Who should be the author(s) of an academic paper? This question is becoming increasingly pressing, due to the increasing prevalence and scale of scientific collaboration, and the corresponding diversity of authorship practices in different disciplines and subdisciplines. This paper addresses the conceptual issues underlying authorship, with an eye to ameliorating authorship practices. The first part of the paper distinguishes five roles played by authorship attributions: allocating credit, constructing a speaker, enabling credibility judgements, supporting accountability, and creating an intellectual marketplace. The second part of the paper argues that distinguishing these functions helps us see that at least some of the confusions around authorship are due to tensions between these functions. The final part of the paper suggests a way to resolve these conceptual confusions, which we will call the CSWG proposal. This proposal suggests replacing authorship with a bundle of roles tailored to the functions of authorship-contributor, spokesperson, writer, and guarantor-which can be distributed in a number of different ways. as a collective author. 2 In many disciplines (especially in interdisciplinary collaborations) it will be unclear or indeterminate what the norms for ascribing authorship are.Researchers have identified a number of problems caused by the combination of scale of collaborations and diversity of authorship practice: i) confusion around disciplinary norms (Street et al. [2010], Mitcheson [2011], Macfarlane [2017]); ii) deliberate flouting of disciplinary norms (in a 2005 survey of NIH grant-holders 10% of respondents admitted assigning authorship inappropriately (Martinson et al. [2005], see also Pignatelli et al. [2005], Rohwer et al. [2017])), iii) the emergence of ghost authors (writers who are not listed as authors, often to hide commercial interests) who haunt as much as a fifth of papers in medical journals (Wislar et al. [2011] see also Flanagin et al. [1998], Mowatt et al. [2002]); iv) a lack of consensus about how to resolve disagreements (Macfarlane [2017]); v) disagreements about authorship (Mitcheson et al. [2011]); and vi) problems in reading a byline (Shaw [2016]). Given these problems, it is no surprise that authors often make subversive and unruly authorship attributions (Penders and Shaw [2020]).