Tilburg University
Proper Nouns and Pronouns
van Vliet, S.M.K.
Publication date:
2008
Document Version
Publisher's PDF, also known as Version of record
Link to publication in Tilburg University Research Portal
Citation for published version (APA):
van Vliet, S. M. K. (2008). Proper Nouns and Pronouns. LOT.
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal
Take down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Proper
nouns and
pronouns
Trans 10 fax: +3130253 6000
3512JKUtrecht e-mail:lot@let.uu.nl The Netherlands http://www. lotschool.nl
Cover Illustration: © Sol LeWitt, 1997,Wavylineswith blackborder (detail), c/o
Pictoright 2008.
ISBN 978-90-78328-49-0
NI.JR 616
Proper
nouns and
pronouns
The production
of
referential
expressions
in
narrative
discourse
Eigennamen en Pronomina
De productie van referentiele uitdrukkingen in narratieve teksten
Proefschrift
ter verkrijging vande graadvandoctor
aandeUniversiteitvanTilburg, op gezag vanderector magnificus,
prof. dr. F.A. van derDuynSchouten,
in het openbaarteverdedigen
ten overstaan van een door het collegevoorpromoties
aangewezencommissie
in de aula vandeUniversiteit
op vrijdag28maart 2008
om 16.15 uur
door
Sarah Maria Karmijn van Vliet
Prof. dr. A.A. Maes
Copromotor: Dr. J. Schilperoord
Acknowledgement ..11
Chapter1.Introduction... 13
1.1 The use of referentialexpressionsinnarrative discourse... 13
1.2 A referencepointapproach todiscourse reference... 18
1.3 Approachand set-up
ofthe study. 70
1.4 Researchquestions... 1.5 Contentsoverview . 77
Chapter2. Reference pointsanddominionsinnarratives.Adiscourselevel exploration ofthereferencepointmodel
of
anaphora... 252.1 Introduction...25
2.2Discourse referentsasmental entities...25
2.3The referencepointmodel
of
anaphora......„„„..282.3.1NominalsemantirR 78
2.3.2Sentential anaphora constraints 2.4 Adiscourse level referencepointmodel
of
anaphora... 352.4.1 Prerequisites foradiscourse level referencepointmodel... 35
2.4.2 Characteristics
of
referencepointsanddominions in narratives... 362.4.3An illustration ofthenarrative referencepointmodel... 38
2.5 Factors inthenarrative referencepointmodel... 38
2.5.1 Protagonists 19
2.5.2 Intervening reference .. 40
2.5.3Syntacticfunction.. 2.5.4Linear position... 2.5.5Referential distance... „... ...43
2.5.6 Episode structureinnarrative€ 44
2.5.7 Character perspective ..50
2.5.8 Interaction
of
factors...-522.6 Summary S4
Chapter
3.Operationalizationof
factorsaffectingreferential choice innarrativeg -57
3.1 Introduction...
3.3Interveningreference 58 3.4Syntactic
function 59
3.5 Linear position... 3.6Referential distance... 3.7Episodestruct,irp 61 3.7.1 Episodeshifts 62
3.7.2 Episodeanddistancefactors 62
3.7.3 Episodeanddiscourse markers ... 63
3.8Visualviewpointandcharacterperspective 64
3.8.1Visualviewpoint... 3.8.2Character perspective 66
3.9 Perceptualattention 70
3.10 Summary 71 Chapter4.Collecting production data.Elicitation ofacorpusofwritten narrativeR 73 4.1 Introdurtinn 73 4.2 A model
of
languageproduction... 4.3Visual stimuli:verbalizing comic picturesasproduction task... 754.3.1 Characteristics ofthecomicgenre... 76
4.3.2 Readingandverbalizing
comics 85
4.4 The visual stimuli ... 4.4.1 Episode
shifts 89
4.4.2Viewpointshifts .... 4.4.3 PagebreakQ 90
4.5 The construction ofacorpus
ofwritten
narratives... 4.5.1 Variationin narrative production... 914.5.2 Elicitation t,qk 92
4.6 Analysis ofasingle story 93
Chapter 5. Thedistribution
ofproper
nounsandpronouns.A
frequency analysis ofthecollected corpus 103 5.1 Introduction 103 5.2 Characterization ofthecorpus.......1035.2.1 General characteristicsandselections 103 5.2.2 Types
of
expression used in reference totheprotagonist... 1075.3 Syntactic function 110 5.4Linear position...-.-....111
5.5 Episode structure... ... 112
5.5.1 Episode shifts 112 5.5.2 Episodeanddistancefactors 114 5.5.3 Episodeanddiscourse markers...115
5.5.4Situation model factors...116
5.7 Pagebre91(4 124
5.8 Discusginn 125
Chapter 6.Theprobability
of
proper nouns.A
regressionanalysis of thecollected cnrpi,9 131
6.1 Introduction- .131
6.2 Motivation fortheregressionanalygk 131
6.3 Description ofthelogisticregressionanalysis ... 133
6.4Predictor variablesandhypotheses ...135
6.5Material...„„„„„„.... 138
6.6Logisticregression models ... ... .140
6.6.1 Regression modelsbasedon constant... 140
6.6.2Regression modelsbasedonreferential distance... 142
6.6.3Regression modelsbasedondiscourse-structural factors and intervening reference... 146
6.6.4Regression modelsbasedon syntacticfunctionandotherfactors... 152
6.6.5Regression modelbased on all significantfactors ... 154
6.7Discus inn 156 Chapter7.Meaning and useofproper nounsandpronouns. Conclusion and Discussion... 161
7.1Concluqinn 161 7.2 Epilogue... „... 169
Appendices... „... 173
References...„„„„„„„„„„„„„...187
Samenvatting inhetNederlandg 193
Acknowledgements
I consider myself very fortunate to have such wonderful supervisors, colleagues,
family andfriends. Let me take this opportunity to express my gratitude to some of
the peoplewho playedanimportant role in this project.
First, I wishto thank my daily supervisor Joost Schilperoord. Thank you
for your excellent way
of
explainingthings, for
your patience inspite of my
stubbornness attimes, forthetimesyouprevented mefrommaking methodological mistakes, and
for
keeping me ontrack. Most of all I wish
tothank you for your
enthusiasm for this project and our inspirational talks onallmatterslinguistic.
I am very grateful tomy promotor LeoNoordman for his supervision and
support. Your meticulous comments were tremendously helpful, and often forced
me to clarify my line
of
thinking. Thank youfor
pointing out mynon-sequiturs and for askingsuchdifficult
questions.I am alsovery grateful to mypromotor Fons Maes for his many invaluable comments and his passionate speeches about where the dissertation should be headed.Thank you, also,
for
alwaysbeing sohelpful throughoutthisproject, both in practicalandlinguisticmatters.Dear Joost, Leo, and Fons, I will miss the talks wehad among the four of
us. Our meetings were always pleasant and constructive, and I have learned a lot from all of you.
A special word
of
thanks is duetoFrederike van der Leek. Her lectures on 'language,world, and mind'initially
inspired me to graduate in linguistics, and hersupervision of my masters thesis made me so enthusiastic that I applied for the
'referentialexpressions'project, resulting in this dissertation.
Huub vandenBerghdeserves amillion thank yous forhis invaluable help
with the multi-level analysis, which allowed me to
write what I
now think is themost interesting part of this book, namelytheregressionanalysisreported in chapter 6. Thank you ever so much.
Throughout my Ph.D. project I have hadmany helpful comments from or
interesting conversationswithscholars during conferences, LOTcourses,orthrough
email correspondence. For their various contributions to my
thinking - some of
which did and some
of
which didn't end up in this work - I wish to thank WallaceChafe, Barbara Dancygier, Catherine Emmott, TheoJanssen, Leonoor Oversteegen,
Ellen Prince, Russell Tomlin,and Carel van Wijk, with apologies to those I might
have forgotten.
I am grateful to themembers ofthedissertation committee,
for
taking thetime to readthemanuscript and
for
theirhelpfulcommentsatvarious points duringthis project: Huub van
den Bergh, Frederike van der Leek, Jost Sanders, ArieVerhagen,andWietske Vonk.
I must also mention my colleagues at the Discourse Studies department,
Pashiera Barkhuysen,andothers.Thank youfor making me feel at homeatTilburg
University from theveryfirst day Istartedworkingthere.
I also wish to thank my fellow (former) Ph.D. students outside Tilburg,
Ninke Stukker, Maaike Belien, Mirna Pit, Jacqueline Evers-Vermeul, and Esther
Pascual,for making Ph.D. lifesopleasurable during conferences, LOTcourses, and othermeetings.
Of course I must thank my non-linguistic friends
for
being such goodfriends. Among the lovely
people I know I wish
to thank especially Annemarie Jongkoen, Maartje Henket, Gerda de Dreu, Lavie van Ingen, Anne Marie Koper,PantherContent,Yolanda Jolie,Erwin Smit,andMartinePrange.
A special word
of
thanks is also due to Liesbeth Verschure, my father'swife, andtheloveliest step mom in theworld. Thank you for your loveandsupport throughoutthosemanyyears, and
for
being suchagreat'oma'toDavid.And
of
course my father, Pieter van Vliet, deserves mentioning. Daddydearest, you are a greatfather. It means so much to me to have a father who has
always had
faith in me, and
has always encouraged me to do thethings I am
passionate about.Andnever adull moment...
And last, butcertainlynot least, I wish tothank my love, StefKegel. Thank you so much fortaking such good care
of
David andme. Thanks to your undyingdedication andsupport, I was able to keep onworking morning, noon, and
night, if
necessary. I trustthere will nowbeplenty of time forthe three of ustorelax, enjoy,
Chapter 1
Introduction1.1 The useof referentialexpressionsinnarrativediscourse
This study addresses the phenomenon
of
reference maintenance in discourse: throughout the study Iwill
focus on consecutive reference to protagonists innarratives. More specifically, I willinvestigate the conceptual, discourse-structural,
and linguistic factors that determine a narrator's referential choices during the
production
ofDutch
writtennarrative discourse.In discourse,speakers andwritersusevariousmeans torefertopeopleandthings: a single conceptualizedperson, forexample, can be referred to using various types of
referential expression. Considerthefollowingsentences:
(1) a. A girl in a red coat just leftamessage for you at thecounter
b.77;e
girlyou
werejust mlking to leftamessage for you atthecounterc. 77;at girlover there leftamessage for you at the counter d. 77,is girl just leftamessage for you atthecounter e. 77:e girl leftamessage for you atthecounter f. ZadieSmith leftamessage for you atthecounter g. Zadie leftamessage for you atthecounter
h. The 24-years old best-selling novelist from North London leftamessage for
YOU
i. She leftamessage for you atthecounter j. She leftamessage for you atthecounter
k. She wroteamessage for you and 0 left it atthecounter
The italicizednoun phrases in thesesentences can all be used
for
reference to the same person, the author Zadie Smith, depending on the context inwhich the
sentence isuttered. Letmedescribe just a few ofthedifferentcontexts: Sentence (a)
maybeuttered byaperson who does not knoworrecognizetheauthor, but saw her
deliveramessageandsubsequently reports it tothe addressee. Example (c) may be used whenthereferent Zadie Smithisunknown tothediscourse participantsbut
still
intheir vicinity, thatis,withinthephysicalcontext ofthe discourse.
A
sentence like(g) may be used
for
example in a situation in which the discourse participants arefamiliar orevenfriends withtheauthor. Sentence (i) can be used whentheauthor is
currently the
topic of
a conversation between the discourseparticipants, in a
sequencelike: I just saw Zadie Smith leaving the building. She left a message for
expressions used in different contexts range from elaborate relative clause NPs to pronouns orzero anaphora.
In narrative discourse, narrators may refer tocharactersin various ways. In
fact, it isvery commonforcharacters tobeconsecutivelycodedbydifferent types of
referential expressions, at differentpoints during the discourse. Let me give a few briefexamples: In thefollowingexcerpt fromashort storythenarrator uses both full
NPs (proper nouns) and pronouns in reference to the protagonist (in this and the
following examples, target references are given in bold or initalia):
(2) MYERSwas travelingthroughFrance in afirst-class rail car on his way
to visit his son in Strasbourg, who was astudent at the university there.
He hadn't seen the boy in eight years. There had been no phone calls
between them during this time, not even a postcard since Myers and the boy'smother had gone their separate ways - the boy staying with her. The
finalbreak-up was hastenedalong, Myers alwaysbelieved, by the boy's
malign interference in their personal affairs. (RC.TC: 47)1
As you can see in the example above, the main character (in bold) isintroduced by
a
full
nominalphrase (a proper noun), followed by pronominal references, in turnfollowed by tneresumption offull nominal reference. Also note that reference to a
secondary character, in the same excerpt (in italics), is maintained through the
continued use
of
definite full NPs.Instead
of
using full NPs(propernouns) andpronouns,narrators may also use an alternative description to refer to a character. Take the following examplefromachildren's book byRoaldDahl (Preceding this excerpt isapassageinwhich
the parents of four year
old James Henry Trotter are eaten up by an angry rhinoceros):(3) Now this, as you can
well
imagine, was arathernastyexperience for two such gentle parents. But in the long run it was far nastierfor
Jamesthan itwas for them. Theirtroubles were all over in ajiffy; they were dead and
gone in thirty-five seconds flat. Poor James, ontheother hand, was still
very much alive, and all at once, he foundhimselfaloneandfrightened in
avastunfriendly world. (RD.JGP: 7)
Consider anotherexample froma Roald Dahlstory, inwhich littleGeorge
is referred to by pronouns:
(4) Oh, how
he hated Grandma!He
really hated [italics in original] thathorridoldwitchywoman (RD.GMM: 12)
1
Examples takenfrompopular andliterary fictionare labelledbyabbreviations. The sourcesarelisted in
Introduction 15
Inthefirstsentence,
it
seems as if wehearGeorge'svoice rather thanthenarrator's.Nevertheless, the narrator is notquotingthe main character, but refers to him with
the third person pronounhe. Notice that in this sentence involvingthe character's 'voice', the use of the propername,George, would sound somewhat awkward.
So far we have seen that whenspeakers
/
writers referto charactersorother referents in discourse, they may doso using various typesof
referential expression types, such as (in)definite NPs, proper nouns, demonstrative NPs, pronouns or zeroes.Throughout this work we will find that in narrative discourse, reference to protagoniststypically displays what I will callapropernoun / pronoun alternation.
Also in colloquial, non-literary narratives, topical characters are often either
explicitlyreferred to by means of
full
descriptions(mostly propernames),orcodedbyless informative forms such aspronouns (orzero forms). This alternation can be
illustrated bythefollowing(attested) narrativetext2:
(5)
(a) De zomervakantie isinzicht. The summer holidays are coming
(b)
Maarje
heeft een idee Maartje has an idea(c) en 0 wil nietgestoord worden.
and 0 doesn't want to be interrupted (d) Ze wileendraak ineenknutselen
She wants to fabricate a dragon
(e) en daar is zeurenlang zoet mee.
Which keeps her (subjecO busy for hours on end.
(f) Nu moetdedraak nog geschilderd worden
Now the dragon has to be painted (g) en dan is hij af.
and then it is finished. (h) Ziezo,dedraakisklaar.
There it is, the dragon is finished.
(i) De vakantieiseindelijkbegonnen The holidays have finally started
0) en Maarge gaat op stap methaarnieuwedraak. and Maartje goes out with her new dragon
(k) Ze wildemensenlatenschrikken She wants to give the people a scare
(1) en 0 hangt haar draak voor het raam van een huis.
And 0 puts up her dragon in front of a window
(m) De man diebinnen in het huis ligtte slapen,
2
This excerptisbased on one ofthetexts from theelicitedCorpus(cf chapter4), slightly adapted for
The man sleeping inside the house (n) begrijpt niet wat er aan de hand is.
Doesn't understand what is going on.
(o) Hij kijkt door het raam
He looks through the window
(p) en 0(hij)ziet:niemand! And 0 sees: nobody!
(q) OndertussenisMaartje atlang uit het zicht.
In the meantime, Maartje has long since disappeared.
(r) Enkele dagenlaterheeftzeopnieuw een idee.
A couple of days later, she has another idea.
(.-)
(s) Maartie is eenslim meisje
Maartje is a clever girl
(t) En ze weetal precies wat ze wil gaan doen.
And she already knows exactly what she wants to do.
Now let us take a look at how
the narrator refers to the protagonist Maar(ie,throughoutthecourse of her stor),3.Theintroduction ofthecharacteroccurs through
the explicit mention of her name.
After
theintroduction of
the protagonist, thenarrator uses pronouns or zeroes, but from time to time she again uses a proper
name, Maarge, to refer to the protagonist. In other words, at times the narrator repeats the proper noun to refer to the main character ofthe
story. Now, is this
repetition oftheproper nounarandom matter, or is
it
possible to detectsomesystem in this phenomenon?In this story, repeated
explicit
reference occurs three times, namely inclause (j) after 7 intervening clauses, then again in clause (q) after 6 clauses, and
also in clause (s) in a different part of the same story. Looking at the conceptual
content ofthe story, wecanobserve thatthenarratorrepeats theprotagonist's name
atthefollowingpoints:
(i) In clause (i) the textconveysashift in timeand situation: It is the
beginning of
the holidays, which represents achange in the
situation of
the protagonist. This narrativeshift in time and
circumstancesis immediately followed, in the next clause (j), by a repeatedproper noun to refer totheprotagonist.
(ii)
Thebeginning ofthestory isallabout themain character, Maartje.In clause (m), a
new character is introduced in the story, theneighbour. The nextfewlinesareconcerned with the situation and
experiences of this neighbour, leaving out
reference to the
protagonist
altogether: Four
lines later, in clause (q) the story returns to the protagonist; the focus ofthe storyhasshifted from3 1 willusefemininegender toindicatethe language producer, i.e. the speakeror writer, andmasculine gender for the hearerorreader.
4 Note that in line 16'nobody' isanindirectreference totheprotagonist:Although theneighbour does
Introduction 17
the protagonist to another character, and then shifts back to the
protagonist; and this character shift is
accompanied by the
resumption oftheproper name.
(Hi) In clause (s) the narrator gives us her own interpretation rather
than a report of the main character's ongoing activities and the
events she is involved in. That is, this so-callednarrator comment
represents a description of a permanent and inherent property of
the protagonist, which is accompanied by a repeated use of the
protagonist's name.
What may strike us when observing this pattern isthat, strictly speaking, it does not
seem at allnecessary torepeat explicitreference totheprotagonist.Throughout the story,
it
shouldbe clear that the narrative revolves aroundthe central character. In addition,ambiguity is ruled out (there is no other female character, so thepronoun she cannot but refer to the protagonist). So why repeat the proper name at all?Judging from the idea thatonlyambiguityneeds to beavoided, repeatingtheproper
noun might seem redundant. In the texts shown above, however, the narrators do
alternate between using names and pronouns (or zero anaphora) for the main
characters in theirstories. Asweshall seein laterchapters, animportant role can be ascribedto referentsalience within the embedding context.
The aim of
this study is to explain the patternsof
alternating full andattenuated nominals, in consecutive reference to narrative
characters, in a
cognitively plausible way. The present study focuses on referential
choice, i.e. it
takes a production rather than comprehension or processing
perspective. I will
investigate the process of the production
of
referential expressions, rather than the (more widely investigated) processof
anaphora resolution. Iwill restrict my
research mostly toone specific area: consecutiver ferenceto topical characters in Dutch written narrative discourse.The literature on discourse reference reveals numerous factors that can be
associated with the choice
of
referential form.However, most of
the existingresearch consists
of
qualitative case studies, and involves carefully edited texts.Most
of
thesestudies, that is, are notbasedon systematic statistical analysis of data collected undermaximallycontrolled experimental conditions. Further, most studiesfocus on one or two of the many factors that mightbe assumed to be relevant to
referentialchoice. Not all ofthem, lastly,involveindependently established theories
of
therepresentationofdiscourse.This study
of
coursebuildson importanthypotheses andinsightsoffered inthe previousliteratureon discourse reference, and aims to complement the existing
literature in a number of ways: by offering statistical analyses
of
experimentallyelicited production data; by considering several factors simultaneously; and by
attempting to situate thedata
within
atheoretical frameworkfor
nominal categoriesand narrative representation. In doing so, this study aims to assess the relative
influence
of
several factors affecting referential choice, and to shedlight on the
The remainder of this introductory chapter consists offour sections: Section 1.2
brieflypresents thetheoretical background.
In
section 1.3 Iwill
describe the set-upand approach ofthestudy; section 1.4 summarizes the mainresearch questions; in section 1.5,lastly, I
will
presentanoverview ofthe contentsof
chapters2through 7.1.2 Areferencepoint approach to discourse reference
Various studies have described the form and meaning of the different
types of
referential expression, as well as the mechanisms and conditions
of
their use and interpretation. The central aimof
thesereference theories is toaccount for the form and interpretationof
(anaphoric) discourse referents. That is, they attempttoexplain (i) the way aspeakerintroduces anew referententity intothediscourse and the wayshe subsequently maintains reference to it in the discourse, using varioustypes of
referential expression; and (ii) the way a hearer resolves what discourse entity the
speaker
refers to, and the way
he integrates a referential expression into thediscoursemodel.
In most theories
of
discourse reference, referential form is related to the information status or cognitive smtusof
referents (Chafe 1976, 1987, 1994; Giv6n1979,1983; Prince 1981;
Ariel
1988,1990,2001; Tomlin 1987;Tomlin & Pu 1991;Kibrik
1999;Gundel et al. 1993; Van Hoek 1992, 1995, 1997). That is, in
accounting
for
online referential choice and anaphora resolution,there is an
important role for cognitive processing factors such as concept activation, or, the
(assumed) accessibility or salience
of
mental representations. The guiding assumptions forthepresent studyarebasedprimarily on acognitive semantic viewof nominal categories in terms
of
salience (Van Hoek 1997). Proper nouns andpronouns, on this view, reflect different degrees
of
referent saliencewithin the
immediate context, as represented in the
minds of
the discourse participants: Apronoun represents a high degree ofreferent salience within the embedding context, and a proper noun represents a relatively low degree of referent salience within the
context.
The theoreticalapproach taken in this studyis largelybased ona
discourse-level application of Van Hoek's (1997) reference point model of anaphora. Van
Hoek's model, which comprises a conceptual-semantic account
of
sententialanaphoraconstraints,isexpanded so astoaccount
for
referential choices atthelevelof
discourse. Within the proposed discourse-level model, nominal entities serve asconceptual reference points (,cf. also Langacker 1990), against which the
surrounding context is to be interpreted. The stretch
of
context within which a referent is the most salient entity and functions as conceptual reference point istermedadominion.
Intheextension of this model to the level
of
discourse, I presentacognitivesemantic characterization
of
proper noun/
pronounpatterns in discourse. I proposeIntroduction 19
context (what Van Hoek (1997) terms reference point
/
dominion organization)keeps track ofthe evolving relation between topic referents andlocalcontext. This organization in turn determines the alternation of
full
(proper noun) andattenuated(pronominaD reference.
Therelation between reference points, dominionsandreferential choice can be describedasfollows: As long astheintendedreferentfunctionsasreferencepoint
within the embedding context, a speaker or writer continues to pronominalize the
referent; if forsome reason thereferent is no longer (tobe presented as) thecentral
element,or, referencepoint intheembedding context, the speakerorwriterrepeats explicitreference,
for
exampleby usingaproper noun.It is hypothesized that the extent of a referential dominion, that is, the
extent ofhigh referentsaliencewithinacertain stretch
of
discourse context,dependson a number
of
factors:thetopicality of
thereferentwithin
theentire discourse; thereferent's salience
within
the clause (its syntactic function and its linearposition);the linear textual distance to the previous mention ofthe referent(in terms of the
number
of
intervening words and/orclauses); the presenceof
intervening referents;the embedding narrative structure; and characterperspective. One ofthe relevant factors, narrative structure,isillustrated below:
(6) As George sat there pondering this interesting problem, his eye fell
uponthebottle
of
Grandma'sbrownmedicine standing on thesideboard.Rotten
stuff it
seemed to be. Four times a day a large spoonful of it wasshovelled into hermouth andit didn't do hertheslightest bit of good. She
was always justashorridafter she'd had it as she'dbeenbefore. Thewhole
point
of
medicine, surely, was to make apersonbetter. Ifit
didn't do that,then itwasquiteuseless.
So-ho!Thought Georgesuddenly.Ah-ha! Ho-hum!Iknowexactly
what 1'11 do. I shall make her a new medicine, one that is so strong and so fierce andso fantastic it willeither cure hercompletely or blow off the top
of her head. I'll make her amagic medicine,a medicine no doctor in the
world has evermadebefore.
George looked atthekitchenclock. It said five pastten.There was nearly an hourleftbefore Grandma's next dose was dueat eleven.
'Here we go, then!' cried George,jumping up from the table. 'A
magic medicine
it
shall be!'(RD.GMM: 14)
Episodetransitions (representedherethrough paragraphs)involveaconceptualshift
inthenarrative representation, or, situation model (cf. Zwaan
&
Radvansky 1998); achange in (one or more) narrative characteristics such as location, time, cause,
character
and/or
motivation (ibid.)may close off areferentialdominionandtriggerthe repeated use ofa proper noun. The narrator can alsoexploitthe referencepoint
principles sketched above, for communicative purposes: continued
pronominalization enhances conceptual continuity, whereas repetition ofa proper
The proposed model is tailored specificallytotopic maintenance,
for
whichpropernouns andpronouns, itisargued, arethetypical categories. It does not apply to referents that are 'new' inadiscourse(segment), orto casesinwhichthereferring
expression itself is used to add information about the referent, as in example (3)
above.
The corpus analysis presented in chapters 5 and 6 addresses the extent to
which therelevant factors distinguished on thebasis ofthis model indeed underlie
theactualreferential choices made
by
narratorsin online narrative production.13 Approach and set-up of the study
The present study is conducted within the overall framework
of
CognitiveLinguistics
(Lakoff
1987, Langacker 1987, 1991, Talmy 2000, Fauconnier 1994,Barlow
&
Kemmer 2000). Research conducted within this framework, whichdeveloped in the past two decades, focuses on the relations between language use,
languagestructure,andgeneralcognition.
Given its theoretical orientation, Cognitive Linguistics can be considered closely related to the field
of
psycholinguistics. Because of the emphasis ona usage-based language system, and on the important role ascribed to general (non-linguistic) cognition, cognitive linguistics naturally lends itself to investigations involving linguistic processing, and to empirical validation through (quantitative) psycholinguistic methods5. Pioneers in the fieldof
cognitive linguistics, such asLakoff (1987)
and Langacker (1987, 1991) have largely relied on establishedlinguistic analytical methods. Recent years have witnessed more research in
theoretical linguistics employing psycholinguistic methods suchasexperimentation
and corpus analysis (Geeraerts 1999, Verhagen 2005, Goldberg 2006 inter alia).
Both analytical and psycholinguistic (quantitative) approaches seem to have their
own advantages: Fauconnier (1994) remarks that a qualitative investigation of extraordinary cases might bringto light the processes that are normaUy at work in
language. Geeraerts (1999) argues that adopting a usage-basedframework such as Cognitive Linguistics implies that its methodology needs to include quantitative
research methods such as those used in corpus research and psycholinguistic
experiments. In my view, the
inclusion of both
an analytical and a quantitativeapproach
might be
an advantage. AsTalmy (2005) puts it, "each of the
methodologies nowbeing applied to cognitive linguisticshasunique capacities that
make it necessary for our overall understanding
of
conceptual structuring inlanguage, as well
as havinglimitations that make
the other methodologiesadditionallynecessary for this understanding" (Talmy 2005: 11).
This study
is situated at the crossroadsof
cognitive linguistics andpsycholinguistics. I
will
employ both qualitative and quantitative methods in theIndeed one oftheobjectionsagainst mainstream generative grammar has been that, although it aims to accountfor language as abiologicaland psychological phenomenon, it does notlenditselfeasily to
Introduction 21
following way: Qualitative analyses are
employed for
thedevelopment of
hypotheses concerning usage
of
proper nouns and pronouns. Using mostly literaryexamples, I
will
describethevarious factors,listed in 1.2above, that canbe assumedto affect referentialchoice. Thesefactors will be put to the test in anempiricalstudy
of
Dutch narrative production. Due to the fact that several factors canbeassumed to(simultaneously) influence referential choice, a quantitative analysis is especially
warranted, in order toassesstherelativeimportance of each
of
thesefactors, and toassess whether they cancel each other out. Therefore, two quantitative methods, namely frequency analysis and regression analysis, are used in the analysis of an
elicited corpus: I present afrequencyanalysis ofthe distribution
of
referential form relative to the factors at issue(chapter 5); I
also present a regression analysisassessing theweight oftheindividualfactors (chapter 6).
In order to obtain relevant production data forthe quantitative (frequency
and regression)
analysis, I elicited
a corpusof
written Dutch narrative texts.Participants were askedto produceawritten narrative on thebasis ofaseries of 25 pictures. The pictures told a children's story about one protagonist
involved in
several events. Some of the factors assumed to affect referential form - episode
boundaries, viewpoint boundaries and perceptual attention
shifts - were
implemented in the picture series. In this way I obtained a corpus ofstructurally
similar,comparablewrittennarratives.
Inthe corpus analysis, I investigate the way the relevant factors influence referential choice during the onlineproduction
of
written narrative texts. On thebasis of the results, it is assessed whether the findings are compatible with the
proposedsalience-basedcharacterization ofthecategoriesproper nounandpronoun.
I will also
briefly address the production rules or principles that might guide theonline selection
of
referential form. Withthisstudy, then, I hope tocontribute bothto a psycholinguistic model oflanguage production, and to a theoretical model of
nominalsemanticsandpragmatics.
1.4Research Questions
Theresearchquestions that willbecentral to the present study areasfollows:
1. How can
we account for the patternsof
alternating proper nouns andpronouns in narratives in
terms of
a conceptual-semantic salience-basedcharacterization
of
nominal categories? (Chapter 2).2. Given that
this salience-based characterization relates to the embeddingcontext, which conceptual and discourse-structural characteristics of
narrative discourse shouldbedistinguished? (Chapter 2)
3. During narrative discourse production, what are the relevant factors for the
distribution
of
proper nounsversuspronouns, referring to topical referents?(Chapter 5). More
specifically, what is
the (relative)influence of the
4. Based on the observed factors, is
it
possible to characterize the linguisticknowledge, rules, and /or communicative strategies that govern the
real-time production
of
proper nounsandpronouns? (Chapter 7).1.5Contents overview
In this chapter I haveprovided an outline ofthe issues to beaddressed inthis study. Thissectionbrieflydescribes thecontents
of
chapters2through 7.In chapter 2 I presentthetheoretical background tothestudy. Itishypothesized that referent salience and narrative context can account for the coding
of
consecutive topic reference in narrative discourse. The chapter consists ofthefollowing: (i) it
presents discourse referents as mental entities; (ii) it gives a description of the
nominalcategoriesproper noun andpronounin terms
of
salience; (iii)it
presents anoverview ofVanHoek's (sentential) referencepointmodel
of
anaphora (1997); (iv)it presents an extension of this model to the level
of
discourse, distinguishing a numberof
grammatical, discourse andnarrative structural factors assumedto affectreferential anaphoric patterns in narratives; and (v) it provides a description and
illustration
of
these factors, using examples from English fictional narratives. Themodel presented in this chapter forms the basis ofthe hypotheses for referential
choiceinnarratives, tobeinvestigated inthecorpus study.
Chapter3 presentstheoperationalization ofthefactorsassumedtoaffectreferential
choice in Dutch written narratives. The relevant factors are all
based on a
characterization ofthe nominal categories proper noun and pronoun in
terms of
referent saliencewithinthecontext, asproposed in chapter 2. Some
of
thesefactorsare arather straightforwardapplication ofthe discourse level referencepointmodel of anaphora, someare specific to discourse (rather than sentential) contexts, some
are basedon previousresearch.
Chapter 4 presents themethodological validation ofthe research.
It
describes how the production data havebeencollected. Iwill
argue thatthemethod used tocollectthe corpus-elicitationthroughvisualstimuli (comics) - is avalid way
of
obtainingnarrative production data. I
will
demonstrate that, inthegenreof
comics, discourse-structural characteristics suchas episodic structure can beconveyed throughvisualmeans. I
will
describethe implementation ofsuch factors inthevisual stimuli, andthe productiontask
ofwriting
narratives on thebasisof
thesevisualstimuli.Chapter 5 addresses thequestion
of
which factorsarerelevanttoreferential choice (i.e. the first partof
research question3 above). To this end, I present a frequencyanalysis of the distribution
of
referential expressions in the collected corpus,Introduction 23 First, I present a general characterization ofthe corpus, establishing the linguistic
topic status oftheprotagonist. The main part ofthe chapter isconcerned
with
testinga number
of
hypotheses presented in chapter 3.It
presents the distribution ofreferential form relative to the clause-internal factors syntactic function and linear
position;
it
analyses the discourse-structural factors episode boundaries, viewpointshifts, andpage breaks (reflecting perceptual attention), implemented in the visual
stimuli; it
alsopresents a more fine-grained analysisof
discourse-structural factorsin terms ofthesituation modelparameterscharacter, location and time; the chapter also addresses the hypothesis that repeated proper nouns after episode boundaries may serve the function
of
signalling the discourse structure (apart from the basicidentifying function); lastly, the chapter includes an analysis
of
referential formrelative todegreeofcharacterperspective.
Chapter 6 addresses the question oftherelative importance ofvarious significant factors in referential choice (i.e. the second part
of
researchquestion3 above). It isconcerned
with
modeUingtheproper noun/
pronoun alternation.Whereasin chapter5 I report
the observed frequenciesof
proper nouns and pronouns in differentdiscourse situations, in
chapter 6 I
present estimations of theindividual and
cornbined contribution
of
these factors totheprobability that a narrator will use a repeatedproper noun (rather thanapronoun), throughalogistic regression analysis.In this
way, several factors that usually co-occur in discourse production can bedisentangled. Using a
selected part of
the collected corpus, Iwill
report theindividual and relative
effect of
the factors referential distance (in words andclauses), episode boundaries, viewpoint shifts, intervening referents, and syntactic
function.
Chapter 7 presents the conclusions ofthe study. I
will
propose that the saliencecharacteristics ofthereferentialcategoriespropernounandpronoun, combined with
the conceptual structure ofthe narrative contexts in which they occur, can explain the observed proper noun /pronoun patterns in consecutive reference to narrative
characters. The chapter also offers an
outline of
a processmodel of
the choice between proper noun and pronoun, involving two basic mechanisms, based onA
discourse level
exploration of
the
reference
point
model of
anaphora
2.1 Introduction
This chapter forms the theoretical background to the empirical study presented in
subsequentchapters.Itpresents salience as thecentralnotion
for
topic maintenance,and describesthecharacteristics inthenarrative contextwhichmightaffect referent
salience.
The set-up ofthechapter is asfollows: Section2.2 presents theconceptual conditionsfor linguisticreferencetodiscourseentities
(following
Jackendoff 2002). Section 2.3 presents the processconditions for particular Opes of reference: it
presents a conceptual-semantic description ofthe nominal categories proper noun
and pronoun,asadopted in Van Hoek (1997). Section 2.3also presents thereference
pointmodel
of
sentential anaphora constraints, introducing the factors linear order, conceptualconnectivityandpoint of
view,which affectreferent saliencewithin theclause andwhich will turn out to playanimportant roleindiscourselevel anaphora
aswell. Section 2.4presents thebasic arguments
for
treatingsentenceanddiscourse level anaphora in the same way, and for adopting this particular theory for thedevelopment
of
hypothesesconcerning referential patterns in narrative. Italsogivesa description ofthe representation
of
reference points and dominions in narrative discourse. In 2.5,thediscourselevelfactorsforreferential choiceare described and illustrated with examples from English fictional narrative. Section 2.6, lastly,presents thesummary ofthe chapter.
2.2Discoursereferentsasmental entities
This section presents theconceptualconditions for our ability torefer to things and persons in real orfictionaldomains. Theattachment ofaso-called indexicalfeature
to a perceived or conceived entity, which singles outan entity as an individual, is
what allowsfor(consecutive)linguisticreference. Withinthecontext ofthe present
study, the attachment of an indexical feature to perceived or conceived entities allows narrators to produce chains
of
linguisticreference to a single character in anarrative. In the elicitation task used here (cf. chapter 4), narrators track the
16 Chapter 2
present the protagonist as a single individual throughoutthestory, even when it is
absentin certain pictures,or depicted fromadifferentvisualangle.
Bothin discourse and in everyday life,people attribute identitiestopeople, animalsandthings, invery differentsituations: Takeforexample the case
of
seeing,hearing, strokingandtalking about one's favourite cat. This involvesvery different
sensory and conceptual processes. Moreover, even within a single modality, the
same cat can be seen from very differentangles,yielding different and incomplete images.Yet people somehow manage to identify the cat as one and thesame
'thing',
throughout the course of time. In this section I use
Jackendoffs (2002) view of
linguisticreferencetodemonstratehowconceptsand percepts give risetochains of
referencein discourse, throughtheattachment
of
indexical features toaperceived or conceived object.As a starting point, I adopt the now widely accepted view that linguistic
referenceprimarily involves the relation between linguistic expressionsand mental
concepts, rather thantherelation between language andreality (the outside world)1
(Johnson-Laird 1983, Langacker 1987, Jackendoff 2002). As Jackendoff (2002)
points out, people often refer to 'objects' that cannot easily be defined as realistic
objects in 'the world'. He illustrates this claim by numerous examples: fictional
characters suchas Sherlock Holmes', geographical objects such as Wyoming, social
entities such asthevalueOfmy watch; auditorilyperceived objects suchasMahler's
Second Symphony, andvirtual objects such as thesquare formed by four dots (cf.
figure 1 below). (An extensivereview
of
these andother examples canbe found inJackendoff 2002: 300-3).
Figure 1 (Jackendoff 2002: 301 )
These examples illustrate that linguistic expressions refer to conceptualstructures
rather thantoobjects in theworld, independent of the mind.
Inwhatfollows 1 will use Jackendoffs (2002) example
of
visual perception to illustrate how chainsof
perception andreference comeabout.Jackendoffargues that percepts and concepts arelinked to
an 'index' feature,which in
turn enablesconsecutive verbal reference to it.
Jackendoffs (2002: 306) example runs as
follows:
(1) Hey, look atthat!
Ipointingl
When a speaker utters a sentence such as
(1),the
understandingof
deictic 'that',which contains little descriptive content,
involves more than just
the linguisticprocessing ofthe sentence;the hearer must make use ofthevisualsystemin order to
1
identifythereferent. It is not theeyes,however, thatidentifywhatever the speaker is pointing at: The immediateretinal imageis sensitive onlytostimuli suchas colour,
contrast,
intensity etc; it does
notitself
distinguish specific objects or objectlocations. As Jackendoff puts it:
"inboard from here it's
all computation" (2002:307): It is
the brain that transforms the retinalimage to
a"percept", a
"cognitive/neural structure that distinguishes
individuals in
the perceived environment and that permits one to attend to oneoranother of them. One can stop attending to aperceived individual andthenreturn to it; one can track aperceivedindividual as
it
moves through the perceived environment and asit
changesproperties such as orientation, color and shape" (2002:307). What are the crucial
features of suchapercept, andwhatmakes(repeated) linguisticreference to such a percept possible?
Jackendoff elaborates on the exampleHey, look at that! by exploring what
is involved when that refers to adisgusting bug that iscrawlingacrossthefloor. He
explains thata percept that thehearer mustidentify in orderto connectthe deictic expression to an entity perceived "out there", contains at least the
following
features:First, the percept containsanumberof
descriptivefeatures. The features ofavisual percept such as the oneillustratedhereconsist ofsize,color, shape, as well as location and motion. The descriptive features are not necessarily
visual. In a
variant of this example, Did you hear that?, the featuresareauditory. Percepts come
indifferentmodalities, such asseeing,hearing,smellingandfeeling.
Another importantaspect ofapercept is that itisperceived as a (moving)
figure
against a surrounding background. The fact that aperceived entity in someway stands out from itsperceived surroundings forms the initial basis
for
buildingan indexical feature of the percept,which 'marks' the percept (orconceptwhen it is
established through language) asan individual entity: "[the indexical feature] gives
the [...]minda 'something'towhichdescriptive features canbebound" (Jackendoff
2002: 311). This feature, once established, is no longer dependent on perceptual
input, but allows ustofurther
'track'
theentity throughout time; even when it is nolonger perceived itisassumedto continue to exist, and canbeperceived again, and
recognized as the same entity, in any of the existing modalities. The indexical
feature, then, is modality-neutral and is used
for
example in tracking movingpercepts, and inidentifying perceptsorconcepts thatdisappear andre-appear from the focus
of
(visual,verbal,auditoryetc)attention:Jackendoff argues that
"[t]he
indexical feature ofa percept is the crucialfeature for linguistic reference" (Jackendoff 2002:314). These indexical features,
whether established perceptually or linguistically, are also necessary for chains
of
linguistic cor€ference.
A
proper name, which denotes anindividual, "has an
indexical in
its associated concept"(Jackendoff 2002: 318).
An attenuated expression such asapronoun does not contain suchanindexical feature; apronoun2 Anindexical feature can also besplit ormerged:forexample,asingle objectmay break into two or
more pieces, ortwo lumps of clay canbemoldedtogether(evoking evolutivereferents, cf. Maes 2001).
Othercharacteristicsofpercepts andconcepts, such as itsvaluation,itsclassificationasself-produced vs notself-produced, meaningful vs not meaningful and familiar vs.novel,are described inJackendoff
28 Chapter 2
such as she merely designates a female animate entity. Nevertheless, an indexical
feature should somehow be present in order to produce the
pronoun in any
meaningful way. The production and interpretation
of
pronouns is thereforedependent on context for the attachment of an indexical feature. As we shall see
below,a prerequisiteforusing such an attenuated expression,which in
itself
lacksanindexicalfeature, is that itbehighlysalient
within
thecurrent context.2.3Thereference
point
modelof
anaphoraThis sectionpresentsan overview ofVan Hoek's (1995,1997) sentential r€terence point model of anaphora.
Van
Hoek's characterization of nominal semantics will beaddressed in2.3.1 below; theconstraints on
full
nominalandpronominal sentential anaphora willbe addressedin section 2.3.2.2.3.1 Nominal Semantics
Whereas the previous section addressed the conceptual prerequisites
for
linguisticreference, this section discussestheprocess conditions for choosing a particular type
of
referential expression. To this end, this section presents Van Hoek's (1997) semantic characterization of the nominal category, which largely draws on the notionaccessibilio'(Ariel 1988),equivalent to VanHoek's notionsalienc/.
During the production and comprehension
of
discourse, attention flowsfrom one focus to the next and concepts
continually move into and out of the
immediate focus
of
consciousness (Chafe 1987, 1994; Langacker 2001). Chafe (1994) argues that only acertain amountof
information can be'mentally active' orfocused on at acertain time. Therefore, conceptshavedifferent activation states in people's minds at differentpoints in thediscourse. Linguisticcategories respond to
thesedistinctions: referential expression types such as(in)definitenominals, proper nounsandpronouns codeinformation pertaining tothe currentmentalaccessibility of adiscourse entity,asrepresented intheminds ofthediscourse participants (Chafe
1987, Ariel 1990 inter alia). The nominalcategoriesdefinite full nominal (including
proper noun) and pronoun code different degrees of referent salience or
accessibility: Full nominal expressions such as proper nouns are so-called low accessibilitymarkers, i.e. they indicate thattheintended referent has alowdegree of accessibility. Pronouns, on the other hand, are high accessibility markers, in that
they indicate thatthereferentiscurrently highlyaccessible.
On this view,the choice ofa certainreferential expression largely depends on the (assumed) salience ofa referent within the current context of use.
A
recently' Although Van Hoek (1997) adopts Ariel's (1988) notionofaccessibdityin characterizing nominal categories, in her own analyses she mostly uses the equivalent term salience (and sometimes,
prominence). Iwill therefore also use the term sahence. The readershould keep in mind however, that
mentionedreferent
for
exampleisassumed to be intheforefront oftheinterlocutor'sconsciousness, and may thereforebeaccorded a high degree
of
salience. Therefore,it suffices to use a pronoun to refer to that referent. A referent may also be in the
forefront
of
awareness of the discourse participants due to other, non-linguisticcircumstances, witness the following discourse-initial referencetoSaddam Hussein duringapressconference4:
(2) LadiesandGentlemen: We ['ve] got 'em
At the press conference announcing his capture, the referent Saddam Hussein was
on everybody's mind, since rumours ofhis capture were alreadycirculating and it
was considered very important news. The pronoun reflects the referent's high
saliencewithinthecontext ofthe pressconference.
Van Hoek(1997) distinguishes anotheraspect
of
nominal semantics - onethat shecharacterizesas analogous tothenotion
of
salience:full
nominals (such as proper nouns) and pronouns differ with respect to the degreeof
subjectivity orobjectivi)
they impose on thereferent(cf. Langacker 1990). Van Hoek's particulardistinction between subjective and objective construal refers to what Langacker
(1990) calls the stage model, i.e. amodel of the discourse situation in which both
discourseparticipantsand conceptualised entities
differ in
the degree towhich theyare themselves portrayed as 'conceptualizer' (in Langacker's terms),
viewer or
cognizer ofthedescribedsituation.
For currentpurposes, the subjectivity / objectivity distinction in nominals
can be described as follows: In perspectivized contexts, a (subjective) pronoun
portraysits referentasconceptualizer ofthepredication
of
which it is part whereasan (objective) proper noun portrays its referent as the object
of
conceptualisation,from an 'outside' perspective, i.e. that
of
eitherthenarrator oranother (secondary)character.
Van Hoek
also describes the subjectivity distinctions as follows:"Reference via name implies greater distance between the conceived referent and
the[discourse] participants, andacorrespondingly more objective conception of the referent. A pronoun portrays the referent as conceptually closer to the discourse participants, andcorrespondingly as more subjectivelyconstrued" (Van Hoek 1997:
219).
As we shall seeinsection 2.5.7 and in chapter 5, I
will
relate thisnotion tothecharacterization givenin Sanders(1994),which offersreliablelinguistic criteria for distinguishing ways inwhichthe consciousness
of
discourse participants as wellasembeddedcharacters canberepresented in the text
4 Cf. Ariel (1988, 1998)forfurtherdiscussion on the relevanceofreferentaccessibility/saliencerather than thereferent's 'geographical context' in explaining referentialpatterns (geographicalcontext referring
30 Chapter 2
2.3.2 Sentential anaphora constraints
Within Van Hoek's reference point model
of
anaphora, thefelicitous use of a
coreferential
full
nominalorpronoun depends onthe salience ofits referentwithinthe immediate context,
which is in
turn determined by three factors: conceptual connectivity, linear order, andpoint of view.
In order to plausibly relate (pro)nominal anaphora to the relevant characterizationof context, Van Hoek
develops a model of semantic relations between nominals, in terms
of
referencepoints and dominions. Reference points function as local topics within a semantic
domain, or dominion.
A
dominion is defined as amentalspace (Fauconnier 1994),consisting of the (conceptual, non-linguistic) context within which a particular
referent is themost salient element. Van Hoek puts itasfollows:
Reference points are elements which are prominent
within the
discourse and so serve to set up the contexts
within
which theconceptualizer makes mental contact with other entities. The
dominion of
a reference point consists of the elements that areconceptually locatedrelative to the referencepoint,whose construal is shaped by their association with the reference point (Van Hoek
1995: 313).
A pronoun can only be used
if
there is a salient antecedent that can function asreferencepoint forthe interpretation
of
thatpronoun. Afull
nominal can be used ifitsimmediatecontextfalls outside thedominion ofacorresponding referencepoint.
Using a full nominal that necessarily fallswithinthe dominion of a corresponding
referencepointsendsthewrongsignalconcerning the referent'sretrievability within thecontext, preventinganintended coreference reading.
This is illustrated in (3) below: The full
nominal Steve falls within thedominion ofthe pronoun he, which, as clausal subject, represents the most salient
entity and the reference
point for the rest of
the clause (i.e., with the rest of theclause in itsdominion). The full nominal Steve can therefore notbe interpreted as coreferential to the subject pronoun.
(3) # He put the money inSteve'spocket
Drawing onanumber
of
theoriesof
discourse reference(Giv6n 1983,Ariel
1990 inter alia), Van Hoek describes the
initial
selection ofa reference point asfollows: "X islikely tobe taken asareferencepoint
relative to Y if X
is[salient] inthe contextwhichincludes Y. Thisreflects thebasic nature ofareference point as
something which is selected on the basis
of
salience and used as a starting pointfrom which to
make mental contact with other, less salient entities" (Van Hoek1997: 58).
The representation ofa salient nominal
within
its embedding (sentential)context, demarcatingtheextent ofthe domain inwhichthenominal remainshighly
organization. The sentential organization
of
reference pointsanddominionsdepends on the salience ofa nominal (its reference point status) and (the strength of) itsconceptual-semantic connections with coreferential
nominals. Van Hoek
distinguishes three factors that shape this organization, once an entity has beenselected as reference point:
(i)
conceptual connectivity;(ii)
linear order; and (iii)point of
view. These factors determine the extent ofa referencepoint's dominion,within
which corresponding referents are necessarilypronominalized. I will now
describethesefactors in turn.
Conceptual Connectivity
Conceptual connectivity pertains to the strength
of
relations between nominal elements, beit
within a singleclause, asentence oran entire discourse. Van Hoek defines conceptual connectivity as"theextenttowhichtwo elementsare conceivedas participating together in a larger conceptual unit." (Van Hoek 1997: 61). Such
units correlateforinstancewithverb-argument structures,sentences, orparagraphs.
The degree
of
conceptualconnectivity partlydeterminesthepossibility forareferentto 'escape' the dominion of a coreferential nominal, and to be coded by a full
nominal.
Conceptual connections within the clause are reflected in the grammatical relations between verb, complements,and modifiers. To describe the configuration
of the verb and its complements, Van Hoek uses the notion
of
complement chain.The complement chain draws on the grammatical relations hierarchy: subject >
direct object> indirect object> oblique (Keenan
&
Comrie 1977 inter alia). Thisnotation reflects the hierarchy
of
prominence within the central clause: a clausal subject is more prominent thanadirectobject,which in turn ismore prominent thananindirectobject,which in turn ismore prominent thananoblique complement. But
note that Cognitive Grammar views such syntactic relations as surface
manifestations
of
underlyingconceptual-semantic structure (cf. van Hoek 1997: 10,66). Apart
from these verb complements, there may be (coreferential) nominalswithin
various typesof
sentential modifiers. These represent the least prominent entitieswith
respecttoother entities inthecentralclause.Constraints on coreference
within
the clause largelyfollow from the
prominence asymmetries reflectedinthese patterns; the loweranominal'sposition
on the complement
chain, the lower
its conceptual connectivity to nominals elsewhere intheclause, andthehigherthepossibilitiesfor
construingthereferent asoutside the
dominion of
a corresponding referentwithin
that clause, andconsequently,
for
using afull
nominal.To give
an example, the subject is the most prominent element within thecomplement chain, and therefore functions as the reference
point for
the centralprocessdescribed bythe clause, withallotherentities in theclause in its dominion. This explainstheunacceptability
of
sentence(5) underacoreferential reading:(4) Johnlikeshamother
32 Chapter 2
Van Hoek explains that,thesubjectbeing the most saliententity inthe sentence, he
in (5) isthe mainreferencepoint within the complement chain. Any corresponding
nominal within that chain, such as the possessive
nominal John's within the modifier, must be highly salient as well, because it isconceptualizedrelativeto the subjectreferencepoint. Consequently, the use ofaproper noun, which signals lowsalience, preventsacoreferential reading. Van Hoek puts itasfollows: "Because the subject isareferencepoint with the rest ofthe clause in its dominion,apronominal
subject cannotcorrespond with a
full
nominal elsewhere in theclause" (Van Hoek1997: 66).
Van Hoek argues that clausal conceptual connectivity is not just a
notational variant
of
syntactic (c-command) relations. The basic patterns arisingfromthecomplement chain - such asthetendency fortheclausal subject tofunction
as reference point for other entities in the clause (which are therefore necessarily
pronominalized if theyarecoreferential to it), andthetendency
for
complements5 ofthemain clause verbtofunctionas referencepoint
for
correspondingentities in thesubclause - can be overridden by conceptual-semantic
factors. This can be
illustrated bythefollowingexamples:
(6) * Mary hit him
just
before John got up(7) Mary hithim before John hadachance to get up
(Brugmanand
Lakoff
1987,cited in Van Hoek 1997: 92)Thefirstexample isungrammatical underacoreferential reading, because any main
clause verb complement normallyfunctions as reference
point for
a correspondingnominal in thesubclause, andthe latter should thereforebe coded byapronoun. In
the second sentence theconceptual break improves coreferencepossibilities. That is,
the irrealis subclause describes a situation which does not in reality occur(John gettingup), and thereby presents a conceptual break with respect to the main clause,
which describes a situation which does occur in reality (Mary hitting John). This
allows the referent John to
fall
outside thedominion of
the coreferential object nominal.Van Hoek provides many examples which indicate that semantic characteristics suchasthose in (7) overridethe clause-level generalizations usually captured in terms
of
c-command. Examples like these demonstrate thefundamentally semantic
nature of
the anaphora constraints, which allows for the possibilitythatsententialanddiscourse anaphorabeaccounted for in the same way.Moreover, conceptual connectivitybetween nominal elements represents a continuum typically ranging from the strong connectivity between complements of
the verb,
to weaker connectivity found with clausal modifiers, to weakest connectivity, reflected in discourseunitboundaries. This weakest typeof
conceptualconnectivity is relevant to reference point