Content uploaded by Gordon Bower
Author content
All content in this area was uploaded by Gordon Bower on Dec 11, 2016
Content may be subject to copyright.
Narrative stories as mediators for serial learning1
GORDON
H.
BOWER and MICHAL
C.
CLARK, Stanford
University, Stanford, Calif 94305
Subjects learned 12 serial lists
of
10 nouns
by
one
of
two
methods: a control method
ofnormal
study and rehearsal,
ora
narrative-chaining method, where S
was
instructed to
construct a meaningful story woven around the words to be
remembered.
Each
Narrative S
was
permitted
as
much time
as
he needed for constructing each story, typically taking
1-2 min.
His
yoked
control mate
was
given the same amount
of
time to study
each
list. Recall
of
each
list immediately after
it
was
studied
was
perfect for all
Ss.
However, after the 12 lists
had been studied, a probe test for longer-term recall
of
each
list showed Narrative
Ss
recalling six-seven times
as
much
as
their
yoked
Controls. Average median recall
was
93%
vs
/3%
for the Narrative and Control
Ss,
respectively.
The
results
appear to demonstrate the role
of
thematic organization in
increasing leaming, decreasing interlist
in
terferen
ce,
and
guiding reconstructive
recall.
A technique recommended by mnemonists (e.g., Young &
Gibson, 1962) for learning serial lists is the "chaining"
method, whereby S is enjoined
to
construct a narrative story
around the critical words to
be
remembered. The critical
words are to
be
woven
in
to the story
in
the order they are
to
be
recalled, and these words should
be
emphasized
in
some
manner, e.g.,
by
vocal stress, pausing, or by making them the
main actors or objects
in
the narrative. The prescriptions
perrnit a wide latitude in constructive details (e.g., the number
of
critical words per sentence) depending upon the ease
of
organizing the particular list
of
words to
be
leamed. A
common additional prescription is that S should try
to
visualize the scenes he is constructing for linking the successive
words.
Such procedures request that S generate very many
contextual verbal responses. And
it
is
not at
all
obvious why
these should aid memory rather than compete or interfere
with recall
of
the critical words. Further, a prior study by
Jensen & Rohwer (1963) found no effect
of
sentence
mediators (linking words n and n +
I)
upon rote leaming by·
the serial anticipation method. However, the Jensen and
Rohwer study had several features not conducive to showing
verbal mediation effects: (a) successive linking sentences were
unrelated and were provided by E only on the first trial, and
(b) the
Ss
were mental retardates learning a picture series.
For
several reasons, then, our initial study with the chaining
technique
was
done simply to
see
whether it "worked"
efficiently in circumstances for which it plausibly might
be
efficient. These circumstances were (a) self-paced exposure to
the complete serial list, (b) the critical recall units were
content words (nouns), and (c) S had a large number
of
lists to
leam and remember,
so
that massive interference and
forgetting would normally
be
expected for control
Ss
not
using the narrative chaining technique.
METHOD
Each S studied and recalIed
12
successive serial lists
consisting
of
10 concrete nouns chosen to
be
apparently
unrelated.
All
Ss
were run individual1y; they first received
general instructions for the
se
rial leaming task. The Narrative
Ss
were then briefly instructed on the mnemonic technique,
as
folIows: "A good
way
to leam a list
of
items
is
to make up a
story relating the items to one another. Specifical1y, start with
the first item and
put
it in a setting which
will
al10w
other
items
to
be added to it. Then, add the other items
to
the story
in the same order as the items appear.
Make
each story
Psychon. Sei., 1969, Vol. 14 (4)
meaningful
to
yourself. Then, when you are asked to recall the
items, you can simply
go
through your story and pull
out
the
proper items in their correct order."
The Narrative S
was
handed the first list
of
10 words and
told
to
make up his story.
He
did not
have
to
say
his story
aloud, and he could take
as
long
as
he needed.
When
S
was
finished, he handed the list back to E (who recorded the time
taken by S), and then S immediately recalled the seriallistjust
studied. Then the second through twelfth lists were gone
through in the same way.
For
each Narrative
S,
a yoked
Control S
was
ron who received the 12lists in the same order,
each for a study-time equal to that taken by the Narrative
S.
The Control S
was
told simply
to
study and leam each serial
list, and he also did
an
immediate recall
of
each list just after
he had studied it. After the twelfth list had been studied and
immediately recalled, S
was
asked
to
recall the first list again,
then the second list, and
all
subsequent lists. The
cue
for recall
of
a list was the first word in that list; S
was
asked to recall the
remaining nine words
of
that list in their correct order.
The
Ss
were 24 undergraduates fulfilling a service
requirement for their introductory psychology course.
Alternate
Ss
were assigned to
be
in the Narrative
vs
yoked
Control condition. Each pair
of
Ss
received the
12
lists in a
different order within the day.
RESULTS
The times taken by the Narrative
Ss
to construct their story
varied from 40 sec to 199 sec with
agrand
mean
of
104 sec.
Fifty-seven per cent had times between I and 2 min. These
times
grew
shorter over the first four lists,
as
though
Ss
were
becoming more proficient at concocting their stories.
Neither group experienced any difficulty in the immediate
recall test that followed study
of
a list. Median percentages
recalled were 99.9% and 99.1% for the Narrative and Control
Ss,
respectively. However, the differential learning and/or
forgetting for the Narrative
Ss
showed up strongly
in
their
later recalI, when S tried to recall
all
12
lists. The median
percentages
of
words recalled
in
their correct list and correct
absolute position are shown in Fig. I for the two groups for
the
12
lists. There
is
a tremendous difference, with the
Narrative
Ss
recalling
six
to seven times more than their yoked
Controls. There
was
no overlap
in
recall scores
of
the two
100
0
80
w
.J
.J
«
ß60
IX:
I-
ffi40
o
IX:
LLI
0.
20
NARRATIVE
CONTROL
0~~2~--4~~6~~8~~10~~1=2--
LIST
POSITION
Fig. I. Median perceotages
recaIted
o.er
the 12 lists.
18t
Table I
SampIe
Stories
A LUMBERJACK DARTed
out
of
a forest, SKA TEd around a
HEDGE past a COLONY
of
DUCKs.
He
tripped
on
some FURNITURE,
tearing bis STOCKING while hastening toward the PILLOW where bis
MISTRESS lay.
A VEGETABLE
can
be a useful INSTRUMENT for a COLLEGE stu-
dent. A carrot
can
be a NAIL for your FENCE or BASIN. But a
MERCHANT
of
the QUEEN would SCALE that fence and feed the
carrot to a GOAT.
One night at DINNER I had the NERVE
to
bring
my
TEACHER.
There
had
been a FLOOD
that
day, and the rain BARREL was sure
to
RATTLE. There was, however, a VESSEL in the HARBOR carrying tbis
ARTIST
to
my
CASTLE.
groups on any list; the
average
of
the median scores
was
93%
for the Narrative
Ss
vs
13%
for their yoked Controls.
The pieture
is
mueh the same
if
one scores reeallieniently,
eounting a word eorreet regardless
of
the order or intended list
in which it was reealled.
For
Control
Ss
the list words
have
simply beeome unavailable, whereas this has been prevented
in
some way by the narrative-story eonstruetions.
There are small effeets due to list order apparent in Fig.
I,
an
improvement over early lists for Narrative Ss, and a slight
serial-position eurve for Control
Ss.
These are ancillary
fmdings
of
no interest here.
We
next examined the relationship between study-time
on
a
list and later reeall
of
that list. First, for eaeh
S,
reeall
of
the
six
Iists with his longer study times
was
eompared to reeall
of
the
six
Iists with his shorter study times. In this within-S
eomparison, there
was
no suggestion
of
a short
vs
long
difference in recall for either the Narrative or Control
Ss.
This
may
have
been because variation
of
an
S's study times
was
relatively small. Second, over the
12
Ss
by
12
Iists in eaeh
condition, the 144
eases
were divided at the median study
time, and average reeall scores computed for the shorter
vs
longer times. For the Narrative
Ss,
average reeall for the lists
with the shorter study times
was
88%
vs
92%
for the lists with
the longer study times. For the Control
Ss,
average
reeall
was
12%
for the shorter-time and
41
% for the longer-time lists;
these differ significantly, indicating that Control
Ss
yoked
with fast Narrative
Ss
recalled less than those yoked with
slow
Narrative
Ss.
These comparisons reveal that recall
of
Control
Ss
was
affected by study time, while that
of
Narrative
Ss
was
not.
However, this effeet
of
study time on Control
Ss
is
still far
182
from aceounting for the main effeet
of
the narrative
elaboration. (Ineidentally, Control
Ss
always feIt that they had
more than enough time to leam eaeh Iist-until the final recall
tests.)
Stories were taken from a
few
Narrative
Ss
after their final
reeall; a sampIe
of
these are shown in Table I with the
10
critieal words eapitaIized. These have a eertain "stream
of
eonsciousness" sense and unity about them, and they are not
bad solutions to the task
of
connecting
10
unrelated nouns in
a specified order.
DISCUSSION
We
think the effeet in this experiment
is
probably due
to
thematie organization. The person generates meaningful
sentences to relate successive words, and he tries
to
relate
suecessive sentences
of
his generated text around some eentral
theme or action imagery. The sentences and themes from
successive Iists
are
different and probably are kept distinct
from one another in memory. The first-word eue prompts
reeall
of
the theme, and from that the person appears to
reconstruct the sentences and pull out the critical words. The
reeonstruction appears
to
be
hierarehical, from theme
to
sentences to eritical words.
We
would presume that this
thematic organization affeets leaming and that it
also
reduces
interference between the many Iists S
is
leaming. Further
studies
of
this effect could yield more useful information by
recording Ss' total verbal behavior ("thinking aloud") at study
and at recall.
A remarkable aspeet
of
the performance
of
Narrative
Ss
is
that they rarely intruded nonIist words in their reeall (less than
.5
per S). One might first suppose that this discrimination
between critieal
vs
context words
was
based on form class,
since
all
critical words were eonerete nouns. But a glanee at
the sampie stories in Table I shows (a) some eontext words are
nouns, and (b) some eritical words are used
as
verbs or
adjeetives in the stories. The basis for this high level
diserimination between eritieal and eontext words added by S
remains somewhat
of
a mystery.
REFERENCES
IENSEN,
A.
R., & ROWHER,
W.
D.,
IR.
Verbal mediation in
paired-associate and serial leaming.
Ioumal
of
Verbal Learning &
Verbal Behavior, 1963, 1,346-352.
YOUNG,
M.
N., & GIBSON,
W.
B.
How to develop
an
exceptional
memory. Hollywood, CaJifomia: Wilsbire Book Co., 1962.
NOTE
I.
This research was supported by Grant MH-13950
to
the first author
from the National Institute
of
Mental Health.
Psychon. Sei., 1969, Vol.
14
(4)