HUMAN FACTORS IN COMPUTER SYSTEMS : GAITHERSBURG CONFERENCE CONTINUE D (Review of Session on Evaluating Text Editors ) W .J . Hanse n Three Rivers Computer Corporation "An Ease of Use Evaluation of an Integrate d Document Processing System" by Michael Good , Massachusetts Institute of Technology . At the time of the experiment reported i n this paper, the Etude system was merely a prototype, so there were no expert user s available to apply Robert ' s methodology . Instead, subjects were trained on Etude an d performed a short task on both Etude and a typewriter . In the event, the typewrite r proved faster for the task . However, th e paper is able to show that the users, none o f whom has prior computer exposure, could us e the Etude system without anxiety and wit h positive attitudes . Evaluating Text Editors (Session 3B), chaired b y Henry Ledgard, Human Factors Limited, and Joh n Whitside, Digital Equipment Corporation . Tex t editing is probably the one activity that user s spend the most time performing, so it i s appropriate to devote considerable attention t o improving its human factors . This sessio n covered several techniques for evaluation o f text editors . " An Analysis of Line Numbering Strategies i n Text Editors " by M .L . Schneider, S . Nudelman , and K . Hirsh-Pasek, Sperry Univac . In " fractional line numbering, " 1 .3 is th e same as 1 .30 and follows 1 .25 ; i n "hierarchical line numbering," 1 .3 precede s 1 .25 and intermediate numbers includ e additional levels of hierarchy : 1 .3 .1 , 1 .3 .2, . . . . Fractional line numbers were shown to b e faster to use, even though the subjects wer e familiar with hierarchical numbers . (Although this result may be of limite d application to screen editors, it may impl y that section identifiers for hierarchica l text ought not to be purely numeric . ) " Evaluation of Text Editors " by Teresa Roberts , Xerox Systems Development Department, and Thoma s Moran, Xerox Palo Alto Research Center . The evaluation methods reported in this pape r can be carried out by non-psychologists with a minimum of equipment and preparation . Fou r factors are measured for each editor : (a ) time to perform a specified set of edits (b ) frequency of errors by experts with th e editor (c) learning time for novices, and (d ) functionality, the possibility of performin g a specified set of tasks easily within th e editor's command set . (The set of tasks fo r the functionality measure is itself a valuable contribution to the design of tex t editors .) The results reported for eigh t text editors accord well with my ow n intuitive evaluations of these editors, s o the methodology does seem to be producin g valid results . In general, the full screen editors evaluated at twice the speed of th e line-oriented editors : the most complex editor, TECO, was slowest to use, even fo r experts at its use . The fastest editor , Gypsy, uses only a mouse and a five ke y handset, so it scored the lowest score fo r functionality (though only slightly lowe r than TECO .) When I plotted time versu s functionality for the full screen editors, I found that the fuller functioned editors wer e indeed slower, though only by a small margin . Because it provides a methodology that can b e widely and usefully applied, this paper--an d the thesis on which it is based--merits a wide readership . " Can We Expect to Improve Text Editin g Performance?" by David Embley and George Nagy , The University of Nebraska-Lincoln . A better title for this paper might be ' Ho w we are trying to evaluate text edito r performance .' The paper reports the curren t status of a developing laboratory fo r analysis of editor usage . This laborator y includes instrumenting editors to captur e command-by-command times . A real innovatio n is the capturing of data from students a s they are using the editor for course work . Such field experiments can go a long wa y toward validating the results of more limite d experiments . Although not in the paper, the presentation at the conference reported on a technique for evaluating the cognitive loa d of individual commands . An analysis of the matrix of transition times from one command type to another can serve to divide th e inter-command time into time for absorptio n of the result of one command and time t o prepare for the next . It will be interesting to see results from this analysis when the work is complete .

/lp/association-for-computing-machinery/human-factors-in-computer-systems-gaithersburg-conference-continued-m5hThl6Ktv