## Introduction

## Method

### Participants

### Global–local task

### Procedure

### Statistical analysis

## Results

### Global–local task

_{ p }

^{2}= 0.70, reflecting the well-known global precedence effect (Navon, 1977), that is, faster responses to globally than locally defined targets (421 vs. 467 ms). Second, the effect of congruency, F(1,34) = 31.70, p < .0001, MSE = 1502.73, η

_{ p }

^{2}= 0.48, reflecting interference of the irrelevant target level, as indicated by a faster RT on congruent as compared to incongruent trials (431 vs. 457 ms). Third, the effect of switching, F(1,34) = 27.59, p < .0001, MSE = 7429.21, η

_{ p }

^{2}= 0.45, which revealed that repeating the task allowed for faster responding than switching between target levels (418 vs. 471 ms). Fourth, the interaction of task switch and congruency, F(1,34) = 4.22, p < .05, MSE = 1459.05, η

_{ p }

^{2}= 0.11, indicated that the congruency effect was larger when target levels were switching (35 ms) than when they were repeated (17 ms). Most importantly for our purposes, target level interacted with group, F(1,34) = 4.21, p < .05, MSE = 1879.08, η

_{ p }

^{2}= 0.11 (see Fig. 2): the size of the global precedence effect was significantly smaller in the gamma group (36 ms) than in the control group (57 ms). Last, both switch and congruency did not interact significantly with group, F(1,34) = 1.61, p > .05, MSE = 7429.21, η

_{ p }

^{2}= 0.04 and F < 1, respectively. Given that conventional null-hypothesis significance testing (NHST) cannot be used to provide evidence in favor of the null hypothesis (H

_{0}), we calculated the Bayesian (posterior) probability associated with the occurrence of the null hypothesis [p(H

_{0}|D)] to validate the absence of any interaction between the factors group and congruency (crucial for our second hypothesis). To this end we used the method proposed by Wagenmakers (2007) and Masson (2011). This method uses Bayesian information criteria (BIC), calculated using a simple transformation of sum-of-squares values generated by the standard ANOVA, to estimate the Bayes factor and generate p(H

_{0}|D), assuming a “unit information prior” (for further details, see Kass & Wasserman, 1995; see also Jarosz & Wiley, 2014).

Group | Control | Gamma | ||
---|---|---|---|---|

Mean RT | Mean error | Mean RT | Mean error | |

Switch | 495 (22.2) | 8.2 (1.3) | 447 (22.2) | 5.6 (1.3) |

Repetition | 428 (11.2) | 6.2 (0.9) | 407 (11.2) | 6.1 (0.9) |

Switch cost | 67 ms | 40 ms | ||

Local target | 490 (16.1) | 8.8 (1.5) | 445 (16.1) | 6.3 (1.5) |

Global target | 433 (16.8) | 5.6 (0.9) | 409 (16.8) | 5.5 (0.9) |

Global precedence effect | 57 ms | 36 ms | ||

Incongruent | 476 (18.3) | 11.4 (1.4) | 439 (18.3) | 8.9 (1.4) |

Congruent | 448 (14.1) | 3.0 (0.8) | 415 (14.1) | 2.8 (0.8) |

Congruency effect | 28 ms | 24 ms |

_{0}|D) provides the exact probability of the occurrence of H

_{0}. The analysis revealed that the p(H

_{0}|D) was 0.84, hence, on the basis of the guidelines proposed by Raftery (1995), represents positive evidence in favor of H

_{0}.

_{ p }

^{2}= 0.67, reflecting interference from the irrelevant target level, as indicated by a smaller proportion of errors in congruent than incongruent trials (2.9 vs. 10.2 %).

### Mood and arousal

_{ p }

^{2}= 0.11, but not mood, F ≤ 1. LSD Fisher post hoc analyses revealed that, for the control group, arousal levels were comparable across the two measurements (arousal levels were 0.5 and 0.3, for the first and second measurement, respectively, p = .61). In contrast, for the gamma group, a significant difference between the two time points indicated an increase from the first to the second measurement (0.3 vs. 1.1, p = .02), suggesting that our manipulation worked as expected.