varieties of pure alexia: the case of failure to ... - William James Hall

11 downloads 0 Views 2MB Size Report
a selective and severe deficit in naming visual objects and in reading letters, words, and numerals. Three sets ... semantic information from pictures and Arabic numerals. Based on this ... COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2), 203- 238. Ó 1998 ..... C, E, I, O, U, and i, m, o, u accounted for about. 40% of correct ...
COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2), 203- 238

VARIETIES OF PURE ALEXIA: THE CASE OF FAILURE TO ACCESS GRAPHEMIC REPRESENTATIONS Michele Miozzo and Alfonso Caramazza Harvard University, Cambridge, USA

We d ocu m ent the case of a patien t (GV) w ho, follow ing a left p osterior brain lesion , show ed a selectiv e and sev ere d eficit in nam ing v isual objects and in read ing letters, w ords, and num erals. Three sets of find ings are critical for the interp retation of the patient’s alexia. First, d espite intact v isu al p rocessin g abilities and preserv ed ability to recogn ise the shape and orientation of letters, GV cou ld not d eterm ine w hether a p air of letters had the sam e nam e. Second, she should not access the orthographic stru ctu re and m eaning of v isu ally presen ted w ords, althou gh she cou ld access m ean ing from orally spelled w ord s and she cou ld access orthograp hic stru ctu re from m ean ing in w ritten w ords. Third , GV cou ld access p artial sem antic inform ation from p ictu res and Arabic num erals. Based on this p attern of resu lts, w e conclud e that the form of alexia m anifested by our patien t resu lts from failu re to access the grap hem ic represen tation s of letters and w ord s from norm ally p rocessed visu al inpu t. The find ings further su gg est that access to letter form s and g rap hem e rep resentation s are sequ entially ord ered stag es of processing in w ord recog nition. The resu lts also su gg est that g raphem ic p rocessin g m ay be a d istinct property of the left hem isp here.

INTRODUCTION

1994; Dam asio & Damasio, 1983; Déjèrine, 1892; Gesch wind & Fusillo, 1966). The residual

The terms “alexia withou t ag raphia” and

read ing abilities m anifes ted by patients with pure alexia m ay v ary extens ively . Som e patients retain the ability to identify words

“pure alexia” refer to a class of sev ere read ing disorders appearing in the context of (relativ ely ) intact written lang uag e prod uction. These d isord ers are typically associated with lesions of the left posterior cortical and subcor-

correctly through a laboriou s process of recognition of the indiv idual letters. The hallm ark of this “letter-by -letter” approach to read ing is an

tical areas (Benson, 1985: Black & Behrm ann,

increase in read ing latencies as a function of

Requests for rep rints should be ad dressed to Alfonso Caram azz a or Michele Miozz o, Dep artm ent of Psy chology , William Jam es Hall, Harv ard University , 33 Kirklan d St., Cam bridg e, MA 02138, USA. The res earch rep orted here w as supported in part by an N IH g rant N S 34073 to AC. We thank Jeff Anderson and Ged eon Pad w a for their assistance in m aterial preparation and test ad m inis tration and Gail C. Ross for her ed itorial help. We are profound ly in debt to GV for her inexhau stible patience and g ood hu mour over the in num erable hours of testing.

Ó 1998 Psycholog y Press Ltd

203

MIOZZO AND CARAMAZZA

word leng th. Other patients seem to be com that is com m only accompanied by a sev ere

Warring ton (1962) found that three letter-by letter read ers had norm al recognition thresholds for sing le letters and pictures but

im pairm ent in nam ing letters. Two approaches hav e been taken in d ev el-

abnorm al threshold s for m ultiple stimuli presented sim ultaneou sly. And others hav e ar-

opin g exp lanations for the different forms of pure alexia. One group of exp lanations has

gued that all early stag es of visual processing of letters are intact and that the deficit arises becau se of a faulty access to intact lexical orthog raphic form s (Patterson & Kay, 1982), or

pletely unable to read words — a form of d eficit

attem pted to characterise the v ariou s form s of the disord er in term s of dam ag e to one or m ore stag es of processing in a m od el of the read ing process. Within this approach sev eral proposals hav e been m ad e. Som e authors hav e argued that the d eficit inv olv es relativ ely early stag es of processing (Farah & Wallace, 1991; Fried m an & Alexand er, 1984; Montant, N az ir, & Poncet, this issue; Rapp & Caram azza, 1991; Seku ler & Behrm ann, 1996; and see Behrm ann, Plau t, & Nelson, this issue, for a com prehensive review of this proposal). On this v iew , the deficit is not specific to read ing but to all types of visual stimuli (numerals, sym bols, pictured objects). The reduced processing efficiency resulting from the perceptual d eficit forces the patient to adopt a “divid e and conqu er” strateg y resulting in the letter-by -letter behav iour. Other authors hav e located the source of d eficit at a stag e of processing specific to letters (Arguin & Bub, 1993; Bub & Arguin, 1995; Howard, 1991; Kay & Hanley , 1991; Reuter-Lorenz & Brunn, 1990). On this view, the d eficit specifically inv olv es the efficient recognition of letters, forcing the patients to adopt a serial letter recognition strateg y. Others still hav e suggested that the d eficit specifically inv olv es processing multiple objects simultaneou sly (Kinsbourne & Warring ton, 1962; Lev ine & Calv anio, 1978). For exam ple, Kinsbou rne and

204

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

becau se the lexical orthog raph ic representations them selv es are d am ag ed (Warrington & Shallice, 1980). Of course, as has been noted prev iously (e.g . Farah & Wallace, 1991; Fried m an & Alexand er, 1984; Price & Humphrey s, 1992; Rapp & Caram az za, 1991), these exp lan ations of pure alexia are not necessarily m utually exclu sive. They could represent correct accoun ts for different sub-types of the disorder. That is, there m ay be patients who are alexic becau se they fail to process norm ally all types of visual input, others who are alexic becau se of a selectiv e d eficit in recognising letters, and so on. A d ifferent approach to exp laining pure alexia has been taken by Coslett and Saffran (1989a, 1989b, 1992, 1994; see also Coslett & Monsul, 1994; Saffran & Coslett, this issue) based on consid erations of the neu roanatom ical correlates of this disord er. Déjèrine (1892) and Gesh wind and Fusillo (1966) had earlier proposed that pure alexia reflects the disconnection of intact v isual processes in the right hem isphere from the lang uag e areas of the left hem isphere. Coslett and Saffran hav e added to this claim the assum ption that the perform ance of pure alexics principally reflects the read ing capacities of the right hem isphere.

VARIETIES OF PURE ALEXIA

And since a number of pure alexics appear to hav e considerable read ing skills (e.g . Coslett & Saffran, 1994; Shallice & Saffran, 1986), the

The right-hem isphere hypothes is of pure alexia exp lains the observ ed variations in performance am ong these patients by proposing

read ing capabilities of the right hem isphere must be equally consid erable. In fact, these

that the variation reflects differential d eg rees of integ rity of interhem ispheric connections.

authors hav e proposed that the right hem isphere has processes that can support the recognition of the orthog raphic and sem antic content of written word s. Howev er, compared

On this view, letter-by -letter read ers are those who hav e intact interhem ispheric connections between neu ral structures d ed icated to the processing of objects’ visual form s. The left

to the left hem isphere, the right hem isphere’s capabilities are lim ited in at leas t two respects: It cannot process abstract words and gram -

hem isphere can thus interp ret the visual form of each letter of a word that it receiv es from the right hem isphere as a distinct v isual object and

matical m orphem es and it cannot process the phon olog ical structure of words (see also

it can then retriev e the letter nam es. Once av ailable, the nam es of the letters are used to

Coltheart, 1980; Saffran, Bogyo, Sch wartz, & Marin, 1980; Zaidel & Peters, 1981; Zaidel & Sch weig er, 1984; but see Patters on & Besner, 1984 for criticism ).

recognise the word (in som e unspecified way ). By contrast, those pure alexic patients who fail to read word s altog ether are assum ed to hav e dam ag e to the interhem ispheric connections

Coslett and Saffran (1994) also m ake two hig hly specific assum ptions concerning interhem ispheric connections . One assump -

that transfer visual inform ation. This form of dam ag e prev ents the left-hem isphere read ing m echanism s from receiv ing any input, with

tion is that only structures that are functionally hom olog ous in the two hem ispheres are

the consequ ence that these patients are unable to read at all.

connected (directly). Thus, for exam ple, the sem antic system in one hem isphere is connected only to the sem antic system in the

Coslett and Saffran (1994) also argued that their proposal can exp lain why some alexic patients perform so well in lexical decision and

other hem isphere and not to any other com ponent. The other assum ption is that not all processing m echanis ms in one hem isphere

sem antic judgem ent tasks with written words despite their poor perform ance in pronou ncing them . Their exp lanation of these contrast-

are connected d irectly to their counterp arts in the other hem isphere. The priv ileg e of direct connection is restricted to those system s that

ing effects goes as follo ws: Word pronunciation is impaired becaus e this task requ ires access to left-hem isphere m echanis ms

the authors claim hav e a “venerable phylog enetic history”— nam ely , early perceptual

which, by hypothesis , are dam ag ed in these patients; howev er, lexical decision and sem an-

processes, and the representations of the structural and the sem antic featu res of objects.

tic judgem ent tasks are relativ ely spared because they can be perform ed directly by the right hem isphere.

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

205

MIOZZO AND CARAMAZZA

In this paper, we address som e of the issues raised by the d ifferent theories of pure alexia by considering the perform ance of a patient

problem atic for the hypothes is that there are duplicate word recog nition m echanis ms in the left and right hem ispheres . Howev er, with re-

(GV) who presents with a particularly clear set of dissociations as the result of d am ag e to the

spect to object recognition, the data are consistent with the hypothes is that some form of

left posterior cortical areas and the corpus callosum . GV was virtually unable to read and was im paired in read ing letters and Arabic num erals as well as in nam ing pictures. In

rudimentary concep tual kno wled ge, accessible through v isual representations of objects, is encod ed in the right hem isphere.

contrast, her ability to recog nise orally spelled word s was intact, suggesting that g raphem ic inform ation is preserved but not accessible

CASE REPORTS

through the v isual m od ality . We also replicated and extend ed the recent finding of Co-

GV is an 84-y ear-old , right-han ded wom an with a colleg e education. She worked as a

hen and Dehaene (1995, 1996) of a d issociation between im paired read ing of Arabic numerals and preserv ed recognition of their quantity valu e. Finally , we were able to show that GV

nurse for m any years prior to retirem ent. In October 1994, she had an infarct inv olv ing the left posterior cerebral artery . An enh anced CT scan (June 1995) shows a lesion in the left oc-

could access partial sem antic inform ation from pictures but not from written words. Based on these results, we propose that GV’s read ing

cipital and posterior tem poral areas , extend ing to the corpus callosu m (see Fig. 1). An add itional area of hypod ensity, probably due to a

im pairm ent reflects a failu re to access the abstract (graphem ic) representations that specify

silent, previous CVA appears in the right parietal region. The patient presents with a right

the id entity of letters and , consequ ently , failure to access the orthog raphic form of word s and their m eanin g. A d ifferent functional im -

v isual field cut. The only d eficits revealed by the neurop sycholog ical testing were nam ing difficulties for visually presented stim uli, in-

pairm ent is responsible for GV’s deficit in processing pictures and Arabic num erals. With this type of stim uli, access to sem antic

cluding objects, pictures, colours, and read ing / nam ing d ifficulties for letters, num erals, and words. A series of tests was adm inistered

inform ation is still possible, thou gh lim ited and insufficient to support the selection of the 1 correct nam es . We argue that our d ata are

to d eterm ine whether, as a consequ ence of her right parietal lesion, there were signs of visual neg lect. N o ev idence of v isual neg lect em erged

1

Obv iously the term “pure alexia” cap tures only one asp ect of GV’s im pairm ent — that related to read in g. Other term s m ig ht be used to des cribe the ad ditional im pairm ents show n by ou r patient (e.g . “optic ap hasia” for her deficit in processing objects). Nev ertheless, the term “pure alexia” cap tures one asp ect of GV’s read in g im pairm en t— the fact that her read ing d ifficu lty is not accom panied by dy sg rap hia or ap hasia.

206

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

VARIETIES OF PURE ALEXIA

Fig. 1. Representative cut of GV’s CT scan (the left hemisphere is shown on the right in the picture). The picture shows a lesion of left occipi tal and posterior temporal cortical areas extending to the splenium of the corpus callosum. An area of hypodensity is also present in the right pariet al region.

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

207

MIOZZO AND CARAMAZZA

from any of the tests (line bisection, variou s cancellation tasks, dra wing from copy and from mem ory).

tem pts to approach the correct response. With tasks requ iring nonv erbal responses, the first response was retain ed for further analy ses. (It

The stud y reported here was started 3 months after the onset of the illnes s and con-

shou ld be stressed, howev er, that the pattern of results rem ained essentially unchang ed

tinued for 6 m onths. During this period GV’s condition remained neu rolog ically and functionally stable. The inv estig ation focused on her abilities to process letters, words, numer-

with different scoring proced ures.) Whenev er the sam e list of stimuli was adm inistered m ultiple times or was used in different tasks, ord er of item presentation was counterbalanced .

als, and pictured objects. The results of the variou s exp erim ental tasks are presented in separate sections organised as follo ws: (1) gen-

Each adm inistration of the exp erim ent proper was preced ed by a practice block. Alphanu m eric stimuli were printed in 34-p oint

eral nam ing , read ing , and spelling perform ance; (2) visual object processing ; (3) letter and

Helv etica font. (Essentially the sam e pattern of perform ance was found when stim uli were

word processing; (4) num ber processing . Each section conclud es with a sum mary of the relevant finding s. We beg in with a description of the general proced ures used in the stud y.

presented in a sm aller size.) The size of lowercase letters was norm alised to the size of upper-case stim uli. Only letters judged to hav e differen t shap e in upper and lo wer case (Boles

GENERAL PROCEDURES

& Clifford , 1989) were used in tasks where letters were sho wn in differen t form ats. Specifically , we selected the follo wing pairs of

GV was asked to perform two types of tasks: (1) tasks in which the stim uli were presented for unlim ited exp osure, and (2) computerised tasks in which stim uli were shown for lim ited duration. The proced ures in these t wo ty pes of tasks were slightly different, and hence they will be described separately .

Time-unconstrained Tasks In these tasks, two types of responses were produced — either a nam ing or a nonv erbal response (e.g . pointing ). In the nam ing tasks, the patient’s last response to each item was scored ; this procedure ackno wled g es the possibility that the patien t’s initial responses were at-

208

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

letters: A-a, B-b, D-d, E-e, F-f, G-g, H-h, N-n, P-p, Q-q, R-r and T-t .

Time-constrained Tasks The patient was seated at a distance of approxim ately 45cm from a com puter screen. A trial had the follo wing structure: The patient initiated a trial by pressing the space bar; then a fixation point (a cross) appeared for 700m sec, which was im med iately replaced by the stim ulus. Because of GV’s visual cut, targets were al way s shown at the left of fixation. To respond , the patien t pressed one of t wo key s with either the right or left hand , on the right and left of the bod y m idline, respectiv ely . Pilot tests rev ealed that with this proced ure the

VARIETIES OF PURE ALEXIA

patient did not miss any stimulus. The assignment of key s to type of response was counterbalanced across testing sessions. Stim ulus

production. A deficit in nam ing letters and num erals m ight originate from a difficulty in accessing the phonolog ical form of the stimuli.

presentation and record ing of response latencies were both controlled by PsychLab soft-

To exam ine this hy poth esis, we adm inistered tasks that requ ired the oral production of letter

ware (Bub & Gum , University of Victoria, Victoria, British Colum bia, Canad a). In ord er to obtain a substantial num ber of observations and to reduce response variability , the com -

nam es (oral spelling ) or num ber nam es (e.g . retriev al of numerical facts). The analy sis of her perform ance in variou s spelling tasks also provid es the basis for d ra wing conclusions

puterised tasks were adm inistered multiple tim es across sev eral testing sessions. The ord er of presentation of the exp erim ental blocks was

abou t the integ rity of her kno wled ge of a word’s orthog raphic representation.

varied system atically for each adm inistration of the test. Correct response latencies exceed ing three stand ard dev iation s were replaced by their condition’s m ean. To hav e an equ al num ber of responses in each cell, the latter procedure also was ad opted with erroneou s responses. Different sorts of stim uli were shown v ia compu ter presentation : pictures, letters, word s, and d igits. The m axim u m v isual ang le of pictures was 5°. Alphanu m eric characters were in 34-p oint Helv etica font.

NAMING, READING AND SPELLING The results presented in this section aim to establis h the selectiv ity of GV’s nam ing d eficit. Specifically , we inv estig ated whether her im pairm ent is restricted to the v isual m odality and whether, within this m od ality , it extend s to d ifferen t sorts of stim uli (pictures, colours, letters, word s, and num erals). Com parisons of her nam ing abilities across m odalities were used to determ ine the intactnes s of sem antic kno wled ge and of the m echanis ms for lexical

Picture Naming GV’s ability to nam e pictures was form ally tested by m eans of Snod grass & Vander wart’s (1980) picture set (N = 260). This set was adm inistered twice. Response m od ality varied in the two adm inistrations (oral vs. written). As shown in Table 1, GV was sev erely impaired . She was able to nam e only abou t 50% of the pictures successfully. Furtherm ore, her lev el of accuracy remained unchang ed across output m od alities. The majority of her erroneou s responses can be broad ly classified as semantic confusions, which accou nt for 82% and 71% of her errors in oral and written nam ing , respectively . Within this categ ory of responses we included sem antic substitutions (e.g . table ® “seat”, tie ® “belt”), superordinate nam es (e.g . necklace ® “piece of jewellery ”, pepper ® “veg etable”) and functional d escriptions of the objects (e.g . saw ® “used for chopping wood ”, ruler , ® “to take measu res”). Visually sim ilar errors (e.g . balloon ® “flo wer”, clock ® “button”) and persev erations were produced far less frequ ently .

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

209

MIOZZO AND CARAMAZZA

Table 1 . Oral and Written Picture Nami ngÐ

GV’s Responses No. of Responses/Task

——————–—————————–

Type of Response Correct Errors Sem antically related Visually sim ilar Perseveration s Others

Example

ear ® “nose” pepper ® “veg etable” balloon ® “flower” toaster ®

“stairs”

Colour Naming GV’s sev ere nam ing deficit extend ed to colours. She successfully nam ed the colour of only 4/ 12 crayons. All errors consisted of colour nam e substitutions. Naming in Nonvisual Modalities In nam ing a set of objects sho wn for tactile exp loration (from Hillis & Caram az za, 1995a), GV respond ed correctly to 46 out of 47 item s (98%). She also perform ed well in nam ing orally presented d efinitions of the sam e objects (96% correct). Howev er, in nam ing pictures of these objects she was correct only 62% of the tim e. Her errors (17/ 18) were alm ost all semantic confusions (e.g . sock ® “belt”, spatula ® “a kitchen thing ”). The d ifferen ce am ong tasks is statistically reliable [Cochran, Q(2) = 28.7, P < .001]. The contrast bet ween the performance in the v isual confrontation and the other nam ing tasks clearly dem onstrates the modality / specificity of her nam ing disturban ce. Furtherm ore, GV’s good perform ance in the tactile and nam ing -to-definition tasks

210

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

Oral Naming

Writing Naming

—————–

———————

No.

%

No.

%

125

(48)

132

(51)

111

(42)

91

(35)

(4) (5) (1)

12 19 6

(5) (7) (2)

10 13 1

exclu des the possibility that her v isual nam ing deficit results from dam ag e to the sem antic system .

Letter Naming In sev eral testing sessions, GV nam ed each upper- and lo wer-case letter of the English alp habet 20 times. Upper- and lo wer-case letters were presented in separate blocks of 26 stim uli each. GV was impaired in nam ing the stim uli in both formats, alth ough her accuracy was slightly hig her with upper- compared to 2 lo wer-case letters [44% vs. 30%; c (1) = 10.7, P < .001]. This asymm etry, also noticed in other patients with impaired nam ing of letters (e.g . Perri, Bartolom eo, & Silv eri, 1996), presum ably reflects d ifferen ces in the d eg ree of confusability am ong letters in the t wo types of formats. Correct responses were produced for a restricted group of letters — the characters A, C, E, I, O, U, and i, m, o, u accoun ted for abou t 40% of correct responses in each set. The probability with which individ ual letters were produced as incorrect responses was also unequ al,

VARIETIES OF PURE ALEXIA

reflecting a tend ency to persev erate with a subset of letter nam es.

Table 2 . Naming of 1- to 5 -di git Arabi c Numeral sÐ Correct Nami ng and ª Syntacticª Errors Correct

Reading Aloud GV was virtually unable to read words. Presented with a set of 70 word s controlled for frequ ency and leng th (Goodm an & Caram az za, 1986), she correctly read only one word , chair. Six months later, her perform ance rem ained unchang ed (0/ 70 correct). GV’s attem pts to read were v ery laboriou s. Usually , she started by trying to identify sing le letters, then tried to concatenate them into syllables, som etim es producing a word (e.g . ruin ® “guru”). Occasionally , responses contain ed parts of the stim ulus (e.g . chipmunk ® “check” or special ® “deal”). Persev erativ e responses were also noticed . Naming of Numerals Different lists of num erals were used . In one task, each d igit in the rang e 0 –9 was presented , in a rand om order, 10 tim es. GV successfully nam ed 60% of the item s. Correct responses were unequ ally distributed across target d igits; for exam ple, 8 was nam ed correctly nine tim es while 3 was nam ed correctly only three tim es. She had a bias to respond with the digits 4, 5 and 6, which accounted for 72% of her error responses. Her lev el of perform ance decreased dram atically with 2- to 5-digit numerals: She correctly nam ed only 29/ 280 (10%) item s ov erall, and virtually none of the 4- and 5-digit num erals (see Table 2). In this task, numerals were presented without comm as, and the patient was instructed to read each stimulus as a “single num ber”, rather than to read the digits

Syntactic Errors

—————– Target 1-d igit 2-d igit (teens) 2-d igit 3-d igit 4-d igit 5-d igit

No. 17/ 12/ 10/ 6/ 1/

40 40 40 40 40

0/ 120

GV’s

———————

%

No.

(42) (30) (25) (15) (5)

5 4 1 0 0

(0)

6

sequ entially . Errors typically consisted of digit substitutions (e.g . 647 ® “827”, 39087 ® “49021” ). “Syntactic errors” — characteris ed by the deletion (e.g . 647 ® “67” or “6”) and / or addition (e.g . 647 ® “1647” or “64007” ) of one or m ore d igits — were observed v ery rarely (< 1%). An analy sis was carried out to d eterm ine whether GV’s nam ing errors tend ed to be “close” in m ag nitude to the exp ected response (e.g . 3 ® “4” or 8 ® “7”). We analy sed a larg e corpus of errors (N = 369; details abou t these errors will be presented later in the Mag nitude Judgem ent Task). There were only eig ht expected responses in the corpus used for this analy sis (the digits 0 and 5 were nev er presented ). The distribution of the observ ed errors as a function of their numerical distance from the target was contrasted with the distribution exp ected by chance (i.e. when each digit has an equ al probability of being selected ; Cohen & Dehaene, 1996; see Fig. 2). Responses that were one or t wo d igits larg er/ smaller than the target were considered as “close errors”. The observed frequ ency of close errors was not significantly differen t from that exp ected by

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

211

MIOZZO AND CARAMAZZA

Fig. 2. Distribution of GV’s errors (N) as a function of

sonal numerical inform ation (e.g . her y ear of

their numerical difference from the target (sol id line) in the Oral Naming Task with Arabic digits. The figure also shows the distribution expected if responses were randomly selected (broken line). Note that the digits 0 and 5 were never presented in this task.

birth). Further ev id ence that access to number nam es was intact is provided by her ability to

chance [c (1) < 1]. This find ing ind icates that num erical distance is not am ong the factors 2

significantly affecting the selection of the erroneou s responses in this task.

Numerical Facts GV was asked to retriev e variou s num erical facts such as “ho w m any pennies are in a dollar?” or “ what is the year of the d iscov ery of Am erica?” She perform ed fla wlessly with oral questions of this sort. Excellent perform ance was also found with questions concernin g per-

212

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

solv e sim ple calculations (e.g . 2 + 3; 6 × 7) with orally presented stim uli, These results indicate that her number nam ing deficit is not due to dam ag e to lexical prod uction m echan ism s or the num ber sem antic system.

Aural Recognition of Orally Spelled Words and Spelling Tasks The list of words used to assess read ing was also used in a recog nition task of orally spelled words and in written and oral spelling -to-d ictation tasks. As sho wn in Table 3, GV performed fla wlessly in recognising orally spelled words and m ade v ery few errors (4.3%) in the t wo tasks. The few errors (6/ 140) all inv olv ed

VARIETIES OF PURE ALEXIA

Table 3. Spelling TasksÐ

GV’s Correct Responses Correct

a

————— Task

No.

%

Written spelling Oral spellin g Aural recog nition of orally spelled w ord s

66 68 70

(94) (97) (100)

a

Total = 70.

lo w-frequ ency word s and were phonolog ically plau sible realisation s of the target (e.g . pirate ® PYRATE, pigeon ® PIGION), with one excep tion (the omission of “t” in instinct ® INSTIN K). The results of these tasks indicate that GV has norm al access to the orthog raphic structure of word s both in recog nition (aural recog nition of orally spelled words) and in production (spelling ).

Discussion The results of the tasks reported in this section reveal the extent and sev erity of GV’s nam ing im pairm ent with v isual stim uli: Nam ing was sev erely disrupted for pictures, colours, letters, words, and numerals (see summ aries in Tables 1 and 2). Her contrasting perform ance in tasks with nonv isual stim uli allo ws us to exclud e some hy poth eses abou t the functional lev el of her deficit. Specifically , becau se GV successfully nam ed objects from tactile exp loration and from verbal description, we can exclud e as causes for her visual nam ing d eficit both d am ag e to the sem antic system and dam ag e to the word production m echanism s. We can also exclu de a deficit in prod ucing the nam es of letters, since she could retriev e them

norm ally in the Oral Spelling Task. Sim ilarly , since GV successfully nam ed numerical facts in response to orally presented stimuli, we can infer that she has intact access to the lexical forms of num bers. GV’s norm al ability to recognise orally spelled word s sho ws that her orthog raphic input lexicon is intact (Caram az za & Hillis, 1990; though see Warrington & Shallice, 1980, for a contrasting viewpoint).

VISUAL OBJECT PROCESSING The tasks reported in this section ad dress three points. First, a com parison of the results with v ariou s visual processing tasks allo ws the id entification of the locus of functional dam ag e responsible for GV’s im pairm ent with visual stim uli. We ad dress whether such d am ag e arises (a) at the lev el of perceptual analy sis, (b) at the lev el of access to inform ation abou t the shape of fam iliar objects (structural d escription), or (c) at the lev el of sem antic access. Second , we attem pt to d eterm ine whether sem antic inform ation is fully accessed with pictures. (In a later section, the accessibility of sem antic inform ation with pictures will be contrasted to that found with v isually presented words.) Third, since sev eral of the tasks presented in this section hav e previously been giv en to other patients with alexia, the results of these tests allo w a direct com parison of the perform ance of our patient with that of prev iously docum ented cases.

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

213

MIOZZO AND CARAMAZZA

Visual Perceptual Tests The percentag e of correct responses produced by GV in a series of visuopercep tual d iscrim ination tasks from the BORB (Rid doch & Hum phrey s, 1993) is com parable to that of a g roup of 39 neurolog ically intact control subjects (ag e 50–80; see summ ary in Table 4). These tasks requ ire v ery fine discriminations of line leng ths, line orientation, size of squares, and the position of small gaps in circles. GV’s good perform ance in these tasks rules out a lo wlev el perceptual d eficit as the m ain cause of her modality -specific impairm ent. In addition , her ability to trace the outlines of ov erlapp ing stim uli indicates that she can segreg ate visual objects. Matching Objects from Different Views The structural description of an object is a m ental representation specifying the visual features of a canonically oriented fam iliar object (e.g . Riddoch & Hump hrey s, 1987a). Access to these representations can be ev alu ated by means of a recognition test in which one has to decid e whether the sam e object is pictured from different perspectiv es. The m aterial and the norm s for this task were from Riddoch and Hum phrey s’ (1993) BORB test. Subjects were Table 4 . Visual Discrimi nation TasksÐ

GV and Controls

% Correct

——————–— Feature

GV

Controls

z -scores

Line leng th

80

89

– 1.8

Line orientation Square siz e Gap position

90 83 82

82 91 87

+ 0.8 – 0.9 – 0.5

214

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

presented with three pictu res: t wo of the pictures represented d ifferent view s of the sam e object; the third picture, a foil, sho wed an object v isually similar to one view of the target. GV’s accuracy (46/ 50; 92%) d id not d iffer from that of controls (age 50 –80; 89.8%).

Reality Decision GV was asked to discrim inate between real and chimaeric v isual objects. The latter objects were created by replacing one part of one object with a part from another object of the sam e categ ory (e.g . the head of a donkey was m ounted on the bod y of a d og). Stim uli (N = 64) and norm s were from the BORB (Ridd och & Humphrey s, 1993). GV perform ed within the norm al rang e: 83% v s. 81% correct for GV and the m ean of controls, respectively . The com ments occasionally made by the patient rev ealed that she was able to identify stim uli ev en when she failed to nam e them . For exam ple, she correctly accepted the picture of a pig as real, but was only able to describe it as “a farm anim al,” and althou gh she correctly rejected the chim aeric snake/ turtle as unreal, she described it as hav ing “the head of a goat.” Her good perform ance in the Reality Judgem ent Task, paired with her good ability to m atch objects from d ifferent views, prov ides conv incing ev id ence that GV can successfully access the stored structural descriptions of objects. Picture Categorisation GV perform ed fla wlessly in sorting 30 pictures into the categ ories animals vs. plants. She also correctly distinguished (48/ 52; 92%) between

VARIETIES OF PURE ALEXIA

pictures of food (e.g . sandwich, apple) and nonfood item s, includ ing kitchen utensils.

Picture Association (Visually Similar Foils) In this task, GV selected the two associated pictures out of a set of three. The d istractor was visually similar to one of the targets (e.g . pear was presented with light bulb and light s witch). The stim uli were from Hillis and Caram azza (1995a). GV correctly matched 32/ 33 (97%) of the item s. In a separate session, she nam ed only 34% of the pictures correctly. Picture Association (Related Foils) Picture categ orisation and picture association tasks hav e prev iously been used to ev alu ate the intactness of sem antic access (e.g . Coslett & Saffran, 1989b; Iorio, Falang a, Frag rassi, & Grossi, 1992; Manning & Cam pbell, 1992). Consistent with the interp retation reached by these earlier inv estig ators, we should conclude that GV’s access to the conceptual representation of objects is intact. Ho wev er, such a conclusion would be unjustified if based only on such perform ance, since access to incom plete concep tual representation of objects might be sufficient to support good perform ance in the kind of recognition tasks presented earlier (e.g . Hillis & Caram az za, 1995a). Access to an intact (complete) sem antic representation is not need ed for the rather coarse discrim ination bet ween anim als and plants; and it is not need ed for the finer-g rained edible vs. ined ible discrimination either. The retriev al of an incomplete representation that includ es the right featu res is sufficient for accepting an object as edible (e.g . , , ) or ined ible (e.g . , , ). Analog ous argum ents can be made to account for good perform ance in the triad ic picture association task. Correct matching in this task may simply result from a differential judgem ent: The associated item s share m any m ore featu res with each other than either one does with the unrelated item . A judgem ent of this sort does not requ ire access to a com plete sem antic description. All that is need ed is that associated objects should activ ate a larg e enou gh num ber of com m on featu res to d istinguish between related and unrelated item s. In the tasks used in previous researches , the foil were al way s sem antically unrelated to the other t wo item s. Therefore, the foils v ery rarely activ ated sem antic featu res in com m on with the two other item s, a fact that shou ld guarantee a correct m atch ev en in condition s of partial disruption of sem antic access. Thus, the results presented do not exclu de the possibility that GV has a deficit in accessing sem antic representations. And , in fact, when GV is given tasks that requ ire access to d etailed sem antic inform ation, her perform ance deteriorates dram atically . Hillis and Caram azza (1995a) introd uced the follo wing modification to the triad ic m atchin g task: The visually similar foil was replaced by a related , thou gh less associated , foil. To illustrate, the related pair light bulb and light switch, which in the original task was shown with the v isual foil pear, was no w shown with the sem antically related foil traffic light . GV’s perform ance in the task with sem antically related foils differed dram atically

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

215

MIOZZO AND CARAMAZZA

from the original triad ic association task: In the new task, she respond ed correctly on only 15/ 33 trials (45%; com pare this to 97% correct

the Snod grass & Vander wart (1980) set were presented with three different words on three separate occasions: the nam e of the picture, a

in the previous task). The lo wer perform ance in the related foils task is not becau se there is

sem antically related nou n (e.g . grapes –cherry , guitar –violin), and a phonolog ically sim ilar

am big uity as to the correct choice for the new item s. When GV was given an aural version of this task (i.e. she was g iven the nam es of the objects in a triad ), she perform ed fla wlessly

nou n (e.g . eag le–needle, foot –flute). An item was scored as correct only if the subject accepted the correct nam e and rejected the two foils. GV perform ed rather poorly in this task:

(33/ 33). Thus, we hav e ev idence that GV’s sem antic system is intact but that access to representation s through the v isual m odality

She respond ed correctly only 64% (167/ 260) of the time. The vast m ajority of her errors (81/ 96; 84%) inv olv ed the acceptance of a sem antic

(i.e. for pictures) is im paired .

foil. The few tim es (8/ 96; 8%) in which GV accepted a phon olog ical foil, the foil was visually sim ilar to the target (e.g . anchor ®

Property Judgement Task GV ans wered yes/ no questions (N = 160) abou t specific sem antic featu res of objects (e.g . Is it (lemon) sour? Does it (eagle) sing?). For each item there were two “yes” and two “no” spoken question s. A wid e rang e of featu res were queried : visual, functional, associativ e, and encyclopaed ic. Care was taken to exclu de questions whose ans wers could be d eriv ed from featu res contained in the dra wing s. In one condition a picture was shown, in the other the nam e of the picture was orally presented . A striking d issociation was found bet ween conditions: When targets’ nam es were presented orally she was alm ost inv ariably correct (97%); with pictures, ho wev er, she respond ed correctly only 83% of the time [McNemar, P < .001]. Spoken Word/Picture Verification In this task (from Hillis & Caram azza, 1995a), GV was asked to recog nise whether a spoken nam e m atched a given picture. Pictu res from

216

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

“hang er”). Thus, we hav e conv erging ev id ence that GV’s perform ance is impaired when the task requ ires her to m ake fine sem antic distinctions.

Discussion The results reported in this section seem to exclu de a v isual perceptual d eficit as the basis for GV’s poor nam ing perform ance. N o indication s of dam ag e to the v isual processing system em erged from tasks requ iring very precise discrimination s of perceptual featu res such as orientation , leng th, size, and position (see Table 4). Furtherm ore, she perform ed flawlessly in object recognition tasks (e.g . the Reality Decision Task), which requ ire excellent perceptual abilities for normal execution. N ev ertheless , one could object that these tasks were not particularly stringent. More stringent tests of GV’s ability to process visual stim uli are presented later in the paper.

VARIETIES OF PURE ALEXIA

Two other important find ing s are reported in this section (see Table 5 for sum m ary): (a) intact perform ance in tasks requ iring access to

dam ag e to sem antic kno wled ge itself, since she perform ed nam ing and other sem antic tasks norm ally when the input inv olv ed the aud i-

structural kno wled ge of v isual objects, and (b) im paired perform ance in visually accessing se-

tory and the tactile modalities . Sev eral of the tasks presented in this section

mantic inform ation . The latter finding s inv ite the conclusion that up to the point of access of stored structural descriptions of objects, visual processing is intact; it is the subsequ ent access

are identical to those giv en to other patients with an “isolated ” right brain for v isual processing (patients with selectiv e dam ag e of the left posterior areas and of the corpus callo-

to sem antic inform ation that is d isrupted. Howev er, access to sem antics is not com pletely impossible, but partially lim ited . This

sum), including some of the patien ts with alexia reported by Coslett & Saffran (1989b, 1992). These cases all tend to sho w intact access

characteris ation of her ability to access sem antics through visual m od ality is in ag reem ent

to the structural representation s of objects (see Iorio et al., 1992 for review). In studies where

with the type of errors she comm only produced in picture nam ing — sem antic substitutions. Errors such as ruler ® “to take measures” or fox ® “ wild anim al” rev eal that

sem antic access from visual objects was exam ined exten siv ely (e.g . De Renz i & Saetti, 1997; Hillis & Caram azza, 1995a; Iorio et al., 1992; Riddoch & Hum phrey s, 1987b), the results are

despite her failu re to retriev e the entire set of target’s sem antic featu res, som e of these features were accessed successfully. N ote, ho w-

id entical to those observ ed with GV: Access was only partial. In sev eral other studies the sem antic tasks

ev er, that GV’s d eficit cannot be attribu ted to

used (e.g . picture classification) did not inv olv e the discrimination of detailed sem antic

Table 5 . GV’s Perf ormance in Semantic TasksÐ

Summary

% Correct

———————— Task

Pictures

Words

100 92

100 96

Association task Vis ual foils Related foils

97

100

45

100

Property judgement

83

87

Word/Picture verifications

64



Categorisation task Anim als v s. Plants Food v s. N onfood item s

inform ation of v isual objects. Thus, it is not possible to m ake direct comparisons across all relev ant cases on this aspect of lang uag e processing. Howev er, som e of the tasks used in the latter studies (e.g . picture association with visual foils) were also used with GV (and patient DHY studied by Hillis & Caram az za, 1995a). For these tasks, similar results were found across stud ies: Patients perform ed well. Howev er, as we hav e seen in GV, these tasks are not sufficiently sensitiv e to reveal an impairm ent of sem antic access with visual objects. Therefore, it cannot be exclu d ed that in prev iously docum ented cases (includ ing those reported

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

217

MIOZZO AND CARAMAZZA

by Coslett & Saffran, 1989a, 1992) complete sem antic access from visual objects was im paired .

WRITTEN LETTERS AND WORDS In the preceding section , we inv estig ated the integ rity of access to the structural representation of fam iliar objects. We no w inv estigate whether analog ous structural inform ation is av ailable for written letters. We also exam ine the av ailability of inform ation abou t abstract letter identity (graphem es) and the exten t to which GV can access inform ation abou t the orthog raph ic and sem antic content of written word s. In d esig ning the latter tasks we follo wed the proced ures described by Coslett & Saffran (1994).

Reality Decision GV was asked to disting uish between fam iliar upper/ lo wer-case letters (N = 52) and pseudoletters (N = 38). The latter are stim uli that look like letters; they were created by rearrang ing the strokes of letters of the Rom an alp habet (see exam ples in Fig. 3). GV correctly classified all the stimuli, d espite the fact that she frequently prod uced the wrong nam e when she spontaneou sly nam ed the letters (it was not requ ired by the task; incorrect nam ing was also observ ed when GV perform ed the Reality Judgem ent Task with objects). Orientation Decision This task, dev ised by Cooper & Shep ard (1973), was used to exam ine the av ailability of infor-

218

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

Fig. 3. Examples of pseudoletters used in the Reality Decision Task.

m ation abou t the canonical orientation of letters and digits. The task requ ires the discrimination bet ween norm ally orientated and reflected characters. Stim uli were presented in each of six orien tations , equ ally spaced by 60° starting from 0° (see exam ples in Fig. 4). The asym m etrical letters and digits R, J, G, 2, 5 and

7 were used in this task. Each stim ulus appeared t wice in each orien tation, once in the canonical and once in the noncanonical (reflected ) orientation. As is ev ident from Fig. 4, GV’s perform ance in this task is ind isting uishable from that of an ag e-m atched control subject (85% and 87% correct judgem ents, respectiv ely ). Such a good lev el of perform ance cannot hav e resulted from the fortuitous selection of letters and digits easily identifiable by the patient. In a separate session, she suc2 cessfully nam ed only 15% of the stimuli . The results of the orientation task provide support for the hy poth esis that GV has access to the structural representations of letters and d igits.

VARIETIES OF PURE ALEXIA

pairs. Letters were sho wn one at the time for 200m sec, separated by a 400m sec inter-stim ulus interv al. On each trial, the stim uli appeared at t wo distinct locations — in the left quadrant abov e or belo w fixation. The order of presentation was counterbalanc ed between the two locations . As illustrated in Fig. 5, the patient respond ed accu rately only with PI pairs (93% correct); with NI and D pairs her perform ance was m uch less accurate (63% and 61% correct, 2 respectiv ely ) althou gh abov e chance [c (1) = 5.1 and 6.7; for both, P < .001]. These results ind icate that access to inform ation abou t graphem e identity is sev erely impaired .

Fig. 4. Correct performance (%) for GV and a matched control subject in a letter/di git orientation task (Cooper and Shepard, 1 973). Examples of normall y oriented and ª reflectedª stimuli are show.

Same/Different Name Decision Task The letter matching task dev ised by Posner and Mitchell (1967) was given to GV. In this task, the response “sam e” is expected with pairs of letters that are eith er physically identical (PI; A–A, a–a) or nam e id entical (N I; A–a, B–b). When the two letters had a d ifferent nam e (D), the subject was instructed to respond “different.” On 50% (N = 288) of the trials, D pairs were sho wn; the remaining trials consisted on an equ al num ber of PI and NI

Letter Transcoding Ad ditional ev idence on the av ailability of information abou t graphem e identity can be obtain ed with a task in which visual letters are to be reproduced (written) in a d ifferent format (e.g . A ® a, d ® D). GV successfully transcoded 31/ 84 (37%) of the upper/ lo wer-case letters (only the 14 characters with different shapes in the two formats were used ). Errors consisted of well-form ed but incorrect letters (e.g . A ® p). The results of the letter (and digit) tasks presented thus far support the conclusion that GV has normal access to the structu ral d escriptions of letters (and d igits) but not to inform ation abou t abstract letter identities or graphem es.

2

(Opposite). The reliability of these resu lts is further supported by the find ings of another Orientation Judg em ent Task. In this task, almost the entire set of upper- and lo wer-case letters w as used (N = 41). Each character w as presented in tw o orientations (canonical and noncanonical). Noncanonical letters w ere “upsid e-dow n” (rotated 180° on the fronto-parallel plane) or “reflected .” GV alw ay s discrim inated the stim uli correctly.

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

219

MIOZZO AND CARAMAZZA

Word Categorisation Task A word categ orisation task (animal vs. artefact) was used to assess GV’s ability to access m eanin g from print. Stim uli were three- and four-letter, highly fam iliar words. The stim uli were presented for unlim ited time on separate cards. GV was asked to sort them into two piles. She was requ ired not to nam e the words, but to “concentrate” on their m eaning . Her lev el of accuracy in this task d oes not differ from chance (23/ 40, 57% correct responses), a result clearly indicating sev ere d ifficulty in accessing the m eaning of word s. Fig. 5. GV’ s correct performance (%) in the Same vs. Different Name Decision Task (after Posner & Mitchell, 196 7).

Lexical Decision Task Successful discrimination of fam iliar v s. unfamiliar words is based on the retriev al of appropriate lexical orthog raphic inform ation. Lexical d ecision perform ance can thus reveal the extent to which orthograp hic inform ation is av ailable. GV was shown a set of 20 hig h-frequency words and 20 non word s. The latter were created by chang ing one letter in very fam iliar Eng lish word s (e.g . home ® HAME). All stim uli were four letters long and were printed in upper case. GV was asked to decid e whether the string of letters formed a word. She was also given strict instructions not to nam e the stim uli. She was unable to d istinguish bet ween word s and non words (55% correct). This result inv ites the conclusion that GV cannot access the orthog raphic lexicon from visual input.

220

COGNITIVE NEUROPSYCHOLOGY, 1998, 15 (1/2)

Priming Task Coslett & Saffran (1989a, 1989b) dem onstrated that the read ing perform ance of patients with pure alexia m ay im prove when stim uli are presented for lim ited exp osure. We exp lored whether GV’s perform ance m ight sim ilarly im prove if written stim uli were presented v ery briefly . We used a pictu re categ orisation task (anim als vs. artefact) that used briefly presented words as prim es. If GV can access the m eanin g of words, her categ orisation latencies would be exp ected to be faster with identical and related (sam e-categ ory) primes than with unrelated primes. In a separate exp erim ent, pictures were used as primes. The latter exp erim ent serv ed as a control to d eterm ine whether a priming effect could be obtained with primes other than words. The results of the picture categ orisation tasks d iscussed in the prev ious section conv incing ly dem onstrate that GV can access sufficient sem antic inform ation to categ orise objects. We, thus, exp ect that GV would show

VARIETIES OF PURE ALEXIA

picture/ picture sem antic priming . Howev er, given prev ious results with word s, which hav e shown failu re to access lexical representations ,

[t (35) = 5.6, P < .0001]. In ad dition, she respond ed faster to pictu res preceded by the id entical than related prim es [t (35) = 2.8, P