Acta
Psychologica
53 (1983)
231-241
231
North-Holland
THE EFFECTS OF CONCURRENT SECONDARY THE USE OF IMAGERY IN A FREE RECALL TASK W.A. MATTHEWS
*
Unrversrtg of Southampton,
UK
Accepted
February
An attempt to that encoding effects
dimension
of items during
of high
trials,
can
in the
Paivio’s
be ascribed High-l
of secondary
secondary
Recall
materials.
An
One secondary
of the High-I
items
task. As performance
it was accompanied
view that to the
the superior
fact
that
by improved
impaired
to prevent
compared
investigated of equal
the
numbers
the use of imagery
in the presence
task improved
of the High-l
words
as an additional
which
lists consisting
task was intended
was selectively
of High-I
is available
is reported
of word
on the secondary recall
recall
imagery
experiment
task on the free recall
and low imagery.
acquisition.
inhibiting
to evaluate
words
of two types
ON
1983
was made
of Low-l
TASKS
in some
of the image subjects
over 3
words.
It is widely recognised that the extent to which words are capable of evoking visual images reliably predicts the efficiency with which they will be recalled. This has been fully documented in varied learning and recall contexts by Richardson (1980). Paivio (1971) has argued that such findings are attributable to the fact that imagery provides an additional dimension to the normal verbal/semantic code along which the items can be coded. Thus two codes are available for the processing of high imagery words (High-I), one based upon imagery and the other upon verbal/semantic characteristics, but only a single verbal/semantic code is thought to be available for low imagery items (Low-I). However the fact that ratings of words on the dimensions of imagery and concreteness are highly correlated, allow alternative explanations to that provided by Paivio for the facilitation found in recall. Baddeley et al. (1975) have proposed that reductions in concreteness ratings lead to differences in the encoded verbal/semantic characteristics and that * Author’s
address:
ton SO9 5NH.
W.A.
Matthews.
Dept.
of Psychology,
University
of Southampton,
UK.
OOOl-6918/83/$3.00
‘C 1983, Elsevier
Science
Publishers
B.V. (North-Holland)
Southamp-
imagery
processes
have
little
direct
effect
A number of investigations have elucidate this issue using performance
on either
coding
or recall.
been carried out to attempt to on a concurrent secondary task
during either learning or recall. as a means of minimising the influence of imagery processes. While differing in detail. the studies have been designed to test the hypothesis that the facilitative effects of imagery on learning or recall are reduced when the learner is carrying out a concurrent task which involves visual/imagery processes. However the results of these studies have produced contradictory outcomes. Janssen (1976a. b) reported a small but statistically significant selective reduction in the recall of High-I words when the secondary task required the subject to report which item of a visually presented digit pair was absent. Auditory presentation of the digits produced no such effect. Similar findings were reported by Atwood ( 197 1) who employed the same secondary task as Janssen. However attempts to replicate Atwood’s findings have been unsuccessful (Baddeley et al. 1975). Failures to obtain selective interference in the recall of High-I words have been reported by Baddeley et al. (1975) and Warren (1977). using a pursuit-rotor tracking track as the source of interference. Baddeley et al. reported three experiments. The first two were aimed at establishing that a pursuit rotor task did interact with other tasks that used the visual modality by measuring error rates on the tracking task while carrying out either visual or verbal secondary tasks. It was claimed that the results of these experiments indicated that time on the target declined while carrying out visual tasks and that this established the appropriateness of the task as one that interfered with visual/imagery processes. However there are problems with this view. for in experiments 1 and 2, subjects were required to stay on target for a longer period while carrying out the visual component of the secondary task than the verbal component. Thus task difficulty was not equated and it is a moot point whether the time differences on target are a function of the modality of the secondary task, as is claimed, or of the differences in time required to be on target during the visual and verbal tasks. The third experiment then used tracking as the secondary task during the paired-associate learning of verbal material of varying imagery values. The results of experiment 3 in which subjects learned lists of 15 noun-adjective pairs which were either concrete or abstract (i.e. of High-I or Low-I values) in the presence or absence of the tracking task revealed no selective interference in the recall of the concrete (High-I)
word pairs. In the absence of the tracking task, performance was superior on the concrete (High-I) items (88% correct compared to 38%) but in its presence, performance declined by a similar percentage in both concrete (High-I) and abstract (Low-I) conditions. This failure to obtain an interaction between tracking task and the recall of High I items was repeated by Warren (1977: exp. 3) who studied the free recall of a mixed list of High-I and Low-I words. Baddeley et al. used these results to question Paivio’s claim that dual coding was the basis for the reliably obtained superiority in recall of High-I over Low-I items. They attempted to reconcile their failure to find differential loss of concrete (High-I) items in the presence of the tracking task by suggesting that there are different consequences of the correlated measures on the imagery and the abstractness-concreteness dimensions. It was claimed that the abstractness-concreteness dimension reflected the absence or presence of semantic features that were encoded automatically during learning but that imagery, as it increased. allowed the operation of optional strategies. The facilitation in recall of the concrete (High-I) words could be ascribed to automatically used semantic processes: the failure to find loss of the concrete (High-I) items in the presence of the secondary task was due to the fact that subjects were not using imagery-based strategies. However their failure to find any differential loss of concrete (High-I) items may be due to the inappropriateness of the pursuit-rotor task, as a source of interference. Thus the secondary task used by both Baddeley et al. and Warren may not have prevented the generation and use of images and may have done no more than provide general interference for both High-I and Low-I items. This experiment reports an attempt to throw more light on this issue using the effects of different secondary tasks from those mentioned earlier. The word lists contained 10 High-I and 10 Low-I items and on the assumption that the ‘image’ is a component in the encoded version of each item which is then used to facilitate recall, sought to prevent the formation and/or use of images with the High-I items during presentation. In addition, an assessment of the effects of practice on the secondary task on subsequent recall was included by having 3 presentation trials on each list. The order of the presented items was randomly re-arranged on each trial to inhibit the use of grouping strategies based on serial order or imagery based linking strategies. The characteristics of the secondary task were the critical problems
for the task must be an effective inhibitor of the formation of images during learning; must allow some measure of the accuracy of the performance of subjects; and must allow the derivation of a secondary control task which is as similar as possible to the experimental task in terms of perceptual and motor demands. but does not require the maintenance of image inhibiting activity during the encoding operations on the words. This was accomplished by visually presenting a series of outline shapes in synchrony with the auditory presentation of the items in the word lists. In each sequence there were 4 blocks of 5 related shapes; in some conditions. 2 of the 4 shapes that were presented at 5 set intervals after the presentation of the first slide (which was one of a square, rectangle, triangle or circle) would reconstruct the first shape when they were superimposed. The 4 shapes were designed to be difficult to label verbally and with size and orientation as well as shape being relevant dimensions. the maintenance of the information by image-like representation was necessary. Such images had to be retained throughout the 5 slide series as the second slide of any pair that reconstituted the first shape always occurred in the 5th position. The subject had to indicate a decision (yes/no) after each successive set of 5 slides so giving a measure of performance accuracy. The control task involved the presentation of shapes in exactly the same way. In this condition a small break might occur anywhere in the contour and the subject had to scan the shape, identify the presence or absence of a break and count how many slides in each set of 5 contained such contour breaks. This number, which varied between 0 and 3 had to be written after each successive set of 5 items. Thus the input and output requirements in the secondary tasks were almost identical and each task contained a memorial component; they differed in the characteristics of that memorial component with the experimental condition requiring the retention of images. If, therefore, the first of these secondary tasks utilised image processing capacity, performance on the High-I items should be worse in the experimental condition than in the control condition. Low I items should be little affected (if at all) by these variations, if Paivio’s views on dual coding have some validity. A pilot experiment using 34 subjects suggested that the secondary tasks were effective in the ways suggested above, and following minor modifications to the secondary task, the experiment reported here was carried out.
235
Method Material.7 (u) Lists
Two lists of 20 bisyllabic nouns were constructed from the norms produced by Toglia and Battig (1978). In each list were IO words that were highly rated on both imagery (I) and concreteness (C), and 10 words that were low on both scales. The lists were very similar on familiarity (FAM) ratings. The mean values were: List A:
High-I Low-I List B: High-I Low-I
5.98; 3.54; 5.94; 3.47;
High-C Low-C High-C Low-C
6.18; 3.49; 5.98; 3.59:
FAM FAM FAM FAM
6.25. 5.74. 6.21. 5.78.
The lists were recorded on magnetic tape, in a male voice at a rate of one word every 5 sec. The order of High-I and Low-I items was generally alternating, with occasional variations to destroy a regular pattern, but similar numbers of High-I and Low-I items were present at the beginnings and ends of the lists. There were three presentations of each list with word order being randomised between presentations within the constraints specified above. (h) Slides (i)
Three sets of 20 outline shapes were constructed and prepared as slides for projection by a Kodak Carousel projector. Within each set of 20 were 4 groups of 5 related shapes, with each group constructed on the following principles. The first of the 5 slides contained an outline shape; 4 other slides followed. with each containing an outline of a different component of the first presented shape. In some groups, the combination of two of these 4 outlines with shape, orientation and size relevant attributes reconstructed the first presented shape. In other groups the 4 slides did not contain any outlines which reconstructed the initial shape. Where 2 outlines reconstituted the initial shape, the second of the 2 outlines always occurred on the 5th slide with the other member of the pair occurring randomly in positions 2, 3. or 4. The 4 outlines used in the 1st. 6th, I Ith and 16th position in each set were square, circle, triangle and rectangle with order varying between sets. The components in the other positions were designed to be difficult to describe by a simple verbal label. Within each set, 2 of the 4 groups could produce a reconstructed complete shape and 2 could not. The position of these 2 groups in the set, the shapes that could be reconstructed, and those that were effective in producing a reconstruction were varied between the 3 sets. (Matching task.) (ii) 3 sets of 20 outline shapes were constructed and prepared as slides. The same outlines were used as in b(i) but in some instances, there was a break in the contour of the outline. The break was located in various positions and was never greater than 4% of the length of the contour. The 4 groups of 5 slides that were related to the square, circle, triangle and rectangle were mixed together in this condition; what was varied in
each
group
of 5 slides
was
contour.
The
slides.
(Counting
task.)
(iri)
2 practice
sequences
described
number
in (b) (i) and
40 undergraduates of 20 as part
Each
S had
3 learning
was
presented
outline and
boxes
whether how
was
allowed
was
many
before
the
box
counting
of the error
second
Ss were
told
a subsequent nously
with
square,
circle.
that them
their
item
stood
counting count
in each
task.
on the response with
words
was
a 5 min
of these
were
told
of these sheet given
interval
and
when between
sheet
presentations trials
slide
Ss had
to indicate
of the
5th slide
group.
to indicate
List A wab used
had the
the matching
order
effects
secondary within
of an
a response
5th slide
each
contours.
group
task.
*as
was task
for the
condition
reversed.
carried
from
other.
;I
by
the first
3% of each
An
out
to the
giving
no
to
for
they
slides into
in each
on
auditorily
presented
that
and
they
to decide
for
synchrowc)uld
was
20th
see a
visually
in each
them
shape
but they
projected
and
the signal.
were
to
IO practice
01
Before
the
shape,
to
to v. rite the number
Ss understood
of the task.
size and
by an assistant
prc>Jectcd.
of 5 slides that
2 of
The occurrence
signalled
contour
and
any
slides.
sheet. were
to ensure
lrhcther
not only
15th and
the outlines
sequence
the 2 components
told
the response
broken
gave
presented being
\\ere
account
IOth,
where ;I
items
1at. 6th. 1I th and 16th positions.
of 5 slide5
condition
initial
to the Lvords being
space
the assistant
in each
the
group
to the visual
taking
to look
each
(O--3) after
were
of the 5th.
close
after
the projection
transfer.
shape.
occurred
of the 3 matching
transfer
folIoming
blocks
each Lvord
each
other
four
of the group,
after each
representing
on each
in the appropriate
Ss were
how many
group5
conditions.
trial
rows
in one of the
the presentation
in front
principles
learning
with
One
the matching the
On each
task
5’ was
broken
to attend
test,
the critical
After
a tick or a cross
the last who
to scan
reconstructed
was
or rectangle
was
of 5
1x35 required uaa
had
the
counting
of the
condition.
asymmetrical
recall
Before
triangle
task
the
presentation
a digit
The decrements
task
free
and
reconstructed
of asymmetrical
their
the words.
orientation. enter
that
in the
group
as Ss in 2 separate
sheets
alternative
had
decrements
position.
written
using
front
During
to enter
for of
rate
to the possibility
In
of 5 slides
presence
presentation
support
slides
condition;
comparison
this
a Yes/No
Ss had
in the group
of
with
were 4 boxes
list B for the counting
assessment
acted
test.
5 sec.
of 5 slides.
under
trials
the
a break
3 in each
constructed
response
the recall
for
of which
blocks
in each
shapes and
contained 0 and
of the matching
separate
for
visible
in each
the counting
matching
each
on
2 of the 4 presented
During
were
of Southampton
under
recall
representing a tick
which
between
requirement.
free
which
of 6 rows
to place
each
for 5 sec. Synchronised
shape
consisting
IO slides
of
trials
set
60
of figures varied
of the University
written
presentation.
number instances
(b) (ii) above.
of a course
Immediate
the
of these
slidus the task.
paired There
Thus in the matching task, visual information that could not be adequately verbalised had to be retained for use at 4 points during the presentation of the word lists of High- and Low-l items, whereas in the counting task retention of visual information was not required during the presentation of a similar list, but the observation requirements and temporal constraints were identical and response requirements almost so.
Results
and statistical
treatments
Earlier work reported by Richardson (1978: exp. 3) has indicated that facilitative effects of imagery are absent in the free recall of recency items (Waugh and Norman 1965; Tulving and Colotla 1970). Estimating the size of the recency component in these 20 word lists was carried out using 3 methods. Method 1 employed the procedures and values specified by Tulving and Colotla (1970) who proposed that an item was retrieved from the recency component if no more than 6 items intervened (as either input or output) between presentation and recall. In the 2 remaining methods, the recency component was estimated using the criterion of either the last 3 presented items recalled in any of the first 3 positions of the recall protocol (Method 2). or the last 2 presented items recalled in the first 2 positions in the recall protocol (Method 3). The general pattern of the derived data was similar using all methods; however the 2 latter methods are likely to include larger numbers of recency items which should remain unaffected by the secondary task so that trends in the data indicating the effects of secondary task on High-l items may become increasingly obscured as one moves from Method 1 to Method 3. Split-plot analyses of variance (Kirk 1968) were conducted on
5Mean number
of
4-
words
recalled 3-
2-
l-
~
I
-
Matchmg Counting
1
Tl
T3
T2 TEIIS
Fig.
1. Mean
Method
number
I: recency
of words items
recalled
excluded.
by trial
and
condition
(40 subjects).
Data
produced
by
the 3 forms very
of the data.
similar
based
and
on data
The
mean
pattern
scores
Significant with
effects
cases).
However,
effects
and
p < 0.01)
type
that
between
type
decreasing
were
I and task,
the
are
not
in fig.
analyses reported.
1. (A similar
trials
3. The
on the task
main
effects
significant
order
than
of
in both the main
( F( 1,38) = I 1.54,
I was significant
was
of
rather
more
p i 0.01)
> ~0.05
= 7.97. p < 0.05).
trials
with
= 196.52.
( F values
interactions
and
(F(2.76)
I and
p < 0.01)
( F(2.76)
for trials
not significant
is centred
of secondary
between
and
I( F( 1.38) = 21.24.
and
1 to trial
trial
secondary
effects
1 are presented
for
items
task
interest
between
as was
from
of secondary
I and 2 w’ere
by Methods As expected.
experiment.)
obtained Low-I
derived
reported.
use of Method
in the pilot
than
primary
that
interaction
the
improving
and
are
3 showed
were
recalled
performance
presentation,
by
on the data
I data
by Method
derived
being
based
Method
was obtained
main
items
from
derived
of results
High-I
The results
those
Finally.
(F(2.76)
the triple = 7.45.
p <
0.01). Performance counting
on the High-I
condition,
effects
focused
trials.
on
On trial
this
the
effect
I recall
items
in the matching
difference
decreased
of the
of the High-I
3. performance
was
better
15.05, p < 0.05).
With
the counting
items
on every
the Low-I
respectively, A
analysis
of
these
to an understanding
Low-I
items.
occurred
One
because
High-I
and
worse
if this were
performance
on
becoming
the
number
the case
a post
was
6.10
(Max
errors
was
3.25.
However.
already
rates.
A further
than
in the over
of the Low
on trial
2 but on
(F(1.76 =
items
superior
main
items
that
Low-I
18.80
on the High-I
on trials
I. 2 and
3
there
which
is
into
Ss within
group
rate
It is clear
each data
by type mean from
ment
in the recall
while
carrying
out
showing
and
the figure
little task
this
that
task
should
practice
make
improved
processing
a positive
items
resources
relationship
between
out based
on
group
the mean
number
of
in the other
group
the mean
number
of
main
items
the counting
task.
mean
by
replicated
group
showed
of error
was carried
being
6.0. 4.0 and
2.3
An analysis
of
and
interaction
recall
effects errors
scores.
differences of the data
3 with
task
rates
in error
I by trials
arrangement
the significance
or interactive error
significant
by trial
was carried
of matching
variation
the low error
of the High-I
2 blocks
of variance
of 7 with
the earlier
from
High-I
on the learning
In one
no significant
of secondary
scores
is that
to additional
task.
the analysis
3 groups
repeated
on the
in the recall
produce
of Ss as a function
21 Ss arranged
trials
of the High-I
concentrated
possibility
of Ss into
3 trials)
were
rearrangement
and
scorers
leading
information
efficiency.
although
and
The
thus
on the matching
= 12 over
reported,
on these
the higher
some over
task
so this should task
out using
error
Low-I
in the acquisition
improvement
hoc rearrangement made
errors
provide
involved
An alternative
task,
secondary
of errors
levels
then
task.
If this were
and
Accordingly,
p c 0.01).
the
than
of simple
difference
than
was significantly
may
is that
secondary
available.
efficiency
variance
recall
out of the secondary
on the secondary
errors
with
items
(F(1.76) = 18.80. 22.98,
changes
possibility
Ss opted
most
cant
on
was significantly
of the processes
component;
recall
task,
trial
worse
A test
task
High-I
was
trials.
p < 0.01).
detailed
relevant
the
items
on the
than
and
matching
condition over
(F(1.76) = 14.19, p < 0.05). There was no significant
I items trial
but
added (F(4.36)
a signifi= 40.00.
are presented
in fig. 2.
a substantial
improve-
at least
as efficient
as it was
239
2(a) Matching
condltlon
,
6-
5-
of words recalled
4-
3-
21 LOW -1
~ l-
High
-
-
- - - -
TI
numbers
M&urn Low
of errorS
number number
of errors
of errors
TZ
2(b)
Counting
T3
condltlon
7-
6-
5Mt?W number of words recalled
4-
3-
2”
__ l-
-
High -
- - - -
Tl
LOW -1 number
Medium Low
of errors
number
number
of errors
of errors
T2
T3
TWIS
Fig. 2. Mean subjects). counting
Data
number
of words
produced
condition.
recalled
by Method
by trial. I: recency
condition. items
and
excluded.
error 2(a):
rate
on matchmg
matching
condition;
task
(21 2(b):
Discussion This experiment has demonstrated that High-I words are better recalled than Low-I words when presented in a mixed list with the subject simultaneously carrying out a secondary visual task which does not require the retention of images for its completion (i.e. the counting task). It could be argued that in its essentials, the counting task had similar characteristics to the pursuit-rotor tracking task used by Baddeley et al.. in that visual scanning of the presented stimuli was required and a response had to be made. albeit intermittently. rather than continuously. What these secondary tasks did not require was the generation and maintenance of images, and the fact that selective interference with High-I words failed to appear in both cases provides support for the view that the task used by Baddeley et al. was simply inappropriate for the purpose, and that their results have no direct bearing on the issues raised by Paivio’s dual coding hypothesis. Allport (1980) has suggested that there is a need to specify the components of the cognitive system in a far more precise way than has been attempted (as well as emphasising the difficulties of investigating such systems) and it seems reasonable to expect that any investigation of the suppression of imagery processes using secondary task techniques should involve imagery processes and not just the visual system in both tasks. When this was done in the matching task in this experiment. so that images had to be retained during the presentation of the High-I and Low-I words, a selective loss of the High-1 items in the non-recency part of the list was obtained, and prevented the appearance of the recall superiority of High-I items that is normally found without such interference (e.g. Warren 1977; Baddeley et al. 1975). This result lends support to the view that the availability and use of images is at least one component in the acquisition and recall of High-I words. However, this experiment included three trials in each condition to allow the time course of such interference to be investigated. When these data were considered along with the error rate on the secondary matching task, it became clear that in addition to the expected improvement over trials on all items, there was a differential improvement in the recall of the High-I items as the error rate on the matching task decreased. Thus, in addition to confirming the earlier demonstration of selective interference by Janssen which was shown on a single presenta-
tion
and
recall
trial,
this
pattern
of performance
can persist
over three
indicates
with imagery
on the petence
secondary task decreases, indicating a higher degree on that task, so recall of High-I words improves. This
that capacity becomes increasingly dimension as error rate declines
trials.
that
ference
However.
inter-
as error
rate
of comsuggests
available to use an image encoding on the secondary task and provides
support for a view which claims that imagery that acquired from a verbal-semantic code recall.
based information adds to to facilitate learning and
References Allport.
D.A.,
London: .Atwood.
1980.
‘Attention
Routledge
G..
and
and
Kegan
performance’.
In:
G.
Claxton.
rd..
Cognitive
psychology.
Paul.
1971. An experimental
study
of visual
imagination
and
memory.
Cognitive
Psychol-
and usual
-orking
memory’.
V. London:
Academic
ogy 2. 239-289. Baddelq. In:
A.D.. P.M.A.
S. Grant. R&bitt
E. Wight and
S.
and N. Thomson,
Dornic.
197.5. ‘Imagery
eda.. Attention
and
performance.
Pre.\s. Janswn.
W.H..
the image. Janssen.
W.H..
Journal Kirk.
1976a. Selective Acta
1976b.
Selective
of Experimental
R.E..
1968.
interference
Paychologica
interference
Psychology
Experimental
in pared
asboclate
and free recall
learning:
messing
up
40. 35-48. during
the
rrtrleval
of
visual
images.
Quarterly
28, 535- 539.
design:
procedures
for
the
behavloural
sciences.
Belmont,
CA:
Brooks/Cole. Paivio.
A.. 1971. Imagery
Richardson.
J.T..
memory. Richardson,
197X.
Quarterly J.T..
and verbal Mental
Journal
Toglia.
M.P. and
Tulving.
E. and
V. Colotla.
Warren.
M.W..
1977. The effects
recall. Waugh.
Memory N.C‘. and
1970. Free
the
recall
Holt.
distinction
Psychology
and human
1978. Handbook
and Cognition D. Norman.
imagery
New York:
and
of Experimental
1980. Mental W. Battig.
processes.
imagery
of trilingual
of recall-concurrent
and Wmston. primary
and
secondary
30. 47 l-485.
memory.
of semantic
Rinehart betureen
London:
MacMillan.
Hard
norms.
lists.
Cognitive
vi.\ual-motor
Hillhdale.
NJ:
Psychology
distraction
Erlhaum. I, 86-89.
on picture
and uord
5. 362-370. 1965. Primary
memory.
Psychologxal
Review
72. X9%104.