Journal of Pragmatics 39 (2007) 292–304

www.elsevier.com/locate/pragma

A textual perspective on Givón’s quantity principle

Shaojun Ji *
Department of English, Chinese Culture University, Taipei 111, Taiwan
Received 11 November 2005; received in revised form 17 January 2006; accepted 22 January 2006

Abstract
This paper reports on a study conducted to test in written texts the applicability of one of Givón’s iconic
coding principles, the quantity principle, which states that less predictable information will be given more
coding material. The study bases its investigation on English narrative texts, with attention focused on
comparing the linguistic constructions that are used at episode- and subepisode-initial positions in these
texts. The results of the study support the quantity principle. They specifically indicate that the episode-
initial constructions, which function to signal major breaks in the thematic structures of the narrative texts,
tend to carry more pieces of new (and therefore unpredictable) information than their subepisode-initial
counterparts and in correspondence to this difference the former use more coding material than the latter.
# 2006 Elsevier B.V. All rights reserved.

Keywords: Iconicity; Narrative; Episode; Information

1. Introduction

Givón (1995) provides an elaborate discussion on three principles of iconic coding, one of
which is the quantity principle. An implication of this principle, which states that ‘‘less
predictable information will be given more coding material’’ (Ibid.:49), can be observed in the
following examples:

(1) a. Once there was a wizard. He lived in Africa.

He went to China to [Ø] get a lamp. The wizard . . .
b. Joe told Bill and then [Ø] Sally.
c. Joe told Bill, then he told Sally.
d. Joe told Bill, then he told Sally. (Ibid.:50)

* Tel.: +886 2 2862 5894; fax: +886 2 2861 8279.

E-mail address: sjji_ca@yahoo.ca.

0378-2166/$ – see front matter # 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.pragma.2006.01.010
 S. Ji / Journal of Pragmatics 39 (2007) 292–304 293

These examples show that the full nouns, independent or unstressed pronouns, and zero
anaphoras are used in a way that corresponds to an increasing degree of informational
predictability. The full nouns are used to express the least predictable information, the
independent or unstressed pronouns are preferred for coding the somewhat more predictable
information, and the zero anaphoras occur in contexts where the information is most predictable.
The quantity principle as proposed by Givón reveals what is more commonly known as
diagrammatic iconicity (Haiman, 1980). Such iconicity involves no resemblance between a
linguistic sign and its referent as in the case of onomatopoeic words. It is mainly reflected in the
way the relationship between linguistic signs mirrors the relationship between their referents. An
example of this type of iconicity can be found in Tai’s observation (Tai, 1985) that in Chinese
syntax, the order of linguistic constructions such as phrases or clauses corresponds with the
temporal order of the events they refer to. What Givón’s quantity principle highlights, as
illustrated by the above examples, is an iconic relationship in which the referents with higher or
lower degrees of informational predictability are coded by the linguistic constructions of smaller
or larger sizes, respectively.
The quantity principle points out something very fundamental about the way language is used.
Although it was first put forward as a syntactic coding principle, it is worthwhile to find out if the
same principle also applies in written texts. The experimental study reported below is conducted
for such a purpose.

2. The quantity principle from a textual perspective

To find out about the applicability of the quantity principle in written texts, this study bases its
investigation on narrative texts, with attention focused on the linguistic constructions that are
used at the beginning of episodes and subepisodes in these texts. The rationale for this approach
is as follows: In a narrative text, the writer typically describes series of events from an actual or
fictional world in the past. These events often take place in different episodes that may involve
different participants doing things at different places and different times. The empirical support
for the validity of the episode has been provided by some psycholinguistic studies which
indicate that people recalling stories treat the information of an episode as an integral unit (e.g.,
Black and Bower, 1979; Glenn, 1978; Thorndyke, 1977). This finding is in keeping with the
results of other studies (e.g., Haberlandt et al., 1980; Mandler and Goodman, 1982) that have
found the so-called episode-shift effect: readers pause longer in processing episode-initial
sentences than episode-internal sentences. Taken together, these studies suggest that the episode
as an intermediate unit in a narrative text is psychologically real and recognizable on thematic
grounds.
To help the reader to perceive the organization of a narrative text the writer signals transitions
from one episode to another. Based on an analysis of narrative texts in the form of fairy-tales and
travel-guide articles, Virtanen (1992) shows us how clause-initial adverbial phrases of time and
place are used to signal shifts of textual units of various sizes such as sections, episodes, and
moves. According to Prideaux and Hogan (1993), in both oral and written narratives, preposed
subordinate clauses of time (characterized as a marked construction) occur statistically far more
frequently than nonpreposed ones (characterized as an unmarked construction) at the beginning
of an episode. Their analysis indicates that as a device for thematic reorientation, preposed
subordinate clauses are frequently used to code the beginning of a new discourse unit such as an
episode. Givón (1993:315) makes a similar observation. In addition, he points out that preposed
participle clauses may serve the same purpose. In his view, all these sentence-initial
294 S. Ji / Journal of Pragmatics 39 (2007) 292–304

constructions have their coherence links in two directions, anaphoric and cataphoric. Their
anaphoric links can reach back to the thematic information anywhere in a preceding textual unit,
while their cataphoric links anchor themselves nicely to the main clause, which then launches a
new unit.
As the intermediate units of narrative texts, episodes, however, are not monolithic in structure.
They often develop in subsections resulting from all kinds of ways events take place and relate to
one another. In addition to narrating the main events in these subsections, episodes may also
embed subsections that are devoted to recounting things that happen at other times, describing the
participants in the story and their personalities, expressing the narrator’s or a participant’s views
on various subjects, depicting natural environments as part of the story world, evaluating certain
situations or consequences of events, explaining various phenomena, and so forth. All these
subsections, which may be called subepisodes, form coherent units in their own right. In an
experimental study exploring the conceptual basis of episode transitions and their linguistic
expressions, Ji (2002) finds that in identifying episodes the subjects also pick out these
subepisodes on similar grounds and that the beginnings of subepisodes are linguistically
expressed in similar ways as those of episodes.
From the viewpoint of this study, the information carried by the linguistic constructions
used at episode- and subepisode-initial positions may be unpredictable to varying degrees as it
is often associated with changes of various discourse elements.1 Since an episode is more
often set in a larger story world than a subepisode with more different discourse elements
involved, the information of these changes at episode-initial positions is likely to be more
unpredictable than its counterpart at subepisode-initial positions. If we regard unpredictable
information as being equivalent to new information, what has been stated may also mean that
the episode-initial constructions are likely to contain more new information than their
subepisode-initial counterparts. In the light of Givón’s quantity principle and for the reason
that has just been mentioned, this study hypothesizes that the linguistic constructions used at
episode-initial positions will be coded with more words in statistical terms than those used at
subepisode-initial positions. Consequently, the research question for this study is: Are the
beginnings of episodes given more coding material than the beginnings of subepisodes in
narrative texts?
As a first step to find an answer to this question, this study acquired the necessary data for
analysis by conducting a linguistic experiment in which 20 native speakers of English were asked
to segment 10 unparagraphed English narrative texts into episodes.2 The narrative texts used in
the experiment were typical personal accounts of past experience taken from Reader’s Digest.3

1
According to the results of some previous studies (e.g., Chafe, 1980; Virtanen, 1992; Prideaux and Hogan, 1993; Ji,
2002), transitions between textual units in narrative texts are frequently found to be characterized, among other things, by
changes in time, place, participant, theme, or event structure.
2
This study does not rely on the original paragraph divisions in the narrative texts in deciding on the episode divisions
of the texts because, according to some previous studies (e.g., Braddock, 1974; Halliday and Hasan, 1976; Longacre,
1979), not all paragraph divisions correspond with the thematic structures of the texts. Some paragraph divisions may be
made as a result of stylistic concern or personal preference. The acquisition of the episode divisions of the narrative texts
in the way used in this study assures us that they are made consistently on thematic grounds, i.e., on the basis of the
episodic developments of the narrative texts.
3
These texts from Reader’s Digest are: Brothers Under the Skin (April 2003, pp. 58–64), Tommie’s Rainbow (May
2003, pp. 81–85), To Russia with Love (September 2003, pp. 46–53), Entwined (November 2003, pp. 34–39), Operation:
Whale Rescue (November 2003, pp. 72–80), Soaring the Glory (December 2003, pp. 62–67), One Wrong Turn (December
2003, pp. 74–80), Disaster on the Aegean Sea (February 2004, pp. 84–90), For Amie (March 2004, pp. 62–69), and Hope
Floats (August 2004, pp. 20–25).
 S. Ji / Journal of Pragmatics 39 (2007) 292–304 295

An episode was defined to them as a portion of a narrative text that relates to an event or a series of
connected events and forms a coherent unit in itself. The subjects, consisting of 10 male and 10
female undergraduate students at the University of Alberta, Canada, identified the beginning of
each episode in a narrative text by placing a slash before the first word of an episode.
By carrying out the experiment, this study obtains divisions of the narrative texts consistently
made on the basis of their episodic developments. Furthermore, as they are made by ordinary
language users rather than the conductor of this study, these divisions represent their unbiased
views of the episodic structures of the segmented texts. Based on their segmentations, episodes
and subepisodes are then differentiated, followed by a comparative analysis of the linguistic
constructions used at their initial positions.

3. Results and discussion

After the segmented texts were collected, all the points of episode divisions were tabulated. As
expected, the subjects did not make exactly the same choices. However, a large number of their
choices do converge to varying degrees. These shared choices form the data for analysis in this
study.
As mentioned earlier, a differentiation between episodes and subepisodes is necessary
before data analysis can be conducted. In making this differentiation, this study treats the
episode divisions shared by 65% or more of the subjects as the boundaries of representative
episode units identified in the segmented texts. There are two reasons for choosing this
percentage: First of all, the episode divisions classified into this group enjoy a reasonable
majority, i.e., they are identified by more than half of the subjects. Secondly, this reasonable
majority can give us more tokens of episode divisions to be analyzed than greater majorities as
represented by 70% or 80%. Some of the episode divisions made by less than 65% of the
subjects are used as another group of data for contrastive analysis. These are the episode
divisions made by 30–64% of the subjects.4 Because of the lower identification rates these
divisions are considered less definite cases of episode units. However, since they are identified
by about half of the subjects on average, it is plausible to assume that these divisions are
identified on similar grounds as those in the first group but somehow differ from them in certain
aspects. Based on this reasoning, these less definite cases are treated as the subepisodes within
the episodes identified in the first group.
Once episodes and subepisodes are differentiated, the focus of our attention turns to the initial
constructions of their initial sentences, namely the subjects of the sentences and the adverbials
occurring before the subjects. Functionally these constructions are the key elements for
indicating the beginnings of episodes or subepisodes (as suggested by some of the studies
mentioned above). The adverbials are mainly responsible for establishing new temporal or spatial
settings or thematic reorientations of various kinds whereas the subjects are referring expressions
of the participants and other nonhuman referents in the stories. Semantically the information they
carry may be unpredictable to varying degrees in different parts of the stories. Since the adverbial
and subject play different roles, the constructions comprising an adverbial and those comprising a
subject are dealt with in two separate groups. Identifying these constructions and recording the
number of words used in each sentence constitute part of the data analysis.

4
The episode divisions made by less than 30% of the subjects are considered too unrepresentative to be included as part
of the data for further analysis.
296 S. Ji / Journal of Pragmatics 39 (2007) 292–304

The following examples of episode- and subepisode-initial sentences from the story entitled
Disaster on the Aegean Sea are provided to illustrate this part of the data analysis.5

(2) (E) A little after 5 p.m., <the Samina> cast off with more than 500 passengers aboard,
most of them Greeks but also tourists from Canada, Britain, France, Germany and
Australia. The ferry was scheduled to reach Páros, its first stop, at 10:15 p.m. (followed
by a description of some activities going on the ship)
(3) (E) Some hours later <Lajoie, comfortable on a sofa,> checked her watch. We should
be at Páros in about 15 minutes, she thought. (followed by an account of how the first
sign of a sea disaster was noticed)
(4) (E) <The harbor boundaries for Páros> are marked by the Portes Islets, two large
rocks about 15 meters high. The taller one, about two kilometers from shore, is topped
with a navigation light. (followed by an account of how the captain handled the ship
and what happened to it when it approached the harbor)
(5) (E) Down in the lounge, <a woman with a baby in her arms> screamed. Then the light
went out. (followed by a description of how people were running for their lives as the
disaster struck)
(6) (SE) Suddenly, <the friends> spotted a safety exit under the bridge, with a wooden
ladder hanging from it. In the midst of shoving and pushing, Lajoie began to climb,
followed by Désautels. (followed by a further revelation that the ship was sinking)
(7) (E) Strolling through the streets of Páros, <Steve Parry> had just notified his crew of
amateur sailors that they would have to spend an extra day on the island if the Force 8
wind kept up. ‘‘It’d be crazy to leave under these conditions,’’ he said. (followed by a
description of him as an experienced sailor)
(8) (SE) Shortly before midnight, <Parry> received a call from one of his sailors on his
portable VHF: ‘‘Can you come to the port right away? The authorities are asking all
boat captains to put out to sea.’’ (followed by an account of how he heard about the
sinking ship)
(9) (SE) As they cleared the harbor, <Parry> assigned duties. One man would sweep the
water’s surface with a giant floodlight, another would shine a light on the instrument
panel, and the passengers would tie rope ladders. (followed by a description of the
dangerous environment the rescuers had to face)
(10) (E) After jumping into the water, <Désautels> had been sucked under twice. I’m
going to die, she thought. (followed by an account of what was going on in her mind
and out on the sea)
(11) (SE) ‘‘Sophie!’’ cried <Lajoie>, meters away. The relieved friends embraced, then
swam to a capsized life raft to which two other victims were clinging. (followed by an
account of how Lajoie and another passenger were struggling in the sea)

5
In the presentation of these examples, an episode-initial sentence is labeled as (E) whereas a subepisode-initial
sentence is labeled as (SE). Initial adverbial structures are marked by underlining, while initial subjects are indicated by
angled brackets.
 S. Ji / Journal of Pragmatics 39 (2007) 292–304 297

(12) (E) Before diving from the Samina, <George Douris> had found Nicos. ‘‘If the
undertow throws us against those rocks, we’re dead,’’ he told his brother. (followed by
an account of how the two brothers were struggling for their lives in water)
(13) (E) Tossed about, her knees trembling and her palms raw, <Lajoie> felt her strength
waning. ‘‘Hold on,’’ Désautels encouraged her. (followed by an account of how the two
women got rescued)
(14) (E) At about 1 a.m., three hours after the Samina had gone down, <the Margarita>
passed several sailboats and fishing boats returning from the site of the tragedy. Then
she entered the disaster zone. (followed by an account of how more people got rescued)
(15) (SE) <They> spotted one victim floating like a cock. ‘‘It’s all right,’’ Parry yelled to
him in Greek, before Yannis tossed him a rope. (followed by an account of how this
man was rescued)
(16) (SE) From about 100 meters away, <Douris> has spotted Parry’s sailboat. It’s our last
chance, he told himself, while Nicos moaned quietly beside him. (followed by an
account of how the two men got rescued)

In the data analysis as exemplified above, the adverbial and subject constructions in the
episode- and subepisode-initial sentences are tabulated in two groups, while each group contains
several categories. Tabulating the constructions in each group in a set of categories allows us to
show the difference in their distribution across the two groups.
Following this approach, the adverbial constructions are tabulated as adverb, adverbial
phrase, and adverbial clause. A typical example of the adverbs is a word like then, which
provides a close link in time between two events, or a word like slowly, unexpectedly, or suddenly
(as in (6)), which indicates the manner in which an event takes place. The adverbial phrases are
often expressed by a noun phrase such as one day, the next morning, or some hours later (as in (3))
or a preposition phrase (as in (2), (5), (8), (14),6 and (16)). The adverbial clauses may be
represented by a subordinate clause (as in (9) and (14)), a participle clause (as in (7), (10), (12),
and (13)),7 or a verbless clause such as the and-clause in the adverbial construction of (13).
In tabulating the subject constructions, we classify them in the categories of pronoun, single
noun, noun phrase, and modified noun phrase. The category of pronoun is pretty straightforward.
It involves the use of either a personal pronoun (as in (15)) or a demonstrative pronoun. Equally
straightforward are the single nouns, which consist of either a class noun or, in most cases, a
person’s first or last name (as in (3), (8), (9), (10), (11),8 (13), and (16)). In contrast to the
simplicity of the first two categories, the noun phrases involve more variations. They may be
expressed by a person’s full name (as in (7) and (12)) or a compound noun such as school
counselor, doctor John Mullen, operating rooms, and so forth. They may also be represented by a
structure containing a noun preceded by a possessive pronoun or an article (as in (2), (6), and

6
The first part of the adverbial in (14) at about 1 a.m. is tabulated as a preposition phrase whereas the second part three
hours after the Samina had gone down as a subordinate clause.
7
The first two clauses of the adverbial in (13) are regarded as a past participle clause and a present participle clause,
respectively.
8
Sentence (11) does not start with the subject as in most of the cases but with what someone said. From the viewpoint of
textual development, however, the sentence seems to be used to indicate a change of participant in the story in a more
dramatic way. For this reason, although the subject of the sentence does not occur at the beginning of the sentence, it is
recorded as if it did.
298 S. Ji / Journal of Pragmatics 39 (2007) 292–304

Table 1
T-test comparison of the number of words used as adverbials
Group n Mean S.D. t d.f. p
E 107 5.09 4.80 6.2087 218 .0001
SE 113 1.86 2.69
Note: S.D. = standard deviation; t = t value; d.f. = degrees of freedom; p = probability level.

(14)). The category of modified noun phrase characterizes the type of noun phrase that has some
modifying elements, such as an adjective phrase (as in (3)), a preposition phrase (as in (4) and
(5)), or an attributive clause.
Setting up the category of single noun and keeping it distinct from that of noun phrase is
primarily for the purpose of showing the frequent use of a person’s first or last name versus the
use of his/her full name. It is often observed in the data that when a major participant is
introduced into a story for the first time, typically in an episode-initial sentence such as (7), the
full name of this participant is used. When this person is mentioned subsequently in subepisode-
initial sentences like (8) and (9), it is only the last name that is used.9 The tabulations of the first or
last name versus the full name in the two categories reflect this difference in distribution across
the episode- and subepisode-initial sentences.
Compared with the noun phrases, the modified noun phrases (as in (3)–(5)) generally contain
more words as the modifiers are providers of some extra information. As we can see from these
examples, the extra pieces of information help to define the referents in more detailed ways,
which are relevant to the immediate context.
In the ways presented above, the adverbial and subject constructions are identified and the
words used in each of them are counted and recorded. Take (2) as an example. Its sentence-initial
adverbial is identified and recorded as an adverbial phrase made up of five words10 and its subject
as a noun phrase consisting of two words. When two or more constructions are used as the
adverbial of a sentence (as in (13) and (14)), all of the constructions are tabulated and the total
number of words are added up. In cases where the adverbial part is missing, as exemplified by (4),
(11), and (15), the number of words for this part is recorded as zero.11
After the tabulation is done, the episode- and subepisode-initial sentences are compared
through unpaired t-tests in terms of the number of words used in the categories of adverbial and
subject, respectively. Tables 1 and 2 give the t-test results.
As we can see from the two tables, there are 107 episode-initial sentences (labeled as E) and
113 subepisode-initial sentences (labeled as SE) in the narrative texts. The t values indicate that at
the probability level of .0001, there is a significant difference in the number of words used for
both the adverbial and subject between the episode- and subepisode-initial sentences. That is, for
both categories more words are used in the episode-initial sentences than in the subepisode-initial
sentences. The differences between the two groups of data become more concrete if we list the
numbers of the specific constructions used in each category.

9
In many similar cases it is the first name that is used.
10
When a construction is recorded, we simply count all of the words that are used to form the construction without
distinguishing them in any way.
11
In our analysis, the zero number of words recorded does not simply mean that in this sentence no adverbial
construction is used. Rather, our interpretation is that in such a case the information normally carried by the adverbial part
is most predictable in the immediate context and therefore requires no linguistic expression.
 S. Ji / Journal of Pragmatics 39 (2007) 292–304 299

Table 2
t-test comparison of the number of words used as subjects
Group n Mean S.D. t d.f. p
E 107 3.10 3.36 4.0780 218 .0001
SE 113 1.74 1.10

Table 3
Adverbial constructions in episode- and subepisode-initial sentences
Adverbial constructions Episode-initial sentences Subepisode-initial sentences
Adverb 4 (4%) 14 (12%)
Adverbial phrase 56 (52%) 29 (26%)
Adverbial clause 36 (34%) 21 (19%)
Total 96 (90%) 64 (57%)a
a
Because some episode- and subepisode-initial sentences contain no adverbial constructions, the total number of the
adverbial constructions in each column does not add up to the total number of the initial sentences in each group.

Table 4
Subject constructions in episode- and subepisode-initial sentences
Subject constructions Episode-initial sentences Subepisode-initial sentences
Pronoun 12 (11%) 19 (17%)
Single noun 32 (30%) 50 (44%)
Noun phrase 36 (34%) 27 (24%)
Modified noun phrase 27 (25%) 17 (15%)
Total 107 (100%) 113 (100%)

Tables 3 and 4 list the numbers of the adverbial and subject constructions in the two groups.
Take the use of adverbial phrases in the episode-initial sentences as an example. Table 3 shows
that 56 adverbial phrases are used in a total of 107 episode-initial sentences, which means they
occur in 52% of all episode-initial sentences. Based on the numbers given in Table 3, we can
observe that not only is the total percentage (90%) of episode-initial sentences that contain
adverbial constructions considerably higher than that (57%) of subepisode-initial sentences, but
the episode-initial sentences also use more adverbial phrases and adverbial clauses if we compare
the adverbial constructions that are used in the two groups of sentences. From Table 4 we find
that the episode-initial sentences use more noun phrases and modified noun phrases for their
subjects than their subepisode-initial counterparts. The differences between the two groups in the
use of these constructions are obviously the major factor that contributes to the results from the
t-tests.
Our interpretation of the differences revealed here, in keeping with the hypothesis of this
study, is that the episode-initial constructions, which are used to signal major breaks in the
thematic structures of the narrative texts, tend to carry more unpredictable information than their
subepisode-initial counterparts. If we equate predictable and unpredictable information with
given and new information, respectively, the interpretation suggests that the episode-initial
constructions tend to be loaded with more pieces of new information than their subepisode-initial
counterparts. To find out if this is the case, we examined each construction in the two groups of
300 S. Ji / Journal of Pragmatics 39 (2007) 292–304

data with regard to whether its information status is given or new. Guiding our evaluation is the
analytic framework offered by Prince (1981, 1992).
Based on the idea of ‘assumed familiarity’, Prince distinguishes between three types of
information status: ‘evoked’, ‘inferrable’, and ‘new’. The evoked information refers to the
information that has become known to the hearer/reader through linguistic contexts or
communicative situations and for this reason is equal to what is more commonly known as given
information. The inferrable information, as characterized by Prince, is the type of information
that the speaker/writer would expect the hearer/reader to be able to infer through other discourse-
old entities. The new information, as the third type, may represent a kind of ‘brand-new’
information that the hearer/reader cannot possibly know through linguistic contexts or
communicative situations, or it may be the kind of information associated with an ‘unused’ entity.
Compared with a brand-new entity, an unused entity is assumed by the speaker/writer to be
somewhat more familiar to the hearer/reader and therefore identifiable.12 Both brand-new and
inferrable information may be ‘anchored’ to an evoked entity or expressed in an ‘unanchored’
way.
Although the three types of information characterized by Prince were first proposed for the
analysis of noun phrases, they have been applied suitably in analyzing various kinds of English
syntactic constructions such as subordinate clauses (Abraham, 1991), adverbial phrases of time
and place (Virtanen, 1992), and inverted sentences (Birner, 1994, 1997). Following a similar
approach as in these studies, this study takes a close look at each episode- and subepisode-initial
construction in its data and evaluates the information status of its content in terms of Prince’s
taxonomy. To illustrate this evaluation, the adverbial and subject constructions used in
Examples (2) through (16) are presented again below with their information status indicated and
explained:

(2) (E) A little after 5 p.m.: inferrable information; <the Samina>: evoked information.
Explanation: Since the time when passengers got on board was indicated earlier, the
time of departure mentioned here is regarded as inferrable information. The name of
the ship is evoked because it was already introduced at the beginning of the story.
(3) (E) Some hours later: inferrable information; <Lajorie, comfortable on a sofa>: new
information.
Explanation: The time indicated here is inferrable from the time provided earlier.
Although Lajorie has already been introduced, the adjective phrase that modifies the
name provides some additional new information about the way she is at the moment.
(4) (E) <The harbor boundaries for Páros>: inferrable information.
Explanation: Since the coastal town Páros has already been mentioned, the fact that it
has a harbor is inferrable information.
(5) (E) Down in the lounge: inferrable information; <A woman with a baby in her arms>:
new information.
Explanation: The fact that a passenger ship may have a lounge should be inferrable on
the part of the reader. The fact that one particular passenger is cited as an example to
illustrate the panic reactions from the passengers on board represents new information.

12
Following Prince, the information of an unused entity is coded as new information in our data analysis.
 S. Ji / Journal of Pragmatics 39 (2007) 292–304 301

(6) (SE) Suddenly: inferrable information; <The friends>: evoked information.

Explanation: The way that the action takes place as described by the adverb is treated
as inferrable information.13 The friends refers to Lajoie and Désautels who have
already been introduced as such.
(7) (E) Strolling through the streets of Páros: new information; <Steven Parry>: new
information.
Explanation: What the participle clause indicates has no connection with the pre-
ceding episodic development and therefore describes a totally unpredictable action
performed by a character introduced into the story for the first time.
(8) (SE) Shortly before midnight: inferrable information; <Parry>: evoked information.
Explanation: The adverbial part in this sentence is treated as inferrable information for
reasons similar to those given in (2). As Steven Parry is mentioned here again, only his
last name is used.
(9) (SE) As they cleared the harbor: inferrable information; <Parry>: evoked informa-
tion.
Explanation: Based on the episodic development, it is easily inferrable that the
rescuers are on their way to the sinking ship. Steven Parry is mentioned again as the
leader of the operation.
(10) (E) After jumping into the water: inferrable information; <Désautels>: evoked
information.
Explanation: The action described by the participle clause can be inferred from the
fact mentioned earlier that the ship was going down. Désautels who takes the action
was also mentioned earlier.
(11) (SE) <Lajoie>: evoked information
Explanation: Lajoie has been mentioned before. She is the focus of attention again in
the story.
(12) (E) Before diving from the Samina: inferrable information; <George Douris>: evoked
information.
Explanation: The adverbial part is coded as inferrable information for the same reason
as given for the adverbial part in (10). George Douris, who was mentioned earlier, gets
the attention again as the story develops.
(13) (E) Tossed about, her knees trembling and her palms raw: new information;
<Lajoie>: evoked information.
Explanation: The information provided by the three adverbial clauses is unpredictable
from the previous contexts. It announces a new turn in the story that involves one of the
Canadian women.
(14) (E) At about 1 a.m., three hours after the Samina had gone down: inferrable
information; <the Margarita>: evoked information.

13
The rationale for treating an adverb such as suddenly as a carrier of inferrable information is that the various ways that
events take place as typically described by different adverbs are familiar to language users through their cognitive
experience and therefore identifiable in the right contexts of a story.
302 S. Ji / Journal of Pragmatics 39 (2007) 292–304

Explanation: The time indicated by the preposition phrase and the following speci-
fication of it in relation to the sinking of the ship are both inferrable from the
development of the scenarios. The rescuing boat the Margarita is already known to the
reader.
(15) (SE) <They>: evoked information.
Explanation: The personal pronoun They refers to Parry and some other rescuers who
have already become familiar to the reader.
(16) (SE) From about 100 meters away: new information; <Douris>: evoked information.
Explanation: The exact location of Douris is coded as new information because of its
unpredictability. Douris has already been mentioned a few times.

Summarized in Tables 5 and 6 are the results of the examination of all episode- and
subepisode-initial constructions with regard to the information status of their content.
The results clearly indicate that the episode-initial sentences tend to carry more new
information than the subepisode-initial sentences in their sentence-initial adverbials and subjects.
The main reason, as may be perceived from the analysis of the given examples, is that at the
beginnings of some episodes (such as (7) and (13)) totally new settings of various nature need to
be established and new participants introduced, sometimes with extra information specifying
their identity or the way they are in a particular situation (as in (3) and (5)). The newness of such
information apparently requires more linguistic material to code.
What we may also notice from the above results is that a considerable number of adverbials in
both episode- and subepisode-initial sentences are characterized as containing inferrable
information. As we can see from the analysis of the given examples, these adverbials, because of
their thematic connection to certain discourse elements that are already evoked, serve nicely as
bridges between two textual units. In other words, they can signal the start of a new unit by
providing information inferrable from the previous discourse. As for the subjects, once they are
introduced, they become known and remain as such in the rest of the stories. This explains why in
both episode- and subepisode-initial sentences the majority of the subjects are characterized as

Table 5
Information distribution in episode-initial constructions
Episode-initial sentences
Adverbial Subject
Evoked Inferrable New Evoked Inferrable New
1 (1%) 37 (35%) 58 (54%) 65 (61%) 7 (6%) 35 (33%)

Table 6
Information distribution in subepisode-initial constructions
Subepisode-initial sentences
Adverbial Subject
Evoked Inferrable New Evoked Inferrable New
7 (6%) 36 (32%) 21 (19%) 88 (78%) 9 (8%) 16 (14%)
 S. Ji / Journal of Pragmatics 39 (2007) 292–304 303

carrying evoked information. From time to time (as the small numbers of the inferrable tokens
show) they may have to be inferred when they are subsequently replaced by a formally different
yet semantically synonymous construction.

4. Conclusion

The results presented above support the hypothesis of this study and prove Givón’s quantity
principle to be applicable in written texts. They enable us to see the iconic nature of the
correspondence between higher degrees of informational unpredictability and more linguistic
coding material in English narrative texts. A comparison of the linguistic constructions used at
episode- and subepisode-initial positions indicates that the former, which tend to contain more
pieces of new information associated with certain changes of various discourse elements, use
more words than the latter. Such a finding is evidenced in the episode-initial adverbials not only
by the use of more adverbial constructions as a whole, but also by the use of more adverbial
phrases and clauses in particular. It is also supported by the use of more noun phrases and
modified noun phrases in the episode-initial subjects.

Acknowledgements

The author wishes to express his deep gratitude to the National Science Council, R. O. China
for providing a research grant for the study reported in this paper. He also wants to thank the two
anonymous reviewers for their detailed comments and valuable suggestions on an earlier version
of this paper.

References

Abraham, Elyse, 1991. Why because? The management of given/new information as a constraint on the selection of
causal alternatives. Text 11 (3), 323–339.
Birner, Betty J., 1994. Information status and word order: an analysis of English inversion. Language 70 (2), 233–
259.
Birner, Betty J., 1997. The linguistic realization of inferrable information. Language and Communication 17 (2), 133–
147.
Black, John G., Bower, Gordon H., 1979. Episodes as chunks in narrative memory. Journal of Verbal Learning and Verbal
Behavior 18, 309–331.
Braddock, R., 1974. The frequency and placement of topic sentences in expository prose. Research in the Teaching of
English 8, 287–302.
Chafe, Wallace, 1980. The deployment of consciousness in the production of a narrative. In: Chafe, W. (Ed.), The Pear
Stories: Cognitive, Cultural, and Linguistic Aspects of Narrative Production. Ablex, Norwood, NJ, pp. 9–50.
Givón, Talmy, 1993. English Grammar A Function-Based Introduction, vol. II. John Benjamins, Amsterdam and
Philadelphia.
Givón, Talmy, 1995. Isomorphism in the grammatical code. In: Simone, R. (Ed.), Iconicity in Language. John Benjamins,
Amsterdam and Philadelphia, pp. 47–76.
Glenn, Christine G., 1978. The role of episodic structure and of story length in children’s recall of simple stories. Journal
of Verbal Learning and Verbal Behavior 17, 229–247.
Haberlandt, Karl, Berian, Clair, Sandson, Jennifer, 1980. The episode schema in story processing. Journal of Verbal
Learning and Verbal Behavior 19, 635–651.
Haiman, John, 1980. The iconicity of grammar: isomorphism and motivation. Language 56, 515–540.
Halliday, Michael A.K., Hasan, Ruqaiya, 1976. Cohesion in English. Longman, London.
Ji, Shaojun, 2002. Identifying episode transitions. Journal of Pragmatics 34, 1257–1271.
Longacre, Robert E., 1979. The paragraph as a grammatical unit. In: Givón, T. (Ed.), Discourse and Syntax. Syntax and
Semantics, vol. 12. Academic Press, New York, pp. 115–134.
304 S. Ji / Journal of Pragmatics 39 (2007) 292–304

Mandler, Jean M., Goodman, Marsha S., 1982. On the psychological validity of story structure. Journal of Verbal Learning
and Verbal Behavior 21, 507–523.
Prideaux, Gary D., Hogan, John T., 1993. Markedness as a discourse management device: the role of alternative adverbial
clause orders. Word 44 (3), 397–411.
Prince, Ellen F., 1981. Towards a taxonomy of given-new information. In: Cole, P. (Ed.), Radical Pragmatics. Academic
Press, New York, pp. 223–255.
Prince, Ellen F., 1992. The ZPG letter: subjects, definiteness, and information-status. In: Thompson, S., Mann, W. (Eds.),
Discourse Description: Diverse Analyses of a Fundraising Text. Benjamins, Amsterdam and Philadelphia, pp. 295–
325.
Tai, James H.-Y., 1985. Temporal sequence and Chinese word order. In: Haiman, J. (Ed.), Iconicity in Syntax. John
Benjamins, Amsterdam and Philadelphia, pp. 49–72.
Thorndyke, P.W., 1977. Cognitive structure in comprehension and memory of narrative discourse. Cognitive Psychology
9, 77–110.
Virtanen, Tuija, 1992. Given and new information in adverbials: clause-initial adverbials of time and place. Journal of
Pragmatics 17, 99–115.

Shaojun Ji received his M.Phil. in modern English language from the University of Glasgow, U.K. and his Ph.D. in
linguistics from the University of Alberta, Canada. He is currently teaching in the Department of English at Chinese
Culture University, Taiwan. His research interests are in the areas of discourse analysis, pragmatics, and functional syntax.