Carr’s “delinkification” critique is part of a larger argument contained in his book The Shallows. I read the book this summer and plan to write about it more. But for now let’s zero in on Carr’s case against links, on pages 126-129 of his book as well as in his “delinkification” post. … The nub of Carr’s argument is that every link in a text imposes “a little cognitive load” that makes reading less efficient. Each link forces us to ask, “Should I click?” As a result, Carr wrote in the “delinkification” post, “People who read hypertext comprehend and learn less, studies show, than those who read the same material in printed form.” … [The] original conception of hypertext fathered two lines of descent. One adopted hypertext as a practical tool for organizing and cross-associating information; the other embraced it as an experimental art form, which might transform the essentially linear nature of our reading into a branching game, puzzle or poem, in which the reader collaborates with the author. … The pragmatic linkers have thrived in the Web era; the literary linkers have so far largely failed to reach anyone outside the academy. The Web has given us a hypertext world in which links providing useful pointers outnumber links with artistic intent a million to one. If we are going to study the impact of hypertext on our brains and our culture, surely we should look at the reality of the Web, not the dream of the hypertext artists and theorists.
The other big problem with Carr’s case against links lies in that ever-suspect phrase, “studies show.” Any time you hear those words your brain-alarm should sound … Carr’s critique of links employs a bait-and-switch dodge: He sets out to persuade us that Web links — practical, informational links — are brain-sucking attention scourges robbing us of the clarity of print … The researchers Carr cites divided a group of readers into two groups. Both were provided with the text of Bowen’s story split into paragraph-sized chunks on a computer screen. (There’s no paper, no print, anywhere.) For the first group, each chunk concluded with a single link reading “next” that took them to the next paragraph. For the other group, the researchers took each of Bowen’s paragraphs and embedded three different links in each section — which seemed to branch in some meaningful way but actually all led the reader on to the same next paragraph. (The researchers didn’t provide readers with a “back” button, so they had no opportunity to explore the hypertext space — or discover that their links all pointed to the same destination.) … . They didn’t turn the story into a genuine literary hypertext fiction, a maze of story chunks that demands you assemble your own meaning. Nor did they transform it into something resembling a piece of contemporary Web writing, with an occasional link thrown in to provide context or offer depth. No, what the researchers did was to muck up a perfectly good story with meaningless links. Of course the readers of this version had a rougher time than the control group, who got to read a much more sensibly organized version. All this study proved was something we already knew: that badly executed hypertext can indeed ruin the process of reading. So, of course, can badly executed narrative structure, or grammar, or punctuation. Carr also makes reference to a meta-analysis … none of the studies the meta-analysis compiles looked at Web-style links. They all drew comparisons between linear hypertexts (screens with “next” links, not printed articles) on one side, and on the other, literary-style hypertexts broken up into multiple nodes where “participants had many choices in sequencing their reading.”
It’s certainly possible that hypertext (even of the moderate version associated say, with blogs that set out to inform rather than to create SEO-clickfarms) imposes a high enough cognitive burden to outweigh e.g. the benefits of being able to look at the source material that authors are drawing on. It’s also possible that Carr’s proposed solution – of having a selected group of links at the end of the text, along the line of footnotes or endnotes, is a superior one. I’d love to hear from people who have seriously thought about these issues (Anthony Grafton – if you are reading this post, consider this to be an explicit invitation to weigh in). But if Rosenberg’s summation is on target, these studies (and in particular the study that he singles out) are simply not informative in the ways that Carr suggests that they are informative (it would of course be nice to see studies which were informative on this question). Rosenberg notes that it is rather peculiar that he seems to have been the first person to actually look up the studies that Carr draws upon. He doesn’t explicitly note the irony that if Carr had actually linked to the studies, people would have been more likely to have clicked through and read them. But I imagine the thought has occurred to him.
This can be generalized into a broader point on the role that research should play in public debate (see also Fernando Pereira on this ). I’m not inclined to be as harshly critical of Carr as Pereira is – he is not unique. My understanding is that people who want to write non-fiction bestsellers are, shall we say, strongly encouraged to make strong forthright arguments without cavils and hesitations if they want to see their work published and promoted. This does not provide incentives for trustworthy engagement with the existing body of research, which (on most interesting questions in the social sciences and communication studies) is shot through with hedges, doubts, disagreements and qualifications. But this also means, I think, that work which draws upon academic sources should be hyperlinked – and hyperlinked quite extensively too. This in principle would allow people to read the relevant studies for themselves (if they have enough training to make sense of it), and make it more likely that they will see the alternative interpretations of those who do have such training, if they themselves do not. Of course, this also points to the broader need for ungated academic research – if academic work is to have an impact in general public argument, it needs to be broadly accessible to people engaged in these arguments.
Update: See also Mark Liberman .