Вы находитесь на странице: 1из 8

Psychology and Aging 2013 American Psychological Association

2013, Vol. 28, No. 4, 902909 0882-7974/13/$12.00 DOI: 10.1037/a0033263

Does Aging Affect Recall More Than Recognition Memory?


Stacey L. Danckert and Fergus I. M. Craik
Rotman Research Institute at Baycrest, Toronto, Ontario, Canada

Although it is generally agreed that recall performance declines more than recognition memory perfor-
mance in the course of normal aging, there are some dissenting voices. There are also a few empirical
findings that cast doubt on that conclusion. In light of these ambiguities the present experiments were
designed to answer the question in a more definitive fashion. Over a series of 3 experiments, groups of
younger and older adults performed recall and recognition tests successively on the same lists of words.
Several analyses of the resulting data converge on the conclusion that there is a consistent age-related
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

decrement in recall that is disproportionately greater than the age-related decrement in recognition. This
This document is copyrighted by the American Psychological Association or one of its allied publishers.

conclusion is in line with several theoretical accounts of age-related differences in cognitive processing
and also with emerging evidence from cognitive neuroscience.

Keywords: recall, recognition, aging, verbal memory, successive testing

It is widely accepted that the processes of recall and recognition and resource demanding than those involved in recognition, and
memory differ in significant ways. At the cognitive level, one that older adults are less able to muster and deploy such processing
influential view holds that whereas recognition can be accom- resources. A further age-related difficulty with recall tasks may be
plished by either of two independent processes, familiarity and associated with a declining ability to self-initiate retrieval pro-
recollection (Mandler, 1980), successful recall depends largely on cesses, although this problem may be circumvented by the provi-
conscious recollection (Jacoby, Toth, & Yonelinas, 1993). There is sion of increased environmental support during recognition testing
also good evidence that across the adult life span age-related losses (Craik, 1983, 1986).
are greater for recollection than for familiarity (Jennings & Jacoby, Other reports are less supportive, however. Uttl has suggested
1993, 1997), and this result suggests that age-related declines in that the small effects of aging on recognition memory may reflect
memory performance should be greater in recall than in recogni- the fact that recognition tests are relatively easy and so result in
tion tasks. There is evidence to support this conclusion (Craik & ceiling effects for younger participants; a meta-analysis confirmed
McDowd, 1987), but the question of whether the memory impair- this suspicion, yielding the conclusion that age-related declines in
ments associated with aging are more strongly linked to losses in recognition memory are larger than previously thought (Uttl,
recall still suffers from conflicting results. Henry, & Baltimore, 2007). Other studies have also suggested that
In support of the view that age-related memory impairments are recall and recognition fall off proportionately with age. For exam-
less in recognition than in recall, Schonfield and Robertson (1966) ple, Baddeley (1996) reported results on the Doors and People
reported a study that found substantial age-related drops in recall test (Baddeley, Emslie, & Nimmo-Smith, 1994), which yields
but relatively small losses in recognition memory. Craik and recall and recognition scores for visual and verbal materials. He
McDowd (1987) confirmed the finding of greater losses in cued recall tested groups of middle-aged and elderly participants and found
than in recognition with increasing age; they also used secondary task essentially equivalent age-related declines for recall and recogni-
performance as a measure of resource demands and found greater tion. It should be noted that Baddeley transformed the raw scores
secondary task costs for recall compared with recognition, and that to z-scores, however, and that this may yield a different conclusion
recall costs were differentially greater for older adults. This last from an analysis based on raw scores (see also Kopelman et al.,
result implies that the processes of recollection are more effortful 2007; Park et al., 2002). It is possible that recognition and cued
recall are associated with reduced amounts of variance compared
with free recall, in which case age-related declines in the various
measures would appear more similar when expressed as z-scores.
This article was published Online First August 26, 2013. Given these ambiguities, it seemed worthwhile to examine the
Stacey L. Danckert and Fergus I. M. Craik, Rotman Research Institute at question of whether memory problems in older adulthood can be
Baycrest, Toronto, Ontario, Canada. linked differentially to losses in recall and recognition in a slightly
The work reported was supported by a grant from the Natural Sciences different way. Specifically, the present studies were designed to
and Engineering Research Council of Canada to Fergus I. M. Craik. We are assess relative age-related declines in recall and recognition under
grateful for helpful comments on the manuscript from Paul Duberstein and
a variety of encoding conditions. In addition, we attempted to
two reviewers, from colleagues Ellen Bialystok, Nigel Gopie and Nathan
Rose, and for technical help from Lindsay Delima, Brittany Fredericks,
ensure that ceiling effects could not account for the results (Uttl et
Karen Lau, Waqar Qadir, and Sandra Oziel. al., 2007). One way in which the present procedures differed from
Correspondence concerning this article should be addressed to Fergus most previous experiments of this type was that recall and recog-
I. M. Craik, Rotman Research Institute, Baycrest Centre, 3560 Bathurst St., nition tests were performed successively on the same lists. In this
Toronto, ON, Canada M6A 2E1. E-mail: fcraik@research.baycrest.org way we kept both materials and participants constant across tests
902
RECALL, RECOGNITION, AND AGING 903

and were able to explore various contingent relations at the single Materials. Two sets of 10 lists of 20 common nouns were
word levelfor example, given that a word was recognized by used. Half of the participants studied one set of lists and the other
participants, what was the probability that it was also recalled? half studied the other set of lists. In addition to the 200 studied
words, the recognition test featured 200 lure words, making a total
Experiment 1 of 400 words that were randomized into 10 additional lists of 40
words. Two versions of the recognition lists were also constructed,
This experiment was designed to evaluate age-related changes with half of the participants receiving each version. Each encoding
in recall and recognition. Recall and recognition tests were per- and recognition list was matched for word frequency using the
formed on the same list of words, which allowed calculation of the KuceraFrancis frequency norms (Kucera & Francis, 1967). Fi-
conditional probability of recalling a word given that it was rec- nally, word order in each presentation and recognition list was
ognized. In addition, we used a large set of items with the aim of randomized separately for each participant.
keeping overall recognition scores off ceiling. Procedure. Words in the study lists were presented visually at
a 3-s rate, and each list was followed by a free recall test and also
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Method a recognition test, as described below. After presentation of each


This document is copyrighted by the American Psychological Association or one of its allied publishers.

study list, participants were given an interference task in which


Participants. A total of 48 individuals participated in the
they counted backward by three from a random 3-digit number for
experiment after providing informed consent, 24 older adults (18
30 s to eliminate recency effects. This was immediately followed
females) ranging in age from 60 to 80 years (M 69.54 years) and
by the recall test in which participants were given 60 s for written
24 younger adults (21 females) ranging from 18 to 35 years (M
free recall of words from the preceding list. Participants were
20.71 years). All participants were fluent English speakers and
informed that they had 1 min for recall; in the very few cases in
received either monetary compensation or course credit for their
which they were still writing at the end of this period they were
participation. The younger participants were all students at the
allowed to complete writing any recalled words they had still not
University of Toronto. The older participants were community-
written down. This procedure was repeated for all 10 lists. Fol-
living adults who traveled in to our lab to take part in the study;
lowing all 10 presentation and recall phases, participants com-
they were drawn from a pool of volunteers maintained by the
pleted a 5-min questionnaire conducted for another purpose; this
Psychology Department at the University of Toronto. In terms of
provided a break between the recall and recognition phases.
occupation, the 24 older participants had worked at a variety of
The recognition tests were then administered. Ten lists of 40
professional (19) or technical/secretarial (five) jobs and 22 were
words were presented on the computer screen in a prerandomized
now retired. Occupations included managerial/financial/marketing
order. The 200 words from the presentation lists were mixed with
(eight), secretarial (five), and teaching (four); the group did not
200 new lure words, and the resulting 400 words sorted into 10
include any retired university faculty members. With regard to
lists, each containing 20 targets and 20 lures. For each list partic-
health, only two of the young adults were taking medications on a
ipants were asked to indicate any words they recognized as having
regular basis (birth control pills). In contrast, 20 of the older adults
been on one of the study lists by pressing the z or m key, which
were on some form of medication, including long-term prophy-
were labeled yes and no, respectively, on the keyboard. To reduce
lactic medications for blood pressure (eight), cholesterol (eight),
both stress and boredom, recognition was self-paced. Thus, par-
bone disease (five), prostate (two), and arthritis (nine). All partic-
ticipants first recalled and then recognized the same lists of 20
ipants also rated their current health on a 10-point scale, where 1
words. Finally the participants completed the Shipley vocabulary
was poor and 10 was excellent; ratings for the young and old
test to provide a measure of verbal ability. The experiment lasted
groups were 8.2 and 7.8, respectively, t(46) 1.00, p .32. The
approximately 60 min.
older adults had received 15.3 years of education on average and
had an average score of 36.7 on the Shipley vocabulary test
Results
(Zachary, 1986). The younger adults had received 14.9 years of
education on average, and their mean Shipley score was 31.5. Thus Table 1 displays the means for recall and recognition as a
the older group was rather high functioning, with somewhat more function of age. For recognition, the table shows hits, false alarms,
education and higher vocabulary scores than the younger group. and corrected recognition (hits false alarms). The salient aspect

Table 1
Experiment 1: Probabilities of Recall and Recognition as a Function of Age for Both Full and Matched Data Sets

Recognition
n Recall Hits False alarms Hits false alarms RC|RG

Full data set


Young 24 .49 (.16) .80 (.11) .12 (.12) .68 (.20) .59 (.14)
Old 24 .29 (.10) .76 (.11) .21 (.14) .55 (.15) .36 (.13)
Matched data set
Young 18 .46 (.15) .76 (.11) .15 (.12) .62 (.20) .56 (.14)
Old 18 .31 (.10) .80 (.08) .19 (.11) .61 (.12) .37 (.13)
Note. RC|RG column refers to the probability of recalling items given recognition. Mean probabilities are shown in bold type; standard deviations are
in parentheses.
904 DANCKERT AND CRAIK

of Table 1 is that the age difference in recall scores (0.20) was led to greater interference from previous lists. This point was
larger than the corresponding age difference in corrected recogni- addressed in Experiments 2 and 3 by reducing the number of
tion scores (0.13). A 2-way analysis of variance (ANOVA) involv- words to be encoded. One other possible concern in Experiment 1
ing age group (young, old) and memory test (recall, recognition) stems from the fact that participants were given only 1 min for
was carried out on the means in Table 1. The analysis yielded main recall, whereas recognition was unpaced; might this have penal-
effects of age, F(1, 46) 16.04, p .001, 2 0.26, and memory ized the older participants in particular? The mean number of
test, F(1, 23) 173.96, p .001, 2 0.79, showing that young words recalled by the older adults was 5.8 (SD 2.0); 1 min is
adults scored higher than older adults and that recognition was sufficient time to write down at least 10 words (young participants
higher than recall. There was also a significant Age Memory recalled 9.8 words on average, with a range of two to 18; the range
interaction, F(1, 46) 5.79, p .05, 2 0.11, showing that the for old participants was zero to 14). In addition, most participants
age difference in recall was greater than the age difference in stopped writing after 30 45 s, and no one complained that they
recognition. needed extra time. Nonetheless, it is possible that older partici-
One advantage of the present design, in which participants pants felt under time pressure, so longer recall times were used in
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

attempted to recall and recognize the same lists of words, is that it the next two studies. Finally, cross-sectional experiments are al-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

is possible to calculate conditional recall probabilities for specific ways vulnerable to the concern that some other factor apart from
words that were later recognized. Given that a word is correctly the age difference is at least partly responsible for the results. We
recognized, what is the probability that it was also recalled, and are argue that the present two samples were comparable in terms of
there age-group differences in these probabilities? The overall education, health, occupation, and socioeconomic status (see Par-
conditional probabilities for recall given recognition are 0.59 for ticipants section); however, the point is discussed more fully in the
the young group and 0.36 for the old group, a highly significant General Discussion. Experiments 2 and 3 were originally designed
difference, t(46) 5.65, p .001. It is clear from these data that in part to explore age-related differences in encoding strategies,
the ability to recall recognized words is substantially poorer in but the results are reported here in the context of differences
older than in younger adults, despite the fact that recognition levels between recall and recognition.
themselves did not differ markedly.
However, it is still possible that ceiling effects mask group
Experiment 2
difference in recognition scores; for example, five younger and
three older participants had hit rates of 0.89 or higher. To over- Experiment 2 retained the feature of having younger and older
come this possible problem and to equate recognition scores be- participants recall and then recognize the same lists of words, but
tween the groups, we repeated the analyses after removing the six reduced the number of study lists to two in the expectation that this
highest performing younger adults and the six lowest performing would reduce the age difference in recognition scores. The exper-
older adults, based on overall accuracy (hitsfalse alarms). Table iment was also designed to examine age differences in memory
1 shows that recognition hit rates were now slightly greater in the following an incidental deep encoding task compared with an
old group (0.80 vs. 0.76) and that the corrected recognition scores intentional learning task; previous work by Troyer, Hfliger,
were virtually identical in the two groups (0.62 and 0.61 for young Cadieux, and Craik (2006) had shown that a deep processing task
and old participants respectively). Consistent with the full data set, at encoding had the effect of eliminating age differences in rec-
the 2-way ANOVA comparing age group and memory test yielded ognition. In this study we therefore compared two encoding types:
a main effect of memory test, F(1, 34) 156.08, p .001, 2 intentional learning and incidental deep encoding, again with sub-
0.821, showing that recognition was higher than recall. There is a sequent recall and recognition tests on the same word lists.
trend toward an effect of age, F(1, 34) 3.09, p .09, but a
significant Age Memory interaction, F(1, 34) 15.62, p
Method
.001, 2 0.315, signifying an age difference in recall despite the
elimination of the age difference in recognition. This was sup- Participants. The study included 25 younger adults (17 fe-
ported by the tests of simple effects; t(34) 3.66, p .001, and males), aged 18 25 years and 28 older adults (16 females), aged
t(34) 1.0, for recall and recognition, respectively. 60 80 years. One young male participant was excluded based on
performance (16% correct on the recognition test) and four older
participants were also excluded (one male, three female), one for
Discussion
extremely low performance (zero items recalled), and three for
These data confirm the claim that age-related decrements in failing to complete the task. The remaining samples were thus 24
recall are considerably greater than age-related decrements in in each age group; mean ages were 20.9 years and 70.0 years for
recognition memory. When we eliminated the difference in rec- the young and old groups, respectively. As in Experiment 1,
ognition while also removing the possibility of ceiling and floor participants were drawn from the University of Toronto under-
effects, the differences in recall were still substantial. These data graduate subject pool and the University of Toronto adult volun-
thus contradict the suggestion that ceiling effects in recognition teer pool. Younger and older adults had completed 14.7 and 15.7
have masked the true picture of parallel declines in recall and years of education, respectively. Regarding occupation, all of the
recognition with age. Although there are disproportionate deficits younger participants were university students. In the older group,
in recall, small age-related differences were still present in the 21 had worked in professional occupations and three had worked
recognition data. It is thus possible that the study design led to in secretarial/technical positions; 21 of the 24 were now retired.
decreased recognition scores for the older adults owing to the large With regard to health status, 21 of the younger participants were on
number of lists they were required to remember, and this may have no regular medications, of the remaining three, two were taking
RECALL, RECOGNITION, AND AGING 905

birth control pills and one was taking medication for an ADHD sizable age-related decrements in recall. We conducted a 2 2
condition. In the older group, only two participants were taking no 2 mixed-design ANOVA including Encoding Task (deep, learn)
regular medications; 15 were taking preventative medications for Memory Test Type (recall, recognition; both within-subject vari-
potential or actual circulatory problems, six for arthritis, five for ables) Age (young, old). This analysis yielded main effects of
cholesterol, four for depression or anxiety, and two for asthma. age, F(1, 46) 4.49, p .05, 2 0.089, and test type, F(1,
Mean scores on the 10-point self-rated health scale were 8.0 for the 46) 294.67, p .001, 2 0.865, demonstrating that younger
young group (n 22) and 7.8 for the old group, t(44) 1.0. adults performed better overall than older adults and that recogni-
Materials. Two 18-word study lists were used in the experi- tion scores were higher than recall scores. The encoding task main
ment plus a list of 144 words for the recognition test, making a effect was not significant, but the Encoding Task Test Type and
total of 180 common concrete nouns. We used the Cattell Culture
Age Test Type interactions were both significant; F(1, 46)
Fair Intelligence Test (Cattell & Cattell, 1960) and the alpha span
14.47, p .001, 2 0.239, and F(1, 46) 18.95, p .001, 2
task (Craik, 1986) to assess nonverbal IQ and working memory
0.292, respectively. Table 2 shows that the Encoding Task Test
capacity, respectively.
Type interaction reflects the finding that deep encoding was su-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Procedure. Participants were first asked to attend to one of


perior to intentional learning for recognition performance, but
This document is copyrighted by the American Psychological Association or one of its allied publishers.

two (counterbalanced) lists of 18 words (displayed serially on a


computer monitor at a rate of 4 s per word) and to either perform intentional learning was somewhat superior for recall. In addition,
a deep encoding task (Indicate for each word whether it is younger adults performed significantly better than older adults on
something that is pleasant or unpleasant to you) or to simply recall whereas there were no age differences on the recognition
memorize it for a forthcoming memory test. Immediately follow- test. Using t tests to further clarify these effects, younger adults
ing the encoding task, participants were asked to count backward showed an advantage over older adults on both deep and memo-
by 3 s for 30 s (to eliminate recency effects) and were then given rized recall tasks, t(46) 4.22, p .001, t(46) 2.36, p .05,
up to 3 min to write down all words they could recall. They then respectively, whereas there were no significant differences be-
performed the second encoding task and recall test. Following both tween the groups on the recognition tests, t(46) 0.69, p .49,
recall tests, they performed a self-paced recognition memory test t(46 0.18, p .86, respectively). Age-adjusted scores on the
in which they were asked to identify those words that had previ- Cattell Test were 27.5 and 25.6 for younger and older participants,
ously appeared on the encoding lists from a total of 180 serially respectively, t(46) 1.44, p .16, and scores on the alpha span
presented words. Finally, participants completed the Cattell Cul- test were 29.9 and 23.9 for younger and older adults, respectively,
ture Fair Intelligence Test which tests nonverbal IQ through a t(46) 3.04, p .01. Thus, the age groups were equivalent on the
series of visual problem solving tasks. The alpha span task was Cattell measure of fluid intelligence, but younger participants were
then given; short lists of words are read aloud and the participant superior on verbal working memory.
must repeat the words back in correct alphabetical order. Two lists The overall conditional probability for recall given recognition
are provided at each list length, progressing from two to eight was 0.56 for the young group and 0.43 for the old group following
words per list, and the task ends when neither list at a particular
deep encoding, and 0.62 for the young and 0.47 for the old group
level is correctly recalled. The score taken was the total number of
following intentional learning, a significant difference for both
words recalled in the correct position.
deep and intentional learning conditions, t(46) 3.48, p .005,
t(46) 2.81, p .01, respectively. It is clear from these data that
Results the ability to recall recognized words is substantially poorer in
Table 2 shows the means for each condition as a function of age, older than in younger adults, despite the fact that recognition levels
encoding task and memory type. In this study, age differences in themselves did not differ; the mean corrected recognition scores
recognition memory were small to nonexistent, but there were still were 0.71 and 0.70 for the young and older groups, respectively.

Table 2
Experiment 2: Probabilities of Recall and Recognition as a Function of Age and Encoding Type for the Full Data Set

Recognition
n Recall Hits False alarms Hits false alarms RC|RG

Full data set


Youngdeep 24 .51 (.15) .82 (.10) .07 (.08) .76 (.11) .55 (.12)
Younglearn 24 .52 (.18) .74 (.17) .07 (.08) .67 (.17) .65 (.18)
Olddeep 24 .35 (.12) .82 (.15) .09 (.08) .73 (.15) .41 (.16)
Oldlearn 24 .40 (.17) .77 (.19) .09 (.08) .68 (.19) .45 (.17)
Split data set
Younghigh 12 .60 (.12) .86 (.07) .04 (.03) .81 (.06) .60 (.11)
Younglow 12 .44 (.08) .71 (.10) .09 (.10) .61 (.07) .59 (.13)
Oldhigh 12 .44 (.12) .92 (.06) .08 (.06) .82 (.08) .48 (.13)
Oldlow 12 .32 (.11) .69 (.15) .11 (.10) .58 (.11) .43 (.13)
Note. RC|RG column refers to the probability of recalling items given recognition. Data are also included for the partial data set, combined across
encoding conditions. Mean probabilities are shown in bold type; standard deviations are in parentheses.
906 DANCKERT AND CRAIK

To address concerns that the interaction between age and test Experiment 3
type may reflect a ceiling effect in young recognition scores, we
This experiment provided two optimal techniques for learning
split the data into high and low performing groups based on the
word lists, including one that encouraged organization and one that
results of the recognition tests. To accomplish this, we calculated
focused attention on each word individuallysimilar to a deep
each participants mean recall and recognition scores collapsed
encoding task, albeit with intentional learning. These two tech-
over deep and learn conditions, and then split each age group into
niques were suggested to different groups of younger and older
the 12 highest and 12 lowest scorers on the recognition test (hits adults, and the learning phase was followed first by recall and then
false alarms score). The resulting means are shown in Table 2. The by a recognition test for the same list of words. Participants were
recognition hit rates for the low groups are 0.61 and 0.58 for young given a brief practice session on the provided technique. We
and old participants, respectively, well below ceiling. An ANOVA expected that this manipulation would increase the consistency of
on all recall and adjusted recognition scores in Table 2 (both low the method used for memorization across participants.
and high groups) showed significant effects of performance (high,
low), F(1, 44) 55.52, p .001, 2 0.558, memory type Method
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

(recall, recognition), F(1, 44) 350.1, p .001, 2 0.888, and


This document is copyrighted by the American Psychological Association or one of its allied publishers.

Participants. Participants included 41 younger adults, rang-


age, F(1, 44) 9.71, p .005, 2 0.181, signifying higher ing from 18 26 years and 40 older adults aged from 65 81 years.
scores for high groups, recognition, and younger participants, Participants were drawn from the University of Toronto under-
respectively. The Age Memory Type interaction was reliable, graduate subject pool and the adult volunteer pool. The data from
F(1, 44) 22.5, p .001, 2 0.339, but the three-way one older adult was removed from the study for very poor perfor-
interaction between age, test type, and performance was not, F(1, mance (one item recalled) and one younger participant was re-
44) 2.14, p .15, 2 .05, showing again that recall scores fell moved for failure to follow instructions. The mean age of the
more than recognition scores from young to old participants, and young adult group (n 40, F:M 24:16) was 20.7 years, and the
that this effect was found in both high and low subsets of the data. mean age of the old adult group (n 39, F:M 32:7) was 72.8
Finally, the Memory Type Performance interaction was reliable, years. The average number of years of education was 14.1 years
F(1, 44) 8.51, p .01, 2 0.162, reflecting the fact that for the young participants and 16.3 for the older participants. With
recognition scores fell more than recall scores from high to low respect to occupational status, all of the younger participants were
subsets. A further ANOVA on the low subsets only showed that students at the University of Toronto. In the older group, 33 were
the Memory Type Age Contrast was again significant, F(1, now retired and six were semi-retired; their occupations included
22) 4.56, p .05, 2 0.172. Subsequent t tests showed that 28 professional positions and 11 secretarial/technical positions. In
there was a significant effect of age on recall scores, t(22) 3.13, terms of health data, 32 of the younger participants reported no
p .005, whereas the effect of age on recognition was nonsignif- regular medications; the remainder reported being on birth control
icant, t(22) 0.83, p .42. pills (three), medications for chronic pain (three), for a skin con-
Finally, the values of recall given recognition shown in Table 2 dition (one), and treatment for a cyst (one). In the older group, four
also showed reliable age differences in favor of the younger were on no regular medications; of the remainder, there were 16
cases of treatment for blood pressure and circulatory problems, 15
participants, t(22) 2.42, p .05 for the high group, and t(22)
for arthritis, 13 for cholesterol, 8 for depression/anxiety, and 6 for
2.95, p .01 for the low group.
bone disease. The average scores on the 10-point self-rated health
scale were 7.3 for young and 8.0 for old, t(77) 2.36, p .02.
Thus, on the face of it, members of the older group were in
Discussion
significantly better health according to their own assessments.
With a simpler design, we found equivalent recognition scores It should be noted, however, that many older adults commented
in the two age groups. We also ensured that ceiling effects in the that they rated their health relative to their own age group, so it
young recognition scores could not be masking age-related differ- would not be correct to assume that there was an absolute
ences in recognition by splitting the data based on performance benefit to the older group. Within each age group, half of the
participants were given one type of encoding task and half the
and looking at both higher and lower performing subjects. The
other type.
interaction between age and memory test was found in both
Materials. As in Experiment 2, two lists of 18 words were
subsets, despite the fact that adjusted recognition scores were in
used as study materials, with an additional list of 10 words used for
the region of 0.60 for the lower subset. The conditional probabil-
the practice session. The recognition list again contained the 36
ities of recall given recognition were also significantly smaller for target words plus 144 distracters. All words were drawn randomly
the older participants. The significant Encoding Task Test Type from a set of 190 words.
interaction shown in Table 2 probably reflects the fact that inten- Procedure. The task was very similar to that in Experiment 2
tional learning promotes associations and organization, which is with the following exceptions. At encoding, participants were told
better for subsequent recall, whereas deep encoding encourages that they should attempt to remember the items for a later memory test
individual item encoding, which is compatible with recognition and were offered one of two memorization techniques (counterbal-
testing. The final experiment attempted to constrain encoding anced across subjects). The first technique was to memorize the
strategies to reduce any age-related differences in this respect. items by grouping them together in some way (e.g., by forming a
Otherwise the design generally replicated the design of Experi- story containing all of them; Grouped), the second was to
ment 2. memorize items individually, (e.g., think about whether it is rele-
RECALL, RECOGNITION, AND AGING 907

vant to you or picture the item; Individual). A short practice list but the younger adults were superior on a test of verbal working
of 10 words was provided before they continued on to the two memory.
study/recall task sessions (both using the same memorization tech- As in Experiment 2, we again split the groups based on their
nique). Following both encoding and recall sessions, participants performance on the recognition memory test to ensure that ceiling
performed the Cattell Culture Fair Test. Participants then per- effects were not responsible for the findings. Taking the lowest 15
formed the recognition test in a self-paced fashion and finally were performers in each of the 4 groups resulted in matched corrected
given the alpha span test. recognition scores of 0.71. Analyses of these matched groups
showed a significant main effect of memory test, F(1, 56) 250.5,
Results p .001, 2 0.82, and a marginally significant effect of age,
F(1, 56) 3.4, p .07, 2 0.06, with higher performance
Table 3 shows that there were no marked differences between overall on the recognition test and in younger participants. The
the two encoding tasks in either recall or recognition scores.
Memory Test Age interaction was also significant F(1, 56)
Recognition scores (hits false alarms) were again very similar
21.6, p .001, 2 0.28, with younger adults outperforming
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

between younger and older groups, but recall scores showed an


older adults on the recall test but not on the recognition test. The
This document is copyrighted by the American Psychological Association or one of its allied publishers.

age-related advantage to younger adults of 15%, over all. We


simple effects showed that older adults performed less well than
conducted a 2 2 2 mixed-design ANOVA that included Age
younger adults for recall on the grouped encoding task, t(28)
(young, old) x Encoding Task (grouped, individual) x Memory
3.9, p .001, and showed a trend in the same direction for the
Test Type (recall, recognition). This analysis yielded main effects
individually encoded task, t(28) 1.9, p .07. Again, there were
of age, F(1, 75) 5.9, p .05, 2 0.07, and test type, F(1,
no differences in recognition for either grouped, t(28) 0.2, p
75) 388.35, p .001, 2 0.838, demonstrating that younger
.83 or for individual task condition, t(28) 1.4, p .16 across the
adults performed better than older adults and that recognition
age groups. The overall conditional probability for recall given
scores were higher than recall scores. The encoding task main
recognition was 0.66 for the young group and 0.49 for the old
effect was not significant, F(1, 75) 0.44, p .5; however the
group, a highly significant difference, t(77) 5.40, p .001.
Encoding Task Age interaction was marginally significant, F(1,
75) 3.4, p .07, probably reflecting the boost in performance
for older adults on the individual encoding task for both tests that Discussion
the younger adults did not experience. The Memory Test Age
interaction was significant, F(1, 75) 29.1, p .001, 2 0.28; Once again, these data support the idea that older adults are
older adults performed reliably less well on recall than younger more impaired on recall tests than on recognition tests, and that
adults whereas they performed equivalently on the recognition test. ceiling effects in recognition are not driving the Age Memory
Using simple effects to further disambiguate these results, younger Test interaction. Providing participants with consistent memoriza-
adults performed better than older adults on grouped items during tion techniques in the present experiment had little effect on the
recall, t(37) 4.5, p .001, whereas no age difference was found overall pattern of results. There was a boost overall in both young
on the recognition test, t(37 1.4, p .65. This pattern was and old groups when they were provided specific techniques, as
replicated for the individual item task: recall, t(38) 2.3, p .05; compared to the free learning condition in Experiment 2 in which
recognition, t(38) 1.1, p .26. The age-adjusted scores for no further aids were provided (a jump from .74 to .81 on recall for
Cattell Culture Fair Intelligence test were 26.2 and 27.3 for younger adults and .67 to .77 on recognition accuracy, and an
younger and older adults, respectively, t(77) 0.99, p .32, and increase from .77 to .87 on recall for older adults and .68 to .82 on
scores on Alpha Span were 27.7 and 20.3 for younger and older recognition). Taken together, the three experiments show the ro-
adults, respectively, t(77) 3.91, p .001. Thus again the groups bust replicability of the finding that recall performance declines
were equivalent on the age-adjusted measure of fluid intelligence, more than recognition performance from young to older adulthood,

Table 3
Experiment 3: Probabilities of Recall and Recognition as a Function of Age and Encoding Type for Both Full and Partial Data Sets

Recognition
n Recall Hits False alarms Hits false alarms RC|RG

Full Data Set


Younggrouped 20 .58 (.16) .81 (.11) .04 (.06) .77 (.15) .70 (.14)
Youngindividual 20 .52 (.12) .81 (13) .04 (.05) .77 (.15) .64 (.14)
Oldgrouped 19 .36 (.13) .85 (.11) .10 (.11) .75 (.16) .48 (.18)
Oldindividual 20 .43 (.11) .89 (.10) .07 (.07) .82 (.13) .50 (.13)
Lowest performers
Younggrouped 15 .52 (.14) .77 (.10) .05 (.06) .71 (.13) .66 (.13)
Youngindividual 15 .51 (.14) .76 (.11) .05 (.06) .71 (.13) .67 (.14)
Oldgrouped 15 .34 (12) .82 (.11) .12 (.12) .71 (.15) .47 (.19)
Oldindividual 15 .42 (.12) .87 (.10) .09 (.07) .77 (.13) .50 (.15)
Note. RC|RG column refers to the probability of recalling items given recognition. Mean probabilities are shown in bold type; standard deviations are
in parentheses.
908 DANCKERT AND CRAIK

and that this result holds under a variety of encoding manipulations recall for 60 s or a recognition test consisting of 20 targets plus 20
and with all evidence of ceiling effects removed. distracters; the alternative test (i.e., recognition or recall) for that
list was then given immediately. Recall levels were 0.44 and 0.23
for younger and older groups, respectively, and corrected recog-
General Discussion
nition scores (hits minus false alarms) were 0.82 and 0.78, respec-
The purpose of the preceding experiments was to shed further tively. When the recognition test was first, recall scores increased
light on relative performance levels on recall and recognition tests by 0.06 in the young group and by 0.02 in the older group. When
of verbal material in younger and older adults. Many older adults recall was first, recognition increased by 0.04 in the younger group
complain about difficulties with recall, especially with highly and by 0.05 in the older group. Thus, the second test in each trial
specific information such as names, whereas the ability to recog- did benefit slightly from performance on the first test, but there
nize wanted items in a recognition test is typically less problematic were no interactions with age in this respect. It is also worth noting
for them. This type of anecdotal evidence is in line with Craiks that tests for recall and recognition were both given immediately
(1983, 1986) suggestion that recall requires self-initiated process- after list presentation in this experiment, yet gave the same pattern
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ing which becomes less effective with increasing age in adulthood, of results as the current experiments. That is, the longer retention
This document is copyrighted by the American Psychological Association or one of its allied publishers.

and with findings that the familiarity component of recognition is interval for recognition than for recall in the current experiments
relatively unimpaired by the aging process, thereby permitting did not seem to be crucial in shaping the pattern of results. The
older adults to perform virtually as well as younger adults on method of successive testing may therefore be recommended for
recognition tests of item memory (Jennings & Jacoby, 1993, use in studies of memory differences in a variety of developmental
1997). These analyses are also in line with some results in the and neuropsychological situations in which it is important to
literature (Craik & McDowd, 1987; Schonfield & Robertson, contrast relative differences in recall and recognition.
1966) but not with others (Baddeley, 1996). In addition, Uttl and As a comment on the samples of participants involved in the
colleagues (2007) have called attention to the likelihood of ceiling present experiments, it should be noted that the young adults were
effects in many recognition tasks performed by younger partici- all university students, whereas the older adults were mostly re-
pants. tired individuals, living in the community, who volunteer to par-
Using the somewhat unorthodox technique of giving partici- ticipate in psychological studies, and who travel in to the univer-
pants both a recall test and a recognition test successively on the sity to do so. As in many studies of cognitive aging, there are
same words in three experiments, we were able to show consis- therefore a number of differences between the groups apart from
tently that age-related decrements were larger in recall than in their differences in age, and caution should be observed in gener-
recognition. This conclusion followed from the overall analyses alizing the results to entire populations. On the other hand, we
based on the means shown in Tables 13 and from analyses of believe that our samples were reasonably well matched. The ma-
recall levels contingent upon successful recognition. Addressing jority of the older participants had worked in professional occu-
Uttl and colleagues (2007) concerns about ceiling effects in pational or skilled trades and it is likely that the younger adults will
recognition in young adults, we either matched participants in such go on to follow similar types of work. In this sense the groups were
a way as to remove the highest scoring individuals or ran data on at least potentially well matched in terms of educational level,
the lower performing groups. We thereby removed the potential socioeconomic factors, and lifestyle. There is a clear mismatch,
for ceiling effects without any differences in the overall findings. however, in terms of medication; most of our younger adults were
The unequivocal conclusionat least for these types of partici- not taking regular medication whereas 89% of the older adults
pants, materials, and testsis that age-related decrements are were taking at least one medication. This does not mean that the
greater in recall than in recognition memory. older adults were chronically sick, however. Many of the medica-
One factor that may qualify this general conclusion is the extent tion are to guard against possible health problems (e.g., high blood
to which a specific test of recognition memory involves processes pressure, cholesterol buildup, osteoporosis, prostate enlargement)
of recollection as opposed to familiarity. We argue that the finding and have no deleterious effects on cognitive performance as far as
of well-maintained recognition memory in older adults depends we know. In fact there was no evidence of poorer health in our
largely on unimpaired familiarity, whereas recollection does show older groups from the self-rated health assessments; the only
age-related losses. It follows that age-related declines in recogni- significant difference was in favor of better health in the older
tion should be found in recognition paradigms in which, for group (Experiment 3). One remaining factor to note is that both
example, recollection of specific encoding contexts is a salient age group samples in the present study were drawn from better
feature. educated, upper socioeconomic populations. We know of no the-
The procedure of giving participants two successive tests on the oretical reason why a different result should be expected in other
same materials is not entirely novel (e.g., Tulving, 1974; Yoneli- socioeconomic populations, yet for the moment caution should be
nas et al., 2002) but is unusual. In fact, the method gave few observed in generalizing to other groups of younger and older
problems of analysis and interpretation: In unpublished data, we adults.
have found that when the order of recall and recognition was As a final comment, given the growing interest in relating
counterbalanced, both recall and recognition showed higher levels cognitive processes to underlying neural structures and processes,
of performance when tested second, but both improved by the it would be fruitful to link the present age-related differences in
same relatively small amount. In greater detail, the experiment was recall and recognition to underlying brain changes with age. Spe-
one similar to Experiment 1 in the current series, with 20 older and cifically, it would be interesting to establish whether aging typi-
20 younger participants being presented with 10 lists of 20 words. cally affects those brain areas associated with recall more than
Each list was followed by 30 s of interference, and then by either those associated with recognition. With regard to the brain corre-
RECALL, RECOGNITION, AND AGING 909

lates of success on these two types of memory task, Wheeler, Jacoby, L. L., Toth, J. P., & Yonelinas, A. P. (1993). Separating conscious
Stuss, and Tulving (1995) reviewed evidence to show that frontal and unconscious influences of memory: Measuring recollection. Journal
lobe damage was associated with much greater impairments in of Experimental Psychology: General, 122, 139 154. doi:10.1037/0096-
recall than in recognition. Second, Aggleton and Brown (1999) 3445.122.2.139
Jennings, J. M., & Jacoby, L. L. (1993). Automatic versus intentional uses
presented results in support of the view that the hippocampus
of memory: Aging, attention, and control. Psychology and Aging, 8,
supports episodic recall but is not required for recognition judg-
283293. doi:10.1037/0882-7974.8.2.283
ments. Third, the medial temporal lobe cortex (including entorhi- Jennings, J. M., & Jacoby, L. L. (1997). An opposition procedure for
nal, perirhinal, and parahippocampal regions) has been associated detecting age-related deficits in recollection: Telling effects of repeti-
with the familiarity component of recognition judgments (Bowles tion. Psychology and Aging, 12, 352361. doi:10.1037/0882-7974.12.2
et al., 2007; Norman & OReilly, 2003). Regarding studies of brain .352
changes in older adults, Raz and colleagues (2005) have presented Kopelman, M. D., Bright, P., Buckman, J., Fradera, A., Yoshimasu, H.,
findings showing that healthy aging is associated with substantial Jacobson, C., & Colchester, A. C. F. (2007). Recall and recognition
declines in regional brain volumes of the hippocampus and frontal memory in amnesia: Patients with hippocampal, medial temporal, tem-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

cortex, but minimal age-related changes in the entorhinal cortex. poral lobe or frontal pathology. Neuropsychologia, 45, 12321246.
doi:10.1016/j.neuropsychologia.2006.10.005
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Complementing these findings, Yonelinas and colleagues (2007)


Kucera, H., & Francis, W. N. (1967). Computational Analysis of Present-
examined 157 participants aged between 65 and 80 years and
day American English. Providence, RI: Brown University Press.
found that performance on recall and recognition tests was pre- Mandler, G. (1980). Recognizing: The judgment of previous occurrence.
dicted by hippocampal and entorhinal volumes, respectively. They Psychological Review, 87, 252271. doi:10.1037/0033-295X.87.3.252
also found that hippocampal volume declined with age whereas Norman, K. A., & OReilly, R. C. (2003). Modeling hippocampal and
entorhinal cortex volume was relatively unaffected by the aging neocortical contributions to recognition memory: A complementary-
process. Current findings from neuroanatomy and from the cog- learning-systems approach. Psychological Review, 110, 611 646. doi:
nitive neuroscience of aging are thus in good agreement with the 10.1037/0033-295X.110.4.611
present results to yield the overall conclusion that normal aging is Park, D. C., Lautenschlager, G., Hedden, T., Davidson, N. S., Smith, A. D.,
associated with substantial declines in volume of the hippocampus & Smith, P. K. (2002). Models of visuospatial and verbal memory across
the adult life span. Psychology and Aging, 17, 299 320. doi:10.1037/
and frontal brain areas, but negligible changes in entorhinal cortex;
0882-7974.17.2.299
in turn, this pattern of decline and maintenance is associated
Raz, N., Lindenberger, U., Rodrigue, K. M., Kennedy, K. M., Head, D.,
respectively with an age-related decline in recall but relative Williamson, A., Acker, J. D. (2005). Regional brain changes in aging
maintenance of recognition. healthy adults: General trends, individual differences and modifiers.
Cerebral Cortex, 15, 1676 1689. doi:10.1093/cercor/bhi044
References Schonfield, D., & Robertson, B. A. (1966). Memory storage and aging.
Canadian Journal of Psychology, 20, 228 236. doi:10.1037/h0082941
Aggleton, J. P., & Brown, M. W. (1999). Episodic memory, amnesia, and Troyer, A. K., Hfliger, A., Cadieux, M. J., & Craik, F. I. M. (2006). Name
the hippocampal-anterior thalamic axis. Behavioral and Brain Sciences, and face learning in older adults: The effects of level of processing,
22, 425 489. doi:10.1017/S0140525X99002034 self-generation, and intention to learn. Journal of Gerontology: Psycho-
Baddeley, A. D. (1996). Applying the psychology of memory to clinical logical Sciences, 61, P67P74. doi:10.1093/geronb/61.2.P67
problems. In D. Herrmann, C. Hertzog, P. Hertel, M. K. Johnson, & C. Tulving, E. (1974). Cue-dependent forgetting. American Scientist, 62,
McEvoy (Eds.), Basic and Applied Memory Research: Theory in Con- 74 82.
text (vol. 1, pp. 195219). Mahwah, NJ: Erlbaum. Uttl, B., Henry, M., & Baltimore, K. (2007). Are smaller age differences on
Baddeley, A. D., Emslie, H., & Nimmo-Smith, I. (1994). Doors and old/new recognition versus free recall tests artifacts of easy memory
people: A test of visual and verbal recall and recognition. Bury St. tests? Canadian Journal of Experimental Psychology, 61, 374.
Edmunds, UK: Thames Valley Test Company. Wheeler, M. A., Stuss, D. T., & Tulving, E. (1995). Toward a theory of
Bowles, B., Crupi, C., Mirsattari, S. M., Pigott, S. E., Parrent, A. G., episodic memory: The frontal lobes and autonoetic consciousness. Psy-
Pruessner, J. C. . . . Khler, S. (2007). Impaired familiarity with chological Bulletin, 121, 331354. doi:10.1037/0033-2909.121.3.331
preserved recollection after anterior temporal-lobe resection that spares Yonelinas, A. P., Kroll, N. E. A., Quamme, J. R., Lazzara, M. M., Sauve,
the hippocampus. Proceedings of the National Academy of Sciences, M.-J., Widaman, K. F., & Knight, R. T. (2002). Effects of extensive
USA, 104, 1638216387. doi:10.1073/pnas.0705273104 temporal lobe damage or mild hypoxia on recollection and familiarity.
Cattell, R. B., & Cattell, A. K. S. (1960). Culture Fair Intelligence Test, Nature Neuroscience, 5, 1236 1241. doi:10.1038/nn961
Scale 2. Champaign, IL: Institute for Personality and Ability Testing. Yonelinas, A. P., Widaman, K., Mungas, D., Reed, B., Weiner, M. W., &
Craik, F. I. M. (1983). On the transfer of information from temporary to Chui, H. C. (2007). Memory in the aging brain: Doubly dissociating the
permanent memory. Philosophical Transactions of the Royal Society, contribution of the hippocampus and entorhinal cortex. Hippocampus,
London, Series B, 302, 341359. 17, 1134 1140. doi:10.1002/hipo.20341
Craik, F. I. M. (1986). A functional account of age differences in memory. Zachary, R. A. (1986). Shipley Institute of Living Scale: Revised Manual.
In F. Klix & H. Hagendorf (Eds.), Human memory and cognitive Los Angeles, CA: Western Psychological Services.
capabilities, mechanisms and performances (pp. 409 422). North-
Holland, The Netherlands: Elsevier.
Craik, F. I. M., & McDowd, J. M. (1987). Age differences in recall and Received January 8, 2013
recognition. Journal of Experimental Psychology: Learning, Memory, Revision received April 4, 2013
and Cognition, 13, 474 479. doi:10.1037/0278-7393.13.3.474 Accepted April 26, 2013

Вам также может понравиться