Journal of Experimental Psychology:Learning, Memory, and Cognition1990, Vol. 16, No. 6, 1068-1076

In the public domain

Intact Text-Specific Reading Skill in Amnesia

Gail MusenSan Diego Veterans Affairs Medical Center

and Department of Psychiatry,University of California, San Diego

Arthur P. ShimamuraUniversity of California, Berkeley

Larry R. SquireSan Diego Veterans Affairs Medical Center

and Department of Psychiatry, University of California, San Diego

Amnesic patients were studied to determine whether the acquisition and retention of item-specific skills can be supported by nondeclarative (implicit) memory. In Experiment 1, subjectsread 2 different passages 3 times in succession. Reading speed improved at a similar rate in bothamnesic patients and normal subjects and was specific to the text that was read. In Experiment2, amnesic patients and normal subjects read a passage 3 successive times and then reread thesame passage after a 0-s, 10-min, 2-hr, or 1-day delay. In both groups, facilitation persisted forat least 10 min and disappeared within 2 hr. It is suggested that facilitated reading speed dependsimportantly on both semantic and perceptual information and that such information can besupported by nondeclarative memory.

A considerable body of experimental work has accumulatedin recent years to support the idea that memory is not aunitary property of the mind but is composed of multiple,different processes or systems (Hintzman, 1990; Richardson-Klavehn & Bjork, 1988; Schacter, 1987; Squire, 1982; 1987;Tulving, 1985). One important kind of evidence comes fromthe study of human amnesia, which seems to reveal a naturaldivision between different kinds of learning and memory.Although amnesic patients are markedly impaired on tests ofrecall and recognition, they can learn certain perceptuomotorskills (Brooks & Baddeley, 1976; Nissen & Bullemer, 1987),perceptual skills (Cohen & Squire, 1980; Nichelli, Bahman-ian-Behbahani, Gentilini, & Vecchi, 1989), and cognitiveskills (Squire & Frambach, 1990) at a normal rate. Further-more, amnesic patients show a normal shift in preferencefollowing exposure to novel material (Johnson, Kim, & Risse,1985). They also show normal adaptation-level effects (Benz-ing & Squire, 1989), normal classical conditioning (Daum,Channon, & Canavar, 1989), and intact word priming (Shi-mamura, 1986; Tulving & Schacter, 1990).

Memory tasks can be divided into two broad categories.The memory tasks on which amnesic patients fail requireconscious recollection of previous encounters or previously

This research was supported by the Medical Research Service ofthe Department of Veterans Affairs, National Institute of MentalHealth (NIMH) Grant MH24600, the Office of Naval Research, theMcKnight Foundation, a Bioscience Grant for International JointResearch from NEDO, Japan, and an NIMH postdoctoral fellowshipto Gail Musen (MH09858).

We thank Joyce Zouzounis, Matthew MacKinnon, Teresa Dok-sum, and Kathleen Fuchs for research assistance. We are grateful toBetty Ann Levy, Michael Masson, and Daniel Schacter for usefulcomments on an earlier version of this article.

Correspondence concerning this article should be addressed to GailMusen, Veterans Affairs Medical Center (V116A), 3350 La JollaVillage Drive, San Diego, California 92161.

acquired facts. Such tasks have been considered to depend ondeclarative (also called representational or explicit) memoryand to require the integrity of medial temporal lobe or dien-cephalic brain structures that are damaged in amnesia (Cohen1984; Shimamura, 1989; Squire, 1986). In contrast, intactlearning in amnesia is implicit and accessible only throughperformance. The tasks that can be learned include a hetero-geneous collection of skillful behaviors, conditioning tasks,and other cases in which experience changes the facility foroperating in the world but without requiring the consciousrecollection of any previous event. The learning in these casesis nondeclarative (implicit) and independent of the brainstructures damaged in amnesia.

A number of issues continue to be explored concerning thekind of information that is available to declarative and non-declarative memory. Amnesia provides an especially favora-ble opportunity for addressing such issues because in amnesianondeclarative memory can be studied without much likeli-hood that conscious memory strategies will contaminate taskperformance. Accordingly, a finding of entirely intact per-formance in amnesia provides particularly good evidence thata task does not materially depend on declarative memory.

We used this approach to consider what kind of informationunderlies learned skillful behavior. Acquired skills could po-tentially depend on both generic information, which is ac-quired cumulatively across many study trials, and specificinformation, which is obtained from the particular kinds ofstudy trials that are encountered. The question is whetherboth kinds of information are nondeclarative, that is, whetheramnesic patients can acquire both generic and specific com-ponents of skill learning. This question was initially addressedby assessing the ability of amnesic patients to learn to readmirror-reversed word triplets that were either unique or re-peated (Cohen & Squire, 1980). With unique word triplets,the amnesic patients improved at a normal rate, but withrepeated triplets the normal subjects performed better thanthe amnesic patients. This finding could be taken to suggest

1068

SPECIFIC READING SKILL IN AMNESIA 1069

that information about specific encounters was not availableto amnesic patients. However, as others have pointed out(Masson, 1986; Moscovitch, Winocur, & McLachlan, 1986;Nichelli et al, 1989), the same words were always presentedtogether in the same order in each repeated word triplet.Accordingly, normal subjects might have used declarativememory strategies to retrieve the second and third words fromthe triplet after deciphering the first word. Thus, amnesicpatients may have performed worse because they could notuse this strategy, not because they were unable to acquireitem-specific skills. In a recent study of one amnesic patient,Nichelli et al. (1989) reduced the possible influence of decla-rative memory by presenting only one mirror-reversed wordat a time. The patient improved at reading both unique andrepeated words to a similar extent, though it was difficult todetermine if the patient's performance was fully normal.

A considerable body of research with normal subjects sug-gests that specific information is retained as part of skillfulbehavior. For example, in experiments by Kolers (1976),subjects practiced reading passages of geometrically trans-formed text (e.g., inverted or mirror-imaged.) Thirteen to 15months later, subjects read either familiar passages or newpassages, both presented in the same typography that hadbeen practiced earlier. Familiar passages were read morequickly than new ones. Kolers suggested that the advantageof familiar over novel passages was due to the retention oftext-specific, pattern-analyzing skills.

In similar experiments, Masson (1986) measured readingspeed for word triplets that were presented in mirror-reversedform and in alternating small and capital letters. In the studyphase, all letters were encountered equally often in bothuppercase and lowercase. At test, subjects read faster whenletters were presented in the same alternating case pattern thatthey were in at study than when the letters appeared in adifferent alternation. Moreover, reading speed was faster inboth of these conditions than when subjects read entirely newwords. Thus, the skilled performance of these subjects re-flected information about the exact visual detail of specificinstances that had been studied.

Although the experiments reported by Masson (1986) showthat facilitated reading of single words can depend on highlyspecific information, the evidence for text reading is not soclear. In one experiment, auditory presentation of textualmaterial facilitated reading when the same material was sub-sequently presented visually in a transformed typography(Kolers, 1975). This finding shows that semantic, conceptualinformation can contribute to facilitated reading speed. Inanother study (Tardif & Craik, 1989), subjects read differenttexts in two unfamiliar typographies (e.g., mirror-reversed orinverted) and in normal typography. One week later, subjectsread the old passages and new passages in both the practicedtypographies and in a third, unfamiliar typography (e.g.,reversed spelling). Passages were read faster if they had beenread previously and if the typography had been practiced. Inaddition, verbatim passages were reread faster than para-phrased passages, which in turn were read faster than newpassages. However, there was no additional benefit fromrereading a passage in the same typography that it had ap-peared in during the study session. Tardif and Craik suggested

that facilitated rereading of text passages depends on theretention of general pattern-analyzing skills (not on instance-specific pattern-analyzing skills), together with the retentionof conceptual information about the text (both gist and spe-cific lexical information). Others have also concluded thatsemantic, conceptual factors are important in rereading (Carr,Brown, & Charalambous, 1989; Graf & Levy, 1984; Horton,1985; Levy & Burns, in press; Levy & Kirsner, 1989; Levy,Newell, Snyder, & Timmins, 1986; Masson & Sala, 1978).

If the rereading of text does depend on specific factorsrelated to the particular words in the text, the question arisesas to whether these factors are examples of nondeclarative(implicit) memory. One way to address this question is to askwhether amnesic patients would show facilitated readingspeed for textual material and whether this effect is text-specific. This question was addressed recently in a study ofreading speed involving a mixed group of memory-impairedpatients (Moscovitch et al., 1986). At study, subjects read setsof sentences in either normal or inverted typography. At test,which occurred at two retention intervals for each subject (1-2 hr and 2 weeks), subjects were asked to read old and newsentences as rapidly as possible. Both groups read old sen-tences more quickly than new sentences. However, becausecontrol subjects and memory-disordered subjects had differ-ent initial reading speeds (i.e., baseline speeds), it was difficultto determine if the two groups improved to a similar degree.

In the present experiments we pursued further the issuestudied by Moscovitch et al. (1986)—namely, whether am-nesic patients can acquire and retain text-specific readingskills. In the first experiment, we asked whether amnesicpatients would improve their reading speed at the same rateas normal subjects. We also asked whether improvement wasdue to a general skill—for example, an increased speed ofreading aloud—or whether improvement was specific to thetext that was read. In the second experiment we examinedhow long facilitated reading speed persists.

Experiment 1

Amnesic patients and normal subjects read two differentstories three times in succession. Facilitated reading speedwould be reflected in progressively faster times for successivereadings of the first story. If facilitated reading speed is specificto the practiced text, then the first reading of the second storyshould be slower than the third and final reading of the firststory and at about the same speed as the first reading of thefirst story.

Method

Amnesic subjects. We tested 8 amnesic patients (see Table 1), allof whom have been studied in our laboratory on many occasionsduring the past few years. Patients were selected who had shown noevidence of slowness on previous tests in which they had participated.We used this selection factor in order to try to equate control subjectsand amnesic patients for baseline reading speed.

Four of the patients had alcoholic Korsakoff s syndrome. Theyhad all participated in either a magnetic resonance (MR) imaging

1070 G. MUSEN, A. SHIMAMURA, AND L. SQUIRE

Table 1Patient Characteristics

Patient

V. F.D. M.P. N.R. C.N. A.

M. G.

J. L.L. M.

M

Age

7055627351

57

705962.1

Etiology

KorsakoffKorsakoffKorsakoffKorsakoffPenetrating

brain injuryThalaraic in-

farctionUnknownAnoxia

WAIS-R

10310194

106120

111

116111107.8

Attention

1019281

115102

113

122132107

Verbal

7855777667

89

738775.3

WMS-R

Visual

7264739789

84

839682.3

General

7250678068

86

749073.4

Delay

6651537271

63

586562.4

Note. WAIS-R = Wechsler Adult Intelligence Scale-Revised; WMS-R = Wechsler Memory Scale-Revised. The WAIS-R and each of thefive indices of the WMS-R yield a mean score of 100 in the normal population with a standard deviation of 15.

study, which demonstrated marked reductions in the volume of themammillary nuclei (Squire, Amaral, & Press, 1990), or in a quanti-tative computerized tomography (CT) study, which demonstratedreduced thalamic density together with frontal lobe atrophy (Shima-mura, Jernigan, & Squire, in press). Of the 4 other patients, 2 (L.M.and J.L.) have hippocampal pathology identified with MR imaging(Press, Amaral, & Squire, 1989). Patient L.M. became amnesic in1984 as the result of a respiratory arrest that occurred during anepileptic seizure. J.L. became amnesic gradually during a period ofabout 2 years (from early 1985 to early 1987); his cognitive status hasremained stable since that time. The 2 other amnesic patients (N.A.and M.G.) have diencephalic lesions confirmed by MR imaging (forN.A., Squire, Amaral, Zola-Morgan, Kritchevsky, & Press, 1989; forM.G., unpublished observations). Patient N.A. became amnesic, pri-marily for verbal material, when he sustained in 1960 a stab woundto the left diencephalic region with a miniature fencing foil (Teuber,Milner, & Vaughan, 1968). Patient M.G. became amnesic in 1986following a bilateral medial thalamic infarction. The present studywas concerned with the overall performance of amnesic patients, andthe patients are therefore treated here as a single group.

The 8 amnesic patients averaged 62.1 years of age and 12.1 yearsof education. They had an average Wechsler Adult Intelligence Scale-Revised (WAIS-R) IQ of 107.8. Individual IQ and Wechsler MemoryScale-Revised (WMS-R) index scores appear in Table 1. Immediateand delayed (12 min) recall of a short prose passage averaged 5.5 and0 respectively (21 segments total; Gilbert, Levee, & Catalano, 1968).Scores on other memory tests appear in Table 2. Note that the scoreson the word recall test in Table 2 are above zero because on this testof immediate recall, several items can be retrieved from immediatememory, which is intact in amnesia. In addition, the mean score onthe Dementia Rating Scale (Mattis, 1976) was 133 points (maximum= 144, range = 119-143), with most of the points lost on the memorysubportion of the test (6.6 points). The average score on the BostonNaming Test was 54.6 (maximum = 60, range = 48-59). Scores fornormal subjects on these same tests can be found elsewhere (Jan-owsky, Shimamura, Kritchevsky, & Squire, 1989; Squire et al., inpress).

Healthy control subjects. Five men and 4 women served as acontrol group for the amnesic patients. All were employees or vol-unteers at the San Diego Veterans Affairs Medical Center. They

Table 2Performance on Standard Memory Tests

Patients/controlsPatients

V. F.D. M.P. N.R. C.N.A.M. G.J.L.L. M.

MControls (n = 8)

Diagramrecall

8023

1701

115.2

20.6

Pairedassociates

0-0-00-0-21-1-10-0-30-0-20-0-20-0-01-1-3

0.3-0.3-1.66.0-7.6-8.9

Wordrecall (%)

27322919493340443471

Wordrecognition (%)

91568385937193988497

50 words

272427373430313030.041.1

50 faces

312938304234203732.638.1

Note. The diagram recall score is based on delayed (12-min) reproduction of the Rey-Osterrieth figure (Osterrieth, 1944; maximum score =36). The average score for copying the figure was 26.5, a normal score (Kritchevsky, Squire, & Zonzounis, 1988). The paired associate score isthe number of word pairs recalled on three successive trials (maximum score = 10/trial). The word recall score is the percentage of wordsrecalled out of 15 across five successive study-test trials (Rey, 1964). The word recognition score is the percentage of words identified correctlyacross five successive study-test trials (yes/no recognition of 15 new words and 15 old words). The score for words and faces is based on a 24-hr recognition test of 50 words or 50 faces (modified from Warrington, 1984; maximum score = 50, chance = 25). The mean scores for normalsubjects shown for these tests are from Squire and Shimamura (1986). Note that N. A. is not severely impaired on nonverbal memory testsbecause his brain injury is primarily left unilateral.

SPECIFIC READING SKILL IN AMNESIA 1071

averaged 60.9 years of age, had 13.7 years of education, and obtainedWAIS-R subtest scores of 22 for Information (compared with 20 forthe patients) and 57.6 for Vocabulary (compared with 57.9 for thepatients). Immediate and delayed (12 min) recall of a short prosepassage averaged 7.0 and 6.4, respectively.

Materials. Excerpts of two short stories from Reader's Digest andEsquire Magazine, each containing about 20 lines of text, wereselected as study material (204 or 223 words in length).

Procedure. Subjects were told that they would be asked to readstories several times and that they would then be asked questionsabout their content. Subjects then read each story aloud three timesin succession. The order of the two stories was counterbalanced acrosssubjects. Subjects were told to read the stories as quickly as possiblebut not so quickly that they would not understand what they werereading. The interval between each reading was just long enough toremind subjects to read as quickly as possible (5-10 s).

After the third reading of the first story, subjects read the secondstory three times following the same procedure. An interval of about10-15 s separated the readings of the two stories. The time needed tocomplete each reading was recorded with a stopwatch. After the threereadings of the second story were completed, subjects were given 10,three-alternative, forced-choice questions about the first story fol-lowed by 10 similar questions about the second story.

no effect of group (F< 1) and no interaction, F(2, 30) = 2.1,p > .10. Second, a single analysis of variance for the datafrom each story (Group x Story X Reading Time) showedonly the expected effect of repeated reading, F{2, 30) = 10.92,p < .01, and no effect of story (F < 1) or group (F < 1) andno interaction (all Fs < 2.45, ps > .10). Finally, for eachgroup the first reading of the second story was accomplishedat about the same speed as the first reading of the first story(ts < 1.0) and was significantly slower than the third readingof the first story: For control subjects, t(S) = 3.55, p < .01;for amnesic patients, t(l) = 2.37, p < .05.

Thus, amnesic patients and control subjects both read morerapidly with repeated presentations of each story, but therewas no transfer of reading speed from the first to the secondstory. The control subjects performed much better than theamnesic patients on the questions about story content: Story1, 90% versus 68.9%; Story 2, 88.9% versus 63.9%; ft(15) >2.8,/*<.01.

Discussion

Results

The time required to read each story declined with repeatedreadings (see Figure 1). Reading speeds appeared to changesimilarly for amnesic patients and control subjects. The av-erage reading time for both groups was 69.4 s for the firstreading of the first story and 63.3 s for the third reading.There was no discernible transfer of reading speed to thesecond story (for both groups the first reading of the secondstory averaged 69.7 s; the third reading averaged 64.0 s).

Statistical analysis confirmed these impressions. First, sep-arate two-way analyses of variance showed a significant re-duction in reading speeds for both stories: Story 1, F(2, 30)= 10.75, p < .01; Story 2, F(2, 30) = 7.92, p < .01. There was

Amnesic patients improved their reading speed at a normalrate despite impaired declarative memory for the content ofwhat was read. The facilitation was specific to the text thatwas read. These findings support the idea that facilitatedreading speed depends on nondeclarative (implicit) memory.

Experiment 2

In this experiment, we attempted to measure the timecourse of forgetting of the text-specific reading skill. Onereason for examining forgetting was to construct an additionaltest of the hypothesis that facilitated reading speed is nonde-clarative. If this skill is supported entirely by nondeclarative

80 r

S 7 5

o

%p 70

II<= 6 5

60

100

80 jj

oo

60

401 2 3

Story 1

1 2 3

Story 2

AMN CON

Figure 1. Time required to read aloud two different stories, each presented three times in succession.(AMN = amnesic patients, n = 8; CON = control subjects, n = 9. Across the six data points, standarderrors averaged 3.1 [range = 2.8-3.4] for control subjects and 2.8 [range = 2.2-3.7] for amnesic patients.The bars show performance of each group on a test of story content given immediately after the finalreading of the second story [chance = 33%]. Brackets show standard errors of the mean.)

1072 G. MUSEN, A. SHIMAMURA, AND L. SQUIRE

(implicit) memory, then amnesic patients and normal subjectsshould lose the skill at the same rate. A second reason forexamining forgetting was that information about the timecourse of the skill might suggest what kind of process underliesit. After reading text three times in succession, subjects weretested on a fourth occasion either 0 s, 10 min, 2 hr, or 1 daylater. The 10-min delay was included to determine if facili-tated reading speed persisted beyond the range of immediatememory. The 2-hr delay was included so that the results forfacilitated reading speed could be compared to previous resultsobtained for word priming. Many stem-completion and word-association priming effects have been found to disappearwithin 2 hr in normal subjects and amnesic patients (Dia-mond & Rozin, 1984; Graf & Mandler, 1984; Graf, Squire,& Mandler, 1984; Shimamura & Squire, 1984; Squire, Shi-mamura, & Graf, 1987). Finally, a 1-day delay was includedto determine whether facilitated reading speed might be aslong-lived as other skills such as mirror reading (Cohen &Squire, 1980; Nichelli et al., 1989) or the ability to resolverandom-dot stereograms (Benzing & Squire, 1989).

Method

Subjects. The same 8 amnesic patients were tested as in Experi-ment 1. Eight healthy control subjects were also tested, 3 of whomhad participated in Experiment 1. All were employees or volunteersat the San Diego Veterans Affairs Medical Center. As a group, thecontrol subjects (5 men and 3 women) averaged 62 years of age, had13.2 years of education, and obtained WAIS-R subtest scores of 21.5for Information and 56.1 for Vocabulary. Immediate and delayed(12-min) recall of a short prose passage (Gilbert et al., 1968) averaged6.6 and 6.1, respectively.

Materials. Excerpts of four short stories from Reader's Digestand Esquire Magazine, each containing about 20 lines of text, wereselected for study material (211, 218, 230, or 233 words in length).The stores and the subject matter were different from the stories usedin Experiment 1.

Procedure. Subjects were asked to read a story three times insuccession following the same procedure that was used in Experiment

1. As in Experiment 1, subjects were told that they should read aloudas rapidly as possible but not so rapidly that they would not under-stand what they were reading. In this experiment, a fourth readingwas scheduled at one of four different delays after the third reading:immediately, 10 min, 2 hr, or 1 day. All subjects participated in allfour delay conditions, which were given on different days, separatedby at least 1 day. The stories and the delays were counterbalancedsuch that all of the stories were used in all four delay conditions andthe four delays occurred in a different order for each subject. Thus,on each of four different occasions, subjects read a story three timesin succession and then read the same story again after a delay (0 s,10 min, 2 hr, or 1 day later). After the final (delayed) reading of eachstory was completed, subjects were given a written test consisting of10, three-alternative, forced-choice questions about the content of thestory. To determine baseline guessing rates for the content of the fourstories, the multiple-choice tests were administered to 6 subjects whohad not read any of the stories. Their mean scores for the four storieswere close to the expected score of 33.3% for a three-alternative test(38.3%, 36.7%, 33.3%, and 40.0%).

Results

The measure of reading speed for the first three readingswas obtained by averaging the times for all four stories (seeFigure 2). For both groups, reading times decreased withrepeated presentations of the stories (from 79.9 s to 67.0 s forthe amnesic patients and from 78.9 s to 73.3 s for the controlsubjects, F{2, 28) = 24.2, p < .001. There was no effect ofgroup (F < 1.0), but amnesic patients improved at a morerapid rate than control subjects as reflected in the significantGroup x Repetition interaction, F(2, 28) = 3.71, p < .05.

The time required to complete the fourth reading, whichwas scheduled at a variable delay after the third reading,provided information about how long facilitated readingspeed persisted. The amnesic patients and control subjectsperformed similarly across these four delay intervals. Ananalysis of variance (Group x Delay) revealed an effect ofdelay interval, indicating that reading speed gradually slowedacross the four delays, F(3, 42) = 211.2, p < .001. There wasno effect of group (F < 1) and no interaction (F < 1). Thus,

80

11t -

•D

I

I 70

602 3 0 min 10 min 2 hr

Story ' Delay

1 day AMN CON

Figure 2. Time required to read aloud four different stories, each of which was presented three timesin succession and then a fourth time at the indicated delay. (AMN = amnesic patients, n = 8; CON =control subjects, n = 8. Across the seven data points, standard errors averaged 4.2 [range = 3.5-4.7] forcontrol subjects and 3.0 [range = 2.5-4.4] for amnesic patients. The bars show performance foreach group on a test of story content given immediately after the fourth reading of the story[chance = 33%]. Brackets show standard errors of the mean.)

SPECIFIC READING SKILL IN AMNESIA 1073

both amnesic patients and control subjects increased theirreading speed across three repetitions of the same text. Astime passed after the third repetition, reading speed graduallyslowed at the same rate in both groups.

To determine how long facilitated reading speed persistedafter the third reading, we compared the time required foreach fourth reading (at the 0-s, 10-min, 2-hr, or 1-day delay)to the time required for the first reading (averaged across allfour stories). In these comparisons, facilitated reading speedwould be demonstrated by finding a significantly shorterreading time for the fourth reading compared to the timerequired to read the story for the first time. Reading times forthe fourth reading were significantly faster in the 0-s delaycondition: amnesic patients, t(l) = 4.36, p < .01; controlsubjects, t{l) = 3.6, p < .01. They were also significantly fasterin the 10-min delay condition: amnesic patients, t(7) = 4.6, p< .01; healthy control subjects, t{l) = 2.54, p < .05. Afterdelays of 2 hr or 1 day, reading speeds had returned to aboutthe level that was observed during the first reading (for bothgroups, ps > .10). These comparisons indicate that facilitatedreading speed produced by reading a short passage three timesin succession persisted for at least 10 min and was notdetectable after 2 hr.

This conclusion is also supported by comparing the timesrequired for each fourth reading to the times required for thethird reading. At the 0-s and 10-min delays, during the timethat the facilitation persisted, the reading speeds for bothgroups were similar to the times recorded for the third reading(average difference = 0.1 s). In contrast, after the 2-hr and 1 -day delays, reading times were slower (though not alwayssignificantly so) than for the third reading of each story(average difference = 5.92 s): amnesic patients, 2-hr delay,t{l) = 2.06, p = .08,1 -day delay, /(7) = 6.41, p < .001; controlsubjects, 2-hr delay, t(7) = 3.37, p < .05, 1-day delay, t(l) =1.67, p= .14.

Despite the fact that amnesic patients and control subjectsbenefited similarly from their exposure to the text as measuredby reading speed, the amnesic patients performed more poorlythan the control subjects on the questions about story content(73.4% vs. 89.4%): ;(14) = 3.58, p < .01. Questionnaireperformance did not differ as a function of delay interval; atwo-way analysis of variance showed an effect of group, F( 1,14) = 11.9, p < .01, but not of delay (F < 1). This findingprobably reflects the fact that the questionnaire was alwaysgiven after the fourth reading. Thus, subjects always had theopportunity to read the story just prior to answering questionsabout it.

Discussion

This experiment demonstrated that facilitated readingspeed persisted at least 10 min and disappeared within 2 hr.Because 10 min is outside the range of immediate memory,the effect cannot be attributed to the fact that immediatememory is preserved in amnesia. The finding that facilitatedreading speed was lost at the same rate in both amnesicpatients and normal subjects provides strong evidence thatthe facilitation does not depend materially on declarative(explicit) memory.

General Discussion

The two experiments reported here show that, with repeatedreading of the same text, amnesic patients improve theirreading speed at least as much as normal subjects. The facil-itation was not due simply to an improved ability to read outloud but was specific to the text that was read. In both normalsubjects and amnesic patients, the effect persisted for at least10 min and then disappeared within 2 hr. Finally, for theamnesic patients, improved text reading occurred despite amarked impairment in the ability to recollect what was read.

In Experiment 2, facilitated reading speed developed morequickly in amnesic patients than in control subjects. It is notentirely clear how to interpret this finding because it was notobserved in Experiment 1. One possibility is that this effect,when it occurs, is due to the fact that all subjects are subse-quently asked questions about the stories. The control subjectsmay sometimes slow down in order to assure better compre-hension and retention of the stories, whereas the amnesicpatients may forget that memory for the stories will be tested.In any case, the effect was not a reliable one.

In their study of mirror-reading, Cohen and Squire (1980)found that although amnesic patients learned the generalpattern-analyzing skill of reading mirror-reversed word tri-plets, amnesic patients did not improve as much as normalsubjects at reading repeated triplets. One possibility is that themirror-reading of repeated words depends on declarative (ex-plicit) memory. In this view, the advantage that amnesicpatients were able to demonstrate for repeated words com-pared to unique words was due to the fact that the patientshad some residual declarative memory ability (see Shimamura& Squire, 1988, for an illustration of the explanatory useful-ness of residual memory abilities in amnesia). Another pos-sibility is that amnesic patients did acquire item-specific skillsfor reading the repeated words but could not obtain a normalperformance score because they were unable to use the dec-larative (explicit) memory strategies available to normal sub-jects (e.g., normal subjects could have recalled the second andthird words of each triplet after reading the first word). Thesecond possibility seems the likely one in light of the presentfindings of intact text-specific reading skills in amnesia. It isalso consistent with the results reported recently for a singleamnesic patient who showed good day-to-day retention of anitem-specific mirror-reading skill when the opportunity to usedeclarative memory was reduced (Nichelli et al., 1989).

The declarative-procedural distinction was originally ap-plied to amnesia as a way to understand what kinds ofmemory tasks were failed and what kinds were performedwell (Cohen, 1984; Cohen & Squire, 1980; Squire, 1982).Nichelli et al. (1989) suggested that the declarative-proceduraldistinction cannot accommodate the finding of intact item-specific effects in amnesia. However, such findings are fullyconsistent with the declarative-procedural distinction if skills(procedures) are recognized to depend importantly on item-specific information. Indeed, much of the recent experimentalliterature concerning the acquisition of skilled behavior innormal subjects emphasizes the importance of the specificencounters that cumulate in skilled performance (Jacoby,1983; Jacoby & Brooks, 1984; Kolers, 1976; Masson, 1986;

1074 G. MUSEN, A. SHIMAMURA, AND L. SQUIRE

Moscovitch et al., 1986). Amnesic patients can acquire manyskills normally (Brooks & Baddeley, 1976; Cohen & Squire,1980; Nissen & Bullemer, 1987; Squire & Frambach, 1990).What appears to limit their success is not whether the skillreflects multiple exposures to repeated (as opposed to unique)items but whether normal subjects can additionally benefitby engaging declarative memory strategies.

More recently, we used the term nondeclarative (Squire &Zola-Morgan, 1988) to refer to the kind of learning thatamnesic patients can accomplish, recognizing that what isspared includes not only the sort of skillful behavior that fitsthe term procedural, but also other kinds of implicit (uncon-scious) memory abilities such as priming and classical condi-tioning. These memory abilities are all experience-dependentand proceed without declarative knowledge of previouslyacquired facts or events.

Tardif and Craik (1989) found that facilitated reading speedpersisted for at least 1 week in their normal college-agesubjects, whereas our effect disappeared within 2 hr. Ourresults cannot be directly compared to theirs because theirsubjects always reread in transformed typography, whereasour subjects studied and reread in normal typography. Thesemethodological differences might be important in determin-ing the longevity of the effect. Alternatively, it is possible thatthe effect does not persist as long in older subjects (mean agein our study = 61 years) as in college students.

The text-specific reading skill demonstrated in the presentexperiment is also less long-lived than other perceptual skillssuch as mirror-reading (which was still evident after 3 months,Cohen & Squire, 1980) or the ability to perceive stereoscopicdepth (which was weakly evident after 7 days, Benzing &Squire, 1989). The longevity of the effect may depend on theamount of practice. In the present study, subjects read eachtext only three times. In contrast, the mirror-reading skill wasacquired during 150 trials given over a period of 3 consecutivedays, and the ability to perceive stereoscopic depth was ac-quired during 10 trials on 2 consecutive days. Thus, additionalpractice reading the same text might help to maintain aspecific reading skill.

The longevity we found for text-specific reading skill issimilar to the longevity found for word-completion primingand word-association priming, which have frequently beenfound to disappear within 2 hr (Diamond & Rozin, 1984;Graf & Mandler, 1984; Graf et al., 1984; Shimamura &Squire, 1984; Squire et al., 1987; for examples of primingeffects that appear to last longer, see Jacoby, 1983; Jacoby &Dallas, 1981; Light, Singh, & Capps, 1986; Tulving, Schacter,& Stark, 1982). Word-completion priming was assessed byaccuracy rather than speed, so comparisons between wordpriming and reading speed effects should be made cautiously.Nevertheless, the fact that text-specific reading skills and wordpriming have similar time courses raises the possibility thatpriming may be partly responsible for the effect. Thus, facili-tated reading of normal text may depend importantly on thepriming of words and associations in the text.

This idea is also consistent with previous findings showingfaster rereading of verbatim rather than paraphrased text andfaster reading when the first presentation is visual rather thanauditory (Kolers, 1975; Masson, 1984). The fact that it is

advantageous to see the specific words that will later be readsuggested to Kolers (1975) that subjects carry out detailedgraphemic analyses "of the words as patterns on the pagerather than of the words as meanings" (p. 296). Such a viewcould include two possible kinds of analyses: Words could betreated as lexical items or as neutral objects with particularsurface features, that is, as typographical objects consisting ofcharacteristic angles and curves.

Recent studies of visual processing of single words usingpositron emission tomography (PET) have identified a leftposterior word-form area, which is activated both by realwords and pronounceable pseudowords (Petersen, Fox, Sny-der, & Raichle, 1989). In contrast, presentation of letter strings(nonpronounceable pseudowords) or nonletter shapes acti-vates bilaterally a posterior, more lateral area. It is possiblethat facilitated rereading of same-word text involves nonlexi-cal systems of the kind activated by nonpronounceable letterstrings or nonletter shapes (e.g., to the extent that factors liketype font or case influence rereading, Jacoby & Hayman,1987; Masson, 1986; but see Carr et al., 1989). Alternatively,we propose that facilitated rereading depends importantly onperceptual priming within the word-form area, that is, onperceptual priming of lexical information. Perceptual priminghas been proposed previously to operate on a presemantic,word-form system (Schacter, in press) and to facilitate theidentification of previously encountered words. Note that ifperceptual priming is an important factor in facilitated re-reading, then text-specific effects are not due entirely tosemantic, conceptual factors as has sometimes been suggested(Carr et al., 1989; Tardif & Craik, 1989).

In addition to perceptual priming, semantic priming mayalso contribute to text rereading (for this distinction, seeGabrieli, in press; Schacter, in press; Tulving & Schacter,1990). Perceptual priming is modality bound; for example, itwas reduced in both normal subjects and amnesic patientswhen study and test phases were in different modalities (Graf.Shimamura, & Squire, 1985; Jacoby & Dallas, 1981). Seman-tic priming operates at the level of meaning and facilitatesassociations between presented words and semantically re-lated words (Shimamura & Squire, 1984: Schacter, 1985).Perceptual priming could explain why rereading is better whenthe same words are re-presented than when paraphrased textis presented and why visual-visual testing is more advanta-geous than auditory-visual testing. Semantic priming couldexplain why text rereading is facilitated by auditory presen-tation of the same text and by visually presented, paraphrasedtext (also see Levy & Kirsner. 1989, for a multiple-factor viewof text reading).

It is important to note that these considerations of thefactors underlying text-specific reading skills do not requireor assume that all of the conceptual and linguistic informationavailable in text is reflected in the reading skill. Word-associ-ation priming in amnesic patients occurs only for preexisting,unitized associations (Schacter, 1985; Shimamura & Squire,1984). Priming of new associations as measured by wordcompletion (Graf & Schacter, 1985) is weak or absent inamnesia, and an essential component of the phenomenonappears to depend on declarative memory (Cermak, Bleich,& Blackford, 1988; Mayes & Gooding, 1989; Shimamura &

SPECIFIC READING SKILL IN AMNESIA 1075

Squire, 1989). Finally, although the formation of new associ-ations between previously unrelated words has been reportedto facilitate rereading (Moscovitch et al., 1986), this effect hasbeen difficult to replicate (Musen & Squire, in press).

In summary, amnesic patients showed intact retention oftext-specific information as measured by reading speed. Thiseffect was relatively short-lasting—disappearing within 2 hrin both amnesic patients and normal subjects. The time courseof the effect is similar to the time course of word-completionand word-association priming effects. Previous research hassuggested that facilitated reading speed depends significantlyon semantic, conceptual information about the text (i.e., gistand lexical information). The possible contribution of non-lexical, visual-feature information has also been consideredpreviously. We have pointed out in addition the possibleimportance of perceptual priming within a presemantic, word-form system. Whatever the nature of the information respon-sible for text-specific effects (i.e., perceptual, semantic, orboth), this information can be supported by nondeclarative(implicit) memory.

References

Benzing, W. C, & Squire, L. R. (1989). Preserved learning andmemory in amnesia: Intact adaptation-level effects and learning ofstereoscopic depth. Behavioral Neuroscience, 103, 538-547.

Brooks, D. N., & Baddeley, A. (1976). What can amnesic patientslearn? Neuropsychologia, 14, 111-122.

Carr, T. H., Brown, J. S., & Charalambous, A. (1989). Repetitionand reading: Perceptual encoding mechanisms are very abstractbut not very interactive. Journal of Experimental Psychology:Learning, Memory, and Cognition, 15, 763-768.

Cermak, L. S., Bleich, R. P., & Blackford, S. P. (1988). Deficits inimplicit retention of new associates by alcoholic Korsakoff patients.Brain and Cognition, 1, 145-156.

Cohen, N. J. (1984). Preserved learning capacity in amnesia: Evidencefor multiple memory systems. In L. R. Squire & N. Butters (Eds.),Neuropsychology of memory (pp. 83-103). New York: GuilfordPress.

Cohen, N. J., & Squire, L. R. (1980). Preserved learning and retentionof pattern-analyzing skill in amnesia: Dissociation of knowing howand knowing that. Science, 210, 207-210.

Daum, I., Channon, S., & Canavar, A. (1989). Classical conditioningin patients with severe memory problems. Journal of Neurology,Neurosurgery, and Psychiatry, 52, 47.

Diamond, R., & Rozin, P. (1984). Activation of existing memoriesin anterograde amnesia. Journal of Abnormal Psychology, 93, 98-105.

Gabrieli, J. D. E. (in press). Differential effects of aging and age-related neurological diseases on memory subsystems of the brain.In F. Boiler & J. Grafman (Eds.), Handbook of neuropsychology.New York: Elsevier.

Gilbert, J., Levee, R., & Catalano, K. (1968). A preliminary reporton a new memory scale. Perceptual and Motor Skills, 27, 277-278.

Graf, P., & Levy, B. A. (1984). Reading and remembering: Concep-tual and perceptual processing involved in reading rotated passages.Journal of Verbal Learning and Verbal Behavior, 23, 405-424.

Graf, P., & Mandler, G. (1984). Activation makes words moreaccessible but not necessarily more retrievable. Journal of VerbalLearning and Verbal Behavior, 23, 553-568.

Graf, P., & Schacter, D. L. (1985). Implicit and explicit memory fornew associations in normal and amnesic subjects. Journal of Ex-

perimental Psychology: Learning, Memory, and Cognition, 11,501-518.

Graf, P., Shimamura, A. P., & Squire, L. R. (1985). Priming acrossmodalities and across category levels: Extending the domain ofpreserved function in amnesia. Journal of Experimental Psychol-ogy: Learning, Memory, and Cognition, 11, 385-395.

Graf, P., Squire, L. R., & Mandler, G. (1984). The information thatamnesic patients do not forget. Journal of Experimental Psychol-ogy: Learning, Memory, and Cognition, 10, 164-178.

Hintzman, D. L. (1990). Human learning and memory: Connectionsand dissociations. Annual Review of Psychology, 41, 109-139.

Horton, K. D. (1985). The role of semantic information in readingspatially transformed text. Cognitive Psychology, 17, 66-88.

Jacoby, L. L. (1983). Remembering the data: Analyzing interactiveprocesses in reading. Journal of Verbal Learning and Verbal Be-havior, 22, 485-508.

Jacoby, L. L., & Brooks, L. R. (1984). Nonanalytic cognition: Mem-ory, perception and concept learning. In G. L. Bower (Ed.), Thepsychology of learning and motivation: Advances in research andtheory (Vol. 18, pp. 1-47). San Diego, CA: Academic Press.

Jacoby, L. L., & Dallas, M. (1981). On the relationship betweenautobiographical memory and perceptual learning. Journal of Ex-perimental Psychology: General, 110, 306-340.

Jacoby, L. L., & Hayman, C. A. G. (1987). Specific visual transfer inword identification. Journal of Experimental Psychology: Learning,Memory, and Cognition, 13, 456-463.

Janowsky, J. S., Shimamura, A. P., Kritchevsky, M., & Squire, L. R.(1989). Cognitive impairment following frontal lobe damage andits relevance to human amnesia. Behavioral Neuroscience, 103,548-560.

Johnson, M. K., Kim, J. K., & Risse, G. (1985). Do alcoholicKorsakoff s syndrome patients acquire affective reactions? Journalof Experimental Psychology: Learning, Memory, and Cognition,11, 22-36.

Kolers, P. A. (1975). Specificity of operations in sentence recognition.Cognitive Psychology, 7, 289-366.

Kolers, P. A. (1976). Reading a year later. Journal of ExperimentalPsychology: Human Learning and Memory, 2, 554-565.

Kritchevsky, M., Squire, L. R., & Zouzounis, J. A. (1988). Transientglobal amnesia: Characterization of anterograde and retrogradeamnesia. Neurology, 38, 213-219.

Levy, B. A., & Burns, K. I. (in press). Reprocessing text: Contributionsfrom conceptually-driven processes. Canadian Journal of Psychol-ogy-

Levy, B. A., & Kirsner, K. (1989). Reprocessing text: Indirect meas-ures of word and message level processes. Journal of ExperimentalPsychology: Learning, Memory, and Cognition, 15, 407-417.

Levy, B. A., Newell, S., Snyder, J., & Timmins, K. (1986). Processingchanges across reading encounters. Journal of Experimental Psy-chology: Learning, Memory, and Cognition, 12, 467-478.

Light, L. L., Singh, A., & Capps, J. L. (1986). Dissociation of memoryand awareness in older adults. Journal of Clinical and ExperimentalPsychology, 8, 62-64.

Masson, M. E. J. (1984). Memory for the surface structure of sen-tences: Remembering with and without awareness. Journal of Ver-bal Learning and Verbal Behavior, 23, 579-592.

Masson, M. E. J. (1986). Identification of typographically transformedwords: Instance-based skill acquisition. Journal of ExperimentalPsychology: Learning, Memory, and Cognition, 12, 479-488.

Masson, M. E. J., & Sala, L. S. (1978). Interactive processes insentence comprehension and recognition. Cognitive Psychology,10, 244-270.

Mattis, S. (1976). Dementia Rating Scale. In R. Bellack & B. Keraso(Eds.), Geriatric psychiatry (pp. 77-121). New York: Grune &Stratton.

1076 G. MUSEN, A. SHIMAMURA, AND L. SQUIRE

Mayes, A. R., & Gooding, P. (1989). Enhancement of word comple-tion priming in amnesics by cueing with previously novel associates.Neuropsychologia, 27, 1027-1072.

Moscovitch, M., Winocur, G., & McLachlan, D. (1986). Memory asassessed by recognition and reading time in normal and memory-impaired people with Alzheimer's disease and other neurologicaldisorders. Journal of Experimental Psychology: General, 115, 331-347.

Musen, G., & Squire, L. R. (in press). Implicit memory: No evidencefor rapid acquisition of new associations in amnesic patients ornormal subjects. Society for Neuroscience Abstracts, 16.

Nichelli, P., Bahmanian-Behbahani, G., Gentilini, M., & Vecchi, A.(1989). Preserved memory abilities in thalamic amnesia. Brain,111, 1337-1353.

Nissen, M. J., & Bullemer, P. (1987). Attentional requirements oflearning: Evidence from performance measures. Cognitive Psy-chology, 19, 1-32.

Osterrieth, P. A. (1944). Le test de copie d'une figure complexe [Thetest of copying a complex figure]. Archives de Psychologie, 30, 206-356.

Peterson, S. E., Fox, P. T., Snyder, A. Z., & Raichle, M. E. (1990).Activation of extrastriate and frontal cortical areas by visual wordsand word-like stimuli. Science, 249, 1041-1044.

Press, G., Amaral, D. G., & Squire, L. R. (1989). Hippocampalabnormalities in amnesic patients revealed by high-resolution mag-netic resonance imaging. Nature, 341, 54-57.

Rey, A. (1964). L'examen clinique psychologie. Paris: Presses Uni-versitaires de France.

Richardson-Klavehn, A., & Bjork, R. A. (1988). Measures of memory.Annual Review of Psychology, 39, 475-543.

Schacter, D. L. (1985). Multiple forms of memory in humans andanimals. In N. M. Weinberger, J. M. McGaugh, & G. Lynch (Eds.),Memory systems of the brain: Animal and human cognitive proc-esses (pp. 351-379). New York: Guilford Press.

Schacter, D. L. (1987). Implicit memory: History and current status.Journal of Experimental Psychology: Learning, Memory, and Cog-nition, 13, 501-518.

Schacter, D. L. (in press). Perceptual representation systems andimplicit memory: Toward a resolution of the multiple memorysystems debate. In A. Diamond (Ed.), Development and neuralbases of higher cognitive function. Annals of the New York Academyof Science.

Shimamura, A. P. (1986). Priming effects in amnesia: Evidence for adissociable memory function. Quarterly Journal of ExperimentalPsychology, 38A, 619-644.

Shimamura, A. P. (1989). Disorders of memory: A cognitive scienceperspective. In F. Boiler & J. Grafman (Eds.), The handbook ofneuropsychology (Vol. 3, pp. 35-73). New York: Elsevier.

Shimamura, A. P., Jernigan, T. L., & Squire, L. R. (1988). Korsakoff ssyndrome: Radiological (CT) findings and neurological correlates.Journal of Neuroscience, 11, 4400-4410.

Shimamura, A. P., & Squire, L. R. (1984). Paired associate learning

and priming effects in amnesia: A neuropsychological study. Jour-nal of Experimental Psychology: General, 113, 556-570.

Shimamura, A. P., & Squire, L. R. (1988). Long-term memory inamnesia: Cued recall recognition and confidence ratings. Journalof Experimental Psychology, 14, 763-770.

Shimamura, A. P., & Squire, L. R. (1989). Impaired priming of newassociations in amnesia. Journal of Experimental Psychology:Learning, Memory, and Cognition, 15, 721-728.

Squire, L. R. (1982). The neuropsychology of human memory. An-nual Review of Neuroscience, 5, 241-273.

Squire, L. R. (1986). Mechanisms of memory. Science, 232, 1612-1619.

Squire, L. R. (1987). Memory and brain. New York: Oxford Univer-sity Press.

Squire, L. R., Amaral, D. G., & Press, G. A. (in press). Magneticresonance measurements of hippocampal formation and mammil-lary nuclei distinguish medial temporal lobe and diencephalicamnesia. Journal of Neuroscience.

Squire, L. R., Amaral, D. G., Zola-Morgan, S., Kritchevsky, M., &Press, G. (1989). Description of brain injury in the amnesic ofpatient N. A. based on magnetic resonance imaging. ExperimentalNeurology, 105, 23-25.

Squire, L. R., & Frambach, M. (1990). Cognitive skill learning inamnesia. Psychobiology, 18, 109-117.

Squire, L. R., & Shimamura, A. P. (1986). Characterizing amnesicpatients for neurobehavioral study. Behavioral Neuroscience, 100,866-877.

Squire, L. R., Shimamura, A. P., & Graf, P. (1987). Strength andduration of priming effects in normal subjects and amnesic patients.Neuropsychologia, 25, 195-210.

Squire, L. R., & Zola-Morgan, S. (1988). Memory: Brain systems andbehavior. Trends in Neuroscience, 11, 125-127.

Tardif, T., & Craik, F. I. M. (1989). Reading a week later: Perceptualand conceptual factors. Journal of Memory and Language, 28,107-125.

Teuber, H. L., Milner, B., & Vaughan, H. G. (1968). Persistentanterograde amnesia after stab wound of the basal brain. Neuro-psychologia, 6, 267-282.

Tulving, E. (1985). How many memory systems are there? AmericanPsychologist, 40, 385-398.

Tulving, E., & Schacter, D. L. (1990). Priming and human memorysystems. Science, 247, 301-306.

Tulving, E., Schacter, D. L., & Stark, H. (1982). Priming effects inword-fragment completion are independent of recognition mem-ory. Journal of Experimental Psychology: Learning, Memory, andCognition, 8, 336-342.

Warrington, E. K. (1984). Recognition Memory Test. Windsor,ONT:NFER-Nelson.

Received March 21, 1990Revision received May 14, 1990

Accepted May 24, 1990 •