Supplementary Materials1: Mixed versus blocked control

In the experimental conditions of Experiment 1 and 2 supraliminal and subliminal trials were intermixed, whereas only flash-suppressed cues were presented in the control tasks. Possibly, the intermixed visible arrow cues facilitated breakthrough of the suppressed arrow cues in Experiment 1 and 2 (target detection task), as a result of which the control task (cue detection task) would have underestimated the visibility of the suppressed arrow cues. To test for this possibility we ran an additional control experiment (N = 8) in which participants were instructed to report the direction of the arrow as soon as they could discern it. One experimental session was comprised of 120 flash-suppressed trials (blocked condition), and was identical to the control condition in Experiment 1 and 2. In the other experimental session visible trials were intermixed (intermixed condition) in such a way that the stimulus presentation was identical to the target-detection task of Experiment 1.

Overall, 85% of the subliminal arrow cues were detected within 6 seconds, after which a new trial was initiated if no response was given. These trials yielded an average detection time of 1.7 seconds (SD = 1.1). Importantly, there was no difference between the blocked and mixed presentation on detection times of arrow directions;neither mean detection time of arrow cues that were perceived within 6 seconds (all SOA’s: p > .5), nor the number of masked arrows that were not seen within 6 seconds (p= .08) differed between presentation conditions. However, there was a trend that more (masked) arrows were perceived in the blocked design, as compared with the mixed design. This suggests that, by using a blocked design in the cue detection task (the control task) we might have overestimated the number of unsuccessfully masked arrow cues in the (mixed) target detection task of Experiment 1 and 2. Hence, our initial approach appears somewhat conservative in detecting non-conscious cueing effects.

Finally, this supplementary control experiment was used to explicitly test our assumption that supraliminally presented arrow cues were detected faster than subliminally presented cues. This assumption was confirmed in a repeated measures ANOVA with the factors SOA and Cue Visibility by a main effect of Cue Visibility, F(1, 7) = 141.891, p < .001, η2 = .953. Specifically, subsequent paired-samples t-tests revealed that in the SOA 100 condition, subliminal cues were detected 1191 ms (SD = 264) later than supraliminal cues, t(7) = 12.772, p < .001. Similarly, in the SOA 500 condition, subliminal cues were detected 1241 ms (SD = 370) later than supraliminal cues, t(7) = 9.471, p < .001.

Supplementary Materials2: Analyses of subliminal trials in Experiment 2

In the result section of Experiment 2 only the supraliminal conditions were analyzed. Firstly, because the main objective was to test whether subliminally presented statistical context would affect the utilization of intermixed supraliminal cues. Secondly, because we mainly provide analyses of cueing effects (congruent RTs subtracted from incongruent RTs) and there were no incongruent trials in the subliminal predictive condition. While investigating cueing effects on subliminal trials was not the main purpose of Experiment 2, our general conclusions lead us to assume that the predictive value of the cues would have no effect on RTs in subliminal trials. To test this assumption we compared the RTs of the three subliminal conditions of Experiment 2 as a supplementary analysis: predictive congruent, non-predictive congruent and non-predictive incongruent. Since an SOA of 100 ms led to longer RTs (M = 379 ms, SD = 42.5) than an SOA of 500 ms (M = 350 ms, SD = 40.8), F(1, 12) = 39.902, p < .001, η2 = .77, we analyzed both SOA conditions separately.

Within the SOA 100 condition none of the three pairwise comparisons were significant: Congruent RTs in the predictive condition did not differ from congruent RTs in the non-predictive condition (M = 2.44 ms, SD = 27.9), t(12) = .314, p = .759, d = .08; Congruent RTs in the non-predictive condition did not differ from incongruent RTs in the non-predictive condition (M = .77 ms, SD = 8.71), t(12) = .320, p = .755, d = .09; Congruent RTs in the predictive condition did not differ from incongruent RTs in the non-predictive condition (M = 3.21 ms, SD = 25.5), t(12) = .454, p = .658, d = .13.

Next, we compared RTs in the three subliminal conditions (predictive congruent, non-predictive congruent and non-predictive incongruent) for trials with an SOA of 500 ms. Again, none of the pairwise comparisons yielded significant results: Congruent RTs in the predictive condition did not differ from congruent RTs in the non-predictive condition (M = 1.34 ms, SD = 19.2), t(12) = .250, p = .806, d = .07; Congruent RTs in the non-predictive condition did not differ from incongruent RTs in the non-predictive condition (M = -.79 ms, SD = 6.14), t(12) = .464, p = .651, d = .13; Congruent RTs in the predictive condition did not differ from incongruent RTs in the non-predictive condition (M = .55 ms, SD = 18.3), t(12) = .108, p = .916, d = .03.

In sum, there was no effect of predictive value on subliminal trials, nor was there an effect of cue-target congruency. This is in line with our general conclusion that visible statistical relevance is required for subliminal cue utilization to occur.

Supplementary Table 1. Mean RTs in ms (SD) and accuracy in percentage of errors (SD) for all conditions in Experiment 1.

100 ms SOA
Supraliminal / Subliminal
Non-predictive / Predictive / Non-predictive / Predictive
Congr. / Incon. / Congr. / Incon. / Congr. / Incon. / Congr. / Incon.
Errors / .90 (1.20) / .83 (1.02) / .28 (.34) / 3.85(4.41) / .77 (1.58) / .96 (2.40) / .96 (1.63) / .96 (1.63)
R.T. / 390 (62) / 400 (62) / 390 (48) / 422 (49) / 396 (61) / 400 (62) / 406 (45) / 410 (44)
500 ms SOA
Errors / .96 (.89) / 1.67 (2.41) / .48 (.99) / 3.53 (4.02) / .77 (1.58) / 1.35 (1.65) / .77 (1.58) / .58 (1.10)
R.T. / 352 (50) / 384 (49) / 337 (38) / 420 (42) / 371 (46) / 378 (45) / 378 (53) / 394 (54)

Supplementary Table 2.Mean RTs in ms (SD) and accuracy in percentage of errors (SD) for all conditions in Experiment 2.

100 ms SOA
Supraliminal / Subliminal
Non-predictive / Predictive / Non-predictive / Predictive
Congr. / Incon. / Congr. / Incon. / Congr. / Incon. / Congr. / Incon.
Errors / .77 (2.14) / .77 (1.58) / 1.15 (2.42) / 1.35 (1.94) / 1.09 (1.54) / 1.09 (1.84) / 0.58 (0.79) / n.a.
R.T. / 370 (48) / 385 (47) / 377 (44) / 391 (43) / 379 (45) / 378 (45) / 381 (43) / n.a.
500 ms SOA
Errors / .96 (2.17) / 2.88 (4.98) / .96 (2.17) / 4.23 (6.57) / 1.47 (2.41) / 1.79 (2.38) / 1.22 (1.39) / n.a.
R.T. / 334 (48) / 364 (41) / 336 (44) / 363 (38) / 349 (40) / 350 (41) / 351 (44) / n.a.