Are subliminal effects permanent

New publication: Evidence against subliminal influence from anchor effects

For a decade now, as in many empirical sciences, it has also become clear in psychology that many findings cannot be replicated. In order to raise awareness of this replication crisis among the students as well, various causes and solutions were discussed with the students during the seminar "The replication crisis in psychology" in the winter semester 2019/2020, and a replication was carried out according to current scientific standards. This study has now been published in the prestigious Journal of Experimental Social Psychology.

Anchor effects are a well-established phenomenon: we are influenced in our estimates by values ​​that are put into the room in this context. But does something like this also work below the threshold of consciousness - that is, subliminally? On the question of subliminal anchor effects, there are studies by Mussweiler and Englich (2005) and by van-Rooijen and Daamen (2006) in which it was possible to show that numerical estimates could be influenced by a subliminal number. While Mussweiler and Englich (2005) argued that the effect only occurs when the test subjects took the time to estimate, van-Rooijen and Daamen emphasized that the effect only occurs under time pressure. In the spirit of the Open Science movement, we preregistered and performed two direct replications. We were able to show that neither of the two studies can be replicated. In this respect, the original findings are probably incidental findings. There is no evidence that numerical estimates are influenced by subliminal numbers.

All materials from the studies, the data, analysis scripts, pre-registrations and the documents for the seminar are permanently available online here:
The article can be downloaded here free of charge until December 5, 2020.

Reference: Röseler, L., Schütz, A., Blank, P. A., Dück, M., Fels, S., Kupfer, J., Scheelje, L., & Seida, C. (2021). Evidence against subliminal anchoring: Two close, highly powered, preregistered, and failed replication attempts. Journal of Experimental Social Psychology, 92, 104066.