateral (matching) hand. In this way, the input and output would be

restricted to one hemisphere and minimize the involvement of the other hemisphere.

However, the ANOVA that incorporated hemisphere as an independent variable showed no

main effects nor first- or second-order interactions with Hemisphere. Consequently, the

results were collapsed across this variable. Nevertheless, this variable may become

prominent if the analysis were extended to search for an electrophysiological correlate for

Linguistic Empathy.

3.4.3 Procedure. Participants first completed an inventory of linguistic and cultural

background, followed by an EQ test (Baron-Cohen & Wheelwright, 2004) to produce a

measure of Psychological Empathy. During the main cognitive task that followed,

participants were positioned with their eyes 57.3cm from the monitor while resting on a

chin rest. At that distance, one degree of visual angle is equal to one centimeter of distance

on the screen. Participants responded to each trial by pressing one of four keys

corresponding to the level of sentence acceptability, with cartoon faces shown on the

response screen at the end of each trial to illustrate the levels of acceptability of each

response key. Participants responded with the hand ipsilateral to the visual field target;

thus, if the target was left-lateralized, they should respond with the left hand using the

keys “x”, “d”, “f”, or “v”, corresponding to “worst”, “bad”, “good”, and “best” ratings, while if

the target was right-lateralized, they should respond with the right hand using the keys,

“b”, h”, “j”, or “m”, again corresponding to “worst”, “bad”, “good”, and “best” ratings.

Participants were instructed to indicate how “natural or acceptable” they judged the target

sentence to be if they were to hear it from an external speaker.

The experimental session began with a practice block of twelve trials, followed by

two blocks with 104 trials each. The trial began with a fixation cross flashed in the center of

the screen for 180ms. Then, a context sentence appeared centrally for 1500ms above the

fixation cross; this sentence then disappeared and another brief (100ms) fixation cross

remained. Next, the first part of the target sentence appeared centrally for 1500ms, but

with a blank line for the critical final word. Lastly, the final word of the sentence, or target

word, appeared lateralized for 180ms at one degree of visual angle to the right or the left of

the edge closer to fixation, subtending from two to five degrees. Immediately after the

target word, a screen appeared that reminded the participants of the response options,

which remained until the response was given. At the end of each trial, a screen appeared

that reminded the participant of the correct response pattern of the keys (

x, d, f, v

for the

left hand,

b, h, j, m

for the right hand). When this screen appeared, this unfortunately

created variable delay between processing the display and responding to the target, which

contaminated response time as a potential variable. Consequently, reaction time was not

considered as a variable to measure performance in this task.

Figure 3.1

below visually

demonstrates the structure of each trial.

Figure 3.1:

This figure illustrates the screens presented to participants during a

single trial of the experiment. This particular sample trial consists of a null

context for an unmarked reciprocal item. Since the target is flashed on the right

side of the screen, the response to this item should be with the right hand.

Figure 3.1: Sample Trial of the Experiment.

CHAPTER 4

EXPERIMENT: CORRELATIONS BETWEEN

LINGUISTIC AND PSYCHOLOGICAL EMPATHY

This chapter provides the results, discussion, and extensions for the psycholinguistic

experiment described in the previous chapter. The results are split into three sections:

results for items with reciprocal verbs, results for items with active/passive sentences, and

the combined results. A discussion of the results follows that interprets the significance of

these findings with respect to the Linguistic Empathy Hierarchies (EHs) and Psychological

Empathy. This chapter concludes with the limitations and several potential extensions of

this experiment.

4.1 Results

Empathy Quotients (EQs) were obtained for 32 of the 34 participants (19 females, 13

males). Scores ranged from 25-73 out of a possible range of 0-80, with a mean score of 46.8.

The mean score for females was 49.7, and the mean score for males was 42.7; this gender

difference missed significance,

(30) = 1.81,

= .08, but such a difference would be

consistent with prior literature. Baron-Cohen and Wheelwright (2004) found the average

EQ ratings for females to be 47.2 (SD = 10.2) and for males to be 41.8 (SD = 11.2).

4.1.1 Reciprocal items and experiment validation.

4.1.1.1 Reciprocal design.

In chapter 3, the predictions established that the

experiment design was considered successful if the results showed a significant effect of

Context (Topic EH) x Subject [of the target sentence] (Person EH). A 2 x 2 ANOVA of

Context (Null, Priming) x Subject (First Person, Third Person), with mean acceptability

rating as the dependent variable, showed a main effect of Subject. Target sentences with

First Person subjects (M = 3.53, SE = .063) were rated higher than with Third person

subjects (M = 2.68, SE = .107),

(1, 33) = 62.21,

MSE

= .391,

< .001, η

= .65. There was

also a main effect of Context, with

the Null context (M = 3.16, SE =

.078) rated higher than the

Priming context, (M = 3.05, SE =

.068),

(1, 33) = 7.72,

MSE

= .057,

= .009, η

= .19. The critical

Context x Subject interaction was

significant,

(1, 33) = 16.11,

MSE

= .092,

< .001, η

= .33 (

Figure

4.1

). This validated the design by

showing that it successfully

implemented the intended rating

order of the defining conditions

(i.e., the four stimulus types), such

that A > {B, C, D} and {A, B, C} >

D. Below, example sentences with

reciprocal verbs for each of the

defining conditions are listed in

Table 4.1.

A: + +

B: - +

C: + -

D: - -

Figure 4.1.

Acceptability ratings for First Person

versus Third Person grammatical subjects in Null

or Priming contexts. Error bars represent +/- 1 SE.

Points ABCD correspond to stimuli in which: A-

both EHs observed (++); D-both EHs violated (- -);

B-Topic EH violated, Person EH observed (-+); C-

Topic EH observed, Person EH violated (+-).

Figure 4.1: Mean Ratings, Reciprocal Items

4.1.1.2. Defining conditions.

The results for the defining conditions are summarized in

Table 4.2

and

Table 4.3.

First,

Table 4.2

provides mean ratings for the Reciprocal items and their correlations with EQ.

The t-tests showed that all the defining conditions (A, B, C and D) were significantly

different from zero. In this table, as in all tables that follow, a cell is highlighted when the

datum therein is statistically significant. Paired-samples t-tests also showed that in the

Priming context, target sentences

with First person subjects (condition

B) had significantly higher ratings

than target sentences with Third

person subjects (condition C).

Table

4.3

contains the measures of the

differences between the defining

conditions. The second column in

Table 4.3

lists the meaning of the

Table 4.1.

The four defining conditions (A, B, C, D) for the Reciprocal group with

example contexts and target sentences, and indication of the Topic and Person

Hierarchy observation (+) or violation (-).

Defining

Condition

Context

Target sentence

Violations

T P

Guess what happened this morning?

I fought Matt.

+ +

Matt can be such a bully. This

morning,

I fought Matt.

- +

Matt can be such a bully. This

morning,

Matt fought me.

+ -

Guess what happened this morning?

Matt fought me.

- -

Table 4.1: Sample Reciprocal Sentences for the Defining Conditions

Table 4.2

. Mean ratings for the defining

conditions (ABCD) of the Reciprocal sentences

and their correlations with EQ.

Defining

Conditions

Mean

Rating

Correlations

with EQ

3.69

0.268

0.137

3.37

0.156

0.395

2.73

-0.35

0.05

2.64

-0.496

0.004

Table 4.2: Mean Ratings and Correlations,

Reciprocal Items

difference measures in terms of the conditions of the two hierarchies that are being

assessed. For example, A-B is the same as A: (T+ P+) – B: (T- P+). Thus, this shows the

difference between T+ and T- when P+ remains constant, written as T

P+

. Of the difference

measures, there was a significant effect (i.e., greater than zero) for A-D, A-C, and A-B

Taken together, these data validate the defining conditions. C-D showed only a trend

towards significance, thus, the prediction of a significant difference was not supported.

There was no prediction about B-C, yet the data showed that B was significantly higher

than C, suggesting that the violations of the Person EH impacted the participants to a

greater extent than the violations of the Topic EH in this experiment design (t = 5.232, p <

0.001) when there was an explicit entity that was topicalized, and the target sentence

contained a default topic.

4.1.1.3 Linguistic Empathy sensitivity and correlations with Psychological Empathy

(EQ).

There was a prediction that stated that A-D would be the crudest measure of

Design

Empathy

Hierarchy

Mean

Rating

t-test

Correlation with

A - D

(T∑

+ P∑) / 2

1.081

8.908

< 0.001

0.647

<.001

A - B

P+

0.329

4.607

< 0.001

0.11

0.548

C - D

P-

0.095

1.392

0.174

0.335

0.061

B - C

(T- P+) – (T+P-)

0.657

5.232

< 0.001

0.38

0.032

A - C

T+

0.986

9.269

< 0.001

0.523

0.002

B - D

T-

0.753

5.777

< 0.001

0.542

0.001

Table 4.3.

Mean difference measures in the defining conditions (A, B, C, D) for the

Reciprocal sentences, t-tests and correlations with EQ. The Empathy Hierarchies

that correspond with the difference measures are listed.

Table 4.3: Mean Ratings and Correlations, Reciprocal Difference Measures

sensitivity to Linguistic Empathy, and that a person’s sensitivity to Linguistic Empathy

should be related to Psychological Empathy (EQ) by a significant negative correlation

between the differences of ratings of A-D and the EQ of the participant. To demonstrate

this, EQ was correlated with the four defining conditions (A, B, C and D,

Table 4.2

) as well

as with the four differences between these conditions (A-D, A-B, C-D as well as B-C,

Table

4.3

Among the defining conditions, only C and D showed significant correlations with

EQ. Among the difference measures, A-B did not show a significant correlation with EQ (r =

0.11, p = 0.548) and the correlation of C-D with EQ missed significance (r = 0.335, p =

0.061). A-D and A-C showed significant correlations with EQ (i.e., both exhibited differences

between conditions when the Person EH was observed and violated). Furthermore, A-D

(i.e., difference between the conditions when both EHs are observed and when both EHs are

violated) showed the highest correlation with EQ (r = 0.647, p = <0.001). A-C does not

involve a violation of the Topic EH, but it does involve a difference between observing and

violating the Person EH. In other words, A-C measures the Person EH (P+ minus P-) in

conditions when the Topic EH is observed (T+). Note that C is smaller or equal to D, and

therefore A-C is smaller than A-D. Thus, A-D is the largest and presumably the best

measure of the two empathy hierarchies. In other words, A-C measures the Person EH (P+

minus P-) in conditions when the Topic EH is observed (T+) since both A and C observe it.

Since A-D accommodates a greater violation of Linguistic Empathy, it is a more accurate

and more sensitive measure of Linguistic Empathy than A - C, and consequently it should

and did correlate more significantly with EQ. Similarly, the correlation of EQ with A-D was

also higher than the correlation for D alone (r = 0.496, p), confirming the prediction that the

difference measure A-D is a more accurate and sensitive measure of Linguistic Empathy

than D.

The difference between the Topic EH and Person EH, B-C, was described as a

promising measure of the asymmetry between the two hierarchies. The results showed that

B-C was significantly different from chance, suggesting that the Person EH was stronger

than the Topic EH in this design. The results also showed that A-B (i.e., T

P+

: difference in

Topic EH violation when Person EH is observed) is always larger than C-D (i.e., T

P-

difference in Topic EH when Person EH is violated).

Taken together, the correlation data suggest that A-D is the most promising

measure of a person’s sensitivity to Linguistic Empathy, combining violations of the Topic

and Person EHs. As expected, the correlations with EQ were largest for terms that include

violations of both EHs (difference measure A-D) rather than violations of only the Person

EH (difference measure A-C), as shown in

Figure 4.2

Figure 4.2.

Correlation between EQ score and sensitivity to Linguistic Empathy

when determining acceptability for reciprocal sentences, with Null context (A - D,

left), or Priming context (A - C, right). Both correlations are significant, but A - D

demonstrates a greater correlation.

Figure 4.2: Correlations of EQ and Linguistic Empathy, Reciprocal Sentences

4.1.2 Active/Passive items and experiment validation.

4.1.2.1. Active/Passive design.

The experiment design was

considered successful if the results

showed a significant effect of Context

(Topic EH) x Structure [Active

versus Passive] (Person EH). The

results for the Active/Passive items

were analyzed using the same

procedure that was used for the

Reciprocal items. A 2 x 2 ANOVA of

Context (Null, Priming) x Structure

(Active: First person subject,

Passive: Third Person subject)

revealed a main effect of Structure,

where Active sentences (M = 3.48,

SE = .077) were rated higher than

Passive sentences (M = 2.44, SE =

Sentences in this group are referred to as Active and Passive to avoid confusion with the

corresponding variable in the Reciprocal group. The “Actives” in this group have first

person grammatical subjects, and the “Passives” have third person grammatical subjects,

but the “First person” and “Third person labels were reserved for the Reciprocals.

Figure 4.3.

Mean acceptability rating for

Active/Passive sentences with Active (first person

subject) versus Passive (third person subject)

sentences in Null or Priming contexts. Error bars

represent +/- 1 SE. Points ABCD correspond to

stimuli in which: A-both EHs observed (++); D-

both EHs violated (--); B-Topic EH violated,

Person EH observed (+-); C-Topic EH observed,

Person EH violated (-+).

A: + +

B: - +

C: + -

D: - -

Figure 4.3: Mean Ratings, Active/Passive Items

.102),

(1, 33) = 83.93,

MSE

= .44,

< .001, η

= .74. The main effect of Context was not

significant,

(1, 33) < 1. The critical interaction of Context x Structure was significant,

(1,

33) = 20.62,

MSE

= .095,

< .001, η

= .39. Follow-up comparisons showed that Active

sentences (M = 3.62, SE = .070) were rated as more acceptable than Passive sentences in

the Null context (M = 2.34, SE = .117),

(33) = 10.01,

< .001, as well as in the Priming

context (Active: M = 3.34, SE = .090, Passive: M = 2.54, SE = .102),

(33) = 6.53,

< .001.

These data are displayed in

Figure 4.3

The interaction indicated that the effect of Structure in the Null context was

stronger than in the Priming context. There was a significant effect of Context in the Active

sentences (First person subjects),

(33) = 5.39,

< .001, where ratings were higher in the

Null context (A) than in the Priming context (B). Passive sentences (Third person subjects)

showed the opposite pattern, which was predicted: the ratings were significantly higher in

the Priming context (C) than in the Null context (D),

(33) = 2.55,

= .015. This validated

the design by showing that it successfully implemented the intended rating order of the

Table 4.4.

The four defining conditions (A, B, C, D) for the Active/Passive group with

example contexts and target sentences, and indication of the Topic and Person

Hierarchy observation (+) or violation (-).

Defining

Condition

Context

Target sentence

Violations

T P

Good to see you. Can I tell you about

what happened last night?

I massaged Sue.

+ +

Sue and Dan have had quite a busy

day. Dan got a haircut, and

I massaged Sue.

- +

Sue and Dan have had quite a busy

day. Dan got a haircut, and

Sue was

massaged by me.

+ -

Good to see you. Can I tell you about

what happened last night?

Sue was

massaged by me.

- -

Table 4.4: Sample Active/Passive Sentences for Defining Conditions

experimental conditions, i.e., A > {B, C, D} and {A, B, C} > D. Example sentences for each of

the defining conditions are listed in

Table 4.4.

4.1.2.2. Defining

conditions.

The results for the

follow-up tests are summarized

Tables 4.5

and

4.6.

First,

Table 4.5

provides the mean

ratings for the defining

conditions (A, B, C and D), and

the t-tests showed that all the

defining conditions were

Defining

Conditions

Mean

Ratings

Correlations

with EQ

3.62

0.116

0.527

3.34

0.004

0.985

2.54

-0.169

0.356

2.34

-0.371

0.037

Table 4.5

. Mean ratings for the defining conditions

(A, B, C, D) of the Active/Passive sentences and

their correlations with EQ.

Table 4.5: Mean Ratings and Correlations,

Active/Passive Items

Table 4.6.

Mean difference measures in the defining conditions (ABCD) for the

Active/Passive sentences, t-tests and correlations with EQ. The Empathy Hierarchies

that correspond with the difference measures are listed.

Design

Empathy

Hierarchy

Mean

Ratin

t-test

Correlation with

A - D

(T∑

+ P∑) / 2

1.261

9.79

< 0.001

0.408

0.02

A - B

P+

0.281

5.234

< 0.001

0.153

0.404

C - D

P-

0.196

2.322

0.027

0.314

0.08

B - C

(T- P+) - (T+P-)

0.785

6.15

< 0.001

0.14

0.445

A - C

T+

1.065

9.786

< 0.001

0.239

0.187

B - D

T-

-0.981

7.179

< 0.001

0.325

0.07

Table 4.6: Mean Ratings and Correlations, Active/Passive Difference Measures

significantly different from zero.

Table 4.6

provides the difference measures, and

demonstrates that there was a significant effect (i.e., greater than zero) for all difference

measures. Paired-samples t-tests also showed that in the Priming context, Active target

sentences (B) had significantly higher ratings than Passive target sentences (C). This

difference between the Topic EH and Person EH, B-C, was described as a promising

measure of the asymmetry between the two hierarchies, and it suggests that the violations

of the Person EH impacted the participant greater than violations of the Topic EH in this

experiment design (t = 6.15, p < 0.001). The results also showed that A-B is always larger

than C-D.

4.1.2.3 Linguistic Empathy sensitivity and correlations with Psychological Empathy

(EQ).

As with the Reciprocal group, the predictions stated that A-D in the Active/Passive

group would be the best measure of sensitivity to Linguistic Empathy, and that a person’s

sensitivity to Linguistic Empathy should be related to Psychological Empathy (EQ). To

investigate the validity of this prediction, EQ was first correlated with the four defining

conditions (A, B, C and D,

Table 4.5

), with the four differences between these conditions (A-

D, A-B, C-D and B-C,

Table 4.6

). Of the four defining conditions, only the acceptability

ratings for Passive items in the Null context (D) correlated significantly (and negatively)

with EQ score. Of the differences in these conditions, only A-D showed a significant

correlation with EQ (

Table 4.6

The correlations with EQ for the individual defining conditions are shown in

Table

4.4

, and the correlations with EQ for the difference measures are shown in

Table 4.6

Among the defining conditions, only D showed significant correlations with EQ. Among the

difference measures, A-B did not show a significant correlation with EQ (r = 0.153, p =

0.404), and the correlation of C-D with EQ missed significance (r = 0.314, p = 0.08), and B-C

did not show a significant correlation (r = 0.14, p = 0.445). In fact, A-D was the only

difference to show significant correlations with EQ (r = 0.408, p = 0.02). The correlation of

EQ with A-D was also higher than the correlation for D alone (r = -0.371, p = 0.037),

confirming the prediction that the difference measure A-D was a more accurate reflection of

sensitivity to Linguistic Empathy than the absolute measure, D.

As with the Reciprocal group, the correlation data suggest that A-D is the most

promising measure of a person’s sensitivity to Linguistic Empathy. This measure combines

the violations of the Topic and Person EHs (condition D), and compares them with

corresponding stimuli with no EH violation. As expected, the correlations with EQ were

largest for terms that include violations of both the Topic EH and the Person EH (difference

measure A-D) rather than violations of only the Person EH (difference measure A-C), as

shown in

Figure 4.4

Figure 4.4.

Correlation between EQ and Linguistic Empathy for Active/Passive

sentences, with the Null context (A-D, left, significant), and the Priming context

(A-C, right, not significant).

Figure 4.4: Correlation of EQ and Linguistic Empathy, Active/Passive Sentences

4.1.3 Comparison across sentence types. Finally, the correlation was examined for

each of the defining conditions and the difference measures of Linguistic Empathy for the

Reciprocal and Active/Passive item groups together. The correlation between A-D in the

Reciprocals and A-D in the Active/Passive was highly significant,

= 0.71, p < .001. The

correlation of B-C in the two type of sentences was also a significant but smaller,

= 0.51, p

= .003. A Fisher’s Z test showed that the correlation coefficients for A-D and B-C did not

differ significantly from each other, z = 1.22, p = 0.22. More generally, all the difference

measures that were analyzed (see

Table 4.3

and

Table 4.6

) correlated significantly across

the two sentence types, but they did not correlate significantly with each other within each

sentence type. The pattern of the results demonstrated that a measure of Linguistic

Empathy in the Reciprocal item group correlated significantly with the same measure in

the Active/Passive item group.

In comparing respective ratings for the Active/Passive stimuli with the Reciprocal

stimuli, inspection of

Figure 4.2

and

Figure 4.4

suggests that the ratings for points C and D

were lower in the Active/Passive sentences than for the Reciprocal sentences, whereas the

ratings for points A and B were similarly high for the Reciprocal and Active/Passive

sentences. More formally, condition D was significantly lower in the Active/Passive than in

the Reciprocal sentences,

(33) = 3.45,

= .002. Point C was almost significantly lower in

the Active/Passive than in the Reciprocal sentences,

(33) = 1.89,

= .068. By contrast, the

difference in mean ratings for the two types of sentences was not significant for either point

A (

(33) = 1.36,

= .183) or point B (

(33) < 1).

4.2 Discussion

4.2.1 Defining Conditions. The findings confirm that comparisons across

acceptability ratings yield valid and stable measures of sensitivity to violations in

Linguistic Empathy. The results found that for both Reciprocal and Active/Passive groups,

with first person subjects in a null context (condition A:

Guess what happened this

morning? I fought Matt

.) were rated higher than sentences with third person subjects

(condition D:

Guess what happened this morning? Matt fought me.

), as was predicted by

the combined effect of violations to the Person EH and the Topic EH. In the priming

context, the sentences with first person grammatical subjects (condition B:

Matt can be

such a bully. This morning, I fought Matt

) were less acceptable than in the null contexts

(condition A). The sentences with third person grammatical subjects were more acceptable

in priming contexts (condition C:

Matt can be such a bully. This morning, Matt fought me.

)

than in the null contexts (condition D). Thus, the acceptability ratings confirmed the

predictions that A > {B, C, D} and {A, B, C} > D. These ratings also validated the principles

stipulated in Kuno and Kaburaki (1977) and Kuno (1987) regarding the role of the two EHs

in sentence acceptability. In other words, when there is a conflict in the Empathy relations

between the Person EH and the Surface Structure EH, the Topic EH and the Surface

Structure EH, or both the Person EH and the Topic EH with the Surface Structure EH,

sentence acceptability becomes more marginal.

The correlation of the measurements with EQ is more complex. Firstly, let us

consider the correlations with defining conditions for both Reciprocal and Active/Passive

groups. For sentences that observed the Person EH, no correlation with EQ score was found

for null context (condition A:

Guess what happened this morning? I fought Matt

. and

Good

to see you. Can I tell you what happened last night? I massaged Sue.

). This finding suggests

that sentences with no EH violation should be deemed acceptable by all participants,

regardless of EQ. Next, in sentences that observed the Person EH, correlation with EQ was

slightly higher but not significant for the priming context either (condition B:

Matt can be

such a bully. This morning, I fought Matt.

and

Sue and Dan have been quite busy today.

Dan got a haircut, and I massaged Sue.

). Sentences in which the Person EH was violated

were statistically significant in the priming context for the Reciprocals (condition C:

Matt

can be such a bully. This morning, Matt fought me.

), but not for the Active/Passives

(condition C:

Good to see you. Can I tell you what happened last night? Sue was massaged

by me.

). Reciprocals and Active/Passives with third person subjects in the null context

displayed the most significance (condition D:

Guess what happened this morning? Matt

fought me.

). These findings suggest that as the stimuli increase in violations, they

gradually become judged as more unacceptable by people with higher EQ.

4.2.2 Difference Measures. For the Reciprocal items in both null and priming

contexts, the mean acceptability ratings correlated negatively with EQ score for the target

sentences with third person grammatical subjects. These correlations were strongest in the

differences between ratings of conditions A (

Guess what happened this morning? I fought

Matt

.) and D (

Guess what happened this morning? Matt fought me.

). Correlations were also

significant between conditions A and C (

Matt can be such a bully. This morning, Matt

fought me.

), though less than A-D. The correlations demonstrated that the higher the EQ

measure is in a participant, then the more sensitive that person is to Linguistic Empathy

violations of both the Person EH and Topic EH. This matched the expectation that a

correlation with EQ would be most significant in sentences with a more violations of

Linguistic Empathy. In other words, a higher EQ score correlates with a larger discrepancy

between ratings of sentences that observe and sentences that violate the EHs. The lower

the EQ is for the participant, the less sensitive the participant is to violations of the Person

EH and Topic EH together (condition D:

Guess what happened this morning? Matt fought

me.

). This sensitivity to violations is less powerful but remains significant when the Topic

EH was observed in priming contexts (condition C:

Matt can be such a bully. This morning,

Matt fought me.

). The fact that the correlation is less significant when only one EH is

violated confirms that violating multiple hierarchies is additive: the violations compile to

make an utterance sound less acceptable. This also indicates that violations of the Person

EH in this design affect participants with higher EQ regardless of the type of context

sentence, and comparatively, violations of the Person EH have less of an effect on

participants with lower EQ.

For the Active/Passive items, the mean acceptability ratings correlated negatively

with EQ score for the passive sentences only in the null context. This correlation was the

difference between the ratings of conditions A (

Good to see you. Can I tell you what

happened last night? I massaged Sue.

) and D (

Good to see you. Can I tell you about what

happened last night? Sue was massaged by me.

). The implication must be the same as with

the Reciprocal items: the lower the EQ is for a particular participant, the more inclined that

person is to ignore violations of the Person EH in Null contexts. However, there was no

significant correlation between EQ score and ratings of stimuli that violated the Person EH

and observed the Topic EH (i.e., difference between condition A:

Good to see you. Can I tell

you about what happened last night? I massaged Sue.

and condition C:

Sue and Dan have

been quite busy today. Dan got a haircut, and Sue was massaged by me.

). The lack of

significant correlation with EQ score indicates that the observation (

I massaged Sue.

) or

violation (

Sue was massaged by me.

) of the Person EH in the Active/Passive items overrode

the effect of the Topic EH for all participants despite EQ. This can be interpreted either as

participants with lower EQ scores are relatively less sensitive to violations of the Person

EH and more sensitive to violations of the Topic EH in the Reciprocal items versus the

Active/Passive items, or as participants with higher EQ scores are less sensitive to the

effects of Topic EH in the Active/Passive items. In either case, the interactive effects of both

Linguistic EHs that correlated across Reciprocal and Active/Passive item groups support

the hypothesis that sensitivity to Linguistic Empathy violations is related to Psychological

Empathy.

4.2.3 Linguistic Empathy measure. I argue that the specific rating of a marginal

sentence is not as important as the difference between how this sentence sounds compared

to a felicitous sentence. For instance, if a person rates sentences with EH violations as a 1

out of 4, and rates sentences with no violations as a 3 out of 4, then this person seems to be

affected by the EHs to the same extent as a person who rates sentences with EH violations

as a 2 out of 4, and rates sentences with no violations as a 4 out of 4. Since the difference is

2 in both instances, these raters are influenced by the effects of EH violation to the same

degree. However, if the measure of Linguistic Empathy sensitivity were recorded by an

individual’s rating of sentences with EH violations without such a comparison, this

distinction would be lost. Indeed, the results demonstrated that EQ was correlated more

strongly with the difference of A-D than with D alone (A:

Guess what happened this

morning? I fought Matt

versus D:

Guess what happened this morning? Matt fought me.

This supports the claim that a difference between two defining conditions is a more

accurate measure of EQ than a proposed fixed value at a single point.

If a valid measure of Linguistic Empathy sensitivity is required to be positively

correlating with EQ, then the three difference measures A-D (A:

Guess what happened this

morning? I fought Matt.

minus D:

Guess what happened this morning? Matt fought me.

), A-

C (A:

Guess what happened this morning? I fought Matt.

minus C:

Matt can be such a

bully. This morning, Matt fought me.

), and B-D (B:

Matt can be such a bully. This morning,

I fought Matt.

minus D:

Guess what happened this morning? Matt fought me.

) are valid

candidates as measures of sensitivity to Linguistic Empathy because there is an imbalance

in the amount of EH violation. However, since A-C and B-D only contain the violation of a

singular EH, these differences measures are less likely candidates. Further, if a difference

measure that has the highest correlation with EQ is assumed to be the best estimate of

sensitivity to Linguistic Empathy, then A-D (A:

Guess what happened this morning? I

fought Matt.

versus D:

Guess what happened this morning? Matt fought me.

) is clearly the

preferred measure, at least when contrasting the Topic and the Person EHs under the

restricted conditions specified by the experimental design. Indeed, this difference is also the

crudest in that it demonstrates the difference between observing both EHs and violating

both EHs. The significant correlation between Linguistic Empathy sensitivity and EQ is

consistent with the view that Linguistic Empathy sensitivity is a meaningful individual

difference.

4.2.4 Implications for the Linguistic Empathy Hierarchies. Difference in the relative

strength of the Topic EH vis à vis the Person EH has been found among different

languages. In this experiment the Person EH overrode the Topic EH (witness that cases

Matt can be such a bully. This morning, I fought Matt

were rated better than

Matt can

be such a bully. This morning, Matt fought me.)

, although strong heavy contextual priming

can significantly improve the acceptability of sentences that violate the Person EH (e.g.,

Guess what, big news! Remember Basil, the guy I met last summer in England, that cool

handsome aristocrat? Well, Basil is marrying me!).

Nevertheless, in the null context, odd

English sentences like

Sue was massaged by me

are still considered grammatical. In

Lummi (a Native American language), however,

He was hit by me

(although structurally

equivalent to

Sue was massaged by me

) is judged to be ungrammatical (Oshima, 2007a, p.

757). Evidently, in this language the Person EH is stronger than the Topic EH, and

Empathy Hierarchies have the status of grammatical rules. Considering the varying status

of Linguistic Empathy among the languages of the world, the results obtained from the

participants with lower EQ in this experiment are by no means impossible.

The investigation of which stimulus type demonstrated the best measure of the

Topic EH is intricate. Across both Reciprocal and Active/Passive item groups, the difference

of A-B (A:

Guess what happened this morning? I fought Matt.

versus B:

Matt can be such a

bully. This morning, I fought Matt.

) is not equal to the difference of C-D (C:

Matt can be

such a bully. This morning, Matt fought me.

versus D:

Guess what happened this morning?

Matt fought me.

). The comparison of the EH values is as follows: A-B = (T+ P+) – (T- P+),

and C-D = (T+ P-) – (T- P-). These differences represent comparisons of when the Person

EH remained constant (P+ & P+ or P- & P-) while the status of the Topic EH changed from

observed to violated. Thus, in these differences, the Topic EH is the only EH whose status

changes. The difference in the status of the Topic EH was larger in A-B (A:

Guess what

happened this morning? I fought Matt.

versus B:

Matt can be such a bully. This morning, I

fought Matt.

) than in C-D (C:

Matt can be such a bully. This morning, Matt fought me.

versus D:

Guess what happened this morning? Matt fought me.

). This suggests that the

violation of the Topic EH when the Person EH is observed (condition B:

Matt can be such a

bully. This morning, I fought Matt

) made the violation of the Topic EH more obvious to the

rater; contrarily, a violation of the Topic EH when the Person EH is also violated (D:

Guess

what happened this morning? Matt fought me.

) made the violation of the Topic EH less

obvious. In effect, this demonstrated that violation of the Person EH influenced the ratings