Erratum: Predator and prey activity levels jointly influence the outcome of long-term foraging bouts (Behavioral Ecology (2013) 24 (1205-1210) DOI: 10.1093/beheco/art052

Abstract

The authors of the paper "Predator and prey activity levels jointly influence the outcome of long-term foraging bouts"published by Sweeney et al. in Behavioral Ecology in 2013 (doi:10.1093/beheco/ art052) recently investigated some irregularities in the raw data. This experiment was conducted primarily by undergraduates in the Pruitt Lab in 2012. Sweeney, Cusack, Armagost, and O'Brien conducted single behavioral assays on crickets and multiple behavioral assays on spiders. These data were used to determine the repeatability of spider activity level (Spider ID numbers PLE-04 to PLE-83). The data were organized into excel sheets and sent to Pruitt to curate. A second pool of spiders was collected by Pruitt and Cusack for staged predator-prey interaction trials (Spider ID numbers PLE-143 to PLE-158), but these new spiders were assayed only a single time each. Keiser helped perform statistical analyses and wrote portions of the manuscript. The original methods section did not state that new spiders were collected for the mesocosm portion of the experiment. After investigating the raw data, we found that: 1) body condition indices were not calculated correctly and 2) body size measurements for the new set of spiders were duplicated from the original set of spiders. Thus, the body sizes of the spiders used the predator-prey interaction trials are unknown. We reanalyzed the data using the same ordinal logistic regression as before, but after removing body condition as an independent variable. The results of the revised analysis remain the same: cricket mortality was influenced jointly by the activity levels of interacting predators and prey (χ2 = 21.77, P < 0.001; Table 1). Although we are hesitant to interpret data that lacks predator body size information, we also used the original data collected by undergraduate coauthors to verify that spider activity level does not correlate with body size (linear regression: F1,41 = 0.21, P = 0.65, R2 = 0.005). The raw data originally provided to Keiser for analysis are now freely available on Figshare (https://doi.org/10.6084/ m9.figshare.11778540.v1). Provided therein are two datasheets: one contains the data as originally provided by Pruitt to Keiser for analyses; the other contains the same data but with errors highlighted and described.