Unidimensionality Reading AbMty Working Memory

14
Kantokoshinetsu Association of Teacher of English, KATE NII-Electronic Library Service KantokoshinetsuAssociation of Teacher of English, KATE Unidimensionality of Reading AbMty and Working Memory Capacity: Implications forL2 Reading Tests SHIZUKA [rletsuhito eukushima National Conege of Technology> ABSTRAcr This paper reviews the literature and integrates the findings in divisibilityfidentifiability of L1 and L2 reading skills, and presents a hypothesis cenceTning L2 reading tests. Whether reading ability can be brokendow'n into component skiils isan issue which has long divided reading researchers. Review of the literature reveals what can be almost termed a chaos of results. However, ifwe distinguish between Psycholag:'cal and Psyehemetrie unidimensionality, the apparcntly conflicting results could converge nicely into a clear picture. That is, reading ability could be 'considered PsychalogicxzUy rnultidimensiQnal but rwhamstricxiU7 unidimensional. It may follow that at least fbrmeasurernent purposes, itisnot too umeasonable to treatreading as a unidimensional traiL What underlies thisvinual unidimensionality of reading abiliry rnight be thefact that different levels of reading processes draw on the same cognitive resource, namely, working memory capacity. The capacity theory of comprehension proposed by Justand Carpenter (1992) provides a plausible account forthis unidimensionality in the psychometric sense. Since one's working memory capaciry is a function of hislher lower-order processing eMciency intheir ffarnework, one hypothesis for L2 reading tests would be that for the purpose ef revealing individual diffeiences in overall readingability, we may well coneenttateon measuring the Teader's lower-order precessing ethciency. 1.INTRODUCTION The current practice among English teachers,materials writers, and testconstructors of talking about, tryingto develop, and attempting to measure `reading ski11s' separately--such as the ski11 of understanding the details of thecontent accurately, as opposed to that of grasping the gist of a discourse, etc.--is implicitly based on theview that there exist separately teachable and testable reading ski11s, i.e. the view that reading isa multi-divisible trait. However, if thisisnot thecase, namely. if reading can not be 1rroken down into separable ski11s, then such a practice would not be -65-

Transcript of Unidimensionality Reading AbMty Working Memory

Page 1: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation of Teacher of English, KATE

Unidimensionality of Reading AbMty and Working Memory

Capacity: Implications for L2 Reading Tests

SHIZUKA [rletsuhito

eukushima National Conege of Technology>

ABSTRAcr

This paper reviews the literature and integrates the findings in divisibilityfidentifiability of L1

and L2 reading skills, and presents a hypothesis cenceTning L2 reading tests. Whether reading

ability can be broken dow'n into component skiils is an issue which has long divided reading

researchers. Review of the literature reveals what can be almost termed a chaos of results.

However, if we distinguish between Psycholag:'cal and Psyehemetrie unidimensionality, the

apparcntly conflicting results could converge nicely into a clear picture. That is, reading ability

could be 'considered

PsychalogicxzUy rnultidimensiQnal but rwhamstricxiU7 unidimensional. It may

follow that at least fbr measurernent purposes, it is not too umeasonable to treat reading as a

unidimensional traiL

What underlies this vinual unidimensionality of reading abiliry rnight be the fact that different

levels of reading processes draw on the same cognitive resource, namely, working memory

capacity. The capacity theory of comprehension proposed by Just and Carpenter (1992) provides a

plausible account for this unidimensionality in the psychometric sense. Since one's working

memory capaciry is a function of hislher lower-order processing eMciency in their ffarnework, one

hypothesis for L2 reading tests would be that for the purpose ef revealing individual diffeiences in

overall reading ability, we may well coneenttate on measuring the Teader's lower-order precessing

ethciency.

1. INTRODUCTION

The current practice among English teachers, materials writers, and test constructors of talking

about, trying to develop, and attempting to measure `reading

ski11s' separately--such as the ski11

of understanding the details of the content accurately, as opposed to that of grasping the gist of a

discourse, etc.--is implicitly based on the view that there exist separately teachable and testable

reading ski11s, i.e. the view that reading is a multi-divisible trait. However, if this is not the case,

namely. if reading can not be 1rroken down into separable ski11s, then such a practice would not be

-65-

Page 2: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation of Teacher of English, KATE

justifiabte, This is an issue which has long puzzled and continues to puzzle reading researchers,

Review of the literatuie be!ow seems to reyeal alrnost uninterpretably confusing results.

2. Ll FACTOR ANALYTIC STVDIES

One of the first factor analytic study of Ll reading comprehension was conducted by Davis

(1941, reported in Davis 1968). Analyzing the scores of 9 subtests, he conclUded that 5 fbetors

were distinguishable, but that word knowledge and reasoning in reading accounted for virtually al1

of the variance of comprehension. However, Thurstone (1946) reanalyzed Davis' (1941) data and

concluded that one cornmon factor adequately accounted for the data.

Denick (1953) attempted but fai1ed to discern (a) the ability to answer fat tual questions, (b) theability to read-between-the-lines, and (c) the ahility to evaluate logical soundness and rhetcxrical

effectiveness, ffom the test scores of 457 fieshmen at Chicago Junior College.

Hunt (1957, reported in Davis 1968) reported that, in his test, only the vocabulary items were

measuring a skill in comprehension that was significantly different ffom the other items. This

implied that reading comprehension involved two ski!ls: word knewledge and paragraphcotnprehension.

Davis' (1968) claimed that he had succeeded in distinguishing the fo11owing five factors frdm

his data of 988 twelfth graders: (1) recalling word meanings, (2) drawing inferences from theconteng (3) finding answers to questions asked explicitly or merely in pasaphrase in the content, (4)recognizing a writer's purpose, atritude, tone, and rnood, and (5) fo11owing the structure of a

passage.

This finding by Davis was challenged by 'Miorndike

(1971), who presented the results of a

components analysis of Davis' (1968) data. He showed that the non-chance variance in the sets of

eight tests could be cempletely accounted for by threc factors, the first factor accounting fOr more

than 93 percent of the variance. Ms paper emphasized the importance of the first factor in tests of

reading cornprehension, which he believed to be largely a measure of reasoning.

A stilt further refbotorization of Davis' data was made by Spearritt (1972). Tbe resuks indicated

that four skills of (1) recalling word meanings, (2) drawing inferences from the content, (3)recognizing a wTiter's purpose, auitude, tone, and mood, (4) following the structure of a passago,were separately distinguishable skilis. However, Spearritt stressed that `although

certain

comprehension subski11s can be differentiated, present types of reading comprehension tests, . .

1argely measure one dasic ability' (p. 1 10).

Tlie confusion around the interptetation of Dayis' data seems to have been somewhat mediated

by Andrich & Godftey (1978-9), who did not rely on factor analysis. Applying Rasch analysis to

the scores of 188 9th to 13th graders in Austraiia, they concluded:

on the one hand, at what might be termed a inacro-level analysis, Davis' test provides

-66-

Page 3: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation of Teacher of English, KATE

a measure ef a unitary trait On the other hand, at what might be terrned a micrv-level

analysis, it appears to measure 4 major ski11s. (p.199)

Lunzer et al. (1979) attempted to devise reading comprehension tests aimed at assessing primary

school children's abilities to understand text at different levels ef comprehension. They

administered four separate tests designed to measure the foliowing separate skills: (1) word

meaning (2) words in context (3) literal comptehension (4) drawing inferences from single strings

(5) drawing inferences from multiple snings (6) interpretation of metapher (7) finding salient ormain ideas, and (8) formingjudgements. However, their fhctor analysis fai1ed to find evidence for

the separabildy of such skills, the single fhctor accounting for 81 percent of the tota1 varianoe. Ihis

1ed them to conclude that `one must reject the hypothesis that the several tasks llsed in tests of

reading comprehension call on distinct ski11s which can he differentially assessect and taught' (p.S9) and that

`individnal differences in reading...reflect only one general aptitude: this being the

pupil's abiliry and willingriess to reflect on whatever it is he is reading' (p. 64).

Zwick (1987) lent another suppert for single trait hypothesis. She assessed the dimensionality of

reading data collected in the 1983-84 NAEP (National Assessment of Educatiet)al Rogtess) survey

in the United States and concluded that `overall,.it is not unreasonable to treat the data as

ilnidimensional.'(p.306).

Carver (1992) investigated the fhctors measured by several standardized reading comprehension

tests adrninistered to over 3oo students in (hades 3-8 and to 64 college students. Four different

factor analyses of these data consistently res"lted in a single fhctor - which Carver interpreted as

Efficiency Level fhctor - in case of elementary and junior high school subjects, and in two factors -

which he temied Accuracy Level factor and Rate Level factor - when the subjects were college

students. Carver has concluded that the most important fat tor involved in tests that measure reading

cornprehension is efficiency level, which includes the ability to read fdst.

Rost (1993) analyut reading compsehension scores of 220 second graders who were learning to

read. Factor analyses yielded either one hroad fhctor, `general

reading comprehension' accounting

for 8S% of the nonchance variance, or, at most two fbctors `vocahulary'

accounting rotated for

55% of the reliable variance and `inferential

reading' accounting fbr 36%. Based on these findings,

Rost (1993: 89) concluded that `(a)dministering

several tests (or subtests) fot the assessment of

reading cemprehension subskills may be a waste of time.'

3. L2 EACTOR ANALYTIC STUDIES

In 1970s, much research was done to test John Oller's `Unitary

Competence Hypothesis'(UCH),

i.e., a hypothesis that language ability could not be broken down into such component parts as

speaking ability, reading ability, ete. Although the hypothesis was eyentually shown to be false

(Oller 1983; Hughes & Porter 1983), the rejoction of the hypothesis that 1anguage is a unitary ski11

would not necessarily imply the Tejection of the hypothesis that reading is a unitary ski11, which

-67-

Page 4: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation of Teacher of English, KATE

concerns us in the psesent paper. [hus, from three studies (Hinofotis 1976, Anderson 1976, Oller &

Perkins 1978) on UCH, those parts related to reading ski11s are reviewed below.

Hinofods (1976, reported in Oller 1979) investigated the pattern of relationships among the

various parts of the placement test used at Center fbr English as a Second Language at Carbondale,

nlinois, the TOEFL, the five subscales on the Foteign Service Institute Oral Interview, plus a cloze

test, Part of the results indicated no eyidence fbr separable components in reading. Similar result

was obtained by Anders`?n (1976, reported in Oller 1979).

Oller (1979) reports an earlier study (Oller & Perkins 1978), in which three kmds of reading test

data were factor analyzed to a principal components solution and then to a varimax solution.

wneTeas the loadings on the general fhctor accounted for 47% of the total avai1able variance, the

interpreted leadings on the four fhctor rotated solution only accounted for an additional 6% (for atota1 of 53%). [his indicated that the indivisibility bypothesis was to be preferred.

Jafarpur (1987) developed a short-context test (SCI) and investigated if it measured different

traits than other traditional reading tests. The data 1ed Jafarpur to adqpt threc ftctor solutions and to

conclude that SCT is factorially different fivm the cloze procedure or traditional reading tests.

However, this interpretation by Jafarpur could be open to question since the first fat tor accounts ftrr

87% of the total variance in the tests, the second for 8.9% and the third for 4.2%. Exarnining

Jafarpur's data, Hateh & Lazaraton (1991: 495-496) argries that `(firom

the eigenvalues, it appears

that the tests are unidimensional'.

Negishi (1996) examined tbe perfbrmarrce by a t(rtal of 460 EFL students in Japan on the tests of

reading in the roEFL and the Cambridge FCE test. Using recently developed technique of fu11

information ftctor analysis (Bock, Gibbons, & Muraki 1985), he succeeded in extracting several

discemible fhctors from his data. However, he concludes that `(i)n any event, the number of rt ading

traits that the current L2 rertding tests are able to discern is much more resnicted than traditionallists of L2 reacling ski11s or L2 reading test spocifications suggest' (p.139).

4. L2 IDENTIFIABILITY STUDIES

The research reyiewed so fhr has been concemed whether statistically distinguishable ski11s

could be extracted ffom reading test scores. There has also been a somewhat different strand of

research which could be terrned `identifiability

studies'. These studies have mainly investigated to

what extent we can agree tihat a test item is mcasuring the specific ski11(s) that it claims to measure.

In their review of recent L2 reading research, Williams & Moran (1989: 223) note that although

whether reading is a single holistic prvcess or whether it consists of a number of component ski11s

remains unresolved, materials abound which claim to promote different reading skills and

strategies. One study that seems to have promoted such an approach is the work of John Munby

(1978). A!though his extensive list of language `microskills'

has been strongly criticized for being

subjective (Davies 1981; Mead 1982), its impact on subsequent syllabus design, materials writing,

.asm

Page 5: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation of Teacher of English, KATE

and test construction has been considerable (e.g., Grellet, 1981; Yalden, 1987; Heaton 1988),

It was Alderson & Lukmani (1989) who triggeTed the debate over the validlry of subski11s

taxonomy, such as Munby's, in relation to L2 testing, In their study, teachers at Lancaster

University were presented with a reading test used at Bombay University and re(luired to rate the

items as tests of 'lower',

`middle'

or `higher'

order abilities. In 27 out of 41 items, there was little

agreernent on the levels being tested. in addition, for approximately the same number of iterns .

judges disagreed considerably over the ski11s being tested. Alderson & Lukrnani argued that `one

must surely begin to question whether definably separate or different levels do exist' (p.264).

Alderson (1990a) was an attempt to replicate Alderson & Lukmani (1989) with a mere carefu11y

researched test. Again, in most cases there was little agreement between judges. From the results of

these two stildies, Nderson (1990a: 436) has concluded that `(e)ven

if tbere are separable ski11s in

the reading process which one could identify by a rational process of analysis of one's own reading

behavior, it appears to be extrernely dirncult if not impossibie to isolate them for the sake of testing

or research'.

Alderson's (1990a) above claims were strengly eriticized by Weir, Hughes, & Porter (1990).T hcy pointed out that Alderson's methodology was defective in that no evidence was given that his

judges had had any experience in the task they had to do, they had been given no training, and the

terrns `High'

and `Low'

had been insufficiently defirred.

Lumley (1993) made ari attempt to sort out this dispute by using a much more carefu11y defined

skitl taxenomy and trained judges. He investigated the level of agreernent shown by a gtvup of five

experienced ESL teachers on the identification of selected reading ski11s with particular iteTris in a

reading test. Contrary to Al(lerson & Lukmani (1989) and Alderson (1990a), fo11owing diseussion

and definition of terms,. almost complete agreement was reached between the raters for ail items.

Thus it seems to haye been shown thatjudges are able to agree on what an item is testing if they

are trained using a well-defined taxonomy. Nevemhcless, the implications of this finding are not

very clear; the fact that judges can agree what an item is testing may not guaraiitee that the item

does test the ski11 when actually attempted by an examinee, It might be even possible that what

skill(s) an itern elicits varies from one test taker from another (Alderson & Lukmani 1989;

Alderson 1990b). As a matter of fhct, Negishi (1996) has shown that although we can agree, to

some extent on what a reading test item is testing, based on test specifications, the trait(s) that will

actually be tapped by a reading test item cannot be predicted from test specifications.

5. INTEGRATION OF TIIE FINDINGS

Having reviewed quantitative and qualitative studies in L1 and L2 on reading ski11s divisibility,

it seems in order at this point to attempt an integration of the findings. The fo11owing

summarization seems possible:

(1) Three or more ski11s could be distinguished occasionally (Davis, 1941, ] 968; Spearritt 1972;

-69-

Page 6: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation ofTeacher of English, KATE

Andrich and Godftey 1978-9; Negishi 1996), but

(2) Most of the variance could be accounted for by word knewledge and reasoning (Davis 1941,

1968; Hunt 1957; Spearritt 1972 , Rost 1993).

(3) Often Qne generul factor emerged to the extent that reading seemed a unitary skill (Thurstone

1946; Denick 1953; Tliorndike 1971; Andrich and Godfrey 1978-9; Lunzer 1979; Zwick

1987; Carver 1992; Rost 1993; Hinofotis 1976; Anderso: 1976; Oller & Perkins 1978;

Jafarpur 1987),

This sumarization, which at first sight rniglrt appear to be nothing more thafi a mixed bag,

could be rephrased as that a ski11 like reading is multidimensional and unldimensional at the same

time. How can this be?

A test is cal1ed unidimensional when `only

one ability or trait is neoessary to "explain" or

"account!'

for examinee perfbrmance' (Harnbleton & Swaminathan 19g5: 16). This does not imply

that performance on the items is due to a single psychological psocess. As long as al1 the iterns

function in unison - that is, the perfbrmance on each itern is affected by the same processes and in

the sarne form - unidimensionality will hold (Tl]{midike 1982; Bejar 1983, both cited in Baker

1989).

However, as Baker (1989) notes, the concept of unidimensionality has tended to be confused

with psychological simplicity (e.g., Canale 1986). With the intent of solving such

rnisunderstandings, Henning (1992) has suggested distinguishing between Psycholagt'catunidimensionality and Psycbot)retrib unidimensionaliry. According to Henning,

`(p)sychological

unidimensionality in a test impiies that the test scores are intended to be interpreted as refiective of

the extent of the presence of some known unitary psyeboIogical construct or trait' (p. 2), wlrereas`psychometric

unidimensionality can be present when the test measures a vaTiety of correlated

underlying psychological dmenslons ( p.3). McNamara (1991; 1996) rnakes the same distinction.

This distinction seems suitable when we attempt to interpret the above `mixed

bag'

summarization, There is no doubt whatsoever that reading is a complex ski11 consisting of multiple

cognitive activities, e,g., word Tecognition, lexical access, parsing, proposition construction,

discourse-level integration, etc, In this psychological sense, reading skill should he considerd multi-

dimensional arKl rnulti-divisible, But at the same time, it seems also true that those various skills in

rmy cases tend to act in unisop, as if one unitary ski11 to the extent that they form a virtually single

measurement (limension. Thus in this psychomeuic sense, reading tends to be unidimensional and .unltary.

[lhere are two irnportarit features of unidimensionality in the psychomeuic sense that should be

pointed out here. One is that it is a sample-dependent concept (Hambleton & Swaminathan 1985;

Henning 1992). Thus, although in most of samples researched to date psychometricunidirnensionality has been observed, it should be possible to find a sarnple in which reading is

psychometrically multi-dimertsional. The other characteristics is that unidimensionality is not a

catesorical concept but should rather be considered as one end of a continuum (Henning 1992;

-70-

Page 7: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation of Teacher of English, KATE

Chei & Bachnan 1992; Blais & Laurier 1995) Thus one may talk about one test being more

unidirnensional than another or one sample, e.g. adult native speakers, good readers, etc., showing

more unidimensional perfOrTmance than another sarnple, e.g., non-native speakers, poor readers,

children, ete.

The view that reading is a psychologically multidimensional but psychometricallyunidimensional ski11 seems consistent with observadons by cognitive psycbologists. Downing &Leong (l982:t4-18) list 21 major generalizations about the acquisition of skills, the first two of

which are:

1. The perfbrTnance o £ any skill involves a highly cornplex pattern of behavior,

2, A ski11ed perfbrmer executes this complex pattern smoothly, without faltering.

As a conclusion, it seems that reading, which psychologically could be broken down intocomponcnts, tends to be psychornenically a unitary process. So for measurement purposes, it would

not seem too unreasonable to treat it as a single construct.

6. WORKING MEMORY CAPACITY

We have noted that reading ski11s appear to `function

in unison' (Bejar 1983: 31). This

phenomenon could be well aocounted for if we assume a common cognitive factor that affects all

the levels of ski11s (Just & Carpenter 1992). A current the(rry of L1 reading seems to indicate that

this factor might be working mernory. Working mernory

'is an information processing construct

which has processing and storage functions that compete for a limited capacity (Baddeley & Hitch,

1974; Case, 1978; Daneman & Carpenter 1980; Danernan 1987; Rayner and Pol!atsek 1989; Just

and Carpenter 1992). It can be considered a flexible workspace whose limited capacity can be

allocated to either storage or processing.

It was Baddeley and Hiteh (Baddely and Hitch 1974; Hitch & Baddely 1976) who first proposedthe concept of working memory as an alternative to traditional short-term memory concept, which

had storage function only (Atkinson and Shifltin, 1968, 1971; Posner and Rossman, 1965; Waugh

and Norman, 1965, all reported in Daneman 1987). Building on their work, Daneman and

Carpenter (l980) (tevised a test with heavy processing and storage demands to measure the trade-

off between proeessing and storage functions of working memory. The working memory capacity

as measured in this test - reading span - correlated with threc reading comprehension measuros,

including verbal SM (lying between ,S and .6) and tests involving imt renieyal foetween .7 and .9)

and pronominal reference. Also, it was found that subjects with 1arger spans were better at

abstracting a theme ffom a written or spoken narrative passage. Encovraged by these findings, M,

A. Just, P. A. Carpenter and their co-workers have (teveloped a comprehensive theory of !eading

comprehension with working memory as one of its key concepts (Daneman & Carpenter 1980,

1983; Carpenter & Just 1986, 1989; Just & Carpenter 1980, 1987,1992; Daneman 1987; King &

-71-

Page 8: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation of Teacher of English, KATE

Jllst 1991, MacDonald, Just, & Carpenter 1992).

According to Just & Carpenter (1992), while working memory plays a central role in al1 forTn$

of complex thinking, its fimction in language eomprehension is especially evident. Cornprehension

entails ptocessing a sequence of symbols perceived and storing their partial and final products over

time. At lexical level, a reader must be ahle to quickly retrieye the eaclier representation of earlier

words and phraFes in a sentence to integrate tirem to later words and phrases. Storage demands also

occur at discourse-level. The reader must store the theme of the text, the propositions from

preceding sentenoes, cte.

Since storage and processing draw on the same capacity, they affect each othgr. In this

framework, `capacity' of working memory can be expressed as the maximum amount of activation

available to support either storage or processing. When the total amount of activation available to

the system is exceeded by the sum of activation needed for storage and that required for the

ongoing processing, both the activation used for maintenance and the activation allotted for

processing will be scaled back proportionately, Thus, the larger the capacity, the better the

comprehenslon task will be executed.

rt should be peinted out that in this context the capacity of working memory and processing

efficiency are. as it were, two sides of the same coin. By interfering with storage, the inefficient

processes of the poor reader will be functionally equiyalent to a smaller storage capacity (Danemanan(l Carpenter 1983: 562). In other words, indivldual differences reside in the trade-off between the

processing and storage functions, with the rnajer source of individual differenoes residing in the

processing component (Danernan 1987: 61). If prQcessing is more efficienL it wi11 take up less of

the cabacity, 1eaving mere resource for storage and for highcr order cornputation. If prooessing is

less ethcient, it will take llp more of the limited capacity, leaving less resource for storage and for

higher order computation. This means that the preceding relevant infbrmation cannot be retained in

working memory or that little resource can be allocated to higher order computation like

integration, or both, In either case, there would be no way successfu1 comprehension would take

place.

7. EMPIRICAL EVIDENCE

The above theory has considerable empirical support in case of Ll readers. It has been shown

that readers with larger working rnemory capacities can rnore readily recover from

misiinterpretations caused by `garden

path' sentences, i.e., sentences that are characteristically

misinterpreted on their first reading, either as a result of the syntactic structure or as a result of a

lexical ambiguiry (Daneman & Carpemer 1983). Larger capacity readers have also been shown to

better process complex syntactic structures (King & Just 199t) and arnbiguous syntactic structures

(MacDonald, Just & Carpenter 1992), make better use of pragrriatic cues on syntactic processing

(Just & Carpentcr 1992), than those with smaller capacities. At discourse level, 1arger capacity

readers can more successfully find an antecedent for a pronoun, better abstract a theme from a

-72-

Page 9: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation of Teacher of English, KATE

passage Oarieman & Carpenter l980), and more easily integrate infbrmatien over distances in atext (Yuill, Oakhill & Parkin, reported in Just & Carpenter 1992),

Differences in working rnemory capacity or processing efficiency are implied in several age-

related differences as well. The decline in language perfbrmance in the elderly Ll readers has been

shown yo fbcus on scntences whose syntax makes 1arge demands on working memory (Kemper1986). Comprehension differences between younger and okler children mainly seem to reside in

their understanding of implicit informadon rather than in explicit information (Paris & Upton 1976;

Pais & Lindauer 1976, reported in Mitchell 1982). Since developmenta1 research suggests that

children have much slower and less ethcient proeesses (Case, 1978; Chi, 1976; Huttenlocher &

Burke 1976, reported in Daneman & Carpenter l980), this finding could be interpreted as

indicating that ptocessing inefficiency may lead to less ability to work out implied meanings. It has

been also shown that young readers seem to put much less effort into the prooess of organizing and

linking the sentences in a text than adults do (Mitchel; 19g4), which could be taken to mean that

processing inethciency may also mean less ability to integtate information. Older adults (6S-79years of age) show relatively greater deficits than do younger aduks when they must make an

inference that requires integrating infoTmation across sentences (Cohen 1979). This result was not

interpreted as indicating a loss of inferential reasoning ahility per se, but as sbowing that it is a

vulnerable stage of comprehension which tends to drop out when total ptocessing demands exceed .capac1ty.

8. HYPOTHESIS CONCERNING L2 REAI)ING TESTS

Although the theory of capacity constrained comprehension was originally developed in Ll

reading,contexq what the theory predicts would happen and what evidence has shown does happen

with less efficient L1 readers would also seem a plausible account fct what is often observed with

many L2 readers. The fo11owing statement by an L2 reader (Berman 1984: 143) could be

considered expressing difficulty processing one whole sentence due to overloaded working

memory:

`I

was so busy working on this part of the sentence, I forgot it was connected with something

else.'

Perkins & Brutten (1992: 72-3) comment that their own wark indicates that second language

readers' memory span is tco little to be able to psoccss aiKl to recall large quantity of text. Pafan

(1994) reports a finding by Tajika (1990) that readers chose inappropriate cultura1 schemata when

reading in L2 but were able to avoid that when reading the same text in Ll. This could be acase

where dealing with L2 specific problems took up more of their working memory that thcre was not

sufficient cognitive resources left for appropriate higher order operation to take place, In other

words, insufficient processing ethciency may have prevented their desirable Ll reading strategy -

-73-

Page 10: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation ofTeacher of English, KATE

activation of appropriate schemata in this case - to be transferTed to L2 reading. My own experience

also indicates many L2 readers often fhil to make generalizations of to draw inferenoes fVvm the

materials they read. It would not seem reasonable to assume that they 1ack generalization or

inference ability perse since they seem to be much better at such operations when reading in their

L1. What seems to be happening is that those readers are devoting too much capacity to proeessingthe incoming words to be engaged in higher order processing. Borrowing Clarke's (1980: 206)

expression, `1imited

control over the language "short circuits" the good reader's system, causing

himther to revert to poor reader strategies when confVonted with a dirncult or confusing task in the

second1anguage'.

There is strong indication that what enables Ll reading strategies and acquired cognitive

abilities to be transferred to L2 reading is threshold level of L2 ahility (Cummins 1976; Clarke

1979; Alderson 1984). The implication, then, would be that individual differences in L2 reading

ability may essentially exist in lower order processing efficiency rather than in higher orderabilities. 111is woula lead us to hypothesize that ineasuring L2 specific ability - lower order ski11s -

is an ethcient way 6f predich'ng overall L2 reading proficiency because a rz ader's lower erder ski11s

wou{d tell us also about the extent to which she is likely to be able to execute higher or(ler ski11s or

cognitive ability when reading in L2. Whether this hypothesis is empincally supported remains to

be investigated in the future.

BIBLIOGRAPHY

Alderson, J. 1984. Reading in a foreign langtiage: a reading prvblern or a language preblem? In

Alderson, J. C & Unguhart, A. H. (Eds,) Ruting in. a Fbreigpi Latrgucrge. Harlow: Longrnan.

Alderson. J. C. 1990a. Testing reading ccFmprehension ski11s (Part one), Reeding in a Forelgri

la,rguage 6.2. pp. 425-438.

Alderson, J. C. 1990b. Tlesting reading compTehension skills (Part two).Getting students to talk

ahout taking a reading test (a pilot study). Retiding in a Foreigti Lt ngutrge 7, 1. pp. 465-503.Alderson, J. C. & Lukmani, Y. 1989, Cognition and reading: cognitive levels as embodied in test

questions. Readiqg in a Foreigv: Ltitrguage 5,2. pp. 2S3-270.

Anderson, J. 1976. ilsJTholitrguistic Etpert'tnents in Foreign Latrguqge 7lasting. Santa Lucia,

Queensland, Australia: University of Queensland Press.

Andrich, D. & Godffey. J. R. 197g-9. Hierarchies in the skills of Davis' Reading Comprehension

Test, Fonn D: An empirical investigation using a latent trait model. Reading Research

Qxarterty XIV. 2. pp. 1g2-2oo.

Atkinson, R. C., and Shifffin, R. M. 1968. Human memory; A proposed system and its control

ptocesscs. Ip Spence, K. W. and Spence, J. T. (Eds.). The 1 tsyehotqgy ofLeaming and

MbtitAats'on: Advances in Rescarch and 1heot), Vhl. 2. New Yotk: Academic Press.

-74-

Page 11: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation of Teacher of English, KATE

Atkinson, R. C., and Shiffrin, R. M. 1971. The control of short-term memory. Scientijic

Amert'aan 225. New Yotk: Academic Press, pp. 82-90.

Baddeley, A. D. and H{teh, G. 1974. Workirtg memory. In Bower, G. H, (Ed.) 71he Ibychology of

Lecrvting and Malithation. (Vol. g) San Diego, CA: Academic Press,

Baker, R. L. 1989. An investigation of the Rasch madel in its apPtication to foreign langucrge

pmfibientty testing. Unpublished doctoul dissenation, University ofMinburgh.

Bejar, L I. 1983. Achiet-etnent Testitrg: Recent Advatrces. Beverly Hills: Sage Publications,

Berman, R. A. 1984. Syntactic components of tbe FL reading process. In Alderson, J. C &

Unguhart, A. H. (Eds.), Rutitrg in a Fbreign Ltingvetage, Harlow: Longrnan.

Blais, J. and Laurier, M. D, 1995. The dmensionality of a placement test from several analytical

perspectives. Latrgutrge Testitrg 12. 1 , pp. 72-98.

Bock, R. D., Gibbons, R. D., and Muraki, E. 1985. FtiU-hVfOrmation Item Factot' Analysis O(RC

Report No,85-1 [Revised]). Chicago: }gational Opinion Research Center.

Canale, M. 1986. The promise and threat of computerized adaptiye assessment of reading

compreherision. In Stansfield, C. W. CEd.). 7;echttt)lttgy and Latrgut4ge 7;asting. Washington,

DC:Teachers of English to Speakers of Other Languages.

Carpenter, P. A. & Just, M. A. 1986. Cogniti-ve Processes in Reading. In Orasanu, J. Md.), Ruting

dmpahension: Fro"i Researth to P)rdw Hillsdale, NJ: Lawrence Erlbaum.

Carpenter, P. A. & Just, M. A. 1989. [[he role of working memory in 1anguage compreliension. In

Klahr, D. & Kotovsky, K. (Eds.), Cbmptex hofbr,nation Ptocwssi,rg: 11ie imPact ofHerbert A.

Simon. Hillsdale, NJ: Erlbaum.

Carver, R. P. 1992. wnat d'o standardized tests of reading comprehension measure in terms of

efficiency, accuracy,, and fate? Recuiing Research Qtearterly 27. 4. pp. 346-359.

Case, R, 1978. Inte11ectual develcrpment ffom birth to adulthood: A Neo-Piagetian interpretation. In

Seigler, R. (Ed.), C)iildrenls 11iinking VVhctt IlevefizPs? Hillsdale, N, J.: Lawrence Erlballm

Associates.

Chi, M. T. H. 1976. The short-term rnemory limitations in children: Capacity or processing

deficits? Memot y and thgnuton 4. pp. 559-580.

Choi, I. & Bachman, L. F. 1992. An investigation into the adequacy of three IRT models for data

ffom two EFL reading tests. hanguage 71astirg 9.1. pp. 51-78.

Clarke, M. A. 1979. Reading in Spanish and Engtish: Evidence from adult ESL students. languqge

Luarvsing29.pp.121-150.

Clark, M. A. 1980. The short circuit hypothesis of ESL reading-r when language competence

interferes with reading perfbrmance. MtNlervt Ltingv"rgefotcrvttal 64. pp. 203-2{}9,

-75-

Page 12: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation ofTeacher of English, KATE

Cohen, G. 1979. Language comprehension in old age. Cognutve 1ts),choLqgy l l. pp. 412-429.

Cummins, J. 1976. The infiuence of bilingualism on cognitive growth: a synthesis of research

findings and explanatory hypotheses. vabrking Rtpers on Bilinguatism 9. pp. 2-43.

Daneman, M. 1987. Reading and working rnemory. In Beech, J. R, & Colley, A. M. (Eds.).

Ciog7titive Aoprotrches to Reading. John Wiley & Sons: Chichester.

Daneman, M. & Carpenter, P. A. 1980. lndividual differences in working memory and reading,

Jbund of labcti Lcamingand Vlerb`ti Behavier 19. pp. 450466.

Daneman, M. & Carpenter, P. A. 1983. Individual differences in integrating information between

and wimin sentences.foetrnal ofEtperi'mental 1)bycholqgy: Letm:ing Metnot )t and Ct)gnition.

pp.561-584.

Davies, A. 1981. (Reyiew ofi Communicative Syllabus Design. 1;ESQL (?veelrterly 15. pp, 332-336.

Davis, E B. 1941. Fbentlamentalfoctors in cemPrehension in mading' . Unpublished docteral

dissertation; Harvard University.

Davis, F. B. 1968. Research in comprehension in reading. Reading Rescarch Qucirterly 3.4. pp. 499-545.

Denick C. 1953. ITiree tzspects ofretlding comptchension as masured by tests ofdute,ent le,rgtks.

Unpublished doetoral dissertation; University of Chicago.

Downing, J. and Leong, C, K. 1982. Rsychology of Reading. New York: Macmillan.

Grellet, F.1981. Developing Reading Skills: A Ptactical Guide to Reading Cemprehension

Ewvises. Cambridge: Carnbridge University Press.

Hambleton, R. K. and Swaminathan, H. 1985. netn Response lheoty: IVinciptes and 4aPliartions. Boston: Kluwer-Nijhoff

Heaton, J. B. 1988. varititrg' E)rgtish Lamp"rge 1lasts. New Edition. London: bongrnan,

Henning, G. 1992. Dimensionality and construct validity of language tests. Latrgmrge 11estitrg 9.1.

pp.1-11,

Hinofotis, F, A. B. 1976. An inthestigutibn ofthe concttrrerti thcdlddy of alate testing as a tncasure of

otigraU pm]Zcienc), in Eirglish as a secotid language. Unpublished doctoral dissertation, South

IllinoisUniversity.

Hiteh, G. J. and Baddeley, A. D. 1976. Verbal reasoning and working mernory. Quarterlylbtttvtat

ofEtper±'mental

Psphoklgy 28. pp, 603-62 1 .

Hughes, A. & Porter, D. (Eds.), 1983. Current Devel(rpments in Language 11estitrg, London:

Academic Press.

Hunt, L. C. 1957. Can we measure specific factors associated with reading comprehension?

Jlotcrnal ofliUtcaationtzl Rescarch S1. pp, 161-171.

Huttenlocher, J. & Butke. D. 1976. VVhy does memory span increase with age? Jbttf?tal of lerbtzl

-76-

Page 13: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation of Teacher of English, KATE

Lcarningand Verb`d Behavior 8. pp. 1-3 1.

Jafarpur, A. 1987. The short-context technique: an alternative fbr testing reading cornprehensien.

Leingwage Zesting 4.2. pp.193-220.

Just, M. A. & Carpenter, P. A. 1980, A theory of reading: ffom eye fixations to cemprehension.

Pb)ehategicat Rev:'etv 87,4. pp. 329-354,

inst, M. A. & Carpenter, P. A. 1987. 11he PlsyshtMagy ofRuting and in,rgutrge Co,npmhension.

Boston: Allyn & Bacon.

Jusu M. A. & Carpenter, P. A. 1992. A capacity theory of cornprehension: individual differences in

working memory. Psy¢hcilegictze Revlew 99.1. pp. 122-149.

Kemper, S. 1986. Imitation of complex syntactic constructions by elderly adults. APPtied

Asycholotinguistias 7. pp. 277-288.

King, J. & Just M. A, 1991. IndividuEl difiierences in syntactic processing: the role of working

memory. found ofMe,not y and Langutrge 30. pp. 586602.

Lumiey, T, 1993, The notion of subski11s in reading comprehension tests: 'an

EAP example.

Langvarge 71esti,rg 10.3. pp. 21 1-234.

Lunzer, E., Waite, M. & bolan, T. 1on9. Comprehension and comprelrension tests. ln Lunzer, E. &

Gardner, K. (Eds). 1)ie Etrutve tlse oflaeeding. London: Heinemann Educational Books.

MacDonald, M. C., Just, M. A. & Carpenter, P. A. 1992. Wcrrking memory constraints on the

processing of syntactic aTnbiguity. Ckrgn"iee twholclgy 24. pp. 56-98.

McNamara, T. 1991. Test dimensionality: IRT analysis of ari ESP listening test Ltingt"4ge 1;asting

82. pp. 139-159.

McNamara, T. 1996. Mcasun'ng Second Lt"rguage Pwforpnance. Essex: Addison Wesley bongman

Lrd.

Mead. R. 1982. Review of Munby, J. Cemmunicative Syllabus Design. 4Pptied Linguistic s 3.1 . pp.

7078.

Mitehelll, D. C. 1982. 17te P7acess ofRuting: A Ctrgnive Amztysis ofFlnent Ruting and Learning

to RutL Chichester: John Wiley & Sons.

Mitchell, D. C, 1984, An eyaluation ef subject-paced reading tasks and other methods for

investigating immediate process in reading. In Kieras, D.E. and Just, M. A. (Eds,). IVIew

Mbthads in Reading CbmPrehension Rescarch. Hillsdale, New Jersey: Lawrence Erlbaum

Associates.

Munby, J. 1978. Communicative S),Uabus Design. Cambridge: Cambridge University Press.

Negishi, M. 1996. 71he Latent 1hait Sbucture ofL2 Rcaditrg dntptelimtsibtt 11ests. Unpublished

doetoral dissertation, Centre for Applied Language Studies, University of Rcading.

Oller, J. W., Jr. 1979. Latrgmrge 11ests at thhool. London: Longrnan.

-77-

Page 14: Unidimensionality Reading AbMty Working Memory

Kantokoshinetsu Association of Teacher of English, KATE

NII-Electronic Library Service

KantokoshinetsuAssociation ofTeacher of English, KATE

Oller, J. W., Jr. (Eds.) 1983. lsswes in Latrgucrge Tbstirrg Research. Rowley, Massachusetts:

Newbury House,

Oller, J. W. Jr. & Perkins, K. 1978. Levngutrge in ElaVeecation: 71asting the 11asts. Rgwly, Mass.:

Newbury House.

Paran, -A. 1994. Rutitrg Proaesses in a ,first atid a second latrgiqge: an on-line compan'son of

E)rgtish and Hleblew. Unpublished doctaral dissertatioru Ur!iversity of Reading.

Paris, S. G. & Lindauer, B. K 1976. The role of inference in children's comprehension and

merriory fbr sentences. C)tegnutve Ilsysholtagty 8. pp. 217-227.Paris, S. G. & Upton, L. R 1976. Children's mernory for inferential relationships in peose. Chua

Detheltipment 47. pp. 660-668.

Perkins, K. & Brutten, S. R. 1988. An item discriminability study of textually expllcit, textually

impliciL and scripta11y implicit question. RELCJbumal 19.2 pp.1-1 1 .

Posner, M. I. and Rossman, E 1965. Etfects of size and location of infbrmational transforms upon

short-term retention.Jburvtal ofE} ptrt'"iental Pk"et hology 70. pp. 496-50S.Rayner, K & PcSlatsek A, 1989. 11re Rspfichaltlgy ofRuting: Englewood Cliffs, N. J.: Prentice Hal1,Rosg D. H. t993. Assessing different components of reading comprehension: fact or fiction?

La,rgmrge 71asting 1O. 1. pp. 79-92.

Speanitg D. 1972. Identification of subski11s of reading cornprehension by maximum likclihood

factor artalysis. RecedingResueTrch Quarterly 8.1, pp, 92-11 1.

Tajika, M. 1990. EFL reading comprehension as an interactive ptocess. Paper presented at the 24th

msOL Convention, San Elrancisco, CA, March 6-IO 1990.

1horndike, R. L. 1971. Reading as reasoning. Address delivered to Division 15, Arnerican

Psychological Assoeiation, Washington D. C.

Tliorndike, R. L. 1982. Ecti)cational measurement: theory and practice. In Spearitt, D. (Ed.) 11he

imProvetnent of Mleasurement in Education and JFts)ichology. Hawthorn, Victoria: The

Australian Council fictr Educational Research.

Thurstone, L. L. 1946. Note on a reanalysis of Davis' resding tests. Pbychometnlha 1 1. pp. 185-188.Waugh, N. C. and NorTnan, D, A. 1965. Prirnary memory. Rs)-chologitzzt Review 72, pp. 89-104.Weir, C, J., Hughes, A. and Porter, D. 1990. Reading skills: hierarchies, implicational relationships

and identifiability. Reeditrg in a Fbtreigvt Latrgi"4ge 7. 1 . pp. 505-51O.Williams, E. & Moran, C. 1989. Reading in a foreign language at intennediate and advanced levels

with particular reference to English. Lzinguqge Tcaching 22.4. pp, 217-228.Yalden, J. 1987. AincsPles of Course Dest'gvt for Ltingwtrge 71eaching. Cambridge: Cambridge

University Press.

Yui11, N., Oakhill, J, & Parkin, A. 1990, Working memory, cornprehension ability afKl the resolution

of text anomaly. Unpublished Manuscript. University of Sussex. Brighton, East Sussex,

England.

Zwick, R. 1987. Assessing the dimensionality of NAEP reading data. fournal of Educational

Maczsurement 24, pp. 293-308.

-78-