Effects of Link Arrangements on Search Efficiency
As a result of this investigation, several pieces of information can be found which could be of use to the expanding web-design community. Upon revisiting the five salient findings in the discussion, each yields some valuable information. Furthermore, it is important to not limit the re-visitation to statistically significant data.
"Changing the column layout of a small number of links does not have as profound an impact on search time as changing the column layout of a large number of links".
This could be a message to designers of large and complex pages to consider, with caution, the layout and design of the page. It could also be interpreted as a green light for web designers of pages with few links. In general, the rule for web page design that results from this piece of data is "the more links on the page, the more consideration should be taken." This motto is most likely in keeping with current industry standards. Those involved in data intensive web sites (i.e.: search engines, libraries) are probably more likely to concern themselves with orderly presentation in general, than are smaller, narrowly focused web sites.
A user forced to scroll though a lengthy list will take significantly more time to find a given web link than a user with a short, non-scrolling list.
Partially the result of scrolling and partially the result of increased search difficulty, this information is simultaneously obvious and important. In general, the practitioner should always realize that any portion of data that is part of an obscured set will take a longer time to find. More obviously, the more links on the page, the better hidden individual links become.
"Results indicate that column treatments of a large numbers of links has an effect on search time."
This statement could imply that if the designer is given no choice with regards to page depth, that no tested improvement can be made on the search time. In other words, nothing shows that arranging many links in particular formats is going to generate lower search times, so choose the best one from a design point of view.
Another possible interpretation for the practitioner is to couple the above statement with the first statement and say "if the web designer has a depth choice, use a format which mixes and matches the weight of the significant interaction effect with the weight of the insignificant column treatments."
If attempts to duplicate this experiment are undertaken, there are several considerations that should be made that were not made. A retrospective view of the entire process first points to the need for more subjects. This implementation did not afford researchers the opportunity to involve more subjects, however, it is possible that more comprehensible results would have been obtained with more subjects.
Another consideration to make is search task control. Although some attempt to control where the search tasks took the subject, there were undoubtedly free variables such as word familiarity and subjective prominence of the word in a list. In other words, a subject may be given the word "cat" and asked to find it in a list of words which are mostly 4 characters long. Such a search would be more difficult that a search for the word "dictionary" in the same list.
Still other considerations are to change the variables more radically. Given the marked difference in the results in each of the link arrangements, it could be argued that web pages using 10, 100 and 1000 links per page could yield stronger results with respect to the findings of this experiment.
It may be worthwhile to complete a second experiment using the same treatments -- one-, two-, and three-column formats with a modification to the independent variables using an alphabetized and/or categorized link arrangement style.
Given the comments of subjects and the nature of the data generated by this experiment, it seems possible to expand this project into a more courageous undertaking:
WWW Text Link Navigation: An analysis of
link organization, cardinality
and familiarity and its impact on user proficiency and
satisfaction.
Independent Variables:
Organized: columnar, alphabetized, grouped
Not Organized: paragraphs, random(x, y), ungrouped
Advanced User: Web site visits greater than pre-determined amount
Novice User: Web site visits less than or equal to pre-determined
amount
Lots of Links: Link number greater than pre-determined amount
(always scrolling)
Few Links: Link number less than or equal to pre-determined
amount (never scrolling)
Dependent Variables:
Time: Elapsed time from visual presentation to choice (objective)
User Satisfaction: Questionnaire based user analysis (subjective)
The most interesting piece of information to emerge from this experiment is probably the interaction effect between columns and links. This interaction effect, as discussed, says that changes in column layout have more dramatic effects on the search time of large data sets than the search time of small data sets. This is certainly a refined (if not new) hypothesis. However, these tests alone are probably not enough to support the new hypothesis. To be able to support this hypothesis with any vehemence would require much more research.
The data for this entire experiment was the result of two collection trips to the Naval Academy in Annapolis. Surely subsequent implementations of this project could and should include a more diverse sample population. Another suggestion is to consider a similar study which also takes into account color contrast -- a user customizable feature.
Forward to Acknowledgements
Back to Discussion
Back to Title Page