Re-Thinking In-Line Linking: DITA Devotees Take Note!

by Mark Baker

The question of whether links on a web page should be inline in the text or relegated to one of the margins is not a new one. It deserves re-examination because the increasing popularity of the Darwin Information Typing Architecture (DITA) introduces a technical bias into the question. Managing inline linking in reusable content in DITA is complex, which makes it expensive, which makes it rare. It is commonly considered a DITA best practice to avoid inline links in favor of links created by external reltables, which means that the links are listed at the bottom of the page. (DITA 1.2 keyrefs provide some support for inline linking, but at the cost of significant complexity).

This strikes me as the tail wagging the dog — a design approach being dictated by the limits of a particular tool. Naturally, DITA advocates justify this approach by referring to those studies and style guides that come down on the no inline-linking side of the question. But I think it is a design approach with serious flaws and dubious justification, and this limitation should give people pause when they consider jumping on the DITA bandwagon.

There have been several small studies on the appropriate location of links in a web page, and they have produced conflicting results. There are a few reasons for this, including the time at which the studies were done, the design of the studies, and the presumptions that the researchers made about what kind of reading web pages should be optimized for.

In a comment on my blog post Are We Causing Readers to Forget, Larry Kunz made reference to the IBM style guide (I don’t have a copy myself, so I am grateful to Larry for the reference):

The IBM Style Guide (a very comprehensive guide for writing software documentation, and now available to the general public) recommends placing links in a list at the bottom of a topic (page) rather than inline. Their research has found that a reader is less likely to become distracted when the links are at the bottom, where they won’t entice the reader to go off-topic but where they still can be used if needed.

I don’t know how old the research is on which this recommendation is based, but I do know that 15 years ago I certainly found inline links distracting. My brain saw the underlining and the color change as emphasis in the text. But my brain is now fully used to seeing links (which are generally more subtly presented today) and I read right past them without any sense that the text is emphasised or inflected in any way. Concern on that front, therefore, seems somewhat out of date.

There are other ways links could be considered distractions, but I will get to them later. Another criticism of inline linking is that readers may fail to see the links. This is the exact opposite worry: the worry that inline links are so inconspicuous that people won’t even notice them, let alone be distracted by them. This opposite fear leads, interestingly, to the same conclusion: place the links at the bottom of the page.

This, alas, runs afoul of another common web study: eyetracking studies, which show that people’s eyes follow a F pattern on a page. Combine this with Jacob Nielsen’s unsurprising findings that people largely ignore the navigational elements of a page and go straight to the text, and it looks like putting links at the bottom of the page is going to be a great way of making sure no one ever sees them.

This conclusion is contradicted, however, by a 2001 study from SURL (The Software Usability Research Lab at Wichita State University), which found that links were equally easy to find, no matter where they were located. Here is where we can begin to see how the design of the experimental influences the results. In this study, the participants were asked questions which required them to follow the links to find the answers. Thus they had to explicitly look for links. This suggests that when people have to find links, it doesn’t much matter where they are. But the success rate of people explicitly looking for links does not tell us whether people not looking for them would be more likely to notice them if they were inline or in the margins.

Another problem with this type of study is that, precisely because it is a study, and the participants know that it is a study, they will expect that the answers are to be found within the content or its links. The text being studied exists in an implicit box, from which no participant is likely to break out by Googling for the answers on the test. But in real life, where there is no guarantee that the text before you contains all the answers you seek, Google beckons constantly.

Interestingly, while the study found no difference in the ability to find links based on their location, it found a strong anecdotal preference for inline linking:

[T]here were significant subjective differences between the link arrangements favoring the embedded links. That is, participants indicated that they believed that embedding the links within a document made it easier to navigate, easier to recognize key information, easier to follow the main idea of the passages, and promoted comprehension. Moreover, participants significantly preferred the embedded link arrangement to the other arrangements. Conversely, placing links at the bottom of a document was perceived as being the least navigable arrangement, and was consequently least preferred.

This strong user preference is apparently at odds with studies that actually set out to test comprehension and retention, at least according to Nicholas Carr’s alarmist The Shallows (an excerpt here from Wired). Carr worries that the web is full of distractions, such as inline links, which are rewiring our brains and causing us not to think deeply or retain what we read. All this is deeply at odds with the many tracts that tells us that Web users don’t read, they skim, and that therefore we should make every page a frenetic mix of titles, tables, lists, graphics, and other geegaws, lest the sight of two plain paragraphs in succession should send the reader spinning off into space.

On the web, you can’t win, it seems. Readers won’t read plain text, and they won’t retain text filled with distractions.

All of this concern seems to me to rest on two deeply flawed presumptions: that reading on the web should be like reading on paper, and that people read in order to retain the text. Here’s what’s wrong with both these ideas:

Distinguishing information seeking from information consumption

All the studies of reading on the web that I have seen seem to miss one very basic fact. Before information consumption, comes information seeking.

In the book world, information seeking begins with a trip to the library or to the bookstore. It then involves a lot of flipping through card catalogs, a lot of walking through the stacks with your head tilted sideways reading the spines, and a lot of flipping through introductions, TOCs, and indexes as you select which books to actually borrow or buy. Finally, you proceed to the checkout and take yourself back home with your stack, settling into your easy chair, and start reading. That is a lot of time, effort, and cash invested in information seeking before any real reading begins.

By contrast, on the web, information seeking starts with a Google search, followed by opening all the likely looking search results in browser tabs, followed by going through the tabs one by one skimming for promising content, closing unpromising tabs, and finally going back and reading more thoroughly the tabs that seemed promising (at least, that’s my process). All this involves a lot of reading that is pure information seeking, equivalent to the time spent in the card catalog and the stacks at the library or bookstore.

But when people compare web reading to book reading, they count all of the information seeking behavior as “reading” on the web, whereas they only count the time after you settle into your easy chair as “reading” for the book world (if they measure it at all). This is a gross distortion. In the library to find your way to texts; on the web you find your way through texts.

Another difference between the library and the web is that once you have found a likely text, your investment in that text in the book world is much greater than your investment in the web world. Even if the books you got from the bookstore or library prove disappointing at first, you are likely to stick with them because it will take a lot more time, effort, and money to replace them. Sticking with them to see if they will eventually prove at least adequate for your needs is a good economic strategy because the cost of starting over is high.

On the other hand, if you find a text on the web that proves disappointing, your investment is very small, and it makes much better economic sense to look for a better text, since doing so is very inexpensive. This does not represent a loss of the ability to concentrate, just a rational change in behavior based on the reduced cost of acquiring alternate texts.

So, most of the eyeballs on your text are not in settled reading mode, they are in wayfinding mode. They will skip and skim, because that is what you do in wayfinding mode. (This is not a web thing at all, since you do the same thing standing in the stacks at the library. But the eye tracking tools can’t measure you there.)

In fact, the mostly-wayfinding/sometimes-reading pattern is exactly what shows up in the studies. It is simply that the studies all seem to interpret it as a difference between reading paper and reading on the web when a simpler explanation is that information finding in a paper world mostly takes place before the seeker sits down to read.

Some people think of links as exit points, but for someone in wayfinding mode, anything you are not interested in or don’t understand is an exit point. If I am reading (or skimming) a text, and it makes a reference to a concept I don’t understand, or a task I don’t know how to perform, what do I do? My choices are:

1. Go back to Google and find another text written more to my background.
2. Highlight the text that refers to the unfamiliar concept or task, then right click and select “Search Google for…” (that is, make my own link).
3. Go hunting around the margins of the page to see if there are helpful links there.
4. Click on the handy link that the thoughtful author has provided right on the troublesome reference and go to content on the same site that clarifies the concept or describes the task.

Options 1 and 2 both take the reader away from your site, perhaps for good. Option 3 is improbable, given that:

1. Most sites don’t do this, so why would you be in the habit of looking for them?
2. Even if there are links, their relevance may not be obvious, and I have to go through the entire list to find one that might be relevant.
3. Option 2 is easier and works universally.

Only option 4 keeps the user on your site. A useful link turns an exit point into a re-entry point. It also has the added bonus that it is quite difficult to select text that is a link, so it actively discourages option 2. Not providing a link is not likely to keep the disappointed wayfinder on your page; providing a link creates the possibility of keeping them on your site.

The goal of most reading is not retention

Five or six times a year I make a six hour drive to visit one of the scattered parts of my family. Along the way, I read hundreds of road signs, yet by the end of my journey, I don’t remember any of them.

Did those road signs fail as communication devices because I don’t retain them? Of course not. Those signs exist to inform me of specific local situations requiring specific local decisions and actions on my part. For them to do their job, all this is required from me is that I decide and act correctly in the moment. Once I have taken the correct action, there is no reason to remember the sign. Forgetting past signs is actually a good thing, allowing me to concentrate on the immediate road ahead.

Most of our reading is like this — in the moment. When we read a novel, we read for the visceral impact of the imagined events, not to retain the text (unless we are reading for school, which, of course, makes the reading joyless).

When we read technical documentation, we read to complete our immediate task, not to commit the procedure to memory (if we do remember a procedure, it will be through repetition, not reading to retain).

When we read the news or a family letter, what we read may pass into the mass of experiences that shape our hopes and opinions, our sense of how the world works, but they are seldom individually recalled, unless highly affecting and frequently repeated, like the images of the planes hitting the twin towers.

No one reads tweets for retention.

When we read articles or blogs in our professional sphere we read mostly to confirm or to dispute. Let’s be honest about this: if you are pro inline linking, you are reading this to confirm your preference, and if you are anti, you are searching for weaknesses, omissions, and flaws in the argument so that you can rebut. And I would suggest this is a perfectly healthy way to approach this. Antagonism leads to debate, and debate leads to discovery and new understanding.

Even in a school setting, we don’t generally read for lifetime learning; we read for test passing. If we are going to talk about communications media rewiring our brains, we have to acknowledge that the school system rewires our brains to retain texts up till exam time, and to discard them as soon as the exam is complete.

And while we are on the subject of schools, we should note that a great many of the studies on this issue use what is essentially a school model of testing. The subjects are assigned to read a text and then to answer questions, just as they would be if they were in school. As in school, they are not motivated by their own tasks or their own interests, and, as in school, they are consequently easy prey to distraction. Purpose and focus makes all the difference in concentration, retention, and resistance to distractions, but purpose and focus are hard to reproduce in laboratory conditions.

Reading to do

It is comparatively rare, then, that we actually read with the goal or retention and long term learning, either in school or out of it. Most of the time we are reading to complete a specific short term task. Hypertext and search are great for this because they allow us to move very quickly through a large body of content to narrow in on the piece of information we need to complete our immediate task. There is no reason to read thoroughly every text we encounter on this search and no reason to remember any of them once we have completed the task. To complain that people reading for this purpose do not read thoroughly or retain what they have read is to entirely miss the point of why they are reading in the first place. They are reading to do, not to learn.

Reading to learn

On those relatively rare occasions when we do actually sit down to read with the intention of learning and retaining, I think we naturally fall into the mode of reading thoroughly and with attention. At least, I do, and I, at least, don’t find that I even notice links in the text when I am in that mode. I am motivated by a goal to which I am committed and in which I am deeply interested. I’m not seeking distractions, and I am not easily distracted.

On the other hand, if I do come across a concept in what I am reading which is not explained in the text, I do want to take a side trip to understand that concept so that I can return to what I am reading and continue with full comprehension. Again, therefore, an inline link is welcome.

That said, I do find that it is when I am reading in this mode that I tend to move away from the web and pick up a book or an e-book. If I am set upon a course of study, it is generally easier to pick up a good long-form volume that is designed specifically to instruct in that subject area. I will usually keep a tablet nearby in case I need to look up something that the author does not explain, but generally, I do this type of reading off the grid.

This is important too, because if other people are like me in this regard, that means that an even higher percentage of reading on the web is done not for retention but for action, immediate experience, reference, or debate. To condemn inline linking on the grounds that it is detrimental to retention, therefore, is to apply an almost entirely irrelevant standard. The issues should be, does inline linking support navigation, information finding, and getting to action?

Escaping the limits of paper

In the book world, it was reasonable to assume that your reader was going to read your text in a single narrative flow. This was due to the limits of paper and the effort required to switch to another text. Linear reading was an optimal strategy in the paper-and-libraries world, and so it became a cultural norm, something that people were trained to do, not because it was optimal in abstract, but because it was optimal within the limits of the available technology.

On the web, it is not reasonable to expect the same linear reading behavior to be the norm, nor should we wring our hands or imagine that civilization is coming to an end because people are discovering new ways to optimize their information finding in an online world.

Helping the unprepared reader

Readers come to texts from impossibly diverse backgrounds. Even in the most carefully planned book-based curriculum, no two students arrive equally prepared (or identically unprepared) for the same text. With web texts, arrived at by search, link, or social curation, readers, if anything, arrive even more unprepared.

The only way that the unprepared reader is going to successfully comprehend your text is by filling in the gaps in their preparation. Keeping their attention focussed on the single narrative line of your text is not the right answer, because they are not prepared to understand that narrative line.

Forcing them to continue can only produce an uncomprehending recollection of the text itself (a phenomenon anyone who has spent time in a classroom is thoroughly familiar with). But on the Web, there is no way to force them to continue. They can leave anytime they want to. And once they leave, why should they come back to your site, since it failed them? But if you provide them with a rich set of links, you provide them with a way to fill their gaps and equip themselves to comprehend your text. You thus give them a reason to stay, and a reason to return.

Sometimes not linking can be the best strategy

This is not to say that you should always prefer inline linking. It is certainly possible to imagine situations in which inline links should be avoided. Ginny Reddish points out, for instance, that links are distracting for low-literacy readers.

Of course, we don’t design most content for low-literacy readers We don’t design most things for novices. If we did, the Tour de France would be contested on tricycles. We design things for regular users, and expect novices to push through their novice difficulties and learn. But even so, there will be times when you design a site for low-literacy readers.

Different types of content may also call for different approaches to linking. More contemplative content may be better served by fewer links, while more action oriented content may benefit greatly from rich linking. Commercial content clearly benefits from rich linking that keeps the reader browsing and seeing more products. Public service or compliance oriented content, on the other hand, may benefit from providing a fixed path (providing the content itself works well enough to keep the reader on it).

If you choose not to link, however, don’t fool yourself that the mere absence of links will induce people to read your content to the end. Google is always just a swipe and a click away, and everyone but a rank novice web user knows exactly how to make a link for themselves anytime they want one.

A good link strategy is important

Given the choice to link, it is, of course, important to link in the right way. Links should provide context so that the reader has a reasonable expectation about where the link may lead. Links should be useful, rewarding, and deliver on what they promise. A poor or inconsistent linking strategy, or poor selection of link targets, will frustrate the reader and send them away from your site.

Fundamentally, links should fulfill a useful purpose for the reader and/or the writer. Purposes may differ for different sites, but the main purposes I see are these:

• Assist the wayfarer for whom the current text is not their final destination reach the text they really need (preferably on your site).
• Assist the reader who is not fully prepared for the text to fill in gaps in their knowledge or preparation so they can complete their task using your text (as opposed to someone else’s).
• Support your arguments or claims with evidence or argument from other sites. (Yes, this involves sending the reader to another site, but if doing so improves your credibility, that may still be a win.)
• Keep the reader on your site as long as possible. (Your site is what matters here, not the individual page. For most commercial purposes, having the reader visit multiple pages on your site, rather than just one, is highly desirable.)

Achieving a consistent link strategy

I began this essay by complaining about link strategy being dictated by the limitations of tools. Unfortunately, most tools on the market limit your linking strategy in one way or another. Sometimes this means limiting the kinds of links you can create or maintain, but the more common problem is that they simply make the creation and management of links so expensive that in practice organizations cannot afford to provide rich linking.

Whatever your linking strategy, you want to make sure that that strategy is implemented consistently. You don’t want your tools to dictate your strategy to you, and you don’t really want to leave it all in the hands of individual authors either, especially if you are reusing content in different places where you might want to implement different linking strategies.

Link management does not — or should not — mean letting individual authors create links higgledy piggledy and then deploying expensive CMS systems and/or exotic mapping systems to keep them from breaking. What link management should mean is having a consistent linking strategy and being able to apply it in a consistent way.

This is really hard to do if your tools don’t support central control of link creation and you depend on individual authors to adhere to your linking policy. And it creates significant overhead for the author if they have to make these kinds of decisions as they write. If authors have to concern themselves with when to link, what to link to, and where to locate the links on the page, chances are they will not create very many links, and certainly you will not get consistency across the site especially if you have a large number of authors.

One solution is to use a soft linking approach, such as that supported by the SPFE architecture. Soft linking lets authors quickly mark up mentions of concepts, objects, and tasks that you might want to provide links for, without creating actual links. At build time, you can then apply a consistent link policy to these potential link points, selecting when to link, where to place the links, and which resource to link to. All this is done algorithmically based on available metadata, which means you get highly consistent linking, and the correct density and location of links for each media or publication, without having to police individual authors, and without your strategy being dictated by your tools.

We are still adapting to hypermedia systems. This applies equally to readers and writers, thought we all seem to adapt faster as readers than we do a writers, behaving one way when we read, but still writing as if everyone else read the old way. But hypermedia are very different from the old fixed media, and we cannot continue to think of the content we push to the web the same way as the content we used to print and bind and throw in the box. On the web, links are not a decorative gloss that we lay over a static publications. They are the nerves and sinews of hypermedia. We need to stop dismissing links as distractions or exit points, and start using them strategically as mechanisms for navigation, empowerment, and reader retention.

About The Author

Mark Baker, President of Analecta Communications Inc., is a twenty-year veteran of the technical communication industry, with particular experience in developing task-oriented, topic-based content. He has worked as a technical writer, a publications manager, a structured authoring consultant and trainer, and as a designer, architect, and builder of structured authoring systems. He blogs at everypageispageone.com.