Transcript
Page 1: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

This article was downloaded by: [Cornell University Library]On: 19 November 2014, At: 03:21Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

Acta Linguistica Hafniensia: International Journal ofLinguisticsPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/salh20

Givenness revisited indefinite one-anaphora inunscripted Danish dialoguePhilip Diderichsen aa Lund University Cognitive Science , Lund University , Kungshuset, Lundagård, S-222 22,Lund, Sweden E-mail:Published online: 24 Nov 2011.

To cite this article: Philip Diderichsen (2007) Givenness revisited indefinite one-anaphora in unscripted Danish dialogue, ActaLinguistica Hafniensia: International Journal of Linguistics, 39:1, 179-208, DOI: 10.1080/03740463.2007.10414605

To link to this article: http://dx.doi.org/10.1080/03740463.2007.10414605

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Page 2: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

G I V E N N E S S R E V I S I T E D

I N D E H N I T E O N E - A N A P H O R A JN U N S C R I P T E D D A N I S H D I A L O G U E

by

Phil ip Dider ichsen Lund Universi ty Cogn idve Science

Abstrac t Indefinite one-anaphors (e.g. one as in "Phil gave Bob a doughnut because he asked for one ") are at the same time pronominal and indefinite. It is investigated in a small corpus study of unscripted spoken Danish dialogue whether referents of indefinite one-anaphors are as accessible as referents of definite pronouns, and whether they differ in identifiability from referents of definite pronouns. The results suggest that referents of indefinite one-anaphors indeed have approximately the same degree of accessibility as referents of (accented) definite pronouns, but differ with respect to identifiability. These results support a new model of givenness in terms of a multidimensional functional map.

1. Introduct ion Anaphor ic express ions , and pronouns in particular, are well-s tudied in l inguist ics. Definite p ronouns have recieved plenty of a t tendon in different t radit ions, from theoretical syntax, semantics , and pragmatics to more appl ied computa t iona l approaches and psycholinguist ic experimentat ion. However , mos t studies are about 3rd person definite pronouns . Less attention has been given to one-anaphors as in examples (1) and (2) (Dahl , 1985, p . 3 ) ' .

(1) S o m e cot ton t-shirts are expensive , but not the o n e [i.e. (cotton) t-shirt] W e n d y gave Bruce yesterday.

(2) Phil gave B o b a doughnut because he asked for o n e [i.e. a doughnut ] .

Indefinite owe-anaphors are relatively rare, compared to the ubiqui tous definite p ronouns . A s we shall see later, tiiey tend to occur under h ighly special ized c i rcumstances , namely when a category is given in the discourse but no specific ins tance of that category has been individuated. This special nature of indefinite one-anaphors makes it difficult to s tudy them through corpus l inguist ics. T h e intention in the present study is to gather some initial information about attested examples of this type of referring expression. Th i s

' But see (Bhat, 2004), (Haspelmadi, 1997), (McCawley, 1988), (Sag & Hankamer, 1984), and (Sag, 1979).

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 3: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

180 P H I L I P D I D E R I C H S E N

motivates the invest igat ion of a quite intricate sort of negotiat ions, and mus t necessarily m a k e the sample of express ions investigated rather small .

Indefinite one-anaphors are interesting because they possess two features that rarely co-occur : pronominal coding and indefmiteness. In my opinion, none of the exist ing models of g ivenness in the literature account well for this combina t ion of features. For instance, consider Gundel et a l . ' s widely ci ted Givenness Hierarchy, which is a one-dimensional impl icadonal scale of givenness (see figure 1) (Gundel , Hedberg & Zacharski , 1993). Referr ing express ions that require highly accessible ( ' in focus ' ) referents (i.e. p ronouns ) are placed at one ex t reme, whereas express ions that require less accessible and merely ' type identif iable ' referents (full indefinite noun phrases) are p laced at the other ex t reme. T h e ' in focus ' status entails the status 'uniquely idenfifiable' (i.e. the referent is idenfifiable on the basis of the referring expression a lone) , which is a prerequis i te for using a definite form. This in effect predicts that all ' in focus ' referents should also be 'uniquely idenfifiable' . And so on. Genera l ly , each referendal status in the Givenness Hierarchy entails all the statuses to its right on die scale.

uniquely type

in focu.s > activated > familiar > identifiable > referential > identifiable

(/ r/i(,v(N) Ihat^ iheN indef./Au N a N

Figure I. Gundel et al.'s Givenness Hierarchy.

Indefinite one-anaphors require an antecedent jus t like definite pronouns do , and therefore must s o m e h o w be ' in focus ' . Example (3) illustrates this. T h e two ut terances are about equally odd without an available antecedent for the p ronoun .

(3) a. ?Could you take it [what?] with you when you come back? b. ?Could you take o n e [a what?] with you when you come back?

Never theless , indefinite one-anaphors usually refer to entities which are by no means uniquely identifiable, but rather unidenfifiable, or merely ' type identif iable ' in Gundel et al.s terms, jus t like the typical referents of indefinite full noun phrases . This is what makes the ' s ame bird ' readings ( indicated wi th subscripts) in both (4a) and (4b) odd: These ut terances would normally be about a different bird than the first-mentioned one .

(4) There is a b i rd i at the bird feeder right now. a. ? N o w a b i rd i entered the bird house . b . ?No w onei entered the bird house .

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 4: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 181

Based on these l inguisdc intuit ions, indefmite one-anaphors seem to pattern with definite p ronouns with respect to accessibility, but with indefmite full noun phrases with respect to identifiability. This leads to a notion of givenness whe re accessibili ty and identifiability have separate consequences for the use of different referring express ions . I will call this the dissociated v iew of g ivenness . T h e prevalent view of givenness is one where accessibili ty and identifiability are combined into a one-dimensional scale, as exemplif ied by Gundel et al.s Givenness Hierarchy. I will call this the conflated view of givenness .

T h e ques t ion in this s tudy is whether indefinite one-anaphora will ei ther pattern consistent ly with, or differ consistently from, definite p ronouns widi respect to both accessibili ty and identifiability, as would be expected from the conflated v iew, or whether the pattern derived from the linguistic in tu idons above is reflected in text-based measures of accessibili ty and idendfiabil i ty, which wou ld support die dissociated view.

If the dissociated v iew is correct, there should be no difference in accessibili ty be tween the referents of indefinite one-anaphors and the referents of definite p ronouns . Indefmite one-anaphors would s imply be the indefmite counterpar ts of definite p ronouns . O n the other hand, something has to l icense the choice of an indefinite over a definite form, and differences would be expected relat ing to the identifiability of indefinite one-anaphors and definite p ronouns .

These considerat ions can be formulated as two hypotheses , which will be tested in the small corpus study reported below:

(5) Access ib i l i ty h y p o t h e s i s : There are no significant differences be tween relevant indicators of the accessibili ty of referents of definite p ronouns and indefinite one-anaphors .

(6) Iden t i f i ab i l i ty h y p o t h e s i s : There are significant differences be tween re levant indicators of the identifiability of referents of definite p ronouns and indefinite one-anaphors .

These hypotheses will be further refined and operat ional ized on the basis of the theoretical considerat ions below.

2 . C o r p u s T h e corpus used in the study was die dia logue part of the DanPass unscr ip ted spoken Dan i sh corpus^ (Gr0nnum, 2006) . This part of the corpus consists of c lose to 48 ,000 words .

^ Obtained from prof Nina Gr0nnum, Copenhagen University

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 5: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

182 P H I L I P D I D E R I C H S E N

T h e corpus was created in a laboratory setting with 11 dyads of part icipants (acquainted with each other) complet ing Edinburg Univers i ty ' s Human C o m m u n i c a d o n Research Cen t re ' s map tasks (Anderson et al., 1991). Each par t ic ipant had a m a p (see examples of the maps in secdon 2.3 be low) . One , the i n s t m c d o n giver, had a route on his or her map . The other , the i n s tmcdon fol lower, did not. The i r goal was to collaborate so as to reproduce the i n s tmcdon g ive r ' s route on the in s tmcdon fol lower 's map. T h e maps are not exacdy ident ical : Landmarks are missing on one or the other map , a l andmark may appear in t w o different locations on one map but not the other; and the s a m e landmark may have s l ighdy different labels on the two maps . This gives rise to a t m e negotiat ion, with quest ions and answers , backtracks, e tc . Part icipants were explicit ly in formed about these irregularities in written ins tmct ions prior to the recording. It was left to them, however , to discover how and where the maps or the labels differed, and to supply the missing i tems and correct labels on their respect ive maps . Each pair of speakers completed four different sets of maps^. The or thographic transcription of the corpus was searched semi-automat ical ly for occurrences of anaphor ic enlet. Indefinite one-anaphors are accented by default in Danish , but at the t ime of data collection, the corpus had not yet been annotated wi th stress, so this feature could not be used. Instead, a part-of-speech tagger was employed in order to find all occurrences of enlet fol lowed by anything that was not a noun, an adjective, or a participial. These occurrences were then manual ly searched for indefinite o«e-anaphora . T h e corpus was later p e m s e d once more , and a few more indefinite ons'-anaphora turned up. These instances were included. A total of 30 instances were found: 26 of the c o m m o n gender {en), 4 of the neuter gender {et).

For compar i son , the corpus was also searched for definite p ronouns of the c o m m o n gende r {den). This search yielded some 350 instances. Given that there are about as m a n y instances of the neuter gender' ' , there should be in the order of 700 definite p ronouns in the corpus , corresponding to a frequency of 146 tokens per 10,000 words as opposed to the indefinite anaphor frequency of 6 tokens per 10,000 words .

A sample of 60 definite pronouns was extracted. Of these, 30 were chosen from the subset of p ronouns judged by the author to be accented, and 30 from the unaccented subset . This categorizat ion is relevant because of the we l l -known effect of accent ing on the referential propert ies of referring express ions (Kris t iansen, 1996; Vendit t i , S tone, Nanda & Tepper , 2001) . The j u d g m e n t s of the author were corrobora ted by 7 other linguists in a small exper iment . At least

For more information, see http://www.danpass.dk/ and http://www.hcrc.ed.ac.uk/maptask/.

* An assumption that has not been verified, and should only be taken as a rough estimate.

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 6: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 183

7 of the 8 coders agreed on 7 9 % of the items. In terms of C o h e n ' s kappa intercoder agreement (Bakeman & Got tman, 1997; Rietveld & van Hout , 1993; Carletta, 1996), this cor responds to K = 0.68, which indicates that the categorizat ion is reasonably reliable.

3 . Access ib i l i ty of def in i te p r o n o u n s a n d IndeHni te o n e - a n a p h o r s How accessible are indefinite one-anaphora compared to definite p ronouns? Several measuremen t s were performed in order to est imate and compare the accessibil i ty of the referent and/or concept referred to with the three types of pronoun. Referential d is tance to the nearest compat ib le antecedent is the mos t c o m m o n l inguist ic measure of accessibili ty in the li terature (Givon, 1983a; Arie l , 1990; Gri ining & Kibrik, 2005) . The more recent the antecedent , the m o r e accessible it can be assumed to be at the t ime when the pronoun is ut tered. Here , referential d is tance is measured in terms of number of ut terances to the left, as opposed to e.g. G i v o n ' s (1983a) number of clauses to the left. The operat ional definition of an ut terance employed here is inspired by Chafe ' s intonation unit (Chafe, 1994) . As a practical heurist ic , a portion of speech was cons idered an ut terance if it could be perceived as completed in isolation. Both in ter locutors ' ut terances were counted, and if the antecedent occurred in the same ut terance , the an tecedent -anaphor dis tance was scored as 0. First posit ion reduplicat ions as in example (7) be low, a c o m m o n occurrence in spoken Danish (Nedergaard Thomsen , 1992), were treated as anaphors immediately preceded by their antecedent , and consequent ly had an antecedent-anaphor distance of 0.

(7) BI: den der jernbaneoverskcering den den har jeg belt oppe ' that rai lway crossing, that I have all the way u p '

B2: i det nordvestlige hj0me afmin tegning ' in the north-west corner of my drawing '

A var iant of the dis tance measure is the distance to the nearest lexical antecedent , i.e. an antecedent with an explicit nominal head. Again, the c loser the nearest lexical antecedent , the more accessible the concept can be a s sumed to be .

Also , repeated ment ion gives an indication of the accessibili ty of a referent (Lappin & Leass , 1994; Mitkov, 1997; Cristea & Postolache, 2005) . T h e more frequently a referent or concept has been evoked prior to the occurrence of the anaphor , the more accessible it can be assumed to be, an assumption that is often m a d e in computa t ional approaches to pronoun resolution.

Furd ie rmore , the syntactic and morphological nature of the immedia te antecedent is an indication of the sal ience, and thereby accessibility, of an enti ty. For ins tance, an antecedent that is itself an anaphor can be assumed to b e more accessible than one that is a full noun phrase (Kaiser, 2005) . Mos t notably, m a n y studies , for ins tance within die Center ing framework, suggest that there is a

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 7: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

184 P H I L I P D I D E R I C H S E N

preference for the grammat ica l subject to be interpreted as the antecedent in case of ambigui ty (Grosz , 1995; Kehler , 1997; Kehler, 2002) .

Finally, the definiteness and lexicality of the antecedent may give a hint about the accessibil i ty of the referent. The assumpt ions here would be that a definite form of the antecedent renders an entity to be anaphorically referred to more accessible than an indefinite form, and that a pronominal form renders an end ty more accessible than a full noun phrase.

3.1 Referential distance to the antecedent T h e mean dis tance measured in number of ut terances between indefinite OAie-anaphors and their antecedent was 1.4 (sd = 0.9)"\ This measure reflects that the antecedent was typically in the immediate ly preceding utterance. The mean antecedent- to-pronoun distance of the accented definite pronouns was exacdy the same, a l though with a higher variance (t(30) = 0, n.s.). There was a larger difference be tween indefinite one-anaphors and unaccented definite p ronous (t(30) = - 1 . 7 6 , p = 0.09, two-tai led) , as well as be tween accented definite pronouns and unaccented definite pronouns (t(30) = - 1 . 6 4 , p = 0 .11 , two-tai led) .

Anapho r type Avg. referential di stance Indefinite one-anaphors 1.4 (sd = 0.9) Accented definite pronouns 1.4 (sd = 1.3) Unaccented definite pronouns 2.1 (sd = 2.1)

Table I: Average referential distance in number of utterances of different types of anaphoric expression.

There are no significant differences in the data, possibly for lack of statistical power , but there is a tendency for accented forms to have shorter referential dis tances . This result suggests that the differences in referential distance a m o n g the three anaphor types is due to accenting rather than definiteness, and thus supports the hypothesis that there is no difference in accessibility be tween indefinite and definite pronominal forms. What is a bit puzzling is that the tendency is in the wrong direction. G ivon ' s (1983a) topic accessibili ty scale and Ar ie l ' s (1990) Accessibi l i ty Marking Scale both predict that the antecedent of an unaccented pronoun would be closer than that of an accented pronoun s ince, according to them, unaccented pronouns code higher accessibil i ty. The opposi te tendency can be observed in the present data. This may be due to the fact that accented p ronouns tend to shift attention away from the referent that would otherwise be the most likely antecedent and instead pick out some contrast ing

^ About the statistical notation: "sd" means standard deviation, "n.s." means non­significant. t(30) represents a t-test with 30 degrees of freedom. A two-tailed test returns the probability of some measure not being unequal to a reference value, as opposed to a one-tailed test, which returns the probability of some measure not being higher (or lower).

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 8: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 185

'o ther g u y ' . For instance, a grajnmatical object antecedent will typical ly be preferred to a grammat ica l subject in parallel clause sequences such as " John hi t Bill and then H E hit Geo rge" , where nat ive speakers interpret the accented HE as Bill rather than John (Vendit t i et al., 2001) . The same switch-reference or topic cont inui ty/discontinui ty pat tern has been documented for spoken Dan i sh (Krist iansen, 1996). W h e n one looks at the data, it is obvious that topic cont inui ty/discontinui ty p lays a role, especially in situations of potent ial interference. Several of the definite pronouns from the unaccented sample ' bypass ' the nearest poss ible antecedent in favor of the current topic expressed as a noun phrase in a more distant ut terance, whereas the accented definite pronouns tend to prefer the mos t recent possible antecedent (which may , however , be several ut terances away) . Another prominent factor is the fact diat accented definite p ronouns relatively frequently figure in first posi t ion reduplicat ions, which yields an antecedent-anaphor distance of 0, and dius cond ibu tes to the difference. It should be noted, by the way, that Givon himself finds a shorter average referential dis tance of demonstra t ive pronouns than of unaccented pronouns in spoken Engl ish (Givon, 1983b). This is consistent wi th the above explanat ions . Indefinite one-anaphors never figure in first posi t ion reduplicat ions, yet they usually occur prompt ly after the antecedent is first ment ioned, and almost never occur in situations of potential interference. Thus , a l though indefinite one-anaphors never have extremely short anaphor-antecedent dis tances, the mean dis tance is still relatively short.

3.2 L i n e a r d i s t a n c e to t h e n e a r e s t lexical a n t e c e d e n t Whi le the immedia te antecedent is typically very close to a pronoun, the necessary type descript ion in the form of a lexical nominal head is usually a bit farther away. Each p ronoun and indefinite one-anaphor extracted from the corpus has at least one such lexical antecedent. The average distance from an indefinite one-anaphor to the mos t recent lexical antecedent is 3.6 ut terances back. Accented definite p ronouns have a lexical antecedent 2.3 ut terances back on average. The difference is non-significant (t(30) = 1.70, p = 0.09). The difference be tween indefinite one-anaphors and unaccented definite p ronouns is significant ( t(30) = - 2 . 0 5 , p < 0.05), and so is the difference be tween accented and unaccented definite p ronouns (t(30) = - 2 . 9 , p < 0.01).

Anaphor type Avg. referential distance Indefinite one-anaphors 3.6 (sd = 3.5) Accen ted definite p ronouns 2.3 (sd = 2.2) Unaccented definite p ronouns 7.1 (sd = 8.8)

Table 2: Distance to the lexical antecedent of different types of anaphoric expression.

This result repeats the pattern from above: Accented forms tend to have lower distances than unaccented forms. Again , there is no evidence of an accessibil i ty

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 9: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

186 P H I L I P D I D E R I C H S E N

difference be tween indefinite and definite pronominal forms, as predicted in the accessibil i ty hypothes is .

T h e relaUve c loseness of lexical antecedents to accented forms sugges ts that enti t ies referred to using accented forms are introduced relatively late. T h i s is consis tent with the observat ion that accented pronouns are typically not used to mainta in reference to the main topic of a discourse, but rather to shift at tention to contrast ing other entities introduced along the way (Venditti et al., 2001) . If this is in fact the case, this would predict lower values of repeated ment ion for referents of accented pronouns . W e turn to measures of repeated ment ion in the fol lowing section.

3.3 R e p e a t e d m e n t i o n The anaphor ic express ions in the corpus often have more than one pr ior reference. T h e number of prior references is indicative of the accessibili ty of the referent and/or concept being reactivated by the anaphor. The higher the count , the more accessible the referent/concept. It follows from the accessibil i ty hypothesis that there should be no statistically significant difference in this respect be tween indefinite and definite pronouns.

This is in fact the case. The average number of prior references of indefinite one-anaphors is 3.2. For accented definite pronouns it is 3.9. This difference is not significant (t(30) = - 0 . 7 2 , p = 0.47, two-tailed). The difference be tween indefinite one-anaphors and detmi te unaccented pronouns is significant (t(30) = - 3 . 3 1 , p < 0 . 0 1 , two-tai led) , whereas the difference between accented and unaccented definite p ronouns is not significant (t(30) = - 1 . 7 2 , p = 0.09, two-tailed).

Anapho r type Avg. referential di stance Indefinite one-anaphors 3.2 (sd = 2.3) Accented definite pronouns 3.9 (sd = 5.3) Unaccented definite pronouns 6.1 (sd = 4.2)

Table 3: Number of prior references of the different types of pronoun.

This result once again groups the two accented forms together, and the most significant difference a m o n g the three anaphor types thus seems to be d u e to accent ing. This means that again the accessibility hypothesis is confirmed: The re is no ev idence of higher accessibil i ty of indefinite or definite anaphors .

No te that the observed pa t tem is in accordance with G ivon ' s and Ar ie l ' s theories of accessibil i ty. The finding that unaccented pronouns tend to have more prior references suggests that this type of pronoun is to a higher extent used to maintain reference to the main topic over longer s tretches of discourse, wh ich is also consis tent with the referential distance results reported above. Fur the rmore , the result suppor ts the idea that referents of accented pronouns are in t roduced relatively late, and function to shift attention away from the main topic.

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 10: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 187

3 .4 T h e s y n t a c t i c ro l e of i m m e d i a t e a n t e c e d e n t s M a n y studies sugges t that syntactic salience plays a role in the interpretation of anaphor ic express ions : T h e antecedent should be high on the grammat ica l role hierarchy (Givon , 1983a; Kehler, 2002 ; Gundel et al., 1993, p . 280: ex. (9) and (10)) . T h e assumpt ion is that grammatical subjects tend to code higher topic continuity, and will therefore be more natural, or accessible, antecedent candidates for cont inued anaphoric reference. The Center ing framework in some of its formulat ions (e.g. Grosz et al. 1995) makes the explicit assumpt ion that syntactic role is an important factor in the ordering of die so-called forward-looking center list, which is a theoretical construct corresponding to a list of potential topics . T h e highest-ranking entity in the forward-looking center list is the most l ikely enti ty to be talked about in the ensuing ut terance, and subjects generally rank h igher than other syntactic roles.

T h e impor tance of syntactic roles is partially confirmed by die grammat ica l roles of the immedia te antecedents in all three anaphor types . T h e majority of an tecedents are subjects or direct objects, whi le few are in obl ique syntactic posi t ions. But there is a clear difference be tween indefinite one-anaphors and definite p ronouns . T h e accented definite p ronouns more frequentiy have subjects as antecedents , and less frequently have objects (F isher ' s exact test p < 0.01)^. T h e same holds for unaccented definite p ronouns (Fisher ' s exact test p < 0.01) . T h e frequency distr ibutions of accented and unaccented definite p ronouns do not differ significantiy (Fisher ' s exact test p = 0.66). Indefinite one-anaphors , on the other hand, have objects as their most frequent antecedent.

A n a p h o r type Grammat ica l role of antecedent A n a p h o r type Subject Object Obl ique

Indefinite one-anaphors 1 2 3 6 Accen ted definite p ronouns 11 12 7 Unaccen ted definite pronouns 15 10 5

Table 4: Grammatical roles of the immediate antecedents of the three anaphor types.

W h y is this? No te that logical subjects of presentat ional der ( ' t he re ' ) const ruct ions were coded as objects. In the midst of the general confusion in the next example , a series of presentat ive der construct ions can be observed (under l ined) . Bo th part icipants come to realize that there are two banana pa lms on the map , and the instruction follower finally tries to s u m up by introducing each of them in their o w n presentative consduct ion in turn.

Fisher's exact test retums the exact probability of observing a certain frequency distribution (or one that is more extreme in the same direction) given a certain set of row and column totals in a 2x2 contingency table. It is precise even for skewed data, which is not Urue of the x2 test.

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 11: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

188 P H I L I P D I D E R I C H S E N

(8) A | : ligger bananpalmen hos dig efterfeltstationen ' is tlie banana pa lm in your map after tlie field s ta t ion? '

B | : ja ' y e s ' A2: okay na men min liggerf0r

' okay , well but mine is before ' B2: jamen jeg fattede at der var to

' bu t I ga thered there were two ' B3: det er der sa ogsd kan jeg se

'wel l so there are, I can see ' A3: ja det er der ogsd

' yes , so there are ' B4: sd der er en bananpalme altsd syd for feltstationen

og sa erderen sydvest for ' so , there is a banana pa lm south of the field s tadon, and then there is o n e to the southwest '

Presentat ive construcUons like these account for a substantial part of the difference. Seven of the indefinite one-anaphors follow presen tadve const ruct ions , and are consequent ly coded as having objects as their immedia te antecedents . N o n e of the definite accented pronouns do, and only one of die definite unaccen ted pronouns does . This suggests that the antecedents of indefinite one-anaphors tend to be more newly introduced than those of the two types of definite pronoun. Presentative der construct ions by definition introduce a new referent, normally using an indefinite Form (al though except ions , like " there were the usual people at the party", have been noted in the literature).

It is not obv ious why antecedents of indefinite one-anaphors would be newer than antecedents of definite pronouns. Indefinite one-anaphors themselves usually in t roduce a new referent—after all, they are indefinite. The explanat ion may be s o m e w h a t task specific. The most likely scenario is that un ique landmarks are routinely introduced using definite forms, which will work fine if the landmark in quest ion is indeed unique in both maps . But if it turns out, like in example (8) above , that the landmark is in fact not unique in one or both of the maps , communica t ion will break down and only cont inue once it has been clarified h o w m a n y instances of the landmark actually exist, and where they are located. T h e function of indefinite one-anapora is to introduce a different (i.e. contrasting) referent of the same type as one jus t ment ioned. This is the ideal context for one or more presentadve der cons tmct ions .

This partial explanat ion has more to do with identifiabihty than accessibi l i ty. T h e modva t ion for using indefinite forms in cases like this is precisely that the referent turns out not to be uniquely idenfifiable—not that the concept is inaccess ible . Therefore, it may be unwise to take the result above as ev idence that concepts referred to using indefinite one-anaphors are less accessible than those referred to using definite p ronouns .

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 12: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 189

3.5 Def in i teness m a r k i n g and lexicality of immediate antecedents Are there any differences in definiteness and lexicality in the antecedents of the diree anaphor types? T h e assumpt ions here are that definite forms will r ender an entity more accessible than indefinite forms, and that pronominal forms wil l render an ent i ty more accessible than lexical forms.

A l though an antecedent with an overt nominal head can usually be found somewhere in the referential chain, this is not the only type of noun phrase that can function as the immedia te antecedent. As the fol lowing examples il lustrate, the an tecedent of an indefinite o/ze-anaphor can be in the form of full noun phrases or p ronouns , definite as well as indefini te—singular as wel l as plural .

(9) A : ndr du er kommet godt syd om- eller godtforbi giraffeme sa svinger du til v- en runding til venstre ned rundt om en lille s0 ' w h e n you have passed well south of- or well past the giraffes then you turn 1- in an arc left down around a little lake '

B : sadan en har jeg ikke ' s uch a one I d o n ' t have '

(10) A: skal jeg ovenover eller nedenunder det forladte pakhus ' s hou ld I go above or be low the abandoned warehouse '

B : se sadan et har jeg ikke men jeg tror du skal oven over det ' s ee , such a one I don ' t have but I think you should go above it '

(11) A: sadan en har jeg ikke ' s uch a one I d o n ' t have '

B: har du ikke sadan en? ' d o n ' t you have such a o n e ? '

(12) A l : den har jeg helt oppe ' tha t one I have all the way up '

A2: i det nordvestlige hj0me afmin tegning ' in the nor th-west c o m e r of my drawing '

B : ndderharjegogsaen ' r ight , there I also have one '

(13) A l : de- det kan vcere der er flere ' i t- it may be that there are more '

B : a-dh ' uh -oh ' A2: men der ligger ethvad skal vi sige vest for den., telefonboks

' bu t there is one , what shall we say, to the west of that., phone boo th '

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 13: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

190 P H I L I P D I D E R I C H S E N

These examples show that the type description needed to interpret an indefinite one-anaphor can be mainta ined without necessarily using a referring express ion that includes an explicit nominal head. Fur thermore , the use of s ingular antecedents shows that it is not necessary for entities of the relevant type to be evoked in the plural in order to use indefinite one-anaphors . Referents of indefinite one-anaphors must not necessarily be non-unique, as long as they are type idenfifiable. By contrast , referents of s ingular definite pronouns mus t necessari ly be uniquely idenfifiable.

The defini teness and lexicality of the antecedents of the three anaphor types are shown in the tables be low.

First, cons ide r the definiteness data presented in table 5. The differences among the frequencies in the " tota l" co lumn are marginal ly significant (F i sher ' s exact test p = 0 .054, two- ta i led) ' . The difference between EN and den is significant (F i sher ' s exact test p = 0.04, two-tai led) , whereas the accented definite p ronouns do not differ statistically signif icandy from either the indefinite one-anaphora or the unaccented definite pronouns {EN vs DEN: Fisher ' s exact test p = 0 . 6 1 , two-tai led; DEN vs den: F i sher ' s exact test p = 0 .18, two-tai led) . However , the values of EN and DEN are closer than those of DEN and den, and thus there is a tendency for the two accented forms to pa t t em together.

Antecedent Singular Plural Total EN Indefinite 14 3 17

Definite 11 2 13 DEN Indefinite 14 0 14

Definite 16 0 16 den Indefinite 8 0

oo

Definite 22 0 22 Table 5: Definiteness of immediate/nearest antecedents of the three anaphor types.

Apparent ly , speakers to a certain extent avoid using accented forms to refer back to referents of definite expressions. This may be because accenting on pronouns tends to switch reference away from the main topic—that is, in many cases one that has been newly evoked with an indefinite form. In any case, the data once again group the three anaphor types according to accenting, and there is no ev idence of a difference in definiteness between antecedents of indefinite one-anaphora and definite pronouns .

N o w cons ider the lexicality results in table 6. F i sher ' s exact test indicates that there are no significant differences among the three anaphor types with respect to the lexicality of their antecedents (p = 0.17, two-tai led).

' This is the probability of observing a frequency distribution that is as extreme or more extreme as the one observed.

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 14: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 191

Antecedent Singular Plural Total EN Lexical 10 2 12

Pronomina l 15 3 18 DEN Lexical 18 0 18

Pronomina l 12 0 12 den Lexical 11 0 11

Pronomina l 19 0 19 Table 6: Lexicality of immediate/nearest antecedents of the three anaphor types.

However , there is a qui te clear tendency for the indefinite o/ie-anaphora and the unaccented definite p ronouns to have more pronominal than full noun phrase antecedents , whereas the accented definite pronouns have more full noun phrase antecedents than pronominal antecedents . I investigated the data more closely, and found that a p rominent difference be tween EN and DEN is that EN has more indefinite o/ ie-anaphors as antcedents , and DEN has more indefinite full noun phrases . T h u s , the difference be tween the two accented forms mainly s tems from differences in lexicali ty within the indefinite antecedents . It is a possibil i ty that the type descr ipt ion which is sufficient for the use of one-anaphor ic express ions tends to be more prominent when the speaker chooses a one-anaphor dian w h e n he chooses an accented definite pronoun, where the focus may to a higher degree be on the individual newly introduced referent. The accessibili ty of referents of these two types of referring expression is distributed slightly differently on type and token, so to speak.

A further reason for the many one-anaphoric antecedents of indefinite one-anaphora may b e structural pr iming, or ' a l ignment ' (Pickering & Garrod , 2 0 0 4 ; Hadel ich , Branigan, Pickering & Crocker , 2004 ; Kreysa, Arai , H a y w o o d & Picker ing, 2006) . A growing body of psycholinguist ic evidence shows that s tructures on all levels of language tend to become ' a l igned ' during dia logue. For instance, if one interlocutor uses a certain syntactic construct ion like the pass ive , the other inter locutor will be more likely to d o so as well . The same m a y well hold for indefinite one-anaphora .

S ince the difference be tween accented definite pronouns and indefinite one-anaphors is not statistically reliable, this finding should not be regarded as conclus ive ev idence against die accessibili ty hypothesis .

3.6 Access ibi l i ty: S u m m a r y of results T h e two var iants of the distance measure used yielded qualitatively similar resul ts : d ie pat terning of accented, indefinite anaphors wi th accented, definite p ronouns , which had a shorter dis tance to the immedia te antedent and the nearest lexical an tecedent than unaccented, definite p ronouns—cont ra ry to what would have been expec ted from Givon ' s and Arie l ' s accessibili ty scales. The repeated

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 15: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

192 P H I L I P D I D E R I C H S E N

mention measure also pat terned accented, indefinite one-anaphors with accented, definite p ronouns , which both had fewer prior references than the unaccented , definite p ronouns . With respect to the syntactic role of the antecedent , the two types of definite pronoun patterned together with a large number of subjects compared to the indefinite one-anaphors . However , the difference seemed to be due mainly to idendfiabil i ty differences, not accessibility differences. As for the definiteness of the antecedent , the results again seemed to pattern the two accented forms together, in that they both had substantial ly fewer definite antecedents than the unaccented definite pronouns did. The lexicality of the antecedent y ie lded no significant difference among the three anaphor types . However , there was a surprising tendency for the accented definite p ronouns to have more full noun phrase antecedents than both indefinite one-anaphors and unaccented definite p ronouns . The difference be tween the indefinite one -anaphors and accented definite pronouns turned out to exist mainly in the indefinite antecedents , and it was speculated that the tendency may have to do with a difference in type accessibility, and possibly also with al ignment.

Thus , in sum, none of the differences found convincingly indicated any accessibil i ty difference between indefinite and definite anapho r s—when accenting is accounted for, that is. The majority of differences were due to accent ing, and thus the results do not contradict the notion that accented forms differ from unaccented forms with respect to accessibili ty. One result seemed to mainly indicate a difference due to identifiability. Identifiabilty is exactly wha t we will turn to in the following.

4. Identif iabil ity of definite pronouns and indefinite anaphors W e now turn to the identifiability hypothesis . If the choice between indefinite and definite anaphor ic forms is determined by identifiability, then one should be able to observe differences in this feature among the referents of the three types of p ronouns .

In the fol lowing, a number of i l lustradve examples will be used to clarify what kinds of express ions can be expected to be used to refer to entities of different sharedness status in the visual c o m m o n ground. In order to do this, a quite e laborate notat ion scheme for visual sharedness must be applied. Th i s notation s cheme will be explained before we proceed to the examples .

T h e m a p task maps were designed to include a number of landmarks that vary in sharedness in the following w a y s ' :

• Shared landmarks: Certain landmarks (the majority) were represented on both maps .

• Absence/presence of landmarks: Certain l andmarks were found on one m a p but not the other.

See http://www.hcrc.ed.ac.uk/dialogue/nnaptask.html

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 16: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 193

• Name change of landmarks: Certain landmarks were identical in form and locat ion but had different labels on the two maps .

• 2:1 landmarks: Certain landmarks appeared twice on the instruction g ive r ' s m a p (one in a posit ion close to the route and one more remote) , whereas the instruction follower had only the distant one .

F igure 2 be low shows an example of a set of maps.

1̂ hiiJ^^

\

I 1

J«rnban«ov<r;l«zrmg

gronrtbrud

yaft»r ntngardi

STAUT

privcT* rtiarktr

ofbrandT skov

by99«l«gepladi

(iSrdsplads in«d htm

Figure 2: One of die maps used (map 1). Instrucdon giver version on the left, instruction follower version on the right.

In the fol lowing, the term ' speaker ' WILL be used to designate the person producing THE u t terance containing a particular expression. Converse ly , ' h s t ener ' des ignates the interlocutor, i.e. the person w h o is not the speaker at THAT momen t .

Ei ther the INSTMCTION GIVER or the INSTRUCTION fol lower can be the speaker W H E N shared landmarks are described, whereas only the insfruction giver has dupl icate l andmarks , which means that IF the speaker has duplicate l andmarks , THE speaker is the giver, and if the speaker is the one WITH THE solitary landmark , the l istener is the giver. In the remaining categories , either instruction giver or INSDTICTION fol lower can be the speaker. This yields the fol lowing logical possibi l i t ies for the visual sharedness status of referents:

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 17: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

194 P H I L I P D I D E R I C H S E N

1. Speaker : unique; Listener: unique (1:1). 2. (a) Speaker (giver): non-unique; Listener (follower): unique (2:1).

(b) Speaker (follower): un ique; Listener (giver): non-unique (1:2). 3. (a) Speaker : absent; Listener: unique (0:1).

(b) Speaker : unique; Listener: absent (1:0). 4 . Speaker : (name: A) unique; Listener: (name: B) unique (1(A): 1(B)).

Let us examine what k ind of referential choices would be expected in these different s i tuadons or sharedness states, given that the referent is not be ing ment ioned for the first fime in the dia logue.

O n the dissociated view of g ivenness , referents of indefinite one-anaphors can be assumed to be non-uniquely identifiable (i.e. merely ' type identifiable ' in Gundel et al.s (1993) terms) and at the same d m e highly accessible. In die m a p task, any visual ly non-shared entity will be non-uniquely idenfifiable for one of the inter locutors , at least until its existence has been verbally established.

In the d ia logues analyzed here, a referent cannot be talked about for very long before visual sharedness ceases to play a role for the choice of referring express ion used. Differences with respect to visual sharedness will soon be negotiated by the interlocutors, thus individuating the referent in quest ion and establishing unique identifiability in the universe of discourse. This quickly makes definite forms a felicitous option regardless of visual sharedness .

Thus , wha t we are looking for in this section are transient ' n iches ' of asymmetr ica l g ivenness caused by discrepancies in sharedness , that may give rise to the use of indefinite one-anaphors . What compl ica tes matters is that prior ment ions affect the identifiability status of anaphoric forms so that it can never be de termined by visual sharedness alone, as opposed to first-mention forms.

4.1 S u b s e q u e n t r e f e r e n c e in t he 0:1 a n d 1:0 s i t u a t i o n s In the fol lowing example of subsequent 0:1 reference by the instmction giver, the ins tmct ion fol lower first introduces the landmark "abandoned w a r e h o u s e " using a definite noun phrase. T h e landmark is absent from the instmction g ive r ' s map , so he responds by referring to a n e w — o r rather, hypothe t ica l—landmark of the same type, namely the one which should be, but in fact is not, on his m a p . In do ing so, he appropriately uses an indefinite one-anaphor .

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 18: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 195

skal jeg ovenover eller nedenunder det forladte pakhus? ' shou ld I go above or be low die abandoned warehouse? ' se sadan et/[detl har jeg ikke men jeg tror du skal ovenover det ' s ee , such a one/[that] I d o n ' t have but I think you should go above it ' na ' o h '

Figure 6: Subsequent reference to a 0:1 landmark by the instrucdon giver (the landmark is absent from the instruction giver's map, and the instruction giver is the speaker of the utterance containing the subsequent mendon—hence 0:1) using an indefinite one-anaphor.

But even though the instruction giver cannot see the landmark, he might as well have used an accented definite p ronoun (in square brackets) , s ince the referent is uniquely identifiable after being introduced by the instruction follower.

After the exchange shown, one would normally expect the instmct ion fol lower to specify where the landmark should be . She might do so by referring to another hypodie t ica l entity: one which she expects the insdnc t ion giver to d raw on his m a p . If so, she may use an indefinite one-anaphor or a definite p ronoun . If she chooses to tell the instruction giver where her own instance is located, she m a y l ikewise use an indefinite one-anaphor or a definite pronoun.

T h u s , in subsequent 0:1 or 1:0 reference, a given referent may elicit ei ther indefinite one-anaphors or definite p ronouns . All this applies whether the l andmark exists on the instruction g iver ' s or the instruction fol lower 's m a p . Therefore , no specific predict ions about die frequency of indefinite vs. definite forms can be m a d e for subsequent 0:1 and 1:0 si tuations.

4.2 S u b s e q u e n t reference in the 1:2 s i tuation In the case of 1:2 (speaker-unique, l is tener-non-unique) mentions, some landmark appears twice (incidentally, a lways on the instruction g iver ' s map) . In this s i tuat ion, whe the r the instruction giver uses an indefinite form depends to a large extent on whetiier die two identical referents are close enough in space to likely b e confused by the listener. Recent psychol inguis t ic research suggests that

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 19: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

196 P H I L I P DIDEFUCHSEN

referential d o m a i n s with respect to a visual scene can be very narrow, which means that the dis tance be tween identical landmarks can be quite small wi thout affecting the use of definite expressions (Chambers , Tanenhaus , Eberhard, Fil ip & Car lson, 2002) . S ince the instruction g iver ' s two instances of a 2:1 landmark are a lways separated by a certain distance, he/she in fact often treats the first 2:1 landmark encounte red as uniquely identifiable.

In subsequen t 1:2 si tuations, it is a lways the instmction follower w h o refers to an a l ready- in t roduced landmark. In figure 7 below, the landmark is in t roduced by the instruction giver in an indefinite full noun phrase , and is subsequent ly men t ioned by the ins tmct ion follower.

G i og udfra malet skulle der vcere en parkeret lastbil ' and in line with the target there should be a parked t ruck '

F l ja men den er helt alter 0verst ' yes but that is all the way at the top '

G2 ja ' y e s ' F2 er den det? ' is i t? ' G3 ja ' y e s '

Figure 7: Subsequent reference to a 1:2 landmark by the instmction follower (the landmark is unique in the instmction follower's map, and the instmction follower is the speaker of the utterance containing the subsequent mention—hence 1:2) using an accented definite pronoun.

T h e landmark being unique in the instmct ion fo l lower ' s map, and being s ingled out by the instruction giver, is then treated by the ins tmct ion fol lower as uniquely identif iable. She refers to it using an accented definite pronoun, bu t in this context , an ut terance containing an indefinite one-anaphor would have been acceptable as wel l . In this case, the reference went well , since the inter locutors were in fact speaking about the same landmark.

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 20: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 197

But in many cases , the landmark introduced by the instruction giver will be the first ins tance on the route, as in the fol lowing figure. In such cases , the l andmark is l ikely to be treated as absent from the instruction fo l lower ' s m a p , s ince the s ingular instance on the instruction fo l lower ' s m a p will be p laced far away from the current location on the route .

i fa

G i og der fortscetter du.. 0hh et godt stik- stykke sydpd ' and there you continue. , err some distance to the soutii '

G2 indtil der kommer en parkeret lastbil 'until you reach a parked truck '

Fl sadan en/fden] har jeg ikke ' such a one/[that one] I don ' t have ' [...]

G4 nej, okay 'oh, ok '

Figure 8: Subsequent reference to a 1:2 landmark using an indefinite one-anaphor.

In figure 8, the parked truck is introduced by the instruction giver using an indefinite noun phrase . In this case , since the landmark ment ioned is the one miss ing from the instruction fo l lower ' s map , the instruction fol lower appropr ia te ly reacts by treating the parked truck as absent . She refers to the kind of l andmark ment ioned by the instruction giver by using the indefinite one-anaphor sadan en, lit. ' such a o n e ' . However , she might acceptably have used den 'd iat o n e ' instead, as indicated with square brackets .

Fur ther consider ing the 1:2 situation, if the referent is in t roduced by the instruct ion follower, for w h o m the landmark looks un ique , the instruction giver is l ikely to fail to not ice that he/she actually has another instance of the men t ioned landmark somewhere far away. This other instance will p robably be located outs ide the current referential domain , and therefore the instruction giver will be unl ikely to notice it. The turn comes back to the instruction follower, w h o

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 21: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

198 P H I L I P D I D E R I C H S E N

is now in a posi t ion to refer to a given, seemingly unique landmark. Under these c i rcumstances , the occurrence of indefinite one-anaphors seems highly unlikely.

Let us s u m up at this point. Because of the narrow scope of referendal domains (i.e. the interlocutors literally focus on a narrow area within die map , with respect to which referential expressions are produced and interpreted, c f the discussion above) , subsequent 1:2 situations will often resemble either 1:1 or 0:1 s i tuat ions. T h e l ikel ihood of encounter ing indefinite one-anaphors would be expected to be highest when the subsequent 1:2 situation resembles the subsequent 0:1 situation, that is, when the solitary landmark of the ins tmct ion fol lower is outs ide the current referential domain (i.e. located in a different area of the m a p than the area currently talked about) , and the landmark therefore appears to be absent from the instmct ion fol lower 's map .

4 .3 S u b s e q u e n t r e f e r e n c e in t h e 2 :1 s i t u a t i o n Let us n o w turn to the 2:1 situation. A subsequent ment ion 2:1 si tuation resembles the subsequent 1:2 situation. The main difference is that the turn has c o m e back to the instruction giver. A subsequent 2:1 situation is one where the instruction giver ment ions an already-introduced 2:1 landmark. If the landmark is in t roduced by the instruction follower (to w h o m it is unique) , using a definite form, the si tuation is s imilar to the one discussed above: Because of the narrow scope of the referential domain , the instruction giver is unlikely to notice that there are mul t ip le instances of the landmark in quest ion, and it is thus likely to be treated as a 1:1 landmark, yielding few indefinite one-anaphors .

In the other conceivable case of a 2:1 situation, where the instruction giver introduces a non-unique landmark, the instmct ion follower says something , and the turn then comes back to the instmction giver, the course of the exchange will depend to a large degree on which instance of the 2:1 landmark the ins tmct ion giver originally introduced. If it is the one that is present on the instruction fo l lower ' s map , the instmct ion follower will have no reason to quest ion anyth ing , and the instmct ion giver will be likely to use definite forms for subsequen t ment ions , since the landmark will now have been establ ished as uniquely identifiable within the referential domain . Again , this situation will resemble the 1:1 situation.

Bu t if the landmark originally introduced by the ins tmct ion giver is the instance that the ins tmct ion follower does not have on his/her m a p , the ins tmct ion fol lower is likely to protest. The instmction follower can be expected to d o one of two things. Ei ther state that he/she does not have the instance in ques t ion on his/her map , or ask whether the ins tmct ion giver means the other instance located farther away. Either way, either a definite form or an indefinite form m a y be used.

In the example in figure 9 below, the ins tmct ion fol lower is about to say that he canno t see the large rock ment ioned by the ins tmct ion giver, but then impl ic idy asks whether she in fact means the one located "all the way down

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 22: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 199

there" . W h e n the instruction fol lower ment ions the landmark that does exist on his map , the instruction giver is a lmost forced to refer to the other instance, i.e. a different l andmark of the same type. This is the perfect context for an indefmite one-anaphor , and she indeed chooses this type of referring expression.

G i du skal nemlig ned omkring syd for en stor klippe ' you see, you should go south around a large rock '

F) den kan jeg ikke-na helt der nede ' that I can ' t - oh all the way down there '

F2 jeg har en stor klippe sadan altsa ' I have a large rock like about '

G2 du har den meget langt nede ' you have it very far d o w n '

F3 ja ' y e s ' G3 ja ' y e s ' G4 nej jeg har en til

' no , I have another one ' G5 og den har jeg lige vest for indianerlejren

' and that one I have jus t west of the Indian c a m p '

Figure 9: Subsequent 2:1 reference using an indefinite one-anaphor.

4.4 S u b s e q u e n t reference in the 1(A):1(B) s i tuation

Turn ing n o w to the 1(A); 1(B) situation, where the instruction giver or the ins tmct ion fol lower has in t roduced a 1(A):1(B) landmark, whe the r the inter locutor subsequent ly uses a definite or an indefinite form about h is /her version of the landmark will depend in part on whether he/she interprets it as the same l andmark , in which case the speaker-ment ioned landmark will be un ique in the universe of discourse (typically prompt ing a defmite form), or as a different l andmark , in which case the speaker ' s landmark will be absent or located in a

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 23: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

200 P H I L I P D I D E R I C H S E N

different posi t ion on his/her map , and the l is tener 's version of the landmark must be treated as a new entity (typically prompt ing an indefinite form).

E x a m p l e 10 illustrates both possibil i t ies. T h e instrucdon fol lower introduces a 1(A): 1(B) landmark using an indefinite noun phrase. The instruction giver refers to the instruction fo l lower ' s landmark, which is by now uniquely idendfiable , using an accented definite pronoun, a l though he might as well have used an indefinite one-anaphor in this context. He then refers to his o w n vers ion of the landmark , which has a different label, using an indefinite noun phrase , which s ignals that he considers the landmark ' B r a n d - N e w ' , i.e. a different i tem.

F i ja sd- be- [...] gar jeg mod noget der hedder en klippe have ' yes then [...] 1 go toward something called a rock garden '

Gi ja.. den/[sadan en / har jeg heller ikke 'yes. , that one/[such a one] I don ' t have ei ther '

G2 har du noset der hedder et stenscerde i ncerheden ' do you have something called a stone wall nearby? '

F2 b- nej det har jeg ikke ' no , I d o n ' t '

Figure 10: Subsequent reference to a 1(A); 1(B) landmark using an accented defmite pronoun.The instruction giver is under the impression that the landmark is absent from his map, but this is not correct. It just has a different name.

In cases whe re two versions of the same landmark are seen as different l andmarks al together, one would generally expect fewer anaphoric express ions and more full noun phrases .

4.5 S u b s e q u e n t reference in the 1:1 situation Finally, in the 1:1 situation, the only s i tuadon involving a truly 'doubly un ique ' landmark, one would normally expect mainly definite pronouns on subsequent

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 24: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 201

mention. It is reasonable to assume that by default the speaker ' s menta l representat ion of the speech si tuation includes the belief that the l istener has perceptual access to the s ame objects as the speaker (Keysar & Barr , 2005) . Thus , if a l andmark is unique in the speaker ' s visual field, he/she will probably assume, if there is no ev idence to the contrary, that die object is un ique to the listener as wel l . T h e typical referring expression in diis situation is a full definite noun phrase . S ince the landmark is also unique on the l is tener 's map, chances are that the reference will go unchal lenged, and will be cont inued using definite p ronouns . However , indefinite one-anaphors are not precluded from this context , s ince both the ins tmct ion giver and the instmction follower may choose to treat the landmark on his or her o w n m a p as a new entity, for instance if the location of the l andmark is not clear from the inter locutor 's description.

4.6 P r e d i c t i o n s W e can n o w finally ask which predict ions the above analyses give rise to. Is it possible to predict which anaphoric forms will be used on the basis of the visual sharedness status of landmarks? Under some ci rcumstances , yes .

T o s u m up the analyses in the previous sections, most of die contexts for subsequent ment ion allow both definite pronouns and indefinite one-anaphors to be used. T h e only case in which there is good reason to expect a h igher frequency of indefinite one-anaphors is the version of the 2:1 simation when a landmark is introduced by the ins tmct ion giver, and the instmction fol lower then asks whether he/she means the 'o ther ' instance, i.e. the solitary instance present somewhere else on the ins tmct ion fol lower 's map . In this context, it will be difficult for the instruction giver to use a definite form without referring to the non-intended instance.

The identifiability hypothesis , presented as (6) earlier, is repeated as (14):

(14) Iden t i f i ab i l i ty h y p o t h e s i s : There are significant differences be tween relevant indicators of the identifiability of referents of definite p ronouns and indefinite one-anaphors .

This can n o w be refined to the fol lowing testable hypothesis :

(15) A s y m m e t r i c u n i q u e n e s s h y p o t h e s i s : There are significantly more indefinite one-anaphora than definite pronouns among speaker-non-unique/ l i s tener-unique (2:1) landmarks .

This difference would support the idea that identifiability plays an independent role in the choice of referential express ions , and const imtes a sort of m i n i m u m requi rement for concluding that the idenfitiability hypothesis is correct . Shou ld there be no such difference, this would be quite strong evidence against the identifiability hypodies is .

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 25: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

202 P H I L I P D I D E R I C H S E N

4 .7 I den t i f i ab i l i t y : R e s u l t s a n d d i scuss ion T h e results per ta ining to identifiability and definiteness are shown in table 7.

Anaphor type Visual sharedness Indef. one- Accented def. Unaccented def. Sum status anaphors (en) pronouns(den) pronouns (den) 1:1 8 10 10 28 2:1 9 5 2 16 1:2 4 4 3 11 0:1 3 4 2 9 1;0 2 6 9 17 1(A):1(B) 0 1 4 5 Sum 26 30 30 86

Table 7: The visual sharedness of referents of different types of anaphoric expressions. In the indefinite one-anaphors column, two instances have been left out because the referents were not on the maps and thus did not have a visual sharedness status. Two further instances of indefinite one-anaphors are disregarded since they seem to be production error-like repetitions.

The predicted difference shows up in the 2:1 row: There are indeed more speaker -non-unique landmarks referred to by indefinite one-anaphors than by accented definite p ronouns . However , the amount of data is too small for any of the differences to be stadstically significant.

A curiosi ty in the data is the fact that so many referents of indefinite one-anaphors are doubly unique (1:1). This sharedness status would normal ly be expected mainly of referents of definite forms. On closer scrutiny, it turns out that many of these referents were assumed to be non-unique or absent at the t ime of ut terance. T h r e e of the 8 indefinite one-anaphors in this category are used to refer to a banana pa lm that is clearly assumed by the interlocutors to be non-unique, but is in fact unique on both maps . Another indefinite one-anaphor is used about an abandoned monastery which the interlocutors suspect may be one of several ins tances . Another indefinite one-anaphor is used in a situation where the speaker has temporar i ly over looked the landmark in quest ion because there is another l andmark before it on his map. These cases account for more than half of the indefinite one-anaphors in the 1:1 category, and can be argued to be long in a d isdnct ca tegory of asymmetr ic uniqueness (the inter locutors ' suspicion be ing that each has a un ique instance of the category in a different locat ion) . Th i s provides a kind of indirect support for the asymmetr ic uniqueness hypothes is . Fur thermore , two of the referents in this category are used as proper indefmite p ronouns in the sense of for instance Haspelmath (1997) , i.e. with the in tended mean ing ' s o m e o n e ' . In this way, 7 of the 8 referents in this category can be accounted for by alternative classifications.

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 26: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 2 0 3

T h e 1:0 category contains many landmarks referred to using unaccented definite p ronouns . The re is no obvious explanation for this, but note the following example , whe re three definite pronouns are used in close success ion. One might h a v e been sufficient. They all showed up in the sample , and obviously have the s ame sharedness s tams.

(16) G | : altsd den er to takker oven for vceltet stengcerde ' so it is two steps above o v e r m m e d stone wal l '

G2: og den er tre tra- tre takker h0- oven for granitbruddet og sa er den sadan cirka midt imellem ' and it is three st- steps a- over the granite quarry and then it is about midways between the t w o '

T h e idendfiabil i ty results clearly tend to support the asymmetr ic un iqueness hypothes is , especial ly when the data are scmtinized more closely. T h u s , whi le the data are too sparse to consti tute conclusive evidence for or against the identifiability hypodies i s , die hypothesis is not refuted by the data analysed.

5. Discuss ion T h e empir ical invesUgation of die two hypotheses yielded encouraging results . The idendfiabil i ty hypothes is was supported by clear, albeit stafisdcally non­significant tendencies in the data. The accessibility hypothesis was suppor ted by most measu res (dis tance, repeated mention, and definiteness of antecedents) , where the observed differences seemed to have to do with accent ing rather than accessibil i ty. Differences in the syntactic characteristics of antecedents seemed to be a matter of identifiability, whereas statistically non-significant differences in antecedent lexicali ty lacked a straightforward explanat ion.

Thus , indefinite one-anaphors neither pa t tem consistently with, nor differ consistently from, accented definite pronouns with respect to accessibil i ty and identifiability, but rather share the accessibility of p ronouns and, p resumably , the identifiability of indefinite noun phrases, in accordance with the dissociated v iew of g ivenness .

Based on these findings it seems sensible to separate accessibil i ty and identifiability. The fol lowing two-dimensional mode l of g ivenness emerges , which can be seen as (a fragment of) a functional m a p of referential mean ing showing the aspects under discussion in this article, a long the l ines of the functional m a p s shown in studies such as (Croft, 2001) :

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 27: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

204 P H I L I P D I D E R I C H S E N

fiN Df iN

en banan- banan-pa lme p a l m e n

Type Identif iabi l i ty Unique

Figure 11: Two-dimensional model of cognitive statuses motivating referential form. Assumed accessibility in the listener motivates pronominalization and assumed idendfiability in the listener motivates definiteness. Gloss: 'a banana palm'/'the banana palm'.

In this model , the accessibili ty of an entity in the mind of a listener is free to vary regardless of whe the r this entity is a uniquely identifiable referent or rather a (non-s ingleton) set or a merely type identifiable category/ type. T h e model elegantly a c c o m m o d a t e s indefinite o«e-anaphors , which map onto highly accessible, but merely type identifiable entities.

T h e d imens ion of identifiability is presumably binary, since a referent may ei ther b e uniquely identifiable for the listener or not. The accessibil i ty d imens ion is probably more cont inuous . It has to do with the prominence of a referent or type in the mind of the listener, which presumably varies in a graded manner .

Referr ing express ions must necessarily m a p onto this accessibil i ty con t inuum in a categorial fashion. This does not mean that this mapp ing is straightforward (for example , corresponding express ions in different languages are likely to cover different por t ions of the accessibili ty cont inuum, see e.g. Gundel et al. (1993)) , but this is a different story. T h e claim here is s imply that indefinite one -anaphora and accented definite p ronouns cover about the same port ion of the con t inuum in spontaneous spoken Danish .

No te that the model does not account for accenting. I suspect that accent ing reflects yet another independent d imens ion or referential mean ing . Th i s d imens ion might be represented as a 'Z -ax i s ' in the plot. Accented and unaccented express ions obviously need to b e differentiated. As we have seen several t imes , there are clear differences be tween den and DEN (and see

High

Accessibi l i ty

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 28: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 205

Krist iansen, 1996). This quest ion goes beyond the scope of the present s tudy, but the fact that accent ing often plays a significant part in the pat terning of referring express ions is very important to keep in mind for future invest igations of anaphora in Danish .

T h e a m o u n t of tokens analyzed is unfortunately not sufficient to show statistically signif icant resul ts . As it happens , the indefinite anaphor enlet has several different mean ings or functions, and not all of them are even real ly anaphor ic . T h u s , in die 30- token sample in this s tudy, there are instances of en meaning ' any person , s o m e o n e ' , and there are modified versions {sddan en lit. ' such a o n e ' , en til ' one m o r e ' , and en der... ' one that /who . . . ' ) , as well as m o r e numeral- l ike uses {en mean ing ' (only) one as opposed to two or more ' ) ) . T h e s e are not ' p u r e ' forms of anaphor ic en, and are thus somewhat problemat ic to include in a direct compar i son with pure exemplars of stressed and unstressed definite p r o n o u n s . However , s ince all the above-ment ioned variants of enlet do have an an teceden t ' , and can thus be assumed to have a degree of accessibil i ty comparab le to that of a p ronoun , it was decided to include them.

Final ly, it remains to be demonst ra ted whether accessibili ty correlates with lexicali ty, i.e. whe ther the antecedent distance of anaphoric full noun phrases is h igher than that of pronouns , for instance. This would be s trongly expected based on any theory of accessibili ty. Also the repeated ment ion measure wou ld be expected to show a clear effect. Cmcia l ly , the model outl ined above predicts that such a correlat ion should not be influenced by identifiability and thus be equal ly s trong for indefinite and definite forms. These quest ions will not be further addressed here .

T h e da ta presented are admittedly too sparse to be taken as conclus ive ev idence—a c o m m o n prob lem in corpus studies. Tha t said, the predictions of the model seem to hold reasonably well . Thus , at least in this quite distinct type of spoken d ia logue , indefinite one-anaphors do seem to have approximately the same degree of accessibil i ty as odier (accented) pronouns , and are clearly different as far as identifiability is concerned. Whe the r the results can b e replicated in other spoken styles such as more informal conversat ion, or even in written narra t ive , are open quest ions . But intuitively it seems that the function of at least indefinite one-anaphora should remain the same across styles and genres : the express ion of a referent whose category is given, but which itself has not yet been individuated. T h e basic function of definite pronominal forms would l ikewise b e expec ted to be relatively robust: the expression of a referent that is accessible A N D uniquely identifiable in the universe of discourse. On the other hand, I will not speculate about h o w the effect of accent ing might modula te results in o ther genres . This quest ion deserves a study in its o w n right.

Even the 'someone' uses can be analyzed as having an antecedent.

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 29: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

206 P H I L I P D I D E R I C H S E N

These results are promis ing , and it seems worthwhi le to investigate the matter further, both using other corpus resources (other genres, e.g. oral narrative), and psychol inguis t ic exper imentat ion.

R e f e r e n c e s Anderson, Anne , Mi les Bader , Ellen Bard, Elizabeth Boyle, Gwyne th Doher ty ,

Gar rod , S imon , Stephen Isard, Jacquel ine Kowtko , Jan McAllister , K i m Miller , Cather ine Sodl lo , Henry T h o m p s o n & Regina Weinert . 1991. T h e H C R C m a p task corpus . Language and speech, 34, 3 5 1 - 3 6 6 .

Ariel , Mira. 1990. Accessing noun-phrase antecedents. London and New York: Rou t l edge .

Bakeman, Roge r & John M. Got tman. 1997. Observing interaction: an introduction to sequential analysis. Cambr idge: Cambr idge Universi ty Press .

Bhat, Darbhe N. S. 2004 . Pronouns. Oxford: Oxford Universi ty Press. Carletta, Jean 1996. Assess ing agreement on classification tasks: the kappa

statistic. Computational linguistics, 22(2), 2 4 9 - 2 5 4 . Chafe, W. (1994) . Discourse, consciousness, and time. Chicago: The Universi ty

of Ch icago Press. Chambers , Craig G., Michael K. Tanenhaus , Kathleen M. Eberhard, Hana Filip,

& Greg N . Carlson. 2002 . Circumscr ibing referential domains dur ing real- t ime language comprehens ion . Journal of Memory and Lcmguage, 47, 3 0 - 4 7 .

Cristea, Dan & Postolache, Oana-Diana . 2005 . How to deal with wicked anaphora? In Anton io Branco, Tony McEnery & Ruslan Mitkov (Eds.) , Anaphora processing: Linguistic, cognitive, and computational modelling. Vo lume 263 Amsterdam studies in the theory and history of linguistic science pp . 17 -46 . Amsterdam/Phi ladelphia : John Benjamins .

Croft, Wi l l i am A. 2 0 0 1 . Radical construction grammar: Syntactic theory in typological perspective. Oxford: Oxford Universi ty Press.

Dahl , Deborah 1985. The structure and function of one-anaphora in English. Blooming ton , Indiana: Indiana Universi ty Linguist ics Club.

Givon , T a l m y (Ed.) . 1983a. Topic continuity in discourse: A quantitative cross-language study. V o l u m e 3 of Typological studies in language. Amsterdam/Phi lade lph ia : John Benjamins.

Givon, T a l m y 1983b. Topic continuity in spoken english. In Ta lmy Givon (Ed.) , Topic continuity in discourse. Vo lume 3 of Typological studies in language, pp . 3 4 7 - 3 6 3 . Amsterdam/Phi lade lphia : John Benjamins.

G r 0 n n u m , Nina 2006 . Danpass - a danish phonetical ly annotated spontaneous speech corpus . In Proceedings of the 5th International Conference on Language Resources and Evaluation. Genova .

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 30: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

GIVENNESS REVISITED 207

Griining, Andre & Andrej A. Kibrik. 2005 . Model l ing referendal choice in d icourse : a cogni t ive calculative approach and a neural ne twork approach. In An ton io Branco , Tony McEnery & Ruslan Mi tkov (Eds.) , Anaphora processing: Linguistic, cognitive, and computational modelling. V o l u m e 2 6 3 of Amsterdam studies in the theory and history of linguistic science pp . 1 6 3 - 1 9 7 . Ams te rdam/ Philadelphia: John Benjamins .

Grosz, Barbara J., Aravind K. Joshi & Scott Weinste in (1995) . Center ing: a f ramework for model ing the local coherence of discourse. Computational linguistics, 21(2), 2 0 3 - 2 2 5 .

Gundel , Jeanet te K., Nancy Hedberg , & Ron Zacharski 1993. Cogni t ive status and the form of referring expressions in discourse. Language, 69, 274— 307 .

Hadel ich, Kerst in , Holly Branigan, Mart in Pickering, & Mat thew Crocker . 2004 . A l ignmen t in dia logue: effects of visual versus verbal feedback. In Proceedings of the 8th Workshop on the Semantics and Pragmatics of Dialogue (Catalog'04). Barcelona, Spain.

Haspe lmath , Mar t in 1997. Indefinite pronouns. Oxford: Oxford Universi ty Press . Kaiser , Elsi . 2 0 0 5 . Different forms have different referential propert ies:

Impl ica t ions for the nodon of ' sa l ience ' . In Antonio Branco , T o n y M c E n e r y & Ruslan Mitkov (Eds.) , Anaphora processing: Linguistic, cognitive, and computational modelling. Vo lume 263 of Amsterdam studies in the theory and history of linguistic science pp . 2 6 1 - 2 8 2 . Amsterdam/Phi lade lph ia : John Benjamins.

Kehler , A n d r e w (1997) . Current theories of centering for pronoun interpretation: a critical evaluat ion. Computational linguistics, 23(3) .

Kehler , A n d r e w 2002 . Coherence, reference, and the theory of grammar. Stanford: CSLI Publicat ions.

Keysar , Boaz & Dale J. Barr. 2005 . Coord inadon of acdon and belief in c o m m u n i c a d o n . In John C. Tmeswe l l & Michael K. Tanenhaus (Eds.) , Approaches to studying world-situated language use: Bridging the language-as-product and language-as-action traditions chapter 3 , pp . 7 1 -94 . Cambr idge , Massachuset ts : T h e M I T Press.

ICreysa, He lene , Manabu Aral , Sarah L. Haywood , & Mart in J. Pickering. 2006 . Does syntactic a l ignment affect eye movemen t s?

Kris t iansen, Erik (1996) . Topic continuity and prosody: A n exper imental s m d y in danish. In E. Engberg-Pedersen , M. For tesque, P. Harder , L. Heltoft & L. F . Jakobsen (Eds.) , Content, expression, and s tmcmre : studies in Danish functional grammar . Vo lume 29 of Studies in language c o m p a n i o n series (pp. 2 6 1 - 2 8 1 ) . Amsterdam/Phi lade lphia : John Benjamins Publishing Company .

Lappin , S h a l o m & Herbert J. Leass . 1994. A n algor i thm for pronomina l anaphora reso ludon. Computational linguistics, 20(4), 5 3 5 - 5 6 1 .

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14

Page 31: Givenness revisited               indefinite one-anaphora in unscripted Danish dialogue

208 P H I L I P D I D E R I C H S E N

M c C a w l e y , J ames D. 1988. The syntactic phenomena of English. The Ch icago Univers i ty Press.

Mitkov, Ruslan 1997. Factors in anaphora resolution: they are not the only thing that mat ter . A case study based on two different approaches . In Ruslan Mi tkov & Branimir Boguraev (Eds.) , Proceedings of the workshop "Operational factors in practical, robuust anaphora resolution for unrestricted texts" pp . 1 4 - 2 1 . Madr id : Univers idad Nacional de Educac ion a Distancia.

Nedergaard T h o m s e n , O. (1992) . Pronouns , word order, and prosody. Informat ion structuring in preverbal slots in spoken danish. Copenhagen work ing papers in l inguisdcs , 2, 119-232 .

Picker ing, Mar t in &. S imon Garrod. 2004 . Toward a mechanist ic psychology of d ia logue . Behavioral and brain sciences, 27, 169 -225 .

Prince, El len F. (1981) . Toward a taxonomy of g iven-new information. In P. Cole (Ed.) , Radical pragmatics. N e w York: Academic Press.

Rietveld, Ton i & Roeland van Hout . 1993. Statistical Techniques for the Study of Language and Language Behaviour. Berlin: Mouton .

Sag, Ivan A . 1979. The nonuni ty of anaphora . Linguistic Inquiry, 10(1), 1 5 2 -164.

Sag, Ivan A. & Jorge Hankamer . 1984. Toward a theory of anaphoric process ing. Linguistics and philosophy, 7, 3 2 5 - 3 4 5 .

Vendit t i , Jennifer J., Ma t thew Stone, Preetham Nanda, & Paul Tepper . 2 0 0 1 . T o w a r d an account of accented pronoun interpretation in discourse context : Evidence from eye-tracking. Technical report, Rutgers Center for Cogni t ive Science .

A c k n o w l e d g e m e n t s T h e author wishes to thank Ta lmy Givon, Michael For tesque, Peter Harder , and two a n o n y m o u s reviewers for helpful comments on an earlier draft of this article. Also thanks to professor Nina G r 0 n n u m for permiss ion to use the D a n P a s s corpus even before it was finished.

A b o u t the a u t h o r Phil ip Dider ichsen Lund Univers i ty Cogni t ive Science Lund Univers i ty Kungshuse t , Lundaga rd S-222 22 Lund S w e d e n e-mail: phi l ip dot d ider ichsen at lues dot lu dot se

Dow

nloa

ded

by [

Cor

nell

Uni

vers

ity L

ibra

ry]

at 0

3:21

19

Nov

embe

r 20

14


Top Related