With the information we obtained from the menu experiments, books and Cmsc434 class lectures, we designed an experiment which tried to prove the following hypothesis:
The time, number of errors and complexity of retrieving information from the internet is directly proportional to the depth of the tree structure of the arrangement of the links of the web page. The three variables, that is, time, errors and complexity will decrease as the breadth of the tree structure increases and depth decrease.
Even though our experiment did not test this, we believe that there is a maximum size of breadth. This maximum breadth is reached once the breadth is so large that the page looks unorganized, or too much scrolling is needed to read the entire page.
2.2 Subjects
We conducted our experiment on 17 subjects, 6 females and 11 males. The
subjects' age ranged from 19 to 45. Most of them were University of Maryland
students. All subjects had some knowledge of the Internet and they had used some
kind of a web browse before our experiment. results to make the analysis of the
experiment.
2.3 Materials,
Database Material
We had to create our own web pages in order to conduct the experiment. In trying to make the pages as similar as possible we excluded such things as frames and pictures. There is only one .gif file which is loaded on the cache during the down loading of the introduction page of the experiment and remains there, hence, it does not cause any time delay to the experiment. The rest of the files range from 0.5K to 1K, the difference of 0.5K did not bring any significant difference in downloading time. We took all these precaution so as to try to eliminate any false results that could have been caused by different file sizes, large gif files, and network delays.
The database was made up of 64 files with information about the island of Cyprus. We constructed 5 different designs of accessing that database.
Other Materials
Out of the 17 subjects whose results were retrievable from the server, 15 used workstations at the university of Maryland laboratories. 2 subject used Pcs outside of Maryland but connected to the network. There was no need of typing, the subject used only the mouse and screen. The screens ranged from 13 inch to 17 inch. We did not account for this difference in screen size in our analysis of the results.
Procedures
Each subject in the experiment was presented with an introduction web page which had the instructions on how to do the experiment. After reading the page, the subjects were given time to ask any question. Subjects then started the experiment. No questions were answered and no help was given during the experimentation.
Each subject was given 10 tasks. The subjects had to perform 2 tasks on each of the models we presented to them. The tasks we selected were easy to understand and we believe that all the questions were of the same difficulty. The following are two examples of the kind of questions we asked :
Pilot Study
After conducting the pilot study we saw that there were two very crucial problems. First, we had a program that randomly selected which model the subject was going to use, but since subjects had the tendency to back track to the page where the randomizing program was triggered, a subject could accidentally access another experiment and not finish the one he/she was working on without knowing. We decided to eliminate the random selection of modules and just created five different questionnaires with the different models evenly distributed in them. The second problem was that the subject simply forgot to scroll and so they kept searching with only half of the contents of the page visible. During our actual experiment we made the browser windows as large as possible and reminded the subject at the beginning of the experiment to scroll if there is more information off the screen.
Continue
Return to the
Title Page for "Depth vs Breadth in the Arrangement of Web Links"