Text Menu

Anchor density is basically the number of anchors on a node. The more anchors, the more choices are presented to a reader. This is not an either/or, but a scale--the question is how many anchors per node tips the scale over from sparse to dense, and which anchors count.

Anchor density vs. link density

Discussions about link density usually assume a one-to-one correspondence between anchor and link--that each anchor will lead to only one destination. Therefore, when analysts use this assumption to examine link density, they are really counting anchors and determining density based on the amount of visible anchors.

We hold, however, that a single anchor can connect to multiple links (e.g. Samplers [37]). Reagan Library [47] and What We Will [66]both use a central anchor that users can spin around to access many destinations. Him [9] and War Games--Catch The LandMine!! [40] use the same 5 anchors and 1 anchor respectively to trigger randomly different content. Thus, a node with only one anchor can still be quite dense in terms of the number of links while remaining sparse in anchor density.

What are we counting?

There are several strategies to count anchors:

• Compare a ratio of anchoral to nonanchoral content. Extrapolating from Jakob Nielsen, a densely linked site is one where just about every bit of the content in the node is an anchor. On the other hand, a sparse site is one where less than ten percent of the content is an anchor [101] .
  
• Look at the numbers of anchors per page. A 1993 study conducted by Welsh et al, considered a high density site to be forty anchors per node while a low density site was half as many, twenty anchors per node [124]. Yet this study found that regardless of the number of anchors per site, subjects activated fewer than ten per node. Kushal Khan and Craig Locatis state that this study illustrates that there is a ceiling to the number of anchors readers will follow. They state that “too many links may also be confusing” [93, p. 177] . As a consequence from this contention, in their own study, Khan and Locatis held that a high-density site had six anchors per node while a low-density site had only three anchors per node.

  Only a very few of the sites we surveyed (both efferent and aesthetic) had low densities as defined by Khan and Locatis. Popup ads are an odd case, as they are a single anchor--but the entire node is an anchor. Diagrams Series 5 [57], Reach [28], and Charmin' Cleary [16] also consist only of anchoral content, but have very different navigational schemes and Reach and Diagrams Series are densely linked by any definition. The Jew's Daughter [45], provides a single anchor that, when acted on, reconfigures the same node. Zenobia, Queen of Palmyria [10], has a single non-distinguished anchor on each node.

  Most efferent sites in our survey had at least six anchor choices in menus, although only three had anchors embedded in the content.

Overall node to anchor density

Even though Wyoming and Washington D.C. have about the same number of inhabitants, the population density is quite different. In the same manner, density must also be considered by ratio of anchors per node in the work. While Sand Loves [38] has only two links per node, the anchor ratio to node is high, as there are only eight nodes. Conversely, Earthtrends [12] has over 40 anchors in most secondary nodes, yet the anchor to node ratio is low as the site offers material from a database with over 500 variables that provides information from over 200 countries.

Spatial density and layering

Continuing with the population analogy, the same population can occupy a very different amount of space--think of 100 people in an apartment building stacked on top of each other or the same 100 people spread out in large ranchettes. In the same way, anchors can take up a great deal of individual space on the screen or they can be layered in same space. Diagrams Series 5 [57] is perhaps the most extreme version of layering, as all the nodes are also anchors, and all the nodes lie on top of each other in a transparent stack of words that merge into black splotches and must be teased apart one by one.

Where are we counting?

While all anchor properties are intertwingled, discussions of density depend on location (isolated, embedded anchors or anchor clusters in menus and maps). The numbers of anchors, and even the ratio of anchoral content to nonanchoral content is meaningless to defining a reading experience without considering the overall context of where those anchors are. For example, if we use the extrapolated Nielsen definition in determining anchor density, then a site that places anchors in anchor clusters and avoids embedded links (e.g. Nevada Division of Environmental Protection [50]), will always be seen as a sparsely anchored site. If we consider anchor density only in terms of the anchor's relation to the text, this may be quite correct. If we consider anchor density in terms of reader choices on a node, this seems incorrect--as the menus provide many choices per node.

Readers presented with anchors choose whether or not to follow the anchor. The function of anchors congregated in a menu or map seems to be fundamentally different from the function of embedded or content anchors. These anchors show the relationship of the content to the entire site or work, usually with hierarchical substructures. Expecting readers to trigger all of these anchors would be as absurd as expecting restaurant goers to order everything on the menu. So Khan and Locatis' assumptions of a connection between anchor density and confusion may not hold in densely anchored menus or maps as it does in sites with many anchors embedded in the content.

So what definition should we use?

As we only found a few sites that would truly qualify as sparse under even a liberal reading of Khan and Locatis or Nielsen (10 percent or less of the node's elements are anchoral), dividing sites into merely sparse and dense does not mean a great deal.

Efferent sites that had higher anchor densities tended to use menus rather than embedded anchors. If anchor density is counted merely as the ratio of anchors to text, then sites with menus and maps are dense sites. However, if anchor density is counted only in embedded content, then these sites are sparse, as they rarely had embedded anchors. Anchors embedded in the content involve more connective tissue and connotative relationships. Readers must stop their content reading to ponder the importance of the anchor and its relationship to the content.

Thus, density must take anchor clustering (in maps or menus) and anchor isolation (embedded anchors) into account. We would thus break density definitions down further into:

• Menu-driven density: sites where navigational anchors are congregated (in maps or menus) but with sparse anchors in the content.
  
• Content-driven density: sites with many anchors embedded in the nonanchoral content.

With this division, we can see that most of the efferent sites in our survey have a menu-driven density. This reliance may be because efferent sites are serving information that can readily fit into a hierarchical menu system, and thus can offer readers the particular information they are seeking. Aesthetic sites, on the other hand are more evenly divided between menu, content, and sparse strategies.