a visually based developmental reading...

A Visually Based Developmental Reading Deficit

Journal of Memory and Language43, 157–181 (2000)doi:10.1006/jmla.2000.2724, available online at http://www.idealibrary.com on

Michael McCloskey and Brenda Rapp

Johns Hopkins University

This study describes a developmental reading disability and identifies a deficit of visual perceptionas the underlying cause. A.H., a university student with apparently normal reading comprehension,was severely impaired in reading aloud isolated words (e.g.,dear3 “pear”) and sequences ofunrelated words (e.g.,pouch cedar culture jacket3 “cedar pouch jacket culture”). Eight experimentsinvolving several visual presentation conditions convincingly linked her impaired reading perfor-mance to a developmental deficit in perceiving the location and orientation of visual stimuli. Fouradditional experiments demonstrated that A.H. achieves good comprehension for meaningful materialby exploiting knowledge-based constraints (e.g., syntactic constraints) to “repair” the errors intro-duced by her visual system. These results have implications for research on developmental dyslexia,normal reading, and normal vision.© 2000 Academic Press

Key Words:cognitive neuropsychology; dyslexia; developmental dyslexia; reading; reading dis-ability; vision; visual localization.

As this special issue attests, recent yearstivts.d olts

iortatoamveiallsideem-caisefi

cits merit greater attention from cognitive sci-ese

ndorydinged

ngox-ac-ores,ingim-se-

aseor-owel-ndficitthetud-glyhertheofas

thernaicaentforelisr hefor

essce,nd

have seen a growing interest among cogniscientists in the study of cognitive deficiFor the most part, this interest has centeredeficits resulting from brain damage in aduwhose cognitive functioning was normal prto some neurological insult. Developmendeficits, in which a cognitive function failsdevelop normally, have less often been exined from a cognitive science perspectidespite the fact that these deficits— especthe developmental dyslexias—are of conerable societal concern and have long bmajor foci of attention in other research comunities (e.g., neurology, clinical and edutional psychology). One aim of this articleto suggest that developmental cognitive d

This study was supported in part by grants fromNational Institutes of Health (NS21047) and the Intetional Association of Learning Disabilities. We thank ErPalmer for her assistance with several of the experimMatt Goldrick for testing control subjects, Jill Egethconstructing stimulus sentences and scoring results, MOles for assistance with data analyses, and Jill Folk fohelpful comments. We are especially grateful to A.H.her patience with our seemingly endless testing.

Correspondence concerning this article should be addrto Michael McCloskey, Department of Cognitive ScienKrieger Hall, Johns Hopkins University, Baltimore, Maryla21218. E-mail: [email protected].

157

e

n

l

-,y-n

-

-

entists, both because of the insights thdeficits may offer into normal cognition abecause of the contributions cognitive theand methods may make to the understanof the deficits, and ultimately to improvdiagnosis and treatment.

We report a single-subject study of a youwoman, A.H., who presented with a paradical pattern of reading performance: Herademic achievement, standardized test scand self-report suggested that her readability was above average, yet on certain sple reading tasks her performance wasverely impaired. We first present a brief chistory and describe A.H.’s impaired perfmance on two reading tasks. We then shthat A.H. has a profound but selective devopmental deficit in visual perception apresent evidence that this perceptual deunderlies her impaired performance onreading tasks. Next, we describe several sies exploring how A.H. can show seeminnormal reading comprehension despiteperceptual deficit, and finally we considerimplications of our results for the studydevelopmental cognitive deficits as wellfor research on normal cognition.

-

s,

sar

ed

0749-596X/00 $35.00Copyright © 2000 by Academic Press

All rights of reproduction in any form reserved.

A.H.

alhisnoicaag

eleas

lityultesothsstingall

uatgrendu

ofas

unin-s aarat-nand). Iooemees ishm

ooloreim-lityntiltic

bu-

rdn-ffi-ard

ized test report a graphic presentation oftop

testmsr-toredvenng,para-uch

ce in

ai-on-re todi-in

Inper,ok,24ef-

theitordark

ulianditor.

tedm.ap-

lly

.H.,eseuni-rel.

d-rds

158 MCCLOSKEY AND RAPP

A.H. is a right-handed woman with normvisual acuity and visual fields. She has notory of neurological injury or disease, andabnormalities were apparent on neurologexam, electroencephalogram, or structural mnetic resonance imaging. Throughout hermentary and secondary schooling A.H. wassumed to have some form of learning disabibecause she experienced substantial difficwith spelling, math, and foreign languagHowever, her academic performance waserwise good, and she was not placed in clafor learning-disabled students. After graduafrom high school, she entered an academicselective university as a regular undergradstudent, and she completed a bachelor’s dein history in the usual four years. She thentered law school and graduated on scheafter three years.

A.H. was studied in our lab for a periodnearly four years, during which time she w18 –21 years old and an undergraduateversity student. In an interview at the begning of the study A.H. described herself avery bad speller, and she also reported vous difficulties with numbers and mathemics (e.g., difficulty memorizing multiplicatiotables, difficulty learning her new addressphone number when she entered collegecontrast, she described herself as a greader, asserting that she had no probllearning to read and had consistently bplaced in the most advanced reading groupelementary school. A.H. also stated thatread a great deal and had no difficulty copleting her reading assignments in schA.H.’s grades and standardized test scfrom elementary school through college silarly suggest above-average reading abiFor example, she scored in the 94th perceon the Verbal section of the SAT (ScholasAptitude Test), a challenging test of vocalary and reading comprehension.

Nevertheless, both A.H.’s academic recoand her comments in the initial interview cotain occasional hints of potential reading diculty. For example, on a tenth-grade stand

-

l---

,y.-

es

yeee

le

i-

i-

nds

nne-.s

.e

s

-

vocabulary subtest scores places A.H. at theof the scale on a Synonyms in Context subbut at the lower end of the scale on Synonyin Isolation. Interestingly, A.H. mentioned duing the interview that she occasionally failedrecognize familiar words when she encountethem out of context. She also stated that efor material that was not especially challengishe sometimes had to read sentences orgraphs more than once before they made msense. These points take on some significanlight of the results we report.

EXPERIMENTAL STUDIES

We tested A.H.’s reading performance invariety of tasks with stimuli ranging from indvidual letters to paragraphs. All tasks were nspeeded; that is, A.H. was under no pressurespond quickly. Except where otherwise incated, the stimulus on each trial remainedview until A.H. completed her response.some tasks the stimuli were presented on paprinted in Times Roman, Century Schoolboor Courier fonts ranging in size from 12 topoints. (These variations had no discerniblefect on A.H.’s performance.) In other tasksstimuli were presented on a computer monin a sans serif font, as white characters on abackground. Unless otherwise indicated, stimwere presented in lower case characterswere centered on the page or computer monA.H. viewed the computer displays in a lighroom from a distance of approximately 40 cAt this distance each character subtendedproximately 0.5° of visual angle horizontaand between 0.5° and 1° vertically.

For some of the tasks presented to Anormal control subjects were also tested. Thsubjects were undergraduate students at theversity attended by A.H., and they wematched to her in age and educational leve

IMPAIRED READING PERFORMANCE

In a preliminary exploration of A.H.’s reaing we asked her to read aloud individual woand sequences of unrelated words.

Reading Aloud Individual Words

obden,4/ratuske

totim

ordreonlour i

lslis

thead

iteria

aco

sitahe

neatee-d,rdsim-

rdscitreuld

hicing

cal-rd,rd,tion

ili-readastlyto

.H.er

theedre-man-entrmri-tthevi-ualen-iliar

he928

eadon.listthe

.’s

.g.,b d tA kenr beti da

TABLE 1

159DEVELOPMENTAL READING DEFICIT

Stimuli were the names of 144 concretejects selected from the Snodgrass and Vanwart (1980) pictures (e.g.,sun, envelope, crowrefrigerator).1 A.H. was 100% correct (14144) in naming the pictures orally, demonsting that she was familiar with the stimulwords and was able to produce them in spoform.

The words, which varied in length from 312 letters, were presented on paper one at ain random order, and A.H. read each waloud. The stimulus list was presented thtimes, once in each of three testing sessiFour control subjects each read the words aonce; these subjects collectively made 1 erro576 trials (0.2%).

A.H., however, erred on 53 of 432 tria(12%) across the three presentations of theThe 53 errors were distributed over 39 of144 stimulus words: 28 words were misreonce, 8 were misread twice, and three elicerrors on all three presentations. On some tA.H. hesitated noticeably before producingresponse. However, most responses, bothrect and erroneous, were made without hetion, and A.H. usually seemed unaware oferrors.

Table 1 presents examples of A.H.’s erroous responses. As the examples illustrA.H.’s errors were not restricted to low-frquency words such asanchor;she also misreashort, high-frequency words such asdog, penandhand.All of the erroneous response wowere visually similar to the corresponding stulus words.

Impaired performance in reading woaloud could conceivably arise from a defiaffecting any of the cognitive processesquired by the task. For example, errors cooccur in constructing visual or orthograprepresentations of a stimulus word, match

1 We use italics to indicate visual stimulus words (eell); it should be understood that the words presente.H. were not italicized. We use quotation marks for spo

esponses and a horizontal arrow for the relationshipween stimulus and response. For example,bell 3 “dell”ndicates that the visual stimulus wordbell was read alous “dell.”

-r-

-

n

e

es.dn

t.

dls

r--

r

-,

-

the stimulus representation to a stored lexiorthographic representation of a familiar woretrieving semantic information about the woor accessing the phonological representaneeded to generate a spoken response.

In an effort to narrow the range of possibties, we asked A.H. to define words as shethem aloud. The question of interest wwhether definitions for words read incorrecwould correspond to the stimulus or insteadthe erroneous response. For example, if Aread dog as “bog,” would she define the formor the latter? A definition corresponding tostimulusdog would suggest that A.H. matchthe stimulus word to the correct lexical repsentation, and accessed the appropriate setic information, but erred in some subsequprocess (e.g., retrieving the phonological foof the word). In contrast, a definition appropate to the response wordbogwould indicate thaprocessing had gone awry prior to accessingword’s meaning, perhaps in constructing asual or orthographic representation of the visstimulus, or in matching the stimulus represtation to the stored representation of a famword.

Stimuli were the 144-word list used in tprevious word-reading task and a list ofcommon nouns varying in length from 3 toletters. In one presentation of each list A.H. rthe word aloud and then gave a brief definitiIn a second presentation of the 92-wordA.H. first gave the definition and then readword aloud.

The results demonstrated clearly that A.H

o

-

Examples of A.H.’s Errors in Reading Words Aloud

Stimulus Response

dog bogpen denlamp lambsnail nailschain cabinhand bandnose noisechurch cherishapple appeal

thstsialshed

plni-i-

ingnsn oatusec

ouina

creithed-mlyceitheac

adtl-

s-he60

ntlyg er-

nder of

ofecttedhat.’s

n intheurehatndat

nceid-henm-

a-iouss-&

ings

hedh : Sg shr

TABLE 2


word-reading errors arose prior to accessingmeaning of the word. Collapsing across liA.H. made reading errors on 75 of the 328 tr(23%), and in all cases her definition matcher erroneous reading response. For examshe readbreadas “beard,” and gave the defition “man’s facial hair.” (See Table 2 for addtional examples.) Twenty-one of the readerrors with corresponding incorrect definitiooccurred in the define-then-read presentatiothe 92-word list, ruling out the possibility therrors in defining words occurred only becaA.H. was confused by hearing her incorrreading response.2

Reading Aloud Word Sequences

In this task A.H. read aloud sequences of funrelated words (e.g.,peacock comet napkdolphin) presented in a horizontal row oncomputer monitor. Twenty sequences wereated from 80 5- to 7-letter nouns. Words wsalient visually similar neighbors were avoidin order to minimize A.H.’s difficulty with individual words. Words were assigned randoto sequences and to positions within sequenTwo blocks of 20 trials were presented, weach word sequence appearing once inblock.

On two-thirds of the trials (27/40) A.H. rethe words out of sequence (e.g.,peacock comenapkin dolphin3 “peacock napkin comet do

2 As one would expect, A.H.’s definitions also matcer reading responses for words she read accuratelyave an adequate definition for 251 of the 253 wordsead correctly (99%).

Examples of A.H.’s Errors

Stimulus Reading response

dog bogbone donepig digstar tarsrib ripsun nunsskirt skitdust dusk

e,

e,

f

et

r

-

,

s.

h

phin”). All but one of the errors involved tranposition of adjacent words. In addition to tordering errors, A.H. misread 13 of the 1individual words (e.g.,recipe3 “receipt”). Onmost trials A.H. read the word sequence flueand seemed unaware that she was makinrors.

The results from the individual-word aword-sequence reading tasks raise a numbquestions. For example: What is the naturethe underlying deficit? Does this deficit affA.H.’s reading of material other than isolawords and arbitrary word sequences? If so, ware the implications for understanding A.Hseemingly excellent reading comprehensioher academic pursuits and daily life? Infollowing sections we first consider the natof the deficit, describing results showing tA.H. is impaired in perceiving the location aorientation of visual stimuli, and arguing ththis deficit underlies her impaired performaon various reading tasks, including the indivual-word and word-sequence tasks. We texplore A.H.’s apparently normal reading coprehension in light of these arguments.

IMPAIRED VISUAL PERCEPTION OFLOCATION AND ORIENTATION

A.H.’s deficit in visual location and orienttion perception has been described in prevpublications (McCloskey et al., 1995; McClokey & Rapp, 2000; see also McCloskeyPalmer, 1996); here we summarize the findrelevant to the present study.

hee

Reading and Defining Words

Definition

rky swampen an activity is completedake a hole in the earth

en someone puts tar on somethingeartholic women who have given certain vows

y short play, usually without props or costumese of day just after the sun sets

in

muwh

to mwh

to tCa

vertim

Impaired Performance in Localizing

nceuale

inanente-of

taronthe

imal-teht,anglefen

4)imr

r a

oul-Inapa

h -u rod . Oo het hec lec

heo ana tero usl w-

h on-s iewu thet

uliw sksr rmo m-p en-t ndF dea thet ont ectc ndV ta-t rys li.

er-m too nce,i ne t ash vi-d mp-t a

onstX:n

caluceden-py,


Visual Stimuli

A.H. showed severely impaired performaon several tasks requiring localization of visstimuli. In one task she closed her eyes whismall wooden block was placed on a tablefront of her. She then opened her eyesreached for the block with a ballistic movem(i.e., without changing direction in midmovment). When the target was directly in fronther, A.H. reached accurately. However, forgets on her left or right she reached to the wrside of the table (e.g., right for a target onleft) on two-thirds of the trials (63/96).

Results from other experiments revealedpairment in up/down as well as left/right locization. In one experiment an X was presenon a computer monitor at a position left, rigup, or down from the center of the screen,A.H. indicated the location of the X by movina mouse (e.g., move left for a stimulus at thelocation). Whereas control subjects performthe task virtually without error, A.H. erred oover half of the left and right stimuli (75/14and on over one-third of the up and down stuli (51/144). All of her errors were left/right oup/down confusions (e.g., moving down fostimulus at the up location).

A.H.’s visual localization deficit was alsapparent in tasks requiring localization of mtiple visual stimuli relative to one another.one experiment sequences of 3–5 plastic sh(e.g.,circle triangle rectangle) were placed in

orizontal row in front of A.H. While the stimlus sequence remained in view, A.H. repuced it using her own set of shape tokensver half of the trials (16/30) she arranged

okens in the wrong order; for example, sopied the sequencesquare rectangle triangircle as square rectangle circle triangle.

Impaired Performance in VisualOrientation Judgments

A.H. also showed impairment in judging trientation of visual stimuli. In one taskrrowhead (. or ,) was presented at the cenf a computer screen, and A.H. moved a mo

eft or right to indicate the direction the arro

la

d

-g

-

d

d

td

-

es

-nr

e

ead was pointing. Although the task was npeeded, and the stimulus remained in vntil she responded, she erred on 40% of

rials (38/96).In several tasks A.H. copied visual stimhile the stimuli remained in view. These ta

evealed frequent orientation errors, in the fof left–right or up–down reflections. For exale, Figure 1A shows a stimulus from the B

on Visual Retention Test (Sivan, 1992), aigure 1B shows A.H.’s direct copy. A.H. man up–down reflection in copying one of

wo large forms, and a left–right reflectionhe other. Figure 2, which shows A.H.’s diropy of a picture from the Snodgrass aanderwart (1980) set, illustrates that orien

ion errors occurred not only for arbitrahapes but also for meaningful visual stimuGiven that an object’s orientation is detined by the locations of its parts relativene another and to some frame(s) of refere

t seems likely that A.H.’s visual orientatiorrors stem from the same underlying deficier errors in localizing visual stimuli (and eence discussed below supports this assu

ion). Accordingly, we refer to the deficit as

FIG. 1. (A) Stimulus from the Benton Visual RetentiTest (Sivan, 1992). FromBenton Visual Retention Te(Form D, Design 6), by A.B. Sivan, 1992, San Antonio, TThe Psychological Corporation.Benton Visual RetentioTest,Fifth Edition. Copyright © 1992 by The PsychologiCorporation, a Harcourt Assessment Company. Reprodby permission. All rights reserved. “Benton Visual Rettion Test” is a registered trademark. (B) A.H.’s direct comade while the stimulus remained in view.

d-ul-ep

ecionar

mi-denighorkbeob

ofithionpleratendtofelst,owonra

ion

processing across sensory modalities, or a mo-n-s.

ofcu-nta-er-ing

inal

di-rongng

not

ereorr-hernta-beFor

ina

ralthearly

s ahatori-af-ofof

ef-as

usft or

ofrrorion,(17000

hen

artlus


visual localization deficit, with the understaning that this term encompasses A.H.’s difficties with orientation as well as location perction.

Longstanding Impairment

Samples of A.H.’s elementary through sondary schoolwork indicate that the localizatdeficit was present throughout her school yeFor instance, striking mislocalizations andsorientations of people and objects were eviin sketches of paintings A.H. made during hschool visits to art museums. The schoolwerrors suggest that A.H.’s impairment maycongenital, especially given the absence ofvious postnatal neurological insult.

Selective Visual Impairment

A.H.’s impairment is a selective deficitvisual perception. In a variety of tasks wnonvisual stimuli her location and orientatjudgments were highly accurate. For examwith eyes closed she was uniformly accu(186/186) in pointing to the source of a souon her left or right and in drawing an arrowmatch the orientation of a wooden arrow sheby placing her palm over it (32/32). (In contrashe was wrong half of the time when the arrwas presented visually.) These findings demstrate that A.H.’s impairment is not a genespatial deficit affecting location and orientat

FIG. 2. (A) Picture from the Snodgrass and Vanderw(1980) set. (B) A.H.’s direct copy, made while the stimuremained in view.

-

-

s.

t

-

,

t

-l

tor deficit affecting the ability to carry out itended movements when making response

Other findings argue against the possibilitya visual–motor disconnection in which (acrate) visual location and orientation represetions fail to make contact with (perhaps othwise intact) mechanisms for executmovements. In particular, A.H. was impairedindicating the location or orientation of visustimuli even in tasks not requiring spatiallyrected movements. For example, she was wover one-third of the time (24/64) in sayi“left” or “right” to indicate the location of an Xon a computer monitor. (These errors didreflect any difficulty with the wordsleft andright, because verbal left–right responses wuniformly correct in tasks involving auditorytactile stimuli.) Finally, A.H. showed impaiment in tasks with visual stimuli even wheneyes remained fixated during stimulus presetion, indicating that her errors could notattributed to an eye movement disorder.example, her error rate was 56% (45/80)localizing an X to the left or right side ofcomputer monitor while fixating on a centpoint. Taken together, the findings fromvarious visual and nonvisual tasks point cleto a selective deficit of visual perception.

Effects of Visual Variables

Further evidence that A.H.’s impairment ivisual deficit comes from results showing ther accuracy in perceiving the location andentation of visual stimuli was dramaticallyfected by several visual variables. Twothese—exposure duration and flicker—areparticular relevance to the present study.

Several tasks revealed a counterintuitivefect of exposure duration: A.H.’s accuracy wmuch higher for brief than for long stimulpresentations. For example, in pressing a leright response button to indicate the locationa dot displayed on a computer monitor, her erate increased steadily with exposure duratfrom less than 5% at the briefest exposurems) to nearly 50% at the longest exposure (1ms).

The same phenomenon was observed w

A.H. moved a mouse left or right to indicate thew

e nex-

p-er:ingss.

to00

teressthedi--mst toofhe

.’sushe2)deoferm

f 25H.ctly

er,acta-r-forheer

hee

upnd

am

ef-uan iss,

transient subsystem specialized for processingr-temuli.entsub-ectori-of

atean-ck-d)

as inaryc-

an-rawsnc-.g.,r,&

ne3;r-

t toA suals n oro el ofp hent incor-r lus,r pre-t wass sys-t Onee hichA s oft re-s tingt whena r ex-a arlya butt farl ist t thel her,s pecificl


orientation of an arrowhead. When the arrohead was presented for 50 ms, she maderrors in 48 trials; however, for a 1000-msposure her error rate was 46% (22/48).

A.H.’s visual location and orientation percetion was also profoundly affected by flickShe was much more accurate for flickerstimuli than for stimuli of constant brightneIn one experiment A.H. moved a mouseindicate the location of an X presented for 10ms on the left or right side of a compumonitor. In the Steady condition the brightnof the stimulus was constant throughoutdisplay interval, whereas in the Flicker contion the stimulus was turned on and off at 50intervals (i.e., 50 ms on, 50 ms off), causing iflicker visibly. Whereas A.H. erred on halfthe trials in the Steady condition (32/64), smade no errors in the Flicker condition.

Flicker also had a dramatic effect on A.Hdirect copying (i.e., copying while the stimulremained in view) of stimulus figures from tBenton Visual Retention Test (Sivan, 199Under normal lighting conditions, A.H. malocalization and/or orientation errors on 2130 trials (see Fig. 1 for an example). Howevwhen the only illumination in the testing roocame from a strobe light flashing at a rate oHz (thereby producing a strong flicker), A.made no errors, copying the stimulus correon 20 of 20 trials.

In addition to exposure duration and flickseveral other variables also affected A.H.’scuracy in perceiving the location and oriention of visual stimuli. For example, her perfomance was much better for moving thanstationary stimuli (McCloskey et al., 1995). Teffects of the visual variables provide furthevidence that A.H.’s deficit is visual. Also, tfinding that these variables have the samefects on location and orientation perception sports the assumption that A.H.’s location aorientation errors are manifestations of the sunderlying deficit.

McCloskey et al. (1995) interpreted thefects of visual variables by proposing that visprocessing of spatial location and orientatiocarried out in each of two visual subsystem

-o

.

,

-

f--

e

l

a

rapidly changing (e.g., moving, brief, flickeing) visual stimuli, and a sustained subsysmost sensitive to static longer-duration stimIn A.H., McCloskey et al. argued, the transisubsystem is intact, whereas the sustainedsystem is impaired, often producing incorrrepresentations of stimulus location and/orentation. On this account A.H.’s perceptionlocation and orientation is usually accurwhen stimulus conditions favor the (intact) trsient subsystem, as with brief, moving, or fliering stimuli. However, when the (impairesustained subsystem is strongly activated—the case of steady, long-duration, stationstimuli—localization and orientation errors ocur.3

The hypothesized distinction between trsient and sustained visual subsystems dupon previous proposals concerning the futional architecture of the visual system (eBreitmeyer & Ganz, 1976; Goodale & Milne1992; Lennie, Trevarthen, Van Essen,Wassle, 1990; Livingstone, 1990; Livingsto& Hubel, 1988; Merigan & Maunsell, 199Mishkin, Ungerleider, & Macko, 1983; Ungeleider & Mishkin, 1982), but it differs from

3 McCloskey et al. (1995) argued that the impairmen.H.’s sustained subsystem is not one in which vitimulus information cannot be used to compute locatiorientation, and a random choice is made at some levrocessing. Instead, A.H.’s errors apparently occur w

he sustained subsystem systematically computes anect location or orientation representation from the stimueversing left and right and/or up and down. This interation was suggested by several findings. First, A.H.ystematically below chance in some tasks, suggestingematic miscomputation rather than random choice. (xample is the reaching task described earlier, in w.H. reached to the wrong side of the table on two-third

he trials.) Also, A.H.’s localization errors consistently perved certain aspects of the stimulus location, indicahat at least some stimulus information was used evenn incorrect location representation was generated. Fomple, when A.H. erred in the reaching task, she nelways reached to a location with the correct distance

he wrong direction relative to the midline (e.g., reachingeft for an object on her far right). Also worth mentioninghat A.H. does not report experiencing uncertainty abouocation or orientation of objects she is looking at. Rathe says that she always perceives objects as having socations and orientations.

these proposals in some respects. For example,tin

msh-ed

om

eid-

ts obina-

alad-theraldie-

ersing

evio

ori-

r ee

inr tho cth thac et-ta -

ingugonted

rd-

sequence reading tasks are suggestive, thesethatseForhatson.hipereri-rds,ween-sults

u-

ys

u-

ys

u-

ys

red

pt-

esis-

re-ouldinalre-

icala

ter-o . Wo rs ina


whereas some theorists have proposed a distion between “what” and “where” subsyste(e.g., Mishkin et al., 1983; Ungerleider & Miskin, 1982), McCloskey et al. (1995) assumthat both visual subsystems make “where” cputations.

Regardless of the merit of the McCloskeyal. (1995) interpretation, the important conseration for present purposes is that the effecvisual variables can be used as tools for propotential links between A.H.’s visual localiztion deficit and her reading.

READING AND THE VISUALLOCALIZATION DEFICIT

What are the implications of A.H.’s visulocalization deficit for understanding her reing abilities and disabilities? Conceivably,visual deficit could affect reading in seveways, at least under ordinary reading contions. First, A.H.’s visual system might somtimes misrepresent the orientations of lettleading to confusions among letters differonly in orientation (e.g.,b andd). The individ-ual-word reading task provides suggestivedence on this point: A.H. made a numbererrors that could have resulted from letterentation confusions (e.g.,bone 3 “done”;ib3 “rip”; pig3 “dig”; see also the first threxamples in Table 1).Second, A.H.’s visual system might err

epresenting the relative locations—that is,rdering—of letters within a word, and in faer word-reading errors include severalould have resulted wholly or in part from ler-sequence confusions (e.g.,snail3 “nails”;pple3 “appeal”).4 Third, the visual localiza

tion deficit could lead to errors in representthe ordering of words in a sequence, as sgested by A.H.’s frequent word sequence cfusions in reading aloud strings of unrelawords.

Although the data from the word and wo

4 Many of A.H.’s word-reading errors were neither letrientation confusions nor letter-sequence confusionsffer a possible interpretation for these additional errolater section.

c-

-

t

fg

-

,

-f

e

t

--

data do not constitute strong evidenceA.H.’s visual localization deficit was the cauof her impaired performance on the tasks.example, we cannot rule out the possibility terrors such asrib 3 “rip” occurred for reasonother than misperception of letter orientatiTo explore more systematically the relationsbetween A.H.’s localization deficit and hreading, we carried out a number of expments in which she read aloud letters, woand word sequences. In light of the resultsthen examined her reading of coherent stences and paragraphs. Our report on the reof this testing is organized as follows:

Naming LettersExperiment 1 Upper vs lower caseExperiment 2 Brief vs long exposure d

rationExperiment 3 Steady vs flickering displa

Reading Single WordsExperiment 4 Upper vs lower caseExperiment 5 Brief vs long exposure d


Reading Word SequencesExperiment 7 Brief vs long exposure d


Reading ParagraphsExperiment 9 Normal vs sequence-alte

paragraphsReading Sentences

Experiment 10 Acceptable vs unacceable sentences

Experiment 11 Low constraint sentencExperiment 12 Steady vs flickering d

plays

Experiments 1–3: Letter Naming

If A.H.’s visual system sometimes misrepsents the orientations of letters, then she shmake letter orientation confusions not onlyreading words but also in naming individuletters. For a number of lower-case letters,flecting the letter across a horizontal or vertaxis will yield a visual form corresponding to

e

different letter. Forb, d, p, and q a left–rightr illt Foebu -t n,a ldsts rmom

r non orie -m d/ov mr iond -s p–d en-t ater -b oo blw ep cew et-t

E

re-s iglp alli dw ione htld hel rit-ia onfp beb ter

and that her upper-case errors would be re-s

er-c r ona ls.A erepw ean ec-t g-n

t-t 546t 68)f ,m n-nb et-t ple,a d“ .S dm ora too ots let-tt

arer-asse

ori-

r-c ofu r-c forn

sthatasepre-

t


eflection, an up–down reflection, or both, wransform one of the letters into another.xample,b becomesd by left–right reflection,py up–down reflection, andq by left–right plusp–down reflection. Also,n and u are poten

ially confusable through up–down reflectios arem andw. We predicted that A.H. wouhow impairment in naming these eightorien-ation-critical letters (b, d, p, q, n, u, m, w), andpecifically that her errors would take the fof letter orientation confusions (e.g.,b3 “d”;3 “w”).For the remaining letters (e.g.,e, k, o) mis-

epresenting the letter’s orientation wouldecessarily result in a naming error. Somentation-noncritical letters (e.g.,c, o) are symetrical, or nearly so, across horizontal an

ertical axes; for these letters, at least soeflections would yield a visual representatiffering little if at all from the correct repreentation. Even when a left–right and/or uown reflection would yield a visual repres

ation substantially different from an accurepresentation (e.g.,ə), the letter would probaly still be identifiable, because regardlessrientation the shape would not be confusaith that of a different letter. Accordingly, wredicted that A.H.’s naming performanould be better for orientation-noncritical l

ers than for orientation-critical letters.

xperiment 1: Naming Lower- andUpper-Case Letters

In this experiment A.H. named letters pented in lower and upper case. Whereas eower-case letters are orientation-critical (b, d,, q, n, u, m, w), only 3 upper-case letters f

nto this category.M andW could be confuseith one another through up/down reflectrrors (even though these letters have sligifferent shapes in most fonts). The only ot

etter that might be considered orientation-ccal in upper case isP, for which left–rightnd/or up–down reflections could lead to c

usion with lower-caseb, d, or q. We thereforeredicted that A.H.’s performance wouldetter for upper-case than for lower-case let

r

t-

re

fe

ht

yr

-

s,

tricted largely toM, W, and perhapsP.Method. The 26 upper-case and 26 low

ase letters were presented in random ordecomputer monitor in blocks of 52 tria

cross two testing sessions, 21 trial blocks wresented. The lower case lettersb, d, p,andqere identical except for orientation, as wermndw, andn andu. Upper-caseM andW wereon-identical in shape, but an up–down refl

ion of either letter would yield a form recoizable as the other letter.Results and discussion.For lower-case le

ers, A.H. made naming errors on 42 of therials (7.7%). Her error rate was 23% (39/1or the orientation-critical letters (b, d, p, q, n, u, w) but less than 1% (3/378) for orientatiooncritical letters,t(20) 5 7.37,p , .001. Allut one of the errors on orientation-critical l

ers were orientation confusions. For examcross the 21 presentations ofd, A.H. respondeb” six times, “p” three times, and “q” twiceimilarly, she namedw as “m” on five trials anas “w” twice. In the one remaining error f

n orientation-critical letter A.H. respondedne presentation ofd by saying that she was nure. Across the 18 orientation-noncriticalers, A.H. namedv as “u” once andz as “n”wice.5

For upper-case letters, most of whichorientation-noncritical, A.H. made only 12 erors in 546 trials (2.2%). This error rate wreliably lower than the 7.7% rate for lower caletters, t(20) 5 5.60, p , .001. Nine of thetwelve upper-case errors involved the threeentation-critical letters (M, W, and P). A.H.namedM as “w” once,W as “m” twice, andPas “b,” “d,” or “q” six times.6 The remaining

5 The naming ofz as “n,” which also occurred for uppeaseZ, may be related to the fact that a 90° rotationpper- or lower-casez yields a form very similar to uppeaseN. However, A.H. almost never made 90° rotationsonverbal stimuli.

6 Our assumption that the errors forP resulted fromconfusion of upper-caseP with lower-caseb, d,or q seemplausible in the context of the present experiment, giveneach block of trials included both upper- and lower-cletters. However, if upper- and lower-case letters weresented in separate trial blocks, so thatb, d, andq were nopossible stimuli in blocks containing upper-caseP, we

errors wereI named as “l,”V named as “w,” and

iene.,eerncet-th

theccu

enngualrst,on-lmns

er-tere.

sepo-H.onbae

e-terch

e ithee-

dures were otherwise the same as in Experi-

by20/

nlyn,

reta-

f

seadyeon-

ntreeachkerters.8

,alsings ofandms.ndt—off

ter-

-.’s

ck-bet) inthe

asnly

all

ewasolvcks

tatiopredi-re

u


Z named as “n.”Considering just those letters that are or

tation-critical in lower but not upper case (i.b, d, q, n, u), A.H. made 27 errors on thlower-case forms but no errors on the uppcase forms. This result provides clear evidethat the errors in naming orientation-critical lters arose in processing the visual forms ofletters and not, for example, in retrievingnames of letters that had been identified arately.

Taken together, the results of this experimargue strongly that A.H.’s deficit in perceivithe left–right and up–down orientation of visstimuli affects her perception of letters. Fialmost all of her errors occurred for orientaticritical letters. Second, the errors overwheingly took the form of orientation confusio(e.g.,d 3 “b”; M 3 “w”; p 3 “d”). Finally,performance was dramatically worse for lowcase forms than for upper-case forms of letthat are orientation-critical only in lower cas

Experiment 2: Exposure Duration

In this experiment A.H. named lower-caletters that were flashed for 50 ms (Brief Exsure condition) or remained in view until A.made her response (Long Exposure conditiWe predicted that as in tasks with nonvervisual stimuli, A.H.’s performance would bbetter for brief than for long exposures.

Method. Six blocks of 52 trials were prsented. Each block consisted of randomly inmixed brief- and long-exposure trials, with eaof the 26 lower-case letters appearing oncthe Brief Exposure condition and once inLong Exposure condition. Stimuli and proc

might expect A.H. to nameP correctly even if it werperceived at an incorrect orientation. This predictionconfirmed in two blocked-presentation experiments inving a total of 16 upper-case and 16 lower-case trial bloFor lower-case letters, and upper-caseM and W, blockedpresentation had no apparent effect on the rate of orienconfusions. However, whereas A.H. erred on 6 of 21sentations of upper-caseP under mixed-presentation contions, she made no errors onP over 16 presentations in pu

pper-case blocks.

-

-e

e

-

t

-

s

).l

-

n

ment 1.Results and discussion.As predicted, A.H.’s

performance was substantially improvedbrief exposures. Her error rate was 13% (156) in the Long Exposure condition but o1% (2/156) in the Brief Exposure conditiot(25) 5 2.51, p , .05. In the Long Exposucondition, 19 of A.H.’s 20 errors were oriention confusions (e.g.,q 3 “b”; n 3 “u”); theremaining error wast named as “j.” In the BrieExposure condition, A.H.’s only errors werepnamed as “d” andh named as “b.”

Experiment 3: Flicker

In this experiment A.H. named lower-caletters presented either continuously (Stecondition) or flickering (Flicker condition). Wpredicted better performance in the Flicker cdition than in the Steady condition.

Method. Two versions of the experimewere run. In the full-alphabet version, thblocks of 52 trials were presented, with eblock including one Steady and one Flicpresentation of each of the 26 lower-case letIn the orientation-critical version only theorientation-critical lower-case letters (b, d, p, qn, u, m, w) were tested. Three blocks of 48 triwere presented, with each block containthree Steady and three Flicker presentationeach letter. In both versions the steadyflickering letters were presented for 1000The flickering displays—in this experiment ain the other flicker experiments we reporwere created by turning the stimulus on andat 50 ms intervals throughout the display inval.

Results and discussion.In both the full-alphabet and the orientation-critical versions, A.Hperformance was dramatically better for fliering than for steady letters. In the full-alphaversion she erred on 19% of the trials (15/78the Steady condition but made no errors inFlicker condition,t(25) 5 3.11,p , .01. In theorientation-critical version her error rate w38% (27/72) in the Steady condition, but o1% (1/72) in the Flicker condition,t(8) 5 4.48,p , .01. In both versions of the experiment

-.

n-

of A.H.’s errors were orientation confusions for

ndngmin

on-

oo

chulm

fnc

.andsee.tiordser

-

hdisthatose

t

gdssease

ti-the

The 80 words were randomly divided into twointas-

n aewls. Iner

e; ins to

-ter

eer-

on.g.,naslsl

m-

asate

n-ors

ticaler-

rt

l inas

bution,21tionthatorsers,lted


orientation-critical letters.In conjunction with the effects of case a

exposure duration, these findings argue strothat A.H.’s letter-recognition errors stem frothe same underlying deficit as her errorsperceiving the location and orientation of nverbal visual stimuli.

Experiments 4–6: Reading AloudIndividual Words

These experiments explored the effectsletter case, exposure duration, and flickerA.H.’s reading of individual words.

Experiment 4: Reading Aloud Upper- andLower-Case Words

A.H. was presented with words for whiletter orientation or letter sequence errors colead to confusion with another word. For exaple, the orientation-critical wordduckcould beconfused withbuckor puckif the orientation othe d were misrepresented, and the sequecritical wordsawcould be confused withwasifthe ordering of letters were misrepresented

Words were presented in both upperlower case. For most words we expected cahave little or no effect on A.H.’s performancHowever, because some letters are orientacritical only in lower case, some stimulus wowere orientation-critical in lower but not uppcase. For example, lower-caseduck is orientation-critical, but upper-caseDUCK is not.Whereas thed in duckmight be confused witb or p, the D in DUCK should not be confusewith any other letter even if its orientationmisperceived. Hence, we expected wordswere orientation-critical only in lower caseelicit letter orientation errors with lower-cabut not upper-case presentation, leadinghigher error rates for the lower-case forms.

Method.The stimuli were 80 words ranginin length from 3 to 12 letters. Forty of the worwere orientation-critical in lower case. Of the10 were also orientation-critical in upper c(e.g., WAIL), and 30 were not (e.g.,DUCK).Twenty words (e.g.,saw) were sequence-crical (in both upper and lower case), andremaining 20 words were fillers (e.g.,curve).

ly

fn

d-

e-

to

n-

t

o

,

sets of40—sets A and B—with the constrathat half of the words of each type weresigned to each set.

Words were presented in random order ocomputer monitor, and they remained in viuntil A.H. responded. Four blocks of 80 triawere presented over two testing sessionsBlocks 1 and 4 the set A words were in lowcase and the set B words were in upper casBlocks 2 and 3 the assignment of word setupper and lower case was reversed.

Results and discussion.For lower-case stimuli A.H.’s error rate was 27% (43/160). Letorientation errors (e.g.,bill 3 “pill”; dean3“bean”; wink3 “mink”) occurred on 25 of th80 orientation-critical trials (31%), and oneror on a sequence-critical word (pets 3“debts”) apparently involved an orientaticonfusion. Letter sequence confusions (esaw3 “was”; spot3 “pots”) were observed o8 of the 40 sequence-critical trials (20%),well as on one of the orientation-critical tria(lips 3 “lisp”). A.H. also made 8 additionavisually similar word errors on lower-case stiuli (e.g., thorough3 “through”).

For upper-case stimuli A.H.’s error rate w15% (24/160), reliably lower than the 27% rfor lower-case stimuli,t(79) 5 2.61, p , .05.Letter orientation errors (e.g.,MADE 3“wade”) occurred on 4 of the 20 orientatiocritical trials (20%), and letter sequence err(e.g.,BOARD3 “broad”; SILVER3 “sliver”)were observed for 7 of the 40 sequence-critrials (18%). A.H. also made 13 additionalrors (e.g., PETS 3 “pits”), one of which(FRIEND3 “fired”) may have resulted in pafrom a letter sequence confusion.

For the words that were orientation-criticalower but not upper case, A.H.’s error rate w35% (21/60) with lower-case presentation,only 8% (5/60) with upper-case presentatt(29) 5 3.40, p , .01. Seventeen of thelower-case errors (81%) were letter orientaconfusions. These findings argue stronglyat least some of A.H.’s word-reading errarose in processing the visual forms of lettand specifically that some of the errors resu

from misrepresentations of letter orientation.on-gttetad-heep-

r-rieo-

m.g.,irs

( nl

f-a thfi re-s theo urc t ow edP Ep

s tioo timu ngE theB1

de2

g.,a r

offu-ionheha


alady).s inionothted

rsurehe4)

3 hem .g.,w rs( es er-o

sured cet msf

ofe or-m latedw andfl ucet

E

rds ong(

h airso g,r orts encec

. Ine e int theL ted


More generally, the results for the orientaticritical items, in conjunction with the findinthat A.H. made a substantial number of lesequence errors (e.g.,saw3 “was”), supporthe assumption that A.H.’s impairment in reing words aloud is due at least in part todeficit in visual location and orientation perction.

Experiment 5: Exposure Duration

In this experiment A.H. read aloud lowecase words presented for either 90 ms (BExposure condition) or 1000 ms (Long Expsure condition).

Method.Stimulus lists were constructed fro72 word pairs: 48 orientation-critical pairs (emall and wall) and 24 sequence-critical pae.g.,arid andraid). Individual words ranged iength from 3 to 7 letters.

Two blocks of randomly intermixed briend long-exposure trials were presented. Inrst block one word from each pair was pented in the Brief Exposure condition, andther word was presented in the Long Exposondition; in the second block the assignmenords to exposure conditions was reversrocedures were otherwise the same as ineriment 4.Results and discussion.A.H.’s word reading

howed the same effect of exposure durabserved with letters and nonverbal visual sli: Her error rate was 24% (34/144) in the Loxposure condition but only 9% (13/144) inrief Exposure condition,x2 (1, N 5 288) 51.21,p , .001.In the Long Exposure condition A.H. ma

1 letter-orientation errors (e.g.,bell3 “dell”;mind3 “wind”), 8 letter-sequence errors (e.

rm 3 “ram”; raid 3 “arid”), and 5 othevisually similar word errors (e.g.,cub 3“club”). In the Brief Exposure condition, 5A.H.’s 13 errors were letter-orientation consions, and four were letter-sequence confusOn three of the remaining four error trials sindicated that the stimulus presentationbeen too brief for her to identify the word.

r

r

f

e

ef.

x-

n-

s.

d

In this experiment A.H. read aloud individuwords presented either continuously (Stecondition) or flickering (Flicker conditionStimuli and procedures were the same aExperiment 5, except that the Flicker conditreplaced the Brief Exposure condition. In bconditions the stimulus words were presenfor 1000 ms.

Results and discussion.The effects of flickewere even more dramatic than those of expoduration: A.H.’s error rate was 23% in tSteady condition (33/144) but only 1% (1/14in the Flicker condition,x2(1, N 5 288) 5

4.15, p , .001. In the Steady condition sade 26 letter-orientation errors (eound3 “mound”), 2 letter-sequence erro

e.g., alter 3 “later), and 5 other errors. Thingle error in the Flicker condition was a lettrientation confusion.Taken together, the effects of case, expo

uration, and flicker provide strong evidenhat A.H.’s word reading impairment sterom her visual localization deficit.

Experiments 7–8: Reading AloudWord Sequences

These experiments explored the effectsxposure duration and flicker on A.H.’s perfance in reading aloud sequences of unreords. We predicted that brief exposuresicker would reduce her tendency to prodhe words out of order.

xperiment 7: Exposure Duration

In this experiment A.H. read aloud two-woequences presented in Brief (90 ms) or L1500 ms) Exposure conditions.

Method.The 72 stimulus sequences (e.g.,antat) were generated by taking all possible pf 9 3-letter words (ant, egg, fox, hat, key, loim, set, van). (We used short words and shequences to ensure that the entire sequould be perceived in a 90-ms exposure.)Three blocks of 144 trials were presented

ach block the 72 word pairs occurred onche Brief Exposure condition and once inong Exposure condition. Pairs were presen

in random order, with brief and long exposurelusar-

ee.g).he

nsin

rdsg.,-

r e-p sq

E

rds cod

nsw oms alm a-t erea ost

ithF edw esw thr ond t os eat iewu

i-tuk als( hem

.’sw feco in

visual location and orientation tasks with non-v thath inga uall

uess ita rdsa evi-d er int tlyn lifea

ext,n in-f uta t atm ss,1 ;O as,A &M 94;S est& &G nceoh a-p un-l ofa dgei e an hick thew lw tl ofk byi ont m-u ask

erh theu r by


trials intermixed. On each trial the stimuwords were arrayed horizontally, with two chacter spaces between words.

Results and discussion.In the Long Exposurcondition A.H. made word sequence errors (fox ant3 ant fox) on 55 of the 216 trials (25%In the Brief Exposure condition, however, smade only 1 sequence error (0.5%),t(71) 59.87,p , .001.

In addition to the word sequence confusioA.H. made 5 other errors. In 4 of these (2each condition) A.H. replaced one of the wowith another word from the 9-word pool (e.key ant3 “log ant”). In the fifth error, occuring in the Brief Exposure condition, A.H. rorted that she had not seen the stimulusuence.

xperiment 8: Flicker

In this experiment A.H. read aloud five-woequences presented in Steady and Flickeritions.Method. Two hundred 4- to 7-letter nouere used to construct 40 five-word randequences (e.g.,earth paint puzzle chemicuscle). An effort was made to avoid orient

ion- and sequence-critical words. Words wssigned randomly to sequences and to p

ions within sequences.Two blocks of 40 trials were presented, w

licker and Steady trials randomly intermixithin blocks. In the first block 20 sequencere presented in the Steady condition, and

emaining 20 were presented in the Flicker cition; in the second block the assignmenequences to conditions was reversed. Onrial the stimulus sequence remained in vntil A.H. completed her response.Results and discussion.In the Steady cond

ion, A.H. made word-sequence errors (e.g.,fig-re soccer circle knife cactus3 “figure circlenife soccer cactus”) on 33 of the 40 tri84%). In the Flicker condition, however, sade only 1 word-sequence error (3%),x2(1,

N 5 80) 5 52.4,p , .001.Experiments 7 and 8 demonstrate that A.Hord-sequence reading shows the same eff exposure duration and flicker observed

.,

,

e-

n-

i-

e-fch

ts

erbal stimuli. These results argue stronglyer word sequence errors, like her letter-namnd word-reading errors, arise from her vis

ocalization deficit.

READING COMPREHENSION AND THEVISUAL LOCALIZATION DEFICIT

The evidence we have presented argtrongly that A.H.’s visual localization deficffects her reading, at least for individual wond sequences of unrelated words. Thisence raises a question we introduced earli

he article: How can A.H. achieve apparenormal reading comprehension in her dailynd academic endeavors?

Knowledge-Based Constraints in Reading

In comprehending words, sentences, or tormal readers rely not only on the visual

ormation extracted from the stimulus itself blso on knowledge of the regularities presenany levels in written language (e.g., Fo982; Morris, 1994; Morris & Folk, 1998’Seagdha, 1989, 1997; Potter, Moryadbrams, & Noel, 1993; Potter, Stiefbold,oryadas, 1998; Rayner, Pacht, & Duffy, 19harkey & Sharkey, 1992; Tabossi, 1988; WStanovich, 1986; Williams, 1988; Wrightarrett, 1984). Consider the following sequef words from a hypothetical sentence:theorse’s hoo_.Knowledge of English orthogrhy suggests that the missing letter is very

ikely to be a vowel, although it could be anynumber of consonants. Syntactic knowle

ndicates that the incomplete word may boun, adjective, or adverb; lexical-orthograpnowledge suggests more specifically thatord may behoot, hoof,or hoop; and, generaorld knowledge suggests thathoof is the mos

ikely candidate. These and other sortsnowledge may facilitate comprehensionmposing (typically probabilistic) constraintshe interpretation of information from the stilus itself. We will refer to these constraintsnowledge-based constraints.Several points of clarification may be in ord

ere. First, we do not intend to suggest thatse of knowledge-based constraints—eithe

normal readers or by A.H.—is a consciousthe

totinane-uluves oor

-ly

abtheasn-

l-dartoinis

ratn f

eza-nerthrsis-

sinn-a

calstsew-

be

ingayua

a

weight on knowledge-based constraints. Moreingterialgfulthe

hern-ic,in-

withthe

as

intsase.’sre-

ualdis-y begbe

,are

texte orssesm’s

ificsesl-

ter-sederalalaintfor-eak.tinger-ex-in

themaythatl


problem-solving process, or even thatknowledge exploited is necessarily availableawareness. Second, although we draw a distion between knowledge-based constraintsstimulus information, applying knowledgbased constraints in interpreting some stimelement (e.g., a word or phrase) often involthe use of information about other elementthe stimulus (e.g., neighboring wordsphrases). Third, we use the termknowledgebased constraintsrather than the seemingmore straightforwardcontextual constraintsbe-cause some constraints that could conceivbe brought to bear are not contextual inusual sense of the term (e.g., constraints bon knowledge of simple letter or word frequecies).

In the normal reader, application of knowedge-based constraints may occasionally leamisrepresentation of stimulus elements thatimprobable in context, as when one failsdetect misspellings or syntactic anomaliesproofreading. However, stimulus informationusually weighted heavily enough that accuinternal representations are generated evelow-probability stimuli (e.g.,the horse’s hoop).

For A.H., we suggest, the situation is somwhat different. Because of her visual localition deficit, the stimulus representations geated by her visual system are unreliable:orientations of letters, the ordering of letteand the ordering of words are frequently mrepresented (and presumably have beenthe time of her first exposure to written laguage). The situation A.H. faces in reading mtherefore be similar to that of a hypothetinormal individual who, from his or her firencounter with written language, was expoonly to error-ridden material such as the folloing:

Stupents exhibiting disruptive pehaviormay evlauated learning fro disaqiliites.

Under these circumstances A.H.’s readprocesses may have developed in such a wto place less-than-normal weight on visstimulus information and greater-than-norm

c-d

ssf

ly

ed

toe

eor

-

-e,

ce

y

d

asll

specifically, we propose that A.H.’s readmechanisms in essence assume that the mabeing processed is well-formed and meanin(even when the representations provided byvisual system indicate otherwise), and furtthat these mechanisms skillfully exploit costraints at multiple levels (e.g., orthographsyntactic, semantic) to arrive at a coherentterpretation that is as consistent as possiblethe stimulus information. On this accountvisual stimulus representations are treatedinformative but potentially unreliable clues.

When strong knowledge-based constracan be brought to bear—as is typically the cwith meaningful sentences and text—A.Hreading processes may usually be able tocover from the errors introduced by her vissystem. (Consider, for example, that thetorted sentence in the above example ma“decoded” without too much effort, yieldinStudents exhibiting disruptive behavior mayevaluated for learning disabilities.) Howeverwhen the knowledge-based constraintsweaker—as for words presented out of conor word sequences lacking syntactic structursemantic coherence—A.H.’s reading procemay be unable to overcome her visual systeerrors.

We are not in a position to offer a specproposal about how A.H.’s reading procescombine stimulus information and knowedge-based constraints in arriving at an inpretation. Nevertheless, the knowledge-baconstraints hypothesis offers at least a genaccount of both A.H.’s apparently normreading comprehension under high-constrconditions and her severely impaired permance when the constraints are few and wThe hypothesis also makes some interespredictions, which we tested in several expiments. However, before reporting theseperiments we briefly discuss A.H.’s errorsreading individual words, suggesting thatknowledge-based constraints hypothesisexplain some aspects of the error patterndo not follow obviously from A.H.’s visualocalization deficit.

A.H.’s Word-Reading Errors

icher

on38ere

ncd

ons.,g.,

)ove

ncethab

ep-lu

rdrroforingeree

, oordysno

ceulu

esiog

tterdtho-noacniin

he

representation generated by the visual systemnceone-ghtif-

ingayer-an

nfu-

I thev -u np itht ryf onsm

ra-ionon-nti-thecticcent

weh rrors.F um-a ittenl irelyn withb le tod enta-t ntro-d r-se-q n oft ver,i ntrib-u tiale caseo ret-i


Collapsing over the various tasks in whA.H. read individual lower-case words undordinary conditions (i.e., long exposures of nflickering stimuli), she made a total of 2errors. Ninety-five of these errors (40%) wpure letter-orientation confusions (e.g.,dear3“bear”), and 22 (9%) were pure letter-sequeconfusions (e.g.,mar3 “ram”). One hundreeight of the errors (45%) were other typesvisually similar word errors: letter deletio(e.g., finger3 “finer”), letter insertions (e.gnose3 “noise”), and letter substitutions (e.ship3 “snip”). The remaining 32 errors (13%appeared to involve two or more of the aberror types (e.g.,sled3 “sleep”; shoulder3“shudder”).

The letter-orientation and letter-sequeconfusions may be interpreted by assumingan orientation or sequence error introducedA.H.’s visual system resulted in a stimulus rresentation corresponding not to the stimuword (e.g.,bone) but instead to a different wo(e.g.,done). However, some aspects of the epattern cannot be interpreted so straightwardly. Most notable among these is the findthat all of A.H.’s erroneous responses wwords. Clearly, we cannot assume that therors introduced by her visual system alwayseven usually, transformed the stimulus winto another word; presumably, the visual-stem errors often resulted in a representationcorresponding exactly to any word (e.g.,finegrfor the stimulusfinger, slepfor sled, rbotherforbrother). Why, then, did A.H. never produnonword responses, or report that a stimwas a nonpronounceable letter string?

The knowledge-based constraint hypothsuggests a possible answer: A.H.’s word recnition process implicitly assumes that any lestring encountered in reading is a familiar woand selects as a match for the stimulus wordmost strongly activated entry in the orthgraphic lexicon, even when this entry doesfully match the stimulus representation. Bycepting imperfect matches, the word recogtion process would probably often succeedcorrectly identifying a stimulus word even if t

-

e

f

ty

s

r-

r-r

-t

s

s-

r,e

t--

contained letter-orientation or letter-sequeerrors. However, in some instances the errous representation of a stimulus word mimost closely match the lexical entry for a dferent word, leading to an error.

This interpretation accounts for the findthat all of A.H.’s errors were words, and malso explain why approximately half of therors were visually similar word errors other thletter-orientation and letter-sequence cosions (e.g.,finger3 “finer”; sled3 “sleep”).f, for example, a letter sequence error inisual processing offinger resulted in the stimlus representationfinegr, the word recognitiorocess might match this representation w

he lexical entry forfiner rather than the entor finger; similarly, the stimulus representatilep might activate the lexical entry forsleepore strongly than the entry forsled.7

Tests of the Knowledge-BasedConstraint Hypothesis

Experiment 9: Reading Paragraphs

In this experiment A.H. read aloud brief pagraphs adapted from various works of fictand nonfiction. Paragraphs in the Normal cdition were syntactically correct and semacally coherent. However, for paragraphs inSequence-Altered condition several syntaerrors were introduced by transposing adjawords, as in the following example:

7 It is certainly possible that factors other than thoseave discussed also contributed to the word-reading eor example, given that A.H.’s visual deficit was presbly present from the time she was first exposed to wr

anguage, her orthographic lexicon may not be entormal (although her good word-reading performancerief or flickering stimuli suggests that she has been abevelop at least reasonably accurate lexical repres

ions). Also, her visual system may have sometimes iuced errors other than letter-orientation and letteuence errors (e.g., errors involving misrepresentatio

he relative locations of features within a letter). Howen the absence of clear evidence that such factors coted to her performance, we will focus on the potenffects of letter-orientation, letter-sequence, and (in thef multiword stimuli) word-sequence errors when interp

ng A.H.’s reading impairments.

The policeman always sat on his horse halfway up the

shedt

wa

ldoth

-issstin

nt ib

the

e-nweoi

–8ereanor

twncrdsara

nteradal

tionwawa

the.

ees

/mig-ra-

graphs the control subjects made an average ofns,ualnotheed

ra-olrors,verr-

, 9

-redate.enedrorser

.g.,d

st,ne ofrors,sestheell

rmal

bil-enell

the-vail-toin

theual

ver-ssi-ord

ra-


first block from subway the station. For as manymornings as he remember could Arthur had alwayand stopped given the horse three sugar cubes. Thorse had learned to him recognize and always paweat the curb and stretched out his eagerly neck aArthur’s approach. It was the nicest in thing Arthur’sday.

A.H.’s task was to read each paragraph as itwritten.

Our principal prediction was that A.H. woufrequently restore the transposed words tsyntactically acceptable order when readingSequence-Altered paragraphs (e.g.,he remember could3 “he could remember”). The basfor this prediction is straightforward: UnleA.H.’s reading processes can somehow disguish a word sequence error actually presea text from a word sequence error introducedher visual system, the former as well aslatter should frequently be repaired.

Method.Stimuli were 10 Normal and 10 Squence-Altered paragraphs, each printed oseparate page. The two sets of paragraphsmatched in number of words and numbersentences. Individual paragraphs rangedlength from 61–94 words and included 4sentences. In each of the 10 Sequence-Altparagraphs 6 pairs of adjacent words were trposed to create syntactically unacceptable wsequences. In all 60 transpositions thetransposed words were in the same senteand in 58 of the transpositions the two woappeared on the same line in the printed pgraph.

The 10 Normal paragraphs were presefirst, followed by the 10 Sequence-Altered pagraphs. (This presentation order was adopteexclude the possibility that reading of Normparagraphs would be influenced by recognithat some paragraphs had errors.) A.H.instructed to read each paragraph aloud as itwritten; no mention was made of errors intexts. Five control subjects were also tested

Results and discussion.Considering first thNormal paragraphs, control subjects read thparagraphs at a mean rate of 219 words(wpm), with rates for individual subjects raning from 196 to 232 wpm. Across the 10 pa

s

ae

-ny

arefn

ds-doe,

-

d-to

ss

en

11.8 reading errors, including word omissioword insertions, and misreadings of individwords. The vast majority of these errors didsignificantly alter the meaning of the text. Tnumber of errors for individual subjects rangfrom 1 to 23.

A.H.’s reading speed for the Normal pagraphs—227 wpm—fell within the contrrange. However, she made 47 reading erfour times the control subjects’ mean, and otwice the control-subject maximum. A.H.’s erors included 9 word omissions (e.g.,he hadplayed with them and had come to know3 “hehad played with them and come to know”)word insertions (e.g.,as if3 “just as if”), and11 misreadings of individual words (e.g.,sleeping3 “sleepy”). These types of errors occurat roughly three times the control-subject rHowever, the most striking difference betweA.H.’s errors and those of the controls involvword sequence confusions. A.H. made 18 erof this type, most involving words whose ordwas not tightly constrained by context (espeed and determination3 “determination anspeed”; fine, yellowed marble3 “yellowed,fine marble”;large and dark and cool3 “dark,cool, and large”). Control subjects, in contraaveraged 1.8 word sequence errors, and nothese subjects made more than 4. A.H.’s erlike those of the control subjects, in most cadid not substantially alter the meaning oftext. Therefore, her comprehension may whave been good despite her higher-than-noerror rate.

Nevertheless, it seems clear that A.H.’s aity to read material verbatim is not normal, evfor connected text. This finding meshes wwith the knowledge-based constraint hyposis. In the case of coherent paragraphs the aable constraints may usually be sufficientallow a reasonably accurate interpretation—particular, an interpretation that preservesmeaning—to be generated from distorted visrepresentations. However, a fully accuratebatim reconstruction may not always be poble (as in the case of underdetermined worders likelarge and dark and cool).

Turning now to the Sequence-Altered pa

graphs, control subjects read these paragraphs atwea

gto

thecocepmhsof

thetheb-

thethe.g.

rs

typ

air. Inthrete14ing

ubs ioil-

one-hadeirsosua

irsrol3 otely

se-g 6rror..H.heir

irFory

d-ty.”cor-er-

ired51ablekedhadara-ast,of

up-ro-tionnd

ts toem.pro-ishorsfore

TABLE 3

l

rror

AC


an average speed of 175 wpm, 44 wpm slothan their mean rate of 219 wpm for the Normparagraphs,t(4) 5 14.83, p , .001. Readinrates for individual subjects ranged from 160191 wpm. A.H. also slowed her reading onaltered paragraphs, but not as much as thetrol subjects: Her reading rate for SequenAltered paragraphs was 205 wpm, only 22 wslower than her rate for the Normal paragrap

Control subjects read the critical pairstransposed words (e.g.,him recognizein Thehorse had learned to him recognize) with amean accuracy of 72%; that is, on averageread 43.2 of the 60 pairs as written, inanomalous order. Accuracy for individual sujects ranged from 57 to 88%. For 24% ofcritical pairs, the control subjects restoredtransposed words to their “proper” order (epeople were who frightened3 “people whowere frightened”); we will refer to such erroassequence repair errors.On the remaining 4%of the pairs the controls made some otherof error.

Individual control subjects’ rates of reperrors ranged from 8% (5/60) to 37% (22/60)31 of the 71 total sequence repair errors,subject realized the error immediately, andread the sequence accurately. The mean rauncorrected sequence repair errors was(8.2/60), with individual subjects’ rates rangfrom 0 to 25%.

These results indicate that the control sjects exploited knowledge-based constraintreading the paragraphs, sometimes to the pof failing to read the material verbatim. Athough the subjects eventually detected 86%the word transpositions, they initially read ofourth of the transposed pairs in an order twhile highly probable, was not the actual orin the text, and for 14% of the transposed pathe controls produced the words in the mprobable order without realizing that the actorder was different.

A.H.’s performance on the critical word pais presented in Table 3, along with the contsubject means. Remarkably, A.H. read onlythe 60 transposed word pairs (5%) accura

rl

n--

.

y

,

e

e-of%

-nnt

f

t,r,tl

-f.

For 51 of the 60 pairs (85%) she made aquence repair error, and for the remaininpairs (10%) she made some other type of eIn 42 of the 51 sequence repair errors Asimply restored the transposed words to t“proper” order (e.g.,favorite their foods3“their favorite foods”); in the other 9 repaerrors she revised the text more extensively.example, she readThis is why so children manwho live in towns are naughty(with transposewordschildrenandmany) as “That is why children who live in many towns are so naughWhereas the control subjects noticed andrected nearly half of their sequence repairrors, A.H. made no such corrections.

A.H. was remarkably fluent as she repathe word transpositions; for only 8 of thesequence repair errors was there any noticehesitation or stumbling over words. When asat the end of the experiment whether shenoticed anything unusual about any of the pgraphs, A.H. said that she had not. (In contrall 5 control subjects were very much awarethe word transpositions.)

The results of this experiment strongly sport the hypothesis that A.H.’s reading pcesses, operating under the implicit assumpthat the material being read is well-formed acoherent, apply knowledge-based constrainrepair errors introduced by her visual systThe results further suggest that the readingcesses have little or no ability to distinguerrors actually present in a text from errproduced by the visual system, and thererepair the former as well as the latter.

Performance of A.H. and Control Subjects on CriticaWord Transpositions in Experiment 9

Subjects

Response distribution on critical wordtranspositions

Accurate Sequence repair error Other e

.H. 5% 85% 10%ontrols 72% 24% 4%

Experiment 10: Sentence Acceptability

atowpremeassthe.hetheulder.

.pt

s-acenabllseaelf

hentr

bUnecthesennacweencby

hare

lydgpt-tra) o

bound morphemes (2 sentences) of an initiallynhise-cingola-ap-

nd

e onandIn-to bethe

re

entrolanden-de anceged%.onother

heyad-ence

en-troler-

d

ces,ses16msor

ncetonac-

washad


Judgments

In Experiment 9 A.H. was not informed ththe stimulus paragraphs contained errors. Hever, in Experiment 10 sentences weresented with the explicit instruction that socontained clear grammatical errors. A.H. wtold to read each sentence aloud exactly awas written, and then to indicate whethersentence was grammatically acceptableA.H.’s reading processes operate withoutawareness or control to correct errors ofsorts introduced by her visual system, we woexpect her to have difficulty detecting suchrors even when explicitly instructed to do so

Three types of sentences were presentedAc-ceptablesentences were syntactically acceable and semantically coherent (e.g.,The boywere swimming across the lake). UnacceptableSequencesentences were syntactically unceptable due to transposition of two adjacwords or morphemes in an otherwise acceptsentence (e.g.,My parents are doing they acan to help me). Unacceptable-Othersentencewere syntactically unacceptable for other rsons (e.g.,My brother learned to dress hersby the age of three).

We expected A.H. to perform well on tAcceptable sentences, because errors iduced by her visual system should usuallyrepaired by her reading processes. For theacceptable-Sequence sentences we expA.H.’s reading processes to repair many ofword sequence errors without her awarenescontrol, leading her to judge many of the stences grammatically acceptable. For the Uceptable-Other sentences our expectationsless clear; because the errors in these sentwere not of the sorts frequently introducedA.H.’s visual system, we were uncertain to wextent her reading processes would effectpairs.

Method. The stimuli were 20 syntacticaland semantically acceptable sentences ansyntactically unacceptable sentences ranginlength from 6 to 13 words. The 16 Unacceable-Sequence sentences were created byposing two adjacent words (14 sentences

--

it

Ifr

-

-

-tle

-

o-e-

ted

or--rees

t-

37in

ns-r

grammatical sentence (e.g.,The dog was whegone we arrived at home; I was surprised byfather’s boynessish). The 21 UnacceptablOther sentences were generated by introduother types of syntactic errors, such as vitions of number or gender agreement, or inpropriate use of prepositions (e.g.,We ran intoseveral friend at the pool this weekend; I fouthe book and gave it at him).

Sentences were presented one at a timpaper. A.H. first read the sentence aloud,then indicated whether it was acceptable.structions stressed that the sentences wereread exactly as they were written, and thaterrors were not subtle flaws such as usingwhoinstead ofwhom. Five control subjects wealso tested.

Results and discussion.For the Acceptabland Unacceptable-Other sentences the cosubjects were 99% correct both in readingjudging acceptability. For each of these stence types, one of the control subjects masingle error. On the Unacceptable-Sequesentences, control-subject performance ranfrom 94% to 100% correct, with a mean of 96Two of the control subjects made no errorsthe Unacceptable-Sequence sentences; the3 controls each made one error in which trepaired the ungrammatical word order in reing a sentence, and then judged the sentacceptable.

A.H.’s performance on the Acceptable stences was comparable to that of the consubjects. She made only one minor readingror (It’s a beautiful day, isn’t it?3 “It’s abeautiful daytoday isn’t it?”), and she judgeall of the sentences acceptable.

For the Ungrammatical-Sequence sentenhowever, A.H. made correct reading responand acceptability judgments for only 3 of thesentences (19%). For the remaining 13 ite(81%) she repaired the ungrammatical wordmorpheme order when reading the sentealoud (e.g.,My parents are doing they all canhelp me3 “My parents are doing all they cato help me”) and then judged the sentenceceptable. In most instances A.H.’s readingfluent, and she appeared unaware that she

altered the ordering of words or morphemes.thapathssH.’thFoheun

) srer-

ge

e-81suomisuhere

sulrrosemantlynaenhe

cethe

g

ainis

adts aseuauldailngthe

Consider the following sentence:The Smiths-

beors

-on-orolv-

e byA

n-ndsandther

tspair

e

n-icalge-f theingpa-und

s

es-n ofrts-

btime

o terA d ont andA tion.F

tsm se-q ticalw Fore ts-b d


These results provide additional evidenceA.H.’s reading processes operate to reanomalies in stimulus representations, andthe repairs occur outside of A.H.’s awarene

On the Unacceptable-Other sentences A.performance was better, although far belowperfect performance of the control subjects.13 of the 21 items in this condition (62%) sread the sentence accurately and judged itacceptable. For the other 8 sentences (38%repaired the syntactic error in her readingsponse (e.g.,My brother learned to dress heself by the age of three3 “My brother learnedto dress himself by the age of three”) and judthe sentence acceptable.

The difference in rate of repair errors btween Unacceptable-Sequence sentences (and Unacceptable-Other sentences (38%)gests that A.H.’s reading processes may in ssense be tuned to the types of errors her vsystem frequently introduces. In particular,reading processes may be highly skilled atpairing word sequence errors, with the rethat she rarely became aware of these ewhen reading the Unacceptable-Sequencetences. In contrast, the reading mechanismsbe much less adept at repairing infrequeencountered errors such as those in the Uceptable-Other sentences, with the consequthat these errors were more likely to come toattention.

In the next experiment we used sentenwithout errors to test another prediction ofknowledge-based constraint hypothesis.

Experiment 11: Reading and ComprehendinSentences

According to the knowledge-based constrhypothesis, A.H.’s reading comprehensiongood for meaningful material because her reing processes can usually exploit constrainmultiple levels (e.g., orthographic, syntactic,mantic) to repair errors introduced by her vissystem. This account predicts that A.H. shoshow impaired comprehension when the avable constraints are insufficient for identifyior repairing visual-system errors that altermeaning of the stimulus material.

tirat.ser

-he-

d

%)g-ealr-trsn-y

c-cer

s

t

-t

-l

-

moved from Pittsburgh to Philadelphia.Syntactic and semantic constraints would probablysufficient for repairing word sequence errinvolving some parts of the sentence (e.g.,Themoved Smiths from. . .). However, in the absence of additional context the available cstraints would be inadequate for identifyingrepairing at least some sequence errors inving the wordsPittsburghandPhiladelphia.For

xample, if these words were transposed.H.’s visual system (yieldingfrom Philadel-

phia to Pittsburgh) knowledge-based costraints would presumably provide no groufor determining that an error had occurred,the sentence would be misinterpreted. For opossible sequence errors (e.g.,from Philadel-phia Pittsburgh to) the available constrainmight be consistent with more than one re(e.g., from Pittsburgh to Philadelphiaor fromPhiladelphia to Pittsburgh), again creating thpossibility of a comprehension error.

Method. Thirty-eight sentences were costructed such that (a) the ordering of two critwords could not be predicted from knowledbased constraints, and (b) the actual order owords was crucial for determining the meanof the sentence. The critical words were serated by one word in 34 sentences, by a bomorpheme in 3 sentences (e.g.,My husband’mother is in town for the game), and by acomma in 1 sentence (According to Ed, Bill iscoming at 4:00 p.m.). For each sentence, a qution was constructed to probe comprehensiothe critical word sequence (e.g.,Where is theinew home?for The Smiths moved from Pit

urgh to Philadelphia).The sentences were presented one at a

n a computer monitor in random order. Af.H. read the sentence aloud, it was replace

he monitor by a comprehension question,.H. gave a spoken response to the quesive control subjects were also tested.Results and discussion.The control subjecade no errors in reading critical worduences. A.H., however, transposed the criords in 9 of the 38 sentences (24%).xample, she readThe Smiths moved from Piturgh to Philadelphiaas “The Smiths move

from Philadelphia to Pittsburgh” andMy hus-

troh aontheinomr 8thas

nsad

aya

thesn-sehe

s,ennd

ing

Exal-ofte

ar tke

inalge-nsialstthrend

block the assignment of sentences to conditions

ate3/

(2/

3 ngob-

romat

an-

ingnt,ntsom-

in

.g.,-

aingulip,oth-is-

terstualher

rd-

ghtrs,1–3ingherfu-

n ofe

seeEx-vi-r—


band’s mother is in town for the gameas “Mymother’s husband is in town for the game.”

On the comprehension questions the conranged in accuracy from 87% to 100%, witmean of 95%. A.H., however, was correctonly 68% of the questions (26/38). For 25 of29 trials in which she read the critical wordsthe correct order (86%), she answered the cprehension question correctly. However, foof the 9 sentences in which she transposedcritical words her answer to the question wconsistent with the erroneous reading respoand therefore incorrect. For example, after reing At the party Mary gave John a birthdpresent as “At the party John gave Marybirthday present” she responded “Mary” toquestion,Whose birthday was it?These resultconfirm the prediction that A.H.’s comprehesion would be impaired when knowledge-baconstraints were too weak to remedy errorsvisual system was likely to introduce.

In addition to transpositions of critical wordA.H. made 10 other errors in reading the stences. Most of these were misreadings of ividual words (e.g.,Bert 3 “Bret”), althoughtwo were word sequence confusions involvnoncritical words (e.g.,The toy car3 “The cartoy”).


If the word-sequence errors observed inperiment 11 resulted from A.H.’s visual locization deficit, then flickering presentationsentences should reduce the error rate. Tothis prediction we presented sentences similthose from Experiment 11 in Steady and Flicconditions.

Method.The stimuli were 120 sentenceswhich the ordering of two or more criticwords was not predictable from knowledbased constraints (e.g.,Lucy said Betty was aextraordinary person). Two blocks of 120 trialwere presented, with Steady and Flicker trrandomly intermixed within blocks. In the firblock half of the sentences were presented inSteady condition, and the other half were psented in the Flicker condition; in the seco

ls

-

e

e,-

dr

-i-

-

stor

s

e-

was reversed.Results and discussion.Flicker dramatically

affected A.H.’s reading performance. The rof critical word transpositions was 28% (3120) in the Steady condition but only 2%120) in the Flicker condition,x2(1, N 5 240)5

2.1, p , .001. This result provides stroevidence that the word sequence errorsserved in the absence of flicker stemmed fA.H.’s visual localization deficit, and hence ththe deficit is implicated in her reading of meingful material.

GENERAL DISCUSSION

We have analyzed in some detail the readabilities and disabilities of a university studeA.H. Although A.H.’s academic achievemeand self-report suggested normal reading cprehension, she was profoundly impairedreading aloud isolated words (e.g.,box 3“pox”) and sequences of unrelated words (eshore lemon picnic fabric3 “lemon shore picnic fabric”).

Our previous studies with A.H. revealedremarkable developmental deficit in perceivthe location and orientation of visual stim(McCloskey et al., 1995; McCloskey & Rap2000). On the basis of these studies we hypesized that when reading, A.H. frequently mperceives the orientation and ordering of letand the ordering of words. These perceperrors, we suggested, were responsible forimpaired performance in the word and wosequence reading tasks.

To test this interpretation we carried out eiexperiments in which A.H. read aloud lettewords, and word sequences. Experimentsdemonstrated that A.H. was impaired in namisolated letters and that the vast majority oferrors took the form of letter orientation consions (e.g.,p3 “b”; m3 “w”). Experiment 1also showed that when accurate perceptioletter orientation was critical for identifying thlower-case form (e.g.,d) but not the upper-caform (e.g.,D) of a letter, A.H.’s performancwas much worse for the lower-case form.periments 2 and 3 showed further that twosual variables—exposure duration and flicke

had the same effects in the letter-naming task asr-aso-usimanndpodinncndg

ead

ionor-mi.’se on-tenneblcteb

ultthes oromusar

, buo

estsarath

ys(e.g

oporso.Hinttha

reliance on knowledge-based constraints was inom-

ad-iredinged

ayatsortn-likese-def-

al

on-icalonp-ntsar-menkef-

forcer-al-

enility.ateriseost

The&

82;thatno-c-

93;ys-em

g aha-um-


in tasks with nonverbal visual stimuli. In paticular, A.H.’s letter-naming performance wdramatically better for brief than for long expsures, and for flickering than for continuodisplays. Experiments 4–6 demonstrated slar effects of case, exposure duration,flicker on A.H.’s reading of isolated words, aExperiments 7–8 demonstrated that brief exsure or flicker virtually eliminated A.H.’s worordering errors in the word-sequence readtask. These results provide compelling evidethat A.H.’s deficit in perceiving the location aorientation of visual stimuli is the underlyincause of her impaired performance on the ring tasks.

We then considered whether this concluscould be reconciled with A.H.’s seemingly nmal reading comprehension in her acadepursuits and daily life. We proposed that A.Hreading processes rely heavily on knowledgregularities at multiple levels in written laguage to constrain interpretation of the potially inaccurate stimulus representations geated by her visual system. When the availaconstraints are strong, as in reading connetext, A.H.’s reading processes may usuallyable to “repair” the visual-system errors, resing in good comprehension. However, whenconstraints are weak, as with isolated wordsequences of unrelated words, recovery fvisual errors may often not be possible. Theof knowledge-based constraints is not, wegued, a conscious problem-solving processrather proceeds without A.H.’s awarenesscontrol.

The knowledge-based constraints hypothwas tested in Experiments 9–12. Experimenand 10 demonstrated that when A.H. read pgraphs or sentences containing errors ofsorts frequently introduced by her visual stem, she spontaneously repaired the errorsfrom subway the station3 “from the subwaystation”). Further, she was usually unawarethe errors and rarely could detect and rethem even when explicitly instructed to doThese experiments provided evidence that Arelies heavily on knowledge-based constrain reading. Experiment 11 demonstrated

i-d

-

ge

-

c

f

-r-ed

e-

r

e-t

r

is9-

e-.,

ft

..st

fact necessary for her to achieve adequate cprehension of meaningful material: A.H.’s reing comprehension was found to be impafor meaningful sentences in which the orderof two critical words could not be determinfrom the available constraints (e.g.,Mary andJohnin At the party Mary gave John a birthdpresent). Finally, Experiment 12 showed thA.H.’s reading errors on sentences of thiswere virtually eliminated with flickering setence displays, indicating that these errors,her errors with letters, words, and wordquences, arose from her visual localizationicit.

Implications for the Study of DevelopmentCognitive Deficits

The present study provides a strong demstration that the theoretical and methodologtools of cognitive science can shed lightunderlying cognitive dysfunctions in develomental cognitive impairments. The experimewe have reported allowed us not only to chacterize A.H.’s reading performance in sodetail but also, and more importantly, to lithis performance convincingly to a specific dicit of visual perception.

What are the implications of our resultsunderstanding developmental dyslexia? Wetainly do not suggest that impaired visual locization is the underlying deficit in all or evmany cases of developmental reading disabHowever, the results from A.H. do demonstrthat a developmental reading deficit can afrom causes other than those posited by mcurrent theories of developmental dyslexia:most widely accepted theory (e.g., BradleyBryant, 1978, 1983, 1985; Liberman, 19Shankweiler & Liberman, 1972) assumesdevelopmental dyslexia is caused by a phological deficit, and the principal alternative acount (e.g., Galaburda & Livingstone, 19Lovegrove, 1993) attributes dyslexia to a dfunction of the transient visual subsyst(which appears to be intact in A.H.).

Reading is a complex process requirinmultitude of perceptual and cognitive mecnisms, and reading impairments could pres

ably arise from deficits affecting any of theseloprabc-

enezeovenetiod iapveesee

ed)ing

aseid-esud&

on9;a

e ofi-

therege

en-n-o-ntatal

hereraairingfi-

prire

ry.rangats ip

would surely also have been misdirected. Cer-.H.

uesof

inginc-sub-nd-

st,cha-the

.H.’sn-

andssesub-rfor-tesrma-

nc-sub-omm-n ofe-

portrdousm in

aithn-tion.rd-

toath-

lsoreults:inusined

on,rief


mechanisms. Indeed, the literature on devemental reading disorders suggests considevariation in the underlying cognitive dysfuntions (e.g., Ellis, 1985; Martin, 1995). Givthese points, the search for a single causdevelopmental dyslexia, which characterimost research in the area, is unlikely to prfruitful. Furthermore, the probable heterogeity of the dyslexia category undercuts the ranale for the group-study approach adoptemost developmental dyslexia research. Thisproach, in which results are averaged ogroups of subjects classified as dyslexic, ron the assumption that the groups (and indthe population from which they were samplare homogeneous at the level of underlydysfunction.

An alternative approach involves single-canalyses of underlying deficits in many indivuals with developmental reading disabilitiThis approach, exemplified by the present st(see also Broom & Doctor, 1995; CastlesColtheart, 1993; Rayner, Murphy, Henders& Pollatsek, 1989; Snowling & Hume, 198Temple & Marshall, 1983), should providebasis for characterizing the extent and naturvariation among dyslexics in underlying decits, and it may also contribute to resolvingapparent contradictions among many of theported results that are based on group avera

Achieving an understanding of the pottially diverse underlying dysfunctions is of cetral importance in developing effective diagnsis and treatment methods for developmedyslexia (or any other form of developmencognitive impairment). For example, duringelementary school years A.H. endured sevefforts to remediate her severe spelling impment (which apparently stems, like her readimpairments, from her visual localization decit). The remediation programs consistedmarily of exercises such as copying wordspeatedly or looking up words in a dictionaGiven that these exercises required accuperception of written words, it is unsurprisithat A.H. derived no apparent benefit. Hadtempts been made to address her difficultiereading letters or isolated words, these attem

-le

ofs

--n-rtsd

.y

,

f

-s.

l

l-

--

te

-nts

tainly, no one would have suggested that Aread under a strobe light.

Implications for Research onNormal Cognition

Our results also have implications for issin the study of normal cognition. The effectsflicker and exposure duration on A.H.’s readprovide new evidence concerning the disttion between sustained and transient visualsystems we drew on the basis of previous fiings from A.H. (McCloskey et al., 1995). Firthe present results suggest that reading menisms have access to information from bothsustained and the transient subsystems. Aimpaired performance with long-duration, cotinuous presentations of letters, words,word sequences implies that reading procereceive input from the (impaired) sustained ssystem, whereas her dramatically better pemance with brief or flickering displays indicathat the reading processes also receive infotion from the (intact) transient subsystem.

The present findings also speak to the futions served by the sustained and transientsystems. McCloskey et al. (1995) argued frA.H.’s performance with nonverbal visual stiuli that both subsystems compute the locatiovisual stimuli. The results from the word-squence reading task provide additional supfor this conclusion. A.H.’s high rate of wosequence errors with long-duration continudisplays implicates the sustained subsystecomputing the relative locations of words instring; her virtually error-free performance wbrief or flickering displays implies that the trasient subsystem also computes relative loca

The results from the letter-naming and woreading tasks go beyond previous findingssuggest that both sustained and transient pways contribute not only to localization but ato recognition of visual stimuli. The logic heis the same as for the word-sequence resA.H.’s frequent misidentification of letters (isolation or in words) with long, continuostimulus exposures suggests that the sustasubsystem is implicated in letter recognitiand her vastly improved performance with b

or flickering displays suggests that the transienttio

cuheinoanal

sinerc-el,salierth-ss-traayfortim

onr-ouansinn-

ance)ithith”in

zet”

bide”incith

thadusesalnctas

ess

ing stream in fact require both. For example,atiall ar-ightrsalob-ast

reascessomeeseourthatem-

ndthe

nce-evi-ay aceead-hatl inheola-

adien-e er-

ter-

er-

d toer-

the.

ined

d asf oonw t iss Thea ct thee rl ivens


subsystem also plays a role in the recogniprocess.

These conclusions raise questions aboutrent conceptions of visual pathways and tfunctions. Two widely adopted (although bymeans universally accepted) assumptionsrelevant for our purposes. The first (e.g., Ugerleider & Mishkin, 1982) is that the dorsvisual pathway mediates spatial proces(“where”) whereas the ventral pathway undlies visual object recognition (“what”). The seond assumption (e.g., Livingstone & Hub1988; Livingstone, 1990) is that the dorstream receives input primarily from an earvisual pathway (the magnocellular, or M, paway) that is specialized for low-acuity proceing of transient stimuli, whereas the venstream’s input comes primarily from a pathw(the parvocellular, or P, pathway) tunedhigh-acuity processing of more sustained suli.

At least in combination, these assumptiare difficult to reconcile with A.H.’s perfomance. The present findings, and those fromprevious studies of A.H., suggest that both trsient and sustained stimuli can drive procesin both “what” tasks (e.g., letter and word idetification) and “where” tasks (e.g., localizingobject on a table or words within a sequenFurther, A.H. shows poor performance wsustained stimuli and good performance wtransient stimuli in both “what” and “wheretasks, a pattern that is difficult to interpretterms of pathology affecting a system organiinto M/dorsal/“where” and P/ventral/“whapathways.

These findings might be accommodatedassuming that both M and P pathways provinput to both dorsal/“what” and ventral/“wherstreams, or by recasting the functional disttion between the two cortical streams. Wrespect to the latter option, one might positthe dorsal and ventral streams are more redant in their functions than current hypothesuppose. Another possibility is that the dorand ventral-stream functions, although distiare such that cognitive processes typicallysumed to require only one or the other proc

n

r-r

re-

g-

l

l

-

s

r-g

.

d

ye

-

tn-s-,--

object recognition requires considerable spprocessing (e.g., in representing the spatiarangement of features in an object). One mtherefore consider the possibility that the doand ventral streams interact in recognizingjects, such that the former carries out at lesome of the required spatial analysis, whethe latter supports other aspects of the pro(perhaps including aspects that are in ssense spatial). Detailed discussion of thpoints is beyond the scope of this article;intent here is merely to suggest directionscould be pursued in future theoretical andpirical work.

Finally, the results from our paragraph- asentence-reading experiments (includingcontrol subjects’ performance on the sequealtered paragraphs) add to the availabledence that knowledge-based constraints plsignificant role in reading. A.H.’s performanin these experiments, as well as her good ring comprehension in daily life, suggest tthese constraints are often quite powerfumeaningful material. Just how powerful tconstraints can be, even for sentences in istion, is illustrated by a task in which A.H. resentences, some of which included letter ortation, letter sequence, and word sequencrors, as in the following examples:

The young boy took the quppy long on afuoon walks, often stopping to fetch paly.

The stock market is storngly affected dy dcebtions the the of pulbic.

The agency uses comquter models ansimluations prebict the effects of human itnaction the with bay.

The computer lab is otfeu durign bnsyweek lats of calsess any in given esmseter8

A.H. was told that some sentences conta

8 Prior to introduction of errors, these sentences reaollows: The young boy took the puppy on long afternalks, often stopping to play fetch; The stock marketrongly affected by the perceptions of the public;gency uses computer models and simulations to prediffects of human interaction with the bay;andThe compute

ab is often busy during the last week of classes in any gemester.

errors and that her task was to read the sen-, saddinrorshth

verita-ithlesw

irbynddi-ariern-

ncegesaya-

conbe

ryard

nds

in

ofisuarma

B talf ay,

C op-

E el-ey.

F

Galaburda, A. M., & Livingstone, M. (1993). Evidence for. In

ofner-pha-

of

L .L.

:.

L ad-.),.

L ve-nat-I.-.

L m,gy,

L sualM.

em-fer-

M en-

M edsualu-

M on,at-. J.,al-

M tionir-,

M llelf

M ).th-

M nce


tences as they were supposed to be; that iswas instructed to correct the errors while reing. She was also asked to indicate after reaeach sentence whether it contained any erHer performance was remarkable. Althoughstumbled occasionally, she read most ofsentences (including the examples above)smoothly, repairing the errors without hestion. Further, for 60% of the sentences werrors (18/30), again including the exampgiven above, she reported that the sentenceentirely correct.

Although impressive, A.H.’s ability to repaerrors can with practice be approximatednormal individuals. Our own experience, ainformal experimentation with colleagues, incates that sentences with multiple errorsquite difficult at first but become much easafter some practice. (Normal individuals cotinue, however, to be aware that the senteare full of errors.) These observations sugthat rather minimal stimulus information moften suffice in the reading of meaningful mterial, and that the use of knowledge-basedstraints to minimize stimulus processing maya significant component of reading skill.

REFERENCES

Bradley, L., & Bryant, P. E. (1978). Difficulties in auditoorganization as a possible cause of reading backwness.Nature,271,746–747.

Bradley, L., & Bryant, P. E. (1983). Categorizing souand learning to read—A causal connection.Nature,301,419–421.

Bradley, L., & Bryant, P. E. (1985).Rhyme and reasonreading and spelling.Ann Arbor, MI: Univ. of Mich-igan Press.

Breitmeyer, B. G., & Ganz, L. (1976). Implicationssustained and transient channels for theories of vpattern masking, saccadic suppression, and infotion processing.Psychological Review,83, 1–36.

room, Y. M., & Doctor, E. A. (1995). Developmensurface dyslexia: A case study of the efficacy oremediation programme.Cognitive Neuropsycholog12, 69–100.

astles, A., & Coltheart, M. (1993). Varieties of develmental dyslexia.Cognition,47, 149–180.

llis, A. (1985). The cognitive neuropsychology of devopmental (and acquired) dyslexia: A critical survCognitive Neuropsychology,2, 169–205.

oss, D. J. (1982). A discourse on semantic priming.Cog-nitive Psychology,14, 590–607.

he-gs.eey

as

e

st

-

-

l-

a magnocellular deficit in developmental dyslexiaP. Tallal, A. M. Galaburda, R. R. Llina´s, & C. VonEuler (Eds.),Annals of the New York AcademySciences: Vol. 682. Temporal processing in thevous system: Special reference to dyslexia and dyssia (pp. 70–82). New York: New York AcademySciences.

ennie, P., Trevarthen, C., Van Essen, D., & Wa¨ssle, H(1990). Parallel processing of visual information. InSpillmann & J. S. Werner (Eds.),Visual perceptionThe neurophysiological foundations(pp. 103–128)New York: Academic Press.

iberman, I. Y. (1982). A language-oriented view of reing and its disabilities. In H. R. Myklebust (EdProgress in learning disabilities, Vol. 5(pp. 81–101)New York: Grune & Stratton.

ivingstone, M. (1990). Segregation of form, color, moment, and depth processing in the visual system: Aomy, physiology, art, and illusion. In B. Cohen &Bodis-Wollner (Eds.),Vision and the brain: The organization of the central visual system(pp. 119–138)New York: Raven Press.

ivingstone, M., & Hubel, D. (1988). Segregation of forcolor, movement, and depth: Anatomy, physioloand perception.Science,240,740–749.

ovegrove, W. (1993). Weakness in the transient visystem: A causal factor in dyslexia? In P. Tallal, A.Galaburda, R. R. Llina´s, & C. Von Euler (Eds.),Annalsof the New York Academy of Sciences: Vol. 682. Tporal processing in the nervous system: Special reence to dyslexia and dysphasia(pp. 57–59). NewYork: New York Academy of Sciences.

artin, R. (1995). Heterogeneity of deficits in developmtal dyslexia and implications for methodology.Psy-chonomic Bulletin & Review,2, 494–500.

cCloskey, M., & Rapp, B. (2000). Attention-referencvisual representations: Evidence from impaired vilocalization.Journal of Experimental Psychology: Hman Perception and Performance,26, 917–933.

cCloskey, M., Rapp, B., Yantis, S., Rubin, G., BacW. F., Dagnelie, G., Gordon, B., Aliminosa, D., Boman, D. F., Badecker, W., Johnson, D. N., Tusa, R& Palmer, E. (1995). A developmental deficit in locizing objects from vision.Psychological Science,6,112–117.

cCloskey, M., & Palmer, E. (1996). Visual representaof object location: Insights from localization impaments.Current Directions in Psychological Science5,25–28.

erigan, W. H., & Maunsell, J. H. R. (1993). How paraare the primate visual pathways?Annual Review oNeuroscience,16, 369–402.

ishkin, M., Ungerleider, L. G., & Macko, K. A. (1983Object vision and spatial vision: Two cortical paways.Trends in Neuroscience,6, 414–417.

orris, R. K. (1994). Lexical and message-level sentecontext effects on fixation times in reading.Journal of

Experimental Psychology: Learning, Memory, and

ual,

rela

nd

ts o-

3).

ory

rdfo

y:

ek,

R ofproom

S Aglyips-

S tex-

tual constraints in text and word priming.Journal of

S ftho-

S se

S izedage

&

T rentd

T vel-f

U al&r

W of

W tiva-d

W in

((


Cognition,20, 92–103.Morris, R. K., & Folk, J. (1998). Focus as a context

priming mechanism in reading.Memory & Cognition26, 1313–1322.

O’Seagdha, P. G. (1989). The dependence of lexicaledness effects on syntactic connectedness.Journal ofExperimental Psychology: Learning, Memory, aCognition,15, 73–87.

O’Seaghda, P. (1997). Conjoint and dissociable effecsyntactic and semantic context.Journal of Experimental Psychology: Learning, Memory, and Cognition,23,807–828.

Potter, M. C., Moryadas, A., Abrams, I., & Noel, A. (199Word perception and misperception in context.Jour-nal of Experimental Psychology: Learning, Memand Cognition,19, 3–22.

Potter, M. C., Stiefbold, D., & Moryadas, A. (1998). Woselection in reading sentences: Preceding versuslowing contexts.Journal of Experimental PsychologLearning, Memory and Cognition,24, 68–100.

Rayner, K., Murphy, L. A., Henderson, J. M., & PollatsA. (1989). Selective attentional dyslexia.CognitiveNeuropsychology,6, 357–378.

ayner, K., Pacht, J. M., & Duffy, S. A. (1994). Effectsprior encounter and global discourse bias on thecessing of lexically ambiguous words: Evidence freye fixations.Journal of Memory and Language,33,527–544.

hankweiler, D., & Liberman, I. Y. (1972). Misreading:search for causes. In J. F. Kavanagh & I. G. Mattin(Eds.),Language by ear and by eye: The relationshbetween speech and reading(pp. 293–317). Cambridge, MA: MIT Press.

harkey, A. J. C., & Sharkey, N. E. (1992). Weak con

t-

f

l-

-

Memory and Language,31, 543–572.ivan, A. B. (1992).Benton Visual Retention Test (fi

edition).San Antonio, TX: The Psychological Corpration.

nowling, M., & Hume, C. (1989). A longitudinal castudy of developmental phonological dyslexia.Cogni-tive Neuropsychology,6, 379–401.

nodgrass, J. G., & Vanderwart, M. (1980). A standardset of 260 pictures: Norms for name agreement, imagreement, familiarity, and visual complexity.Journalof Experimental Psychology: Human LearningMemory,6, 174–215.

abossi, P. (1988). Accessing lexical ambiguity in diffetypes of sentential contexts.Journal of Memory anLanguage,27, 324–340.

emple, C., & Marshall, J. (1983). A case study of deopmental phonological dyslexia.British Journal oPsychology,74, 517–533.

ngerleider, L. G., & Mishkin, M. (1982). Two corticvisual systems. In D. J. Ingle, M. A. Goodale,R. J. W. Mansfield (Eds.),Analysis of visual behavio(pp. 549–586). Cambridge, MA: MIT Press.

est, R. F., & Stanovich, K. E. (1986). Robust effectssyntactic structure on visual word processing.Memory& Cognition, 14, 104–112.

illiams, J. N. (1988). Constraints upon semantic action during sentence comprehension.Language anCognitive Processes,3, 165–206.

right, B., & Garrett, M. F. (1984). Lexical decisionsentences: Effects of syntactic structure.Memory &Cognition,12, 31–45.

Received March 17, 1999)Revision received October 6, 1999)

a visually based developmental reading...

Documents