111 / 218
109
Reward modulation of cognitive function: aging
0.010). A between subject analysis in this subgroup revealed that the older group earned more
reward than did the younger group (mean €9.01 (SE 0.32) vs. mean €10.50 (SE 0.20); t(26.448)
= 11.343, p < 0.001).
We were puzzled by this effect and reasoned that the age-related increase in total earnings
might originate from differences in the response deadline, which was set during a pre-test
practice phase (
methods
). When the response deadline was determined, participants were
instructed to respond as fast and accurately as possible. We reasoned that participants who
put more emphasis on the accuracy instruction would not respond as fast as possible during
practice. This would then result in longer, less stringent response deadlines during the actual
test. In the current paradigm, inaccurate responses, no matter how fast, are never rewarded.
Therefore, adopting such a cautious (slow and accurate) response strategy during practice
may result in higher earnings. For example, imagine two participants (A and B) who are
theoretically both able to respond within 400ms. If participant
A
responds cautiously during
the practice phase, the average response time during practice will be slower (e.g. 900ms)
than that of someone who emphasized speed during practice (participant B, e.g. 500ms). As
a consequence, participant A will have plenty of time to respond accurately on test, thereby
increasing the number of rewarded trials. By contrast, participant B will need to continue
to respond relatively fast. Participant B will thus make more errors, and therefore a lower
number of trials will be rewarded.
To test the idea that the response strategy during practice differed with age and that this would
lead to the observed age-related differences in earnings, we first assessed whether age was
associated with the length of the individually determined response deadlines. We observed
an overall age-related increase in the response deadline (i.e. across 4 trial-types: Arrow/Word
x Switch/Repeat), so that older participants were allowed to respond more slowly on test than
did younger participants (Age x Response deadline: (ρ) = 0.587, p <0.001). One might argue
that the differential Age x Reward effects on repeat and switch trials reported above might
Table 5.2
Reward x Task-switching effects for younger and older
subgroups*
Younger
(n=35)
Older
(n=33)
Difference
RT
-0.164
(-10.50;10.17)
-0.486
(-9.58;8.61)
-0.055;
P > 0.1
Accuracy
0.66
(-3.24;4.56)
-2.21
(-5.70;1.27)
-1.061;
P > 0.1
SAT
0.43
(0.25;0.61)
0.15
(0.01;0.29)
-2.755;
p = 0.006
SAT = Speed-Accuracy-Tradeoff = (z-speed- z-accuracy) / 2 ; RT = response times
* subgroups (study A, C and D in table 1 and figure 3a) were not confounded by
differential reward size