The tradeoff of the hyperlink

Nicholas Carr’s recent post on delinkification explores whether we’d be better off if we didn’t use hyperlinks in-line.  He’s hitting on an old issue among the hypertext crowd, as various kinds of hypertext systems, from Apple’s Hypercard, to Hyper-G, have explored the pros and cons of the entire concept of hypertext.

In ancient times, I did some rudimentary studies on the effect of links on reading in 1995 for IE 1.0 and IE 2.0 – I recall many preceding academic studies on hypertext that went  further than we did (oddly enough, they’re hard to find on the web – still looking). Some of that data will be too crusty to apply to today’s web, but some of it is entirely relevant. I’ll follow up if I can find the good stuff.

A few points worth adding to the debate:

  1. The skill of the author is missing from the conversation. The better the writer, the better the job he/she does at anticipating questions or making sure the links are worth the cognitive cost of forcing the user to decide to click or stay. In similiar fashion we can criticize paragraphs, semi-colons, (parentheticals), fonts, bold/italics,blog templates and many factors that we know impact people’s ability to read, as when they are used poorly they do create problems for readers. All  choices writers make have cognitive tradeoffs. Readability, a simple filter that makes pages easier to read, is a surprisingly good alternative to many website and blog designs, but for the better writers on the web, it takes away more than it gives.
  2. This echoes the debates about footnotes and endnotes. I’ve done anecdotal research on this, and in reading this 30+ comment thread of people’s impressively specific preferences, I concluded there is no final answer. It’s too personal, and often people’s feedback hinges on the endnote/footnote style of the last book they read. It’s easy to forget there are many ways how footnotes are used, much like links, some better and some worse. Some uses, in some situations, earn their cognitive costs more than others.
  3. Good browsers should apply preferences for links, including what Carr describes (holding all links until the end). Markup languages are supposed to allow the browser to choose how to present various things, including links. If the reader wants to view all the links at the end, or on the side, or automatically go and pre-load pages, they should all be part of what a browser does to create a good reading experience. This does create conflicts of artistry (should my words appear as I want?) but the spirit of HTML/CSS or any markup language is to give control to readers as well as writers.
  4. Tabbed browsing changes the risks. For those users who use them, it gives an alternative. I know I and other tab users open links from an article in tabs as I go, and let them wait until I finish the article (or until I get stuck on a fact/reference I hope is addressed in a link).
  5. If minimalism for reading is ideal, web site design is a factor too. Even Carr’s site has a top navigation section, and a sidebar with various links and images of books to be clicked on (not that this negates his argument – less distractions are less distracting). Images are possibly more of a drag on cognitive load than a single hyperlink, and it wouldn’t be hard to do research to find out (I suspect in a reading comprehension comparison of Readability vs. most website designs, Readability wins).
  6. Perception of credibility. Forget the reality – in some cases links show the possibility the writer has done their homework. In a glance I can see the link density of a page – too much and I might pass, but none at all, and I might wonder if the writer has thought much about the topic, since they didn’t bother to show they’d found a reference to support or counter their own claims.

13 Responses to “The tradeoff of the hyperlink”

  1. Adam

    A single, easy browser change could completely change how we view this dispute.

    When I click “open link in new tab”, the new tab should know its browsing history.

    The primary problem we have is lost context. I read a page, open a link in a new tab, close the original page, and start reading the link. Now I want to compare that link back to the original page, but I can’t. The browser knows how I got to the link, but deliberately forgets and breaks my expectation of being able to use the back button.

  2. Scott Berkun


    It’s a good point, but not quite on target for Carr’s concerns (e.g. linking at all has more downsides than upsides).

    I think there are a fair number of browser design issues that could be useful on this. Some exist already as plugins or in Opera. I wouldn’t be surprised if a plugin exists to handle tabs as you suggest.

    The problem for browser design is there are perf issues with copying the entire history stack (which can be quite large) into each new tab. With 10 or 20 tabs, each with the entire history stack, it’s not a trivial performance issue anymore. There are ways around this (perhaps just keep the history url list, not the memory of the pages themselves), but I’ll leave it to people who actually make browsers in this decade to chime in on the challenges.

  3. Kathy Gill

    Thanks, Scott. The skill of the reader is missing from this equation as well.

    We know from research in the 1990s that there is a measurable cognitive load when we visit websites that have flashing things and jumbles of color and that cognitive load has nothing to do with hypertext. It has to do with page design.

    It’s also quite possible to make a printed page “hard to read” (even if there is a reasonable readability measurement for the text on the page). For example, think back to the early days of Wired Magazine, for goodness sakes! Hot pink (or lime green) background with reversed white type set in italics and 16 different font families! (Raise your hand if you remember these designs.) Wired was criticized for making the reading experience “hard.” See for an analysis of 1.1.

    I personally /abhor/ those “double-underline” hyperlinks that denote “this is an ad.” I also frown upon excessive self-referential links in an article (unless the hypertext is unambiguous, such as something like “when I wrote about this last year”). In fact, unambiguous hypertext is a teachable art. Rather than criticizing the tool (hyperlinks), I think Carr’s critique would be more productive if he pointed out to authors how hypertext could be more useful.

    I’d not thought about having smart browsers with settings that adjust where links appear. Although, I’m not firmly convinced that this is a good thing (it’s not quite like having every driver able to customize the controls on the dash or radio – but it’s close), I certainly would like the equivalent of a pop-up blocker for those obnoxious double-underlined ones!

    In my opinion, Carr’s expanded critique of hypertext that appears in in Wired magazine (a promo for his book due out this month) exhibits some of the flaws you outlined in last week’s “Calling BS On A Social Media Guru.”

    * “A 2007 scholarly review of hypertext experiments concluded” …. (no link, no end-of-passage citation)
    * “Another researcher, Erping Zhu, had people read” … (no link, no date, no end-of-passage citation)
    * “In a study published in the journal Media Psychology” … (no link, no date, no end-of-passage citation)

    How do we know that these studies say what Carr says that they say? We have to believe him. How do we know that there aren’t studies that contradict these? We have to trust him.

    The variables associated with “recall” (what Carr’s cited research is trying to measure) are numerous … which leads to a final reminder that *correlation is not causation.*

  4. Adam

    Sorry for the short comment. I got about halfway through the post, remembered this idea I had in a dream weeks ago, and posted it before finishing.

    I think it illustrates an overall point: browsers can solve a lot of our problems. And doing it on the user side allows for more choice. We know that being accessible to a wide audience requires resizable text and layouts that work even with large text sizes. That typographic control is lost in order to gain the benefits of being on the web.

    The same is true of hierarchy, linking, and notes. Some readers want to read, others are trying to explore a wide range of links, and some critically read every piece of text and every bit of supporting links (nested, perhaps, in stacks of four or five) to differentiate facts from theories. There is no reason that browsers can’t allow all of them to achieve their goal.

    For the reader, dim elements marked as header, footer, or sidebar, and lower the contrast of links (like Readability). Of course this requires well-written semantic html.

    For the browser, combine all the links in one place at the top of the article. Include an “open all references in new tabs” option. Of course, this requires that links all have text that is meaningful out of context or appropriate title attributes.

    For the researcher, allow “open page in new side-by-side tab” so that the source linked can be compared against the claim on the original page or two texts read in parallel.

    Some of these ideas are crazy, some won’t work with poorly-written html, but all are examples of how a smarter client can make better use of well-written hypertext.

  5. Scott Berkun


    Excellent points. I too recall the early days of Wired magazine and the adventure in reading their choices created.

    The notion of less sloppy linking is a great one – certainly many bloggers don’t think much about this. But some community blogs, like metafilter, have implied link style guidelines that get at some of the ideas you’re getting at.

    I haven’t read the wired article, but I’ll take a look. As I mentioned, reading, and hypertexts effects on it, has a fairly strong research history, I’ve just failed to find any of the papers I remember yet.

    At the end of the day, I guess i feel this: writing is hard. Good writers do many things to make the work for readers easier, and this includes good use of links. I wish blogging tools did a better job of teaching techniques for this or making better default choices, but the burden will always be on the writer, and on the reader.

  6. Scott Berkun

    Adam: These are interesting ideas, but the fundamental problem is very few people will ever use these things if they’re not on by default, and it’s unlikely they would be.

    To turn them on by default would demand they were worthy of the real estate and learning curve they demanded, but since they’re only of interest to a fraction of users, it’s very hard for a browser designer to justify making them a default part of the experience.

    Way back when, we did experiment with different tricks to help readers with links. We tried icons to show when they were media (especially the dreaded PDF), we tried default tool-tips that show the URL and indicated by color if it was the same site or a different site, but it was never a clear enough win to turn the stuff on automatically.

    So for now I think we’re stuck with plugins – perhaps someone will find a clear enough subset of these modifications that are generally beneficial and lobby for them to become default parts of the experience, but that’s a long hill to climb.

  7. Mark Bernstein


    Carr confounds many different issues in his attack on links. Certainly, it is harder to read a heavily annotated text than it is to read the same text without annotation, just as it’s easier to read a short text than a long one. Often, there is no substitute for reading deeply; links provide more information, and naturally that information requires some thought.

    Effortless and thoughtless reading is not always what we want, and sometimes even though we might want it, it may not be something we can reasonably expect. I’d *like* Misner, Wheeler, and Thorne’s GRAVITATION to be easier reading; if it could be easier, they’d have written it that way. So would Proust.

    Early hypertext research investigated this area intensely; Carr’s call for “delinkification” is very close in spirit to Bob Glushko’s “hypertext engineering” manifesto, except that Glushko’s arguments were based on real (if mis-estimated) economic costs rather than ersatz cognitive science.

  8. Phil Simon

    Great post, Scott. I completely agree with you on points #4 and #5. I’m a recent convert to readiability because most sites (yours excluded, of course) are either poorly laid out or just too damn busy.

  9. The Notes Web

    Hyperlink, it’s become important when a content of a page have a relation to other page or when the page ‘s content need to be explained by other page content. And I think there is many other reasons why a page need to include hyperlink in it.



Leave a Reply

* Required