Proc. R. Soc. B 283, 20152888. (Published Online 27 January 2016). (doi:10.1098/rspb.2015.2888)
The authors of the paper by Pruitt J.N., Wright C.M., Keiser C.N., DeMarco A., Grobis M.M., & Pinter-Wollman N. `The Achilles' heel hypothesis: misinformed keystone individuals impair collective learning and reduce group success' published in 2016 in Proceedings of the Royal Society B 283: 20152888 were recently made aware of problems in the raw data collected by Pruitt J.N. In particular, 86% of the values of the rewarding stimulus in time 9 are separated by an integer (11) from the values in time 11 of the rewarding stimulus. Similarly, 82% of the values in time 10 of the rewarding stimulus are separated by an integer (3) from values in time 12 of the rewarding stimulus. In addition, 21% of the values in time 2 of the rewarding treatment are separated by a fixed integer (8) from values in time 1 of the dangerous treatment.
Furthermore, in the data on individual learning, 80% of the values from time step 5 in the rewarding trial are separated by exactly 10 from the third time point in the unrewarding trials. Similarly, 80% of the values from time step 6 in the rewarding trial are separated by exactly 11 from the second time point in the unrewarding trials. These irregularities are focused on the keystone and generic individuals and no such irregularities are found in the control individuals.
Reanalysis, with these irregularities removed, no longer provides clear support for some of the main conclusions of the study. The authors therefore wish to retract the paper.