Research Data Services

When should data be cited?

All data that were used to carry out your research should be cited to ensure reproducibility of your work. This includes both data that you generated and any data that were previously published (by you or someone else) and re-used in a given study.

How should data citations and references be formatted?

Data should be cited in the same fashion as a journal article; an abbreviated in-text citation should be included wherever appropriate in the main text, and a full reference should be included in the reference. This includes both data that you generated and any data that were previously published (by you or someone else).

How should data citations and references be formatted?

Many of the major reference formats that are used by publishers have specific reference/citation formats for data/datasets, in addition to the better-known styles for journal articles, books, preprints, etc. (e.g., APA). A journal may also have its own specification for how to cite a dataset. In general, dataset references should contain:

• Author(s)
• Publication year
• Title
• Repository
• DOI

Additional information like the dataset version (if relevant) and a bracketed description (e.g., "[Data set]") may also be recommended or required. The in-text citation should be formatted like articles or books (e.g., Smith et al., 2025).

Oftentimes, when researchers are reusing data or software that were previously published, they cite the associated paper that first described the data or software. However, this is a poor scholarly practice for several reasons:

• The paper does not contain the software or the data, it merely uses/analyzes them. As data and software become more common and increase in size and complexity, they are increasingly being separated from the main text of articles, whether in Supplemental Information or in third-party repositories. Therefore, if you need to cite a dataset or a piece of software, it is inappropriate to cite just the paper; it would be like only citing one paper of a certain research group when you actually used four related papers from that group for your argument.
• Legal attribution requires attribution of that object, not a related object. Researchers often apply Creative Commons licenses like CC BY (Attribution) to data and software. This is typically inappropriate for different reasons (e.g., Creative Commons says not to use CC licenses for software), but chief among them is that legal attribution is not the same as scholarly citation. Where content can actually be copyrighted (e.g., most software, some proprietary data), any content that is reused but that only cites the associated paper (or that doesn't cite anything at all) has committed a copyright violation in failing to properly attribute the data or software specifically. If you wanted to cite an idea put forward by one paper, it would not be acceptable to cite a different paper by the same author(s) that doesn't make that actual point; the same is true for data/software and an associated paper.
• Citing the associated paper prevents tracking of reuse of the data/software specifically. Many researchers are interested in knowing how many people may be reusing their data and/or software. Basic web traffic attributes like views and downloads are tracked for publicly accessible deposits, but these don't necessarily indicate reuse - someone might just have a glance at your files and then decide not to use them. The clearest way to indicate reuse of data or software, as when we refer to content directly in articles, is a full in-text citation and reference.