kclpure.kcl.ac.uk  · Web viewIn Korsakoff patients, Pitel, Beaunieux, Guillery-Girard, Witkowski, de la Sayette, et al. (2009) argued for a 2-route theory of new semantic learning

10/12/14

Amnesia in an actor: learning and re-learning of play passages despite severe autobiographical amnesia

Michael D. Kopelman1 John Morton2

1King’s College London 2 University College London

(Institute of Psychiatry) (Institute of Cognitive Neuroscience)

Academic Neuropsychiatry 17-19 Queen Square

3rd Floor, Adamson Centre London

South Wing, St Thomas’s Hospital WC1 3AR

Westminster Bridge Road United Kingdom

London, SE1 7EH [email protected]

United Kingdom

[email protected]

(Corresponding author)

Abstract

We describe the case of an accomplished actor, whom we term AB, who suffered severe

amnesia following a cardiac arrest and hypoxic brain damage, affecting medial temporal and

thalamic structures. His performance on standard episodic memory tests, and on measures of

retrograde amnesia, including autobiographical memory, were severely impaired. When

presented with passages from plays he had not appeared in, AB showed a severe impairment

at the first learning trial, but thereafter showed a ‘normal’ learning curve for this semantically

and syntactically complex material. On being presented with passages from plays he had

performed in the past, AB did not show any recognition of them whatsoever, as one might

expect from his severe episodic memory impairment. However, AB showed a striking

benefit (savings score) in relearning passages he had previously performed, compared with

new passages, despite not having any autobiographical recall of having performed the

relearned passages before. Moreover, although his initial recall performance in learning these

passages was impaired compared with healthy control actors of similar age and experience,

AB demonstrated the same incremental learning rate on subsequent learning trials of the

passages as did the controls. [REWRITE: We conclude that the learning of such verbal

material is not easily categorised in the traditional dichotomy, relying on both semantic and

episodic components, and that, in the absence of any autobiographical recall, AB was

particularly dependent upon his intact semantic/procedural skills.]

Key words

Semantic learning, episodic memory, autobiographical amnesia, actor

2

1. INTRODUCTION

The notion that a (‘sparse’) hippocampal system encodes episodic memories, but that

slower learning (particularly perhaps of semantic information) can occur within a neocortical

system, was first postulated by Marr (1971). It was then incorporated into McClelland,

McNaughton, and O’Reilly’s (1995) connectionist model of complementary learning systems

in the hippocampus and neocortex (see also Kumaran & McClelland, 2012).

In recent years, there have been various attempts to show new semantic learning in

amnesic patients with extensive bilateral hippocampal or medial temporal damage. Tulving,

Hayman, and Macdonald (1991) demonstrated new semantic learning by patient KC of 3-

word sentences. Glisky and Delaney (1996) demonstrated learning of 5 fictitious facts about

famous people by the method of vanishing cues in 4 patients in post-traumatic amnesia.

Bayley and Squire (2002) showed that patient EP showed learning of novel 3-word sentences

using cued-recall and forced-choice recognition testing. Stark, Stark, and Gordon (2005), and

Stark, Gordon, and Stark (2008) demonstrated learning of new 3-word sentences in the

amnesic patient TE. More recently, Sharon, Moscovitch, and Gilboa (2011) demonstrated the

rapid acquisition (‘fast mapping’) of novel word-picture associations, when learned

incidentally and tested by forced-choice recognition.

Other investigations have demonstrated that new semantic knowledge has been

acquired by amnesic patients since the onset of their disorder. Verfaellie, Koseff, and

Alexander (2000) demonstrated familiarity with post-onset vocabulary and famous faces in

an hypoxic patient. Westmacott and Moscovitch (2001) demonstrated that patient KC had

acquired knowledge of famous names and vocabulary since the onset of his amnesia.

Similarly, O’Kane, Kensinger, & Corkin (2004) showed that HM had learned post-onset

famous names and semantic facts. McCarthy, Kopelman, and Warrington (2005)

3

demonstrated that RFR, who was severely amnesic following herpes encephalitis, had

acquired new vocabulary, and was able to make familiarity judgements about famous faces

and names of people who had become prominent since the onset of his disorder 19 years

earlier. Bayley, O’Reilly, and Squire (2008) also showed some (limited) acquisition of new

vocabulary, and knowledge of famous names, famous faces, and new objects around the

home in two amnesic patients. Likewise, Vargha-Khadem, Gadian, Watkins, Connelly, Van

Paesschen, W., et al. (1997) demonstrated that patients with developmental amnesia can show

a very good level of semantic knowledge, acquired in childhood and adolescence, despite

significant impairment in episodic learning. A more recent study of one of these patients

(Jon) showed that he can learn new facts, but that his inter-trial retention was impaired

relative to controls (Gardiner, Brandt, Baddeley, Vargha-Khadem, & Mishkin, 2008). In

Korsakoff patients, Pitel, Beaunieux, Guillery-Girard, Witkowski, de la Sayette, et al. (2009)

argued for a 2-route theory of new semantic learning (cf. Marr, 1971; McClelland et al.,

1995), and that the involvement of episodic memory in semantic learning differs both

according to the nature of the information to-be-learned and to the severity of the episodic

memory impairments (e.g. in Korsakoff vs non-Korsakoff alcoholics).

These various studies have been interpreted as demonstrating ‘slow’ semantic

learning by (anterior/inferior/lateral) temporal lobe neocortex in the absence of functioning

medial temporal lobe tissue, although an alternative hypothesis would be that the inefficient

learning is mediated by any residual medial temporal lobe tissue. However, the

investigations which have examined the learning process in such patients have usually

employed very simple tasks, e.g. the learning of 3-word sentences, and/or they have

demonstrated the acquisition of semantic knowledge in these patients by using familiarity

judgements, forced-choice recognition, or (occasionally) cued-recall, rather than by free

recall.

4

In the present investigation, we have examined a severely amnesic patient to

demonstrate the free recall of more complex, ‘real world’ material. We took advantage of our

patient’s profession as a celebrated actor to carry out this investigation. Moreover, when he

was asked to re-learn lines he had previously performed, he had first learned them during

time-periods for which he now had negligible autobiographical memory.

Ours aims were:

(i) To examine whether he would be able to learn ‘novel’ passages that he had not

performed before, and to compare his learning curves with those of

comparably skilled actors; and

(ii) To investigate his re-learning of ‘old’ passages he had performed in the past to

see if this re-learning facilitated his performance (relative to the learning of

novel passages). We also compared his rate of re-learning with that of

comparably skilled actors re-learning passages that they had previously

performed.

2. CASE DESCRIPTION

Patient AB was 67 at the time of testing. He was a professional actor of major

prominence. He had appeared in very many plays, films, and television programs. He was

perhaps best known for his English stage performances, which encompassed Shakespeare and

the ‘classics’, as well as modern theatre. He had received many outstanding notices

throughout his career, particularly in latter years.

In June 2005, AB was at a public meeting, when he suffered a cardiac arrest. He was

resuscitated by people present at the scene, and then by ambulance personnel. The medical

5

records recorded 5 minutes ‘down time’. He was treated in the Intensive Care unit of a

major acute hospital where 2 coronary stents were inserted, and a cardioverter defibrillator

implanted.

When he first recovered, AB manifested severe clinical amnesia. He was

disorientated in time and place, and frequently confabulated with a grandiose flavour. He

was alternately irritable and apathetic. At that stage, it was recorded in the medical records

that he had very little awareness of the severity of his amnesia. By August, he was

“beginning to recognise family members”. He was transferred to a major London

Rehabilitation Unit.

AB was first seen by one of the authors in November, 2005. He was still in the early

stage of his illness, and he was partially orientated. He was confabulating less floridly, but he

was still severely amnesic. He was completely unaware of his heart attack or cardiac arrest.

He reported his main problem to be “fatigue”, and said that his main goal was to return to

work. He thought that there were current wars in Palestine and the Balkans, but not Iraq.

Ten minutes after discussing a biography he had written, AB had no recall of this

conversation, nor that he had just been told by the clinician that he had read the book.

Examples of AB’s confabulations were that he should be performing in a theatre that

evening; or that he should be at a centre where he used to work; that he had to attend a Court

case, which had occurred 20 years earlier; that he was in Paris; and that his (deceased) mother

had accompanied him to another medical appointment. His wife reported that he would

become very agitated about these matters. Over the next 6 months, AB’s confabulations

subsided. AB was tested in August and September 2006, when he was still severely memory-

impaired, but no longer confabulating. Over the next 4 years, he showed a gradual

improvement; and, despite his handicap, he gave radio, film, and even a couple of stage

performances to great acclaim.

6

Figure 1(a) shows slices from AB’s MRI brain scan in June, 2005, before the cardiac

defibrillator was inserted. Figure 1(b) shows AB’s CT brain scan in November, 2006. It

shows widening of the fronto-temporal sulci and frontal horns of the lateral ventricles. The

coronal slice appears to show enlargement of the temporal horns, suggestive of medial

temporal lobe atrophy.

[FIGURE 1 APPROX HERE]

Figure 2 shows slices from AB’s 18 fluoro-deoxy-glucose PET scan in November,

2006, compared with findings from 20 healthy volunteers (Toosy, A.T., Burbridge, S.E.,

Pitkanen, M., Loyal, A.S., Akanuma, N., et al., 2008). It can be seen that AB showed

reduced metabolism (glucose uptake) in the thalamus bilaterally, and the medial and ventro-

medial frontal, and retrosplenial regions. Reduced thalamic and retrosplenial glucose uptake

in hypoxic brain damage following cardiac arrest has previously been reported by Reed,

Lasserson, Marsden, Lewis, Stanhope, et al. (1999) and by Markowitsch, Weber-

Luxenburger, Ewald, Kessler, and Heiss (1997).

[FIGURE 2 APPROX HERE]

3. METHODS

3.1 Background Neuropsychological Tests

AB was tested during August and September, 2006. He was administered the National

Adult Reading Test-Revised (Nelson & Willison, 1991) as a measure of premorbid IQ , and

the Wechsler Adult Intelligence Scale WAIS-III (Psychological Corporation, 2001) test to

estimate current IQ. The Doors and People battery (Baddeley, Emslie, & Nimmo-Smith,

1994), and a short version of the Recognition Memory Test (Warrington, 1984) were used to

7

evaluate current anterograde memory. Executive function was assessed on FAS verbal

fluency (Benton, 1968), and the Brixton and Hayling tests (Burgess & Shallice, 1997).

Remote memory for famous faces was tested on items from the 1970s to the 2000s (Johnston,

2006), and autobiographical memory on the two components of the Autobiographical

Memory Interview (Kopelman et al., 1990).

3.2 Experimental Tests

We obtained AB’s theatrical curriculum vitae from the 1960s until the time of his

illness in 2005. We selected 2 passages of prose or verse from each of 27 stage plays – 32

passages were from plays in which AB had previously performed, and 22 were from ‘control’

plays, in which he had not performed. 22 passages were from Shakespeare, and 32 were not

from Shakespeare. The mean length of these passages was 46.4 words (±6.2). The plays in

which he had performed covered the period from the 1960s (when he was in his 20s) to the

2000s (when he was in his 60s). The control plays were matched as closely as possible to the

plays in which he had performed. Thus, one Shakespeare tragedy was matched with another,

one Restoration play with another, Ibsen with Ibsen and modern plays were equally matched

by style. Passages were chosen so that there were no direct cues to their origin, such as

proper names or well-known quotations. Our success in making these selections was

evidenced by the fact that, on their passages (see below), only in 2 instances did control

subjects identify a passage that they had not performed as ‘familiar’; all the others they said

were ‘unfamiliar’.

The prose or verse passages were typed out on sheets without indicating where they

were from. An example passage was this from ‘A Woman of No Importance’:

8

One should never take sides in anything. Taking sides is the beginning of sincerity,

and earnestness follows shortly afterwards, and the human being becomes a bore.

However, the House of Commons really does very little harm. You can't make people

good by Act of Parliament—that is something.

The procedure was as follows: AB was shown a passage from a play, and then asked

to read it aloud. The sheet was then turned over. He was then asked a series of questions:

whether the passage was familiar, who was the author? The name of the play? The

character? He was asked whether he had performed this role. If he said he had performed

the role, he was asked a series of context questions: when he had performed it (1 point),

where he had performed it (2 points), who had directed him (1 point), any memories of the

Director’s style (2 points), and he was asked to name 3 other people in the cast (3 points)

[Maximum = 10].

After responding to these items, AB was asked to retrieve and describe in as much

detail as possible two autobiographical memories, one relating to an incident which occurred

during the performance of the play (prompt = to do with other players/the audience), and one

relating to another incident around that time (prompt = involving relationships, family or

work). If he did not initially respond, he was given the appropriate prompt. These were then

scored for descriptive richness and specificity in time and place. Autobiographical memory

was scored on a 0 to 3 scale (half-points allowed) per memory in terms of descriptive

richness and specificity in time and place, analogous to the scoring procedure for the

Autobiographical Memory Interview (Kopelman et al., 1990). As two autobiographical

memories were sampled, the maximum total score was 6. Secondly, autobiographical

memory responses were re-scored when only ‘pure episodic’ memories (scores of 3.0 or 2.5

per memory) were allowed; scores of 2 or less were all re-coded as zero (again two memories

were sampled: maximum score = 6).

9

After responding to all these questions, AB read the passage aloud once more, and

then the sheet was turned over. He was invited to recite the passage from memory as best he

could, and his response was recorded manually and on audio tape. He then read the passage

once more, after which he was again asked to recall it. He was given a maximum of 5

learning trials. Alternatively, the trial was ended after perfect recall. These responses were

later scored in terms of the number of words correct, and then converted to percent correct

scores.

After the learning trials were completed, AB was told what the name of the play was

(not having identified any of them correctly before). He was then again asked whether he

had performed it in the past. If the answer was now ‘yes’, the context and autobiographical

memory questions were asked again (‘post-prompting’).

AB was re-tested on a small subset of the passages at a week’s delay. This subset

consisted of 6 re-learned (performed) and 6 control (novel) passages. At this delay, AB was

given each passage again, and asked to read it aloud. He was then asked to recall it. There

was just one delayed recall trial for each passage.

In a separate session we tested AB’s recognition of what we thought were culturally

familiar quotations. In the design of the learning experiments, we had been at pains to avoid

passages which might have been familiar outside the context of a performance. Having

studied his performance on not very well-known passages, some from plays he had

performed in, we decided to go to the other extreme and we presented AB with famous single

lines, all from Shakespeare. The criterion used was that we would expect anyone with an

interest in English theatre to know not just the play, but the person who spoke the lines and to

be able to produce some continuation. There were 12 such lines, and four others where we

thought he might at least know the play. These quotations are listed in Table 3.

10

3.3 Control Participants

Five control participants were recruited (3 male, 2 female; mean age = 65.40 ± 3.65).

All were professional actors of equivalent eminence and stage experience. From their

theatrical CVs, a comparable selection of plays (performed between the 1960s and 2000s)

was selected, as well as control plays; and then comparable prose or verse passages were

identified. The mean length of their passages was 47.5 words (±5.4), which did not differ

significantly from AB’s passages (t=1.33, P=0.185). The procedure was the same as for AB.

3.4 Analysis

In comparing AB’s performance with the controls’ at baseline (trial 1), we analysed

the data, first, in terms of Crawford and Howell’s (1998) t-test for comparing an individual

against a small sample (patient vs. controls’ mean); and, secondly, with a one-way ANOVA

and Dunnett’s post-hoc test (AB vs each control). As we were testing the hypothesis that AB

performed significantly worse than controls, 1-tailed tests were used. The slopes of the

learning curves were examined, first, by use of Crawford, Garthwaite, and Howell’s (2003)

intra-individual measure of association: in essence, this allowed us to examine whether the

correlation between learning trial and performance observed for AB differed significantly

from the correlations observed in the controls. Secondly, we again used a one-way ANOVA

and Dunnett’s post-hoc tests (AB vs each control), this time to examine difference scores

between trial 1 (T1) and trial 5 (T5). AB never reached ceiling performance in learning new

lines; and, in relearning old lines, he did so infrequently (less than 10% of occasions at trial

11

3). The controls seldom reached ceiling until trial 4 or 5. Consequently, we also examined

difference scores between trial 1 (T1) and trial 3 (T3).

4. RESULTS

4.1 Background tests.

Table 1 shows AB’s scores on the background neuropsychological tests. AB was a

man of high premorbid IQ, but he was severely impaired on standard verbal and visual recall

and visual recognition memory tests (Doors and People battery; RMT). He also showed a

severe impairment in identifying famous faces, extending back to the 1970s. Consistent with

this, on the Autobiographical Memory Interview (AMI), he showed impairments in retrieving

personal semantic facts about his life and autobiographical incidents with a temporal gradient

which was steeper for facts than incidents.

[TABLE 1 APPROX. HERE]

4.2 Experimental tests.

4.2.1 Identifying the plays

Figure 3a shows the patient’s and the controls’ hit rates and false negatives for

identifying whether or not they had performed in the play from which the passage was taken.

Before being told which play each passage was from, AB said ‘No’ to all the passages,

obtaining zero hits and 100% false negatives (and also 100% true negatives for plays he had

not performed). By contrast, Controls obtained 88.6% hits, 11.4% false negatives, and 100%

true negatives. After being told the name of the play at the end of the session (Figure 3b),

12

AB obtained 80.0% hits correctly with 20.0% false negatives. Controls obtained 100% hits.

For plays he had not performed, AB gave 5.9% false positives and 94.1% true negatives.

Across individuals, these distributions differed significantly pre-prompting (Fisher Exact test

= 44.08, P<0.001), and post-prompting (Fisher’s Exact test = 10.44, P=0.006). In summary,

before prompting, AB did not identify any of the passages as familiar, and so concluded he

had not appeared in the plays from which the passages had been taken. The control

participants, on the other hand, were able to make discriminations as to whether a passage

was familiar or not, and whether or not they had played the part for the vast majority of play

passages. Moreover, after being given the name of the play the passage was from, the

controls were always correct about whether they had performed it or not.

[FIGURE 3a,b APPROX HERE]

4.2.2 Context Scores

Figure 4 shows context scores (for when and where the play had been performed, the

director, and other cast members, etc.) pre- and post- prompting with the name of the play

from which each passage had been taken (maximum score =10). Since AB recognised none

of the passages, he could not be given the context questions pre-prompting. After prompting

with the name of the play, his mean score (for passages he had performed) was 3.53 (±2.82).

Controls scored a mean of 7.33 (±3.42) pre-prompting, and 8.60 (±1.23) after prompting (for

passages they had performed). Post-prompting, AB differed significantly from the controls at

t(58) = 9.62, P<0.001.

[FIGURE 4 APPROX HERE]

4.2.3 Autobiographical Memory

13

Figure 5 (left panel) shows autobiographical memory scores and ‘pure episodic’

scores pre-prompting. AB was unable to give any autobiographical memories before being

told the name of the play. Mean scores for controls were 4.17 (±2.28) for total

autobiographical memory score and 3.75 (±2.38) for pure episodic memories. Figure 5 (right

panel) shows autobiographical memory scores after prompting with the name of the play.

AB still showed very few responses. His mean autobiographical memory score was 1.25

(±1.47) versus the controls’ mean of 4.94 (±1.42) (t(58)=8.83,P<.001). Moreover, most of his

memories came from the 1960s to 1980s, whereas the controls showed a relatively uniform

level of performance between the 1960s and 2000s. For pure episodic memories, AB’s mean

score was only 0.375 (±1.02) (he gave only 2 such autobiographical responses), and the

control mean (4.44 ±1.89) remained significantly higher than AB’s (t(58)=8.15,P<0.001). In

short, AB could not remember any autobiographical memories related to the time when he

had performed in these plays before being prompted. Moreover, he had very little

autobiographical memory, even after being prompted with the names of the plays. By

contrast, all five of the control subjects had a rich store of knowledge available concerning

the plays they had appeared in, not only about the context in which they performed the play

itself, but about their personal lives at the time of the play.

[FIGURE 5 APPROX HERE]

4.2.4 AB’s learning of ‘new’ and ‘old’ lines

Figure 6 (upper panel) shows AB’s performance in terms of re-learning lines from

plays he had performed in previously over 5 learning trials, compared with his performance

in learning new lines from plays in which he had not performed (per cent correct scores). It

also shows performance on a one-off recall trial a week later. At trial 1, there was a

significant difference in AB’s performance (t (52) = 3.7, P= 0.001). Overall, there was a

14

significant main effect of condition (F (1,52) = 13.28, P = 0.001) and a significant effect of

learning trial (F (4, 49) = 15.30, P <0.001). There was no significant condition by trial

interaction effect (F(4,49) = 0.43, P = 0.79). In brief, this means that AB showed significantly

better performance in the recall of lines that he had previously performed despite not

spontaneously recognising any of the lines or the plays that they had come from. However,

his learning rate after trial 1 was essentially the same for both relearned and new lines. We

conclude that, with the relearned lines, AB, without awareness, accessed a representation laid

down at the time he learned them. This gives the advantage of relearned over new lines. The

absence of any added advantage on the second and subsequent trials would be consistent with

his accessing the record of the previous trial, rather than going back to the original

representation again. The benefit from re-learning appears to have been constant across

trials.

The figure also shows that AB exhibited very good delayed retention when re-tested

at a week, scoring 53.2% for re-learned lines and 43.0% for new lines. These figures are

significantly better than trial 1 of the original learning. The re-test involved reading the

passage out loud followed by free recall, and we concluded he was able to access the record

of the previous week's learning trials.

Figure 6 (lower panel) shows AB’s performance (mean words learned after 5 trials) in

terms of relearning passages he had performed in across different decades with a regression

line superimposed; it shows that, in terms of re-learning these passages, there was the

suggestion of a (reverse) ‘temporal gradient’, whereby he performed better on passages he

had learned recently, compared with passages he had learned a long while ago, (i.e., the

reverse of the temporal gradient found on the AMI). However, there were no significant

differences detectable between any pair of decades with a Mann-Whitney test, due to high

variability in performance.

15

[FIGURE 6 APPROX HERE]

4.2.5 AB compared with Controls

Figure 7(a) & (b) show AB’s performance compared with the 5 individual control

participants. It will be seen that C3 started at a somewhat lower baseline performance than

the other controls in relearning lines (Figure 7b), but on trials 3 to 5 this actor’s performance

was as good as the others’. This actor also showed the fastest rate of learning new lines

(Figure 7a). The other actors’ learning curves were closely in parallel. Figure 7c shows AB’s

performance in learning new lines compared with the controls’ mean performance (expressed

as mean per cent correct scores). Figure 7d shows the same for re-learned lines.

[FIGURE 7 APPROX HERE]

For the new lines, AB differed significantly from controls at trial 1 (Crawford and

Howell t=-2.53, P<0.032, 1t; Anova F(5,77)=14.62, P<0.001, Dunnett post-hocs, P=0.027 to

P<0.001). Examining performance across trials (T1 to T5) using Crawford et al.’s (2003)

intra-individual measure of association, AB’s correlation between learning trial and

performance was not significantly different from the mean correlation of the controls (t= -

1.09, P(2t)=0.33), indicating that the slopes of the curves did not differ. We also looked at

this in terms of the difference scores between trial 1 (T1) and trial 5 (T5) using a one-way

ANOVA. On this, the participants differed significantly between T1 and T5, F(5,77) = 2.82,

P=0.02, largely due to control subject 3. On Dunnett post-hocs, only C3 differed

significantly (P=0.002) from AB. Other differences from AB were negligible. We also

carried out a one-way ANOVA on difference scores between T1 and T3, which did not give

any significant difference across the participants, F(5,77) = 0.94, P=0.46. Omitting C3, there

16

were no significant differences between the participants across all 5 trials (F(4,60)=0.80

P=0.53) or 3 trials (F(4,60)=0.19 P=0.94).

In comparing performance on re-learned lines at T1, AB differed significantly from

controls (Crawford and Howell t=-2.52, P=0.033; Anova F(5,98)=8.30,P<0.001, Dunnett

post-hocs, AB vs C3 N.S., AB vs other C’s, P=0.014 to <0.001 ). Examining performance

from T1 to T5 again using IIMA (Crawford et al., 2003), there was again no significant

difference between AB’s and the controls’ trial x performance correlations, t= -2.055, P(2t)

=0.109. Using a one-way ANOVA of T5-T1 difference scores, we found that the participants

differed significantly, F(5,98) = 4.087, P=0.002, but again only C3 differed from AB on

Dunnett post-hocs (P<0.001). Likewise, comparing T3-T1, there was a significant

difference, F(5,98)= F=5.001, P<0.001, but again only C3 differed significantly from AB

(P<0.001) on Dunnett post-hocs. Omitting C3, there were no significant differences between

the participants across all 5 trials (F(4,81)=0.73 P=0.57) or 3 trials (F(4,81)=1.003, P=0.41).

Fig 8a shows that AB performed significantly better at his Shakespearean passages

than his non-Shakespearean (F(1,52) =11.17, P=0.002), although, again, there was no

difference in the slopes of the learning curves (F(4,208) = 0.63, NS). By contrast, Fig 8b

shows that the controls performed almost identically on their Shakespearean and non-

Shakespearean passages. The extent of the advantage for Shakespearian passages was the

same for old and new passages as can be seen from Table 2, which gives the mean scores on

the initial trial. The Shakespearean passages were (nearly) all in verse, and this imposed a

structure for recall which AB, with his still extant analytic skills, appeared better able to

exploit. In addition, the difference between old lines and new was the same for the

Shakespearean and other passages. This indicates an additive effect in the factors

contributing to the setting up of the initial schema, i.e. the use of material in the ‘old’ trace

and the Shakespearean structure.

17

[FIGURE 8 APPROX HERE]

[TABLE 2 APPROX HERE]

We also examined AB’s learning of name-face associations (‘People’) across 3

learning trials and delayed recall on the Doors and People test. For comparison (Figure 9),

we used data derived from control participants (mean age = 67.9 ± 8.70, N=30) in a study by

Greene, Baddeley & Hodges (1996) (their Figure 5). These data suggested that AB started

from a low (zero) baseline, and then showed a normal learning rate (approximately in parallel

with controls), with good retention at delayed recall.

[FIG 9 APPROX HERE]

4.2.6 Familiar quotations

Table 3 shows AB’s performance on culturally familiar quotations. It shows that AB

identified the play correctly for 10 of the 16 lines, and he correctly identified the character for

8 of them. He produced the continuation on only 5 occasions. If he could not produce a

continuation spontaneously, he was given a cue consisting of the beginning of the next line.

Thus, with the Henry V line he was cued with "or close the wall up" and completed the line

with "with our English dead." The cue was effective in 5 of 8 trials. In most cases, AB also

volunteered information about whether he had or had not played the role. This included

commenting that he had played the role of Portia at his all-boys school. We tested one of the

control participants with the same lines and he correctly identified all 16 of the plays, but

provided the next line without error in only 3 cases. In other words, although AB had been

unable to identify, or know whether he had performed less familiar passages (see above), he

could recognise and place these culturally more familiar passages, such as "to be or not to

be".

18

[TABLE 3? APPROX HERE]

5. DISCUSSION

In learning new/novel play passages, AB’s performance at an initial learning trial was

well below the controls, as expected, but thereafter he appeared to show a normal learning

curve with good retention at one week. When given passages from plays he had performed,

AB did not identify them as ‘familiar’, recognise that he had performed them, or recall any

related autobiographical facts or incidents. After prompting with the play titles, he could

identify whether he had performed the play in 80.0% of cases, with 20.0% false negatives and

5.9% false positives, but he still had very poor recall of autobiographical contextual facts

(when, where, and with whom) about plays he had performed, and virtually no recall of

autobiographical incidents from that time. All his scores were well below those of healthy

actors for equivalent plays that they had performed in their lifetimes. Despite this, AB

showed a clear ‘sparing’ effect in re-learning lines from plays he had performed in, compared

with learning new lines. In other words, there was an ‘implicit’ saving effect which was

present on the first learning trial and an equivalent ‘sparing’ effect thereafter, i.e. without any

significant difference in the slopes of his learning curves between the two kinds of passage.

New learning of passages

For the new passages there was a statistically significant advantage at the initial

learning trial for the controls over AB. Since all the passages were unfamiliar, this difference

must have been due to cognitive factors. It is likely that both immediate verbal memory and

the efficiency of executive functions played a role and, in these two respects, AB scored very

poorly on the background tests. AB showed a further benefit at the initial trial for passages

from Shakespeare, compared with other passages, an advantage not shown by the control

19

subjects. We conclude that over the years, AB, who was a noted Shakespearean actor, had

built up schema (Ghosh & Gilboa, 2014) for reading, interpreting and remembering

Shakespearean verse, and that he was able to use these schema on the initial trial.

Although AB’s baseline performance was significantly below that of the control

actors, and his scores remained below theirs, the slopes of his learning curves were

essentially parallel with those of 4 out of 5 of the controls, and with the controls’ mean curve.

In other words, AB appeared unimpaired in his rate of incremental learning of these novel

passages beyond the first recall trial. This learning cannot have relied upon autobiographical

memory; and what is striking is that the material learned was lengthier, and much more

complex (semantically and syntactically), than that described in previous reports of new

semantic learning in amnesia (e.g. Tulving et al., 1991; Bayley & Squire, 2002; Stark et al.,

2008).

Re-learning of play passages

In terms of the re-learned (‘old’) passages, we found that both AB and the controls

benefitted greatly, relative to ‘new’ learning. If anything, the advantage for AB was greater

(20.2% vs 11.8% at the first learning trial). This benefit at re-learning must arise from the

utilisation during the first learning trial of information laid down during the original rehearsal

and performance of the speeches. However, AB did not recognise any of the ‘old’ passages,

and could not have been using (explicit) autobiographical information in retrieval. This

indicates that these memory traces must be content-addressable (Morton, Hammersley, &

Bekerian, 1985); and the facilitation on re-learning the passages appears to be effective

whether it is entirely implicit, as in AB, or (at least partially) explicit, as in the controls. In

20

AB, we also found that this old vs. new advantage was additive to his Shakespeare vs. Other

benefit.

Moreover, there was no significant difference between the slopes of AB’s learning

curves for re-learned (‘old’) lines and his curve for new lines. Moreover, his mean learning

curves for relearning ‘old’ lines was parallel to that of controls. It appears that, on the first

learning trial, AB accessed the (implicit) record of the original learning, but that thereafter he

accessed the record of the previous learning trial(s). For example, on the third learning trial,

he accessed what had been laid down in the course of the second trial, and built upon that as

he read the passage again. There was also an advantage for old vs new passages in the

controls; this benefit also accrued at the first learning trial and was then maintained across the

subsequent trials. We conclude that the use of the initial learning trace in augmenting the

subsequent learning of text had not been affected in AB.

In this learning experiment, then, the major difference between AB and the controls

arose almost entirely on the very first learning trial, whether the passage was old or new.

Thereafter, we did not find any differences between AB and the controls in the slopes of the

learning curves. This pattern appears to be broadly consistent with AB’s performance in

learning (novel) face-name associations (Figure 9). Relative to age-matched controls, AB’s

performance on the first recall trial was well below that of the controls, but his learning rate

across 3 trials appeared to be approximately parallel to that of the controls with good

retention at delayed recall.

The Nature of Learned Text

21

The material CR acquired was much lengthier, and semantically and syntactically

complex, than that acquired and retained in previous studies of ‘new semantic learning’

(REFS). It was also different from that rapidly acquired in Sharon et al.’s (2011) study of

‘fast mapping’: although these authors showed rapid and efficient non-hippocampal

(semantic) learning in amnesia, this was demonstrated on (forced-choice) recognition testing,

but not on recall memory.

Let us now concentrate, for the moment, on the material in memory of actors some

time after they have finished appearing in a play. When the passages from plays they had

performed were presented to controls, they usually triggered the name of the play and

whether they had performed in it (88.4% ‘hits’). This reminding could have been automatic

or deductive. For example, the passage could have led them directly to the play name, or

alternatively it could have led to the character, and thence to the name of the play. The play

names, in virtually all cases, led to the control participants being able to produce relevant

contextual and rich autobiographical information (see Figs 4 and 5). Hence, it appears that

the learning of a speech by an actor, in the course of rehearsing a play, is likely to involve not

only the extraction of meaning (semantic learning), but also the anchoring of such learning

within an autobiographical context. To that extent, this learning might be considered to

trigger both semantic and episodic components.

Greenberg and Verfaellie (2010) have recently reviewed how past past

autobiographical knowledge can facilitate semantic retrieval (Westmacott, Black, Freedman,

& Moscovitch, 2004; Renoult et al, 2012), just as semantic learning can facilitate new

episodic learning in amnesia (e.g. Kan, Alexander, & Verfaellie, 2008). Greenberg and

Verfaellie argued for the inter-dependence of these conceptually distinct systems, and a

similar interaction may occur in healthy actors’ relearning of ‘old’ lines. More recently,

Verfaellie, Bousquet, and Keane (2014) have argued for a contribution of hippocampal and

22

(antero-lateral) temporal neocortical contributions in recalling detailed semantic narratives

acquired in early life, with the hippocampal contribution critical for recalling the detail of

these narratives. Others have pointed to interactions between medial frontal and hippocampal

structures in memory and concept formation (Kumaran, Summerfield, Hassabis, & Maguire,

2009; van Kesteren, Ruiter, Fernández, & Henson, 2012).

Moreover, it does not seem controversial to consider very famous lines (e.g., “to be or

not to be …”) as being more purely semantic. They are more complex than the three-word

sentences learned by the densely amnesic patients KC and EP (Tulving et al., 1991; Bayley &

Squire, 2002), but it seems reasonable to suppose that these lines and their source have been

overlearned and decontextualized well beyond the point of semanticisation. Thus, it might

not be surprising that AB, despite his clear autobiographical deficit, was able to identify these

passages, give the name of the play they came from, and to continue a number of them.

What are we to make of the less familiar but relearned (‘old’) passages, which AB had

committed to memory in the past and had produced on a number of occasions on stage?

These passages did not lead him to the appropriate play, nor were they familiar to him. His

contextual knowledge and autobiographical memories for facts and events at the time he

performed these plays remained very poor (unlike the controls), even after prompting with

the name of the play. Although AB had originally learned the ‘old’ passages at a specific time

and occasion in his life, these were unavailable to him explicitly, when he relearned the lines.

But, he must have been able to access and utilise this material, at least implicitly, when asked

to re-learn these lines.

A potential explanation of these findings is that we are seeing slow residual learning

by a cortical route. However, AB’s learning was not ‘slow’ insofar that his learning rate

(after the first trial) was the same as that of the controls. His impairment lay at the first

23

learning trial, but learning thereafter was a matter of building upon what had already been

stored, at which he appears to have been unimpaired. Moreover, he showed good retention of

his learning at a week’s delay.

Whilst previous investigations have demonstrated new semantic learning of relatively

simple material in amnesia, that used in the present investigation was longer and both

syntactically and semantically more complex. What struck both the authors was the highly

skilled manner in which AB could immediately read, parse, and extract the meaning from

even the most complex novel passages, presumably using his long-established actorly

procedural and semantic skills. His superior performance in recalling the Shakespearean

passages could be viewed as consistent with this. AB appeared to show a normal learning

rate from a low baseline in learning name-face associations from the Doors and People

battery. The fact that he also showed preserved learning rates on this complex verbal

material in the absence of any autobiographical or contextual recall may have relied in part

on his ability to ‘chunk’ the passages by means of his exceptional semantic/procedural skills

in verse and prose reading. In addition, there was a form of ‘implicit’ knowledge, originally

embedded in an autobiographical context (no longer available to him), which facilitated his

performance at the first learning trial of previously performed passages. Unfortunately, we

did not have the opportunity to test his implicit knowledge of other remote episodic

information. However, as mentioned above, in subsequent years, AB was able to use these

skills, somewhat remarkably, in a partial resurrection of his career.

Acknowledgements

We are very grateful to the patient and his wife for their enthusiastic participation in

this study. Sadly, he has now died. We are also very grateful to the other eminent actors,

24

who also participated in this study as controls. We are grateful too to Ms Maria Borrelli who

prepared some of the figures and to Ms Asako Yokoya for preparing the manuscript.

INSERT RE PERMISSION TO PUBLISH

25

References

Baddeley, A.D., Emslie, H. & Nimmo-Smith, I. (1994). Doors and People: A Test of Visual

and Verbal Recall and Recognition. Bury St Edmunds: Thames Valley Test Company.

Bayley, P.J., O’Reilly, R.C., Curran, T., & Squire, L.R. (2008). New semantic learning in

patients with large medial temporal lobe lesions. Hippocampus, 18, 575-583.

Bayley, P.J., & Squire, L.R. (2002). Medial temporal lobe amnesia: gradual acquisition of

factual information by nondeclarative memory. J Neurosci, 22, 5741-5748.

Benton, A.L. (1968). Differential behavioral effects in frontal lobe disease.

Neuropsychologia, 6, 53-60.

Burgess, P.W., & Shallice, T. (1997). The Hayling and Brixton Tests. Bury St. Edmunds,

UK: Thames Valley Test Company.

Crawford, J.R., Garthwaite, P., Howell, P.H., & Venneri, A. (2003). Intra-individual

measures of association in neuropsychology: inferential methods for comparing a single-case

with a control or normative sample. J Int Neuropsychol Soc, 9, 989-1000.

Crawford, J.R., & Howell, D.C. (1998). Comparing an individual’s test score against norms

derived from small samples. Clin Neuropsychol, 12, 482-486.

Gardiner, J.M., Brandt, K.R., Baddeley, A.D., Vargha-Khadem, F. & Mishkin, M. (2008).

Charting the acquisition of semantic knowledge in a case of developmental amnesia.

Neuropsychologia, 46, 2865-2868.

Ghosh, V.E., & Gilboa, A. (2014) What is a memory schema? A historical perspective on

current neuroscience literature. Neuropsychologia, 53, 104-114.

26

Glisky, E.L., & Delaney, S.M. (1996). Implicit memory and new semantic learning in

posttraumatic amnesia. J Head Trauma Rehabil, 11, 31-42.

Greenberg, D.L., & Verfaellie, M. (2010). Interdependence of episodic and semantic

memory: evidence from neuropsychology. J Int Neuropsychol Soc, 16, 748-753.

Greene, J.D.W., Baddeley, A.D., & Hodges, J.R. (1996) Analysis of the episodic memory

deficit in early Alzheimer’s disease: evidence from the doors and people test.

Neuropsychologia, 34, 537-551.

Johnston, K. (2006). D.Clin. Psych. Thesis: Neuropsychiatric factors in retrograde amnesia.

University of London.

Kan, I., Alexander, M.P., & Verfaellie, M. (2008). Contribution of prior semantic knowledge

to new episodic learning in amnesia. J Cogn Neurosci, 21, 938-944.

Kopelman, M.D., Wilson, B.A., & Baddeley, A.D. (1990). The Autobiographical Memory

Interview. Bury St Edmunds: Thames Valley Test Company.

Kumaran, D., & McClelland, J.L. (2012). Generalization through the recurrent interaction of

episodic memories: a model of the hippocampal system. Psychol Rev, 119, 573-616.

Kumaran, D., Summerfield, J.J., Hassabis, D., & Maguire, E.A. (2009). Tracking the

emergence of conceptual knowledge during human decision making. Neuron, 63, 889-901.

Markowitsch, H.J., Weber-Luxenburger, G., Ewald, K., Kessler, J., & Heiss, E.-D. (1997).

Patients with heart attacks are not valid models for medial temporal lobe amnesia. A

neuropsychological and FDG-PET study with consequences for memory research. Eur J

Neurol, 4, 178-184.

27

Marr, D. (1971). Simple memory: a theory for archicortex. Philos Trans Royal Soc B Biol

Sci, 262, 23-81.

McCarthy, R.A., Kopelman, M.D., & Warrington, E.K. (2005). Remembering and forgetting

of semantic knowledge in amnesia: a sixteen-year follow-up investigation of RFR.

Neuropsychologia, 43, 356-372

McClelland, J.L., McNaughton, B.L., & O’Reilly, R.C. (1995). Why there are

complementary learning systems in the hippocampus and neocortex: insights from the

successes and failures of connectionist models of learning and memory. Psychol Rev, 102,

419-457.

Morton, J., Hammersley, R.H., & Bekerian, D.A. (1985) Headed records: a model for

memory and its failures. Cognition, 20, 1-23.

Nelson, H.E., & Willison, J.R. (1991). National Adult Reading Test (NART): Test Manual.

Windsor: NFER-Nelson.

O’Kane, G., Kensinger, E.A., & Corkin, S. (2004). Evidence for semantic learning in

profound amnesia: an investigation with patient H.M. Hippocampus, 14, 417-425.

Pitel, A.L., Beaunieux, H., Guillery-Girard, B., Witkowski, T., de la Sayette, V., Viader, F.,

Desgranges, B., & Eustache, F. (2009) How do Korsakoff patients learn new concepts?

Neuropsychologia, 47, 879-886.

Reed, L.J., Lasserson, D., Marsden, P., Lewis, P., Stanhope, N. Guinan, E. and Kopelman,

M.D. (1999). FDG-PET analysis and findings in amnesia resulting from hypoxia. Memory, 7,

599-612.

28

Renoult, L., Davidson, P.S.R., Palombo, D.J., Moscovitch, M., & Levine, B. (2012). Personal

semantics: at the crossroads of semantic and episodic memory. Trends Cogn Sci, 16, 550-

558.

Sharon, T., Moscovitch, M., & Gilboa, A. (2011). Rapid neocortical acquisition of long-term

arbitrary associations independent of the hippocampus. P Natl Acad Sciences USA, 108,

1146-1151.

Stark, C.E.L., Stark, S.M., & Gordon, B. (2005). New semantic learning and generalization in

an amnesic patient. Neuropsychology, 19, 139-151.

Stark, S., Gordon, B., & Stark, C. (2008). A case study of amnesia: exploring a paradigm for

new semantic learning and generalization. Brain Injury, 22, 283-292.

Toosy, A.T., Burbridge, S.E., Pitkanen, M., Loyal, A.S., Akanuma, N., Laing, H., Kopelman,

M.D., & Andrews, T.C. (2008). Functional imaging correlates of fronto-temporal dysfunction

in Morvan’s syndrome. J Neurol Neurosurg Psychiatry, 79, 734-735.

Tulving, E., Hayman, C.A., & Macdonald, C.A. (1991). Long-lasting perceptual priming and

semantic learning in amnesia: a case experiment. J Exp Psychol Learn Mem Cogn, 17, 595-

617.

van Kesteren, M.T.R., Ruiter, D.J., Fernández, G., & Henson, R.N. (2012). How schema and

novelty augment memory formation. Trends Neurosci, 35, 211-219.

Vargha-Khadem, F., Gadian, D.G., Watkins, K.E., Connelly, A., Van Paesschen, W.,

Mishkin, M. (1997). Differential effects of early hippocampal pathology on episodic and

semantic memory. Science, 277, 376-380.

29

Verfaellie, M., Bousquet, K., & Keane, M.M. (2014). Medial temporal and neocortical

contributions to remote memory for semantic narratives: evidence from amnesia.

Neuropsychologia, 61, 105-112.

Verfaellie, M., Koseff, P., & Alexander, M. P. (2000). Acquisition of novel semantic

information in amnesia: effects of lesion location. Neuropsychologia, 38, 484-492.

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

Westamacott , R., & Moscovitch, M. (2001). Names and words without meaning: incidental

postmorbid semantic learning in a person with extensive bilateral medial temporal damage.

Neuropsychology, 15, 586-596.

Westmacott, R., Black, S.E., Freedman, M., & Moscovitch, M. (2004). The contribution of

autobiographical significance to semantic memory: evidence from Alzheimer’s disease,

semantic dementia, and amnesia. Neuropsychologia, 42, 25-48.

30