When arXiv announced it would ban people for hallucinated citations, that is citations of papers that don’t exist, the discussion online got sidetracked by the question of whether academics actually read the papers they cite. Some people proudly insisted that any good scholar always reads every paper they reference, others argued that was ridiculous.

As always, the answer is never that simple. In certain fields, it is enormously important to read the papers you cite if you want to do solid, careful, scholarly work. In others, it’s entirely irrelevant.

It mostly comes down to what citations are for. And luckily, I’ve already written a post about that.

So let’s go through the citation motivations I mention in that post.

First, some citations are about respecting priority, feeding the system by which academics get credit for having an idea first. The incentive system of academia depends on getting this more or less right, but that doesn’t mean every academic has to check things at every step of the way. Besides, if you get this wrong, you’ll find out quickly. Submit a paper to a preprint server like arXiv, and you’ll be sure to get emails telling you that some obscure Soviet researcher figured it all out first.

Other citations are about substantiating claims. These are the most important to get right. Here, you really ought to have read, if not the whole paper, at least the full justification for the claim you’re making. You can have some leeway if the methods are unfamiliar enough, for example a complicated experiment you can’t understand all the details of. Science and technology do require some trust. But you should have at least a sense of where things could go wrong, and why.

Citations to provide context are a different beast. Here, you’re trying to tell a reader where your ideas come from. You can’t show them the conversations you have with your colleagues, the things they value and get you excited about. So you have to show them papers instead. But the papers aren’t the thing you read, they’re just a convenient proxy.

Finally, citations do sometimes just exist to follow social conventions. And yeah, you don’t have to read these, just like you don’t have to say how you’re doing when someone asks you how you’re doing. They’re the academic equivalent of social white lies, and should be taken roughly as seriously, both by their supporters and detractors.