Positive affect, reward, and cognitive control 6
Supplementary Material
Analysis of practice effects in the AX-CPT
To look into possible practice effects, additional analyses including the factor Block (1, 2) in the respective baseline and reward phases were conducted. In line with our analyses reported in the main text, mean error rates and mean median RTs were analyzed in separate ANOVAS for Baseline and Reward phase performance. To avoid redundancy, here we will only report on the respective Block related results.
Baseline - Error data
A 2 (Affect: neutral, positive) x 2 (Block: 1, 2) x 4 (Cue-Probe-condition: AX, AY, BX, BY) mixed factors ANOVA revealed no main effect of Block (F < 1, p = .432). Furthermore, all interactions including Bock did not prove reliable (all F < 1, all p > .42). That is, the analysis did not reveal any performance changes between blocks. Only descriptively we found that errors on AY trial in the neutral group increased form Block 1 to Block 2, whereas the opposite was found in the positive group (i.e., decreased error rates in AY trials from Block 1 to Block 2, see Table S1).
Table S1. Mean error rates in % (SD in parentheses) in Baseline Blocks 1 and Block 2 as a factor of Affect group and Cue-Probe-condition.
neutral / positiveBlock / AX / AY / BX / BY / AX / AY / BX / BY
1 / 0.9 (1.48) / 13.97 (16.01) / 4.06 (6.57) / 0.31 (1.98) / 1.4
(1.8) / 8.13 (14.86) / 1.88 (5.33) / 0.71 (4.52)
2 / 0.95 (1.15) / 15.22 (16.82) / 2.81 (6.0) / 0.31 (1.98) / 1.0 (1.42) / 6.56 (10.2) / 1.56 (5.05) / 0.0
(0.0)
Baseline - RT data
The same ANOVA with mean median RTs showed a significant main effect of Block, F(1, 78) = 30.78, p < .001, η2p = .283, and a significant interaction of Block x Cue-Probe-condition, F(3, 234) = 7.71, p < .001, η2p = .09. Single comparisons showed significantly faster RTs in all cue-Probe-conditions from Baseline block 1 to 2 (all F > 10, all p < .01) except for AY trials (F < 1, p = .39, see Table S2 for full table of means).
Table S2. Mean median RTs in ms (SD in parentheses) in Baseline Block 1 and Block 2 as a factor of Affect group and Cue-Probe-condition.
neutral / positiveBlock / AX / AY / BX / BY / AX / AY / BX / BY
1 / 352 (54.95) / 524 (121.6) / 299 (93.57) / 309 (110.25) / 353 (53.81) / 478 (65.01) / 268 (96.23) / 293 (110.0)
2 / 341 (50.62) / 520 (137.15) / 243 (108.56) / 271 (120.41) / 338 (44.04) / 467 (52.5) / 224 (91.53) / 245 (114.75)
Reward phase - Error data
A 2 (Affect: neutral, positive) x 2 (Contingency: contingent, random) x 2 (Block: 1, 2) x 4 (Cue-Probe-condition: AX, AY, BX, BY) x 2 (Incentive: incentive, non-incentive) mixed factors ANOVA (see Table S3 for full table of means) revealed a main effect of Block, F(1, 76) = 31.34, p < .001, η2p = .292, which was further qualified by a significant interaction of Block x Cue-Probe-condition, F(3, 228) = 15.42, p < .001, η2p = .169. All other interactions with Block did not prove reliable (all F < 2.39, all p > .126). Planned comparisons revealed that the interaction of Block x Cue-Probe-condition is due to increasing error rates from Block 1 to 2 in AY trials only (F = 21.84, p < .001, 25.89 % vs. 32.7 %, all other cue-probe-sequences F < 3.04 and p > .085). This effect suggests that participants learn the utility of the A cue with increasing time on task (which in most of the cases is actually followed by the target).
Table S3. Mean error rates in % (SD in parentheses) in Reward phase Block 1 and Block 2 as a factor of Affect group, Contingency, Cue-Probe-condition, and Incentive.
neutral / positiveContingency / AX / AY / BX / BY / AX / AY / BX / BY
non-incentive Trials
Block 1
Contingent / 0.81 (0.92) / 47.05 (22.73) / 2.59 (5.32) / 0.0
(0.0) / 1.54 (2.06) / 36.25 (27.18) / 2.59 (5.32) / 0.63 (2.8)
Random / 0.45 (1.0) / 7.59 (10.31) / 0.63 (2.8) / 0.63 (2.8) / 0.9 (1.24) / 5.63 (9.49) / 0.0 (0.0) / 0.0
(0.0)
Block2
Contingent / 1.09 (2.07) / 50.8 (26.57) / 1.25 (3.85) / 1.88 (4.58) / 1.17 (3.07) / 48.13 (34.0) / 0.63 (2.8) / 0.71 (3.19)
Random / 0.9 (1.24) / 16.43 (16.39) / 3.13 (6.88) / 0.63 (2.8) / 1.17 (1.34) / 15.0 (16.52) / 1.25 (5.59) / 0.0
(0.0)
incentive Trials
Block 1
Contingent / 2.5 (2.42) / 46.88 (29.21) / 1.25 (5.59) / 1.88
(6.12) / 2.06 (2.34) / 43.13 (34.76) / 1.88 (4.58) / 0.0
(0.0)
Random / 0.99 (1.48) / 10.63 (14.21) / 3.13 (6.88) / 0.0
(0.0) / 0.81 (1.78) / 10.0 (14.96) / 3.13 (8.95) / 0.0
(0.0)
Block2
Contingent / 1.97 (2.65) / 51.88 (27.59) / 1.34 (4.13) / 1.88 (4.59) / 2.95 (9.06) / 51.88 (32.26) / 2.5 (8.7) / 1.88 (8.39)
Random / 0.45 (0.79) / 12.5 (15.17) / 1.88 (4.58) / 0.63 (2.8) / 0.71 (1.35) / 15.0 (13.81) / 0.63 (2.8) / 1.25 (3.85)
Reward phase - RT data
The same ANOVA[1] with mean median RTs revealed only a marginal significant main effect of Block, F(1, 70) = 3.4, p = .069, η2p = .046, which was further qualified by a significant Affect x Block x Cue-Probe-condition, F(3, 210) = 3.34, p < .05, η2p = .046. No other interaction with Block proved reliable (F < 1.64, p > .182). Planned comparisons on the additional interaction of Affect x Block x Cue-Probe-condition revealed faster RTs specifically in B Cue trials from Block 1 to Block 2 in the neutral affect groups only (BX: F = 5.77, p < .05, BY: F = 3.86, p = .053; positive affect: both F < 1, p > .426) and a specific AY benefit under positive affect only in Block 2 (F = 4.76, p < .05, 446 ms vs. 496 ms; Block 1: F = 3.14, p = .081, 455 ms vs. 487 ms). Taken together, this suggests that with more time on task in the reward phase, the influence of positive affect might increase again (see Table S4 for full table of means).
Table S4. Mean median RTs in ms (SD in parentheses) in Reward phase Block 1 and Block 2 as a factor of Affect group, Contingency, Cue-Probe-condition, and Incentive.
neutral / positiveContingency / AX / AY / BX / BY / AX / AY / BX / BY
non-incentive Trials
Block 1
Contingent / 291.62 (30.46) / 443.85 (51.47) / 191.18 (70.32) / 194.06
(66.34) / 299.62 (37.18) / 455.47 (54.3) / 180.26 (72.15) / 183.53 (66.38)
Random / 359.2 (61.67) / 543.5 (127.26) / 261.33 (75.45) / 252.93 (76.38) / 332.0 (34.73) / 483.53 (58.4) / 201.78 (94.84) / 223.38
(100.19)
Block2
Contingent / 269.24 (27.17) / 467.53 (73.46) / 165.35 (56.54) / 167.35 (59.28) / 277.35 (49.86) / 447.53 (70.12) / 170.53 (51.16) / 165.59 (52.87)
Random / 346.6 (57.34) / 542.95 (144.81) / 199.3 (68.33) / 224.98 (80.74) / 327.2 (30.76) / 465.48 (52.76) / 215.8 (167.51) / 191.73
(88.7)
incentive Trials
Block 1
Contingent / 272.65 (30.88) / 433.56 (69.23) / 139.68 (65.54) / 158.71
(75.81) / 271.35 (53.28) / 408.41 (53.98) / 145.59 (72.43) / 153.79
(64.92)
Random / 353.5 (55.2) / 517.6 (105.51) / 242.45 (118.19) / 243.48
(73.4) / 330.8 (34.09) / 470.68 (55.94) / 198.18 (76.46) / 196.9
(106.23)
Block2
Contingent / 258.5 (42.22) / 422.65 (56.21) / 137.74 (73.27) / 146.59 (86.47) / 268.85 (73.59) / 413.5 (152.03) / 152.42 (153.97) / 166.41 (217.82)
Random / 345.13 (53.38) / 550.28 (155.2) / 198.63 (79.0) / 209.45 (64.49) / 326.08 (30.33) / 456.78 (42.78) / 204.08 (113.17) / 192.78 (87.37)
Control for possible group differences in Baseline
In the second phase of the experiment half of the participants of a given affect group were randomly assigned to a performance contingent or non-contingent reward condition. To check for possible preexisting (random) differences between contingency groups, we reran Baseline analyses including the (in this experimental phase) irrelevant additional factor Contingency (contingent, random). Neither error rates nor RT analysis revealed significant main effects between Contingency groups; also, no interactions including Contingency were significant (see Tables S5 and S6 for ANOVA results). We can thus rule out that any differences between Contingency groups found in the reward phase can be ascribed to preexisting differences between groups.
Table S5. ANOVA results for error rates in Baseline phase
Source / df / F / p / η2pMain effects
between
Affect (A) / 1, 76 / 7.74 / < .01 / .092
Contingency (B) / 1, 76 / < 1 / .531 / .005
within
Cue-Probe-Sequence (C) / 3, 288 / 42.43 / < .001 / .358
Interactions
A x B / 1, 76 / < 1 / .576 / .004
A x C / 3, 228 / 5.4 / < .01 / .066
B x C / 3, 288 / < 1 / .875 / .003
A x B x C / 3, 288 / < 1 / .618 / .008
Table S6. ANOVA results for RTs in Baseline phase
Source / df / F / p / η2pMain effects
between
Affect (A) / 1, 76 / 1.95 / .167 / .025
Contingency (B) / 1, 76 / < 1 / .73 / .002
within
Cue-Probe-Sequence (C) / 3, 288 / 252.55 / < .001 / .769
Interactions
A x B / 1, 76 / < 1 / .377 / .010
A x C / 3, 228 / 2.86 / < .05 / .036
B x C / 3, 288 / 2.28 / .08 / .029
A x B x C / 3, 288 / < 1 / .786 / .005
Additional results to reward responsiveness, total reward sum and BAS scores
Extending the results reported in the main text, we show here complete correlation tables (Tables S7-S10) of total reward sum and all BIS/BAS scales separately for each experimental group. Because both random reward groups received a fixed reward sum, only correlations between BIS/BAS ratings are included in the respective tables.
Table S7. Correlations of total reward sum, BIS scale scores, BAS scale scores, and BAS subscale scores in the NeutralReward group.
1 / 2 / 3 / 4 / 5 / 61 Total reward sum / -
2 BIS / -.16 / -
3 BAS / -.03 / -.18 / -
4 BAS Drive / -.04 / -.22 / .68** / -
5 BAS Fun seeking / -.12 / -.1 / .66** / .05 / -
6 BAS Reward responsiveness / .09 / -.05 / .77** / .33 / .36 / -
* p < .05, ** p < .01
Table S8. Correlations of total reward sum, BIS scale scores, BAS scale scores, and BAS subscale scores in the PositiveReward group.
1 / 2 / 3 / 4 / 5 / 61 Total reward sum / -
2 BIS / .29 / -
3 BAS / .23 / .09 / -
4 BAS Drive / .06 / .23 / .65** / -
5 BAS Fun seeking / -.03 / -.20 / .69** / .06 / -
6 BAS Reward responsiveness / .49* / .27 / .69** / .46 / .09 / -
* p < .05, ** p < .01
Table S9. Correlations of BIS scale scores, BAS scale scores, and BAS subscale scores in the NeutralRandomReward group.
1 / 2 / 3 / 4 / 51 BIS / -
2 BAS / -.09 / -
3 BAS Drive / -.05 / .65** / -
4 BAS Fun seeking / -.38 / .77** / .19 / -
5 BAS Reward responsiveness / .26 / .82** / .40 / .43 / -
* p < .05, ** p < .01
Table S10. Correlations of BIS scale scores, BAS scale scores, and BAS subscale scores in the PositiveRandomReward group.
1 / 2 / 3 / 4 / 51 BIS / -
2 BAS / -.25 / -
3 BAS Drive / -.19 / .70** / -
4 BAS Fun seeking / -.19 / .73** / .19 / -
5 BAS Reward responsiveness / -.24 / .83** / .36 / .53 / -
* p < .05, ** p < .01
As can be expected, no significant correlations were found between BIS scale and BAS scale (including BAS subscales), and BAS subscales did not significantly correlate with each other but only positively correlated with total BAS scale (cf., Carver and White, 1994). The only significant correlation between any BIS/BAS scale and total reward sum was the positive correlation of total reward sum and BAS reward responsiveness in the PositiveReward group (see also Results in main text).
Effect sizes of affect effects in AY and BX trials in the present study and the AX-CPT study by Dreisbach (2006, Experiment 2)
As reported in the main text, both the present study and a previous AX-CPT study by Dreisbach (2006) indicate reduced proactive control and/or increased reactive control in association with positive affect as compared to neutral affect. In the present study this modulation of cognitive control by positive affect was primarily indicated by reduced AY interference, while the Dreisbach study demonstrated medium to large effects of increased BX interference. Table S11 gives an overview of these results.
Table S11. Effect sizes (Cohen’s d) and absolute difference scores (RT/errors neutral – positive) of affect effects in AY and BX trials in the present study and the AX-CPT study by Dreisbach (2006, Experiment 2).
present study (N=80) / Dreisbach, 2006a (N=36)Cue-Probe-sequence / Cohen’s d / absolute difference / Cohen’s d / absolute difference
RTs
AY / 0.54 / 55 ms / 0.21 / 21 ms
BX / 0.25 / 22 ms / 0.52 / -91 ms
Errors
AY / 0.57 / 7.26 % / 0.27 / 3.06 %
BX / 0.4 / 1.73 % / 0.95 / -10.28 %
Note. a Effect sizes are given for Experiment 2 only, because therein procedure included distractor letters between Cue and Probes like in the present study.
[1] Including the additional factor Block reduced the sample size in Reward phase RT analysis to N = 76, because three subjects in each performance contingent reward group had 100 % error rates specifically in AY trials in either Block 1 or 2, thereby causing empty cells in RT data.