Neuropsychologia 40 (2002) 1930–1947

Mechanisms underlying perseveration in aphasia:

evidence from a single case study
Stephen J. Gotts a,∗ , Antonio Incisa della Rocchetta b , Lisa Cipolotti b,1
a Department of Psychology—Baker Hall, Carnegie Mellon University & Center for the Neural Basis of Cognition, Pittsburgh, PA 15213, USA
b Neuropsychology Department, The National Hospital for Neurology and Neurosurgery, Queen Square, London WC1N 3BG, UK

Received 18 June 2001; received in revised form 30 April 2002; accepted 30 April 2002

Abstract
Aphasic individuals often inappropriately and unintentionally repeat recent responses, errors termed recurrent perseverations. In a series
of picture naming experiments, we investigated the impact of manipulating stimulus factors on the number of perseverations produced
by an aphasic patient (E.B.) with markedly impaired naming skills. E.B. produced significantly more perseverations to pictures with low
frequency names and following stimulus repetition. In contrast, semantic relatedness and presentation rate failed to influence perseveration.
Our results are considered in the context of theories that relate recurrent perseverations to intact priming mechanisms [Brain 121 (1998)
1641; Aphasiology 12 (1998) 319; J. Exp. Psychol. Learn. Mem. Cogn. 19 (1993) 243]. We conclude that these theories can correctly
predict some but not all aspects of E.B.’s perseverations. In particular, they failed to predict that: (1) perseverations often appeared to reflect
the earlier sequential proximity of stimuli and responses; and (2) perseverations became less likely as more experimental trials intervened,
a trend that did not interact with presentation rate. We review evidence relating recurrent perseverations to neuromodulatory deficits and
we conclude that a theory of the functional role of neuromodulation in the cerebral cortex proposed by Hasselmo [Neural Netw. 7 (1994)
13] is capable of accounting for all aspects of E.B.’s perseverative behavior.
© 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Acetylcholine; Dysphasia; Naming; Neuromodulation; Perseverations; Priming

1. Introduction cessing are not completely clear, their existence seems to

Aphasic patients produce a wide range of errors in lin-
guistic tasks. These errors have been of interest to language
researchers because they provide a relatively unique win-
dow into the contents of information processing and serve
as detailed constraints on theories of the normal language
system [7,13,33,35,67]. One type of aphasic error that is
particularly interesting in this regard is the perseveration
[1,12,37]. Perseveration refers to the inappropriate repeti-
tion or continuation of a previous response when a different
response is expected [31]. For example, in a picture naming
task an aphasic patient who correctly provided the response
“dog” to a picture of a dog, might provide the same re-
sponse several trials later to the picture of a horse. While
the implications of these errors for theories of language pro-
∗ Corresponding author. Tel.: +1-412-521-6176; fax: +1-412-268-2798.
E-mail addresses: gotts@cnbc.cmu.edu (S.J. Gotts),
l.cipolotti@ion.ucl.ac.uk (L. Cipolotti).
1 Co-corresponding author. Tel.: +44-171-837-3611x3295;
fax: +44-171-813-2516.
reflect the persistence of recently processed information and
the failure of current stimulus processing to override this
persistence. To the extent that persistence and competition
are intrinsic features of language processing, perseverations
may provide important insights into the functioning of the
normal language system and how it breaks down in aphasia.
Studies of perseveration in aphasia have yielded a number
of general characteristics. Perhaps the most striking of these
is that a previous response may be provided again after a
number of intervening stimuli or responses [2,12,30,37,64].
The often delayed nature of these perseverations has led
some researchers to refer to them as recurrent, distinguishing
them from types that appear to be an extension or continu-
ation of the immediately preceding response [62,63]. While
it can be difficult to distinguish true recurrent perseverations
from responses repeated by chance due to an impoverished
vocabulary, rigorous methods designed to assess the inci-
dence and time course of perseverations have shown that they
can occur after delays of up to 10–15 trials [12]. However,
they do appear to decrease in likelihood as the number of
intervening trials increases and are most common after little

0028-3932/02/$ – see front matter © 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 2 8 - 3 9 3 2 ( 0 2 ) 0 0 0 6 7 - 2
 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947
or no delay [12]. Recurrent perseverations may be on whole
words, part words or even parts of drawings, sometimes
occurring as a blend of a previous response and the current
target [1,6,12,53,64,65]. Individual aphasic patients may
perseverate on more than one task, although some patients
appear to perseverate only on certain tasks [12,14,53]. For
these patients, tasks that do not elicit perseverations tend to
be those performed at or near ceiling [12,53].
Recent theories developed from psychological and neu-
ropsychological perspectives have attempted to account for
the existence of recurrent perseverations by appealing to
the same mechanisms proposed to underlie priming effects
in normal subjects ([12,37,73]; see also [56]). One such
theory, proposed by Cohen and Dehaene [12], holds that
cognitive representations at many processing levels com-
monly persist for some period of time following initial
activation. Subsequent processing may be facilitated if an
identical or similar stimulus is presented while activity still
persists, a phenomenon referred to as priming. If a different
stimulus is presented, the corresponding input overrides
persistence through a competitive process. When brain
damage degrades or “deafferents” input to a certain level
of processing, representations persisting in activity are not
always effectively overridden, resulting in perseverations.
A related view has emerged from research aimed at iden-
tifying the neurophysiological basis of perseverations. Sand-
son and Albert [63] proposed that recurrent perseverations
result from low levels of acetylcholine. Indeed, recurrent
perseverations have been shown to correlate with choliner-
gic deficits and drugs that mimic acetylcholine can reduce
perseverations [20,54,70]. Many studies have suggested that
the functional role of acetylcholine in the central nervous
system is to modulate the dynamics of cortical processing
and learning, making cells more sensitive to feedforward,
bottom-up sensory inputs (see [26] for a review). Under
a cholinergic deficit, processing becomes “sluggish”, less
dominated by feedforward input and more dominated by
intracortical inputs. Perseverations can then result from the
reduced ability of feedforward input to override residual ac-
tivity. They might also result from heightened intracortical
plasticity and the formation of associations between jointly
activated cortical representations. Later activation of one of
these representations could then inappropriately re-activate
an “associated” representation (see [25] for discussion).
Despite these theoretical developments, relatively few
studies have attempted to assess the impact of stimulus
factors on recurrent perseverations. Halpern [24] studied
the influence of part of speech, level of abstraction, lexical
frequency and word length on the number of persevera-
tions produced by a mixed group of aphasics performing
word-repetition and word-reading tasks. Only word length
was found to have a significant effect with more persever-
ations produced to longer target words. Santo Pietro and
Rigrodsky [64] found that in sentence completion, picture
naming and word-reading tasks, aphasics produced more
perseverations to stimuli of low lexical frequency and when
1931
a fast presentation rate was used compared to a slower rate
(response–stimulus interval (RSI) of 1 s versus 10 s). A few
researchers have suggested that perseverations often share
a phonological or semantic relationship with the stimuli
that elicit them [1,37,65]. However, an obvious relationship
does not always exist. In the only study to date that has
carefully controlled for chance relationships between tar-
gets and perseverations, Hirsh [30] demonstrated that most
perseverations produced by an aphasic patient (C.J.) with
severe naming difficulties were unrelated either semanti-
cally or phonologically to their corresponding target names.
Nevertheless, it was of interest that when all of the target
stimuli that led to the same perseverated responses were
considered, these targets were likely to be semantically
and phonologically related. It should be noted that most
previous studies have not controlled for the chance repeti-
tion of responses, as appropriate methods have only been
developed recently [12].
The current investigation is aimed at understanding the
influence of stimulus factors on aphasic perseveration and
the implications that these findings have for existing theo-
ries. We present the case of an aphasic patient (E.B.) who
was markedly impaired on naming tasks and exhibited a
large number of verbal recurrent perseverations. In three
experiments, we manipulated stimulus factors that have
been shown previously to influence lexical priming effects
in normal subjects, as well as aphasic performance, and as-
sessed their impact on E.B.’s perseverations. This allowed
us to examine the potential relationship between recurrent
perseverations and priming effects and afforded comparison
with other patterns of aphasic performance.
2. Case report
E.B. is a 71-year-old, right-handed homemaker. She was
admitted at the National Hospital in May 1998, with a recent
history of epileptic fits. A malignant left fronto-temporal
meningioma was diagnosed and subsequently excised in
the first week of June. Post-operatively she developed lan-
guage difficulties that are described below. A follow-up
post-operative MRI scan carried out at the time of the ex-
perimental investigation (last week of June 1998) showed
the extent of cortical lesion (see Fig. 1).
The Brodmann’s areas (BAs) unequivocally involved
were: BAs 44, 45, 46, 10, 22 and 6. BAs 20, 38, 39 and 37
were spared. BA 47 also appeared spared.
3. Neuropsychological assessment
E.B. was referred to the Neuropsychology Department
for assessment of her language difficulties. She was as-
sessed on a shortened verbal scale of the WAIS-R; her
verbal IQ score, prorated from four subtests, was in the
defective range (see Table 1).
1932 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947

3.1. Language tasks

E.B.’s spontaneous speech was sparse, marred by fre-

quent anomia, recurrent perseverations and neologisms. The
informational content was rather limited. For example, in
describing the “cookie jar” scene from the BDAE [22], she
said: “this is the cookie jar . . . he’s getting up to cook
(actually a boy on a stool stealing the cookies and about
to fall off) . . . the cookie jar . . . he’s got her cooking
he is ki..ke..kee..ker cooking of . . . ”. Her description of
the “beach” scene (from the Queen Square Screening Test
for cognitive deficits [75]) was: “that’s a small place boat
(boy falling off a boat), this is a small plot plause (boy
building sandcastle) . . . that’s a small plause (father and
son playing ball) . . . ”. E.B.’s perseverations were largely
whole-word, although she exhibited occasional phonemic
perseverations. She appeared to be aware of her persever-
ations, frequently following them by saying “no, no it’s
not”.

3.1.1. Word and sentence repetition

In contrast to her markedly impaired speech output, her
ability to repeat three-syllable, high frequency words and
four–seven word sentences was preserved. The few errors
that she made in the word-repetition task were recurrent
Fig. 1. Post-operative MRI scan of patient E.B. showing the extent of the
cortical lesion. Brodmann’s areas (BAs) 44, 45, 46, 10, 22 and 6 were
perseverations.
affected.
3.1.2. Word retrieval
In contrast, on the Raven’s Colored Progressive Matrices, Her naming skills were gravely impaired. She was barely
she obtained an adequate score. Only her visual memory able to score on the Graded Difficulty Object Naming
functions were assessed on the Short Visual Recognition Test [43] and was only able to name a few items on the
Memory Test [76] and were found to be impaired. Her vi- Oldfield Picture Naming Test [51]. The majority of her
suospatial constructional skills were relatively preserved, as errors in these tests consisted of recurrent perseverations.
indicated by her low-average score on the block design sub- For example, she correctly named in succession the pic-
test of the WAIS-R. tures of a clock and a drum. When presented the picture
of a book, she said: “a drum . . . no a clock . . . no a
Table 1 drum . . . ”.
Neuropsychological assessment
Intellectual functions 3.1.3. Word and sentence comprehension
Verbal IQ 59 Clinically, her comprehension skills appeared relatively
Digit span 4 well preserved. She was able to follow a simple conversa-
Vocabulary 1 tion, to understand task instructions and to execute simple
Arithmetic 3
Similarities Failed to score
verbal commands. On formal testing, however, her perfor-
Block design 7 mance was poor both on a test of single word comprehension
Raven 26/36 75th percentile (British Picture Vocabulary Scale [15]) and on a test of sen-
Memory tence comprehension (Test for the Reception of Grammar
SVRMT for faces 16/25 <5th percentile [4]). In both tasks, she showed a tendency to perseverate in
Language her responses.
GDONT 2/30 Very poor
Oldfield 9/30 Very poor 3.2. Assessment of word retrieval, production,
BPVS 80/150 Poor
comprehension and transcoding skills
TROG 56/80 Poor
Raven: Raven’s Colored Progressive Matrices [60]; SVRMT: Short
In order to assess E.B.’s word processing skills in more
Visual Recognition Memory Test [76]; GDONT: Graded Difficulty Object
Naming Test [43]; Oldfield: Oldfield Picture Naming Test [51]; BPVS: detail, we administered oral and written naming, spoken
British Picture Vocabulary Scale [15]; TROG: Test for the Reception of word–picture matching, reading, writing and repetition tasks
Grammar [4]. across different semantic categories.
 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947
Table 2
Percent correct in oral and written naming, spoken word–picture matching, reading, repetition and writing to dictation as a function of semantic category
Animals Body parts
ONAME 30 60
WNAME 55 60
SWPM 100 90
READ 100 100
REP 95 100
DICT 90 100
Average 78.3 85
ONAME: Oral naming; WNAME: Written naming; SWPM: Spoken word–picture matching; READ: Reading; REP: Repetition; DICT: Writing to dictation.
3.2.1. Materials
There were five categories of nouns: 10 animals, 10 body
parts, 10 colors, 10 countries and 10 objects. The stimuli
were all of high frequency and were partly based on those
used by McKenna and Warrington [42] (for further details
see [9]).
3.2.2. Procedure
Six different conditions were tested: (1) oral naming;
(2) written naming; (3) spoken word–picture matching; (4)
reading; (5) repetition; and (6) writing to dictation. She
completed all these conditions, in this order, on two sepa-
rate sessions on the same day. Written naming, reading and
repetition were tested once again, in this order, in two sepa-
rate sessions in the following days. In all conditions, except
spoken word–picture matching, items were presented in-
dividually and serially, blocked by category. In the spoken
word–picture matching condition, arrays were comprised of
five items from a given category. In the naming tasks, E.B.
was asked to say or write the name corresponding to the pic-
ture. In the reading task, E.B. was asked to read aloud single
words, each typed on a card. In the writing to dictation
task, the examiner said each word once and E.B. was asked
to write it down. In the word-repetition task, the examiner
said each word once and the patient was asked to repeat
the stimulus immediately after the examiner. In the spoken
word–picture matching task, the examiner named each of the
five items in turn and the patient was required to point to the
appropriate item.
3.2.3. Results
The percentage of correct responses in each condition for
each of the five categories is given in Table 2. E.B.’s per-
formance on comprehension, reading, writing and repetition
tasks was almost at ceiling for all of the categories. In sharp
contrast, her oral and written naming performance was poor.
E.B.’s errors in oral and written naming tasks were
analyzed. Errors were classified as: (1) perseveration—
the repetition of a previously emitted response; (2)
semantic—category coordinate, associate or subordinate;
(3) phonological—including at least 50% of the target’s
phonemes; (4) unrelated—real or nonsense word bearing
no semantic or phonological relationship with the target;
1933
Colors Countries Objects Average
80 10 70 50
75 20 45 51
100 50 100 88
95 95 95 97
95 100 95 97
80 100 100 94
87.5 62.5 84.2 79.5
and (5) omission—silence, statement of no response or an
empty comment. The analysis revealed that in both naming
tasks, her errors were mostly whole-word perseverations
(32.5% of total errors; e.g. after having correctly named
the picture of a skirt, she said “skirt” to the two successive
pictures, a tent and a ball) and omissions (31.4% of total
errors; e.g. horse—“. . . this is a lovely.”). A few semantic
errors (14% of total errors; e.g. watch—“clock”) and unre-
lated errors (16.2% of total errors; e.g. arm—“. . . up one
side”) were also present.
3.2.4. Comments
The results from this assessment indicate that E.B. had
severe oral and written naming impairments even for very
high frequency words. In contrast, her performance in the
comprehension task was near ceiling. This pattern suggests
that her naming difficulty with these items was not primarily
attributable to a severe semantic impairment. Furthermore,
her preserved reading, repetition and writing skills indicated
that her oral and written output impairments were not due to
a deficit at the level of motor planning and articulation of a
phonemic or a graphemic sequence. E.B. showed persevera-
tions predominantly in the naming tasks. The purpose of the
following experimental studies is to investigate the underly-
ing mechanism responsible for her recurrent perseverations.
4. Experimental investigation
In the following experiments, we manipulated a number
of stimulus factors and assessed their influence on E.B.’s
tendency to produce recurrent perseverations while naming
pictures. Given the growing theoretical focus on the poten-
tial relationship between perseveration and priming effects,
we decided to concentrate on factors that have been shown to
influence priming effects in normal subjects. Priming theo-
ries of perseveration hold that perseverations result from the
failure of weakened or degraded perceptual input to over-
ride representations that are persisting in activity [12,37,73].
The basic prediction that follows from these theories is that
because priming and perseverations are due to the same
mechanisms (namely persistent activity), stimulus factors
that influence one should also influence the other. While
1934 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947
individual theories do differ somewhat in how activation is
computed and how links or connections are modified with
experience, the similarities across theories allow us to con-
sider at least two general classes of stimulus factors. Factors
that weaken or slow processing of current stimuli should
give rise to more perseverations because representations that
are persistently active are less effectively overridden. Al-
ternatively, factors that enhance the activity levels of active
representations should also give rise to more perseverations
because it is more difficult for input to override them. Along
these lines, we chose to manipulate the following stimulus
factors.
(1) Lexical frequency can be thought of as a factor that al-
ters the strength or speed of current stimulus process-
ing. Words that are low in frequency are less practiced
and are named more slowly than high frequency words
[18,48]. Frequency has also been shown to influence
priming effects across a variety of tasks [3,52,66]. If one
assumes that low frequency words have weaker corre-
sponding input, they should elicit a greater number of
perseverations than high frequency words because it will
be harder for them to override persistently active repre-
sentations. While there has been some support for this in
previous studies [30,64], frequency has yet to be directly
manipulated in a single case study of perseveration.
(2) Presentation rate can be thought of as a factor that in-
fluences the activity levels of persistent representations.
Passive or spontaneous decay of activation is a common
feature of many parallel processing theories ([12,37];
see also [13,40,55]). As persistent activity passively
decays over time, it will eventually reach a low level.
Consistent with this, greater priming effects are ob-
served after shorter delays, showing dramatic reductions
over several seconds [23,44,46]. Perseverations should,
therefore, also be more likely under a fast presentation
rate with shorter delays between stimuli because height-
ened persistent activity will be more difficult to override.
There has also been some support for this prediction in
a large group study of aphasic patients [64], but it has
yet to be tested in a more detailed single case study.
(3) Repetition can be thought of as another factor that in-
creases the activity levels of persistent representations.
Each time a stimulus is repeated, residual activity is
presumed to increase a certain amount, making it easier
to respond the next time the same stimulus is presented.
This suggestion is supported by the observation of rep-
etition priming, faster identification following one or
more stimulus repetitions [52,66,71,81]. Perseverations
should be more likely following repetition because it
will be increasingly difficult to override persistently
active representations.
(4) Semantic relatedness, like presentation rate and rep-
etition, can be thought of as a factor that influences
the activity levels of persistent representations. Normal
subjects are faster at processing a stimulus if it has just
been preceded by one that is related in meaning, a phe-
nomenon referred to as semantic priming [45,47,49].
If a stimulus at least partially activates semantically
related representations, then processing a semantic as-
sociate shortly thereafter may be facilitated by some
residual activity. However, if a persistently active rep-
resentation receives an additional boost from a seman-
tically related current stimulus, there is also a chance
that this representation will inappropriately generate a
perseverative response [72,73]. Perseverations should
be more likely when stimuli are all highly related in
meaning because there will be more opportunities for
a weak current stimulus to boost a semantically related
persistent representation.
While these four stimulus factors have not been manip-
ulated in single case studies of aphasic perseveration, all
four have been manipulated previously in studies of aphasic
comprehension performance. Warrington and Cipolotti [77]
investigated the effects of varying frequency, rate, repeti-
tion and semantic relatedness on the spoken word–picture
matching performance of patients believed to suffer from
either refractory or degraded-store impairments. Refractory
patients were affected by presentation rate, semantic relat-
edness and repetition but not by frequency. Degraded-store
patients, on the other hand, were strongly affected by lexical
frequency but not by rate, semantic relatedness or repetition.
A comparison of E.B.’s pattern of naming performance with
these other documented patterns might yield additional in-
sight into the mechanisms underlying aphasic perseveration
and how they differ from mechanisms revealed by other
neurological conditions.
4.1. General procedures
In all experiments, E.B. was simply required to provide
spoken names for pictures presented to her by the exper-
imenter. If she had not provided a response within 10 s,
an omission was scored and the picture was removed. The
RSI, i.e. the period of time between the offset of her re-
sponse and the presentation of the next picture, was strictly
maintained. As in other studies of aphasic perseveration,
naming trials were structured serially into blocks, and errors
were tabulated across trials [1,12,14,30,37,53,64]. Different
blocks corresponded to particular experimental conditions,
allowing effects of the stimulus factors to be assessed by
comparing error totals for the different conditions (see also
[14,24,53,64,77]). It is important to note that this basic ap-
proach differs somewhat from priming paradigms in which
trials are structured into prime–probe combinations and
properties of the prime stimuli are varied orthogonally from
properties of their probe stimuli [45,46]. In general, the
application of such a paradigm would dramatically increase
the number of naming trials necessary to obtain effects in
a single patient because each “probe” trial on which per-
severations could be elicited would need to be preceded
 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947 1935

by at least one prime stimulus. We have opted to use a straints effectively decreased the size of E.B.’s spared vo-
more standard paradigm for eliciting perseverative errors in cabulary and inflated the estimate of chance perseverations,
which each naming trial within a block serves as a potential resulting in a more conservative estimate of true persevera-
“prime” for subsequent trials and a potential “probe” for tions. Across all experiments, 16/136 perseverations (12%)
previous trials. The basic predictions of priming theories of were thrown out due to this filtering process (Note 1).
perseveration are not expected to depend critically on these
Note 1. Previous methods for estimating chance persever-
methodological differences.
ations have involved randomly scrambling the orders of
trials within blocks and comparing the actual trial-to-trial
4.2. Scoring of errors
time course of perseveration with randomized ones [12,30].
These methods implicitly require that blocks are long
4.2.1. Perseverations
enough to include a representative sampling of the response
A perseveration was scored if the current response con-
dimension being perseverated (e.g. phonemes, words, etc.),
tained at least half of a previous response and was not the
a requirement that is strongly violated in our experiments.
correct response (e.g. having responded “swan” on a previ-
The method that we have developed will yield estimates
ous trial, responding “swan” or “swat” to the picture of a cow
similar to other methods for long blocks but appears more
on the current trial). To simplify scoring, only responses re-
appropriate when shorter blocks are employed, as the other
peated within the current block of trials were counted as per-
methods will heavily overestimate the base probability of
severations, and perseverative responses that were repeated
particular responses in these circumstances.
on a single trial were counted only once. Any spontaneous
self-corrections were allowed, although the error responses
4.2.2. Omissions
were still scored and included. The number of persever-
An omission was scored if E.B. gave no response or indi-
ations was tabulated across trials for each block, and the
cated that she did not know the name of the current picture.
number for each block was further collapsed across blocks
Similarly to perseverations, the number of omissions was
that sampled the same levels of experimentally manipulated
tabulated across trials for each block in each experimental
stimulus factors. As others have pointed out previously
condition.
[12,17,30], errors might be erroneously scored as persever-
ations if patients are simply generating responses randomly
4.2.3. Other errors
from their spared vocabularies, particularly if their spared
All incorrect responses that were not perseverations or
vocabularies are small in size. To control for this possibility,
omissions were scored as other. We initially employed a
we developed a method for filtering out chance persever-
more complicated scoring system that included phonolog-
ations. Chance levels of perseveration were estimated by
ical, semantic and visual errors, as well as circumlocutions
replacing all of E.B.’s error commissions (errors other than
and neologisms. However, each of these error types turned
omissions) with new error responses randomly selected from
out to be relatively infrequent and no individual types ap-
her known vocabulary (i.e. responses that had been correctly
peared to be significantly influenced by the experimental
provided at least once throughout the course of the experi-
manipulations. For the sake of simplicity, the other clas-
ments). We considered it unlikely that E.B. would select re-
sification was adopted. The number of other errors was
sponses completely at random without some constraints due
tabulated across trials for each block in each experimental
to the target stimuli or familiarity with particular responses.
condition. In cases where E.B. spontaneously corrected
Therefore, selection of the new error responses was subject
herself, errors were still tabulated.
to both weighted frequency/familiarity and semantic cate-
gory constraints such that errors were from the same broad
superordinate semantic category as the target stimulus and 4.3. Experiment 1: frequency, presentation rate,
tended to be high in lexical frequency [8]. Chance persever- repetition and semantic relatedness
ations were tabulated in the same way as E.B.’s actual error
responses and the entire process was repeated 100 times to The aim of the first experiment was to determine if fre-
yield a stable estimate of chance. The distance or lag in tri- quency, presentation rate, repetition and semantic related-
als between each perseverative response and its most recent ness would influence E.B.’s tendency to produce recurrent
previous utterance was also recorded, as earlier studies have perseverations in picture naming. In order to assess this, we
shown the time course of perseverative phenomena to be im- modified a word–picture matching test described in detail
portant [12] (discussed later in greater detail). For each trial by Cipolotti and Warrington (Experiment 5 in [10]) and
lag in each experimental condition, the number of actual used it as a picture naming task.
perseverations was compared with its corresponding chance
estimate. If this number exceeded the 95% prediction inter- 4.3.1. Stimuli
val for the chance estimate, the actual perseverations were The stimuli consisted of colored pictures of 16 objects,
retained in the total (see [50] for discussion of prediction 16 animals and 16 foods. Each category of stimuli was di-
intervals). The application of frequency and semantic con- vided into four sets of four items. There were 12 stimulus
1936 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947

sets in total. Half of the sets contained pictures with high
frequency names and the other half contained pictures with
low frequency names. Crossed with frequency and category
was semantic relatedness. Half of the sets contained pictures
that were closely related semantically (i.e. close) and half
contained pictures that were more distantly related, with-
out crossing a major category boundary (i.e. distant). Thus,
there were four types of sets for each of the three categories
(i.e. close high frequency; close low frequency; distant high
frequency; distant low frequency).
4.3.2. Procedure
Each of the 12 stimulus sets was tested both at a fast rate
(RSI = 1 s) and a slow rate of presentation (RSI = 15 s) in
blocks. Pictures were presented individually and in a given
block, each of the four pictures was presented twice in a
pseudo-random order. The 12 sets were administered in a
predetermined order such that fast-rate and slow-rate blocks
alternated. Successive blocks differed in category, frequency
and semantic relatedness.
4.3.3. Results
The incidence of each error type and the correspond-
ing percentage of total error are presented for each factor
individually in Table 3. Comparisons for each error type
were carried out by computing a χ 2 statistic using the total
number of trials that contained a particular error type and
the total number of trials that did not contain the error type,
yet afforded an opportunity to make one (Note 2). This al-
lowed us to detect changes in error number across different
experimental conditions while normalizing for differences
in opportunities to make those errors. We chose this over
an approach using percentage of total error because such
an approach would implicitly assume independence of the
mechanisms that produce different error types.
Note 2. It is important to mention that the independence
assumption of the χ 2 -test may not be strictly satisfied here.
While this assumption is often relaxed in analyses of aphasic
errors [12,67,78,79], we have verified the results for perse-
verations and total errors in each experiment with repeated
measures ANOVAs calculated over the number of errors
elicited per experimental block. Moreover, results that are
highly significant or that replicate across experiments are
unlikely to be strongly undermined by this issue.
Frequency had a significant influence on E.B.’s naming
performance. Pictures with low frequency names elicited
more total errors (χ 2 (d.f. 1) = 54.28, P < 0.001), per-
severations (χ 2 (d.f. 1) = 11.16, P < 0.001), omissions
(χ 2 (d.f. 1) = 9.13, P < 0.005) and other errors (χ 2
(d.f. 1) = 5.27, P < 0.05). Rate had a significant effect
on total errors, with a greater number of errors under a fast
rate (χ 2 (d.f. 1) = 4.70, P < 0.05). However, rate had no
significant effects on individual error types. Perseverations
were more than doubled following repetition, although this
effect just failed to reach significance (χ 2 (d.f. 1) = 2.8,
P < 0.1). Repetition did have a significant influence on
other errors (χ 2 (d.f. 1) = 10.4, P < 0.01). This effect
could not be attributed easily to any single standard speech
error type (e.g. neologism, circumlocution, etc.). Semantic
relatedness failed to have significant effects on any error
types. There were no significant interactions involving any
of the variables.
4.3.4. Comments
Consistent with the results of the group study conducted
by Santo Pietro and Rigrodsky [64], E.B. was more likely to
produce perseverations and other errors when the picture to
be named had a low frequency name. However, in contrast to
their results, rate did not significantly affect perseverations,
influencing instead total errors with a slight trend for more
omissions at a fast rate. This raises the possibility that not all
aphasics who perseverate are similarly affected by presenta-
tion rate. Repetition and semantic relatedness also failed to
have significant effects on perseveration, although there was
a trend for more perseverations following repetition. The
failure to observe these effects led us to conduct a second ex-
periment (consideration of the effects of stimulus factors on
omissions and other errors will be deferred until Section 6).
4.4. Experiment 2: presentation rate, repetition and
semantic relatedness
In Experiment 1, rate, repetition and semantic relatedness
did not have significant effects on E.B.’s perseverations.
Experiment 2 aimed to replicate and extend these results
using a different set of items. A word–picture matching test
of common household objects was modified and used as a
picture naming task.

Table 3
Experiment 1: the effect of frequency, presentation rate, repetition and semantic relatedness on number of errors
Frequency Rate (RSI) Repetition Semantic relatedness

High Low P Fast (1 s) Slow (15 s) P First Second P Close Distant P

Perseverations 7 (25) 25 (32) <0.001 19 (32) 13 (29) NS 9 (19) 23 (43) <0.1 17 (31) 15 (30) NS
Omissions 11 (39) 29 (38) <0.005 25 (42) 15 (33) NS 16 (33) 24 (45) NS 22 (40) 18 (36) NS
Other 10 (36) 23 (30) <0.05 16 (27) 17 (38) NS 23 (48) 6 (11) <0.01 16 (29) 17 (34) NS
Total errors 28 77 <0.001 60 45 <0.05 48 53 NS 55 50 NS
The values given in parentheses are percentage of total.
 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947
4.4.1. Stimuli
Stimuli were all black and white line drawings of com-
mon objects selected from the categories of clothing, house-
hold utensils, kitchen utensils, furniture, office supplies and
vehicles (for further details, see Experiment 6 in [79]). In
contrast to the last experiment, pictures were arranged in ar-
rays of six items each. There were a total of 12 arrays: 6
semantically close in which all items were from the same
subordinate category of objects and 6 semantically distant
arrays that consisted of 1 item from each of the 6 subordi-
nate categories.
4.4.2. Procedure
Each array was tested at both a fast rate (RSI = 1 s)
and a slow rate of presentation (RSI = 10 s) in a blocked
design. On each trial, the experimenter indicated to E.B.
which drawing was to be named by pointing to it in the
array. Each drawing was probed twice in a pseudo-random
order. The 12 arrays were tested in a predetermined order
such that the fast-rate and slow-rate blocks alternated, and
successive arrays differed in semantic relatedness.
4.4.3. Results
The incidence of each error type and the corresponding
percent of total error are presented for each factor individu-
ally in Table 4.
Rate influenced total errors (χ 2 (d.f. 1) = 7.01, P <
0.01) and omissions (χ 2 (d.f. 1) = 13.37, P < 0.001).
However, it again had no significant effect on persevera-
tions. Repetition had a significant effect on total errors (Mc-
Nemar: χ 2 (d.f. 1) = 18.58, P < 0.001) and persevera-
tions (χ 2 (d.f. 1) = 15.13, P < 0.001), and there was a
non-significant trend for other errors (χ 2 (d.f. 1) = 3.15,
P < 0.1). Semantic relatedness had a significant effect on
perseverations (χ 2 (d.f. 1) = 4.62, P < 0.05). Addition-
ally, there was a significant interaction in total errors be-
tween rate and semantic relatedness (χ 2 (d.f. 1) = 4.39,
P < 0.05). E.B.’s performance in the distant condition was
significantly better than in the close condition at a slow rate
(χ 2 (d.f. 1) = 5.45, P < 0.02), but not at a fast rate (χ 2
(d.f. 1) = 2.59, P < 0.2). However, interpretation of this
interaction was complicated by an order effect in testing.
The two worst conditions, fast-distant and close-slow, were
the last two administered (first versus last half: χ 2 (d.f. 1) =
8.34, P < 0.01).
Table 4
Experiment 2: the effect of presentation rate, repetition and semantic relatedness on number of errors
Rate (RSI)
Fast (1 s) Slow (10 s) P First
Perseverations 40 (41) 39 (51) NS
Omissions 30 (31) 8 (11) <0.001
Other 28 (29) 29 (38) NS
Total errors 98 76 <0.01
The values given in parentheses are percentage of total.
1937
4.4.4. Comments
Rate had a significant effect on correct performance
that appeared attributable to more omissions at a fast rate.
However, there was still no rate effect on perseverations.
The effects of repetition and semantic relatedness on perse-
verations did reach significance in this experiment (further
consideration of the rate × semantic relatedness interaction
in total errors will be deferred until Section 6). We decided
to examine the effects of repetition and semantic relatedness
further in a third experiment.
4.5. Experiments 3A and 3B: repetition and
semantic relatedness
In Experiment 2, we found that both repetition and se-
mantic relatedness had significant effects on perseverations.
While there was a trend for more perseverations following
repetition in Experiment 1, neither repetition nor semantic
relatedness yielded significant effects in that experiment.
In this experiment, we attempted to assess the replicability
of the repetition and semantic relatedness effects on perse-
verations. Warrington and Cipolotti [77] have demonstrated
previously that some aphasic patients who do not show
semantic relatedness effects for within-category manipu-
lations may nevertheless show them for between-category
manipulations. Accordingly, we examined broader and more
extreme semantic relatedness manipulations to ensure that
any lack of effect would not be due to the specific semantic
distance used. In Experiment 3A, we compared E.B.’s per-
formance on semantically distant arrays (all pictures in the
same category) with her performance on very distant arrays
(each picture from a different category). Experiment 3B
involved an even more extreme manipulation, comparing
performance on semantically close arrays with very distant
arrays.
4.5.1. Stimuli
The stimuli consisted of colored pictures of 12 animals,
12 objects and 12 foods, all with moderate to high frequency
names (median frequency = 226.5, S.D. = 465.9 [8]). Pic-
tures in each category were divided into four within-category
sets of three pictures each. Half of these within-category
sets were semantically distant and were used in Experiment
3A (e.g. bread, ice cream and tomato), and other half were
semantically close sets used in Experiment 3B (e.g. apple,
Repetition Semantic relatedness
Second P Close Distant P
21 (28) 58 (58) <0.001 48 (54) 31 (36) <0.05
18 (24) 20 (20) NS 17 (19) 21 (25) NS
35 (47) 22 (22) <0.1 24 (27) 33 (39) NS
74 100 <0.001 89 85 NS
1938 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947

banana and orange). Between-category (very distant) sets Table 6

for both Experiments 3A and 3B were formed by taking Experiments 3A and 3B combined (individual error types): the effect of
repetition and semantic relatedness on number of errors
one picture from each of three different within-category sets
(e.g. horse, pipe and tomato). This resulted in a total of six Repetition Semantic relatedness
semantically distant/six very distant sets for Experiment 3A First Second P Distant Very distant P
and six semantically close/six very distant sets for Experi- Perseverations 0 (0) 9 (33) <0.05 5 (26) 4 (24) NS
ment 3B. In each experiment, a given picture occurred in ex- Omissions 4 (44) 12 (44) <0.1 11 (58) 5 (29) NS
actly one within-category set and one between-category set. Other 5 (56) 6 (22) NS 3 (16) 8 (47) NS
Total errors 9 27 <0.001 19 17 NS
4.5.2. Procedure The values given in parentheses are percentage of total.
The testing procedure was the same for both Experiments
3A and 3B. Each stimulus set was tested in a blocked fash-
ion at a fast rate of presentation (RSI = 1 s). Pictures were semantically very distant blocks from both experiments were
presented individually and in a given block, each of the three also pooled to form one very distant condition. The effects
pictures in a set was presented two times in a pseudo-random of repetition and semantic relatedness on individual error
order, resulting in six trials per block. Stimulus sets were types after pooling the data are presented in Table 6. There
presented in a predetermined order with successive sets dif- was an effect of repetition on E.B.’s total errors (McNemar:
fering in semantic relatedness. χ 2 (d.f. 1) = 12.19, P < 0.001) and perseverations (χ 2
(d.f. 1) = 4.81, P < 0.05), as well as a non-significant trend
4.5.3. Results for omissions (χ 2 (d.f. 1) = 3.45, P < 0.1). However, there
The effect of repetition and semantic relatedness on the was still no effect of semantic relatedness on any error type.
incidence of total errors is presented in Table 5 separately
for Experiments 3A and 3B. In Experiment 3A, there was 4.5.4. Comments
an effect of repetition on E.B.’s total errors (McNemar: χ 2 Repetition was found to influence E.B.’s perseverations,
(d.f. 1) = 5.82, P < 0.02), while semantic relatedness replicating the results of the previous experiments. Seman-
failed to yield a significant effect (χ 2 (d.f. 1) = 0.0, P < tic relatedness failed to have a significant effect on errors of
1.0). Unfortunately, the frequencies of individual error types any type even when broader and more extreme manipula-
were too small to reveal any more specific effects. In Ex- tions were used. This suggests that the influence of semantic
periment 3B, repetition again had a significant effect on to- relatedness on E.B.’s poor naming performance was weak
tal errors (McNemar: χ 2 (d.f. 1) = 5.82, P < 0.02), while at the best.
semantic relatedness again failed to elicit any significant ef-
fects (χ 2 (d.f. 1) = 0.34, P < 0.7). Frequencies of individ-
ual error types were also too small in this experiment to yield 5. Error analyses
more specific effects on error types. Given that repetition
and semantic relatedness yielded very similar patterns of re- We carried out three error analyses. These analyses al-
sults in both Experiments 3A and 3B, we pooled the data to lowed us to clarify: (1) the nature of the relationship be-
see if there might be significant effects on individual error tween perseverations and their corresponding target stimuli;
types. The semantically distant blocks from Experiment 3A (2) the relationship between target stimuli leading to the
were pooled with the semantically close blocks from Exper- same perseverative response; and (3) the influence of time
iment 3B into one distant (within-category) condition. The and intervening trials on the persistence of perseverations.

5.1. Analysis 1: the relationship between perseverations

Table 5 and their corresponding target stimuli
Experiments 3A and 3B (all errors): the effect of repetition and semantic
relatedness on number of errors
In the previous experiments, semantic relatedness had no
Repetition significant effect on the number of perseverations. Cohen
First Second Total and Dehaene [12] have recently suggested that persevera-
tions may occur at any of a number of processing levels
Experiment 3A
Distant 3 7 10 according to where in the processing system the input has
Very distant 3 8 11 been degraded or “deafferented”. Following this suggestion,
Total 6 15
if E.B.’s input to semantic processing were not greatly de-
graded, it might not be surprising that semantic relatedness
Experiment 3B
did not influence perseverations significantly. E.B.’s im-
Close 1 8 9
Very distant 2 4 6 pairment may arise instead from damage to a post-semantic
processing level that is necessary for naming, such as pho-
Total 3 12
nology and/or representations mediating between semantics
 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947
Table 7
Number of perseverations by target–perseverate relation and semantic
relatedness condition tabulated across Experiments 1–3
Relation Semantic relatedness condition
High related Low related
Semantic 45 (63) 11 (22)
Phonological 0 (0) 3 (6)
Combined 7 (10) 1 (2)
semantic–
phonological
Unrelated 19 (27) 34 (69)
Total 71 49
The values given in parentheses are percentage of total.
and phonology (see [34] for discussion). If this were the
case, one might expect a high number of perseverations to
share a phonological or combined semantic–phonological
relationship with their corresponding target stimuli. In order
to examine this possibility, we re-analyzed E.B.’s persever-
ations from the previous three experiments. Each persever-
ation was classified as being: (1) semantically related to the
target name (e.g. trousers—“jacket”); (2) phonologically
related to the target name (i.e. sharing 50% of the target’s
phonemes: e.g. coat—“comb”); (3) semantically and phono-
logically related to the target name (e.g. shirt—“skirt”); or
(4) unrelated to the target name (e.g. apple—“horse”) (see
[30,36] for similar approaches to analyzing perseverative er-
rors). Results of this tabulation across all three experiments
are presented in Table 7 by semantic relatedness condition.
An interesting pattern is present when one examines per-
severative errors in the high versus low semantic relatedness
conditions. In the high semantic relatedness conditions, most
perseverations shared a semantic relationship with their tar-
gets, with fewer unrelated perseverations. Phonological and
combined semantic–phonological perseverations were rela-
tively infrequent. However, in the low semantic relatedness
conditions, most of the perseverations were unrelated to
their targets, with a few semantic, phonological and com-
bined semantic–phonological perseverations. The simplest
explanation of this pattern of data is that E.B. was simply
repeating recent responses and was relatively unaffected by
the semantic relatedness manipulation. Recent responses
would happen to be semantically related to current targets
in the high semantic relatedness conditions by virtue of
the way in which stimuli were blocked and they would be
unrelated to current targets in the low semantic relatedness
conditions. In other words, the factors giving rise to E.B.’s
perseverations most of the time appear to be unrelated to
the meaning of the target or the phonology of its name.
5.2. Analysis 2: the relationship between target stimuli
leading to the same perseverative response
In the last analysis, we found that E.B. produced many
perseverations that were unrelated to their corresponding
target stimuli. In line with this observation, Hirsh [30] de-
1939
scribed the performance of another aphasic patient (C.J.)
who also exhibited a high number of unrelated persevera-
tions. Interestingly, when the target stimuli that led to the
same perseverative responses were analyzed, they were
Overall found to be semantically and/or phonologically related. For
56 (47) example, if the stimulus dog elicited the response “desk”,
3 (3) then a subsequent stimulus that was semantically related to
8 (7) dog (e.g. cat) might elicit the response “desk” again—even
though dog and cat share no obvious relationship to desk.
53 (44) One potential explanation of this finding is that a rapid
association was being formed between target stimulus and
120
unrelated error response. The next time the same target or
one similar to it was presented, the same response might
be triggered, yielding a perseveration. Close inspection of
E.B.’s perseverative responses appears to support the idea
that rapid associations were being formed between target
stimuli and error responses, as well as between consecutive
stimuli and responses. Perseverations that followed inter-
vening responses often appeared to be related to the initial
sequential proximity of target and response. Appendix A
presents three examples of E.B.’s responses from experi-
mental blocks that illustrate this point. For instance, Exam-
ple 1 shows that when the stimulus olives was presented,
E.B. incorrectly provided “fig” from two trials back, a per-
severation. The next stimulus dates directly followed olives,
eliciting another perseverative response “a fri-”. When dates
was presented again three trials later, the perseverative re-
sponse “fig” was once again produced. Examples 2 and 3
provide similar instances of this phenomenon.
In order to quantify this phenomenon and evaluate it sta-
tistically, we determined the likelihood that a perseverative
response to a particular target stimulus in the second half
of a block would be given within close temporal proximity
of the same target stimulus in the first half of the block. We
considered only perseverations that followed other inter-
vening responses in the second half of each block because
these perseverations could not be explained simply by as-
suming that E.B. was stuck on an immediately preceding
response. For each of these perseverations, the target stimu-
lus that elicited it was noted. If E.B. had provided the same
response in the first half of the block, its distance in trials
from the earlier presentation of the same target was calcu-
lated and tabulated across perseverations. The results from
these tabulations are presented in Fig. 2 by experiment,
along with two different sets of controls.
For the first control, we randomly scrambled the order
of the trials within the first half of each block 100 times
and re-tabulated the number of responses provided at each
target–response distance. This allowed us to detect any struc-
ture present in the temporal order of targets and responses.
The second control was similar to the first, but only involved
tabulation of correct responses at each target–response dis-
tance. This made it possible to determine how much of any
observed temporal structure was simply an artifact of E.B.
correctly performing the naming task. For both controls,
prediction intervals were generated for responses at each
1940 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947

Fig. 2. Experiments 1 and 2: number of times a perseverative response in the second half of each block was given near the same target stimulus in the
first half of each block. The distance in number of trials between the earlier target and response is given on the abscissa; negative distances indicate that
the response occurred before the target stimulus. Control estimates showing the target–response distances for randomly scrambled trial orders (all trials,
as well as correct only) are provided for comparison.

target–response distance based on variability present in the
scrambled trial orders. If the actual number of responses
given at a certain distance was outside of the control range,
the responses were taken to be significant at the level of the
prediction interval (e.g. P < 0.01 indicates that the number
of responses was outside the 99% prediction interval—the
interval in which 99% of the control values lie). While there
were too few delayed perseverations in Experiment 3 to
analyze, the same basic pattern was found for both Experi-
ments 1 and 2. Perseverative responses given in the second
half of the block appeared to be provided at or near the same
target stimulus earlier, becoming less likely as responses
either substantially preceded or followed the earlier target.
The number of responses at a target–response distance of 0
(i.e. the same target that later elicited the delayed persever-
ation) was significantly greater than the control estimates
based on randomly scrambling the order of all trials in the
first half of the block (Experiment 1: P < 0.01; Experi-
ment 2: P < 0.05). This indicated that the structure present
in the temporal order of targets and responses was indeed
non-random. A fraction of these earlier responses reflected
correct naming responses proximal to the earlier target.
However, the number of perseverative responses that were
provided just before or just after the same target stimulus
earlier exceeded the number explained by correct respond-
ing in both experiments (Experiment 1—target–response
distances of ±1 combined: P < 0.001, ±2 combined: P <
0.005; Experiment 2—target–response distances of ±2 com-
bined: P < 0.03, ±3 combined: P < 0.1). These findings,
taken together, demonstrate that perseverations following
 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947
intervening responses were related to the initial temporal
proximity of target and response. The formation of rapid
associations between stimulus and response might, there-
fore, account for many of E.B.’s delayed and/or unrelated
perseverations.
5.3. Analysis 3: the relationship between time, intervening
trials and the persistence of perseverative responses
In Experiments 1 and 2, manipulating presentation rate
appeared to have no impact on E.B.’s perseverations. In con-
trast, Santo Pietro and Rigrodsky [64] found that a group
of aphasics produced fewer perseverations at a slow rate
(RSI = 10 s versus 1 s). Unfortunately, they did not present
data for individual patients. Their population may have been
heterogeneous with respect to the rate effect. The lack of a
rate effect is also interesting in the context of results reported
by Cohen and Dehaene [12]. They showed that all three of
their patients were less likely to perseverate as the number
of intervening trials increased. If this decrement were due
to elapsed time rather than intervening trials, per se, fewer
perseverations would be expected under a slower presenta-
tion rate. On the other hand, if the decrement were due to
the processing of intervening trials, no rate effect would be
expected. Cohen and Dehaene [12] were unable to distin-
guish between these two possibilities because presentation
Fig. 3. Experiments 1 and 2: distributions of the lag in trials between perseverations and corresponding previous responses for fast and slow rates of
presentation.
1941
rate was not varied experimentally. The rate manipulation
employed in Experiments 1 and 2 allowed us to examine
the relationship between elapsed time, intervening trials and
the persistence of perseverations more explicitly. We carried
out a “lag” analysis similar to the ones carried out by Cohen
and Dehaene [12]. We matched each one of E.B.’s perse-
verative responses with its most recent previous utterance
and then recorded the number of trials separating the two
responses (i.e. the trial lag). The distributions of trial lags
are presented in Fig. 3 by experiment and presentation rate,
along with estimates of the average number of chance perse-
verations at each lag. As in earlier analyses, these persever-
ations represent those that had already survived the method
of filtering chance perseverations described earlier.
For both the fast and slow presentation rates in Experi-
ments 1 and 2, perseverations were most common at short
trial lags. In fact, consecutive strings of the same response
were quite frequent, sometimes including as many as five
or six consecutive trials and accounting for approximately
55% of E.B.’s perseverations. Perseverations became more
infrequent as the number of intervening trials increased,
decrementing rather continuously up to about four or five
intervening trials—a pattern consistent with the results of
Cohen and Dehaene [12]. Interestingly, this decrement was
not significantly different under a slow presentation rate
than under a fast rate (Experiment 1: χ 2 (d.f. 2) = 2.57,
1942 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947
P < 0.3; Experiment 2: χ 2 (d.f. 3) = 2.96, P < 0.5). It
is important to note that with respect to actual time, the
decrement is occurring at very different time courses for
the two rate conditions due to the large difference in RSIs
used. For example, a perseveration in the fast condition that
occurred after two intervening trials would be produced
roughly 12 s after its corresponding previous utterance (as-
suming an average naming latency of 3 s, consistent with
our informal observations of E.B.’s naming times). A per-
severation following the same number of trials in the slow
condition of Experiment 1 would be produced after roughly
54 s. The fact that the trial lag distributions are not very
different across rate conditions for both Experiments 1 and
2 suggests that the variable mediating the decrement is the
number of intervening trials rather than time.
6. General discussion
We investigated the residual naming abilities of an apha-
sic patient (E.B.) who suffered from a marked anomia. E.B.
produced many recurrent perseverations and omissions in
both oral and written naming of pictures with high fre-
quency names. Her comprehension, reading, writing and
repetition abilities were relatively preserved, indicating that
her naming deficit was most likely due to damage affecting a
post-semantic processing level, yet prior to the level of mo-
tor planning and articulation of a phonemic or a graphemic
sequence. This is consistent with the anatomical evidence
available from the post-operative MRI scan (Fig. 1). Cortical
areas most selectively associated with semantic processing
did not appear to be affected by the lesion (i.e. BAs 20,
28/38, 39 and 47; see [57] for a review). Instead, her lesion
appears to include areas that have been argued to reflect
phonological processing and the willful generation of verbal
responses to stimuli that activate many potential response
options (i.e. BAs 44, 45 and 46, or Broca’s area [19,58,61]).
Results of the clinical assessment indicated that E.B.’s
perseverations were largely restricted to naming tasks. This
provides further evidence for task selectivity in aphasic per-
severation (see also [12,14,53]). As in previous studies, the
tasks that elicited the fewest perseverations were also per-
formed near ceiling (e.g. spoken word–picture matching,
reading, repetition and writing to dictation).
In three experiments, we assessed the influence of stim-
ulus factors on E.B.’s perseverative tendencies. We found
that perseverations were more likely for stimuli with low
frequency names compared to those with high frequency
names. Perseverations were also more likely following a
within-block stimulus repetition. While the rate of stimulus
presentation did have a significant influence on E.B.’s total
errors and number of omissions, it failed to influence either
the number of perseverations produced or their persistence
across intervening trials. The within-block semantic relat-
edness of stimuli had a significant impact on perseverations
only in Experiment 2, failing to have a significant effect on
perseverations or other error types in Experiments 1, 3A and
3B. Moreover, there was little evidence that perseverative re-
sponses shared a semantic relationship with the targets that
elicited them. While there was a significant interaction be-
tween rate and semantic relatedness in total errors in Exper-
iment 2, the weight of the evidence suggests that semantic
relatedness was not a critical factor mediating E.B.’s poor
naming performance.
All of the stimulus factors used in the above experi-
ments have been manipulated previously with other aphasic
patients. Patients proposed to suffer from refractory impair-
ments perform poorly in a spoken word–picture matching
task under conditions of fast presentation rate and repeated
stimuli, while their performance is relatively unaffected
by lexical frequency ([10,77]; see also [78,79]). Patients
proposed to suffer from damage directly to semantic rep-
resentations (i.e. a degraded-store impairment) exhibit a
contrasting pattern of effects in that they are unaffected
by rate, they perform consistently across repetitions, but
they have particular difficulty identifying stimuli with low
frequency names [11,74,77,80]. Both patient groups suffer
from semantic impairments and are affected by the semantic
relatedness of distractors (see [59,69,77] for further discus-
sion). Although the main focus of this investigation was
to examine the effects of manipulating stimulus factors on
E.B.’s perseverations, her overall naming performance as
reflected by total errors can be compared with the identifi-
cation performance of these other patients. Similarly to the
refractory patients, E.B. performed worse with a fast rate
and following a within-block repetition. However, like the
degraded-store patients, she was significantly affected by
lexical frequency. Despite her significant repetition effects,
E.B. also tended to be consistent in which pictures she
named correctly (Note 3). Unlike either patient population,
E.B. was unaffected by semantic relatedness and appeared
to have spared comprehension abilities. The mixing of re-
fractory and degraded-store patterns along with marked
perseverative behavior might suggest that E.B. is suffering
from a substantively different type of impairment than ei-
ther group of patient. E.B.’s pattern of performance also
provides evidence that rate and repetition effects are not
exhibited solely by patients with comprehension difficulties
(see also [38]).
Note 3. Analyses identical to those carried out by Cipolotti
and Warrington [10,77] revealed that E.B.’s performance
across within-block repetitions was more consistent than
expected by chance in two out of three experiments (Ex-
periment 1: χ 2 (d.f. 2) = 6.73, P < 0.05; Experiment 2:
χ 2 (d.f. 2) = 6.31, P < 0.05; Experiments 3A and 3B
combined: χ 2 (d.f. 2) = 1.37, P > 0.3).
6.1. Relationship between priming and perseveration
Recent theoretical accounts developed from psychologi-
cal and neuropsychological perspectives have suggested that
 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947
recurrent perseverations result from the same mecha-
nisms that give rise to priming effects in normal subjects
([12,37,73]; see [56] for a similar view). For example, Co-
hen and Dehaene [12] have proposed a general information
processing account that is capable of addressing a broad
range of empirical effects associated with priming and
perseveration. On their theory, cognitive representations
at many processing levels “persist” in activity for some
period of time following initial activation. Persistence is
formulated in a general sense and may ultimately reflect
any of a variety of potential mechanisms (e.g. sustained
activity, transient changes in links or connection strengths,
temporary changes to activation functions, etc.). In normal
subjects, persistence is facilitatory if an identical or similar
stimulus is presented (i.e. priming). If a different stimulus
is presented, the corresponding input overrides persistence
through a competitive process. When brain damage deaffer-
ents input to a certain level of processing, representations
persisting in activity are not always effectively overridden,
resulting in perseverations. The format of the persevera-
tions will then correspond to the format of the deafferented
processing level (e.g. phonemes, words, digits, etc.).
Several aspects of E.B.’s perseverative behavior are pre-
dicted by this theoretical perspective. Factors that weaken
or slow the processing of a current stimulus such as reduced
lexical frequency should give rise to more perseverations
because it is more difficult for weakened input to override
persistent activity. Indeed, E.B. produced more persevera-
tions, as well as other types of errors, to pictures with low
frequency names. Factors that increase the activity levels
of active representations such as repetition should also give
rise to more perseverations because it will be more difficult
for input to override enhanced persistent activity. Consistent
with this prediction, E.B. produced more perseverations
following within-block repetition of the same stimuli. Se-
mantic relatedness might also be expected to influence
perseverations because persistently active representations
can receive an additional boost from semantically related
stimuli, increasing the chance that they will inappropriately
elicit perseverative responses. While we found little evi-
dence for such an influence in our experiments, this is not
necessarily inconsistent with priming accounts of persever-
ation. As mentioned above, if E.B.’s damage did not deaf-
ferent input to semantic processing but instead deafferented
a post-semantic processing locus, one would not necessar-
ily expect to observe an effect of semantic relatedness on
perseveration. Persistent activity at the level of semantic
processing might be overridden in a normal fashion.
However, current priming theories of perseveration failed
to predict other features of E.B.’s perseverative behavior. In
contrast to the results of Santo Pietro and Rigrodsky [64], we
found no evidence for an effect of presentation rate on the
number of perseverations that E.B. produced. We did find
evidence that perseverations became less likely as the num-
ber of intervening trials increased, although this trend did
not interact with presentation rate. We also found evidence
1943
that E.B. produced many perseverations that were unrelated
either semantically or phonologically to their corresponding
targets, often after several intervening stimuli. When ana-
lyzed in more detail, we established that these persevera-
tions appeared to be related to the sequential proximity of
earlier stimuli and responses. What might help to explain
these findings?
6.2. Perseveration and cholinergic deficits
Sandson and Albert [63] proposed that recurrent per-
severations result from cholinergic deficits and low levels
of acetylcholine. Consistent with this proposal, recurrent
perseverations have been shown to be correlated with low
cortical levels of choline acetyltransferase (ChAT) and high
numbers of senile plaques, indicating cholinergic deficits
[20]. Drugs that mimic acetylcholine have also been shown
to reduce the number of perseverations elicited by patients
who are believed to suffer from cholinergic deficits (e.g.
post-encephalitic amnesia [54]; Alzheimer’s disease [70]).
Cholinergic neurons located in subcortical nuclei of the
basal forebrain and brainstem send projections up to the
cortex in segregated fiber bundles that diffusely innervate
different neocortical regions. The segregated nature of these
projections may allow for potentially selective effects of
damage due to a wide variety of pathologies and disease
processes [68].
A range of neuroscience studies has suggested that the
functional role of acetylcholine in the brain is to enhance the
processing of sensory inputs and to facilitate the encoding
of new memories [5,21,28,32]. Hasselmo and co-workers
[25–27,29] have attempted to explain these cognitive-level
functions in the context of a neural network model of
the cortex that incorporates the cellular actions of acetyl-
choline. In general, the model makes a distinction between
intrinsic/feedback and afferent cortical projections. Intrinsic
projections arise from within a given cortical region and
terminate in the same region. Afferent projections arise
from outside a cortical region and pass information along
in a bottom-up fashion from cortical regions involved in
earlier sensory processing. Feedback projections then send
information back toward more sensory regions. Acetyl-
choline largely suppresses transmitter release at intrinsic
and feedback projections while simultaneously making
cortical cells more responsive to excitatory inputs and en-
hancing synaptic plasticity effects. This leads to a dynamic
in cortical processing in which feedforward sensory inputs
play a stronger role in determining the activity of neurons in
a cortical region. It also helps to reduce interference across
stimuli in learning by preventing different neural represen-
tations from becoming co-activated via intrinsic/feedback
projections and associated through synaptic plasticity.
When levels of acetylcholine are low, intrinsic/feedback
projections are no longer strongly suppressed. Cells are
somewhat less excitable overall, although intrinsic/feedback
inputs contribute proportionately more to the activity that
1944 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947
they do exhibit. As a result, synaptic modification will
greater reflect the associations mediated by these recurrent
intrinsic/feedback fiber synapses. This situation may be
problematic if multiple neural representations are active si-
multaneously because spurious associations may be formed
via the intrinsic/feedback projections, potentially corrupt-
ing representations in the long term (see [25] for a detailed
discussion).
These neural mechanisms appear promising in explaining
both general and detailed aspects of E.B.’s perseverative be-
havior. Under a cholinergic deficit, the normal suppression
of intrinsic and feedback projections is removed and cells are
somewhat less excitable overall. One potential impact of this
is that neural activity will sustain itself for longer through
the undamped intrinsic/feedback projections. This will re-
quire that new stimuli override persistent activity at even
longer delays. Perseverations will be more likely, particu-
larly to stimuli with low frequency names, because afferent
input contributes proportionately less to processing, making
it harder to override sustained activity. Many of these perse-
verations will occur consecutively after previous responses
because representations are simply remaining active. Indeed,
these responses make up half of all of E.B.’s perseverations.
Another potential impact of a cholinergic deficit is that
normal learning mechanisms that are mediated by synaptic
plasticity will be somewhat disrupted. While plasticity ef-
fects will be reduced overall when acetylcholine levels are
diminished, some degree of synaptic modification will occur
at excitatory projections each time a stimulus is presented,
influencing processing on a longer time scale. When a stim-
ulus is repeated as in our experiments with E.B., persevera-
tions can become more likely because intrinsic and feedback
connections will be strengthened with each repetition. This
will lead activity to sustain itself more effectively, making
it more difficult for afferent sensory input to override it. As
sensory inputs arrive and start to override sustained activity,
inappropriate associations between different co-activated
neural representations may also be formed though synap-
tic plasticity because intrinsic/feedback connections are no
longer suppressed. When a stimulus activates one of these
neural representations again at a later time, activity may
inappropriately re-activate the associated representation
through the intrinsic/feedback connections. This could help
to explain E.B.’s delayed perseverations, as well as the
phenomenon revealed by Error Analysis 2 showing that de-
layed perseverations reflect the earlier sequential proximity
of targets and responses. Finally, the finding in Error Anal-
ysis 3 that delayed perseverations become less likely with
more intervening trials might reflect an interference effect
in synaptic modification and learning. While repeating the
same stimulus or sequence of stimuli may lead to similar
synaptic changes for each repetition, processing different
stimuli on each trial may reverse or overwrite these changes
to a certain extent. Such interference effects are well-known
in distributed neural network models of learning [39,41].
These effects would not necessarily interact with presenta-
tion rate because the changes depend on the stimulus pro-
cessing that occurs with each trial rather than on time, per se.
Similar effects have been observed in list-learning contexts
in which greater active rehearsal of stimuli during spaced
practice can potentially aid performance [16]. We view the
alternative possibility of active rehearsal during long RSIs
as unlikely in the current context because it would require a
great deal of effort and would not benefit E.B.’s performance
in the naming task.
It is important to point out that while we have focused
discussion on how cholinergic deficits (or deafferentation)
might account for E.B.’s recurrent perseverations, signif-
icant effects in omissions and other errors were also ob-
served. Although these errors did not always change along
with perseverations, it is possible that the mechanisms
that underlie the different error types are partially shared.
For example, one might view an omission as reflecting an
elicited pattern of neural activity that is too different from
any known state to support a response (see [56] for a similar
view). Under a cholinergic deficit, sensory input might par-
tially overcome persistent activity, yielding a novel pattern
of activity that either produces no response (an omission)
or a response blend (a neologism or other error). Activity
patterns that are distorted by neurological damage may also
be close enough to a known pattern to generate a simi-
lar, yet incorrect response (other errors such as semantic,
phonological or visual). Future studies will need to evaluate
the relationship between different error types more explic-
itly across a range of stimulus factors and with a variety of
patients who either perseverate or do not.
It is also important to point out that the cholinergic
deficit hypothesis of recurrent perseveration is not irrecon-
cilable with priming theories of perseveration. Indeed, the
explanatory principles involved are essentially the same,
namely residual activity, longer-term learning, and dimin-
ished sensory input. However, it does introduce important
biologically-based constraints on the form that processing
and learning take, constraints not present in current prim-
ing theories. While we view this approach as promising in
accounting for E.B.’s perseverative behavior, particularly
in its ability to address puzzling aspects such as the high
number of unrelated and delayed perseverations, much
work remains to be done. Other patients who exhibit recur-
rent perseverations need to be tested with these and other
stimulus factors in order to establish better generality of the
findings. Further studies are also needed in order to estab-
lish stronger links between particular neurochemical deficits
and changes in the incidence of perseverations, as well as
changes in the magnitudes of priming effects. Along these
lines, it is interesting to note that norepinephrine and acetyl-
choline have similar cellular actions (see [26] for a review).
Noradrenergic and cholinergic deficits might, therefore, be
expected to elicit similar behavioral consequences. Given
the complexity of the data and candidate mechanisms, it will
also be important to explore these issues more thoroughly
in the context of explicit computational models.
 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947 1945

Acknowledgements References

We wish to thank Dr. Schott for permission to study [1] Albert ML, Sandson J. Perseveration in aphasia. Cortex 1986;22:
E.B., a patient under his care, and for providing us with 103–15.
[2] Allison R, Hurwitz L. On perseveration in aphasics. Brain
facilities to carry out our investigation. Many thanks to 1967;90:429–48.
Peter Rudge, Tim Shallice, David Plaut, Bobby Klatzky, [3] Becker CA. Semantic context and word frequency effects in visual
Craig Haimson and Lisa Gershkoff-Stowe for useful dis- word recognition. Journal of Experimental Psychology: Human
cussions and/or helpful comments on earlier drafts of the Perception and Performance 1979;5:252–9.
[4] Bishop DVM. Test for the reception of grammar (TROG). Abingdon:
paper. S.J.G. was supported by an NSF Graduate Research Thomas Leach, 1982.
Fellowship. [5] Brocher S, Artola A, Singer W. Agonists of cholinergic and
noradrenergic receptors facilitate synergistically the induction of
long-term potentiation in slices of rat visual cortex. Brain Research
Appendix A. 1992;573:27–36.
[6] Buckingham HW, Whitaker H, Whitaker HA. On linguistic
perseveration. In: Whitaker HA, Whitaker H, editors. Studies of
Perseverative errors which appeared to be due to the rapid neurolinguistics, vol. 4. New York: Academic Press, 1979. p. 329–52.
formation of associations between sequentially presented [7] Butterworth B. Hesitation and the production of verbal paraphasias
stimuli/responses. Perseverations of interest, along with their and neologisms in jargon aphasia. Brain and Language 1979;8:
133–61.
presumed corresponding previous stimuli are indicated by [8] Carroll JB, Davies P, Richman B. Word frequency book. Boston:
“−>”. Houghton Mifflin, 1971.
[9] Cipolotti L. Sparing of country and nationality names in a case of
modality specific oral output impairment: implications for theories
of speech production. Cognitive Neuropsychology 2000;17:709–29.
Trial Stimulus Response [10] Cipolotti L, Warrington EK. Towards a unitary account of access
dysphasia: a single case study. Memory 1995;3:309–32.
Example 1 (Experiment 1) [11] Coughlan AK, Warrington EK. The impairment of verbal semantic
1 Figs “A fig” memory: a single case study. Journal of Neurology, Neurosurgery
and Psychiatry 1981;44:1079–83.
2 Rhubarb “Rhubarb” [12] Cohen L, Dehaene S. Competition between past and present:
−>3 Olives “Fig” assessment and interpretation of verbal perseverations. Brain 1998;
−>4 Dates “A fri-” 121:1641–59.
5 Rhubarb “Rhubarb” [13] Dell GS, Schwartz MF, Martin N, Saffran EM, Gagnon DA. Lexical
6 Olives No response access in aphasic and non-aphasic speakers. Psychological Review
1997;104:801–38.
7 Dates −>“Fig” [14] Denes G, Dalla Barba G, Cipolotti L, Semenza C. An unusual case
8 Figs “A pig . . . no” of perseveration sparing body-related tasks. Cortex 1990;26:269–73.
[15] Dunn LM, Dunn LM, Whetton C, Pintilie D. British picture
Example 2 (Experiment 1) vocabulary scale. Windsor: NFER-Nelson, 1982.
−>1 Rabbit “Hare” [16] Ebbinghaus, H. Memory: a contribution to experimental psychology.
−>2 Donkey “Donkey” New York: Teachers College, Columbia University, 1885 [translated
by Ruger HA, Bussenues CE, 1913].
3 Cow “Cow” [17] Ellis AW, Marshall JC. Semantic errors or statistical flukes? A note
4 Lamb No response on Allport’s on knowing the meaning of words we are unable
5 Cow “Cows” to report. Quarterly Journal of Experimental Psychology 1978;30:
6 Rabbit −>“Donkey” 569–75.
7 Lamb No response [18] Forster KI, Chambers SM. Lexical access and naming time. Journal
of Verbal Learning and Verbal Behavior 1973;12:627–35.
8 Donkey “D-” [19] Frith CD, Friston K, Liddle PF, Frackowiak RS. Willed action and the
prefrontal cortex in man: a study with PET. Proceedings of the Royal
Example 3 (Experiment 2) Society of London Series B Biological Sciences 1991;244:241–6.
−>1 Helicopter “Helicopter” [20] Fuld PA, Katzman R, Davies P, Terry RD. Intrusions as a sign of
−>2 Ashtray “Ashtray” Alzheimer’s dementia: chemical and pathological verification. Annals
3 Dress “Dress” of Neurology 1982;11:155–9.
[21] Gil Z, Connors BW, Amitai Y. Differential regulation of neocortical
4 Brush “Hairbrush”
synapses by neuromodulators and activity. Neuron 1997;19:679–86.
5 Chair “Stool” [22] Goodglass H, Kaplan E. Boston diagnostic aphasia examination
6 Glass “Glass” (BDAE). Philadelphia: Lea & Febiger, 1986.
7 Glass “Glass” [23] Gough PB, Alford JA, Holley-Wilcox P. Words and contexts. In:
8 Brush “Hairbrush” Tzeng OJL, Singer H, editors. Perception of print: reading research in
experimental psychology. Hillsdale, NJ: Erlbaum, 1981. p. 85–102.
9 Dress “Dress” [24] Halpern H. Effect of stimulus variables on verbal perseveration of
10 Helicopter −>“Ashtray . . . no” aphasic subjects. Perceptual and Motor Skills 1965;20:421–9.
11 Ashtray “Ashtray” [25] Hasselmo ME. Runaway synaptic modification in models of cortex:
12 Chair “School” implications for Alzheimer’s disease. Neural Networks 1994;7:
13–40.
1946 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947
[26] Hasselmo ME. Neuromodulation and cortical function: modeling the
physiology basis of behavior. Behavioural Brain Research 1995;67:
1–27.
[27] Hasselmo ME, Anderson BP, Bower JM. Cholinergic modulation of
cortical associative memory function. Journal of Neurophysiology
1992;67:1230–46.
[28] Hasselmo ME, Bower JM. Cholinergic suppression specific to
intrinsic not afferent fiber synapses in rat piriform (olfactory) cortex.
Journal of Neurophysiology 1992;67:1222–9.
[29] Hasselmo ME, Bower JM. Acetylcholine and memory. Trends in
Neuroscience 1993;16:218–22.
[30] Hirsh KW. Perseveration and activation in aphasic speech production.
Cognitive Neuropsychology 1998;15:377–88.
[31] Hudson AJ. Perseveration. Brain 1968;91:571–82.
[32] Huerta PT, Lisman JE. Heightened synaptic plasticity of hippocampal
CA1 neurons during a cholinergically induced rhythmic state. Nature
1993;364:723–5.
[33] Kohn SE. The nature of the phonological disorder in conduction
aphasia. Brain and Language 1984;23:97–115.
[34] Lambon Ralph MA. Distributed versus localist representations:
evidence from a study of item consistency in a case of classical
anomia. Brain and Language 1998;64:339–60.
[35] Lecours AR, Lhermitte F. Phonemic paraphasias: linguistic structures
and tentative hypotheses. Cortex 1969;5:193–228.
[36] Lhermitte F, Beauvois MF. A visual–speech disconnection syndrome:
report of a case with optic aphasia, agnosic alexia and color agnosia.
Brain 1973;96:695–714.
[37] Martin N, Roach A, Brecher A, Lowery J. Lexical retrieval
mechanisms underlying whole-word perseveration errors in anomic
aphasia. Aphasiology 1998;12:319–33.
[38] McCarthy RA, Kartsounis LD. Wobbly words: refractory anomia
with preserved semantics. Neurocase 2000;6:487–97.
[39] McClelland JL, McNaughton BL, O’Reilly RC. Why there are
complementary learning systems in the hippocampus and neocortex:
insights from the successes and failures of connectionist models of
learning and memory. Psychological Review 1995;102:419–57.
[40] McClelland JL, Rumelhart DE. An interactive activation model of
context effects in letter perception. Part I. An account of basic
findings. Psychological Review 1981;88:375–407.
[41] McCloskey M, Cohen NJ. Catastrophic interference in connectionist
networks: the sequential learning problem. In: Bower GH, editor.
The psychology of learning and motivation, vol. 24. New York:
Academic Press, 1989. p. 109–65.
[42] McKenna P, Warrington EK. Category specific naming preservation.
Journal of Neurology, Neurosurgery and Psychiatry 1978;41:571–4.
[43] McKenna P, Warrington EK. Testing for nominal dysphasia. Journal
of Neurology, Neurosurgery and Psychiatry 1980;43:781–8.
[44] McKone E. The decay of short-term implicit memory: unpacking
lag. Memory and Cognition 1998;26:1173–86.
[45] Meyer DE, Schvaneveldt RW, Ruddy MG. Activation of lexical
memory. In: Proceedings of the Presentation of Paper at the 13th
Annual Meeting of the Psychonomic Society, St. Louis, MO,
November 1972.
[46] Meyer DE, Schvaneveldt RW, Ruddy MG. Loci of contextual effects
on visual word recognition. In: Rabbitt PMA, Dornic S, editors.
Attention and performance V. New York: Academic Press, 1975.
[47] Meyer DE, Schvaneveldt RW. Facilitation in recognizing pairs of
words: evidence of a dependence between retrieval operations.
Journal of Experimental Psychology 1971;90:227–34.
[48] Monsell S, Doyle MC, Haggard PN. Effects of frequency of visual
word recognition tasks: where are they? Journal of Experimental
Psychology: General 1989;118:43–71.
[49] Neely JH. Semantic priming effects in visual word recognition: a
selective review of current findings and theories. In: Besner D,
Humphreys GW, editors. Basic processes in reading: visual word
recognition. Hillsdale, NJ: Erlbaum, 1991. p. 264–336.
[50] Neter J, Kutner NH, Nachtsheim CJ, Wasserman W. Applied linear
regression models, 4th ed. Chicago, IL: Irwin, 1996.
[51] Oldfield RC, Wingfield AR. Response latencies in naming objects.
Quarterly Journal of Experimental Psychology A 1965;17:273–81.
[52] Ostergaard AL. The effects on priming of word frequency, number of
repetitions and delay depend on the magnitude of priming. Memory
and Cognition 1998;26:40–60.
[53] Papagno C, Basso A. Perseveration in two aphasic patients. Cortex
1996;32:67–82.
[54] Peters BH, Levin HS. Memory enhancement after physostigmine
treatment in the amnesic syndrome. Archives of Neurology
1977;34:215–9.
[55] Plaut DC, McClelland JL, Seidenberg MS, Patterson KE. Under-
standing normal and impaired word reading: computational principles
in quasi-regular domains. Psychological Review 1996;103:56–115.
[56] Plaut DC, Shallice T. Perseverative and semantic influences on visual
object naming errors in optic aphasia: a connectionist account. Journal
of Cognitive Neuroscience 1993;5:89–117.
[57] Price CJ. The functional anatomy of word comprehension and
production. Trends in Cognitive Science 1998;2:281–8.
[58] Pugh KR, Shaywitz BA, Shaywitz SE, Constable RT, Skudlarski P,
Fulbright RK, et al. Cerebral organization of component processes
in reading. Brain 1996;119:1221–38.
[59] Rapp BC, Caramazza A. On the distinction between deficits of
access and deficits of storage: a question of theory. Cognitive
Neuropsychology 1993;10:113–41.
[60] Raven JC. Colored progressive matrices. London: Lewis, 1956.
[61] Robinson G, Blair J, Cipolotti L. Dynamic aphasia: an inability to
select between competing verbal responses? Brain 1998;121:77–89.
[62] Sandson J, Albert ML. Varieties of perseveration. Neuropsychologia
1984;22:715–32.
[63] Sandson J, Albert ML. Perseveration in behavioral neurology.
Neurology 1987;37:1736–41.
[64] Santo Pietro MJ, Rigrodsky S. The effects of temporal and
semantic conditions on the occurrence of the error response of
perseveration in adult aphasics. Journal of Speech and Hearing
Research 1982;25:184–92.
[65] Santo Pietro MJ, Rigrodsky S. Patterns of oral–verbal perseveration
in adult aphasics. Brain and Language 1986;29:1–17.
[66] Scarborough DL, Cortese C, Scarborough HS. Frequency and
repetition effects in lexical memory. Journal of Experimental
Psychology: Human Perception and Performance 1977;3:1–17.
[67] Schwartz MF, Saffran EM, Bloch D, Dell GS. Disordered speech
production in aphasic and normal speakers. Brain and Language
1994;47:52–88.
[68] Selden NR, Gitelman DR, Salamon-Murayama N, Parrish TB,
Mesulam M-M. Trajectories of cholinergic pathways within the
cerebral hemispheres of the human brain. Brain 1998;121:2249–57.
[69] Shallice T. From neuropsychology to mental structure. Cambridge:
Cambridge University Press, 1988.
[70] Smith CM, Swash M. Effects of cholinergic drugs on memory in
patients with Alzheimer’s disease. In: Proceedings of the Presentation
of Paper at the International Society for Neurochemistry Satellite
Meeting on Aging of the Brain and Dementia, Florence, Italy, August
1979.
[71] Tulving E, Schacter DL. Priming and human memory effects. Science
1990;247:301–6.
[72] Vitkovitch M, Humphreys GW. Perseverant responding in speeded
picture naming: it’s in the links. Journal of Experimental Psychology:
Learning, Memory and Cognition 1991;17:664–80.
[73] Vitkovitch M, Humphreys GW, Lloyd-Jones TJ. On naming a
giraffe a zebra: picture naming errors across different object
categories. Journal of Experimental Psychology: Learning, Memory
and Cognition 1993;19:243–59.
[74] Warrington EK. The selective impairment of semantic memory.
Quarterly Journal of Experimental Psychology 1975;27:635–57.
[75] Warrington EK. The Queen Square screening test. Queen Square,
London: Institute of Neurology, 1989.
[76] Warrington EK. The Camden memory tests. Hove, East Sussex:
Psychology Press, 1996.
 S.J. Gotts et al. / Neuropsychologia 40 (2002) 1930–1947 1947

[77] Warrington EK, Cipolotti L. Word comprehension: the distinction
between refractory and storage impairments. Brain 1996;119:611–25.
[78] Warrington EK, McCarthy RA. Category specific access dysphasia.
Brain 1983;106:859–78.
[79] Warrington EK, McCarthy RA. Categories of knowledge: further
fractionations and an attempted integration. Brain 1987;110:1273–96.
[80] Warrington EK, Shallice T. Category specific semantic impairments.
Brain 1984;107:829–54.
[81] Wiggs CL, Martin A, Sunderland T. Monitoring frequency
of occurrence without awareness: evidence from patients with
Alzheimer’s disease. Journal of Clinical and Experimental Neuro-
science 1997;13:235–44.