Breadcrumbs provide a visual indicator that a particular node has been visited, anchor has been activated, or link has been traversed. This helps to prevent the problem noted by Nielsen: "users complained that they would often feel compelled to take hypertext links to places where they had recently been, just to make sure that they did not miss important information" ("The Art of Navigating Through Hypertext" 303).
Nielsen's hypertext, written in HyperCard, marks anchors with a checkmark if they have been activated. (Note that he calls this footprinting, a term we have reserved for a separate feature.)
Martin suggests that the reader might colour-code breadcrumbs according to how useful they perceive the link to be (88). However, this ignores the influence of context on the reader's evaluation process. A link might be ranked low the first time it is encountered. Since it would now be marked with a colour indicating its low interest, the reader would be less likely to activate it on future visits, even though it may then be much more useful. A better solution (for systems in which the overhead is justified) would be to allow reader annotation of links as well as nodes.
Eventually, breadcrumbs accumulate to the point where they are marking most places; at this point their utility is minimal. Bernstein proposes that breadcrumbs disappear after a time. "[C]rumbs represent pages we have read recently, and imaginary birds remove breadcrumbs the reader leaves unvisited for more than thirty pages" ("The Bookmark and the Compass" 43). We recommend further that the decay function governing breadcrumb disappearance be carefully tailored for the given system or hyperbook. In some cases, it might be best to let the reader determine the shelf life of the bread.