Psychology and Aging
2011, Vol. 26, No. 3, 716 –730
© 2011 American Psychological Association
0882-7974/11/$12.00 DOI: 10.1037/a0022513
Decreasing Complexity of Affective Space in Older Adults Lower on
Cognitive Control: Affective Effects in a Nonaffective Task and With
Nonaffective Stimuli
Lee H. Wurm
Wayne State University
Many theoretical accounts predict that as people age, they rely increasingly on affect. At least one
account (Dynamic Integration Theory) makes the additional prediction that an accompanying effect of
aging is a narrowing of affective space. These predictions were tested in the context of the relatively
automatic low-level cognitive process of lexical access (auditory word recognition). Experiment 1 used
emotion words and Experiment 2 used nonemotion words. Both experiments provided support for both
predictions. Compared to younger adults (ns ⫽ 36 and 56), older adults (ns ⫽ 36 and 54) showed larger
but less complex effects of dimensions of affective connotation. In addition, older adults with more
cognitive resources showed a data pattern like that of younger adults, while those with fewer resources
did not. Affective effects emerge even in nonaffective contexts. The tight link between affect and
cognition is discussed.
Keywords: aging, auditory lexical decision, word recognition, emotion, cognitive resources
selves as increased attention to (and memory for) positive information. They are due at least in part to an emphasizing of positive
information by older adults (Charles, Mather, & Carstensen, 2003;
Mather & Carstenson, 2003); there may also be a de-emphasizing
of negative information (e.g., Labouvie-Vief & Medler, 2002;
Mather & Johnson, 2000; Mather & Knight, 2005).
Many models of cognitive-affective aging can predict positivity
effects, including two whose predictions will be examined in the
current study: Socioemotional Selectivity Theory (SST; e.g.,
Carstensen et al., 1999; Carstensen & Mikels, 2005; Mikels, Larkin, Reuter-Lorenz, & Carstensen, 2005) and Dynamic Integration
Theory (DIT; Labouvie-Vief 2003, 2005, 2009; Labouvie-Vief,
Grühn, & Mouras, 2009).
In SST age-related changes in affective processing reflect an
upward shift in the priority of socio-emotional goals, which is
related to a compressed time horizon as end-of-life issues become
more salient. A claim found in recent extensions of SST (though
not central to the original model) is that positivity effects are the
result of controlled processes, and available cognitive resources
are drawn upon to increase such positivity effects. According to
Knight, Seymour, Gaunt, Baker, Nesmith, and Mather (2007),
There is overwhelming evidence of age-related declines in many
cognitive processes (e.g., Mahmood, Adamo, Briceno, & Moffat,
2009; Radvansky, Zacks, & Hasher, 2005; Salthouse, 2004b), and
much work attempts to relate these declines to changes in the
capacity or function of specific brain regions (e.g., BlanchardFields, 2010; Cabeza, 2002; Reuter-Lorenz & Cappell, 2008;
Stern, Zarahn, Hilton, Flynn, DeLaPaz, & Rakitin, 2003). There
are equally well-documented age-related advances in areas relating to affective1 competence, complex problem-solving involving
affect, and emotion regulation (e.g., Blanchard-Fields, 2007, 2009;
Labouvie-Vief, 2003; Mather & Carstensen, 2005). Peters, Hess,
Västfjäll, and Auman (2007) argue that an enhanced role of affect
in cognitive processing among the elderly makes sense because it
is a more implicit form of information that can be processed more
automatically.
This hypothesized increased reliance on affect, coupled with
data and theory suggesting that older adults are driven more by
socio-emotional goals than by informational goals (e.g.,
Carstensen, Isaacowitz, & Charles, 1999; Mather & Carstensen,
2005; Labouvie-Vief, 2003, 2005), provides an explanation for
positivity effects in older adults. Although positivity effects are not
always found (Murphy & Isaacowitz, 2008), they manifest them-
“Although older adults may have fewer cognitive resources overall,
we predict that the resources they do have will be allocated in a
goal-consistent manner that enables successful emotion regulation and
the maintenance of positive emotional experience. Consequently, successful emotion regulation should be a function not only of age, but
also of overall cognitive resources. Those older adults whose cogni-
This article was published Online First March 28, 2011.
This research was supported by a fellowship from the Wayne State
University Institute of Gerontology. I thank Eric Fuller and Sean Seaman
for help with stimulus creation, Sebastiano Fisicaro for a valuable discussion of statistical issues, and Annmarie Cano for helpful criticism of a
previous version of this manuscript.
Correspondence concerning this article should be addressed to Lee
Wurm, Department of Psychology, Wayne State University, 5057
Woodward Avenue (7th floor), Detroit, MI 48202. E-mail:
[email protected]
1
Although emotion and affect are often used interchangeably, in this
paper I will use emotion to refer to discrete states like fear, disgust, or joy,
and affect, affective variable, and affective dimension to refer to other
measurable aspects of experience or judgment, like valence, arousal, or
potency (see below).
716
DECREASING COMPLEXITY OF AFFECTIVE SPACE
tive control abilities are least compromised by age should show the
largest positivity effects. Those older adults whose cognitive control
resources are most compromised should have the greatest difficulty
carrying out emotional goals and so should show the smallest positivity effects” (p. 706).
Mather and Knight (2005) found that older adults do indeed
recruit cognitive resources to accomplish socio-emotional goals,
and note further that
“The idea that older adults’ positivity effects in memory are the result
of cognitive control processes may seem counterintuitive. After all,
many studies have shown that older adults are worse than younger
adults at tasks requiring cognitive control. Nevertheless, even if their
strategic processing is less effective, older adults may be more successful in regulating emotion if they devote a larger proportion of their
resources to such goals” (p. 555).
DIT is an integrative model of emotional development designed
to explain the pattern of both gains and losses in cognitiveaffective functioning across the lifespan. DIT takes a long view of
life histories, interpretations of life events, outlooks about the
future, concepts of the self, and so on, and bears primarily on
overarching constructs such as attachment style and psychological
adjustment. According to DIT, positivity effects are related to
affect optimizing, which is an automatic process associated with
declining cognitive resources in aging (e.g., Labouvie-Vief, 2003).
Positivity effects thus result from automatic rather than controlled
processes.
DIT also posits that older adults exhibit a narrowing of affective
space as a result of cognitive decline. Narrowing refers to a
reduction of the dimensionality of complex situations, meaning
that some aspects of an object, event, or situation are considered at
any one time but other aspects are not. Judging becomes less
cognitively taxing because the representational space is collapsed
(cf. Paulhus & Lim, 1994). According to DIT, older people in
general are less able to maintain complex representations because
of a lowered availability of the cognitive resources such representations require. Those older adults with a greater availability of
cognitive resources use them to maintain complex affective representations more like those of younger adults (i.e., to counter the
trend toward narrowed representations).
Keil and Freund (2009) found what might be such a narrowing
of affective space in a study in which participants rated the valence
and arousal of stimuli. Younger adults’ ratings showed a curvilinear relationship between valence and arousal. Very high-arousal
stimuli were rated as either quite pleasant or quite unpleasant,
while low-arousal stimuli tended to be intermediate on valence.
The relationship for older adults was more linear and unidimensional. Valence ratings decreased as arousal ratings increased, and
this correlation between valence and arousal became stronger with
age.
In contrast, recent research from SST-associated investigators
does not show such a narrowing of affective space. In one study in
which participants had to think about anticipated endings (e.g.,
visiting a meaningful spot for the last time), older adults’ experience of the mixed complex emotion of poignancy did not differ
from that of younger adults (Ersner-Hershfield, Mikels, Sullivan,
& Carstensen, 2008). In another study, Charles (2005) had older
and younger participants view film clips depicting injustice, and
717
then rate their feelings on five negative emotions: anger, contempt,
disgust, fear, and sadness. Younger adults reported stronger feelings of disgust, compared to the other emotions. Older adults, on
the other hand, reported similarly high feelings for anger, disgust,
and contempt. Charles interpreted this to mean that the older
adults’ emotional responses were more heterogeneous than the
younger adults’ responses. In the present context such findings
predict that older adults should show as much affective complexity
than younger adults, or perhaps even more.
Affect and Language Processing
The role of affect in language processing has received little
attention, which is surprising because language shapes perception
and provides the context for emotion and affect (Barrett, Lindquist,
& Gendron, 2007). The few studies to use words as stimuli do not
even agree about basic phenomena such as whether one observes
a positivity effect (e.g., Grühn, Smith, & Baltes, 2005; Kensinger,
2008). In addition, nearly all of this work has focused on the
processing of longer sentences or texts presented visually (e.g.,
Isaacowitz, Löckenhoff, Lane, Wright, Sechrest, Reidel, & Costa,
2007; Stine-Morrow, Miller, & Hertzog, 2006; Titone, Koh, Kjelgaard, Bruce, Speer, & Wingfield, 2006). An apparent exception to
this is a study by Knight, Maines, and Robinson (2002), in which
the authors included a homophone spelling task and an auditory
verbal memory task as two of their measures. However, theirs was
a study of mood congruence effects in memory rather than a study
of language processing, and the participants were mood-induced.
The role of affect in spoken language processing by older adults
thus remains largely unknown. This relative neglect matters because speech remains the primary means by which people interact
and exchange information. This is especially true of older people,
who are less likely than younger people to use technologies such
as the internet and text-messaging.
In this study, I will explore the roles of age and affect in a
low-level cognitive process called lexical access. Lexical access
occurs when a visual pattern or sound wave makes contact with the
mental representation of a word in a person’s mind, a process that
takes place several times per second during reading or listening.
Although there are well-documented declines in some functions
related to spoken language comprehension, such as processing
speed, fluid intelligence, and memory (e.g., Craik & Salthouse,
2000; Salthouse, 1993, 1996, 2004a; Wingfield, Tun, & McCoy,
2005; Zacks, Hasher, & Li, 2000), older adults maintain very good
performance in this area of daily living (Wingfield & StineMorrow, 2000).
At first glance, a relatively automatic process like lexical access
might seem to be a poor choice for investigating the role of
available cognitive resources; however, this method offers several
advantages in studying affective processes in aging. First, it allows
for differentiation between competing explanations of age-related
differences in affective processing, because some models rule out
such effects in automatic processes. Second, it is important to
begin to study this completely unexplored territory because of the
centrality of spoken language in interpersonal relationships. Third,
as will be seen in the next paragraphs, we have repeatedly found
effects of affective variables in lexical access for younger people,
and learning about age differences in these effects can help refine
theoretical models.
718
WURM
Colleagues and I have explored the role of affective dimensions
in lexical access for college-age participants in several studies.
Participants rate stimuli on one or more affective dimensions. A
separate group of participants then hears these words and performs
some task that taps into lexical processing, such as lexical decision
(deciding as quickly as possible whether something is a real word)
or naming (repeating what is heard). We find that the reaction
times (RTs) of the second group are related to the ratings made by
the first group (see below).
Some of the affective dimensions we have studied are evaluation, potency, and activity (Osgood, 1969; Osgood & Suci, 1955).
Evaluation is a good versus bad dimension, activity is a dimension
that carries connotations ranging from slow to fast, and potency is
a dimension ranging from weak to strong. Evaluation is generally
used interchangeably with valence.2 Some example stimuli from
Wurm et al. (2004b) illustrate what the other two dimensions mean
in the context of word connotations: rabbit receives high participant ratings on activity but low ratings on potency, while the
reverse is true for tree; fire is high on both dimensions and egg is
low on both dimensions.
Participants can easily provide their judgment of any word’s
connotation on these dimensions by using a number line (e.g.,
Vakoch & Wurm, 1997; Wurm, Vakoch, & Seaman, 2004b).
There is no a priori reason to suspect that lexical access would be
influenced by affective variables such as these, but it is (Vakoch &
Wurm, 1997; Wurm & Vakoch, 1996; Wurm et al., 2004b; see
also Wurm, 2007; Wurm & Seaman, 2008; Wurm & Vakoch,
2000; Wurm, Vakoch, Seaman, & Buchanan, 2004c; Wurm, Whitman, Seaman, Hill, & Ulstad, 2007). For example, faster lexical
decision times have been found for words high on evaluation (i.e.,
positive rather than negative words; e.g., Wurm & Vakoch, 1996;
Wurm et al., 2004b). Both of these last studies also found a
significant interaction between evaluation, potency, and activity.
Such effects can be understood in terms of a two-pass model of
meaning extraction. The first pass is very fast but relatively crude,
and operates in terms of a small number of behaviorally relevant
dimensions of affective connotation (such as evaluation, potency,
and activity). The effects suggest a strong connection between
perception, affect, and action. Even during the process of recognizing a word, the listener’s perceptual system is engaged in
preparing real-world behavioral responses that might be appropriate (e.g., Wurm, 2007; Wurm & Seaman, 2008).
No similar published work exists for older adults. It is therefore
unknown whether lexical access in older adults is affected by the
same kinds of variables, and in the same ways, as lexical access in
younger adults. Answering this question will have theoretical
implications for models of cognitive aging and models of affective
processing across the lifespan.
Both experiments of the current study address the same two
major questions. The first is whether older people show larger
effects of affective variables in lexical decision, compared to
younger people. Any theoretical account in which older people
place increased reliance on affect would predict this, but it has
never been tested.
The second major question has to do with DIT’s prediction of
narrowed affective space in older adults. As an example of what
this means in the context of the mental representations of words,
thorn and tarantula were given similar low ratings on evaluation
by the participants in Wurm et al. (2004b) (both Ms ⫽ 1.8 on the
8-point evaluation scale) but they were given fairly different
ratings on activity (M ⫽ 2.1 for thorn and 5.3 for tarantula).
Incorporating both pieces of information results in a more complex
representation; thorn is low on evaluation and on activity, and
tarantula is low on evaluation but relatively high on activity. A
narrower way to organize this information in affective space is to
simply classify both items as low on evaluation and to make no use
of their activity difference (or, to classify each item according to its
activity, but to make no use of their evaluation ratings).
It should be noted that neither DIT nor SST makes direct
predictions about affective processing at the level of lexical access.
Nevertheless, a number of experiments involving low-level cognitive processing have garnered support for each theory (e.g., Jain
& Labouvie-Vief, 2010; Knight et al., 2007; Mather & Knight,
2006; Wurm, Labouvie-Vief, Aycock, Rebucal, & Koch, 2004a)
so it is worthwhile to examine some predictions derived from each
one. Experiments 1 and 2 will test the following hypotheses.
Hypotheses
1.
Older adults should show larger effects of affective variables than younger adults. The affective
variables included in Experiment 1 are evaluation, potency, and activity, defined above. Any
model in which older people rely more heavily
on affect or emotion than younger people makes
this prediction.
2.
Older adults should show narrowed affective
space (i.e., less complex effects of affective variables) compared to younger adults. DIT predicts
this but SST does not.
In Experiment 1, I will test these hypotheses with emotion
words (defined below). Experiment 2 builds on Experiment 1 by
using nonemotion words and investigating the cognitive resource
question more directly.
Experiment 1: Emotion Words
In Experiment 1, I examined the auditory processing of pure
emotion words, which Clore, Ortony, and Foss (1987; Ortony,
Clore, & Foss, 1987) defined as emotion words that are relatively
free of cognitive and behavioral connotations. The RT data came
from Experiment 1 of Wurm et al. (2004a), a project driven by
different theoretical questions. With the exception of age group,
the independent variables analyzed here were not examined in the
2004 study.
2
Wurm et al. (2004b) gathered evaluation ratings for 100 nouns. Thirtyseven of the nouns are also listed in the Bradley and Lang (1999a) valence
norms. The correlation between evaluation and valence for these words is
.92, which is as strong as the correlation expected for repeated ratings of
the same items on the same dimension (e.g., Barenboym, Wurm, & Cano,
2010; Wurm, Cano, & Barenboym, 2010).
DECREASING COMPLEXITY OF AFFECTIVE SPACE
Method
Participants.
Participants were 36 younger and 36 older
native speakers of English. All reported normal hearing. Older
adults (29 women and 7 men) ranged in age from 52 to 92 years
(M ⫽ 74, SD ⫽ 10.1), and were recruited from Senior Centers in
the Detroit area. They received no monetary compensation for
their participation. As in Wurm et al. (2004a, Footnote 1), I
excluded data from the two older participants who were younger
than 59, as they might not be representative of older participants.
Younger adults (29 women and 7 men) ranged in age from 18 to
28 years (M ⫽ 22, SD ⫽ 3.2) and were recruited from the subject
pool in the Department of Psychology at Wayne State University.
They received extra credit in a psychology course for their participation. Participants were asked to indicate the highest level of
education attained on a 7-point ordinal scale. Nearly every younger
adult indicated “some college work completed,” which was the
median older response.
Materials. Twelve critical words, and their values on evaluation, activity, and potency, were taken from Morgan and Heise
(1988). An additional 36 items were chosen from the same source.
Twelve were retained as filler items, and the remaining 24 were
changed into legal pronounceable pseudowords (see Appendix A).
Stimuli were recorded at the end of the carrier phrase “When that
happened I felt ________” and saved as individual .wav files.
Wurm et al. (2004a) provide additional details.
Procedure. Participants participated at a time of their own
choosing. Older adults participated between 9:30 a.m. and 2:30
p.m., with the majority between 11:30 a.m. and 1:30 p.m. Younger
adults participated between 9:30 a.m. and 4:00 p.m. Participants
were tested one at a time in a sound-shielded chamber or quiet
room. Digitized speech files were played over headphones at a
comfortable listening level chosen by each participant. Participants
were directed to make a speeded lexical decision about the item in
sentence-final position by pressing one button for words and
another for pseudowords. RTs were measured from the acoustic
onset of each stimulus. Before the main experiment, participants
heard a practice list of similar composition. The practice list
contained 12 stimuli that were not used in the main experiment.
Analytic strategy. I analyzed the log RTs with a multilevel
linear mixed-effects analysis of covariance, with participant as
random effect (e.g., Baayen, 2004; Pinheiro & Bates, 2000). This
is a variant of the repeated-measures regression analysis outlined
by Lorch and Myers (1990), suitable for research in which there
are multiple observations per participant.3
In addition to age group, evaluation, potency, and activity, I
included uniqueness point (UP) location in milliseconds and word
frequency (Baayen, Piepenbrock, & Gullikers, 1995) in the statistical models (both log transformed because of skewed distributions). The UP (Marslen-Wilson, 1987) is the point in the acoustic
signal at which a word diverges from all other words in the
language and represents the earliest moment at which a person
could initiate a positive lexical decision. These variables are included primarily for variance reduction, because they generally
have significant effects in lexical processing studies. They also
help to eliminate alternative explanations of the data (e.g., if UP
location and evaluation were correlated, then any apparent evaluation effect might actually be caused by UP location if the latter
had not been controlled for).
719
Main effects were entered in one step, followed by two-way
interactions, three-way interactions, and the four-way interaction.
At each step entry of effects was simultaneous. Significance levels
were evaluated at the step at which an effect was entered.
I assessed the interactions between age group and each dimension (i.e., age ⫻ evaluation; age ⫻ potency; age ⫻ activity).
Significant interactions in which the older adults’ effects are larger
are taken as evidence in support of the hypothesis of increased
reliance on affect (Hypothesis 1).
Based on previous research (Wurm & Vakoch, 1996; Wurm et
al., 2004b) I expected that RTs for younger adults would depend
on the three-way interaction between evaluation, potency, and
activity. That is, the effect of evaluation on younger adults’ RTs
should be conditional on potency and activity. This would be
evidence of complex underlying lexical representations, because
the combination of all three dimension weights co-determines
lexical access times. The narrowed affective space hypothesis
predicts a four-way interaction: The three-way interaction between
evaluation, potency, and activity should depend further on age. It
should be significantly weaker for older adults, because of their
less complex (i.e., narrower) representations (Hypothesis 2).
Results and Discussion
Data from three older and two younger participants were excluded because of error rates ⬎.25. To improve normality, 2
impossibly fast RTs and 14 RTs slower than 2,500 ms past
stimulus offset were excluded from analysis (totaling 1.9% of the
data; the same cut-off was used for both older and younger adults).
As noted above, all participants reported normal hearing. Nevertheless, poorer accuracies for older adults might indicate hearing
loss, and hearing loss could interact with variables of interest (e.g.,
Wingfield et al., 2005). Fortunately, mean accuracies were very
high and similar (M ⫽ 96.5 and 96.8% for older and younger
participants, respectively; t(63) ⫽ ⫺0.27, p ⫽ .7901). The results
of the analysis are shown in Table 1.
Main effects. The effects of word frequency and potency
were marginally significant, and there was no effect of activity.
There were significant effects of UP location (later UPs predicted
slower RTs), age (older participants were 13% slower), and evaluation (higher evaluation predicted faster RTs). Each of these
significant effects was in the direction expected based on previous
lexical processing research (UP location: Baayen, Wurm, & Aycock, 2007; Wurm, 1997; evaluation: Vakoch & Wurm, 1997;
Wurm et al., 2004b; age: Wurm et al., 2004a).
The evaluation effect can be considered a positivity effect, in
that words with more positive connotations had faster RTs (see
Footnote 2). Older participants were substantially slower than
younger participants. As noted above, older adults’ accuracy was
every bit as good as that of younger adults, so the age difference
was restricted to RTs.
3
There is currently a debate in the statistics literature about the appropriate way to calculate degrees of freedom for this type of analysis. The
lower bound is the number of participants minus the number of IVs in the
model, and the upper bound is the number of participants times the number
of observations, minus the number of IVs in the model. This ongoing
debate means that the procedure for calculating effect sizes is also unclear.
WURM
720
Table 1
Summary of Stepwise Analysis for Variables Predicting Log Auditory Lexical Decision Time
(Experiment 1)
Variable
Step 1
Log word frequency
Log uniqueness point
Age (younger)
Evaluation
Potency
Activity
Step 2
Evaluation ⫻ potency
Evaluation ⫻ activity
Potency ⫻ activity
Younger ⫻ evaluation
Younger ⫻ potency
Younger ⫻ activity
Step 3
Evaluation ⫻ potency ⫻ activity
Younger ⫻ evaluation ⫻ potency
Younger ⫻ evaluation ⫻ activity
Younger ⫻ potency ⫻ activitity
Step 4
Younger ⫻ evaluation ⫻ potency ⫻ activity
Two-way interactions. The two-way interactions between
affective dimensions are of limited importance given the hypotheses of the current study, and none was significant. The theoretically important interactions involve age group. Age interacted
significantly with two of the affective dimensions: evaluation and
activity. These interactions are plotted in Figure 1, which shows
the predictions of the statistical model with all variables set at their
Figure 1. Log RT as a function of age group and evaluation (left panel)
or activity (right panel). Both older slopes are significantly different from
zero. For younger participants, only the evaluation slope is significant.
Evaluation, potency, and activity ratings were taken from Morgan and
Heise (1988).
B
SE B
t
p
0.01
0.73
⫺0.10
⫺0.05
0.01
⫺0.01
0.00
0.01
0.02
0.01
0.01
0.02
1.93
50.71
⫺4.83
⫺8.70
1.74
⫺0.47
.0544
.0000
.0000
.0000
.0832
.6363
0.01
⫺0.02
⫺0.01
0.05
⫺0.02
0.10
0.05
0.04
0.09
0.01
0.01
0.04
0.26
⫺0.40
⫺0.11
5.53
⫺1.36
2.50
.7941
.6927
.9119
.0000
.1746
.0128
⫺0.82
⫺0.00
⫺0.05
0.02
0.31
0.10
0.07
0.15
⫺2.59
⫺0.03
⫺0.70
0.11
.0098
.9766
.4834
.9095
⫺0.96
0.48
⫺2.01
.0453
means except those involved in the interaction. In both cases the
effect was significantly larger for the older participants, supporting
the prediction of increased reliance on affect for older adults
(Hypothesis 1). Inspection of the left-hand panel in Figure 1 shows
that the age difference was largest for words lower on evaluation
(i.e., the more negative words).
Planned comparisons performed separately by age group revealed that for younger participants there was only an effect of
evaluation (B ⫽ ⫺0.02, SE B ⫽ 0.01, p ⫽ .0004). Neither activity
nor potency had significant effects (both ps ⬎ .13). In contrast, all
three affective dimensions had significant main effects for the
older participants (B ⫽ ⫺0.07, SE B ⫽ 0.01, p ⬍ .0001; B ⫽
⫺0.07, SE B ⫽ 0.03, p ⫽ .0401; and B ⫽ 0.02, SE B ⫽ 0.01, p ⫽
.0387 for evaluation, activity, and potency, respectively). These
findings are consistent with any theoretical account positing that
older adults place increased reliance on affect, including both DIT
and SST.
Three-way interactions. The interaction between evaluation,
potency, and activity was significant, as predicted based on previous research (Wurm & Vakoch, 1996; Wurm et al., 2004a). It
will be described in the next section. No other three-way interaction was significant.
Four-way interaction.
The four-way interaction between
evaluation, potency, and activity, and age was significant. I probed
the interaction by assessing the three-way interaction between
evaluation, potency, and activity separately by age group. It was
significant for younger participants (B ⫽ ⫺1.41, SE B ⫽ 0.40, p ⫽
.0005) but not for older participants (B ⫽ ⫺0.20, SE B ⫽ 0.49, p ⫽
.6771). This supports DITs prediction of narrowed affective space
for older adults.
Figure 2 shows the evaluation ⫻ potency ⫻ activity interaction
for younger adults. “High” and “Low” refer to values 1 SD above
and below the mean, respectively. Words were not separated into
categories on any dimension in the current study. Dichotomizing is
DECREASING COMPLEXITY OF AFFECTIVE SPACE
721
merely a convenient way to show the nature of the slope differences. The dimension weights shown in Appendix A were used as
continuous regressors in the analyses, and thus error bars or
confidence intervals on such figures would be meaningless.
As the figure shows, when potency and activity matched, there
was a facilitative effect of evaluation (i.e., words with “good”
connotations had faster RTs). When potency and activity mismatched, there was either no effect of evaluation, or higher evaluation was associated with slower RTs.
The observation that this interaction was significant only for
younger adults is crucial. For them, lexical access time for any
given word depended on the combination of that word’s values on
all three affective variables. The lexical representations for
younger adults must be structured in a way that allows for the
effect of one affective variable (e.g., evaluation) to be conditioned
upon the other two (i.e., potency and activity). Older adults’ lexical
representations are evidently not structured in this complex way.
Thus, both older and younger adults showed an effect of evaluation, but for older adults it was just that (a simple effect of
evaluation) while for younger adults, the effect of evaluation
depended on both potency and activity. The four-way interaction
thus provides support for DITs prediction of narrowed affective
space for older adults.
adults, affective variables (such as evaluation, potency, and activity) affect word recognition performance even for nonemotion
words, and even in tasks like lexical decision that have nothing to
do with affect (e.g., Wurm, 2007). This suggests that the affect/
cognition link might be tighter than generally assumed (Duncan &
Barrett, 2007; Izard, 2007).
In Experiment 2, I also switched from Osgood’s affective dimensions to valence and arousal. This change was made for
pragmatic reasons relating to the availability of ratings and also to
allow more direct contact with theoretical models (most are built
around valence and arousal; few use Osgood’s dimensions). The
change of dimensions does not seem crucial because as will be
seen, the results of the two experiments were consistent. Based on
previous research I predicted larger positivity effects for stimuli
low on arousal, for both age groups (Bradley & Lang, 1999b;
Kensinger, 2008; Purkis, Lipp, Edwards, & Barnes, 2009; see
below). A key theoretical question for the current study will be the
relative sizes of these valence ⫻ arousal interactions.
Another purpose of Experiment 2 was to examine the role
played by the availability of cognitive resources. As a measure of
the available cognitive resources, I used Mather and Knight’s
(2005) cognitive control construct, which is each participant’s
average age-group-specific z-score across three tasks tapping into
different aspects of cognitive processing (see Method for more
detail).
As seen above, recent elaborations of SST predict that people
with higher availability of cognitive resources should show larger
positivity effects. In contrast, in DIT, available cognitive resources
are used to maintain complex affective representations more like
those of younger adults (i.e., to counter the trend toward narrowed
representations). DITs general prediction then is that older adults
with a higher availability of cognitive resources should show
performance more similar to younger adults. DIT thus predicts that
older adults with fewer cognitive resources should show larger
positivity effects, in direct contrast to the prediction of the modified SST. This is because according to DIT, positivity effects are
related to affect optimizing, as discussed above.
There is a separate reason to expect a smaller effect of arousal
for older adults with greater cognitive resources. DIT posits that
older adults find high physiological arousal unpleasant and costly
(see also Keil & Freund, 2009; Kiecolt-Glaser, McGuire, Robles,
& Glaser, 2002), and that they seek to minimize its effects if they
are able. According to DIT, older adults with more resources are
better able to do this, and thus should show smaller arousal effects.
In contrast, SST has little to say about arousal. Mather and
Knight (2005) concluded that arousal did not affect their observed
age ⫻ valence interactions, and that people of all ages showed
similar arousal effects in memory. This view predicts only that any
observed arousal effects will not differ by age.
Experiment 2: Nonemotion Words
Hypotheses
Figure 2. Log RT as a function of potency, activity, and evaluation, for
the younger participants. None of these variables was dichotomized for the
analysis. This interaction was not significant for the older participants.
Evaluation, potency, and activity ratings were taken from Morgan and
Heise (1988).
In Experiment 2 I used nonemotion words, which I defined as
words with no connection to any discrete emotional state. It is
important to assess the hypotheses of the current study with
nonemotion words because nearly all studies on affective processing in aging have used stimuli chosen explicitly to be extreme on
some affective dimension(s), providing what might be a limited
view of the link between affect and routine cognition. In younger
1.
Older adults should show larger effects of affective variables (valence and arousal in Experiment
2) than younger adults. DIT predicts further that
the larger main effects of valence and arousal
will be especially prominent for older adults with
lower availability of cognitive resources. For va-
WURM
722
lence, recent extensions of SST predict the opposite: Higher cognitive control scores should be
associated with a larger positivity effect (the
original model does not make a prediction either
way here). For arousal, SST predicts only that
any observed effect will not be related to age.
2.
Older adults should show narrowed affective
space compared to younger adults. DIT predicts
further that the narrowing should also be especially prominent for older adults with lower cognitive control scores. That is, these older adults
should have smaller interactions than younger
adults. In contrast, older adults with higher cognitive control scores should not show narrowed
affective space compared to younger adults.
Method
Participants.
Participants were 54 older and 56 gendermatched younger adults. All were native speakers of English and
all reported normal hearing. Older adults ranged in age from 60 to
92 (M ⫽ 71, SD ⫽ 8.0), and were recruited through a newsletter
distributed in the Tri-County Metropolitan Detroit area. All were
community-dwelling and received monetary compensation for
their participation. Younger adults ranged in age from 18 to 46
(M ⫽ 22, SD ⫽ 5.9) and were recruited from the subject pool in
the Department of Psychology at Wayne State University. They
received extra credit in a psychology course for participating.
Participants were asked how many years of education they had
completed. One participant in each age group failed to provide a
response. For the 108 remaining participants, education level did
not differ by age group (M ⫽ 15.2, SD ⫽ 2.6 and M ⫽ 14.6, SD ⫽
1.9 for older and younger participants, respectively; t(106) ⫽ 1.45,
p ⫽ .1493).
Materials. Forty common nouns (Appendix B) were selected
from the Bradley and Lang (1999a) norms. An additional forty
nouns were selected from a large dictionary and were made into
pseudowords by changing a randomly determined phoneme. A
male speaker unfamiliar with the purpose or hypotheses of the
study recorded the stimuli, which were digitized at a sampling rate
of 22.05 kHZ and stored in individual disc files. Valence and
arousal values were taken from Bradley and Lang (1999a).
Procedure. Availability of cognitive resources is central to
several hypotheses. As noted above, I used Mather and Knight’s
(2005), cognitive control construct to operationalize the availability of cognitive resources. Specifically, I used each participant’s
average age-group-specific z-score across three tasks:
The refresh task (Johnson, Reeder, Raye, & Mitchell, 2002) is a
measure of how readily participants can bring recently presented
information to mind. Participants performed 165 naming trials,
repeating words into a microphone as each one was printed onscreen. On a random one-fifth of the trials a dot appeared instead
of a word, and the participant had to recall the last word they read.
The mean RT for these refresh trials was divided by the mean RT
on normal trials to create a ratio that controls for the overall slower
responding of the older participants.
The modified sentence span task (Daneman & Carpenter, 1980)
is a measure of working memory. Participants are required to
retain in memory the final word of sentences that have recently
been read, and to provide these words when cued. The number of
sentences up to the cue is gradually increased until the participant
fails two consecutive sets of sentences. The score is the total
number of sentences read up until the stop rule, minus any failures
along the way.
The executive component of the Attention Network Test is a
measure of executive function developed by Fan, McCandliss,
Sommer, Raz, and Posner (2002). Participants are asked to categorize arrows on-screen as either pointing to the left or right,
sometimes in the presence of conflicting information. After discarding incorrect trials, the mean RT for congruent trials is subtracted from the mean RT for incongruent trials. The online version of the test was used (Attention Network Task, 2002).
The procedure for the lexical decision task was the same as in
Experiment 1. Older adults participated at a time of their choosing.
All older adults participated between 9:00 a.m. and 6:00 p.m., with
the majority between 9:00 a.m. and 2:00 p.m. Younger adults
participated at a time of their choosing between 9:00 a.m. and 5:00
p.m.
Analytic strategy. RTs were analyzed by the same method as
in Experiment 1. To test whether older adults show larger effects
of the affective variables than younger adults (Hypothesis 1), I
assessed the interactions between age group and each dimension
(i.e., age ⫻ valence; age ⫻ arousal). DIT’s additional prediction
that these larger effects should be especially prominent in older
adults low on cognitive control was assessed by looking at threeway interactions (age ⫻ control ⫻ valence; age ⫻ control ⫻
arousal).
Narrowed affective space for older people was operationalized
as older people showing a smaller valence ⫻ arousal interaction
than younger people (Hypothesis 2). The more specific prediction
made by DIT is that the three-way interaction between age, valence, and arousal will be moderated by cognitive control. That is,
older people with high cognitive control should have lexical representations as complex as younger people, so the predicted valence ⫻ arousal interaction should be as large for these people as
for younger adults. An age difference is predicted for people low
on cognitive control, though. Here, older people are expected to
show a narrowing of affective space, that is, a weaker interaction
between valence and arousal.
Put another way, according to DIT, the age ⫻ valence ⫻ arousal
interaction should be significant only for people low on cognitive
control. For those high on cognitive control, the valence ⫻ arousal
interaction is predicted to be just as strong for older people as it is
for younger people.
Results and Discussion
Younger adults had significantly better scores on two of the
three components of the cognitive control measure. They had
faster RTs in the Attention Network Test (p ⬍ .001) and significantly longer sentence spans (p ⫽ .0186). There was no difference
in the age groups’ scores on the refresh task (p ⫽ .1542) or in the
composite cognitive control measure itself, t(108) ⫽ 0.20, p ⫽
.8445. Readers are reminded that this is a measure of the cognitive
DECREASING COMPLEXITY OF AFFECTIVE SPACE
resources each participant has available. It is predicted to moderate
several effects for older participants.
To improve the normality of the RT distribution for the main
lexical decision data, three impossibly fast RTs and 10 RTs slower
than 2,500 ms past stimulus offset were excluded from analysis
(totaling ⬍1% of the data; the same cut-off was used for both older
and younger adults).
As noted above, all participants reported normal hearing. Nevertheless, poorer accuracies for older adults might indicate hearing
loss, and such loss might interact with variables of interest (e.g.,
Wingfield et al., 2005). Fortunately, mean accuracies were high
and similar (M ⫽ 94.5 and 92.7% for the older and younger
participants, respectively; t(108) ⫽ 1.40, p ⫽ .1646). The results of
the analysis are shown in Table 2.
Main effects. There were significant main effects of word
frequency (more frequent words had faster RTs), UP location
(words with later UPs had slower RTs), and valence (words with
higher valence [i.e., more positive ratings] had faster RTs). In
addition, younger adults’ RTs were 22% faster than older adults’
RTs. Each of these effects is in the direction expected based on
previous lexical processing research (UP location: Baayen et al.,
2007; Wurm, 1997; valence: Vakoch & Wurm, 1997; Wurm et al.,
2004b; word frequency: Vakoch & Wurm, 1997; age: Wurm et
al., 2004a).
The valence effect is a positivity effect, as words with more
positive connotations had faster RTs. Older participants were
substantially slower than younger participants. As in Experiment
1, this age difference was restricted to the RTs – older adults’
accuracy was as good as that of younger adults.
Two-way interactions. Age group interacted with valence.
Figure 3 shows this interaction, and although the slope difference
is not visually impressive, it is reliable. The positivity effect was
significantly larger for older adults than for younger adults (the
older valence coefficient is 17% stronger than the younger one),
723
Figure 3. Log RT as a function of age group and valence. Valence ratings
were taken from Bradley and Lang (1999a).
supporting the general prediction of greater reliance on affect for
older adults (Hypothesis 1). This finding is consistent with any
theoretical account positing that older adults place increased reliance on affect, including both DIT and SST.
Inspection of Figure 3 shows that the age effect is larger for
lower stimulus valences (i.e., for more negative stimuli). Differences in the log RT values translate to age differences of 211 and
139 ms for valence scores of 2 and 8, respectively. This is essentially the same thing observed in Experiment 1 (see the left panel
Table 2
Summary of Stepwise Analysis for Variables Predicting Log Auditory Lexical Decision Time
(Experiment 2)
Variable
Step 1
Log word frequency
Log uniqueness point
Age (younger)
Valence
Arousal
Control
Step 2
Valence ⫻ arousal
Valence ⫻ control
Arousal ⫻ control
Younger ⫻ valence
Younger ⫻ arousal
Younger ⫻ control
Step 3
Valence ⫻ arousal ⫻ control
Younger ⫻ valence ⫻ arousal
Younger ⫻ control ⫻ valence
Younger ⫻ control ⫻ arousal
Step 4
Younger ⫻ valence ⫻ arousal ⫻ control
B
SE B
t
p
⫺0.03
0.08
⫺0.19
⫺0.02
0.00
⫺0.01
0.00
0.01
0.02
0.00
0.00
0.02
⫺14.76
7.77
⫺9.20
⫺10.12
0.49
⫺0.61
.0000
.0000
.0000
.0000
.6211
.5418
0.02
⫺0.00
⫺0.01
0.01
⫺0.00
⫺0.01
0.00
0.00
0.00
0.00
0.01
0.03
8.81
⫺1.55
⫺2.48
2.06
⫺0.68
⫺0.33
.0000
.1217
.0133
.0391
.4946
.7392
⫺0.00
0.01
⫺0.00
0.00
0.00
0.00
0.01
0.01
⫺1.06
2.02
⫺0.56
0.28
.2887
.0431
.5777
.7790
⫺0.02
0.01
⫺2.70
.0070
724
WURM
of Figure 1, and Footnote 2 on the near equivalence of evaluation
and valence).
Age group did not interact significantly with arousal. The coefficient was twice as strong for older adults, which is in the
direction expected under Hypothesis 1, but the interaction did not
approach significance.
Cognitive control did interact with arousal, as shown in Figure
4. Dichotomizing cognitive control allows for visualization of the
slope change, but cognitive control was not dichotomized for the
analysis. For participants with lower cognitive control scores, RTs
were slower for stimuli higher on arousal. This relationship weakened for participants with higher cognitive control scores.
Thus, the effect of arousal on RTs depended on cognitive
control. Readers are reminded that the cognitive control measure is
a composite of performance on three separate tasks and taken as a
measure of available cognitive resources (Mather & Knight, 2005).
For participants with lower cognitive control scores, increasing
arousal slowed responses. This relationship weakened for participants with higher cognitive control scores. Put another way, high
arousal interfered with lexical processing for people with lower
cognitive control scores, but the interference effect was attenuated
by higher cognitive control scores. DIT predicts this pattern for
older adults but makes no prediction either way for younger adults.
They would have the cognitive resources needed to reduce the
arousal effect, but there is no reason to suppose that this would be
a goal of theirs (i.e., there is no reason to believe they find high
arousal unpleasant).
The valence ⫻ arousal interaction was significant and is shown
in Figure 5. As predicted, the strength of the valence effect
decreased as arousal increased. This pattern maps nicely onto
research from several studies. Bradley and Lang (1999b) asked
people to make speeded judgments of the valence of pictures and
found a RT pattern nearly identical to Figure 5 for judgments of
stimuli related to people, animals, or objects. Purkis et al. (2009)
Figure 4. Log RT as a function of cognitive control and arousal. Cognitive control was not dichotomized for the analysis. Arousal ratings were
taken from Bradley and Lang (1999a).
Figure 5. Log RT as a function of valence and arousal. Arousal was not
dichotomized for the analysis. Valence and arousal ratings were taken from
Bradley and Lang (1999a).
asked participants to make speeded judgments of the valence and
arousal of pictures and words. They found that judgments of
positive stimuli were much faster than judgments of negative
stimuli only if the stimuli were low on arousal. For high-arousal
stimuli the relationship was reversed and weaker. Finally, Kensinger (2008) found a positivity effect in older adults’ memory
performance only for items low on arousal.
Three-way interactions. DIT allowed refinements to Hypothesis 1, such that older adults high on cognitive control should
have valence and arousal main effects like those of younger adults,
while older adults low on cognitive control should have valence
and arousal main effects larger than those of younger adults.
Neither of these predicted interactions (age ⫻ cognitive control ⫻
valence, and age ⫻ cognitive control ⫻ arousal) was significant.
The significant valence ⫻ arousal ⫻ age interaction is shown in
Figure 6. The valence ⫻ arousal interaction was significantly
weaker for older adults, providing support for the general version
of DIT’s narrowed affective space hypothesis (Hypothesis 2).
However, the more detailed version of Hypothesis 2 implies a
four-way interaction. As will be seen in the next section, the
three-way interaction shown in Figure 6 depended further on
cognitive control scores, in the way DIT predicts.
Four-way interaction.
The four-way interaction between
valence, arousal, age, and cognitive control was significant. That
is, the three-way interaction shown in Figure 6 depended significantly on cognitive control scores. To explore this interaction I
divided the participants into those with low versus high cognitive
control scores using a median split (recall that the cognitive control
scores for older and younger participants did not differ), and re-ran
the analysis.
Dichotomizing in this way is artificial and reduces statistical
power to some extent, but it allows a very clear answer to the key
question about the role of cognitive control. DIT predicts that older
people with higher cognitive control should show a valence ⫻
DECREASING COMPLEXITY OF AFFECTIVE SPACE
arousal interaction just as strong as that seen in the younger adults,
because they use their available cognitive resources to maintain
more complex affective representations. Consistent with this prediction, the age ⫻ valence ⫻ arousal interaction was not significant for participants high on cognitive control (B ⫽ 0.00, SE B ⫽
0.01, p ⫽ .5435). Older participants high on cognitive control had
a valence ⫻ arousal interaction that was indistinguishable from
that seen in the younger participants.
In contrast, older people with lower cognitive control should
show a weaker valence ⫻ arousal interaction than younger adults,
because of a decreased ability to maintain complex representations. Indeed, the age ⫻ valence ⫻ arousal interaction was significant for these participants (B ⫽ 0.01, SE B ⫽ 0.01, p ⫽ .0293). As
can be seen in Figure 7, older participants low on cognitive control
had a valence ⫻ arousal interaction that was significantly weaker
than younger participants, indicating affective representations that
are less complex.4
This pattern of results indicates that the general age-dependent
pattern seen in Figure 6 was driven by participants lower on
cognitive control. For low-control participants, older adults had a
smaller valence ⫻ arousal interaction than younger adults (Figure
7). In contrast, for high-control participants, there was no age
difference in the size of the valence ⫻ arousal interaction. This is
consistent with DIT’s view that older adults use cognitive resources to maintain more complex affective representations. It is
only older adults with a lower availability of such cognitive
resources who showed a narrowing of affective space.
Summary and Concluding Discussion
A broad range of cognitive processes has been examined in the
affective aging literature (see Mather, 2006, for a review), but this
has not included the routine process of lexical access. Probing
age-related differences in lexical access complements the larger
literatures that exist on aging and memory, attention, and visual
Figure 6. Log RT as a function of age group, valence, and arousal.
Arousal was not dichotomized for the analysis. Valence and arousal ratings
were taken from Bradley and Lang (1999a).
725
Figure 7. Log RT as a function of age group, valence, and arousal, for
participants below the median on cognitive control. Arousal was not
dichotomized for the analysis. Valence and arousal ratings were taken from
Bradley and Lang (1999a).
search, because spoken word recognition is a low-level automatic
cognitive process that remains largely intact in the elderly (e.g.,
Wingfield & Stine-Morrow, 2000).
Effects of affective variables in lexical processing have been
demonstrated many times with college-age participants (e.g., Vakoch & Wurm, 1997; Wurm, 2007; Wurm & Seaman, 2008; Wurm
& Vakoch, 1996, 2000; Wurm et al., 2003, 2004b, 2004c, 2007),
but only rarely with older adults (Wurm et al., 2004a). As a result
it has not previously been possible to explore some theoretically
important age-related similarities and differences in processing.
This is important not only because of the role language plays as the
context for affect (Barrett et al., 2007) but also because of the fact
that speech remains the primary means by which most older people
interact with others.
Both experiments provided support for both major hypotheses
of the current study. It was hypothesized that older adults would
show larger effects of affective variables compared to younger
adults. The positivity effect was indeed larger for older adults than
for younger adults (Figure 1, left, and Figure 3) and older adults
showed a larger effect of activity than younger adults in Experiment 1 (Figure 1, right). This is in line with data and theory
suggesting that as people age, they increasingly rely on affect. This
may be because affective information is more reliable or more
automatically processed (e.g., Peters et al., 2007), or it may be
4
A reviewer raised the possibility that cognitive control might act as a
proxy for the real underlying variable of interest, education. That is, older
people with more education might simply have more complex lexical
networks, which would have nothing to do with affect per se. Additional
analyses show that this potential explanation is not tenable. First, the
four-way interaction between education, age group, valence, and arousal is
not significant (p ⫽ .47). Thus, cognitive control is not merely a proxy for
years of education. Second, the four-way interaction shown in Table 2
remains significant even if education is added to the statistical model.
726
WURM
because of the changing motivational goals of people as they
age (e.g., Carstensen et al., 1999; Labouvie-Vief, 2003). Whatever
the reasons, older people’s lexical access times are co-determined
by (main effects of) affective variables to a greater degree than are
younger people’s. This conclusion holds both for the pure
emotion words in Experiment 1 and the nonemotion words in
Experiment 2.
The observation that younger adults, too, showed positivity
effects runs counter to literature showing either no such effect for
younger adults, or even a negativity effect (e.g., Kensinger, 2008;
Mather & Knight, 2006). It could be that routine processing of
spoken words is sufficiently different from the other kinds of
cognitive processes studied to produce this different outcome. The
discrepancy might also be because of differences in the nature of
the stimuli and how they are presented. Most researchers use visual
presentation of pictures, but in our previous studies using auditory
presentation of words, we have observed that more positive stimuli
produce faster RTs for younger adults (Vakoch & Wurm, 1997;
Wurm et al., 2004b). It could also have to do with differences in
how stimuli are chosen. In most research on affective processing,
stimuli are chosen to be extreme on one or more affective dimensions. That is not generally true in our research. These possible
explanations await additional research.
The second major hypothesis of the current study was that older
adults would exhibit evidence of less complex (i.e., narrower)
lexical representations. DIT notes that there is typically an agerelated reduction in the availability of cognitive resources, making
older people generally less able to maintain complex affective
representations (e.g., Labouvie-Vief, 2003). In contrast, although
SST does not make any direct predictions about lexical access,
research from SST-associated investigators in related areas of
cognition either predicts no such reduction in affective complexity
(Ersner-Hershfield et al., 2008), or that older adults should show
increased affective complexity compared to younger adults
(Charles, 2005).
I used interaction effects between affective variables as an index
of the complexity of the underlying representations, and the prediction of reduced complexity was supported in both experiments.
Younger adults’ lexical access times were co-determined by a
complex interaction between evaluation, potency, and activity in
Experiment 1. Older adults’ lexical access times were not.
Younger adults’ lexical access times were co-determined by an
interaction between valence and arousal in Experiment 2. Older
adults’ lexical access times were, too, but to a significantly lesser
degree unless the older adults had a higher availability of cognitive
resources.
Certain language acts (e.g., irony and many kinds of jokes) rely
on intentionally mismatched information, and proper understanding requires maintenance of a multi-faceted representation. For
example, to understand sarcasm a listener must maintain and
integrate three kinds of information that are in varying degrees of
conflict: the literal message that was spoken, the way in which it
was spoken, and the overall situational context. Might the results
of the current study mean that subtleties or nuances having to do
with these complex representations are more likely to be missed by
some older people? This question awaits future research.
What do other kinds of models have to say about these results?
There is a speed view which posits that older participants recruit
more cognitive resources to compensate for age-related reductions
in processing speed (e.g., Salthouse, 1996; Schaie, 1994; see also
Cabeza, 2002; Reuter-Lorenz & Cappell, 2008; Stern et al., 2003).
The data do not support this view. First, mean RT for older adults
high on cognitive control (981 ms) was not better than that for
older adults low on cognitive control (975 ms), indicating that the
group with more cognitive resources did not deploy them to
combat a speed deficiency. Second, it is difficult to guess what
sorts of data patterns would have been shown by older adults, if
they had not had to recruit additional cognitive resources. Finally,
somewhat different processes or mechanisms would need to be
invoked to adequately explain the range of effects in the current
study with a speed-based explanation, making them seem very ad
hoc.
Another alternative is the inhibition view (e.g., Hasher & Zacks,
1988; Zacks et al., 2000), which posits that a decline in inhibitory
control is responsible for age-related cognitive deficits. This account can easily accommodate the finding that older adults showed
larger main effects of affective variables. The information carried
by these dimensions is irrelevant for someone performing lexical
decision (i.e., simply deciding if something is a real word in
English), and so the fact that it intrudes more for older adults is
consistent with this view. This account makes the wrong predictions, though, in the context of both four-way interactions. In
Experiment 1, the irrelevant information represented by evaluation, potency, and activity only intruded (if we want to look at it as
an intrusion) for younger participants (Figure 2). In Experiment 2
this account predicts that older adults with fewer cognitive resources should show an amplified valence ⫻ arousal interaction
compared to younger adults, and they showed a significantly
weaker one (Figure 7).
One potential limitation of the current study is its cross-sectional
design. It is impossible to be sure that the age differences I have
found reflect theoretically important age-related changes, or
whether they reflect the different histories, experiences, and characteristics of the two age groups under study. As Salthouse
(2004b) has noted, though, there are also valid criticisms of longitudinal designs, including the possibility of practice and learning
effects, possible effects of theoretically unimportant societal
changes, and so on. I echo his conclusion that it is probably best
for the field to continue to pursue both kinds of approach, and
because the particular issues under study have not previously been
examined at all, the current study is a good starting point.
A second limitation of the current study is that the emotion
ratings used, whether taken from Morgan and Heise (1988) or from
Bradley and Lang (1999a), were obtained from samples of younger
adults. It is possible that these ratings are less appropriate for older
adults than they are for younger adults. This is a problem shared
by all aging studies using these common and well-validated ratings, and I have two reasons to believe that it is not a serious
concern. First, in studies that compare older and younger stimulus
ratings, the correlations are generally extremely high (e.g., Balota,
Pilotti, & Cortese, 2001; Wurm et al., 2004a, Experiment 2).
Second, and more importantly, it is extremely difficult to see how
the patterns observed could be explained by age-inappropriate
ratings. One would need to explain how inappropriate ratings
could produce larger main effects but smaller interactions for older
people (unless the older people are high on cognitive control).
The current study sheds additional light on the mixed picture of
aging in the literature. On the one hand is the intriguing difference
DECREASING COMPLEXITY OF AFFECTIVE SPACE
in the age-related trajectories of cognition and affect. Aging is
accompanied by well-documented and predictable cognitive declines, but it is also often accompanied by improvements in several
areas of functioning related to affect and emotion regulation, and
older adults in general are as happy as or happier than younger
adults (e.g., McNeil, Stones, Kozma, & Andres, 1994; see Diener
& Suh, 1997, for a review of studies from several countries). On
the other hand, the current study adds to the small literature
underscoring just how tightly intertwined cognition and affect are
(Wurm, 2007). Duncan and Barrett (2007) argue that emotion is a
kind of cognition and that no distinction should be made between
them.
Izard (2007) went so far as to say “. . . there is no such thing as
an affectless mind” (p. 270). The current study makes the related
suggestion that there is no such thing as affectless cognitive
processing, even of nonemotion stimuli. This becomes increasingly important because of the age-related shift among older adults
toward socio-emotional goals as motivators. Emotion-related goals
affect the way older people seek information and what they do
with it, even in situations related to health and survival (see
Löckenhoff & Carstensen, 2004, for a review). Overreliance on
simple aspects of affect (possibly with an emphasis on the positive
aspects of a stimulus or response option), combined with an
under-reliance on the more complex interplay between positive
and negative aspects, can easily lead to suboptimal decisionmaking that might have very serious consequences. The current
study shows that this is a general concern, rather than one that only
applies in explicitly affective contexts.
References
Attention Network Task. (2002). Online cognitive test. Retrieved from
http://sacklerinstitute.org/users/jin.fan/ant/ANTFirstPage.html
Baayen, R. H. (2004). Statistics in psycholinguistics: A critique of some
current gold standards. Mental Lexicon Working Papers, 1, 1– 45.
Baayen, R. H., Piepenbrock, R., & Gullikers, L. (1995). The CELEX lexical
database [CD-ROM]. Linguistic Data Consortium. Philadelphia: University of Pennsylvania.
Baayen, R. H., Wurm, L. H., & Aycock. J. (2007). Lexical dynamics for
low-frequency complex words: A regression study across tasks and
modalities. The Mental Lexicon, 2, 419 – 463.
Balota, D. A., Pilotti, M., & Cortese, M. J. (2001). Subjective frequency
estimates for 2,938 monosyllabic words. Memory & Cognition, 29,
639 – 647.
Barenboym, D. A., Wurm, L. H., & Cano, A. (2010). A comparison of
stimulus ratings made online and in person: Gender and method effects.
Behavior Research Methods, 42, 273–285.
Barrett, L. F., Lindquist, K. A., & Gendron, M. (2007). Language as
context for the perception of emotion. Trends in Cognitive Science, 11,
327–332.
Blanchard-Fields, F. (2007). Everyday problem solving and emotion: An
adult developmental perspective. Current Directions in Psychological
Science, 16, 26 –31.
Blanchard-Fields, F. (2009). Flexible and adaptive socio-emotional problem solving in adult development and aging. Restorative Neurology and
Neuroscience, 27, 539 –550.
Blanchard-Fields, F. (2010). Neuroscience and aging. In J. C. Cavanaugh,
C. K. Cavanaugh, J. Berry, & R. West (Eds.), Aging in America, vol. 1:
Psychological aspects (pp. 1–25). Santa Barbara, CA: Praeger/ABCCLIO.
Bradley, M. M., & Lang, P. J. (1999a). Affective norms for English words
(ANEW): Instruction manual and affective ratings. Technical Report
727
C-1, The Center for Research in Psychophysiology, University of
Florida.
Bradley, M. M., & Lang, P. J. (1999b). Fearfulness and affective evaluations of pictures. Motivation and Emotion, 23, 1–13.
Cabeza, R. (2002). Hemispheric asymmetry reduction in older adults: The
HAROLD model. Psychology and Aging, 17, 85–100.
Carstensen, L., Isaacowitz, D. M., & Charles, S. T. (1999). Taking time
seriously: A theory of socioemotional selectivity. American Psychologist, 54, 165–181.
Carstensen, L. L., & Mikels, J. A. (2005). At the intersection of emotion
and cognition: Aging and the positivity effect. Current Directions in
Psychological Science, 14, 117–121.
Charles, S. T. (2005). Viewing injustice: Greater emotion heterogeneity
with age. Psychology and Aging, 20, 159 –164.
Charles, S. T., Mather, M., & Carstensen, L. L. (2003). Aging and emotional memory: The forgettable nature of negative images for older
adults. Journal of Experimental Psychology: General, 132, 310 –324.
Clore, G. L., Ortony, A., & Foss, M. A. (1987). The psychological
foundations of the affective lexicon. Journal of Personality and Social
Psychology, 53, 751–766.
Craik, F. I. M., & Salthouse, T. A. (Eds.). (2000). The handbook of aging
and cognition (2nd ed.). Mahwah, NJ: Erlbaum.
Daneman, M., & Carpenter, P. A. (1980). Individual differences in working
memory and reading. Journal of Verbal Learning and Verbal Behavior,
19, 450 – 466.
Diener, E., & Suh, M. E. (1997). Subjective well-being and age: An
international analysis. In K. W. Schaie & M. P. Lawton (Eds.), Annual
review of gerontology and geriatrics (Vol. 17, pp. 304 –324). New York:
Springer.
Duncan, S., & Barrett, L. F. (2007). Affect is a form of cognition: A
neurobiological analysis. Cognition & Emotion, 21, 1184 –1211.
Ersner-Hershfield, H., Mikels, J. A., Sullivan, S. J., & Carstensen, L. L.
(2008). Poignancy: Mixed emotional experience in the face of meaningful endings. Journal of Personality and Social Psychology, 94, 158 –167.
Fan, J., McCandliss, B. D., Sommer, T., Raz, A., & Posner, M. I. (2002).
Testing the efficiency and independence of attentional networks. Journal
of Cognitive Neuroscience, 14, 340 –347.
Grühn, D., Smith, J., & Baltes, P. B. (2005). No aging bias favoring
memory for positive material: Evidence from a heterogeneity–
homogeneity list paradigm using emotionally toned words. Psychology
and Aging, 20, 579 –588.
Hasher, L., & Zacks, R. T. (1988). Working memory, comprehension, and
aging: A review and a new view. In G. Bower (Ed.), The psychology of
learning and motivation (pp. 193–225). San Diego, CA: Academic
Press.
Isaacowitz, D. M., Löckenhoff, C. E., Lane, R. D., Wright, R., Sechrest, L.,
Reidel, R., & Costa, P. T. (2007). Age differences in recognition of
emotion in lexical stimuli and facial expressions. Psychology and Aging,
22, 147–159.
Izard, C. E. (2007). Basic emotions, natural kinds, emotion schemas, and
a new paradigm. Perspectives on Psychological Science, 2, 260 –280.
Jain, E., & Labouvie-Vief, G. (2010). Compensatory effects of emotion
avoidance in adult development. Biological Psychology, 84, 497–513.
Johnson, M. K., Reeder, J. A., Raye, C. L., & Mitchell, K. J. (2002).
Second thoughts versus second looks: An age-related deficit in reflectively refreshing just-activated information. Psychological Science, 13,
64 – 67.
Keil, A., & Freund, A. M. (2009). Changes in the sensitivity to appetitive
and aversive arousal across adulthood. Psychology and Aging, 24, 668 –
680.
Kensinger, E. A. (2008). Age differences in memory for arousing and
nonarousing emotional words. Journal of Gerontology, 63B, P13–P18.
Kiecolt-Glaser, J. K., McGuire, L., Robles, T. F., & Glaser, R. (2002).
728
WURM
Emotions, morbidity, and mortality: New perspectives from psychoneuroimmunology. Annual Review of Psychology, 53, 83–107.
Knight, B. G., Maines, M. L., & Robinson, G. S. (2002). The effects of sad
mood on memory in older adults: A test of the mood congruence effect.
Psychology and Aging, 17, 653– 661.
Knight, M., Seymour, T. L., Gaunt, J. T., Baker, C., Nesmith, K., &
Mather, M. (2007). Aging and goal-directed emotional attention: Distraction reverses emotional biases. Emotion, 7(4), 705–714.
Labouvie-Vief, G. (2003). Dynamic integration: Affect, cognition, and the
self in adulthood. Current Directions in Psychological Science, 12,
201–206.
Labouvie-Vief, G. (2005). The psychology of emotions and aging. In M. L.
Johnson (Ed.), The Cambridge handbook of age and aging (pp. 229 –
237). Cambridge, UK: Cambridge University Press.
Labouvie-Vief, G. (2009). Dynamic Integration Theory: Emotion, cognition, and equilibrium in later life. In V. L. Bengtson, D. Gans, N. M.
Pulney, & M. Silverstein (Eds.). Handbook of theories of aging (2nd ed.,
pp. 277–293). New York: Springer.
Labouvie-Vief, G., Grühn, D., & Mouras, H. (2009). Dynamic emotioncognition interactions in adult development: Arousal, stress, and the
processing of affect. In H. B. Bosworth & C. Hertzog (Eds.), Aging and
cognition: Research methodologies and empirical advances. Decade of
behavior (2000 –2010) (pp. 181–196). Washington, DC: American Psychological Association.
Labouvie-Vief, G., & Medler, M. (2002). Affect optimization and affect
complexity: Modes and styles of regulation in adulthood. Psychology
and Aging, 17, 571–587.
Löckenhoff, C. E., & Carstensen, L. L. (2004). Socioemotional selectivity
theory, aging, and health: The increasingly delicate balance between
regulating emotions and making tough choices. Journal of Personality,
72, 1395–1424.
Lorch, R. R., & Myers, J. L. (1990). Regression analyses of repeated
measures data in cognitive research. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16, 149 –157.
Mahmood, O., Adamo, D., Briceno, E., & Moffat, S. D. (2009). Age
differences in visual path integration. Behavioural Brain Research, 205,
88 –95.
Marslen-Wilson, W. D. (1987). Functional parallelism in spoken word
recognition. Cognition, 25, 71–102.
Mather, M. (2006). Why memories may become more positive as people
age. In B. Uttl, N. Ohta, & A. L. Siegenthaler (Eds.), Memory and
emotion: Interdisciplinary perspectives (pp. 135–158). Malden, MA:
Blackwell Publishing.
Mather, M., & Carstensen, L. L. (2003). Aging and attentional biases for
emotional faces. Psychological Science, 14, 409 – 415.
Mather, M., & Carstensen, L. L. (2005). Aging and motivated cognition:
The positivity effect in attention and memory. Trends in Cognitive
Sciences, 9, 496 –502.
Mather, M., & Johnson, M. K. (2000). Choice-supportive source monitoring: Do our decisions seem better to us as we age? Psychology and
Aging, 15, 596 – 606.
Mather, M., & Knight, M. (2005). Goal-directed memory: The role of
cognitive control in older adults’ emotional memory. Psychology and
Aging, 20, 554 –570.
Mather, M., & Knight, M. R. (2006). Angry faced get noticed quickly:
Threat detection is not impaired among older adults. Journal of Gerontology, 61B, 54 –57.
McNeil, J. K. Stones, M. J., Kozma, A., & Andres, D. (1994). Age
differences in mood: Structure, mean level, and diurnal variation. Canadian Journal on Aging, 13, 201–220.
Mikels, J. A., Larkin, G. R., Reuter-Lorenz, P. A., & Carstensen, L. L.
(2005). Divergent trajectories in the aging mind: Changes in working
memory for affective versus visual information with age. Psychology
and Aging, 20, 542–553.
Morgan, R. L., & Heise, D. (1988). Structure of emotions. Social Psychology Quarterly, 51, 19 –31.
Murphy, N. A., & Isaacowitz, D. M. (2008). Preferences for emotional
information in older and younger adults: A meta-analysis of memory and
attention tasks. Psychology and Aging, 23, 263–286.
Ortony, A., Clore, G. L., & Foss, M. A. (1987). The referential structure of
the affective lexicon. Cognitive Science, 11, 341–364.
Osgood, C. E. (1969). On the whys and wherefores of E, P, and A. Journal
of Personality and Social Psychology, 12, 194 –199.
Osgood, C. E., & Suci, G. J. (1955). Factor analysis of meaning. Journal
of Experimental Psychology, 50, 325–338 .
Paulhus, D. L., & Lim, D. T. K. (1994). Arousal and evaluative extremity
in social judgments: A dynamic complexity model. European Journal of
Social Psychology, 24, 89 –99.
Peters, E., Hess, T. M., Västfjäll, D., & Auman, C. (2007). Adult age
differences in dual information processes: Implications for the role of
affective and deliberative processes in older adults’ decision making.
Perspectives on Psychological Science, 2, 1–23.
Pinheiro, J. C., & Bates, D. M. (2000). Mixed-effects models in S and
S-PLUS. New York: Springer.
Purkis, H. M., Lipp, O. V., Edwards, M. S., & Barnes, R. (2009). An
increase in stimulus arousal has differential effects on the processing
speed of pleasant and unpleasant stimuli. Motivation and Emotion, 33,
353–361.
Radvansky, G. A., Zacks, R. T., & Hasher, L. (2005). Age and inhibition:
The retrieval of situation models. Journal of Gerontology, 60B(5),
276 –278.
Reuter-Lorenz, P. A., & Cappell, K. A. (2008). Neurocognitive aging and
the compensation hypothesis. Current Directions in Psychological Science, 17, 177–182.
Salthouse, T. A. (1993). Speed and knowledge as determinants of adult age
differences in verbal tasks. Journal of Gerontology, 48, 29 –36.
Salthouse, T. A. (1996). The processing-speed theory of adult age differences in cognition. Psychological Review, 103, 403– 428.
Salthouse, T. A. (2004a). Localizing age-related individual differences in a
hierarchical structure. Intelligence, 32, 541–561.
Salthouse, T. A. (2004b). What and when of cognitive aging. Current
Directions in Psychological Science, 13, 140 –144.
Schaie, K. W. (1994). The course of adult intellectual development. American Psychologist, 49, 304 –313.
Stern, Y., Zarahn, E., Hilton, H. J., Flynn, J., DeLaPaz, R., & Rakitin, B.
(2003). Exploring the neural basis of cognitive reserve. Journal of
Clinical and Experimental Neuropsychology, 25, 691–701.
Stine-Morrow, E. A. L., Miller, L. M. S., & Hertzog, C. (2006). Aging and
self-regulated language processing. Psychological Bulletin, 132, 582–
606.
Titone, D. A., Koh, C. K., Kjelgaard, M. M., Bruce, S., Speer, S. R., &
Wingfield, A. (2006). Age-related impairments in the revision of syntactic misanalyses: Effects of prosody. Language and Speech, 49, 75–99.
Vakoch, D. A., & Wurm, L. H. (1997). Emotional connotation in speech
perception: Semantic associations in the general lexicon. Cognition and
Emotion, 11, 337–349. doi:10.1080/026999397379827
Wingfield, A., & Stine-Morrow, E. A. L. (2000). Language and speech. In
F. I. M. Craik & T. A. Salthouse (Eds.), Handbook of aging and
cognition (2nd. ed., pp. 359 – 416). Mahwah, NJ: Erlbaum.
Wingfield, A., Tun, P. A., & McCoy, S. (2005). Hearing loss in older
adulthood: What it is and how it interacts with cognitive performance.
Current Directions in Psychological Science, 14, 144 –148.
Wurm, L. H. (1997). Auditory processing of prefixed English words is both
continuous and decompositional. Journal of Memory and Language, 37,
438 – 461. doi:10.1006/jmla.1997.2524
Wurm, L. H. (2007). Danger and usefulness: An alternative framework for
understanding rapid evaluation effects in perception? Psychonomic Bulletin & Review, 14, 1218 –1225.
DECREASING COMPLEXITY OF AFFECTIVE SPACE
Wurm, L. H., Cano, A., & Barenboym, D. A. (2010). Ratings gathered
online vs. in person: Different stimulus sets and different statistical
conclusions. Manuscript submitted for publication.
Wurm, L. H., Labouvie-Vief, G., Aycock, J., Rebucal, K. A., & Koch,
H. E. (2004a). Performance in auditory and visual emotional Stroop
tasks: A comparison of older and younger adults. Psychology and Aging,
19, 523–535. doi:10.1037/0882-7974.19.3.523
Wurm, L. H., & Seaman, S. R. (2008). Semantic effects in naming and
perceptual identification, but not in delayed naming: Implications for
models and tasks. Journal of Experimental Psychology: Learning, Memory, and Cognition, 34(2), 381–398. doi:10.1037/0278-7393.34.2.381
Wurm, L. H., & Vakoch, D. A. (1996). Dimensions of speech perception:
Semantic associations in the affective lexicon. Cognition and Emotion,
10, 409 – 423. doi:10.1080/026999396380204
Wurm, L. H., & Vakoch, D. A. (2000). The adaptive value of lexical
connotation in speech perception. Cognition and Emotion, 14, 177–191.
doi:10.1080/026999300378923
729
Wurm, L. H., Vakoch, D. A., Aycock, J., & Childers, R. R. (2003).
Semantic effects in lexical access: Evidence from single-word naming.
Cognition and Emotion, 17, 547–565. doi:10.1080/02699930302302
Wurm, L. H., Vakoch, D. A., & Seaman, S. R. (2004b). Recognition of
spoken words: Semantic effects in lexical access. Language and Speech,
47, 175–204. doi:10.1177/00238309040470020401
Wurm, L. H., Vakoch, D. A., Seaman, S. R., & Buchanan, L. (2004c).
Semantic effects in auditory word recognition. Mental Lexicon Working
Papers, 1, 47– 62.
Wurm, L. H., Whitman, R. D., Seaman, S. R., Hill, L., & Ulstad, H. M.
(2007). Semantic processing in auditory lexical decision: Ear-ofpresentation and sex differences. Cognition and Emotion, 21, 1470 –
1495. doi:10.1080/02699930600980908
Zacks, R. T., Hasher, L., & Li, K. Z. H. (2000). Human memory. In F. I. M.
Craik & T. A. Salthouse (Eds.), Handbook of aging and cognition (2nd
ed., pp. 293–357). Mahwah, NJ: Lawrence Erlbaum.
Appendix A
Stimuli (Experiment 1)
Word
Evaluation
Potency
Activity
Afraid
Annoyed
Contented
Disgusted
Displeased
Glad
Happy
Irked
Petrified
Pleased
Scared
Terrified
⫺0.36
⫺0.72
1.29
⫺0.62
⫺0.67
1.31
1.3
⫺0.61
⫺0.39
1.34
⫺0.29
⫺0.46
⫺0.89
0.46
0.31
0.71
0.49
⫺0.05
0.03
0.45
⫺1.04
0.05
⫺0.91
⫺0.95
⫺0.19
0.22
⫺0.23
0.17
⫺0.17
0.27
0.38
0.29
⫺0.09
0.05
⫺0.01
⫺0.07
Note. Filler words: empty, shaken, outraged, delighted, passionate, unhappy, apprehensive, resentful, agitated,
overjoyed, fearful, deflated. Pseudowords: charned, cheerlace, depressed, downhorted, eloted, emballassed,
exsouted, frighkened, horrisied, hurk, irete, jealoos, joywess, lonefome, lonery, lovezhick, mab, melantoly,
micherable, overyelmed, prud, reliejed, regletful, sickemed. Presentation of all stimuli was auditory. Dimension
ratings were taken from Morgan and Heise (1988). Ratings were not available for the filler items.
(Appendix follows)
WURM
730
Appendix B
Stimuli (Experiment 2)
Word
Valence
Arousal
Bird
Car
Crime
Deceit
Diamond
Doll
Door
Fraud
Frog
Girl
Heart
Honey
Hospital
Hotel
Insect
Kettle
Kitten
Lamp
Lion
Method
Moment
Money
Mountain
Movie
Music
Ocean
Owl
Palace
Priest
Song
Stool
Stove
Thief
Tobacco
Tomb
Umbrella
Village
Window
Woman
Youth
7.27
7.73
2.89
2.9
7.92
6.09
5.13
2.67
5.71
6.87
7.39
6.73
5.04
6
4.07
5.22
6.86
5.41
5.57
5.56
5.76
7.59
6.59
6.86
8.13
7.12
5.8
7.19
6.42
7.1
4.56
4.98
2.13
3.28
2.94
5.16
5.92
5.91
6.64
6.75
3.17
6.24
5.41
5.68
5.53
4.24
3.8
5.75
4.54
4.29
6.34
4.51
5.98
4.8
4.07
3.22
5.08
3.8
6.2
3.85
3.83
5.7
5.49
4.93
5.32
4.95
3.98
5.1
4.41
6.07
4
4.51
6.89
4.83
4.73
3.68
4.08
3.97
5.32
5.67
Note. Pseudowords: appoor, ayven, borgo, boxswain, caset, compach, deecimel, doog, fant, husp, klaxom, lup,
matet, nogul, noigh, noush, paralount, pelfet, poslet, prama, proofile, ral, roggin, rondow, scurp, seconje, sering,
sinul, slesh, slounce, smouge, stoy, tendris, torpilla, treasor, wolt, wug, zam, zence, zollen. Presentation of all
stimuli was auditory. Dimension ratings were taken from Bradley and Lang (1999a).
Received May 14, 2010
Revision received November 30, 2010
Accepted December 6, 2010 䡲