View article

In a recent paper, we reported that cues to visual event segmentation (such as walking through a virtual doorway) can attenuate (or even eliminate) anchoring effects in decision-making (1). We welcome the attention to the strength and reliability of these results (2). In response, we focus first on the (mis) uses of certain statistical approaches in general and then on the reliability of our results in particular. By our count, this author has now published this same short commentary at least 15 times (eg, refs. 3–5), including in PNAS (3), in each case suspecting publication bias based on the omnibus post hoc power test of Ioannidis and Trikalinos (6), and drawing striking conclusions—eg, that 82% of a leading journal’s articles over many years (including widely replicated results) were compromised by publication bias (5). This approach has been widely criticized (eg, refs. 7–11), including by leading experts in statistics and …