The Relative Effectiveness of Hypertext and Text
The Relative Effectiveness of Hypertext and Text 
This study has two main objectives: The comparison of hypertext and printed text, as the title suggests, and the comparison of manual and computer generated indices.
The comparison of hypertext and printed text leads to the result that both are equally good for "reading to learn," but hypertext has an advantage over printed text in "reading to do", especially when the search tasks are complex.
The second comparison, manual vs. computer generated indices, favors the computer generated indices for reference material. The computer generated indices, like a full text index, are less effective than manually generated indices when used with narrative text. This is because the automatic generation creates too many unwanted links, thus making it harder to find the desired information.
Topic very relevant. Automation is what our business is all about.
Detailed description of the study.
The expression "random links" should be explained.
Answer the question, why a one hour practice session with hypertext vs. a lifetime experience with printed text still leads to useful results.
Less statistical data should be used in the main text. If it is necessary, seperate them in boxes.
What different kinds of reading were studied? What were the results?
When does the study suggest that computer generated indices are useful?
After reading all the other articles this one does not come to surprising results. This seems to be "yet another use hypertext for searching" article. A big question of mine still remains unexplained: Why can we expect that people can use hypertext effectively after a one hour training session, and expect it to be more effective than printed, linear text, with which we have a lifetime of experience with?