INTRODUCTION
Diverse neurocognitive deficits affect approximately fifty percent of
individuals with multiple sclerosis (MS) (DeSousa et
al., 2002; Rao et al., 1991). Long-term
memory (LTM) impairment is one of the most consistently reported findings
in MS (Beatty, 1993; Beatty
et al., 1989; Bobholtz & Rao, 2003;
DeLuca et al., 1998; Rao,
1986; for a review see Thornton & Raz,
1997), and appears to be an ecologically valid indicator of daily
functioning (Higginson et al., 2000). Over the
past decade a clearer understanding of LTM function in MS has emerged.
Initially, LTM deficits in MS were attributed to a selective retrieval
failure (e.g., Armstrong et al., 1996; Rao et
al., 1989, 1993), but
recent evidence has also implicated acquisition and encoding processes
(DeLuca et al., 1994, 1998; Demaree et al., 2000;
Gaudino et al., 2001). Variability in the
observed magnitude of LTM impairment (Thornton &
Raz, 1997) may reflect the diversity of memory operations that have
been investigated (e.g., Thornton et al., 2002),
as well as disease-related variables (Arnett et al., 1999a, 1999b; Kessler et al., 1992; Thornton
& Raz, 1997).
To further characterize MS-related memory functioning, the present
study evaluates the contributions of interference to LTM impairment. The
normative pattern of loss of information from the memory store (Brown, 1958; Peterson &
Peterson, 1959) often reflects interference between previously
learned and new information (Keppel & Underwood,
1962). Proactive interference (PI) occurs when prior learning
impedes retention of subsequent materials, and retroactive interference
(RI) refers to the decrement in retention of previously learned
information caused by subsequent learning (Postman
& Underwood, 1973). Interference is heightened when old and new
information is taxonomically related. During encoding, interference
between related words is assumed to reflect active use of the
information's semantic properties (Wickens,
1970).
One paradigm used to elicit PI presents lists of words from the same
semantic category, for one learning trial each. This is followed by a
final trial of words drawn from a new semantic category. Performance on
this Wickens (1970) type paradigm declines
linearly over learning trials as interference between semantically related
items increases. Recall rebounds (“release” from PI) when the
new semantic category is introduced. In clinical samples, intensified
accumulation (but relatively preserved release) of PI has been associated
with diminished working memory (WM) capacity (e.g., Blusewicz et al., 1996), while normal accumulation and
poor release from PI has been attributed to concurrent mnemonic and
executive deficits (Cermak et al., 1974; Randolph et al., 1992; Squire,
1982).
Despite high lesion burden in frontal regions (Sperling et al., 2001) and impairment on both
“frontal” tasks of WM (Pelosi et al.,
1997) and executive abilities (Foong et al.,
1997), patients with MS have shown normal accumulation and release
from PI on Wickens-type paradigms (Beatty et al.,
1989; Johnson et al., 1998; Rao et al., 1993). However, some investigators suggest
that this paradigm may be insensitive to subtle interference abnormalities
(Dobbs et al., 1989). Given that divergence
between MS patients and controls is most evident under cognitive challenge
(D'Esposito et al., 1996; Legenfelder et al., 2003), the potential insensitivity
of the Wickens paradigm might obscure subtle group differences in
susceptibility to interference. The California Verbal Learning Test (CVLT;
Delis et al., 1987) has also been used to
evaluate interference (Kramer & Delis,
1991). The CVLT differs from Wickens-type paradigms in that
accumulation and release from PI occur simultaneously. Specifically, List
B of the CVLT concurrently presents items from the same semantic
categories as List A, inducing PI, and items from new semantic categories,
inducing release from PI. Compared to Wickens-type paradigms, the
simultaneous processing inherent to the CVLT may increase its
sensitivity.
To investigate this possibility, the terms Wickens, interference,
PI, release from PI, clinical, verbal memory, memory impairment, and
CVLT were entered into PsychInfo and Medline. Studies were
retrieved if they included a Wickens-type paradigm with repeated
presentations of a supraspan word list and a final release trial, or if
CVLT interference effects were calculated using some comparison of shared
and unshared items.a
Those studies using a
traditional Brown-Peterson short-term memory task with presentation of
groups of three or four words followed by a counting or sequencing
distracter were not evaluated, as interference between semantically
related words was not the primary focus of those studies.
Table 1 shows that abnormal interference
effects on Wickens-type paradigms were reported in some, but not all,
clinical samples with expected memory dysfunction. Specifically, patients
with Parkinson's disease, traumatic brain injury, and MS did not
differ from healthy controls in susceptibility to or release from PI.
While studies have not systematically examined inference effects on the
CVLT in MS, abnormal PI effects appear widespread on the CVLT in other
patient groups. This pattern suggests that the CVLT may be more sensitive
than Wickens-type paradigms to these effects.
Selected studies of proactive interference effects
Although increased susceptibility to PI has not been detected in MS
using Wickens-type paradigms, susceptibility to PI appears to be an
important determinant of verbal WM in healthy individuals (Lustig et al., 2001). Furthermore, as some evidence
indicates that WM capacity contributes to LTM dysfunction in MS (e.g.,
DeLuca et al., 1994; Litvan
et al., 1988; Thornton et al., 2002), it
may be fruitful to directly evaluate MS-related interference effects and
their relationship to LTM with a more sensitive measure, such as the
CVLT.
An interesting parallel line of inquiry is whether MS participants
experience full recall recovery when interference ceases (i.e., preserved
release from PI). Individuals with MS appear more likely to generally
encode semantic information, and their memory lacks both semantic (Carroll et al., 1984) and contextual (Thornton et al., 2002) distinctiveness. If the loss of
semantic distinctiveness is robust, it should generalize across memory
paradigms. Consequently, MS participants might show diminished sensitivity
to taxonomic shifts under demanding conditions, as evidenced by reduced
release from PI. Finally, while individuals with MS have shown RI on the
CVLT (Kessler et al., 1992; Troyer et al., 1996), these studies did not include
control groups, precluding the evaluation of whether those with MS were
disproportionately susceptible to this type of interference.
Based on the aforementioned rationale, we investigated whether
individuals with MS exhibit differential sensitivity to the semantic
properties of stimulus items and abnormal interference effects on the
CVLT. Furthermore, we assessed the extent to which interference predicts
delayed memory in MS. This later issue is interesting in light of the
relationship between susceptibility to interference on the CVLT and LTM
dysfunction reported in other clinical samples (Vanderploeg et al., 1994). Given MS-related losses in
semantic and contextual distinctiveness (e.g. Carroll
et al., 1984; Thornton et al., 2002) and
WM deficits (e.g. Litvan et al., 1988),
individuals with MS might be particularly susceptible to PI and RI on the
CVLT, a vulnerability that would negatively affect delayed memory.
METHODS
Research Participants
All participants were fully informed of the nature of the study and
consented to participate. Eighty-three participants with MS were recruited
from a local registry of the National Multiple Sclerosis Society, and had
received a diagnosis of multiple sclerosis from their physicians.
Participants were recruited from the community, not from medical
facilities or treatment programs. Exclusion criteria included: age over 60
years, self-reported medical or psychiatric disorders other than MS that
would be expected to affect memory (e.g., active substance abuse, seizure
disorders, head injury), and sensory deficits precluding participation in
the assessment. Multiple sclerosis participants were well matched to a
sample of 80 healthy controls who had volunteered from community sites and
a university participant pool (see Table 2).
Previously, we reported on a subsample of the current participants (Thornton et al., 2002).
Participants with MS completed a questionnaire addressing disease
onset, course, and severity. Seventy-two percent of participants reported
a diagnosis of definite MS from their physician, with fewer individuals
reporting diagnoses of probable (18%) or possible (10%) MS. Those
participants reporting possible MS described at least one discrete symptom
exacerbation, as well as neurological signs consistent with MS.
Participants were classified on the basis of self-reported symptom
onset and course in the following manner: 27% reported a sudden onset of
symptoms with subsequent asymptomatic remission, 46% reported a
fluctuating course with periods of remission, and 27% described a slow
worsening of symptoms. Furthermore, most participants (70%) had not
experienced symptom exacerbation within the past month. Using a
self-report rating scale containing items comparable to those on the
Expanded Disability Status Scale (EDSS; Kurtzke,
1965), MS participants also rated their own level of physical
disability (EDSS equivalent x = 3.1, SD = 2.2). This
self-rating instrument has been associated (r = .82, p
< .001) with experimenter ratings of ambulation (Thornton et al., 2002), and other researchers have
reported strong correlations between self-ratings and neurological exams
in MS patients (Solari et al., 1993; Verdier-Taillefer et al., 1994).
Lastly, participants completed the Beck Depression Inventory (BDI,
Beck et al., 1961), a 21-item self-report
questionnaire of a broad range of depressive symptoms (see Table 2). This measure was scored using the
traditional method and a “partial” scoring technique
eliminating items directly assessing physical functioning and appearance
(items 14, 15, 17, 20, & 21, see Thornton et al.,
2002). This modification addresses the overlap between MS symptoms
and the somatic features of depression tapped by the BDI (consistent with
the recommendations of Minden & Schiffer,
1990). Using traditional BDI scores, there was a trend towards
higher levels of depressive symptoms in those reporting current
exacerbation of neurological symptoms [F(3,67) = 2.4,
p = .08], but when partial BDI scores were used this
difference was nonsignificant [F(3,67) = 1.9, p =
ns]. No other differences in traditional or partial BDI
scores were found for self-reported MS subtype, disease onset and course,
or physical disability level (all ps = ns).
Procedures
The California Verbal Learning Test (CVLT; Delis et
al., 1987) was administered as part of larger neuropsychological
batteries. The CVLT is a verbal list-learning test in which a series of 16
shopping items (List A) is repeated over five learning trials.
Participants are directed to recall as many items as they can after each
trial. After the learning trials, a new 16-item shopping list (List B) is
recited once, and participants are required to recall as many of the new
items as they can remember. Then, without repetition of the list by the
administrator, participants are directed to recall items from List A
(Short Delay Free Recall; SDFR). This is followed by a semantic cueing
trial (e.g., “Tell me the items from the first list that were herbs
and spices”). Following a 20-minute delay, free and cued recall
trials are administered again. Next, participants are required to
discriminate List A items (“yes”) from List B items and other
distracters (“no”) presented in a recognition format.
Interference
Interference measures were calculated based on the procedure outlined
by Kramer and Delis (1991). One-half of the
items on both List A and List B are “shared” (i.e., they are
members of the same semantic categories—fruits and
spices/herbs), while the other half of the items are
“unshared” (they are members of distinct semantic
categories—tools and clothing for List A, fish and kitchen utensils
for List B). Recall of shared and unshared items on the first trial of
List A (List A-1) is compared with recall of shared and unshared items on
List B. Reduced recall of shared items on List B relative to shared item
recall of List A-1 reflects accumulation of PI, while improved recall of
unshared items on List B relative to unshared items on List A-1 indicates
release from PI. Retroactive interference is reflected by decreased
post-interference recall (SDFR) relative to Trial 5 of List A, with
greater declines expected for shared than unshared item recall.
In accordance with Kramer and Delis (1991),
weighted averages of shared and unshared words recalled on Trial 1 were
calculated for each participant in order to control for proportion of
shared words recalled across trials and corresponding differences in the
opportunity to accumulate interference (weighted average of shared
category items = number of items recalled on Trial 1 × [total
number of shared category items recalled on Trials 1–5 ÷
total number of words recalled on Trials 1–5]). Similarly, in
assessing RI, SDFR of shared and unshared items were compared to weighted
averages of List A-5 recall to control for the relative proportion of
shared and unshared items learned.
Individual participant interference scores
To evaluate individual differences in interference effects and the
relationship between these effects and LTM, we created individual
difference scores. The importance of controlling for baseline performance
to avoid possible confounds when calculating trial-to-trial difference
scores to represent interference effects has been previously noted (Torres et al., 2001). Given that the items on the CVLT
are of equivalent recall difficulty, the average of shared and unshared
items recalled on the baseline trial (List A-1 for PI and List A-5 for RI)
provides the most stable estimate of how many shared or unshared items
each individual should be able to recall on any trial. This
estimate can be compared with actual shared or unshared item recall on
Trial B and SDFR for PI and RI, respectively. We therefore calculated
difference scores in the following manner:
These calculations reflect relative rather than absolute
interference-related changes in recall. The rationale for using relative
interference scores is derived from the need to account for general LTM
ability. Absolute (raw) interference difference scores have the inherent
disadvantage of being strongly influenced by baseline recall. For example,
participants with higher List A-1 recall will produce greater raw
interference scores merely because of their strong initial recall. In
contrast, individuals with poor List A-1 recall truncate the extent to
which their recall will decline when under the influence of interference.
Consequently, relative interference scores accounting for initial recall
level were constructed in the same manner as retention scores, which have
been used as an index of forgetting in other research paradigms.
Learning strategies
As the CVLT possesses a latent semantic structure, we assessed the
recognition and use of this structure with the semantic clustering index.
As increased recall itself results in increased clustering, the index was
calculated according to an observed ÷ expected ratio as described
in the manual (see Delis et al., 1987). We also
examined other aspects of learning, including rates of perseverations and
intrusions, recall consistency, and recognition discriminability to
further characterize this sample.
RESULTS
Overall Verbal Memory
Multivariate analysis of variance (MANOVA) indicated that MS
participants recalled significantly fewer items across all five learning
trials [F(1,159) = 16.3, p < .001], at
short-delay free recall [F(1,159) = 18.9, p <
.001] and cued recall [F(1,159) = 17.6, p <
.001], and at long-delay free recall [F(1,159) = 12.2,
p ≤ .001] and cued recall [F(1,159) = 18.7,
p < .001]. Although there were no group differences in
semantic and serial clustering, participants with MS showed reduced middle
region recall [F(1,159) = 5.5, p < .05] and
lower recall consistency [F(1,159) = 21.5, p <
.001]. In addition, MS participants generated more perseverations
during free and cued recall trials [F(1,159) = 4.4,
p < .05] and more cued recall intrusions
[F(1,159) = 6.3, p < .05]. Lastly,
recognition discriminability was lower for MS participants
[F(1,159) = 7.9, p < .01].
Interference Analyses
For each interference analysis presented next, data was evaluated for
distributional normality. In the PI interference data, this inspection
revealed the presence of seven participants (1 MS and 6 controls) who
generated scores in one or more of the recall trials that were either
outlying or extreme (at least two SDs from the mean). As the
scores of these individuals approached either the floor or ceiling of each
recall condition, their inclusion would constrain recall associated with
PI. In addition, their removal improved the distributional properties of
List B recall, reducing skew by approximately half a standard deviation
with minimal effect on kurtosis. Consequently, the seven participants were
removed from the PI analyses reported later.
In the RI data, recall scores of one MS and three control participants
were identified as outlying or extreme. (One of these controls had also
been eliminated from the PI analyses). Removal of these individuals
improved the distribution of short-delay recall scores, reducing skew by
half a standard deviation with minimal effect on kurtosis. These four
outliers were therefore removed from the RI analyses reported later, as
well as an additional control participant who was missing SDFR data.b
Although elimination is not an uncommon
treatment for outliers, numerous authors (e.g., Allison
& Gorman, 1993) suggest running analyses with and without
outliers to evaluate their effect on the results. With the exception of
stabilizing a trend towards greater susceptibility to RI in our MS
participants, the removal of these outliers did not affect the direction
or significance of our results for PI and RI analyses.
Proactive interference effects
Proactive interference effects were analyzed using a 2 × 2
× 2 mixed factorial analysis of variance (ANOVA), with List (List
A-1 or List B) and category (shared or unshared) as within-subject
variables, and group (MS or controls) as the between-subjects factor. A
significant main effect for group [F(1,154) = 6.0,
p < .05] reflected superior overall item recall in the
control group. Both groups remembered fewer items on List B than List A-1
[F(1,154) = 21.1, p < .001], and all
participants demonstrated accumulation of PI and release from PI
[F(1,154) = 268.7, p < .001]. More
importantly, the analyses revealed a List × Category × Group
interaction [F(1,154) = 4.1, p < .05]
indicating differential recall patterns across the PI factors for
participant groups. The divergent PI pattern was decomposed by examining
group differences in (1) recall of shared items in List A-1
versus List B, (2) recall of unshared items in List A-1
versus List B, as well as (3) recall of shared
versus unshared items on List A-1, and (4) recall of
shared versus unshared items on List B, using
the original mixed factorial model.
In terms of accumulation of PI across the Lists, both groups showed a
similar recall decrement for shared items from List A-1 to List B
[F(1,154) = 215.5, p < .001], that did not
interact with group [F(1,154) = 0.3, p =
ns]. Both groups also made significant gains in recall of
unshared items from List A-1 to List B [F(1,154) = 54.0,
p < .001], demonstrating release from PI. Interestingly,
this result was moderated by a significant List × Group interaction
[F(1,154) = 4.7, p < .05]. Figure 1 shows that, relative to control participants,
MS participants exhibited attenuated gains in their recall of unshared
items from List A-1 to List B.
Interference effects in the full multiple sclerosis (MS) sample. (PI =
proactive interference; RI = retroactive interference; SDFR = short-delay
free recall).
The group differences in PI effects were further defined by the number
of shared and unshared items recalled on each list. On the initial List
A-1 trial, unshared items were more readily recalled than shared items
[F(1,154) = 8.5, p < .01]; this pattern did
not vary with Group [F(1,154) = .7, p =
ns]. The superior recall of unshared items persisted through
List B, becoming even more compelling [F(1,154) = 318.4,
p < .001]. In terms of differential release from PI, the
List B data revealed that control participants showed greater facility in
the recall of unshared relative to shared items than did MS participants
[F(1,154) = 5.6, p < .05]. Indeed, the List
B data of Figure 1 shows that control
participant recall was superior only for the unshared items
[F(1,154) = 9.9, p < .01], and not for the
shared items [F(1,154) = .5, p = ns].
This differential pattern of shared versus unshared recall
provides further evidence for attenuated release from PI in MS
participants.
Retroactive interference effects
Retroactive interference effects were analyzed using a 2 × 2
× 2 mixed factorial ANOVA, with List (List A-5 or SDFR) and category
(shared or unshared) as within-subject variables, and group (MS or
controls) as the between-subjects factor. A significant main effect for
group [F(1,156) = 24.6, p < .001] indicated
diminished recall on both trials for MS participants. The List ×
Category interaction [F(1,156) = 45.7, p <
.001] reflected the presence of RI effects for both groups. More
importantly, a List × Category × Group interaction
[F(1,156) = 4.6, p < .05] showed
differential recall patterns across RI factors between participant groups.
This interaction was decomposed by evaluating group differences in (1)
recall of shared items in List A-5 versus SDFR, (2)
recall of unshared items in List A-5 versus SDFR, as
well as (3) recall of shared versus unshared
items on List A-5, and (4) recall of shared versus
unshared items on SDFR, using the original mixed factorial
model.
A substantial decrement in recall of shared items from List A-5 to
SDFR [F(1,156) = 120.1, p < .001] was
observed for all participants, but this result was moderated by a
significant List × Group interaction [F(1,156) = 8.2,
p < .01]. Figure 1 illustrates
that compared to control participants, those with MS experienced a more
precipitous decline in recall of shared items from List A-5 to SDFR. In
terms of unshared item recall, both groups showed a decline from List A-5
to SDFR [F(1,156) = 15.7, p < .001] that
did not differ between groups [F(1,156) = .3, p =
ns]. Lastly, unshared items were more readily recalled than
shared items on List A-5 [F(1,156) = 7.5, p <
.01] and at SDFR [F(1,156) = 66.2, p <
.001], a pattern that did not vary between groups for either List A-5
[F(1,156) = 1.4, p = ns] or SDFR
[F(1,156) = 1.6, p = ns].c
To ensure that our findings were robust to the
possible confounding effects of disease-related variables such as MS
subtype, comorbid depression, medication use, and recent symptom
exacerbation, the previous analyses were repeated in subsamples of
participants reporting they had received a diagnosis of Definite MS
(N = 59), those with Total BDI scores below 20 (N = 71),
those not currently taking anxiolytics (N = 73), and those
reporting symptom remission of at least one month (N = 57). A BDI
cutoff score of 20 was chosen to maximize sensitivity and specificity, in
accordance with the recommendations of Lykouras et al. (1998) for neurological patients. In all analyses, the
patterns of PI and RI effects were comparable with those reported in the
larger group (all ps < .05), with the exception of PI effects
in the symptom remission group, where interference scores were almost
identical to the larger sample, but reduced statistical power attenuated
significance of the between group interaction (p <
.10).
Interference scores and long-term memory
Interference difference scores were calculated according to the
methodology specified in the Procedures section. To evaluate the
relationship between these scores and LTM, linear regression was carried
out separately in MS participants and controls to identify predictors of
long-delay free recall (LDFR). Six independent variables were entered into
the model in the following order: block 1, demographic variables (age and
education); block 2, partial BDI score; block 3, interference indices
(accumulated PI, release from PI, and accumulated RI). Long-delay free
recall was the dependent variable in these analyses. As indicated in Table 3, age and partial BDI score, but none of the
interference indices, predicted LDFR in healthy participants. In contrast,
only accumulated RI predicted LDFR in the MS group.
Predictors of long-delay free recall (LDFR) in MS and control
groups
DISCUSSION
Consistent with past research (e.g., Bobholtz &
Rao, 2003; Rao, 1986), we found
diminished verbal long-term memory in MS participants relative to healthy
controls. Although prior studies have reported similar findings using the
CVLT (Kessler et al., 1992; Troyer et al., 1996), they have not described
corresponding patterns of interference effects, nor their contribution to
LTM. Our findings indicate that while individuals with MS appear no more
susceptible to accumulation of PI than healthy individuals, they
demonstrate attenuated release from PI and intensified RI. These results
appear to be germane to ultimate LTM, as accumulated RI predicted
long-delay free recall in people with MS, but not in healthy participants.
This divergence in interference-memory patterns suggests that there are
unique MS-related processing vulnerabilities that may mediate retention of
information. The nature of these vulnerabilities is detailed in the
following discussion of the pattern of PI and RI effects.
Retroactive Interference Effects
The current results indicate heightened susceptibility to RI in MS
when semantically associated material is processed in memory. An
intriguing subsequent observation is that RI contributes substantially to
the eventual recall of MS participants. After controlling for potential
confounding variables, vulnerability to RI accounted for approximately 37%
of the delayed recall variability of MS participants, but did not reliably
predict memory in healthy individuals. Apparently, individuals with MS
have specific difficulties retaining and retrieving old information after
acquiring new, related information. In aging, this deficit has been
attributed to compromised source monitoring (Hedden
& Park, 2003) and frontal lobe dysfunction (Glisky et al., 2001). Similarly, intensified RI in MS
may indicate a difficulty using source and contextual information to
facilitate retention and retrieval of information.
Although two prior investigations reported SDFR decrements (i.e., RI)
on the CVLT in MS samples (Kessler et al., 1992;
Troyer et al., 1996), neither study evaluated
shared and unshared recall patterns, nor conducted control group
comparisons. Another study reporting no differences between MS and control
participants in postinterference recall used procedures (i.e., RAVLT) that
did not manipulate the semantic relationships between items (Minden et al., 1990). Given that susceptibility to RI
should be strongest for shared item recall (Kramer
& Delis, 1991) these null findings may be artifactual. Indeed,
the current study demonstrated no group differences in the pattern of
unshared item recall across the final learning and postinterference
trials.
Release from Proactive Interference
Previous reports of normal accumulation and release from PI in MS
using Wickens-type paradigms (Beatty et al.,
1989; Johnson et al., 1998; Rao et al., 1993) are in contrast with the current
findings of attenuated release from PI on the CVLT. As opposed to the
singular categorical shift on release trials in Wickens-type paradigms,
two simultaneous events occur during release on the CVLT. Specifically,
successful release from PI occurs when participants register the
distinctiveness of unshared List B items, while simultaneously inhibiting
interference between shared words from List A and to-be-recalled List B
categorical associates. During this ongoing semantic interference,
participants must simultaneously register a categorical shift, a
process requiring processing resources in healthy individuals (Moscovitch, 1994). Normal accumulation and attenuated
release from PI may reflect a selective deficit in the ability to register
and respond to the semantic properties of stimulus items flexibly
in order to facilitate encoding under taxing conditions. A decreased
response to semantic shifts in the stimulus set parallels the loss of
semantic distinctiveness of memory material (Carroll et
al., 1984) and substantially reduced incidental semantic
categorization in LTM (e.g., Arnett et al.,
1997) previously reported in MS samples.
Reduced ability to register distinctiveness in the memory trace could
increase susceptibility to RI. Persons with MS might preferentially
register gist (general) information, rather than contextual and semantic
details (see Goldstein et al., 1992; Thornton et al., 2002). Encoding of general attributes
may make previously learned information more vulnerable to interference
from new entries of a general form, activating similar, superordinate
general target attributes. Conversely, activation of specific and unique
stimulus attributes registering distinctiveness might preserve old
information from the interfering effects of new information (e.g., Hunt & McDaniel, 1993). Diminished release from PI
in MS may therefore reflect a reduced ability to appreciate the specific
(unique) semantic information conveyed in new words.
Attenuated release from PI has often been attributed to information
processing deficits superimposed on memory impairment (Freedman & Cermak, 1986; Squire, 1982). Intact set-shifting (Parkin & Lawrence, 1994) and number of categories
completed on the Wisconsin Card Sorting Test (WCST; Binetti et al., 1995) have been reported as correlates
of release from PI in the elderly and early Alzheimer's patients,
respectively.
Interestingly, CVLT List B recall has also been selectively associated
with executive abilities in neurological patients (Vanderploeg et al., 1994), and reduced release from PI
following frontal versus temporal lobe resection has been
reported in epilepsy samples (McDonald et al.,
2001). Furthermore, Troyer and colleagues (1996) found that the relationship between neurological
impairment, age, and CVLT performance was mediated by total errors on the
WCST in their MS sample. Based on these findings, attenuated release from
PI could be characterized in terms of reduced mental flexibility leading
to diminished encoding efficiency.
Limitations
There are several methodological limitations to consider when placing
our results within the wider context. An important limitation regarding
diagnostic classification was that MS participants did not undergo a
neurological examination. While this may be problematic in correctly
identifying participants as meeting Poser et al. (1983) or McDonald et al. (2001) criteria for MS, it seems improbable that
participants would erroneously report receiving a diagnosis of MS from
their physician. Nonetheless, the validity of the diagnosis, especially
for participants reporting probable or possible MS, is questionable. To
address this limitation, we replicated the pattern of effects in a
subsample of those individuals who had reported a diagnosis of definite
MS.
Secondly, our participant sample was not well characterized medically.
It would have been ideal to access complete medical records of MS
participants to address this limitation. Nevertheless, MS patients appear
to be generally capable of making accurate judgments of their
symptomatology and daily functioning. Other researchers have noted good
agreement (72%–86%) between neurologist and patient ratings of both
MS symptoms and performance of activities of daily living (ADLs) (Gulick et al., 1997).
Finally, several confounding factors could affect the pattern of
interference indices. To address this issue, the current findings were
replicated in participant subsamples who were (1) less apt to be
significantly depressed, (2) whose medications are not specifically
antagonistic to memory, and (3) who were not experiencing symptom
exacerbation. In these analyses, the overall pattern of results held.
Summary and Conclusions
In a community-based MS sample, attenuated release from PI and
intensified RI were observed on the CVLT. For MS participants only, RI
emerged as a strong predictor of LTM; thus, it appears to represent a
central feature of the MS-related memory deficits. We also propose that
these abnormal interference effects in MS may be mediated by limitations
in the contextual and semantic richness of encoding and/or by
decreased mental flexibility and executive functions. Future studies will
be necessary to clarify more precisely the mechanisms underlying abnormal
interference effects in MS.
