Graphic Display of Linguistic Information in English as a Foreign Language Reading

Graphic Display of LinguisticInformation in English as a Foreign Two studies investigated the advantage and instructional effectiveness of the spatial graphic representation of an Englis

Graphic Display of Linguistic

Information in English as a Foreign

Two studies investigated the advantage and instructional effectiveness

of the spatial graphic representation of an English sentence with dinators over a linear sentential representation in English as a foreignlanguage (EFL) reading settings Experiment 1, Study 1, examinedwhether readers studying EFL could better comprehend the sentence—

coor-in which multiple coor-information items are connected by coordcoor-inatcoor-ingconjunctions—when provided with a spatial representation than whenprovided with a sentential representation Experiment 2 of Study 1examined whether the advantage observed in Experiment 1 was due to

Larkin and Simon’s (1987) computational efficacy by testing whether two

different task-completion times would affect the performance of twodifferent display groups Study 2 examined the effectiveness of the in-struction manual, which we compiled, in enabling students to rear-range a linear text with coordinating conjunctions into a spatial displayfor self-study The results indicated that the spatial graphic display en-hanced EFL readers’ comprehension of sentences with coordinatorsmore than the sentential display did, and EFL readers were able toaccurately rearrange a linear sentential text into a spatial display byusing the instruction manual

The methods of displaying information can be distinguished based onhow the configured information is presented and can be broadlydivided into two forms: linear sentential representations and spatialgraphic representations A number of researchers (e.g., Holley &Dansereau, 1984; Larkin & Simon, 1987; Levelt, 1981; Moxley, 1983;

Young & O’Shea, 1981) have discussed the nature of these two methods

of representation

In linear sentential representations, information is displayed in the form

of a sequence, like the propositions in a text (Larkin & Simon, 1987).Examples of linear sentential representations are the English sentencesthat appear in newspapers or books, in which English words are dis-played from left to right Holley and Dansereau (1984) posited threefeatures of linear sentential representations: (a) A linear sentence is aserial expression of a multidimensional knowledge structure (b) It ex-plicitly expresses only a small part of the information the author intends

to present so that much information remains implicit and has to beunderstood by the reader through reasoning (c) In order to project amultidimensional structure (intended information) into a one-dimensional sequence (sentential representations), the authors are re-quired to repeat the nodes of the structure

On the other hand, in a spatial graphic display, the expression of

in-formation corresponds to the components of intended inin-formation sothat each expression includes the information stored in one particularplace in the spatial graphic display; this place contains information aboutrelationships with the adjacent loci (Larkin & Simon, 1987) The infor-mation or knowledge structure that authors intend to express can bedirectly represented in a spatial graphic display A matrix, in which wordsare freely displayed along with ruled lines, is an example of a spatialgraphic display

A number of empirical studies have been conducted to compare tial and sentential representations and have reported the advantage thatspatial representations have over sentential representations in enhanc-ing discourse comprehension (Levie & Lentz, 1982; Robinson & Kiewra,1995; Robinson & Schraw, 1994; Robinson & Skinner, 1996; Waller &Whalley, 1987; Winn, Li, & Schill, 1991)

spa-In explaining why a spatial representation more easily transmits formation than a linear sentential representation does, Waller (1981)

in-proposed the notion of visual argument—the ability to facilitate the

con-struction of conceptual connections during the act of reading Visualargument involves conveying ideas through a spatial graphic display—avisual and spatial arrangement of information—rather than throughwritten linear text Waller (1981) noted that since readers can see—notread—ideas, they are relieved from the burden of interpreting complexrelations described in the linear cobwebs of text Spatial graphic displayssuch as matrices use visual argument to communicate hierarchical rela-tions by placing coordinate concepts below a superordinate one Thisspatial configuration enables learners to make connections more easilythan plain text does On the other hand, by displaying rather than de-scribing the organization or structure of concepts, visual argument al-

lows readers to comprehend the intended message with less effort(Winn, 1990; Winn & Holliday, 1982).

Because of this visual argument, it is expected that readers providedwith a spatial display will process information faster and more effectivelythan those provided with a sentential display With regard to readers’efficiency in processing information, Larkin and Simon (1987) proposed

the notion of computational efficacy Based on the manner in which

read-ers process one-dimensional and two-dimensional displays, they believethat spatial graphic representations can convey conceptual relationshipsfaster and more easily than can linear sentential texts When information

is presented in one-dimensional displays such as texts, readers are quired to perform a linear search When the first relevant element con-cerning a particular concept is found and processed, it has to be stored

re-in the memory before the reader can proceed to search for the next

relevant element The routine of find, process, and remember must continue

in the reader’s working memory until the last element is found andprocessed However, in the case of two-dimensional displays such asdiagrams, once a relevant element is found, the next element will appearnext to or near it This advantageous feature reduces the extent ofsearching required as well as the burden on the working memory be-cause both elements concerning the concept can be viewed simulta-neously in the same place, and, thus, the relationship is representedclearly

Thus far, we have discussed the differences between sentential andspatial representations, the effectiveness of spatial representation in fa-cilitating information processing, and the reason that spatial represen-tations are effective In sum, when reading a linear sentential text, read-ers are required to reason, which implies a successive stepwise combina-tion of small sets of concepts and a rearrangement of these combinedconcepts in the working memory When reasoning, readers are required

to remember the previous steps; thus, the cognitive resources of theirworking memory space can be easily exhausted (Holley & Dansereau,1984; Robinson & Kiewra, 1995) On the other hand, a spatial graphicdisplay presents the most recent concept and the relevant previous con-cepts simultaneously in one or an adjacent location (visual argument).Thus, readers can reason by easily searching and computing the relevantinformation (computational efficacy) Consequently, this spatial displayallows readers to use fewer cognitive resources of their working memo-ries

Almost all these previous studies concerning spatial representationshave examined first language (L1) reading In second language (L2)reading, the cognitive burden on the working memories of L2 readerswho are required to comprehend the text is expected to be greater thanthe cognitive burden on the working memories of L1 readers (Berquist,

1997) In fact, previous studies have reported many differences betweenreading in L1 and L2 with respect to the cognitive burden of the twomodes Some examples of these differences are the problem of lexicalaccess (e.g., Segalowitz & Hebert, 1990); the knowledge of grammar(e.g., Everson & Ke, 1997; Horiba, 1990, 1996; Urquhart & Weir, 1998);the orthographic depth hypothesis (e.g., Koda, 1999); the languagethreshold (e.g., Alderson, 2000); the difficulty L2 readers have in con-structing a situation model (e.g., Zwaan & Brown, 1996); the differentstrategies that L1 readers and L2 readers use to pay attention to texts(e.g., Bernhardt, 1986); and inefficient use of the working memoriespertaining to L2 reading, which is suggested by using the reading spantest as a measure (Harrington, 1991) The general consensus is thatreading texts in L2 places considerable strain on readers, which canprevent them from achieving a completely effective understanding of thetexts (e.g., Miyake & Friedman, 1999; Tsurumi, 2005) Considering L2learners’ limited working memory capacity, it is important and necessary

to determine a way to reduce the cognitive burden on L2 learners whenthey read L2 texts or to create an instructional design to use their limitedworking memory capacity optimally

In this article, we focus on instructional design that is specificallyaimed at enhancing the processing of information items joined by co-

ordinating conjunctions in EFL reading Coordinators, or coordinating conjunctions, such as and, or, and but, combine linguistic units that are

equivalent or of the same rank (Richards & Schmidt, 2002) and are used

to construct coordinate structures—both phrases and clauses (Swan,1998) A coordinating conjunction plays the important role of present-ing information in EFL reading, because it is an indicator of additional,contrasting, alternative, and other such relationships (Swan, 1998) Fur-thermore, Biber, Johansson, Leech, Conrad, and Finegan (1999) arguethat coordinating conjunctions can add one clausal unit to another andthus extend the body of a single sentence In sentences written in En-glish, multiple information units are connected and represented by acoordinating conjunction, thus inevitably lengthening the sentence Be-cause of the limitation of the L2 readers’ working memory capacity(Berquist, 1997), L2 readers would have difficulty fully comprehendingthese lengthened, complicated sentences in which coordinating con-junctions connect many information components

It appears that L2 readers would find it difficult to comprehend asentence containing coordinating conjunctions; for instance, “Two of-ficers had been secretly dismissed from the Kanagawa police force, one

in 1997 and one last year, for allegedly molesting a woman on a train andfor shoplifting, respectively” (“The Price of Police Arrogance,” 1999).This sentence was taken from an actual newspaper article and is difficult

to comprehend because the coordinating conjunctions in this sentence

violate the minimal distance principle, which states that a segment of

lin-guistic units (in this case, “one last year”) must be located next to or at

a minimum distance from the relevant segment of linguistic units (in thiscase, “for shoplifting”) (Richards & Schmidt, 2002) For L2 readers tocomprehend this sentence, especially the latter half, they must (a) knowthat two officers have been dismissed; (b) understand that they weredismissed in different years (one in 1997; the other, last year); (c) reasonthat the first officer was dismissed for molesting a woman and the secondfor shoplifting; (d) when they see the word “respectively,” realize that thetwo officers have been dismissed in different years and for differentreasons that correspond to each other in some way Because the infor-mation items stating when (1997) and why (for molesting a woman) thefirst officer was dismissed are not adjacent in this sentence, and theinformation items about when (last year) and why (shoplifting) the sec-ond officer was dismissed are not adjacent, some readers can get con-fused and thus fail to comprehend these implicit relationships It isexpected that the processing and reasoning of information items that arelocated apart, as in this sentence, places a considerable cognitive burden

on readers, especially L2 readers

With respect to this difficulty, we argue that we can use a spatialgraphic representation of a sentence with coordinating conjunctions,following the spatial representation format suggested by Suzuki (1996).When we use this format, the spatial representation of the sentenceabout the police officers would appear as shown in Figure 1

By reading the information in a spatial graphic representation (seeFigure 1), readers can more easily conceive the interconceptual rela-tions, which were only implicit in a linear sentential representation,using minimal cognitive resources of their working memories This ad-vantage of representing conceptual relationships in the form of a spatialgraphic display can be explained in terms of visual argument In the case

of the sample sentence, the phrases one in 1997 and one last year are located close to each other vertically; similarly, the phrases for allegedly molesting a woman on a train and for shoplifting are also adjacent, one below

the other in a parallel formation Moreover, readers can easily observe

the relationships between the phrases one in 1997 and for molesting a woman on a train as well as between one last year and for shoplifting because

they are also adjacent—one phrase horizontally contiguous to the other.This visual argument (Waller, 1981) of a spatial representation can lead

to computational efficacy (Larkin & Simon, 1987) when L2 readers process

information, which is expected to result in better comprehension ofsentences

Based on the principle of computational efficacy (Larkin & Simon,1987) brought about by using visual argument (Waller, 1981), we predictthat in comprehending sentences in which multiple information items

are joined by coordinating conjunctions, L2 readers provided with tial representations can outperform those provided with linear sententialrepresentations.

spa-To be more specific, the first research question is, on the basis of theadvantage offered by Larkin and Simon’s (1987) computational effi-ciency, can EFL readers better comprehend a sentence in which multipleinformation items are connected by a coordinating conjunction whenprovided with a spatial representation than when provided with a sen-tential representation? This question will be resolved in Study 1 with thehelp of two experiments

In actual EFL reading settings, students are unable to optimize theadvantage of viewing spatial representations of English sentences withcoordinating conjunctions because English sentences with coordinatingconjunctions are normally written in the form of a linear sententialdisplay One solution for this problem is to help the students to rear-range a linear display into a spatial display on their own Moreover, thefirst step of the solution is to compile an instruction manual with guide-lines that would enable the students to independently rearrange thelinear displays into spatial graphic displays

Therefore, the second research question is, are students with an struction manual that provides guidelines on rearranging a linear sen-

in-FIGURE 1

A Linear Sentential Representation and a Spatial Graphic Representation

tential text with coordinating conjunctions into a spatial graphic displaymore successful in implementing the rearrangement independentlythan those without the manual? This question will be answered inStudy 2.

STUDY 1

Study 1 aimed to investigate whether, in EFL reading, a spatial sentation of information that is connected by a coordinating conjunc-tion facilitates better sentence comprehension among students because

repre-of their computational efficacy (Larkin & Simon, 1987)

Experiment 1

Experiment 1 examined whether EFL readers better comprehend asentence in which multiple information components are connected by acoordinating conjunction when provided with a spatial representationthan when provided with a sentential representation

Method

Design and participants Display type (a sentential display or a spatial

display) was a between-subjects factor The dependent variable was thenumber of correct answers to the multiple-choice questions based onfacts that were explicitly stated in four English sentences as materials

(hereafter, this test is referred to as the fact test).

Participants were 56 first-year undergraduates majoring in ceutical science at a Japanese private university (17 males, 39 females).All the participants were given a bookstore voucher (equivalent to 1,000yen) for their participation Based on the results of a general Englishgrammar and vocabulary test, they were assigned to one of two groups(the text group or the spatial group)

pharma-A t test was conducted for the English grammar and vocabulary test

(out of 30 points); however, it did not yield a significant difference in test

scores between the two groups (t (54) = 0.77, p = 0.22) It was confirmed

that the group members were not significantly different in terms of theirEnglish grammar and vocabulary abilities This general English grammarand vocabulary test was taken from the grammar and vocabulary section

of the English Language Proficiency Assessment (ELPA; Association forEnglish Language Proficiency Assessment, 2004) The ELPA is used tomeasure the general English language ability of Japanese students from

the first year of junior high school to the third year of senior high schooland is widely used as a placement test in a number of Japanese univer-sities.

The text group (28 participants) was provided with four English tences, all of which included one or more coordinating conjunctions in

sen-a linesen-ar sententisen-al representsen-ation, sen-and the spsen-atisen-al grsen-aphic displsen-ay group(28 participants) was given the same sentences represented in a spatialdisplay Testing was conducted twice in a typical university classroomwith mixed conditions

Materials We constructed two different types of displays (a linear

sen-tential representation and a spatial graphic representation depending

on the condition) for each of the four sentences, which were taken fromnewspapers or exercise books pertaining to the TOEFL CBT tests Thefollowing four sentences were used as materials in Study 1.1

1 The Earth’s atmosphere consists of water vapor and a mixture ofoxygen, nitrogen, and other gases

2 For many years linguists as well as those who have studied differentcultures have wondered about this possibility

3 Two officers had been secretly dismissed from the Kanagawa policeforce, one in 1997 and one last year, for allegedly molesting a woman

on a train and for shoplifting, respectively

4 The motive for having a Christian wedding is that everyone else isdoing it, or that it’s cheaper than a Shinto wedding, or that it justlooks chic and modern

Figure 1 illustrates an example of a graphically displayed English tence used in the study To eliminate the differences between the stu-dents’ English word knowledge, some words were translated into Japa-nese (up to 6 words for each sentence), which had been selected on thebasis of pilot tests conducted before this experiment, and were glossedwith all four English sentences in Study 1 The test paper was printed ontwo B4 pages For the spatial graphic display group, all linear sententialtexts were rearranged into spatial graphic displays by Suzuki The rear-rangement was based on the principle of visual argument (Waller, 1981):Two-dimensional displays of relevant information were presented simul-taneously in the same place to draw out the latent potential of compu-tational efficacy (Larkin & Simon, 1987), specifically following Suzuki’s(1996) format When observing the linear sentential representation inthe upper part of Figure 1, we see two lines—one connecting the words

sen-1 Sentence 1 is from Suzuki (2001); Sentences 2 and 3 are from Iwamura, Takahashi, & Braven (2000); and Sentence 4 is from “The price of police arrogance” (1999).

one and one and the other connecting for and for These lines, which

indicate what informational elements are connected by each ing conjunction, were included to ensure that the two types of displaysare informationally equivalent (Larkin & Simon, 1987), because twolines have also been used in the spatial graphic representation in thelower part of Figure 1 The number of words used in both representa-tions was also exactly the same to ensure informational equivalenceacross all three experiments In the fact test, we constructed four mul-tiple-choice questions in Japanese, which tested whether the participantscould understand the sentence correctly The following is an example ofone of the questions from the fact test (translated from the Japanese bySuzuki):

coordinat-Question: When and why were the two officers dismissed by the Kanagawapolice? Choose the correct answer from among the following four op-tions

1 One was dismissed last year for molesting a woman; the other, in

Procedure Sessions were conducted for both groups by the first author in

an approximately 40-minute period and were timed by a stopwatch First,all students took the English grammar and vocabulary test for 20 min-utes, and then the first author asked the students to exchange their testpapers with those of others and immediately mark them in accordancewith predetermined keys As noted earlier, the students were then as-signed to either the spatial graphic display group or the text group for abalanced distribution of students as regards test scores Just before thecommencement of the fact test, participants were briefly instructedabout the usage of coordinating conjunctions and informed that theyhad to comprehend the four short sentences well enough to select thecorrect answer from among the four multiple-choice items They werealso told that they could not use dictionaries This brief instruction lastedfor about 10 minutes After the test materials were distributed, studentswere asked to begin the fact test, which would examine how well theycomprehended the four English sentences; there was no time restrictionfor this test The average time taken by the participants to complete thetask was approximately 10 min Students could refer to the sentencesfreely when answering this fact test

Results and Discussion

Means and standard deviations of the correct answers for both tests

appear in Table 1 A t test was conducted for the fact test (out of 4 points), and it yielded a significant difference (t (54) = 3.54, p < 0.01).

Students comprehended the facts described in the four sentences better

when they were provided with a spatial graphic representation (M = 3.18,

SD = 0.76) than with a linear sentential representation (M = 2.32, SD =

1.00) Testing confirmed that the performance of the spatial graphicdisplay group was superior to that of the linear sentential group because

of the configuration of the information displayed, that is, the spatialrepresentation

Larkin and Simon (1987) argue that two representations are tionally equivalent when all the information in one can be inferred from

the other and vice versa; each could be constructed from the tion provided in the other In addition, with regard to computational

informa-equivalence, Larkin and Simon state that two representations are tationally equivalent if any inference can be drawn easily and quickly from

compu-the information given explicitly in eicompu-ther and if compu-they are informationallyequivalent From the perspective of informational equivalence and com-putational equivalence or difference, we can hypothesize that the resultsobtained in Experiment 1 can be attributed to the fact that the linearsentential texts and the spatial graphic representations are information-ally equivalent but computationally different because of a spatial graphicdisplay’s inherent capability: computational efficacy

Experiment 2

Experiment 2 was designed to test whether the results obtained inExperiment 1—in terms of comprehending sentences with coordinatingconjunctions in EFL reading, that spatial representations have an advan-tage over sentential representations—can be attributed to computational

TABLE 1

Means (M) and Standard Deviations (SD) of the Scores of the Correct Answers in the Fact

Test and the English Grammar and Vocabulary Test in Experiment 1, Study 1

efficacy (Larkin & Simon, 1987) In order to investigate computationalefficacy brought about by spatial representations, previous studies—such

as Robinson and Skinner (1996) or Winn, Li, and Shill (1991)—examined how quickly and accurately students could search for andprocess information contained in different types of displays: a linearsentential display or a spatial graphic display Following this minimizing-time paradigm, we hypothesize that students studying with spatialgraphic representations will outperform those studying with linear sen-tential representations when the students are given a minimal short studytime; however, this difference will disappear when they are given a suf-ficiently long study time

Method

Design and participants A 2 × 2 design was adopted in Experiment 2 The

first factor was display type (a spatial graphic representation or a linearsentential representation), and the second factor was study time The

participants in the short-time study groups were required to read one

sen-tence (either spatial or sentential text depending on the condition) and

answer the fact test in 120 s, and those in the long-time study groups were

required to do so in 300 s

Participants were 114 Japanese undergraduates of ages varying from

18 to 20 years (58 males, 56 females); they were given a bookstorevoucher (equivalent to 1,000 yen) for their participation These studentsbelonged to two different universities, and they had not participated inExperiment 1 They were assigned to one of the following four groupsbased on their results in the general English grammar and vocabularytest A one-way ANOVA was conducted for the English grammar andvocabulary test (out of 30 points); however, it did not yield a significant

difference in test scores among the four groups (F (3, 110) = 1.58, p =

0.19) These four groups comprised a spatial graphic display short-time(SS) group (25 participants), a linear sentential display short-time (LS)group (29 participants), a spatial graphic display long-time (SL) group(31 participants), and a linear sentential display long-time (LL) group(29 participants) The dependent variable was the same as that in Ex-periment 1—the number of correct answers to the fact test Testing wasconducted on four different occasions in a university computer roomwith mixed conditions Because this experiment was conducted fourtimes at two universities, the number of participants in each group ateach occasion of testing was not the same However, homogeneity interms of English grammar and vocabulary skills across the four groupswas ensured

We decided to set up two different study-time groups of 120 s and 300

s each, because in Experiment 1, the average amount of time taken by

the participants to complete the task was approximately 10 min Allowingfor pauses between sentences, this completion time implies that thestudents spent approximately 120 s reading a sentence and answeringthe question Accordingly, we hypothesized that 120 s would be thewatershed for examining the benefit of computational efficacy, and weset up the two groups that were given 300 s as study time, a durationregarded as sufficient for the students to devote enough time to readingthe sentence and answering the question.

Materials The study materials were the same as those used in

Experi-ment 1, as were the fact test and the English grammar and vocabularytest

Procedure Sessions were conducted by Suzuki and Sato in an

approxi-mately 40-min period for both short-time study groups (120 s) and in a50-min period for both long-time study groups (300 s) The Englishgrammar and vocabulary test was timed by a stopwatch, and the fact testwas timed by computer software (Microsoft PowerPoint) First, all stu-dents took the English grammar and vocabulary test (ELPA, 2004) for 20min The procedure of marking the test papers and organizing thegroups was the same as in Experiment 1 Based on the test scores, stu-dents were assigned to one of the four groups Just before the com-mencement of the reading and answering period, participants wereasked to turn on their computers and launch Microsoft PowerPoint.Then, 3.5-inch disks, which included the PowerPoint file (created bySuzuki) of the program for this experiment and answer sheets printed

on a single A4 page were distributed After this, students were requested

to open the file, execute the program, and wait for the instructions tobegin

The program comprised 13 slides, and this program led participantsthrough the entire reading and answering period, including the instruc-tions The first slide explained the purpose of this experiment in Japa-nese; this slide was the only one that the participants themselves could

control by pressing the Enter key The computer controlled the viewing

time for the rest of the slides automatically Slides 2 and 3 were inJapanese and provided general grammatical knowledge of coordinatingconjunctions for sentential groups and explained the principle of rear-ranging a sentential English sequence into a spatial representation forspatial groups (each slide was viewed for 120 s) Slides 4–6 presentedexamples of the usage of coordinating conjunctions (sentential repre-sentations or spatial representations depending on the condition); onlySlide 6 displayed an example sentence with a coordinating conjunctionconnecting more than two informational items (each slide for 30 s,except Slide 6, which was viewed for 60 s) In Slide 7 (for 60 s), five

instructions concerning Experiment 2 appeared In Slide 8, an example

of a sentence (either in a sentential representation or a spatial sentation depending on the condition) and a multiple-choice questionwith four options was shown (either for 120 s or 300 s depending on thecondition) From Slides 9 to 12, students took the fact test (either for 120

repre-s or 300 repre-s depending on the condition) Slide 13 marked the conclurepre-sion

of this session After the test, the answer sheets were collected

Results and Discussion

Means and standard deviations for the scores of the correct answersfor both tests appear in Table 2 As for the fact test (out of 4 points), a

2 × 2 (representation type × study time) ANOVA yielded a significant main effect of the representation type (F (1, 110) = 7.54, p < 0.01) The repre- sentation type × study time interaction effect was also significant (F (1, 110) = 7.85, p < 0.01) A posthoc simple effect test indicated that in the fact test, students in the SS group (M = 3.52, SD = 0.50) outperformed those in the

LS group (M = 2.66, SD = 0.99), but the scores of the students in the SL group (M = 3.13, SD = 0.91) were not significantly different from those

in the LL group (M = 3.14, SD = 0.73) This interaction, graphically

displayed in Figure 2, reveals that the type of display (spatial or linear)used to present the information significantly affected comprehension ofthe sentence in the short-time (120 s) condition but not in the long-time(300 s) condition

Experiment 2 was conducted to test whether the results obtained inExperiment 1 can be attributed to computational efficacy (Larkin &Simon, 1987), following the minimizing-time paradigm The results ob-

TABLE 2

Means (M) and Standard Deviations (SD) of the Scores of the Correct Answers in the Fact

Test and the English Grammar and Vocabulary Test in Experiment 2, Study 1

Short time group (120s)

Note For the fact and the English tests, n = 25 in the spatial display short-time (120s) group,

n = 29 in the sentential display short-time group, n = 31 in the spatial display long-time (300s) group, and n = 29 in the sentential display long time group.