You are on page 1of 11

Neuropsychologia 41 (2003) 1230–1240

An investigation of errorless learning in memory-impaired patients:


improving the technique and clarifying theory
Rebecca Tailby a , Catherine Haslam b,∗
a The Australian National University, Canberra, Australia
b School of Psychology, University of Exeter, Exeter EX4 4QG, UK
Received 12 June 2001; received in revised form 7 January 2003; accepted 7 January 2003

Abstract
In rehabilitating individuals who demonstrate severe memory impairment, errorless learning techniques have proven particularly effective.
Prevention of errors during acquisition of information leads to better memory than does learning under errorful conditions. This paper
presents results of a study investigating errorless learning in three patient groups: those demonstrating mild, moderate, and severe memory
impairments. The first goal of the study was to trial a new version of errorless learning, one encouraging more active participation in
learning by patients via the use of elaboration and self-generation. This technique led to significantly better memory performance than
seen under standard errorless conditions. This finding highlights the value of encouraging active and meaningful involvement by patients
in errorless learning, to build upon the benefits flowing from error prevention. A second goal of the study was to clarify the mechanisms
underlying errorless learning. Memory performance under errorless and errorful conditions was compared within and across each group of
patients, to facilitate theoretical insight into the memory processes underlying performance. The pattern of results observed was equivocal.
The data most strongly supported the hypothesis that the benefits seen under errorless learning reflect the operation of residual explicit
memory processes, however a concurrent role for implicit memory processes was not ruled out.
© 2003 Elsevier Science Ltd. All rights reserved.
Keywords: Errorless learning; Memory impairment

1. Introduction tiveness of the technique is to be enhanced, there are two


issues that need to be addressed. The first is the reliance
One of the more successful techniques used in rehabili- on the examiner to generate responses. If individuals can
tating people with memory disorders, especially those with generate their own responses, this in turn may strengthen
more severe forms of impairment, is errorless (EL) learning. memory traces and perhaps facilitate the process. Secondly,
The EL learning technique involves learning or encoding theoretical implications need to be clarified. Presently, ques-
new information without error. To achieve this, individu- tions remain about the processes responsible for improved
als are given the correct information during each learning memory under EL learning conditions. If the mechanisms
episode. In a typical experiment, this involves the examiner that support such learning are clearly identified, we will
providing the same new information to the patient over be in a better position to improve the effectiveness of
multiple learning trials, with the patient repeating or writing the technique. Both issues are addressed in the present
down the information. This can be contrasted with errorful research.
(EF) learning, where individuals are encouraged to guess EL learning originated in behavioural psychology re-
and thus are more likely to produce errors during learning search in the 1960s. Terrace [3,4] and Sidman and Stoddard
[1,2]. Despite some successes, the EL learning technique [5] used a similar technique to teach visual discrimination
is not without limitations. For patients with severe memory to pigeons and to children with intellectual disabilities. Al-
impairment, acquisition is effortful and requires the thera- though new learning was poor in both groups, it was found
pist or examiner to direct learning. There also appears to be that training under errorless conditions facilitated learning
little generalisability across learning situations. If the effec- in these populations. However, the general technique was
not applied to the treatment of memory disorders until
∗ Corresponding author. the 1990s. The focus of this work was both practical and
E-mail address: c.haslam@exeter.ac.uk (C. Haslam). theoretical. Researchers investigated both the effectiveness

0028-3932/03/$ – see front matter © 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0028-3932(03)00036-8
R. Tailby, C. Haslam / Neuropsychologia 41 (2003) 1230–1240 1231

of the technique and the mechanisms or processes that 1.1. Improving the technique
supported learning when applying the technique.
EL learning was first applied to the treatment of memory There is existing evidence to suggest the EL learning
disorders by Wilson et al. [1]. These researchers investigated technique can be improved. Although Evans et al. [8] found
list-learning performance in severely memory-impaired (i.e. a limited advantage when learning face-name associations
amnesic) patients and two non-impaired control groups (a using standard EL conditions, memory performance was
young group and an elderly group). They found that patients best when the EL learning task was extended to include ac-
made significantly more correct responses on a cued recall tive processing of associations via use of a visual imagery
test following EL learning than they did following EF learn- strategy. Further evidence is provided by Squires et al. [6],
ing. Furthermore, the amnesic group received greater benefit who found that the benefits of EL over EF methods were
from the EL condition than did either control group. Bad- greater on a novel association task when compared with
deley and Wilson [2] argued subsequently that EL learning remotely linked word pairs. These researchers argued the
both enhanced learning and reduced forgetting rates in am- added cognitive effort required during learning of novel
nesic patients. associations, involving the creation of an extrinsic link
Many researchers have replicated this finding of supe- between the target words, resulted in a stronger memory
rior performance under EL learning conditions both in representation and better performance at test.
immediate and delayed memory tests. The EL technique Additional evidence from the cognitive literature is rel-
has been shown to be effective when learning remotely evant to this point. Much of the evidence concerning the
linked word pairs and unrelated word pairs [6], as well benefits of active participation in learning comes from re-
as learning extended word lists [7]. It has also been use- search investigating levels of processing (LOP) theory. In
ful in the acquisition of more practical information. In a LOP theory it is argued that the strength or durability of
series of case studies, amnesic patients showed superior a memory representation is dependent on the manner in
learning of object names, names of people, word lists, ori- which the information is processed during encoding [11,12].
entation items (time, place, age) and general knowledge, as Memories are strengthened by elaboration, or by processing
well as electronic diary programming [1]. A recent study information on the basis of its meaning. Current EL learning
found a limited advantage of EL over EF methods on a techniques are akin to rote learning, as they involve repeated
face-name learning task [8]. EL learning has also been exposure to correct information. According to LOP theory,
applied to the restoration of premorbid knowledge [9] and retention of information acquired during EL learning should
the development of procedural skills in a work setting [10]. be improved by increasing the level of active involvement
Although most studies have focused on amnesic patients, or processing engaged by participants during encoding. Re-
the EL learning technique has also been shown to be rea- searchers have shown memory performance is best when in-
sonably successful in facilitating performance in patients formation is learned in a meaningful or semantic context—
suffering from a moderate degree of memory impairment for instance, when processed conceptually or elaboratively
[7]. rather than on the basis of superficial or lexical characteris-
Despite these benefits, there are limitations to the useful- tics. This finding has emerged in both clinical (e.g. [13–15])
ness of the EL technique. First, not all studies have found and non-clinical (e.g. [11]) populations. This evidence high-
the benefits of EL learning to be maintained over time delay lights the value of exploiting semantic memory, which is
(e.g. [6]). Furthermore, there is some evidence to suggest preserved in many people with acquired memory deficits.
that EL learning may not be the best method in all cases, The present research draws upon this notion, and seeks to
particularly when acquiring more complex information. For determine whether EL learning can be improved by encour-
instance, tasks involving multiple steps or requiring the ac- aging participants to engage in elaborative processing during
quisition of more detailed information have been difficult for learning.
amnesic patients even when learning under EL conditions.
Evans et al. [8] found only a limited advantage of EL over 1.2. Theory
EF methods on a face-name learning task, and no advantage
on other complex tasks, namely, programming an electronic There is considerable evidence supporting the case that
organiser and route-finding. EL learning is an effective technique for the rehabilitation
A further limitation of the EL learning technique is that of patients with memory impairment. Yet the cognitive pro-
it remains quite directive. Participants need only listen to cesses responsible for improved retention of information
correct answers, follow written instruction, or on occasion learned under EL conditions are the subject of some debate.
write the answer, over a series of learning trials. Wilson and Two theories have been proposed, and both focus on the
her colleagues have suggested that this may not be sufficient distinction between implicit and explicit memory. The first
to ensure adequate learning [1]. This raises an interesting theory proposes that the benefits seen under EL conditions
question—could the effectiveness of the EL learning tech- are supported by implicit memory. The second theory pro-
nique be improved if patients were somehow able to gener- poses that these benefits are supported by residual explicit
ate their own correct responses during learning? memory.
1232 R. Tailby, C. Haslam / Neuropsychologia 41 (2003) 1230–1240

1.2.1. Implicit memory theory ory. This was first suggested by Hunkin et al. [7]. These
The implicit memory theory was first proposed by Badde- researchers used the same learning procedure as Baddeley
ley and Wilson [2] to account for the facilitation in memory and Wilson [2], but tested memory in two ways—using an
performance observed in amnesic patients who were sub- implicit test (fragment completion) followed by an explicit
jected to the EL learning technique. This theoretical account test (cued recall). They found no correlation between items
is not surprising given evidence of preserved implicit mem- produced in the implicit and explicit memory tests, which
ory in these patients [16,17]. Specifically, Baddeley and is contrary to expectations if performance in cued recall re-
Wilson found that under EF conditions, amnesic patients flects implicit memory. Based on these findings, Hunkin and
tended to repeat the same errors across learning trials. It colleagues argued that the benefits associated with EL learn-
was suggested this occurred because EF learning relies on ing must therefore reflect the operation of residual explicit
explicit memory processes, which amnesic patients cannot memory. This conclusion was strengthened by their findings
apply. Explicit processes enable the conscious monitoring of EL learning advantages in free recall—a task that is ar-
and elimination of errors. Without this capacity, one cannot guably explicit. They thus joined other researchers who ar-
modify and update responses during learning. Hence, under gue that any learning observed in amnesia is only possible
EF conditions any error generated by memory-impaired via the use of residual explicit memory (e.g. [19]).
patients may be repeated, resulting in reinforcement of an Evidence supporting this second theory is also open to
incorrect response. In contrast, implicit memory, attuned to qualification. In the Hunkin et al. [7] study, the accuracy ob-
eliciting production of the strongest response, is reported to served in performance in the fragment completion task was
be intact in amnesic patients (e.g. [16,17]. Implicit learning quite low. Performance at such low rates of accuracy might
is well served under EL conditions, as by eliminating errors be considered to reflect residual explicit memory, which is
during learning the strongest response will be the correct expected to be poor in groups with severe memory impair-
response and this would be the only one reinforced. On the ment. Yet this conclusion poses some difficulties, given the
strength of these arguments and their findings, Baddeley lack of correlation with performance on cued recall (their
and Wilson concluded that the advantages seen under the test of explicit memory). The authors note their failure to
EL technique reflected the operation of implicit memory. find a correlation between implicit and explicit test perfor-
Further, albeit limited, support for the implicit theory was mance may be an artefact resulting from the use of different
provided by Evans et al. [8]. These authors concluded that stimulus materials in their tasks. Performance in fragment
EL learning may only be beneficial in tests of implicit mem- completion (i.e. implicit test) was compared with that on
ory, and may be ineffective when explicit recall of infor- word-stem completion (i.e. explicit test) and the processes
mation is required. They report mixed results in their study, involved in completing these tasks may differ.
and argued that EL learning was only useful in tasks or sit-
uations facilitating retrieval of implicit memory for learned 1.2.3. Summary
material. In sum, error prevention during learning has been found
Critical analysis of the above studies exposes some short- in many cases to be advantageous over techniques that allow
comings. For instance, although Baddeley and Wilson [2] errors to be made in the study period. The benefits of EL
proposed that EL learning must be supported by implicit learning have been shown in a number of learning tasks using
memory, this conclusion was based on the assumption that a variety of materials and information. To date, two theories
implicit memory was preserved in patients with amnesia. Al- have been proposed to explain the mechanisms underlying
though numerous researchers have found that implicit mem- this observed advantage. However the evidence provided to
ory and learning is relatively preserved in amnesia [16,17], date in support of each theory remains open to challenge.
other researchers have argued that implicit memory is not The present research re-examines both theories and attempts
preserved in all tasks performed by amnesic patients [18]. to establish which best fits the data.
This highlights the need to demonstrate preservation of im-
plicit memory with the task being used, rather than simply
accepting an assumption that may be false. The reasoning 2. The present investigation
behind Evans et al.’s [8] explanation is also open to ques-
tion. These authors concluded that EL learning may only be In the present research the two main issues arising from
beneficial in implicit memory tests. Yet in their study (as in the literature are addressed. The first concerns the reliance on
most others) EL learning was beneficial in cued recall tasks, an examiner-directed approach to EL learning and whether
and in one free recall task. Both cued and free recall are the method can be improved by providing semantic cues
measures commonly employed to tap explicit memory. during learning so as to enable participants to generate re-
sponses themselves without error. The second concerns clar-
1.2.2. Residual explicit memory theory ification of the process underlying EL learning.
A subsequent line of empirical research provided support Analysis of the literature from both clinical and non-
for an alternative theory—namely, that the benefits observed clinical domains highlights the value of exploiting seman-
during EL learning are supported by residual explicit mem- tic memory and encouraging deeper or more “meaningful”
R. Tailby, C. Haslam / Neuropsychologia 41 (2003) 1230–1240 1233

processing. In the present study this principle was applied method (i.e. where EL is greater than EF), a simple effect
to the EL learning paradigm. A method was developed for severity in the EF condition and a method by severity
via which participants could generate their own answers interaction is required. The same pattern of results would be
without error (i.e. “self-generated” method) in response to predicted by the residual explicit theory with an additional
semantically rich descriptions of the target words. It was simple effect for severity in the EL condition.
hypothesised that encoding information without error in a
meaningful and active manner would strengthen the mem-
ory representation relative to that formed under the standard 3. Method
EL (i.e. examiner-generated) condition, thus leading to
better performance at test. Hence, increased accuracy and 3.1. Subjects
retention were predicted in the self-generated EL condition.
The present research also investigated whether EL learn- Twenty-four participants took part in the study. Although
ing is supported by implicit memory or residual explicit they differed in pathology (head injury: n = 12, CVA:
memory. This question was addressed by comparing EL and n = 6, hypoxia: n = 3, dementia: n = 1, herpes simplex
EF learning in patients suffering from a range of memory encephalitis: n = 1, Parkinsons: n = 1), most had acquired
deficits—that is, mild or no memory impairment, moder- deficits in explicit memory. A small number of participants
ate memory impairment, and severe memory impairment. in the mild/no impairment group displayed non-impaired
Severity of memory impairment was used as the basis from explicit memory (although all had suffered some neurolog-
which to predict performance according to the two theoreti- ical incident). Participants were divided into three groups,
cal explanations—that is, whether EL learning is supported with eight in each, on the basis of their performance on
by (a) implicit memory, or (b) residual explicit memory. the Verbal Memory Index (VMI) on the Wechsler Memory
Severity was classified on the basis of performance on a stan- Scale-III [20]. The VMI was selected as an appropriate
dardised memory test. A measure of implicit memory was tool for assignment to groups as the experimental tasks in
also taken rather than simply accepting it would be equal the present research were auditory-verbal in nature. A VMI
across groups. This was important given our theoretical ex- score falling between 0 and 69 was classified as severe
ploration required proof of this assumption. (this score range is achieved by less than 2% of the general
If EL learning is primarily supported by implicit memory population and is typical of that achieved by patients suf-
processes, then EF performance should decrease as the level fering from amnesia). Moderate impairment was defined as
of memory impairment increases, as assignment to groups is obtaining a score between 70 and 89 (indicative of perfor-
based on level of explicit memory function. By contrast, if mance between the second and twenty-third percentiles),
learning in the EL condition reflects the operation of implicit and a classification of mild or no impairment was assigned
memory processes then EL performance should be equiva- to participants scoring at or above 90. Participants were
lent, and relatively strong, across all participant groups. The matched as closely as possible across groups in terms of
advantage of EL over EF learning would accordingly be ex- age, gender, premorbid and current intellectual function-
pected to increase as the severity of memory impairment ing. Participants were excluded if deemed to suffer from
increases. depression, which was determined by responses to items
If EL learning is supported by residual explicit memory, in the Beck Depression Inventory-II where a cut-off score
performance under EL conditions would be expected to dif- of 12 was used. Summary demographic information and
fer between groups in line with group differences in the level screening statistics for participants appear in Table 1.
of explicit memory impairment, as is predicted for the EF
condition. However, EL performance would still be supe- 3.2. Materials
rior to EF performance, due to encoding differences. In the
EF condition, an incorrect response is introduced during ac- The materials comprised 96 words of five to six letters
quisition. This should produce interference, making perfor- in length: 48 target words and 48 substitute words (see
mance poorer on a subsequent explicit memory task than in Appendix A for a list of these words). Twenty-four of the
the EL condition, where no such interference is generated target (and substitute) words contained five letters, and
during acquisition. Resolving such interference requires the the remaining twenty-four target (and substitute) words
use of explicit memory, thus the interference effect would contained six-letters. Word length varied to allow for a
be expected to be greatest for participants with the most se- sufficient number of target and substitute words. All target
vere memory impairment, with the size of the interference words were concrete nouns with high imageability. No two
decreasing as the level of memory impairment decreases. target words began with the same two-letter stem, and target
Statistical predictions relevant to each theory can also be words were selected so that the two-letter stem provided for
generated. The one finding that would discriminate between each had at least four possible stem completions for the ap-
these theories is a simple effect for severity in the EL con- propriate word length (i.e. five or six letters). For each of the
dition, which is only predicted by the residual explicit the- 48 target words, a semantically-rich description was created
ory. For the implicit theory to be supported a main effect for for use in the self-generated EL condition. Pilot testing of
1234 R. Tailby, C. Haslam / Neuropsychologia 41 (2003) 1230–1240

Table 1
Demographic information and results of screening tests as a function of severity of memory impairment
Memory impairment group Age Males:females ratio NART-R Estimated premorbid IQa Estimated current IQb
Severe 43; S.D. = 17.2 1:1 27.6; S.D. = 9.9 97.8; S.D. = 10.4 99.9; S.D. = 18.8
Moderate 43.8; S.D. = 13.1 3:1 29.5; S.D. = 10.8 99.8; S.D. = 11.3 106.5; S.D. = 11.3
Mild/no 37.5; S.D. = 15.7 7:1 33; S.D. = 8.1 103.5; S.D. = 8.5 103.4; S.D. = 6.8
a Premorbid IQ estimated using NART-R.
b Current IQ estimated using WAIS-R short form.

the semantic descriptions was conducted on 10 individuals with appropriate counterbalancing of lists to condition across
to ensure the semantic information was sufficient to enable participants within each group as well as counterbalancing
error-free identification of the word by all those tested. Defi- of the order of methods studied. Each participant was seen
nitions for 46 of the words met this criterion. The remaining over two sessions. Two conditions were presented in each
two definitions were then revised and piloted on another session, with a break of 20 min between conditions. There
eight people. The new definitions successfully met the cri- was at least 1 week separating sessions 1 and 2. There were
terion. Finally, each target word had a matched “substitute” two phases to the experimental procedure—a study phase
word, with the same two-letter stem, length and frequency. and a test phase.
These words were used (if required) in the errorful learning
condition only—if participants correctly guessed the target 3.4. Study phase
word on their first attempt, they were told they were in-
correct and the substitute item then became the new target At the commencement of the experimental procedure par-
word. ticipants were told they would be asked to learn some word
The 48 target words were assigned to one of four lists (lists lists, and that each list would be presented a total of three
A, B, C and D). Each list of 12 target words was balanced on times. Participants were also informed that after presenta-
average word frequency (40), and contained an equal number tion of each item in the list they would be asked to write
of five and six-letter words. Where there were multiple target down the word on a piece of paper. In all but the SC task,
words from the same category (e.g. “animals”, “fruit”), these participants were informed that their memory for the words
items were evenly distributed between the lists. would be tested. These latter three conditions (EF, EL(E)
A fifth list, list E, consisted of twelve 2-letter word stems and EL(S)) will be referred to as the “explicit” conditions. In
to be used as distractors in the implicit test, which was a the study phase participants were given the length (i.e. five
stem completion task. This type of implicit memory test was or six letters) and the first two letters of each target word
chosen as it was equivalent to the tasks employed in the rest both verbally, and visually via a flashcard (e.g. BR . . . ).
of the experimental procedure (i.e. the same test materials Following this, instructions varied according to the specific
and procedures—cueing with word stems—were used in the condition as follows.
explicit and implicit tasks). Half of the list E distracter stems
comprised five letters (e.g. SL . . . ), and the remaining half 3.4.1. Errorful learning
comprised six letters (e.g. YE . . . ). All twelve stems had at In this condition participants were asked to guess what the
least one possible completion. word might be. For example, they were told: “I’m thinking
of a five-letter word beginning with “BR”. Can you guess
3.3. Procedure what the word is?” Participants were allowed up to three
guesses (or a maximum of 25 s) and were told whether they
All participants underwent preliminary neuropsycholog- were correct or incorrect after each guess. A failure to re-
ical screening to facilitate subject matching across groups. spond was not accepted, and participants were encouraged
Premorbid and current intellectual functioning were assessed to make a guess. If they had not guessed correctly by the
using the National Adult Reading Test-Revised (NART-R: end of the time limit, they were told the word (e.g. bread).
[21]), and a short form (comprising vocabulary and block If participants provided the correct response on their first
design subtests: [22]) of the Wechsler Adult Intelligence guess, they were told that they were incorrect, and the tar-
Scale-Revised (WAIS-R: [23]) respectively. As mentioned get word was then replaced by the corresponding substitute
previously, allocation to the three experimental groups was word (e.g. brush). This ensured that at least one error was
on the basis of performance on the VMI. made for each stimulus item.
All participants were tested in each of the three learn-
ing conditions—errorful, standard (examiner-generated) er- 3.4.2. Errorless learning (examiner-generated)
rorless (EL(E)), and self-generated errorless (EL(S)). Each and stem completion
participant also completed the stem completion (SC) test In these conditions participants were provided with the
condition. A different target list was used in each condition, correct word. For example, “I’m thinking of a five-letter
R. Tailby, C. Haslam / Neuropsychologia 41 (2003) 1230–1240 1235

word beginning with “BR”, and the word is “BREAD”. (see Appendix B) to identify strategies employed when per-
Please write that down”. forming the stem completion task. As no participant indi-
cated they had deliberately tried to complete the stems with
3.4.3. Errorless learning (self-generated) study list words, it is likely they used implicit processes to
In this condition participants were given a verbal descrip- perform this task.
tion of the target item and asked to name the word (e.g. “I’m
thinking of a five-letter word beginning with “BR”, and this
word describes a food made of flour, liquid and yeast which 4. Results
is baked and then sliced to make sandwiches. What do you
think the word might be?”). Participants were given 25 s to Six “explicit” scores were calculated for each partici-
respond. A failure to respond was not accepted and partici- pant, comprising the number of items recalled correctly
pants were strongly encouraged to provide a response. (out of 12) in each explicit learning condition—EF, EL(E),
Once the target word was identified, participants were EL(S)—at both immediate and delayed test. Mean scores
asked to write down the word on a small piece of paper. and the percentage of accurate responses are provided in
These pieces of paper were removed after each word, to Table 2. As initial testing revealed no effect of time, the
control the amount of exposure to each item. This procedure data were collapsed across time (immediate and delayed)
was carried out for each of the 12 words on the list, and and all analyses conducted using collapsed data. The results
the whole procedure was repeated twice so that participants will be discussed in two parts: the first addressing the prac-
studied each target word three times. Order of presentation tical question of improving the technique, and the second
of the 12 words on successive trials was constant. addressing the question concerning theory.

3.5. Test phase 4.1. Improving the technique

Following completion of the three learning trials in each To determine whether elaborative processing of informa-
condition, a 5 min distracter task was administered (a digit tion enabling self-generation of responses during errorless
span test). Memory was tested both immediately following acquisition was more beneficial than the standard error-
the distracter task (all conditions) and again after a 30 min less learning procedure, performance in self-generated and
delay (“explicit” conditions only). Participants were pro- examiner-generated (i.e. standard) EL learning conditions
vided with a break between immediate and delayed tests. was compared. A two-way mixed ANOVA was conducted,
The testing procedure differed slightly depending on the comparing performance under EL(S) and EL(E) condi-
condition and details are provided below. tions in the three memory-impaired groups (i.e. mild/no,
moderate, severe). As noted above, performance in each
3.5.1. Explicit learning conditions (EF, EL(E) and EL(S)) condition was collapsed over time. A significant main ef-
Memory was tested using a cued recall task. Participants fect of learning method was found, indicating that cued
were presented with the flashcards used in the study phase, recall performance following self-generated EL learning
and were then asked to recall the target word corresponding was significantly better than that seen under standard EL
to the stem cue. These cards were presented in a random conditions (F(1, 21) = 18.74, P < 0.0001). A signifi-
order. Participants were encouraged to provide a response, cant main effect for group was found (F(2, 21) = 13.60,
and if unsure, to guess. Only one response was allowed for P < 0.001), however no method by group interaction was
each word stem. found (F(2, 21) = 1.15, ns). The above results indicate the
modified errorless learning method, which used semantic
3.5.2. Implicit test (stem completion) cues to enable error-free generation of responses by par-
Participants were told that they were required to complete ticipants, produced the best memory performance and the
a list of two-letter word stems of varying length and asked advantage of this method over the standard EL method was
to complete stems with the first word that came to mind. maintained over time and across groups. It is worth noting
It was explained that they had no more than 2 s for each that in both EL conditions, very little information was lost
item, and that the examiner was timing how quickly they following delay (see mean values reported in Table 2).
completed the task. If they got stuck on any stem they were
instructed to leave it and focus on the next stem. Partici- 4.2. Clarifying the mechanisms behind EL learning
pants were not permitted to go back to missed items. Pre-
sentation of the word stems was via the flashcards used in To examine the question of theory, it was important to
the original study phase, randomly interspersed with the 12 determine first whether there was any difference in im-
distracter stems from list E. The experimenter presented the plicit memory functioning (at least on a test equivalent to
flashcards consecutively for 2 s each. A tape recorder and those employed in the experimental conditions) between
microphone were used to record responses. On completion, groups. To achieve this, two scores were calculated for each
participants completed the Memory Strategies Questionnaire participant. First, the number of target stems on the stem
1236 R. Tailby, C. Haslam / Neuropsychologia 41 (2003) 1230–1240

Table 3
Mean priming scores across participant groups
Memory impairment group Mean priming score

Severe 4.50
Moderate 4.25
Mild/no 5.75

completion test completed with items from the original


study list was calculated, giving a “target score”. Second,
the number of distractor stems completed with pre-allocated
baseline items (selected to match target items on word
frequency—see Appendix A for baseline items) was calcu-
lated, giving a “baseline score”. A priming score was then
calculated for each participant, being the difference between
his or her target and baseline scores. Mean priming scores
for each group are presented in Table 3.
A one-way between-groups ANOVA was then conducted
to compare levels of priming across the three participant
groups. Results indicated that the level of priming did
not differ significantly between groups (F(2, 21) = 1.48,
P > 0.05). As priming is considered to be an indicator of
implicit memory functioning, it was concluded that implicit
memory processing on this type of task was comparable
across all participant groups.
To address the question of theory, an analysis comparing
performance under EF and standard EL (i.e. EL(E)) con-
ditions was conducted. Mean performance scores for each
group under the two conditions are provided in Table 2.
The first analysis explored whether data from the present
study replicated earlier findings of an advantage of EL
over EF learning. A two-way mixed ANOVA was con-
ducted, comparing data from the EL(E) and EF conditions
across all participant groups. As expected, the group ef-
fect reached significance (F(2, 21) = 14.63, P < 0.001).
Furthermore, a significant main effect for learning condi-
tion was found, indicating that memory performance was
significantly better following learning under standard EL
conditions (F(1, 21) = 20.76, P < 0.0001). Thus the cur-
rent study provides further confirmation of the superiority
of learning methods incorporating error prevention.
In light of the significant difference that emerged in mem-
ory performance between the EL(E) and EF conditions,
the question remained whether this difference reflected the
operation of implicit memory processes or residual explicit
processes. Resolution of this query was possible by inves-
tigating whether there was a simple effect of severity (i.e.
in memory impairment) in each learning condition. Specif-
ically, the veracity of the implicit theory can be tested by
investigating whether there was a simple effect of severity
for the EL(E) condition. The implicit theory predicts that
all groups should perform equally if EL performance is
supported solely by implicit memory processes, as groups
have been found (via the stem completion test) to show
no difference in implicit processing on an equivalent task.
The implicit theory prediction was tested using a one-way
R. Tailby, C. Haslam / Neuropsychologia 41 (2003) 1230–1240 1237

ANOVA, with performance under standard EL conditions The answer to the first research question was unequi-
as the dependent variable. A significant effect of sever- vocal—memory performance resulting from learning under
ity was found (F(2, 21) = 9.90, P < 0.005) suggesting the new self-generated errorless technique was significantly
that performance in standard EL conditions differed be- better than that under standard errorless conditions. The re-
tween the groups. Post hoc contrasts revealed the moderate sults comparing the standard and self-generated EL methods
group performed significantly better than the severe group indicate there is much to be gained from active as opposed
(t(21) = 2.7, P < 0.05), as did the mild group (t(21) = to passive participation during study. This benefit was seen
4.42, P < 0.001). Although there was a trend towards better in all groups, and the advantage persisted over time. While
performance of the mild group compared to the moderate, there was no loss of information over time with the standard
this was not significant (t(21) = 1.70, ns). Importantly, the technique, the best result was obtained with the modified
finding of an effect of severity in the standard EL condition technique which provided sufficient elaborative cues dur-
does not support the implicit theory, but it is consistent with ing learning to enable self-generation of responses without
the residual explicit theory. error.
A second one-way ANOVA taking performance in the EF Particular mention should be made of performance in the
condition as the dependent variable also revealed a signifi- severely memory-impaired group. These participants expe-
cant effect of severity (F(2, 21) = 15.17, P < 0.001). This rienced the most difficulty in any test of memory. Despite
was as expected given participants were allocated to groups this, at immediate test the severe group recalled two-thirds
based on their performance on a standard explicit memory (68.8%) of the target items learned using the self-generated
measure. Finally, the method by memory impairment sever- EL method, with very little loss of information after delay.
ity interaction in the overall two-way ANOVA failed to reach This was substantially more than that recalled in the stan-
significance (F(2, 21) = 1.26, ns). Although this finding dard EL condition, where around half (52.1%) of the target
was contrary to predictions made under both implicit and items were recalled, and in the EF condition where only
residual explicit theories, it is important to note that a trend a third (33.3%) of target items were recalled. In practical
in the predicted direction emerged. Closer inspection of the terms, this suggests that the self-generated elaborative EL
data in Table 2 shows that the difference between perfor- method has the same advantage over the standard EL condi-
mance under the EL(E) and EF conditions was noticeably tion as the latter has over the EF condition. Although there
greater for the severe group than for the moderate and mild is a marked advantage for standard EL conditions relative to
groups. the trial-and-error method, the present research has shown
The results of statistical analyses, showing a main effect that a similar degree of improvement can be produced
for method and simple effects for severity in both the EL and over the standard EL technique simply by engaging par-
EF conditions, were used to evaluate the competing theories. ticipants in active semantic processing and self-generation
Although the present findings do not find an exact match of responses during study. While this advantage was most
with either of the predicted profiles (as noted in the introduc- dramatic in the severely memory-impaired group, similar
tion), the closest match emerged with the residual explicit gains were seen in the other two groups. These findings,
theory. The implications of the data, including possible al- while confirming previous reports of memory facilitation
ternate explanations for the pattern of results observed, are using the EL learning method, also extend them by high-
discussed below. lighting the value of tapping into the semantic system in
memory-impaired patients, and encouraging active engage-
ment during acquisition of new information to facilitate
5. Discussion stronger memory representations.
Aside from evidence of superiority in terms of the num-
The present study sought to improve the practical use- ber of items recalled and the persistence of these benefits
fulness of an already beneficial rehabilitation technique, over time in the self-generated condition, there were several
and to increase our theoretical understanding of the mecha- flow-on effects that emerged in the course of conducting
nisms supporting errorless learning. The first issue involved this experiment. In many cases participants spontaneously
determining whether the technique could be improved by reported that they found it much easier to recall list words
encouraging more active, or elaborative, processing during at test following learning under EL(S) conditions than un-
errorless acquisition of information. The present findings der the standard EL method (and of course the EF method).
indicate that the self-generation elaborative EL method In addition, the experimenter noted that responding at test
employed resulted in a superior memory performance com- tended to be more rapid following EL(S) learning than the
pared with the standard errorless technique. Performance other two explicit methods. In other words, responses to tar-
was also maintained over time. A second issue concerned get items appeared more accessible. Greater confidence in
the question of theory and in particular, what process is responding was also apparent in responding to these target
responsible for supporting errorless learning. Results were items. While these factors were not the primary focus of this
more equivocal on this matter, with the most consistent experiment, they provide additional evidence of the superi-
match emerging with the residual explicit theory. ority of the self-generated method.
1238 R. Tailby, C. Haslam / Neuropsychologia 41 (2003) 1230–1240

While the gains seen under the EL(S) technique are clear, a processes (although [24], did consider this possibility in a
question arises as to which component of the technique most study on a similar learning technique—the method of van-
facilitated performance. That is, whether it was the elabora- ishing cues). Given the central predictions of both the “single
tive cues provided during target acquisition, the fact that re- process” theories were not fully substantiated by the find-
sponses were actively generated by participants themselves ings of the present study, the combined influence of both
rather than by the examiner, or a combination of the two. Our processes arises as an alternate explanation worthy of con-
findings cannot address this question, and further studies are sideration. Were a combination of implicit and explicit pro-
needed to clarify the relative importance of semantic cueing cesses responsible for improving EL performance, then the
during acquisition, and the act of response-generation itself, predicted difference between EL and EF conditions as in-
in facilitating memory performance in patients demonstrat- dicated by the residual explicit theory would be reduced
ing memory impairment. further. All other predictions would be consistent with the
In relation to theory, the findings appeared to discount residual explicit theory. In this event, an interaction would
the implicit theory proposed by Baddeley and Wilson [2] still be predicted but the effect more difficult to substantiate
but provided some support for the residual explicit theory statistically. Given the predicted patterns of performance un-
proposed by Hunkin et al. [7]. Comparing the results seen der the “residual” and “combined’ theories are so similar, it
under EF and standard EL conditions (Table 2), the first is difficult to distinguish between these accounts using our
thing to note was the superior performance seen in the EL method.
condition in all memory-impaired groups (although only A second and related possibility is that EL learning may
a small advantage was seen for the moderate impairment be supported by different processes in different individuals.
group). A significant overall advantage of EL learning was For instance, it is possible that EL learning relies primarily
found, in line with previous studies and as predicted by on implicit processes in very severely impaired patients who
both theories. The additional finding of a significant simple lack explicit memory abilities, but relies more on explicit
effect of severity in the EL condition suggests that implicit processes in those demonstrating only a mild impairment.
processes alone are not sufficient to explain performance. Finally, interpretation may have been complicated in the
No differences should have emerged between groups in the present study by the ceiling effect observed in the errorless
EL condition were performance due solely to implicit pro- conditions in the mild/no impairment group. Future studies
cesses. This was not found—performance in the EL condi- could avoid the possibility of ceiling effects by increasing
tion increased as severity of memory impairment decreased. list length or having fewer repetitions during acquisition,
It might be argued then that there were differences between which might then reduce performance in the mildly impaired
groups’ implicit memory function. This can be discounted group. This would enable a clearer assessment of the level
given no difference was found between groups in an implicit of benefit provided by EL learning across all groups, and
stem completion task which used equivalent test procedures permit greater confidence in conclusions.
to those employed in the experimental conditions. As such, These considerations notwithstanding, the present study
the most reasonable explanation is that implicit memory has shown that an already beneficial memory rehabilitation
alone cannot explain the performance observed. technique can be improved further with the use of active
Having said this, it must also be noted that the results did or elaborative semantic processing and self-generation dur-
not fully support the residual explicit theory. Of the four sta- ing learning. While the concept itself is not new, it has not
tistical effects predicted by this theory, the only effect not been applied to EL learning previously. This research has
observed in this study was the prediction of a method by only explored one aspect of self-generated elaborative en-
severity interaction. Although not reaching statistical sig- coding. The options for further exploration of this principle
nificance, inspection of Table 2 revealed some differences are numerous. There are many ways in which researchers or
in the mean scores between EL and EF conditions in each therapists might facilitate elaborative processing by patients
group, which is in line with the conceptual predictions of the whilst still preventing errors. Information could be taught in
residual explicit theory. Specifically, in each group accuracy the standard EL fashion (i.e. rote presentation of words) then
in performance was found to be higher under EL conditions patients could be asked to use each word in a meaningful
than under EF and accuracy in the EL conditions increased sentence, create a story around that concept, or some similar
as severity of memory impairment decreased. The failure of strategy—i.e. encouraging self-elaboration. Alternately, one
this trend to reach significance was unexpected, given that could provide the elaborative context for individuals (which
both theoretical accounts predicted a significant interaction may be advantageous if they have deficits in self-directed
between method and severity, but there are several possible cognitive processing) by providing richly elaborated concep-
explanations for this finding. tual information to serve as a cue to the target information.
First, it is possible that EL learning may be supported by Further possibilities exist, and merely await discovery by
a combination of implicit and explicit memory processes. It innovative therapists. For patients with acquired deficits in
is surprising that the theoretical accounts proposed in the EL semantic memory and/or processing, self-generation could
learning literature to date have not considered the combined be encouraged in response to a highly constraining set of
contribution of both implicit and residual explicit memory non-semantic cues (e.g. letters of the word). The method of
R. Tailby, C. Haslam / Neuropsychologia 41 (2003) 1230–1240 1239

vanishing cues, a technique which draws upon this idea, has Appendix A (Continued )
shown promise with patients who demonstrate severe mem-
ory impairment (for an overview, see [25]). Substitute
The issue of which mechanism(s) support EL learning in (6)NE needle Nephew
memory-impaired patients remains a complex one. While (5)RU ruler Ruins
some support for the residual explicit theory emerged from (6)TU turtle Tunnel
the present findings, the data was equivocal. Single theory (5)WH wheel Wheat
explanations may in fact be insufficient to explain the facil- (6)WI window Winter
itation observed when learning information under EL con-
ditions. It is possible that different individuals capitalise on List D
different processes, or that a combination of processes, both (5)BR bread Brush
implicit and residual explicit, contribute to the facilitation (6)CA cannon Cancel
observed. Further research is required to clarify the relative (6)DO donkey Double
contributions of implicit and residual explicit memory pro- (6)FL flower Flames
cessing in EL learning. (5)GL glove Gland
(6)GU guitar Guards
(6)OR orange Organs
Appendix A. Stimulus words (5)PI piano Pilot
(6)SA saddle Savage
Substitute (5)TA table Taken
(5)TI tiger Timid
List A (5)WA watch Wagon
(5)AR arrow Array
(6)BU button Bucket List E (distracter stems and baseline completions)
(6)CH church Change (5)BE beard
(5)CL cloud Clock (5)DA dance
(6)LA ladder Lagoon (5)LO lolly
(6) MO monkey Modify (5)MA match
(5)PA pants Patch (5)NO north
(6)PE peanut Pepper (5)SL slate
(5)SN snail Sniff (6)EM empire
(5)SW swing Sweat (6)FR freeze
(5)TH thumb Thief (6)JU jumper
(6)TO tomato Toffee (6)PR prompt
(6)SI singer
List B (6)YE yellow
(6)AN anchor Antler
(5)AP apple Apron
(6)BO bottle Bottom Appendix B. The memory strategies questionnaire
(5)HE heart Heavy
(5)KN knife Knees This was a series of three questions designed to determine
(6)PO potato Polish whether participants had deliberately tried to complete stems
(6)RA rabbit Ration with words from the study list, or whether they had in fact
(5)SK skirt Skill responded as instructed with the first word coming to mind.
(6)SP spider Sprout
(5)ST stool Stack B.1. Memory strategies questionnaire
(5)TR truck Trust
(6)VI violin Violet 1. What was your general strategy when completing the
stems?
List C
2. Did you deliberately try to complete the stems with words
(6)BA basket Barrel
from the study list?
(5)CR crown Creek
3. Did you respond with the first word coming to mind (its
(5)DR dress Dream
OK if some of these were study list words, so long as
(6)HA hammer Hazard
you weren’t deliberately trying to recall list words)?
(5)HO horse Hotel
(6)JA jacket Jaguar If response to question 2 was “yes”, participants were
(5)LE lemon Lever judged to be using explicit memory to complete the task.
1240 R. Tailby, C. Haslam / Neuropsychologia 41 (2003) 1230–1240

References [13] Tulving E, Hayman C, Macdonald C. Long-lasting perceptual priming


and semantic learning in amnesia: a case experiment. Journal
[1] Wilson BA, Baddeley A, Evans J, Shiel A. Errorless learning in of Experimental Psychology: Learning, Memory and Cognition
the rehabilitation of memory impaired people. Neuropsychological 1991;17:595–617.
Rehabilitation 1994;4:307–26. [14] Milders M, Deelman B, Berg I. Rehabilitation of memory for people’s
[2] Baddeley A, Wilson B. When implicit learning fails: amnesia and names. Memory 1998;6:21–36.
the problem of error elimination. Neuropsychologia 1994;32:53–68. [15] Thoene AIT, Glisky EL. Learning of face-name associations on
[3] Terrace HS. Discrimination learning with and without “errors”. memory impaired patients: a comparison of different training
Journal of Experimental Analysis of Behaviour 1963;6:1–27. procedures. Journal of the International Neuropsychological Society
[4] Terrace HS. Stimulus control. In: Honig WK, editor. Operant 1995;1:29–38.
behaviour: areas of research and application. New York: Appleton [16] Graf P, Squire LR, Mandler G. The information that amnesic
Century Crofts; 1966. p. 271–344. patients do not forget. Journal of Experimental Psychology: Learning,
[5] Sidman M, Stoddard LT. The effectiveness of fading in programming Memory and Cognition 1984;10:164–78.
simultaneous form discrimination for retarded children. Journal of [17] Warrington EK, Weiskrantz L. The effect of prior learning
Experimental Analysis of Behaviour 1967;10:3–15. on subsequent retention in amnesic patients. Neuropsychologia
[6] Squires E, Hunkin N, Parkin A. Errorless learning of novel 1974;12:419–28.
associations in amnesia. Neuropsychologia 1997;35:1103–11. [18] Schacter DL. Implicit memory: history and current status. Journal
[7] Hunkin N, Squires E, Parkin A, Tidy J. Are the benefits of of Experimental Psychology: Learning, Memory and Cognition
errorless learning dependent on implicit memory? Neuropsychologia 1987;13:501–18.
1998;36:25–36. [19] Graf P, Schacter D. Implicit and explicit memory for new associations
[8] Evans J, Wilson B, Schuri U, Andrade J, Baddeley A, Bruna O, in normal and amnesic subjects. Journal of Experimental Psychology:
et al. A comparison of “errorless” and “trial-and-error” learning Learning, Memory and Cognition 1985;11:501–18.
methods for teaching individuals with acquired memory deficits. [20] Wechsler D. Manual for the Wechsler Memory Scale-III. New York:
Neuropsychological Rehabilitation 2000;10:67–101. Psychological Corporation; 1997.
[9] Parkin A, Hunkin N, Squires E. Unlearning John Major: the [21] Crawford J. The estimation of premorbid intelligence. Ph.D. Thesis.
use of errorless learning in the reacquisition of proper names University of Aberdeen; 1990.
following herpes simplex encephalitis. Cognitive Neuropsychology [22] Brooker BH, Cyr JJ. Tables of clinicians to use to convert WAIS-R
1998;15:361–75. short forms. Journal of Clinical Psychology 1986;42:982–6.
[10] Andrewes D, Gielewski E. The work rehabilitation of a [23] Wechsler D. Manual for the Wechsler Adult Intelligence
herpes simplex encephalitis patient with anterograde amnesia. Scale-Revised. San Antonio, TX: Psychological Corporation; 1981.
Neuropsychological Rehabilitation 1999;9:77–99. [24] Glisky E. Acquisition and transfer of word processing skill by
[11] Craik FIM, Lockhart RS. Levels of processing: a framework for an amnesic patient. Neuropsychological Rehabilitation 1995;5:299–
memory research. Journal of Verbal Learning and Verbal Behaviour 318.
1972;11:671–84. [25] Glisky E, Schacter D, Butters M. Domain-specific learning and
[12] Lockhart RS, Craik FIM. Levels of processing: a retrospective remediation of memory disorders. In: Riddoch MJ, Humphreys
commentary on a framework for memory research. Canadian Journal GW, editors. Cognitive neuropsychology and cognitive rehabilitation.
of Psychology 1990;44:87–112. Hove: Lawrence Erlbaum Associates; 1994. p. 527–48.

You might also like