semantic processing and orthographic specificity in hemispatial neglect

Semantic Processing and Orthographic Specificity in Hemispatial Neglect

Regina McGlinchey-Berroth and William P. Milberg Department of Veterans Affairs Medical Center, West Roxbury and Harvard University Medical School

Mieke Verfaellie Boston liniversity School of Medicine

Laura Grande Department of Veterans Affairs Medical Center, West Roxbury

Mark D’Esposito Hospital of the LJniversity of Pennsylvania

Michael Alexander Boston IJniversity School of Medicine

Abstract

Two sets of experiments, each consisting of a semantic priming task and a discrimination task, investigated the pro- ceedings of lexical information in the neglected visual field. In the semantic priming task, subjects made lexical decisions to target words preceded by lateralized word primes; in the discrimination task. they indicated which of two words corre- sponded to a target word presented to the left visual field (LVF) or right visual field (RVF). The first set of experiments indicated that although patients were unable to discriminate words presented in the LVE they showed significant priming when LVF primes were followed by semantically related targets compared to unrelated targets. The second set of experiments further examined the nature of this priming effect by compar- ing priming in a condition in which primes were semantically

INTRODUCTION

An increasing body of evidence suggests that patients with hemispatial neglect following right hemisphere lesions process visual information present in the lcft visual field that they cannot respond to directly. Perhaps the earliest indication that patients with neglect had at least some access to information present in the affected field was described by Kinsbourne and Warrington ( 1 962)- who noted that the reading errors committed by patients with riglit hemisphere lesions maintain the length of the words presented to them in a tachistoscopic reading task. Volpc, Ledoux, and Gazzaniga (1979) found that patients who could not name pictures of objects presented in the neglected field performed above chance on a cross-field matching task. Marshall and Hal-

related to the target word ( e g , TEA-CIJP) and a condition in which primes were unrelated to the target word, but orthographically similar to a related prime ( e g , PEA-CIJP). This experiment replicated the previously established semantic priming effects and demonstrated significant negative priming for targets preceded by LVF primes that were orthographically similar to a semantically related word. Again, patients performed at chance in the forced-choice discrimination task when targets were presented in the LW These findings indicate that semantic processing of neglected lexical information is based on fully specified perceptual and orthographic information. A lateral inhibitory mechanism is proposed that maximizes the probability, albeit unsuccessfully. that neglected ortho- Waphic information will reach awareness.

ligan (1988) showed that a patient’s preference for a picture of a house was influenced by thc presence or absence of “flames” on the left side of the picture even though she judged both pictures to be identical.

More recently, semantic processing of neglected visual information was reported in a study by McGlinchey-Ber- roth, Milberg, Verfaellie, Alexander. and Kilduff (1993). I n that study, implicit processing was assessed within the context of a semantic priming, lexical decision task in which patients judged the lexical status (i.e., word/nonword) of target letter strings that were presented in midline and preceded by lateralized picture primes. On critical trials, the priming displays were composed of one line drawing that was either related or unrelated to the target word and one nonsense figure. Like normal control subjects, neglect patients showed equivalent

priming when related primes were presented to the left visual field 0 or to the right visual field (RVF). A forcedchoice discrimination task was also administered to each subject as an objective measure of awareness of the contralesional stimuli (Cheesman & Merikle, 1986). In this task, the priming displays were followed by two vertically aligned drawings; the target was the drawing that had been contained in the priming display and the second drawing was a foil that did not bear any visual similarity to the target. The discrimination task demonstrated that the same patients who showed semantic priming from LVF primes did not have explicit knowl- edge of the LVF picture primes, as their performance was not better than would be expected by guessing. Similar findings of implicit semantic processing have also been obtained in studies of cross-field matching (Berti & Riz- zolatti, 1992; Verfaellie, Milberg, McGlinchey-Berroth, & Grande, 1995).

Most studies demonstrating intact implicit processing in neglect have used pictures of common objects as stimuli. In part, these earlier studies used pictorial representations because pictures require relatively little transformation before activating representations in the semantic memory system (Potter & Faulconer, 1975). As such, null results would likely not be due to deficits in the ability to encode more abstract representations. However, given that semantic activation has now been clearly demonstrated for neglected pictorial stimuli, the use of abstract lexical information may be helpful in assessing the quality and specificity of the visual information available for processing in neglect. Specifically, the question arises whether the visual representation of neglected stimuli is specific enough to support the formation of orthographic and lexical representations that themselves form the basis of semantic activation. The ability of neglect patients to process orthographic and perhaps lexical information is implied by the length effect in neglect dyslexia (e.g., Kinsbourne & Warrington, 1962). Orthographic processing was also demonstrated in a study by Audet, Bub, and Lecours (1991) who found that their neglect patient was faster to respond to a centrally located letter if it was flanked in the neglected visual field by the same letter compared to a different letter. More directly, Ladavas, Paladini, and Cubelli (1993) reported semantic priming from lexical information in the study of a single patient who could not read aloud the priming words that produced the effect.

To further examine the processing of lexical information in the neglected field, two pairs of experiments were conducted in the current investigation. The first pair was composed of a semantic priming lexical decision task (Experiment la) and a forced-choice discrimination task (Experiment lb). These two experiments are similar to those reported in McGlinchey-Berroth et al. (1993), but use words and nonwords as the priming stimuli rather than pictures. The second pair of experiments, also comprised of a lexical decision task (Experi-

ment 2a) and a discrimination task (Experiment 2b), investigated the extent to which orthographic information is specified in the neglected visual field.

EXPERIMENT la AND lb: SEMANTIC PRIMING AND DISCRIMINATION OF LEXICAL INFORMATION

The purpose of these experiments was to determine whether orthographic representations generated for word stimuli in the neglected field are sufficiently specified so as to provide the basis for lexical encoding and subsequent semantic activation. To assess this possibility, a semantic priming, lexical decision task was employed together with a forced-choice discrimination task. All stimuli were composed of orthographically correct word and nonword letter strings. In the lexical decision task, subjects made wordhonword judgments for centrally located letter strings that were preceded by a lateralized word priming stimulus and a filler stimulus. In the discrimination task, the priming stimuli were followed by two vertically aligned words and the subjects were asked to judge which of the two had been presented in the preceding display. The foil in the discrimination task was selected specifically to be easily discriminable from the target, even if subjects perceived only the rightmost portion of the target in the initial display. As an explicit task, the discrimination task provides a critical control for the semantic priming task in that it uses the same stimuli and presentation conditions but requires awareness of the lateralized stimuli.

Evidence of lexical processing of information in the neglected space was expected to be revealed by an overall semantic priming effect not dependent on the visual field in which the priming stimuli appeared.As this information was not expected to be processed to levels sufficient to support awareness, a visual field effect was expected in the discrimination task such that targets initially presented in the LVF were expected to be at chance and recognized significantly less often than targets initially presented in the RVE

Results and Discussion

The results from the semantic priming task and the discrimination task are examined separately. The dependent variables for these two experiments were accuracy and decision latency in the priming task and accuracy in the discrimination task. To satisfy the assumption of homogeneity of variance, statistical analysis of the decision latency data was conducted on the log transformation of each subject’s raw data. As this transformation did not alter the pattern of the data across conditions, however, Figure 1 displays the mean decision latencies in milli- seconds.

292 Journal of Cognitive Neuroscience Volume 8, Number 3

2ooo 1

Y”” I I

Related Unrelated

Prime Condition

- ‘“1 Nelgect Patients RVF - Lw ] Control Subjects RVF

Figure 1. Mean laical decision latencies as a function of priming condition and visual field in Experiment la for neglect patients and normal control subjects.

Experiment la: Semantic Priming from Lexical Information

Accuracy

Due to the high level of performance in this task, statistical analysis of the accuracy data was not possible. In total, there were only two errors in the critical conditions for the patients; both were made by the same patient (patient 3): one following an LVF unrelated prime and one following an RVF unrelated prime. Similarly for normal subjects; there were only 3 errors in total made by 3 different subjects: 2 occurred following LVF unrelated primes and one following an RVF unrelated prime.

Decision Latency

As the result of generalized attentional deficits that are common in hemispatial neglect (Mesulam, 1985), patients occasionally lost set during the task. For trials in which this occurred (defined individually as decision latencies 2 2 standard deviations above the mean for the condition in which the loss of set occurred), the patients were prompted to respond by guessing and that trial was removed from the data set. This procedure resulted in the loss of 3 trials for 2 of the neglect patients, 2 trials for 3 patients, and 1 trial for one other patient.

Decision latencies were subjected to a 2 x (2 x 2) repeated measures analysis of variance (ANOVA) examining the effects of the Visual Field of the priming stimulus (left versus right) and its Relatedness to the target word (related versus unrelated) across the two groups of subjects (neglect versus control). Table 1 presents the means, standard deviations, and error rates for individual neglect patients and for the normal control group. The means, standard deviations, and error rates for nonword targets and fillers are presented in Table 2.

Overall, neglect patients took 1301 msec to judge the lexical status of targets, whereas the control subjects made these decisions in 722 msec [F(1,15) = 12.17, p < 0.011. Neither the main effect of Visual Field [F(1,15) = 0.371 nor its interaction with Group [F(1,15) = 0.021 was significant. The relationship between the prime and target was highly significant [F(1,15) = 38.70,p < 0.0011. On average,related targets were responded to in 905 msec, whereas unrelated targets were responded to in 1016 msec. This produced an overall priming effect of 111 msec. This priming effect remained stable across Groups [F(1,15) = 2.071 and Vis- ual Fields [F( 1,15) = 1.171. Lastly, the pattern of priming for each subject group was equivalent in the left and right visual field [F(1,15) = 0.781, as depicted in Figure 1.

Experiment Ib: Forced Choice Discrimination

The discrimination data were analyzed in two ways. First, each subject’s percentage correct was determined as a function of the visual field in which the target initially appeared. These data were then used to determine whether patients performed this task better than chance in each visual field (chance determined to lie between 32 and 68%). Additionally, a repeated measures ANOVA was conducted examining the between subjects effect of Group and the within subjects effect of Visual Field to determine whether the pattern of performance across the visual fields was similar in the two groups of subjects.

Table 1 displays the mean percentage correct for individual neglect patients and for the control group. Clearly, neglect patients were unable to decide which word had just been presented to them in the LVF beyond what would be expected by chance. In contrast, they were able to discriminate the target word reliably when it was initially presented in the RVE Normal control subjects performed at ceiling.

Results from the ANOVA revealed that neglect patients’ overall performance of 72% correct was impaired relative to the control subjects’ performance of 94% correct [F(1,15) = 13.81,p < 0.011. Overall, RVF targets were discriminated more accurately (92% correct) than LVF targets (78% correct). However, a significant interaction between Group and Visual Field [F(1,15) = 25.60, p < 0.0011 indicated that this was the case only for

McClincbey et al. 293

Table 1. Individual Patient and Mean Normal Control Data for Lexical Decisions to Real-Word Targets Preceded by Semantically Related and Unrelated Primes in Experiment l a and Percentage Correct Discriminations in Experiment l b

Subject

LVF RVF Discrim ination

Related Unrelated Related Unrelated LVF RVF

Neglect 1

Neglect 2

Neglect 3

Neglect 4

Neglect 5

Neglect 6

Neglect 7

Neglect patients (n = 7)

Normal controls ( n = 10)

103SU 211h

Or

1114 157

0

1722 754

0

1770 730

0

822 292

0

743 237

0

1025 86 0

1176 410

0

682 154

0

1320 555

0

1381 42 1

0

1995 1110

0

1832 866

0

904 205

1

1019 468

0

1135 127

0

1369 409

1

775 206

2

1129 353

0

930 147

0

2214 1252

0

2460 2179

0

680 110

0

518 257

0

1176 158

0

1301 489

0

660 134

0

1279 407

0

1133 321

0

2048 873

0

3144 1644

0

834 174

1

878 414

0

1260 3 18

0

1511 748

1

77 1 246

1

50d

50

55

40

75

85

50

58

93

88

100

100

70

90

100

60

87

95

a Mean decision latencies (msec). Standard deviations. Total errors. ' Percent correct.

neglect patients. Specifically, means comparisons indicated that RVF targets were detected significantly more often than LVF targets for neglect patients [F(1,15) = 52.11, p < 0.0011, but that normal control subjects chose the correct target regardless of visual field [F( 1,15) = 0.551. Neglect patients were impaired relative to control subjects for both LVF [F(1,15) = 87.49,p < 0.0011 and RVF targets [F(1,15) = 4.83,p < 0.051.

The results from the semantic priming experiment indicate that even though neglect patients were slower to judge the lexical status of target items than were normal control subjects, the relative influence of the preceding prime word was equivalent to that obtained in control subjects regardless of whether the prime appeared in their neglected or unaffected visual field. The results from the discrimination data indicated that neglect patients were unable to identtfy a target that had

initially appeared in the LVF beyond what would be expected based simply on a guessing strategy, whereas they performed well above chance for RVF targets. The fact that neglect patients' performance for RVF targets, while reliable, was impaired relative to control subjects, is likely the result of generalized attentional deficits common in parietal lobe damaged patients (Mesulam, 1985) and not associated with a more specific ipsilesional impairment.

Two broad accounts can be advanced to explain the observed semantic facilitation effect. One possibility is that facilitation is based on explicit partial perception of the prime, presumably of the rightmost side of the prime, as has been suggested for other demonstrations of implicit processing in neglect (Farah, Monheit, & Wal- lace, 1991). In this view, patients perceive only the rightmost portion of the prime word and use this information


Table 2. Correct Mean Response Time, Standard Deviation, and Total Errors for Experiments l a and 2a: Semantic Priming in the Neglected Visual Field for Fillers and Nonword Targets

Trial type (pritne/target) ~ ~

Non word Non u lord Word Word Ieft/ujord rigbt/u jord left/nonulord rigbt/non word IVonuwd

Experiment la Neglect patients 1343

0

Normal controls 748

0

( n = 7 ) (508)

(n = 10) (220)

Experiment 2a Semantic Context

Neglect patients 1343 ( I ? = 3) (508)

0

Normal controls 892

0 (I? = 10) (203)

Orthographically mediated context Neglect patients 3434 (n = 3 ) (3075)

0

Normal controls 817

0 (n = 10) ( 183)

1368 (466)

1

726

1 (181)

1368 (466)

1

885 (208)

0

3675 (3088)

6

860 (261)

0

2148 (803)

6

899 (315)

5

2148 (803)

6

1075 (346)

3

2909 (1540)

9

1021

0 (24 1 )

2369 (931)

25

906 (337)

2

2369 (931)

25

1143

1 (438)

2918 ( 1088)

8

1049 (273)

0

2637 ( 1388)

22

965 (420)

11

2637 (1 388)

22

1068 (278)

7 - 255 1 (602)

10

1030 (230)

1

to determine the word’s identity using a combination of automatic and controlled lexical access processes (see Neely, 1991). For example, suppose that when presented with “CAT” in the neglected visual field, patients actually perceive “AT.” It is possible that this partially perceived word segment automatically activates a cohort of representations consistent with the perceived segment’s or- t hograp hic attributes (i.e ., “HAT, ’’ “CAT, ” “ATE, ” “STATE, ” etc.). The patient could then use this cohort to predict or guess the identity of the prime word. One problem with this view is that the orthographic cohort activated based on the perception of “AT is sufficiently large to make the accurate prediction of “CAT” highly improb- able. This position could be maintained, however, by allowing the additional possibility that activation of orthographic representations is restricted to only those lexical representations that match the input in length and letter position. This possibility is supported by studies of neglect dyslexia suggesting that the deficit occurs at a “string” level of representation mddoch, Hum- phreys, Cleton, & Fery, 1990). At this level of representation, letters are thought to be coded in reference to their Spatial position within letter strings, and not in reference to their absolute position in retinotopic space. Thus, word length information may be computed cor-

rectly and could be used to constrain lexical access. despite the inability to specify the letter(s) present in the leftmost positions within the string. In our example, if patients are aware that the letters “AT“ are the last two letters of the prime word and are preceded by one unperceived letter, the cohort activated by “AT” would be reduced to only those representations that are three letters in length and that end in “AT.” This additional constraint might make a guessing strategy more tenable.

One argument against this account is the fact that neglect patients’ performance in the discrimination task was at chance in the LVE If partial perception is sufficient to account for semantic priming, then one would also expect above chance discrimination, as partial information should also be sufficient to discriminate a LVF target from a visually unrelated foil. This argument rests, however, on the assumption that the attentional demands of the priming and discrimination task are identical. It is possible that the attentional requirements of the two tasks are sufficiently different so that partial information is available in one and not the other (see Verfaellie et al., 1995). For example, the requirement that subjects explicitly attend to information in the initial display of the discrimination task and subsequently make a judgment about that information may enhance

the existing attentional bias toward the intact field. In the semantic priming task, this is less likely to occur because there is no explicit demand for the subject to attend or respond to the priming display. In this way, explicit partial perception of lexical information in the LVF may account for semantic priming, even in the absence of above chance discrimination.

A second explanation of semantic facilitation from words in the neglected visual field is that subjects process the prime word implicitly but this processing is not sufficient for awareness. According to this view, neglected lexical information is processed to a level sufficient to activate word meanings, but this processing occurs without the phenomenal awareness that a visual stimulus was present (Kihlstrom, Barnhardt, 8i Tataryn, 1992). This explanation is consistent with past findings of implicit processing of pictorial stimuli in neglect and with findings in normal subjects demonstrating perceptual and semantic processing without awareness (Marcel, 1983; Forster, 1987; Forster, Booker, Schacter, & Davis, 1990).

h 1750 - v

v 1500 - 5 m 1250 - 5 Y

4 c:

1000 -

750 -

.r .- U

8

EXPERIMENT 2: ORTHOGRAPHIC SPECIFICITY IN LEXICAL SEMANTIC PRIMING IN NEGLECT

Experiment 2a was designed to differentiate between an explicit account of priming based on partial perception of the LVF prime and an implicit processing view that assumes that the entire prime is processed at a level sufficient to activate semantic representations without awareness. In this experiment, each patient was tested in two priming conditions: a semantic priming condition and a condition in which semantic priming was mediated by orthographic similarity. The semantic priming condition was a direct replication of Experiment la. In the orthographically mediated condition, the first letter of each prime word from the semantic condition was changed to form a semantically unrelated prime word that was orthographically similar to a semantically related word. For example, if in the semantic priming condition, subjects saw the prime word ‘TEA” and made a lexical decision to the target word ‘‘CUP,” in the orthographic condition they saw the prime word “PEA” and made a lexical decision to the word “CUP.” Each patient was also administered a discrimination task in which they were asked to identlfy a target word that was either orthographically similar or dissimilar to a foil. It was our hypothesis that if the word priming effect in patients with neglect results from explicitly perceived information of the rightmost part of the prime word, priming should be observed equally in both the semantic and orthographically mediated conditions. This is because this account assumes that all completions (including “PEA” and ‘TEA”) would be equally likely prime words. On the other hand, if neglect patients do implicitly process the prime word in total, greater priming should

be revealed in the semantic condition, with only negli- gible priming, if any, in the orthographically mediated condition.

Results and Discussion

As in Experiment 1 the results from the lexical decision task and the discrimination task are examined separately. The dependent variables for these two experiments were accuracy and decision latency in the priming task and accuracy in the discrimination task. Again, statistical analysis of the decision latency data was conducted on the log transformation of each subject’s raw data to satisfy the homogeneity assumption, although Figure 2 displays the results of the critical conditions in millisec- onds. Due to the complexity of the design of this experiment combined with the limited number of neglect patients that were tested, data analysis was performed in

Semantic Context

500 -I Related Unrelated

Prime Condition

- “‘1 Neglect Patients R V F

-i- LVI; Control Subjects 4G- R V F 1

Figure 2. Mean lexical decision latencies as a function of context condition. priming condition, and visual field in Experiment 21 for neglect patients and normal control subjects.


two ways. The first set of analyses was concerned with the between-subjects effect of Group (neglect versus control) across the within-subjects effects of Visual Field (left versus right) and Prime Type (related versus unrelated) for each Context condition (orthographic versus semantic) separately. The second set of analyses was concerned with the effects of Context, Visual Field, and Prime Type for neglect patients and control subjects separately. While somewhat repetitive, we felt that this approach would overcome the limited amount of statistical power in the neglect group, and still provide all theoretically important comparisons.

Experiment 2a: Semantic and Orthographic Priming from Lexical Information

Accuracy

Both the patients and the control subjects performed flawlessly in the semantic condition. The same was true in the orthographic condition for the controls and two of the neglect patients. One patient, however, did have difficulty performing lexical decisions in the orthographically mediated condition, making 16 errors (40%) in the critical conditions. Six of these errors occurred for targets following LVF primes (3 errors each in the orthographically similar and dissimilar condition), and 10 errors occurred for targets following RVF primes ( 5 errors each in the orthographically similar and dissimilar condition). This pattern of error data is most likely in- dicative of fatigue during this testing session, and not due to a qualitative performance difference in the two context conditions. Due to the overall high level of performance, statistical analysis was not conducted on the accuracy data.

Decision Latency

Trials on which patients responded 22 standard deviations from their individual mean for that condition were excluded from the analysis. In total this procedure resulted in a loss of 9 trials. The trials in which this occurred were distributed relatively evenly across conditions and were not systematically related to any of the factors of interest. Table 3 presents the means, standard deviations, and error rates for individual neglect patients and for the normal control group. The means, standard deviations, and error rates for nonword targets and fillers are presented in Table 2.

In the sernaiztic condition, a main effect of Group was revealed [F( 1,ll) = 18.40, p < 0.011, indicating that neglect patients' decision latencies were significantly longer (2248 msec) than those of normal control subjects (890 msec). A main effect of Prime Type [F(1,11) = 21.50, p < 0.001] also reVCAkd that decision latencies to related targets (1 155 msec) were significantly faster than to unrelated targets (1 25 1 msec). This indicates an

overall priming effect of 96 msec. No other effects or interactions were significant.

In the orthographically mediated condition, neglect patients (2743 msec) also responded more slowly [F(l,ll) = 26.94,p < 0.0011 than did normal control subjects (877 msec). The interactions of Visual Field and Group [F(l,ll) = 8.56,p < 0.051 and Visual Field and Prime Type lF(1,ll) = 6.02,~ < 0.051 were significant but were qualified by a marginally significant three-way interaction between Group, Visual Field, and Prime Type [F(1,11) = 4.07,p < 0.071, indicating a different pattern of performance across the two subject groups. In par- ticular, analysis of the neglect patients' data revealed a significant interaction between Prime Type and Visual Field [F(1,2) = 20.9,p < 0.051.This interaction reflected the fact that neglect patients were significantly slower [F(1,2) = 25.51,p < 0.051 in responding to orthographically mediated targets compared to dissimilar targets in the LVF (i.e., negative priming), whereas there was no significant effect of prime type in the RVE Analysis of the normal control subjects' data showed a nonsignificant priming effect for orthographically mediated targets regardless of the visual field in which the prime was presented. The triple interaction between Group, Visual Field, and Prime Type, while marginal statistically, is consistent with the notion that orthographic processing in the neglected visual field differs qualitatively from that in the intact field and from that observed in normal control subjects.

Three additional ANOVAs were conducted to examine the effects of Context (orthographic versus semantic) and Visual Field within each group separately. For normal subjects, decision latencies were subjected to a 2 x 2 x 2 repeated measures ANOVA that examined the effects of Context condition (semantic versus orthographically mediated), Visual Field of the prime (left versus right). and Prime Type (related versus unrelated).

For normal subjects, decision latencies to related targets (869 msec) were significantly faster than to unrc- lated targets I898 msec; F(1,9) = 6.81, p < 0.051. As presented in Figure 2, a significant Context by Prime Type interaction [F(1,9) = 5.36,~ < 0.051 indicated that this priming effect was significant only given a semantically related context [F(l,9) = 5.36,~~ < 0.051 and not given an orthographically similar context [F(l,9) < 1). No other main effects or interactions reached significance.

For neglect patients, this analysis revealed only a significant interaction of Context by Prime Type [F(1,2) = 30.77,p < 0.051, and a significant Context by Visual Field by Prime Type interaction [F(1,2) = 22.23, p < 0.051. The nature of this interaction, as depicted in Figure 2 , was further explored by analyzing the effects of Visual Field and Prime Type separately in the two context conditions. In the semantic context condition, decision latencies were affected only by Prime Type [F(1,2) = 13.08, p < 0.071. While marginal, this effect indicated

McGlincbql el ul. 297

Table 3. Real-Word Targets Preceded by Related or Similar, and Unrelated or Dissimilar Context in Experiment 2a ~~

Semantic Context

LVF RVF

Subject Related Unrelated Related Unrekated

Neglect 1

Neglect 2

Neglect 3


326lU 1506'

oc 1244

323 0

1770 304

0

819 (221)

0

3441 1354

0

1303 446

0

2134 688

0

901 (226)

0

3146 1265

0

1334 356

0

1854 282

0

849 (241)

0

3939 2470

0

1432 345

0

2115 219

0

876 (237)

0

Orthographically Mediated Context

LVF RVF

Subject Similar Dissimilar Similar Dissimilar

Neglect 1

Neglect 2

Neglect 3


4074a 1260'

3c

1581 484

0

2090 568

1

830 (207)

0

3108 1343

3

1321 408

0

1917 235

0

85 1 ( 1 86)

0

5 185 1961

5

1454 266

0

2215 442

0

836 (195)

0

~

6003 3657

5

1532 506

0

2430 224

0

865 (194)

0

'' Mean decision latencies (msec)

' T O ~ ~ I errors. Standard deviations.

that decision latencies to targets preceded by related primes (2 102 msec) were faster than to targets preceded by unrelated primes (2394 msec). Neither the main effect of Visual Field [F(1,2) < 11, nor its interaction with Prime Type [F(1,2) < 11 were significant. In the orthographically mediated context condition, decision latencies were slower to targets preceded by related primes (2767 msec) than by unrelated primes (2718 msec), though this effect failed to reach significance [F(1,2) = 2.771. The interaction between Visual Field and Prime Type [F(1,2) = 20.90,p < 0.051, however, was significant and revealed that decision latencies to targets preceded by LVF primes were significantly slower when the prime was related than when it was unrelated be., negative priming; F(1,2) = 25.51,p < 0.051. In contrast, decision

latencies to targets preceded by RVF primes were slightly faster when the prime was related than when it was unrelated, although this effect was not significant [F(1,2) = 2.01.

In summary, the priming results obtained from the normal control subjects followed the expected pattern whereby significant priming was found in the semantic condition but not in the orthographically mediated condition. In addition, the semantic priming effect did not differ significantly in the left and right visual fields for normal control subjects. The priming results from the neglect patients were more complex. The analyses indicated that neglect patients displayed a typical semantic priming effect in the semantic condition but a negative priming effect in the LVF in the orthographically medi-


ated condition. Response latencies for targets preceded by RVF primes did not differ as a function of orthographic similarity. Examination of the individual means confirmed that this pattern was present in each of the three neglect patients tested.

Experiment 2b: Discrimination of Orthographically Similar Lexical Information

The discrimination data from both testing sessions were collapsed so that the data analysis was conducted with a total of 80 trials per subject. To determine overall level of performance, percentage correct was calculated as a function of visual field and orthographic similarity. As Table 4 indicates, neglect patients performed at chance for LVF targets (chance calculated to lie between 37 and 63%), regardless of whether they were orthographically similar or dissimilar. In contrast, they performed above chance for RVF targets, regardless of whether they were orthographically similar or dissimilar. Normal control subjects performed above chance in all conditions.

Percent correct discrimination was also analyzed in a 2 x (2 x 2) repeated measures ANOVA to determine if orthographic similarity affected either group’s performance differently across left versus right visual field. Significant effects were found for Group [F( 1 , l l ) = 79.80,p < 0.0011,Visual Field [F(1,2) = 2 1 . 3 6 , ~ <0.05], and Orthographic Similarity [F(I,11) = 25.51, p < 0.001]. The interaction between Visual Field and Group was also significant [F(1,2) = 32.731,p < 0.0011. Com- parisons indicated that neglect patients performed significantly more accurately for targets presented in the RVF than for targets presented in the LVF [F(1,2) = 32.31,p < 0.0011, whereas normal control subjects performed similarly regardless of the visual field in which the target was presented [F(l,2) <l] .The three-way interaction was also significant [F(1,2) = 6 . 1 8 , ~ < 0.051. A series of means comparisons revealed that neglect patients’ discrimination performance for LVF targets was more severely impaired when the foil was orthographi-

cally similar compared to orthographically dissimilar [F(1,2) = 12.06,p < 0.011. However, their performance did not exceed chance in either of these conditions.

The findings from the discrimination task again confirm that information that is processed sufficiently to activate semantic information is not specified sufficiently to influence discrimination performance. The absence of an effect of orthographic similarity even for targets initially presented in the RVF suggests that the double simultaneous displays consisting of the target word and filler XXXs were sufficient to prevent attention from being directed exclusively to the rightmost side of the RVF.

GENERAL DISCUSSION

The current study investigated the presence and nature of semantic priming from neglected lexical information in hemispatial neglect patients. Toward this end, two sets of experiments were conducted to determine (1) if orthographic information falling within the neglected visual field is specified sufficiently to activate abstract lexical and semantic representations, and (2) if observed priming effects are based on explicitly perceived orthographic information contained within the rightmost side of the prime word, or based on fully specified, implicitly processed orthographic and semantic representations.

The pattern of results from the first pair of experiments was clear. Even though patients could not explicitly identlfy words appearing in their neglected visual field, these words nevertheless influenced the speed with which subsequent lexical decisions were made. Importantly, the magnitude of the priming effects was equivalent in both visual fields and was similar to that found in normal control subjects. The basic semantic priming effect was replicated in three additional neglect patients in the semantic condition of Experiment 2a, where again, results from the discrimination task indicated that LVF primes were not processed to levels sufficient for visual awareness.

Table 4. Percentage Correct in Experiment 2b

Lefl iiisiiul field Right ilisuul field

Ortlwgruphicully Orthogruphicully

Sirnikir Dissimilar Similur Dissim ilur

Neglect 1

Neglect 2

Neglect 3 Neglect patients (n = 3 )

Control subjects ( n = 10)

35

40

45

40

96

60

55

60

58

97

85

100

60

8 2

91

70

95

85

8S

96

In the orthographically mediated condition of Experi- ment 2a negative priming was observed when primes were presented in the neglected visual field. This finding is inconsistent with the explicit processing account hypothesized above. As discussed previously, if patients processed only the right side of primes presented in the L\T then equivalent priming should have been obtained in the semantic and orthographically mediated condition. As an alternative, we hypothesized that if neglected lexical information is processed in its entirety, but implicitly, priming would be absent in the orthographically mediated condition. This was not the case either. None- theless, we suggest that the presence of negative priming can be accommodated in the context of an implicit processing account and helps to further spec@ the level at which information processing breaks down in neglect.

A possible mechanism by which negative priming from orthographically mediated primes may occur is suggested by a recent model developed by Dagenbach, Carr, and Barnhardt (1990). These authors applied the notion of lateral inhibition to interpret a negative priming effect obtained in a study in which newly acquired vocabulary words served as primes. It was found that when a new word’s meaning is recognizable but not recallable, as in the early stages of learning, responses to other well-known related targets are inhibited. The authors speculated that this inhibitory mechanism provides a way for the semantic memory system to retrieve weakly activated representations during the learning process. According to their view, to retrieve a weakly activated semantic representation, its code is facilitated while previously established and possibly stronger codes for related items are inhibited. In this way, the relative activation of the weak representation is maximized.

A similar process may occur for orthographic information in the neglected visual field. The results from the current study imply that lexical information is specified sufficiently to activate semantic information but not to allow identification. This can be characterized as a retrieval failure in patients with symptoms of neglect whereby orthographic or lexical information is not available to awareness. Analogous to the semantic ambiguity presumed to underlie the negative priming found in the Dagenbach et al. (1990) study, we hypothesize that orthographic or lexical ambiguity may underlie the negative priming found in the LVF of negiect patients. Consider that the cohort of items initially activated upon presentation of a prime consists of all representations that are orthographically similar to the presented item (ix., prime) (e.g., Rosson, 1983; McClelland, 1987). Each of these, in turn, activates its own set of associates in semantic memory. If all of the component features of a lexical item are perceived implicitly and are fully specified, inhibition could spread to lexical items that are inconsistent with the phonological and orthographic representation of the stimulus. This inhibition would spread to the previously activated associates in semantic

memory as well. In this way, all spuriously activated information would be suppressed, thus enhancing the relative activation of the prime word and maximizing the likelihood that it will be retrieved. For example, suppose the prime word ‘SAW” is presented in the LVF followed by the target word “DOG.” Upon presentation of the word ‘SAW,’’ all lexical representations that are orthographically consistent with the prime word are activated (e.g., “SAW,” “PAW,” ‘SAF”’). In semantic memory, this activation spreads to a cohort of associated items (e.g., “CUT,” “TOOL,” “DOG,” “TREE”). To account for negative priming, two additional assumptions need to be made. First, it is hypothesized that the processing of “SAW” leads to a higher level of activation of its corresponding representation than that of orthographically similar words. Second, this activation is insufficient for awareness. Given these assumptions, inhibition is thought to spread to all representations that are not specifically associated with “SAW.” In this way, the target word “DOG” would be inhibited via its association to the orthographically similar word “PAW.”

The inhibitory mechanism proposed above maximizes the probability that neglected primes will reach awareness, but, ultimately, the retrieval mechanism still remains unsuccessful. For primes presented in the RVF, activation levels are sufficiently strong to support full retrieval of the primes and, hence, an inhibitory mechanism need not be invoked. Likewise, in normal subjects, negative priming does not occur because processing is sufficient for awareness.

The results from the present study contribute to existing evidence indicating that a great deal of high level cognitive processing occurs for neglected visual information. This conclusion is strengthened by the use of two analytic strategies. First, a task dissociation strategy was used in which performance was compared between the semantic priming task and the discrimination task. This revealed that the processing of lexical information is intact if existing semantic representations are activated, but this activation is not sufficient for awareness. Second, a manipulation within the priming task was used to indicate that orthographic and subsequent semantic processing is the result of fully specified implicit perceptual processing. Future investigations must continue to stress the limits of preserved implicit processing in neglect. In this way we may further examine issues regard- ing the perceptual integrity of neglected information and begin to understand where in the course of normal processing visual/cognitive representations break down in this disorder.

METHOD Subjects The subjects in Experiments la and l b consisted of 7 patients from the Braintree Rehabilitation Hospital (3 female, 4 male) and 10 normal control subjects. The


patients ranged in age from 64 to 76 years (mean age 71 f 4.74); education ranged from 8 to 13 years (mean education 1 1 f 2.08). The control subjects ranged in age from 59 to 7 3 years (mean age 65 f 4.45); education ranged from 12 to 16 years (mean education 14 f 1.48). Three patients and 10 normal control subjects participated in Experiments 2a and 2b. The patients ranged in age from 70 to 74 years (mean age 72 k 2.08); education ranged from 13 to 14 years (mean education 13 f0.58). The control subjects ranged in age from 59 to 7 3 years (mean age 72 f 3.0) ; education ranged from 12 to 16 years (mean education 14 f 2.0).M subjects were right- handed. For the patients, CT or MRI confirmed a single unilateral right hemisphere infarction and neuropsychological testing revealed contralesional hemispatial neglect. Patients were excluded if there was evidence of a left hemisphere lesion or a history of substance abuse.

Hemispatpal neglect was assessed with a number of tests including horizontal line bisection, figure drawing, word reading (mono- and bisyllabic), detection of double simultaneous stimulation for visual stimuli, and a series of target cancellation tasks (including letters, symbols, and lines). Patients who participated displayed evidence of LVF neglect on at least two of these assessments. Of note, however, is the fact that only one of the patients had mild neglect dyslexia. AU patients had normal visual fields to single stimulation. Table 5 displays demographic, clinical, and lesion localization information for the patients included in this sample.

Apparatus

All experiments employed an Apple Macintosh niicro- computer. The software (Psychlab) was designed for the control of stimulus presentation, timing, and the record- ing of responses with millisecond accuracy. Manual responses were recorded by two telegraph keys interfaced with the computer via a modified keyboard. Face position relative to the computer screen was maintained with a standard opthalmological chin rest centered 12 in. from the screen. The chin rest height was adjustable so that the individual subject’s eyes were at the same level as the fixation point presented on the video monitor.

General Procedure

Each neglect patient was tested in a quiet, softly lit laboratory space located at the Braintree Rehabilitation Hospital, MA. Control subjects were tested at the Brock- tonmest Roxbury Department of Veterans Affairs Medi- cal Center (DVAMC) and the Boston DVAMC. The entire protocol, including the experiments and neuropsychological testing (for the patients), was conducted over a varying number of sessions ranging from one-half to 1 hr each, depending on the patient’s ability to sustain attention. In the initial session(s), background informa-

tion and neuropsychological data were collected. The two sets of experiments were administered in two subsequent sessions. Within each session, the priming task preceded the discrimination task. We felt this to be the most conservative order of testing because it would not permit exposure during the discrimination task to influence the presence/absence of priming during the priming task. All the patients completed the entire protocol within a 2-week period.

Stimuli and Procedure

Experiment la The target stimuli were constructed from 120 high-frequency words chosen from the norms of Francis and Kucera (1982) and were 3 to 5 letters in length. Half of these stimuli were used as real word targets, and the remaining stimuli matched for word length, were changed into nonwords by substituting one letter, but respecting the rules of English orthography. The letter that was substituted occurred equally often on the left and on the right side of the word.

The priming stimuli were also constructed from 120 high-frequency words, 3 to 5 letters in length. For the 60 trials in which the target stimulus was a real word, three different types of primes were used. One-third (i.e., 20) of the primes was semantically related to the target, one-third was unrelated to the target, and the final one- third was composed of nonword filler stimuli. For the 60 trials in which the target stimulus was a nonword, half of the primes were real words, and the other half were nonwords.

All priming displays were composed of a priming stimulus and a neutral “filler” stimulus (string of Xs equal in length to the priming stimuli). The presentation of the prime and filler stimuli was counterbalanced across visual field so that half of the primes in each condition was presented in the LVF and half was presented in the RW. The double simultaneous displays were intended to in- crease the likelihood that attention would be focused centrally. The order of the trials was randomized and remained constant across subjects.

The beginning of each trial was signaled with an asterisk centered on a video display monitor for 500 msec. The asterisk was followed immediately by the presentation of the prime and filler stimuli for 200 msec. The relatively short exposure of the priming stimuli was intended to minimize the possibility of saccadic eye movements. The prime and filler were centered 1.5” to the left and right of the center of the display monitor. Following an interstimulus interval ( IS0 of 400 msec, the target string was displayed centrally and remained on the screen until a response was made by the subject. Yes/No (i.e., word/nonword) decisions and accuracy were recorded automatically by the computer.

The subjects were told that two events would appear on the computer screen, one immediately following the

McGlincbey et al. 301

Table 5. Demographic, Clinical, and Lesion Information for Experiment la

Patient Age Sex Education Hemiparesis Lesion localization

1 73 F 14 Moderate Anterior, posterior PVWM~

2 75 M 8 Moderate Posterior parietal; superior temporal

3 76 M 13 Moderate Posterior parietal; posterior PVWM

4 73 M 10 Mild Posterior parietal

5 65 M 10 Mild Posterior parietal; superior, lateral temporal

6 64 F 10 Severe Frontal; caudate; putamen; ALIC‘; anterior, posterior PVWM

7 72 F 12 Severe Posterior parietal Posterior PVWM

8 70 F 14 Mild Posterior parietal

9 73 F 13 Severe Superior parietal; thalamus; anterior, posterior PVWM

10 74 F 13 Mild Posterior temporal/parietal

a Patients 1-7 participated in Experiment 1; patients 8- 10 participated in Experiment 2; patient 8 had mild neglect dyslexia. Periventricular white matter. Anterior limb of internal capsule.

other. It was explained that the first event would be composed of two “letter strings” that might or might not represent real English words. It was stressed that all the words were common and were not intended to be tricky or difficult to recognize. They were asked to focus on the centrally located fixation point, as the letter strings would be displayed only for a very brief period of time. In addition, they were told that the second event appearing on the screen would be a letter string that might or might not spell a real English word.They were instructed to simply watch the display monitor and to respond only to the second event by depressing, with the index finger of their right hand, the telegraph key marked ‘YES” if the target letter string was a word and to depress the key labeled “NO” if it was not a word. The telegraph keys were oriented vertically so as to minimize the risk of a rightward response bias. Prior to responding, the subjects rested their index finger on a board midway between the two telegraph keys. They were encouraged to respond as accurately as possible.

After practice, subjects were presented with 120 ex- perimental trials, divided into 6 blocks of 20 trials each. During the course of testing, the patients were carefully watched by the examiner to ensure that they maintained attention to the task. Patients were given a brief rest after each block.

Experiment l b

The 120 stimuli in this experiment consisted of 80 words: 40 targets that had appeared as related or unrelated primes in Experiment la and 40 foils that were matched to the targets with regard to length and fre-

quency. Additionally, there were 40 filler stimuli composed of strings of ‘X”s that matched the target with regard to character length. All foils were visually and semantically unrelated to the target. They were selected with the additional proviso that the right side of the foil should be sufficiently different from the target so as to allow correct discrimination even if patients only explore the right side of the word.

During the initial presentation, the correct target word was presented in the LVF on half of the trials and in the RVF on the remaining half of the trials. A filler stimulus was presented in the opposing VE Following each presentation of these lateralized stimuli, a test trial was presented, consisting of the target word as well as a foil, presented vertically, one on top of the other. The positioning of the correct choice on either the top or bottom of the display was counterbalanced across trials. The order of the trials was randomized and remained constant across subjects.

At the onset of each trial, an asterisk was presented on the center of the screen for 500 msec. This was followed by the simultaneous presentation of a target and a fdler stimulus, centered 1.5” to the left and the right of center, for 200 msec. These displays were identical to the priming displays in Experiment la. Following an IS1 of 400 msec, the target stimulus and a real word foil were presented in midline, centered 2’ above and below the center of the screen. To provide the subject with unlimited time to explore these stimuli, they remained on the screen until a response was made. Re- sponse accuracy was recorded automatically by the computer.

Subjects were instructed to keep their eyes on the


fixation point, and to be ready for the presentation of two stimuli, which would be presented only briefly. They were then asked to attend to the two vertically aligned stimuli, and to depress one of two keys, also aligned vertically, to indicate which of the two stimuli appearing on the video monitor they had just viewed. They were asked to depress the telegraph key labeled ‘TOP” if they thought they had just seen the word located on the top and to depress the key labeled “BOTTOM” if they thought they had just seen the word located on the bottom.

After practice, subjects were presented with 40 ex- perimental trials, divided into 2 blocks of 20 trials each. During the course of testing, the patients were carefully watched by the examiner to ensure that they maintained attention to the task. Patients were given a brief rest after each block. AU subjects were instructed to respond as accurately as possible.

Experinzent 2a

In total this experiment was composed of 240 trials; half of the trials comprised the semantic priming condition and the other half comprised the orthographic priming condition. These conditions were blocked and adminis- tercd to subjects on separate testing days.

The target stimuli in the semantic context condition were constructed from 120 high-frequency words chosen from the norms of Francis and Kucera (1982). Half of these stimuli was used as real word targets, and the remaining stimuli matched for word length were changed into nonwords by substituting one letter, but respecting the rules of English orthography.

The priming stimuli were also constructed from 120 high-frequency words. For the 60 trials in which the target stimulus was a real word, three different types of primes were used. One-third (i.e., 20) of thc primes was semantically related to the target, one-third was unrelated to the target, and the final one-third was coniposed of nonword fillcr stimuli. For the 60 trials in which the target stimulus was a nonword, 40 of the primes were real words, and the remaining 20 were nonword primes.

For the orthographically mediated context condition, the first lctter of each prime word in the semantic context condition was replaced to form orthographically similar but semantically unrelated primes that matched the original prime word with regard to word frequency and length. All other stimuli were identical to those in the semantic condition. Experiment 2a was otherwise identical in procedure to Experiment la.

Experimeiit 213

In total, this experiment consisted of 80 trials, split evenly across two testing sessions. Within each session, 40 words served as targets and 40 served as foils. All of the stimuli were words that served as primes in Experi-

ment 2a. Each word was presented once during each testing session.

Within each session, a target word was presented in the left visual field on half of the trials, and in the right visual field on the remaining half of the trials. Following each presentation of these lateralized stimuli, a test trial was presented, consisting of the target word as well as a foil, presented vertically, one on top of the other. On half of the trials ( n = 10 per session), foils were orthographically similar (with the exception of the left-most letter) but semantically unrelated to the target ( e g , corn and horn). On the other half of the trials, foils were semantically unrelated and orthographically dissimilar to the target. The positioning of the correct choice on either the top or bottom of the display was counterbalanced across trials.

Each subject received one list per testing session. A total of 4 lists were prepared that counterbalanced for orthographic similarity and visual field. That is, target items that were orthographically similar in one list were dissimilar in the second list. One list was administered following the semantic priming condition of Experiment 2a and one list was administered following the orthographic priming condition. Experiment 2b was otherwise identical in procedure to Experiment 1 b.

Acknowledgments

This work was supported by Grant NS29342-02 t o William I? Milberg at the GRECC, BrocktonNest Roxbury DVAMC. and Harvard Medical School. We express our thanks to Dr. John Ween for screening our patients. Thanks also t o A. Joelle <;ran& and Elisa Rolton for testing subjects and to Patrick Kilduff for providing software prowamnung and computer support.

Reprint request should be sent t o Regina Mc<;liiichey-Berroth. GRECC/VA Medical Center (182). 1400 VFW Parkway West Roxbury, MA 02132.

REFERENCES Audet, T.. Bub, I).. 81 Lecours, A. R. (1991). Visual neglect and

left-sided context effects. Bruin mid Copition, 16. 1 1-28. Berti, A, , 81 Rizzolatti, G. (1992). Visual processing without

awareness: Evidence from unilateral neglect.Joirma/ of Cogriitive Neuroscietice, 4, 345-351.

scious from unconscious perceptual processes. Caizrrciiari Journal cf Psycbolog~~, 40, 343 -367.

Dagenbach, D., Carr, T. H . , 81 13arnhardt. T. H. (1990). Inhibi- tory semantic priming of lexical decisions due t o failure to retrieve weakly activated codes. Joiirmd of Experinieii- tal Psycholo~y~ 16, 328-340.

scious perception of “extinguished” visual stimuli: Reassess- ing the evidence. Neu ropsyclmlog icr, 29. 94 9 - 9 58 .

Forster, K. I . (1987). Form-priming with masked primes: The best-match hypothesis. In M. Coltheart (Ed.), Attentioil and perffiri?zut?ce (pp. 127-146). Hillsdale, NJ: Erlbaum.

Forster, K. I . , Booker,J., Schacter, D. L . , & Davis. (;. (1990). Masked repetition priming: Lexical activation o r novel

Cheesman. J., 81 Merikle. I? M. (1986). Distinguishing con-

Farah, M. J . , Monheit, M . A. , 81 Wallace, M. A. (1991). Uncon-

memory trace? Bulletin of the Psychonomic Socieq, 28,

Francis, W., & Kucera, H. (1982). Frequency analysis of Eng- fish usage. Boston: Houghton Mifflin.

Khlstrom, J. E, Barnhardt, T. M., & Tataryn, D. J. (1992). Im- plicit perception. In R. E Bornstein & T. S. Pittman (Eds.), Perception without awareness (pp. 17-54). New York: Guilford Press.

Kinsbourne, M., & Warrington, E. K. (1962). A variety of reading disability associated with right hemisphere lesions. Journal of Neurolou, Neurosurgery and Psychiaty, 25, 339-344.

Eidavas, E., Paladini, R., & Cubelli, R. (1993). Implicit associa- tive priming in a patient with left visual neglect. Neuropsy- cbologia, 31, 1307-1320.

Experiments on visual masking and word recognition. Cog- nitive Psycholou, 15, 197-237.

Marshall, J. C., & Halligan, I? W. (1988). Blindsight and insight in visuo-spatial neglect. Nature (London), 336, 766-767.

McClelland, J. L. (1987). The programmable blackboard model of reading. In J. L. McClelland & D. E. Rumelhart (Eds.), Parallel distributed processing (pp. 122- 169). Cam- bridge: MIT Press.

der, M., & Kilduff, I? (1993). Semantic processing in the ne-

341-345.

Marcel, A. J. (1983). Conscious and unconscious perception:

McGlinchey-Berroth, R., Milberg, W. I?, Verfaellie, M., Alexan-

glected visual field: Evidence from a lexical decision task. Clinical NeuropsychoIou, 10, 79- 108.

Mesulam, M.-M. (1985). Attention, confusional states and neglect. In M.-M. Mesulam (Ed.), Principles of behavioral neurolou. Philadelphia: E A. Davis.

Neely, J. H. (1991). Semantic priming effects in visual word recognition: A selective review of current findings and theories. In D. Besner & G. W. Humphreys (Eds.), Basic processes in reading: Visual word recognition (pp. 264- 336). Hillsdale, NJ: Erlbaum.

Potter, M. C., & Faulconer, B. A. (1975). Time to understand pictures and words. Nature (London), 253, 437-438.

Riddoch, J., Humphreys, G., Cleton, I?, & Fery, I? (1990). Inter- action of attentional and lexical processes in neglect dyslexia. Cognitive Neuropsycbolou, 7, 479-517.

Rosson, M. B. (1983). From SOFA to LOUCH: Lexical contribu- tions to pseudoword pronunciation. Memo y and Cogni- tion, 11, 152-160.

Grande, L. (1995). Comparison of cross-field matching and forcedchoice identification in hemispatial neglect. Neum

Volpe, B. T., Ledoux, J. E., & Gazzaniga, M. S. (1979). Informa- tion processing of visual stimuli in an “extinguished” field. Nature (London), 282, 722-724.

Verfaellie, M., Milberg, W. I?, McGlinchey-Berroth, R., &

P ~ C h O l o ~ , 9(4), 427-434.


semantic processing and orthographic specificity in hemispatial neglect

Documents