Professional Documents
Culture Documents
86
decision making, (Chen & Sun, 2003), and resolution of interpersonal predicaments (Crawford & Channon, 2002).
The second theoretical approach suggests that despite the everyday context, problem-solving and decision-making processes
are compiled from primary abilities (such as declarative memory
or inductive reasoning) that have been demonstrated to decline
with age (Marsiske & Willis, 1995). Thus, to the extent that an
everyday task relies on these primary abilities, age-related changes
will emerge (Willis, 1996; Willis & Schaie, 1986). This view is
supported by strong correlations between performances on measures of everyday cognition and traditional psychometric measures
of reasoning and working memory (e.g., Allaire & Marsiske, 1999;
Kirasic, Allen, Dobson, & Binder, 1996) and by reports of significant declines in everyday problem-solving (Denney, Pearce, &
Palmer, 1982) and decision-making effectiveness in older age
(Finucane et al., 2002). Furthermore, several studies have reported
that EPSE peaks in middle age, followed by a modest decline in
later years (Denny & Palmer, 1981; Denney et al., 1982; Denney
& Pearce, 1989). The increase in EPSE reported in some middleaged samples has been attributed to the additive effects of experience in solving everyday problems and the relative preservation
of cognitive capacities (Denney, 1990).
To date, studies investigating the age trajectory of EPSE have
not yielded consistent findings supporting either perspective. The
current study aimed to address the important questions of whether
or not age differences exist in EPSE and whether the magnitude of
such differences approximates that of other age-variant aspects of
cognition, such as memory. Toward this end, we investigated the
role of a number of variables that have been previously advanced
as potentially important moderators of EPSE. These include differences in the scoring criteria applied (Berg & Klaczynski, 1996),
differences in the problem-solving domain (Blanchard-Fields,
Chen, & Norris, 1997; Heidrich & Denney, 1994), differences in
the nature of the to-be-solved problem (Allaire & Marsiske, 2002),
and differences in participant characteristics (Denney & Palmer,
1981; Haught, Hill, Nardi, & Walls, 2000). Each of these are
discussed in turn.
in older adults (Berg, Meegan, & Klaczynski, 1999). Other investigators have de-emphasized fluency demands by examining the
quality of the rationale (Blanchard-Fields, Jahnke, & Camp, 1995;
LaRue & DElia, 1985; Meyer et al., 1995; Stanley et al., 1984;
Zwahr et al., 1999), the effectiveness of the responses (Crawford
& Channon, 2002), or the accuracy of the solutions as compared
with an absolute criteria (Chen & Sun, 2003; Finucane et al., 2002;
Hershey & Farrell, 1999; Hershey & Wilson, 1997; Whitfield,
Baker-Thomas, Heyward, Gatto, & Williams, 1999). Nonetheless,
findings from these studies are not uniform, with some reporting
no age differences (Chen & Sun, 2003; Crawford & Channon,
2002; Hershey & Wilson, 1997; Stanley et al., 1984), and others
reporting significant reductions in EPSE in older adults (Finucane
et al., 2002; Whitfield et al., 1999). The current study can lend
clarity to this issue by examining whether or not the magnitude of
age differences in EPSE is affected by the scoring criteria used. If
age differences are reduced when fluency demands are minimized,
this would support previous contentions that the less exhaustive
and more selective approach exhibited by older adults does not, in
fact, imply a true deficit in EPSE (Berg et al., 1999).
Another unresolved issue involves the comparability of experimenter and participant ratings of EPSE. In contrast to the
experimenter-rated approaches described above, several investigators have asked participants to rate their own satisfaction or confidence in their problem-solving abilities (Artistico, Cervone, &
Pezzuti, 2003; Blanchard-Fields et al., 1997; Crawford & Channon, 2002; Hershey & Farrell, 1999; Hershey & Wilson, 1997) or
to provide likelihood ratings of their typical responses to hypothetical problems of daily life (Blanchard-Fields et al., 1997;
Cornelius & Caspi, 1987; DZurilla et al., 1998; Haught et al.,
2000; Wright, Carscaddon, & Lambert, 2000). Although authors
have acknowledged that such self-report likelihood ratings may
not assess actual problem-solving abilities (Cornelius & Caspi,
1987; Haught et al., 2000), findings from self-rating studies are
commonly presented as evidence of preserved EPSE in aging (e.g.,
Artistico et al., 2003; Berg & Klaczynski, 1996; Crawford &
Channon, 2002; Haught et al., 2000). The present study aims to
lend clarity to the issue of whether conclusions regarding age
differences may vary as a function of the rating system used. Using
meta-analytic methodology, we can compare whether effects derived from studies using different rating criteria (experimenter vs.
participant-derived ratings) are homogeneous and of similar
magnitude.
87
88
to date there has been no systematic investigation of these parameters with multiple age groups. Our meta-analysis can examine the
role of task structure in EPSE by contrasting the magnitude of
effect sizes obtained from studies using well-structured tasks versus those using ill-structured tasks.
Participant Characteristics
Participant characteristics, such as differences in composition of
age groupings or level of formal education, may also underlie
reported age differences, although few primary studies have examined this directly. Our review of the extant literature revealed
that the composition of the age groups involved varied widely. For
example, the age ranges reported for young samples ranged from
as low as 18 years (Heidrich & Denney, 1994) to as high as 64
years (Finucane et al., 2002). Participants classified as middleaged ranged in age from as young as 32 years (Haught et al.,
2000) to as high as 64 years (Meyer et al., 1995), and the ages of
participants classified as old ranged from as low as 55 years
(Patrick, 1995) to as high as 84 years (Devolder, 1993). Such
overlap in age groupings can seriously impede the ability to render
meaningful conclusions from qualitative examination of primary
studies. By ensuring appropriate age group comparisons, the current meta-analysis can provide a more accurate picture of age
differences between comparable groups of young, middle-aged,
and older participants. Another potentially important moderator is
participant educational level. The current meta-analysis can also
address this question by examining the relationship between educational level and effect size magnitude.
Predictions
We addressed the following questions in our meta-analysis: (a)
Are there reliable age differences in EPSE? (b) Are there identifiable methodological variables that moderate these age differences? and (c) Are the findings from the everyday problem-solving
and everyday decision-making literatures distinct, or can they be
meaningfully combined? On the basis of our survey of the literature, we made the prediction that the aging pattern on measures of
EPSE would follow a similar trajectory to that reported on other
cognitive measures, namely, that effect sizes would be largest
between old and young groups and comparatively smaller between
middle-aged versus young and middle-aged versus old groups. In
addition, we examined the following potential moderators of the
expected age differences: the scoring criteria used to establish
EPSE (fluency vs. quality) and the nature of the task (older vs.
younger problem content, interpersonal vs. instrumental content,
well-structured vs. ill-structured task). We predicted that age differences would be smaller under conditions that have been previously argued to bolster EPSE in older age (quality emphasis, older
age problem content, interpersonal problem content, ill-structured
problem) and larger in the other conditions (fluency emphasis,
non-age-specific content, well-structured task). In addition, without specifying the nature of the prediction, we examined whether
the task outcome was dependent on the rater (experimenter vs.
participant) and the focus of the study (i.e., problem solving vs.
decision making). We also examined whether the quality of the
study methodology or the procedures used (i.e., computerized
administration) influenced the effects.
Method
Studies
Three methods were used to locate potentially relevant studies for
inclusion in the analyses: (a) Studies were identified through a computerized search of MEDLINE (1958 2003), PsycINFO (18872003), AgeLine
(1978 2003), and Digital Dissertation Abstracts (1980 2003); (b) manual
searches were conducted of journals identified as having a high number of
potentially relevant articles (Psychology and Aging, 1988 2003; Experimental Aging Research, 19772003; Journals of Gerontology, Series B:
Psychological Sciences and Social Sciences, 19752003); and (c) inspection of references from reviews of everyday problem solving and decision
making in aging. The key terms used for the computer database search
were decision making, problem solving, everyday, expertise, judgment,
strategies, practical, and age (age differences, age groups). Studies were
limited to those that reported results in English. All studies located by these
means were inspected to ensure that they met the following criteria for
inclusion in the meta-analysis.
1. The study needed to allow at least one age group comparison (i.e.,
young vs. old, middle-aged vs. old, middle-aged vs. young). A recent study
found that representative journals typically report the following age groupings: older (58 82), middle-aged (40 59), and young (19 35; Nichols, Rogers, Fisk, & West, 2001). Therefore, to avoid overlap and fill in
the blanks, we adopted the following age ranges as criteria for inclusion
into the analyses: older (60), middle-aged (40 59), and young
(18 39). Two studies with age groups deviating from these ranges by
greater than or equal to 5 years were eliminated (i.e., Fitzgerald, 2000;
Patrick, 1995). A summary of study demographic characteristics is presented in Table 1.
2. Studies were included if data were reported on one or more of the
following outcomes of interest: number of solutions, quality of solutions,
problem-solving/decision effectiveness, problem-solving/decision accuracy, and quality of problem-solving/decision rationale. Because the focus
of this meta-analysis was on EPSE, only studies allowing examination of
directional effects along this dimension (i.e., better vs. worse EPSE) were
included. Studies examining qualitative age differences in everyday
problem-solving processes (i.e., Berg et al., 1998; Blanchard-Fields, Baldi,
& Stein, 1999; Strough, Berg, & Sansone, 1996; Watson & BlanchardFields, 1998) were not included because they addressed a different research
question and did not allow calculation of a directional effect size for EPSE.
In addition, studies of group problem solving were not included (i.e.,
Streufert, Pogash, Piasecki, & Post, 1990) nor were studies investigating
participants ratings of others problem-solving capabilities (i.e., Klaczynski & Robinson, 2000).
3. Included studies were those that were self-identified or identified by
Wendy J. L. Thornton as examining an aspect of problem solving or
decision making within a practical or real-world context. As such, studies
using traditional Piagetian and anagram tasks were excluded, as were those
substituting more realistic stimuli into traditional laboratory tasks (i.e.,
Rabbitt, 1977; Reese & Rodeheaver, 1985) and examining age differences
in skilled (i.e., wisdom, expertise) or game (i.e., chess) performance.
Studies meeting our inclusion criteria were examined to ensure that they
reported sufficient data for a between-age-group effect size to be generated
on a least one variable of interest. If these data were not reported and
contact information was available, study authors were contacted to request
the necessary data. A complete list of excluded studies is available from the
corresponding author on request.
Coding
On the basis of our review of the extant literature, we subsequently
coded variables of interest relevant to the predictions, as well as methodological variables of theoretical interest. The nine coded variables are as
follows.
30 (25.7)
34 (20.9)
69 (28.5)
62 (22.3)
14 (20.1)
(29.2)
30 (25.3)
802 (1720)
28 (28.4)
32 (28.3)
32 (27.9)
48 (27.8)
48 (19.9)
25 (29.1)
20 (20.1)
43 (28.2)
28 (18.6)
80 (2039)
68 (1830)
27 (28.8)
38i (33.7)
154 (1739)
34 (29.2)
1,746
25.7 (4.2)
74 (48.7)
67 (49.7)
126d (43.8)
91 (46.5)f
28 (52.9)
32 (49.7)
32 (49.2)
48 (48.8)
76 (52.4)
239 (40)
20 (41.9)
41 (49)
80 (4059)
59 (4059)
35 (51.5)
79d,j (2347)
34 (5059)
68 (4059)
34 (49.3)
1,263
47.9 (4.7)
M-A
30 (69.7)
27 (72.9)
74 (68.7)
76 (68.3)
14 (71.5)
(65.4)
30 (68.3)
91 (69)f
28 (69.0)
32 (69.2)
32 (68.8)
48 (68.9)
48 (74.9)
106 (74.5)
253 (75)
25 (68.4)
20 (70)
29 (69.5)
32 (71.1)
40 (6069)
58 (6079)
69 (6577)
32 (73.3)
33i (69.2)
198 (6090)
14 (6079)
34 (70.1)
1,473
70.2 (2.7)
Old
14.8 (2.6)
15.1 (3.3)
15.1 (3.1)
()
16.8f (2.6)
14.3 ()
16.5 ()
14 ()
15.8 ()
19.9 ()
15.5 ()
13.5 (2.8)
15.5 (1.7)
15.0 (1.1)
M-A
14.8 (2.5)
15.3 (2.6)
14.1 (1.7)
14.0 (1.4)
15.5 (2.9)
15.1 (1.7)
14.7 ()
14.7 ()
15.6 ()
14.6 ()
13.4 (1.4)
18.7 ()
14.3 ()
15.4 ()
13.90 ()
15.4 (1.2)
15.6 (2.8)
Young
Old
14.5 (1.2)
15.2 (2.5)
13.7 (2.9)
14.7 (3.2)
15.2 (3.3)
15.9 (3.1)
14.0 (2.5)
13.9f (2.6)
13.7 ()
13.4 ()
12.3 ()
13.4 ()
15.1 (3.4)
14.7 ()
17.1 ()
15.9 ()
16.20 ()
14.2 (2.2)
12.8 (3.3)
22.8 ()
22.9 (7.3)
110.0 (8.1)
40.2 (2.5)
54.1 (12.4)
48.9 (15.0)
30.6 (6.0)
Young
.32
31.7 (5.7)
55.6 (10.0)
45.8 (15.7)
33.1 (4.5)
M-A
23.8a,b ()
31.6c (6.1)
114.4e (11.0)
44.1g (2.8)
59.8c (9.7)
48.6h (23.2)
32.8k (4.8)
Old
Note. Dashes indicate that data were not available for these measures. EPSE everyday problem-solving/decision-making effectiveness; N number of participants in each age group; M-A
middle-aged.
a
Thurstone Verbal Meaning Test. b Not included in calculation of d because SDs were not available. c WAISR Vocabulary. d Age treated as continuous variable. e NAART North American
Adult Reading Test. f Demographic data reported in Kant, DZurilla, and Maydeu-Olivares, 1997. g Ammons Quick TestVocabulary. h Quick Word Test. i Ns estimated from ranges provided,
because of missing data. j Data in analysis of young versus middle-aged EPSE only. k Shipley Institute of Living Scale. l The midpoint was used for studies reporting age ranges only. m d
(raw scoreold raw scoreyoung)/pooled SDold & young.
Young
Study
Table 1
Demographic Characteristics of Studies Included in the EPSE Meta-Analysis
89
90
Analysis
We calculated an effect size (d) for each age group comparison (young
vs. middle-aged, young vs. old, middle-aged vs. old) separately by subtracting the younger group mean EPSE from that of the older group and
dividing this score by the pooled standard deviation. Calculation of d
requires information regarding the number of participants, as well as group
performance means and standard deviations (Cohen, 1988). When these
data were not available, we computed d from F statistics with one degree
of freedom, and t statistics. We calculated d from a correlation statistic
(Taylor, 1975) and from a chi-square statistic (Finucane et al., 2002) with
the assistance of Effect Size software (Shadish, Robinson, & Lu, 1999).
Effect sizes were coded so that positive values indicate effects favoring
EPSE in younger age groups, whereas negative values indicate effects
favoring EPSE in older age groups. To correct for upwardly biased estimation of the effect size with small samples, we used Hedges procedures
(Hedges & Olkin, 1985) to calculate an unbiased estimate, g, for each
study, with the assistance of Comprehensive Meta-Analysis software
(Borenstein & Rothstein, 1999). Combining these g values provided the
overall point estimate for each set of effects across all samples. All effect
sizes were calculated according to a fixed effects model of error estimation.
As we were interested in comparing the differences in effect sizes for the
entire collection of effects, it was necessary to determine whether the effect
sizes were homogeneous and could therefore be assumed to originate from
the same population. The first step in the analysis was to determine whether
all studies shared a common effect size by examining resultant QT values
(Borenstein & Rothstein, 1999). The absence of a significant p value in
accordance with examination of Forrest plot dispersion was used to determine homogeneity of the effect sizes. Within-class homogeneity statistics
(QW) were computed to assess whether the effect size estimates obtained
within each group of interest represented a reliable point estimate of the
population value (Borenstein & Rothstein, 1999). If the test of homogeneity failed, studies were partitioned into groupings based on the theoretically derived moderating variables of interest, and between-group homogeneity statistics (QB) were generated to assess the reliability of betweengroup differences (Hedges & Olkin, 1985). The QB statistic is analogous to
an F statistic in testing class differences.
Some studies yielded multiple comparisons of the variable of interest,
and thus, there arose the possibility of a single study contributing more than
one effect size to the meta-analysis. In three of the studies (Denney et al.,
1992; Devolder, 1993; Hershey & Wilson, 1997), multiple comparisons of
EPSE were derived across separate experiments. In these studies, we
pooled the effect estimates to derive an overall study effect size. Another
situation arose for studies that included more than two age groups, potentially allowing more than one age comparison to be entered into the
analysis. Eight of the studies entered into the initial analysis of age
differences offered effects for more than one age group comparison (i.e.,
old vs. young, old vs. middle-aged, middle-aged vs. young). Eliminating
these comparisons would have seriously limited our ability to examine the
principle hypothesis. Therefore, in the initial analysis, we permitted studies
with information from more than two age groups. This allowed a study to
potentially contribute more than one effect to the overall analysis but not
more than one effect to each age group comparison. Thus, although the
estimate of variability within groups (QW) is based on independent observations, the estimate of variability between groups (QB) provides a more
Results
Table 2 lists the effect sizes (Hedgess adjusted g) for each study
by outcome variable, as well as important study characteristics.
We first present findings regarding age differences in EPSE (see
Table 3), followed by examination of the potential moderators of
the observed differences (see Table 4). In interpreting these findings, we have adopted the conventional values of .20, .50, and .80
for small, medium, and large effect sizes for comparing independent groups (Cohen, 1992). In addition, effect size estimates with
their 95% confidence intervals (CI95%) are presented in Table 2.
A significance level of .05 can be inferred when zero is not
contained within the CI95%.
91
Moderator Variables
We examined the potential moderator variables by categorizing
effects on the basis of sample characteristics and examining the
resultant QB and Qw (see Hedges, 1984). The first moderator
variable of interest was the rating criteria variable, which allowed comparison of studies in which performance was evaluated
by the experimenter versus the participant. Twelve additional
effects and six additional studies were entered into this analysis
(Blanchard-Fields et al., 1997; Cornelius & Caspi, 1987; DZurilla
et al., 1998; Feifel & Strack, 1989; Haught et al., 2000; Wright et
al., 2000). A total of six studies contributed both self-ratings and
experimenter ratings of EPSE (Artistico et al., 2003; Berg et al.,
1999; Crawford & Channon, 2002; Hershey & Farrell, 1999;
Hershey & Wilson, 1997; McDonald-Miszczak, 1997). As described above, this resulted in independent observations within
each grouping and a conservative estimation of between-group
variability (QB; see Robinson et al., 1990). Because only three of
these studies allowed separate coding of middle-aged effects
(Blanchard-Fields et al., 1997; DZurilla et al., 1998; Wright et al.,
2000), we combined young and middle-aged effects for these
studies to allow a comparison of younger/middle-aged (ages 18
59) and older (ages 60 and over) groups. Note that although both
participant and experimenter ratings were also provided in Haught
et al. (2000), the experimenter rating score was identified as an
outlier in the previous analysis. Thus, only the participant rating
score was entered into this analysis.
Studies were then grouped according to rater (experimenterrated vs. participant-rated). This grouping resulted in the predicted within-age-group homogeneity within the experimenterrated studies; however, the participant-rated studies exhibited
significant heterogeneity (Qw 43.37, df 11, p .01). The
participant-rated studies were then grouped by type of rating (i.e.,
likelihood rating vs. confidence rating). One study was identified
as a statistical outlier (Crawford & Channon, 2002) and was
removed from this analysis. After removal of the outlier, this
grouping resulted in a homogeneous population of effects for each
of the three groupings: experimenter-rated (Qw 24.06, df 20,
p .05), participant-rated likelihood ratings (Qw 4.50, df 4,
p .05), and participant-rated confidence ratings (Qw 4.52,
df 5, p .05). Removal of this outlier did not impact betweenstudy group heterogeneity, which was significant both with the
outlier included (QB 18.93, df 1, p .01) and excluded
(QB 25.93, df 1, p .01). Examination of the between-groups
differences revealed that effects from experimenter-rated studies
(g .484) were significantly larger (favoring younger adults) than
effects from participant-rated likelihood ratings, g .213,
t(24) 6.73, p .01, but not from participant-rated confidence
ratings, g .349, t(25) 1.41, p .05. Participant-rated likeli-
92
Table 2
Individual Study Characteristics and Weighted Effect Sizes (g)
95% CI
Study
Focus
EPSE determination
Task
structure
Problem content
Outcome
measure
Social/instra
Mix
Social
Instr
Social
Mix
Mix
Mix
Mix
Mix
Instr
Mix
Social/instra
Instr
Instr
Instr
Instr
Instr
Instr
Instr
Instr
Instr
Fluency
Fluency
Quality
Quality
Quality
Fluency
Fluency
Fluency
Fluency
Fluency
Quality
Fluency
Fluency
Quality
Quality
Quality
Quality
Quality
Quality
Quality
Quality
Quality
Lower
Upper
.708
.573
.391
.617
.016
.609
.235
.860
.369
.335
.602
.293
.194
.313
.191
.415
.487
.978
.283
.256
.715
.645
.484
.174
.046
.106
.181
.533
.138
.197
.412
.017
.073
.421
.279
.234
.037
.328
.045
.144
.522
.194
.194
.343
.219
.401
1.242
1.100
0.675
1.415
0.501
1.079
0.666
1.308
0.721
0.744
0.783
0.865
0.621
0.663
0.710
0.785
0.830
1.434
0.761
0.707
1.087
1.072
.567
.409
.061
.235
.503
.390
.213
.785
.277
.533
1.060
.934
.359
.034
.155
.063
.053
.154
.067
.291
.172
.430
.064
.367
.564
.349
.229
.344
.142
.320
.155
.219
.189
.811
.689
.718
.447
.890
.909
.509
Experimenter-rated studies
Artistico et. al. (2003)
Berg et. al. (1999)
Blanchard-Fields et. al. (1995)
Chen & Sun (2003)
Crawford & Channon (2002)
Denney & Palmer (1981)
Denney & Pearce (1989)
Denney et. al. (1982)
Denney et. al. (1992)
Dobson (2000)
Finucane et. al. (2002)
Haught (1985)
Heidrich & Denney (1994)
Hershey & Farrell (1999)
Hershey & Wilson (1997)
LaRue & DElia (1985)
McDonald-Miszczak (1997)
Meyer et. al. (1995)
Stanley et. al. (1984)
Taylor (1975)c
Whitfield et al. (1999)
Zwahr et al. (1999)
Subtotal (k 21)
PS
PS
PS
DM
PS
PS
PS
PS
PS
PS
DM
PS
PS
PS
DM
PS
DM
DM
DM
DM
PS
DM
Ill
Ill
Ill
Well
Ill
Ill
Ill
Ill
Ill
Ill
Well
Ill
Ill
Well
Well
Well
Well
Well
Well
Well
Well
Well
EPSI
SPSIR
LSI
PSI
PSI
Participant-rated confidence
PS
PS
PS
PS
DM
DM
Self-efficacy rating
Perceived effectiveness
Confidence rating
Confidence rating
Confidence rating
Self-efficacy rating
Note. 95% CI the effect estimate meets standard test of statistical significance when zero is not included in the 95% confidence interval; EPSI
Everyday Problem Solving Inventory; SPSIR Social Problem Solving InventoryRevised; PSI Problem Solving Inventory; LSI Life Situations
Inventory; Ill ill-structured task; Well well-structured task; social interpersonal problem content; instr instrumental problem content; mixed
both interpersonal and instrumental problems; EPSE everyday problem-solving/decision-making effectiveness; PS problem solving; DM decision
making; S & E safe and effective.
a
Study allowed separate coding of interpersonal and instrumental problems. b In this study, EPSE was defined as a higher score on the PF and CA scales
and as a lower score on the PD and AD scales. c Effect estimate not included in subtotal, as this study was not included in the final analysis because of
the absence of an older age comparison group. d Study included both experimenter ratings and confidence ratings.
93
Table 3
EPSE Effect Sizes and 95% Confidence Intervals (CIs) for Each Age Group Comparison
No. of
studies
(k)
9
10
17
21
Mean weighted
effect size
(g)
95% CI
Within-category
homogeneity statistic
(Qw)
814
1,035
1,329
2,610
.044
.466
.427
.484
.095.182
.333.598
.317.538
.401.567
12.08*
9.71*
20.93*
24.06*
Mean weighted
effect size
(g)
95% CI
2
10
9
165
1,688
757
.509
.519
.409
.196.821
.413.624
.259.558
6
5
10
439
606
1,565
.413
.388
.542
.220.606
.212.564
.434.650
14
7
1,626
984
.423
.570
.315.532
.441.700
10
11
1,533
1,077
.543
.405
.434.653
.278.532
9
12
800
1,810
.444
.502
.297.592
.402.603
4
10
7
450
1,533
627
.210*
.543*
.456*
.013.407
.434.653
.291.622
p .05). Results also revealed significant between-groups heterogeneity when studies were coded as interpersonal, instrumental,
or mixed (QB 8.43, df 2, p .05). Examination of the
between-groups differences revealed that effects from problems
with interpersonal content (g .210) were significantly smaller
(favoring younger adults) than effects from both studies with
instrumental problem content, g .543, t(12) 2.81, p .05, and
mixed problem content, g .456, t(9) 2.47, p .05. Effect
estimates from studies with instrumental versus mixed problem
contents were not significantly different from each other, t(15)
0.55, p .50. Thus, although EPSE was significantly reduced in
older adults as compared with young/middle-aged adults in all
content conditions, these age differences were substantially reduced when the problem content was interpersonal.
We also investigated three additional planned contrasts (fluency
vs. quality scoring emphasis; age-appropriate problem content;
well-structured vs. ill-structured task), and three unplanned contrasts (quality of the study methodology; problem-solving vs.
decision-making focus; and computerized vs. noncomputerized
administration). Results of these analyses are presented in Table 4.
None of these additional moderator analyses resulted in significant
between-group differences.
We also examined the assertion that educational level is related
to EPSE. Mean participant educational levels by age group were
available from 15 experimenter-rated studies. We computed the
correlation between the mean educational levels of the older participants (M 14.26, SD 1.16) and the effect sizes for EPSE. At
first glance, it appears that a strong relationship exists between the
educational level of older participants and the magnitude of the
effect estimates (r .63, p .01), suggesting that higher levels
of education in older participants resulted in a diminished difference in EPSE between younger and older adults. Nonetheless, the
significance of this relationship was reduced (r .43, p .06)
after one study was removed that was identified as an outlier both
in the previous analysis and on the basis of visual plot inspection
(Haught et al., 2000). It is interesting to note that this study
specifically examined the role of educational level by recruiting a
very highly educated older-adult sample (M education level 17.1
years).
4
16
374
2,112
.507
.473
.292.722
.380.565
Additional Analyses
Variable
Procedure
Computer
Paper/pencil
Read aloud
Quality
Well-controlled
Partially controlled
Not controlled
Focus
Problem solving
Decision making
Task structure
Well structured
Ill structured
Scoring emphasis
Fluency
Quality
Problem content*
Interpersonal
Instrumental
Mixed
Age content
Older
Unspecified
No. of
studies
(k)
94
Discussion
Several findings from the current meta-analysis contribute toward resolution of some of the important issues in this literature.
First, the published literature to date does not support theories of
preserved EPSE in late adulthood, nor does the literature support
the pervasive notion that problem-solving abilities peak in middle
age. Rather, age differences in everyday problem solving are
reliable and of moderate magnitude across methodological and
theoretical domains. Second, very different conclusions regarding
EPSE are derived from likelihood ratings versus experimenter
ratings, suggesting that these criteria assess different constructs.
Third, both the content of the to-be-solved problem and the educational level of the older sample were found to influence the
magnitude of age differences in EPSE. Finally, results appear to
support the conceptual integration of findings from the everyday
problem-solving and everyday decision-making literatures. Each
of these findings is discussed in the following paragraphs.
95
highest predictive validity for real-world problem-solving performance remains an important question for future research.
It is interesting that while older adults rate their likelihood of
effective problem solving as better than younger adults, they are
significantly less confident regarding their actual performance. As
with likelihood ratings, confidence ratings presumably reflect past
experience (Bandura, 1977) and, as such, have been shown to be
closely related to actual problem-solving performance (Artistico et
al., 2003; Haught et al., 2000). The current analysis suggests that
younger/middle-aged adults are more confident than older adults
regarding their problem-solving abilities and that the magnitude of
this age difference is small to medium (g .349). It is important
to note that the current findings do not speak to the relative
accuracy of these various rating systems in older and younger
participants. Rather, they suggest that global estimates of EPSE
vary considerably according to who is determining effectiveness
and depending on whether likelihood or confidence ratings are
used. A potentially fruitful area for future research involves
whether experimental manipulation of self-efficacy impacts older
adults performance on everyday problem-solving and decisionmaking tasks, as has been demonstrated in other cognitive domains
(see Bandura, 1997).
The second moderator detected in the current meta-analysis
involves whether the problem content is interpersonal or instrumental in focus. The current findings suggest that age differences
in EPSE are attenuated when the problem content reflects interpersonal concerns (g .210) as compared with instrumental
concerns (g .548) or those with mixed content (g .456).
Although some caution is necessary given that few studies are
included in the interpersonal content category (k 4), such findings are consistent with those that have highlighted the importance
of contextual factors in problem-solving behaviors. Previous authors have suggested that interpersonal concerns may be more
emotionally salient in older adulthood (Blanchard-Fields et al.,
1995) and that older adults may have more experience solving
interpersonal problems than younger adults (Heidrich & Denney,
1994). In addition, findings indicate that older adults tend to
approach problems with emotionally salient interpersonal content
differently than either middle-aged or younger adults (BlanchardFields et al., 1995, 1997).
The question remains whether older adults engage more fully in
problem-solving tasks that they perceive to be more salient or
more representative of domains with which they have had previous
success. One may predict that self-efficacy in EPSE varies as a
result of task-specific factors, as has been demonstrated in other
cognitive domains (for a review, see Cavanaugh, 1996). Although
there were too few studies providing confidence ratings to effectively assess moderators in the current meta-analysis, it is interesting to note that two studies that investigated satisfaction (Crawford & Channon, 2002) and self-efficacy (Artistico et al., 2003)
regarding interpersonal problem solving reported higher confidence ratings in older adults, as compared with younger adults.
Systematic investigation of self-efficacy across interpersonal and
instrumental problem-solving domains may shed additional light
on this issue.
The findings regarding the impact of educational level are
potentially interesting but require further exploration. To date, few
studies of EPSE have specifically targeted highly educated older
adults (Haught, 1985; Haught et al., 2000), and we are unaware of
96
any study that has compared EPSE in older adults with different
levels of formal education. In fact, most studies attempt to ensure
that older and younger adults have equivalent levels of formal
education, leading to minimal educational variation in the current
literature (see Table 1). Because formal education is highly associated with performance on formal reasoning tasks (Schaie, 1996),
the observed relationship between education and EPSE may not be
surprising. However, a common assumption is that everyday problem solving is at least in part distinct from schooled abilities (see
Berg & Klaczynski, 1996). This poses potentially interesting questions regarding what factors may underlie the apparent relationship
between formal education and EPSE. For example, what aspects of
performance on everyday tasks are impacted by educational attainment, and under what conditions does this relationship
emerge? Future studies of EPSE may benefit from examining the
impact of this potentially important variable.
It is important to note that the lack of a significant difference
between effect estimates derived from studies with a problemsolving versus a decision-making focus supports the commonality
of these literatures. Specifically, the current analyses provide evidence that these findings can be meaningfully combined into a
single homogeneous population of effects. Although this does not
imply that these literatures are interchangeable, it appears that
some factors associated with age impact EPSE in a homogenizing
manner, regardless of objectively defined outcome criteria (i.e.,
decision accuracy, quality of decision rationale, number of safe/
effective solutions, solution quality ratings). A potentially fruitful
contribution of the decision-making literature toward understanding EPSE involves apparent age differences in application of
reasoning biases. Older adults, perhaps in reaction to reductions in
information-processing capabilities, may rely less on analytic processing and more on heuristics to guide decision making (Yates &
Patalano, 1999). Heuristic strategies tend to simplify information
processing by providing shortcuts through complex problems
(Yates & Patalano, 1999) but can also lead to systemic deviations
or biases and, subsequently, less than optimal solutions (Peters
et al., 2000). Although few studies have directly assessed age
differences in susceptibility to bias, findings of age-related increases in illusory correlations (Mutter & Pliske, 1994) have been
suggested to reflect an increase in the use of heuristic processes
(see Peters et al., 2000). In addition, both older and middle-aged
adults are reportedly more prone than younger adults to display
biased reasoning, as evidenced by the uncritical acceptance of
information that reflects their own positions and the rejection of
evidence perceived to be inconsistent with their beliefs (Klaczynski & Robinson, 2000). Whether age-related changes in susceptibility to bias are systematically related to EPSE remains an interesting question for future research.
Our analyses did not reveal systematic differences in effect sizes
with regard to quality of study methodology, method of test
administration (i.e., computerized vs. paper and pencil), or ageappropriate content. In addition, other moderators that have been
suggested by previous investigators (i.e., well- vs. ill-structured
task) were not supported, although this may reflect the interrelatedness of these concepts, as previously identified (Allaire & Marsiske, 2002), and the lack of clear distinctions between these
problem types within the extant literature. The similarity of effect
estimates in studies using quantity versus quality scoring criteria
also deserves special comment. It has been argued that fluency
criteria excessively penalizes older adults performance and mischaracterizes effective problem solving (Berg et al., 1999). Evidence that experts tend to use less information to solve problems
(e.g., Meyer et al., 1995) has been used to support the contention
that older adults use a less exhaustive and more selective approach
to problem solving and therefore may not generate strategies
previously found to be unproductive (Berg et al., 1999). An alternate view holds that a higher number of generated solutions does
in fact reflect better EPSE, as evidenced by the positive relationships observed between the number of generated solutions and
solution quality (see Haught et al., 2000; Heidrich & Denney,
1994). The current findings are consistent with the latter perspective. The magnitude of the age differences was not reduced when
quality was emphasized. Rather, age differences in EPSE were
reliable and consistent regardless of whether the criteria emphasize
number of safe and effective solutions generated, quality of rationale, or decision accuracy.
et al., 1999; Willis, Jay, Diehl, & Marsiske, 1992), and only three
reported data appropriate for between-older-age-group effect size
calculations (Diehl et al., 1995; Marsiske & Willis, 1995; Whitfield et al., 1999). It is possible that everyday problem solving may
reflect different underlying processes with advanced age. It is
interesting to note that, although scoring emphasis did not emerge
as a significant moderator in our comparison of young/middleaged versus older effect estimates, quality versus fluency outcomes
did yield statistically independent results in a study of everyday
cognition in participants aged 60 90 (Allaire & Marsiske, 2002).
Similarly, factor analytic solutions of cognitive measures may
differ with advanced age (Kaufman, Lichtenberger, & McLean,
2001), and reasoning tasks may elicit reliance on different cognitive processes in adults aged 75 and over (Kaufman & Lichtenberger, 2002).
Finally, although effects obtained from the extant literature
present evidence of decreased EPSE in older adults, the current
results do not directly speak to the functional implications of this
age difference. We agree with previous statements that understanding the functional impact of variations in EPSE is a crucial step
toward advancing this literature (e.g., Allaire & Marsiske, 2002).
The few studies to date that have investigated this issue report
strong relationships between performance on everyday problemsolving tasks and functional independence (Allaire & Marsiske,
2002; Diehl et al., 1995). In addition, although much work to date
has focused on problem solving in healthy older adults, little
attention has been focused toward everyday problem solving and
decision making in groups that may be at risk for age-associated
cognitive impairments. If age differences in EPSE reflect a decline
in basic mental abilities, one may predict that groups at risk for
accelerated cognitive decline would be at increased risk for poor
problem solving and decision making. Such findings could have
significant implications for maintenance of optimal functioning.
References
References marked with an asterisk indicate studies included in the
meta-analysis.
Allaire, J. C., & Marsiske, M. (1999). Everyday cognition: Age and
intellectual ability correlates. Psychology and Aging, 14, 627 644.
Allaire, J. C., & Marsiske, M. (2002). Well- and ill-defined measures of
everyday cognition relationship to older adults intellectual ability and
functional status. Psychology and Aging, 17, 101115.
*Artistico, D., Cervone, D., & Pezzuti, L. (2003). Perceived self-efficacy
and everyday problem solving among young and older adults. Psychology and Aging, 18, 68 79.
Baltes, P. B. (1993). The aging mind: Potentials and limits. Gerontologist,
33, 580 594.
Baltes, P. B. (1997). On the incomplete architecture of human ontogeny:
Selection, optimization, and compensation as foundation of developmental theory. American Psychologist, 52, 366 380.
Bandura, A. (1977). Social learning theory. Englewood Cliffs, NJ: Prentice
Hall.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York:
Freeman.
Berg, C. A., & Klaczynski, P. A. (1996). Practical intelligence and problem
solving: Searching for perspectives. In F. Blanchard-Fields & T. M.
Hess (Eds.), Perspectives on cognitive change in adulthood and aging
(pp. 323357). Boston: McGraw-Hill.
*Berg, C. A., Meegan, S. P., & Klaczynski, P. (1999). Age and experiential
differences in strategy generation and information requests for solving
97
98
Diehl, M., Willis, S. L., & Schaie, K. W. (1995). Everyday problem solving
in older adults: Observational assessment and cognitive correlates. Psychology and Aging, 10, 478 491.
*Dobson, S. H. (2000). Precursors of adult age differences in abstract and
practical problem solving. Dissertation Abstracts International: Section
BThe Sciences and Engineering, 60, 3601.
*DZurilla, T. J., Maydeu-Olivares, A., & Kant, G. L. (1998). Age and
gender differences in social problem-solving ability. Personal and Individual Differences, 25, 241252.
Egger, M., Smith, G. D., Schneider, M., & Minder, C. (1997). Bias in
meta-analysis detected by a simple, graphical test. British Medical
Journal, 315, 629 634.
*Feifel, H., & Strack, S. (1989). Coping with conflict situations: Middleaged and elderly men. Psychology and Aging, 4, 26 33.
*Finucane, M. L., Slovic, P., Hibbard, J. H., Peters, E., Mertz, D. K., &
McGregor, D. G. (2002). Aging and decision-making competence: An
analysis of comprehension and consistency skills in older versus
younger adults. Journal of Behavioural Decision Making, 15(2), 141
164.
Fitzgerald, J. M. (2000). Younger and older jurors: The influence of
environmental supports on memory performance and decision making in
complex trials. Journal of Gerontology: Psychological Sciences, 55B,
323331.
*Haught, P. A. (1985). Problem-solving performance of young and older
adults. Dissertation Abstracts International: Section BThe Sciences
and Engineering, 46, 3293.
*Haught, P. A., Hill, L. A., Nardi, A. H., & Walls, R. T. (2000). Perceived
ability and level of education as predictors of traditional and practical
adult problem solving. Experimental Aging Research, 26, 89 101.
Hedges, L. V. (1984). Estimation of effect size under nonrandom sampling:
The effects of censoring studies yielding statistically insignificant mean
differences. Journal of Educational Statistics, 9(1), 61 85.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis.
San Diego, CA: Academic Press.
*Heidrich, S. M., & Denney, N. W. (1994). Does social problem-solving
differ from other types of problem-solving during the adult years?
Experimental Aging Research, 20(2), 105126.
*Hershey, D. A., & Farrell, A. H. (1999). Age differences on a procedurally oriented test of practical problem solving. Journal of Adult Development, 6, 8794.
*Hershey, D. A., & Wilson, J. A. (1997). Age differences in performance
awareness on a complex financial decision-making task. Experimental
Aging Research, 23, 257273.
Johnson, M. M. (1990). Age differences in decision making: A process
methodology for examining strategic information processing. Journal of
Gerontology: Psychological Sciences, 45(2), P75P78.
Johnson, M. M. (1993). Thinking about strategies during, before, and after
making a decision. Psychology and Aging, 8, 231241.
Johnson, M. M. S. (1997). Individual differences in the voluntary use of a
memory aid during decision making. Experimental Aging Research,
23(1), 33 43.
Johnson, M. M. S., & Drungle, S. C. (2000). Purchasing OTC medications:
The influence of age and familiarity. Experimental Aging Research,
26(3), 245261.
Kaufman, A. S., & Lichtenberger, E. O. (2002). Assessing adolescent and
adult intelligence. Boston: Allyn & Bacon.
Kaufman, A. S., Lichtenberger, E. O., & McLean, J. E. (2001). Two- and
three-factor solutions of the WAISIII. Assessment, 8(3), 267280.
Kirasic, K. C., Allen, G. L., Dobson, S. H., & Binder, K. S. (1996). Aging,
cognitive resources, and declarative learning. Psychology and Aging, 11,
658 670.
Klaczynski, P. A., & Robinson, B. (2000). Personal theories, intellectual
ability, and epistemological beliefs: Adult age differences in everyday
reasoning biases. Psychology and Aging, 15, 400 416.
99
Handbook of the psychology of aging (4th ed., pp. 287307). New York.
Academic Press.
Willis, S. L., Jay, G. M., Diehl, M., & Marsiske, M. (1992). Longitudinal
change and prediction of everyday task competence in the elderly.
Research on Aging, 14(1), 68 91.
Willis, S. L., & Marsiske, M. (1993). Manual for the Everyday Problems
Test. University Park, PA: Pennsylvania State University.
Willis, S. L., & Schaie, K. W. (1986). Practical intelligence in later
adulthood. In R. J. Sternberg & R. K. Wagner (Eds.), Practical intelligence: Nature and origins of competence in the everyday world (pp.
236 268). New York: Cambridge University Press.
*Wright, B. M., Carscaddon, D. M., & Lambert, S. D. (2000). Sex,
education, and cautiousness: Implications for counselors. Adultspan
Journal, 2(2), 113122.
Yates, J. F., & Patalano, A. L. (1999). Decision making in aging. In D. C.
Park, R. W. Morrell, & K. Shifren (Eds.), Processing of medical information in aging patients: Cognitive and human factors perspectives (pp.
3154). Mahwah, NJ: Erlbaum.
*Zwahr, M. D., Park, D. C., & Shifren, K. (1999). Judgments about
estrogen replacement therapy: The role of age, cognitive abilities, and
beliefs. Psychology and Aging, 14, 179 191.