.... . • ... . .•" ..... .... . .... 
WHAT BRINGS A HIGHER-ORDER ENTITY INTO FOCUS OF ATTENTION. 9 
S ENTENTIAL PRONOUNS IN ENGLISH AND NORWEGIAN 
Kaja Borthen 
Norwegian University of 
Science and Technology 
bort0013@tc.umn.edu 
Thorstein Fretheim 
Norwegian University of 
Science and Technology 
fretheim@hf.nmu.no 
Jeanette K. Gundel 
University of Minnesota 
gunde003 @maroon.tc.mnn.edu 
ABSTRACT 
This paper investigates the use of sentential pronouns 
in English and Norwegian. We argue that resolution 
of sentential pronouns is sensitive to the distinction 
between forms whose referents must be in focus and 
forms whose referents must only be activated, but not 
necessarily in focus. An investigation of the 
distribution and interpretation of sentential pronouns 
also reveals that the relative salience of a higher order 
discourse endty is influenced by syntactic structure as 
well as extralinguistic factors. 
1 INTRODUCTION 
We can refer to an entity with several different 
expressions, and a single expression can be used to 
refer to many different entities. Two questions that 
naturally arise given these facts are 1) How do we as 
audience manage to understand a speaker's intended 
interpretation of a referring expression? 2) What 
determines the speaker's choice of form when referring 
to an entity? It has turned out to be very fruitful to 
see these two questions as closely connected. A whole 
fieldof research on the distribution and interpretation 
of referring expressions clearly suggests that the 
speaker's choice of form helps the audience pick out 
the intendedreferent of the expression (cf. for example 
Garrod & Sandford 1982, Ariel 1988, Gundel, 
Hedberg and Zacharski 1993). 
In this paper we will look at the distribution 
and interpretation of pronouns that are used to refer to 
higher -order entities like events, facts, or 
propositions, as illustrated in (1). 
(1) John won the race. I know that became I 
saw i_t happen. But it's still hard to believe 
i_t. (Fraurud 19922) 
Our main question in this paper is this: what 
yields an appropriate use of a given pronominal form 
when it refers to a higher-order entity in English and 
Norwegian? We will propose that in both languages 
the form that is most frequendy used is one which 
requires its referent to be activated, but not necessarily 
in focus (in the sense of Gundel, Hedberg and 
Zacharsld 1993). This is so because processing a 
sentence generally brings into focus the major 
partidpants rather than the event or state of affairs 
itsdf. 
2 COGNITIVE STATUS 
Gundel, Hedberg, and Zacharski 1993 (henceforth 
GHZ) propose that a particular form conventionally 
signals that the referent of the expression has a certain 
cognitive (memory or attention) status for the 
addressee. In English, for example, an unstressed 
pronoun such as it signals that the addressee's 
attention is currendy focused on the intended referent. 
Since the addressee's attention will normally be 
focused on only one or two entities at a time, the 
number of possible referent candidates for an 
unstressed pronoun is very low, and referent 
resolution is therefore made possible even though the 
descriptive content of a pronoun is very limited. 
GHZ assume that there are six cognitive 
statuses relevant for the distribution of referring 
expressions: The connection between these statuses 
and their corresponding forms in English is illustrated 
in the Givenness Hierarchy in Figure 1 (because of 
space limitations, we list only the 3 highest statuses 
here.) 
The Givenness Hierarchy 
it that, this that N 
Figure 1 
Figure 1 shows what cognitive status is 
necessary for an expression with a given form to be 
used appropriately. This does not exclude the 
possibility that an entity that has a certain cognitive 
status can be referred to with a form that is to the 
right of this status in the hierarchy since the higher 
statuses (to the left in the table) imply all the lower 
ones. For instance, the demonstrative determiner 
'that' can be used when the referent is familiar, but 
not activated. And it can also be used when the 
88 
'i i 
referent is in focus. An unstressed personal pronoun, 
such as it, on the other hand, can only be used 
appropriately if the speaker's intended referent is in 
the addressee's focus of attention. 
IN FOCUS VS. ACTIVATED 
According to Figure 1, we would expect sentential 
it 1 to be used appropriately only when the entity 
referred to is in the addressee's focus of attention, 
while 'that' or 'this' can be used whenever the entity is 
activated, regardless of whether or not it is also in 
focus. Since higher statuses imply all lower statuses, 
it should be possible to use the demonstrative 
pronouns that and this to refer to entities that have 
the status In Focus; but analyses of naturally 
• occurring discourse show that these forms are used 
primarily when the referent is activated, but not in 
focus. In the statistics presented in GHZ's 1993 
paper, only one out of 33 occurrences of 
demonstrative pronouns in English referred to entities 
that had the status In Focus. GHZ explain this as the 
result of interaction of the Givenness Hierarchy with 
the first part of the Maxim of Quantity "Make your 
contribution as informative as required(for the current 
purpose of the exchange)" (Grice, 1975). That is, 
demonstrative pronouns convey a so-called scalar 
implieature: a weaker form conversationally 
implicates that a slxonger form does not obtain. 
Thus, just as "I have two children" implicates that the 
speaker does not have more than two children, using a 
demonstrative form implicates that the referent of the 
expression does not have a status higher than 
activated, i.e. it is not in focus. 
4 SENTENTIAL PRONOUNS AND THE 
IMPORTANCE OF SUBSEQUENT 
MENTION 
According to GHZ, the entities in focus at a given 
point in the discourse will be that partially-ordered 
subset of activated entities which are likely to be 
continued as topics of subsequent utterances. Thus, 
endties in focus at the beginning of an utterance 
generally include at least the topic of the preceding 
utterance, as well as any still-relevant higher-order 
topics. To the extent that syntactic structure and 
prosodic form encode topic-comment structure and 
1The terms 'sentential it' or 'sentence pronoun' are used 
to refer to a personal pronoun whose antecedent may be a 
sentence, or, more generally: a pronoun that refers to a 
higher-order entity like a fact or a proposition. 
serve to highlight constituents whose referents the 
speaker wants to bring into focus, membership in the 
in-focus set is partially determined by linguistic form. 
For example, the subject of a matrix sentence is 
highly likely to bring an entity into focus, whereas 
this is not the case for elements in $ubordinate 
clauses and prepositional phrases (cf. the centering 
and focusing algorithms of Grosz, Joshi, & 
Weinstein 1986), 
Since objects in the world can naturally he 
referred to with expressions that function as a 
syntactic subject or object, they can obtain the In 
Focus status after being mentioned only once. A 
situation or fact, on the other hand, is typically 
expressed by a whole sentence the first time it is 
referred to. The first mention of situations or facts 
will therefore usually not correspond to expressions 
that have prominent syntactic functions in the 
sentence, and we would expect them to be activated, 
but not yet in focus after being mentioned only 
once. 2 Reference to higher-order entities with ~m 
unstressed pronoun, i.e. a form that requires its 
referent to be in focus, thus typically requires more 
than one previous mention. Consider the examples in 
(2) and 0). 
(2) There was a snake on my desk. It scaredme. 
(3) There was a snake on my desk. That scared 
me, and it scaredmy office mate too. 
In (2), the unstressed pronoun it, which requires its 
referent to be in focus, is most naturally interpreted as 
referring to the snake, not to the situation of a snake 
being on the desk. However, in (3), it is more 
naturally interpreted as referring to the fact that there 
was a snake on the speaker's desk. Since this 
situtation was referred to twice, once in the first 
sentence and once in the second, it is now in focus 
and can be referred to with an unstressed pronoun. 
The distribution of sentence it vs. that in naturally 
occuring discourse also supports the claim that a 
sentence rarely brings the situation or proposition it 
refers to into focus. 
2 Within Centering Theory, which may be seen as 
modelling the process by which an entity is activated 
and brought into focus (see Gundel 1997), situations and 
events introduced by a whole sentence are ranked lower 
as preferred centers than entities introduced by major NP 
arguments in the sentence. 
89 
PROPOSITIONAL ENRICHMENT 
In this section we will report on an experiment which 
examined the influence of some linguistic cues on 
pronoun resolution in Norwegian. Some results from 
this experiment have already been reported on in 
Fretheim & Borthen 1997. We will discuss here only 
the results concerning the interpretation of the 
Norwegian unaccented da and accentedda which are 
the counterparts of English it and that respectively. 
What seems to be the case in the texts that were 
examinedis that a sentence pronoun with two highly 
likely referents is linked to the closest antecedent 
when it is accented (neT) while it is linked to the 
furthest possible antecedent when it is unaccented 
(det). At first glance, these results appear to be 
inconsistent with GHZ's predictions, since it is hard 
to see why the referent of the more distal antecedent 
should be more likely to be in focus than the closest 
one. 
The texts were parts of an auditory listening 
comprehension test on anaphor resolution where 
reaction time was measured. (5) is one example of the 
texts the subjects judged. The subjects' task was to 
decide who was surprised in the last of the three 
sentences. There were 30 subjects altogether. 
(5) (a) Gro spurte Anne om hun var klar over at 
Senterpartiets oppslutnlng var synkende. 
(b) Hun fikk ikke noe svar. 
(c) DET kom sore en overraskelse pli heune. 
(a) 'Gro asked Anne if she was aware that the 
Center Party's popularity was decreasing'. 
(b)' She didn't get any answer.' 
(c) 'That came as a surprise to her.' 
Average reaction time: 1243ms 
30: henne 'her' = Gro and DETfinds its 
antecedentin the (b) sentence. 
0: henne 'her' = Anne and BET finds its 
antecedentin the (a) sentence. 
Even though the interpretation of the sentential 
pronoun was not directly testedin the experiment, we 
can infer the interpretations based on how the subject 
interpreted the feminine personal pronouns. In a 
minimal pair of texts only contrasting in whether the 
sentence pronoun is accented or not we get almost 
opposite results for the two text fragments. In (6), 
which is exactly like (5) except that det is unaccented, 
23 subjects chose Anne as the referentof the feminine 
pronoun henne, while only 7 subjects chose Gro. 
(6) (a) Gro spurte Anne om hun var klar over at 
Senterpartiets oppslutning var synkende. 
(b) Hun fikk ikke noe svar. 
(c) Det kom sore en overraskelse p/t henne. 
(a) 'Gro asked Anne if she was aware that the 
Center Party's popularity was decreasing'. 
(b) 'She didn't get any answer.' 
(c)' It came as a surprise to her.' 
Average reaction time: 1859ms 
7: henne 'her' = Gro and D ET finds its 
antecedentin the (b) sentence. 
23: henne 'her' = Anne andDET finds its 
antecedentin the (a) sentence. 
Since the presence vs. abseuse of accenton det is the 
only difference between the two texts, we can infer 
that this difference must have to do with what 
antecedent the accented sentence pronoun DE'/" (that') 
takes and what antecedent the unaccented det 0t') 
takes. We can assume that when the subjects choose 
Anne as the referent of the feminine pronoun, they 
have selected the antecedent of the sentence pronoun 
det as being found in sentence (a), while when they 
choose Gro, they have selected the antecedent of the 
sentence pronoun as sentence (b). 3 This is so for the 
following reasons. If it was the fact that the Center 
Party's popularity was decreasing that surprised one of 
the women, it should be Anne who was surprised, 
since Gro would not be surprised by the content of 
her own utterance. If it was the fact that Anne didn't 
answer that was surprising, it should be Gro who was 
surprised by this, not Anne. 
While the distinction between accented and 
unaccented sentence pronouns almost led to opposite 
results for (5) and (6), there was a small number of 
subjects who chose Gro as the referent of henne in the 
version with the unaccented det. The average reaction 
time for (6) is also longer than for (5). This may be 
became Gro is dearly in focus, since she is explicitly 
referred to in sentence (b). Note, moreover, that 
according to the Centering algorithm (of. Grosz, Joshi 
and Weinstein 1986, Walker and Prince 1996) the fact 
that Gro was explicitly referred to with a pronoun in 
sentence (b) would necessarily make her the backward 
3We cannot be certain from the informants' answers if it 
is the content of the whole sentence (a) or just the 
complement of 'aware' that is chosen as the referent of 
the unaccented det. but aceourding to our intuitions and 
real world knowledge, it is most probable that it is the 
content of the complement of 'aware' that would be 
interpreted as the referent of det. 
90 
. 
..! 
• ,... , • 
looking center, and therefore most likely the referent 
of henne in sentence (c). Thus, while most subjects 
accomodated in resolving the referent of henne in 
order to get a reading which was consistent with their 
interpretation of the sentence pronoun, a small 
minority chose the interpretation that was most 
consistent with refeaence processing principles for the 
feminine pronoun. 
Further support for the role of accented vs. 
unaccented sentence pronouns in determining 
interpretation of the personal pronouns in these texts 
comes from examples like (7). 
(7) (a) Gro spurte Anne om hun var klar over at 
Senterpartiets oppslutning var synkende. 
(b) Hun svarte ikke. 
(c) Det kom sore en overraskelse p~i henne. 
(a) 'Gro asked Anne if she was aware that the 
Center Party's popularity was decreasing.' 
(b) 'She didn't answer.' 
(c) 'It came as a surprise to her.' 
Average Reaction time: 1196ms 
1: henne 'her' = Gro and det finds its 
antecedentin the (b) sentence. 
29: henne !her' = Anne and det finds its 
antecedentin the (a) sentence. 
In (7), the (b) sentence is changed so that the 
expressed proposition is the same as in (5) and (6) 
(Anne did not answer Gro's question), but the 
pronoun hun ('she') in (b) must be interpreted as 
referring to Anne. 4 In this case, then, there is no 
conflict between the status of Anne as the backward 
looking center (and therefore 'in focus' entity) in (c) 
and the expected readingbased on the interpretation of 
the unstressed sentential pronoun. As predicted, 29 
out of 30 subjects didin fact interpret the text so that 
it was Anne that was surprised. That this is at least 
partly due to the unaccented form of the sentence 
pronoun in (7) is indicated by (8), which is identical 
to (7) except that da is accented. For (8), 22 out of 
30 subjects chose an interpretation that violates the 
Centering principle (andpossibly also the expectation 
that the referent of the unstressed feminine pronoun 
will be an in focus entity) but is in accordance with 
the hypothesis that an accented sentence pronoun will 
be interpreted as referring to a situation or state of 
affairs that is activated, but not in focus: 
4We have assumed that the person who does not answer 
is necessarily interpreted as Anne. while the person who 
does not get any answer is necessarily Gro. 
(a) Gro spurte Anne om hun var klar over at 
Senterpartiets oppslutning var synkende. 
(b) Hun svarte ikke. 
(c) DET kom som en overraskelse ph heune. 
(a) Gro asked Anne if she was aware that the 
Center Party's popularity was decreasing. 
(b) She didn't answer. 
(c) That came as a surprise to her. 
Average Reaction time: 1675ms 
22: henne 'her' = Gro and det finds its 
antecedentin the (b) sentence. 
8: henne 'her' = Anne and det finds its 
antecedentin the (a) sentence. 
For all the fragments in (5)-(8), the result is 
that whenever the sentential pronoun is the 
tmaccented det ('it') the subjects prefer to interpret 
henne ('her') as being Anne, which means that det 
finds its antecedentin the (a) sentence. And whenever 
the sentential pronoun is the accented DET the 
subjects prefer to interpret henne as being Gro, which 
means that DET finds its antecedent in the (b) 
sentence. But why is it that the further antecedentis 
chosen when the sentence pronoun is unaccented and 
the closer one is chosen when it is accented? If the 
predictions of the Givenness Hierarchy are correct, 
why is the content of a further antecedent more likely 
to be in focus than that of a closer one? 
We believe the answer to this question lies 
in the enriched interpretation of the second sentence. 
Even though 'answer' is used intransitively here, there 
is an implicit object which must be recovered when 
this sentence is processed, namely Gro's question 
about whether Anne was aware that the Center Party's 
popularity was decreasing. Thus, since the fact that 
the Center Party's popularity was decreasing is a part 
of the second sentence, this fact is in a sense repeated 
twice (and must be processed twice), and is therefore 
in focus at the beglnnitlg of sentence (c). If the second 
sentence were not connected to the first sentence in 
this way, we would not expect the same effect for the 
interpretation of an unaccented pronoun. This is 
supported by (9). According to our intuitions it is 
impossible for det in the third sentence of (9) to 
referto any content connected to the first sentence. It 
is more likely that det has the second sentence as its 
antecedent, but then the text is not fully coherent. 
This indicates that the implicit content of the second 
sentence in (5), (6), (7), and (8) is crucial for bringing 
the content of some fact presented in the first sentence 
of the text into foctgs. 
91 
(9) (a) Gro spurte Anne om hun var klar over at 
Senterpartiets oppslutning var synkende. 
(b) Dakom Jan plutselig inn i rommet. 
(c) Det kom sore en overraskelse p& heune. 
(a) 'Gro askedAnne if she was awarethat the 
Center Party's popularity was decreasing.' 
(b) 'Then Jan suddenly came into the room.' 
(c) 'It came as a surprise to her.' 
THE SALIENCE PROMOTING ROLE 
OF EXTRALINGUISTIC CONTEXT 
From the examples considered so far it appears that 
one-time-mention is not enough to bring a situation 
into focus. However, speakers do often accept an 
unaccented Norwegian da or English it used 
immediately after a higher-order entity has been 
introduced, and has therefore been mentioned only 
once (see Fretheim 1997). We will consider here 
some of the clearer cases of this type. 
Imagine an interviewer suddenly placing a 
microphone in front of a startledlinguist: 
(10) A: What do you think of the fact that 
linguists usually eamless than 
computer scientists? 
B: It's terrible! 
B': That's terrible! 
According to the native speakers of English we 
consulted, both B and B' are appropriate answers to 
A's question. One possible analysis of (10) while 
maintaining the predictions in the Givenness 
Hierarchy, would be to say that we accommodate and 
this is why (10A-B) is acceptable. But if this is our 
explanation, we would not be able to explain why 
(1 lb) does not sound as good as (10A-B). 
(11) (a) I have heardthat linguists earn less than 
computer scientists, and that's terrible. 
(b) .9.91have heardthat linguists earn less 
than computer scientists, and it's terrible. 
We believe the difference between (10) and (11) 
resides in the fact that some contexts are more 
salience-promoting than others. A question in an 
interviewing-situation can have the effect of 
establishing a discourse topic right away because the 
topic chosen by the interviewer is generally timely 
and/or the interviewee alreadyknows something about 
it. 
It might appear that the difference between 
(10) and (11) is attributable to the fact that the 
situation in question is presented as part of an NP in 
(10) but not in (11); however, note that in Norwegian 
we get exactly the same effect even though the fact in 
(10) is not presentedinside an NP: "Hva synes du om 
at lingvister sore regel tjener mindre penger enn data- 
ingeni0rer?" (lit. 'What do you think of that linguists 
usually earn less money than computer scientists,9') 
Also the next example suggests that the 
salience of a higher-order entity can be promoted 
nonlinguistically. We assume that the difference in 
acceptability between (12A1-B-A2) and (12A1-B-A2') 
can be attributed to the fact that the proposition 
expressed in (12B)is activated, but not in focus. 
(12) AI: Why didn't you come to the rehearsal 
yesterday? 
B: I thought I told you. I hadto help Peter 
move. 
A2: Ah, that's true. 
A2': ??Ah, it's true. 
In (13), in contrast, it is fully appropriate to use it. 
(13) A: Why didn't you come to the rehearsal 
yesterday? 
B: I thought I told you. I hadto help Peter 
move. (Pause)It's true! 
Note that (13A-B)is only coherentff there is a pause 
right before B's third utterance. And B's insistence on 
the truth of what he says will be a result of B 
believing that A does not believe him. This may be 
'activated' through a skeptical look from A, for 
instance. With no such pause or exchange of looks 
B's answer wonldnot sound natural. (13) thus shows 
that cognitive status can be raised by extralinguistic 
means. 
77 CONCLUSIONS 
We have seen how the form of an anaphor referring to 
a higher-order entities may influence the interpretation 
of the anaphor. On the basis of this and on the basis 
of a restrietedanalysis of natural speech and intuitive 
judgments of invented texts, we conclude that the use 
of sentential pronouns in English and Norwegian can 
be explained by the theory proposed in Gundel, 
Hedberg and Zacharski (1993), which links different 
referring forms to the assumed cognitive status of the 
referent. According to this theory, a necessary 
condition for appropriate use of English it and 
Norwegian unaccenteddet is that the referent is in the 
92 
i . 
• . . • . 
i. 
addressee's focus of attention. Explicit subsequent 
mention is one way of bringing a higher-order entity 
into focus of attention. But in addition to that we 
have shown that an account of propositional 
enrichment (based on linguistic or extralingnistic 
context) is necessary in order to adequately capture 
these facts. It also seems to be the case that some 
contexts or situations are more salience-promoting 
than others, and that a person's general knowledge and 
expectations may affect the degree to which 
something is easily brought into focus. It is not clear 
how such factors can be taken into account in a 
computer-based processing system, but propositional 
enrichment is indeed one factor that would lead to a 
more adequateaceount of pronoun interpretation. 

REFERENCES 
Ariel, Mira (1988) "Referring and accessibility." 
Journal of Linguistics 24. 67-87. 
Fmurud, Karl (1992) Processing Noun Phrases in 
Natural Discourse. Ph.D. Thesis. Stockholm 
University, Dept. of Linguistics. 
Fretheim, Thorstein (1997) "Accented and Unaccented 
Situation Anaphom: do they work the way we 
expect them to?" Paper presented at the conference on 
Grammatik i Fokus, Lund, February 6, 1997. 
Fretheim, Thorstein & Kaja Borthen (1996) "Some 
unexpected determinants of local refematial 
(dis)continuity" in S. Bofley et al., eds. Approaches 
to Discourse Resolution Colloquium (DAARC96), 
Lancaster. 
Fretheim, Thorstein & Kaja Borthen (1997) 
"Linguistic Constraints on Relevance in Reference 
Resolution." To appear in Proceedings of the 1997 
Meeting of the Chicago Linguistic Society. 
Garrod, S. C., andA. J. Sandford(1982) "The Mental 
Representation of Discourse in a Focussed Memory 
System: Implications for the Interpretation of 
Anaphofie Noun Phrases." Journal of Semantics 1. 
21-41. 
Gfice, H. P. (1975) "Logic andConversation." In 
P. Cole andJ.L. Morgan, eds., Syntax and Semantics 
3. Speech Acts. 41-58, 
Grosz, Barbara J., Aravind K. Joshi, & Scott 
Weinstein (1986) 'Towards a Computational Theory 
of Discourse Interpretation." Unpubfished ms. 
Gundel, Jeanette K., Nancy Hedberg and Ron 
Zacharski (1993) "Cognitive Status and the Form of 
Referring Expressions in Discourse." l~nL, uage 69. 
274-307. 
Gundel, Jeanette K. (1997) "Centering Theory and the 
Givenness Hierarchy: Towards a Synthesis." In A. 
Joshi, E. F. Prince and M. Walker, eds. Centering 
Theory in Discourse. Oxford University Press. 
Walker, Marilyn A. & Ellen F. Prince (1996) "A 
Bilateral Approach to Giveuness: A Hearer-Status 
Algorithm and a Centering Algorithm." In T. 
Fretheim and J. K. Gundel, eds. Reference and 
Referent Accessibility. Amsterdam: John Benjamins. 
