Exploratory reports: A new article type for Cortex

There aremanyways to find things out. In science, the process of discovery can be divided conceptually into exploratory and confirmatory phases. In the exploratory phase, we observe and explore, generating theories to explain the patterns that we find. Useful theories will support predictions about what we should and should not find in the future if that theory is true. We are then in a position to collect new data, by targeted observation or controlled experiment, confirming or challenging our predictions through critical tests, and revising our theories in light of the outcomes. We may also explore these new data for unanticipated patterns, updating our ideas in an iterative cycle of (inductive) exploration and (deductive) confirmation. Exploratory and confirmatory approaches are complementary, but when we seek to share our work through publication, they are rarely weighed as equal. At most empirical journals (not just those with ‘Experimental’ in their name), the confirmatory approach has special status as the proper mode of science. This preference is pervasive: funding agencies and ethics committees demand that we show clear hypotheses, as an assurance that something might come from the money and the effort; and we learn and we teach that an empirical report has an ordered Introduction, building up to a statement of the hypotheses that the study is designed to test. Exploratory work, which does not fit this template, may be dismissed as ‘insufficiently hypothesis-driven’ e the death sentence for many a hopeful submission. This idealisation of the confirmatory mode creates pressure for published research to conform neatly to the template, even though the reality may be more messy or complex. A preference for positive findings, combined with the expectation that the main results should be predicted a priori, incentivise some ‘questionable’ practices that, whether engaged in consciously or not, seriously distort the scientific record (John, Loewenstein, & Prelec, 2012). High on this list are: p-hacking, whereby analytic flexibility is exploited to probe the data for p-values below the threshold for significance, the fruits of this exploration being reported as if from a planned analysis (Simmons, Nelson, & Simonsohn, 2011; Ware & Munaf o, 2015); and HARKing (Hypothesising After Results are Known; Kerr, 1998), whereby hypotheses suggested by the data are presented as if formulated in advance. The pressure to conform is not only implicit. Some of us may have had well-intentioned advice from editors or reviewers, that we should rethink our introductory rationale to make our paper stronger; and some of us must, at other times, have been that editor or reviewer. Questionable research practices, their causes and consequences, are now widely recognised, and many competent discussions exist (e.g., Ioannidis, 2005). Constructive solutions have also been proposed, and efforts to reform the conduct and publishing of science are gaining traction, especially in Psychology and Neuroscience where the problems have been most acutely apparent (Chambers, 2017; Nosek et al., 2015; Open Science Collaboration, 2012). Cortex was at the forefront of this movement, as one of the first journals to offer Registered Reports, an empirical format that focuses reviewer scrutiny on the importance of the scientific question, and the rigour and power of the methods, before the first data point is collected (Chambers, 2013; Chambers, Dienes, McIntosh, Rotshtein, & Willmes, 2015). Publication is