RESULTS FROM THE FIRST WORLD-WIDE WEB USER SURVEY


GVU's WWW User Survey Home Page

Results From The
First World-Wide Web User Survey


James E. Pitkow
Graphics, Visualization, & Usability Center
Georgia Institute of Technology
Atlanta, GA 30332-0280

Margaret M. Recker
College of Computing
Georgia Institute of Technology
Atlanta, GA 30332-0280

ABSTRACT

The explosion of World-Wide Web (WWW) across the Internet is staggering, both in terms of number of users and the amount of activity. However, to date, no reliable characterization exists of WWW users. In this paper, we report results from a survey that was posted on the Web for a month, in January of 1994. There were several goals motivating our survey. Fir st, we wished to demonstrate a proof of concept for WWW technologies as a useful survey medium. Second, we wanted to bet a-test the design and content of surveys dealing with the Web. Third, as mentioned, we hoped to begin to describe the ra nge of Web users. In one month, we had over 4,700 respondents to our survey. Their responses helped us to begin to chara cterize WWW users, their reasons for using the WWW, and their opinions of WWW tools and technologies.

KEYWORDS

statistics, surveys, demographics, tools, administration

INTRODUCTION

According to recent estimates, the Internet is gaining roughly 150,000 new users per month, joining 20 million exist ing Internet users [New York Times, 1993]. At the same time, the number of WWW packets is growing exponentially. While t hese numbers provide useful approximations on the amount of users and their Web-related activities, little is known abou t their backgrounds, interests, and characteristics.

In this paper, we report results from a survey that was posted on the Web for a month, in January of 1994. There were se veral goals motivating our survey. First, we wished to demonstrate a proof of concept for WWW technologies as a useful s urvey medium. Second, we wanted to beta-test the design and content of surveys dealing with the Web. Third, as mentioned , we hoped to begin to describe the range of Web users.

In one month, we had 4,777 respondents to our survey. Their responses helped us to begin to characterize WWW users, thei r reasons for using the WWW, and their opinions of WWW tools and technologies. However, as we will discuss, these respon ses are representative of a particular class of WWW users, and therefore cannot yet be generalized to the general WWW po pulation. For this reason, we plan to improve the survey and run it periodically in the upcoming years. This will allow us to track changes in the population as WWW evolves and matures. Furthermore, the survey results will help us extrapola te to the general Internet population.

METHOD

Using the Internet to survey users is not new. However, the problem with most of these surveys has typically been an ina dequate level of participation. We believe that the lack of respondents is primarily due to the overhead incurred. That is, most surveys ask the users to report their responses via electronic mail. For example, a survey on WWW technologies that required e-mail responses was posted to comp.infosystems.www in late January 1994. The results of the survey revealed that only fifty-five responses were gathered. While surveys of this type are more effective than conventional mail or phone, they still require considerable end-user effort, which ought to be avoided if possible.

Fortunately, the level of client interactivity supported by Hypertext Markup Language Plus [HTML+], i.e. forms, is suffi cient to enable low-overhead, point and click responses, as well as text entry. Further enhancing the attractiveness of using WWWW technologies is the Hypertext Transfer Protocol [HTTP] server's ability to retrieve documents (the actual sur veys) and execute programs (response logging software). Thus, not only are users able to respond to surveys in an intuitive, non-labor intensive manner, but the results can be logged automatically via the HTTP server.

The design of our survey attempted to exploit WWW technologies. Specifically, we created five separate surveys (with que stions on general background, HTML, HTTP, Mosaic, and WWW Browser/Internet Usage) and rudimentary logging software. The decision to create separate surveys was made for several reasons. First, we wanted users to be able to complete each sur vey in three to five minutes. If we had created one survey with all the questions, we felt that users might have been le ss likely to take the time necessary to complete the entire survey. Second, WWW browsers like Mosaic have difficulty man aging documents with large number of embedded forms. We intentionally designed the surveys to minimize the number of emb edded forms (but still received comments from users that display problems interfered with the overall aesthetics and eff ectiveness of the survey). Finally, categorizing questions facilitated clarity of intention. That is, the users knew bef ore-hand the area each survey was intended to cover and therefore did not end up midway through surveys finding themselv es unable to answer questions.

The logging software parsed the user's responses and append the results to file. Each entry in the logs was time coded, but machine names were not recorded. While this method ensured a moderate degree of animity (correlations between the ti me logged and the times recorded by the server log can still be derived) it primarily resulted from the use of National Center for Supecomputing Applications' (NCSA) httpd_1.0a5 server, which does not pass remote host information into the s ub-shell's environment.

As stated above, the surveys were divided into five categories: 1) general background questions, 2) HTML questions, 3) H TTP questions, 4) Mosaic questions, and 5) WWW Browser/Internet usage. We felt that this stratification was sufficient t o help us characterize WWW users, their reasons for using the WWW, and their opinion of WWW tools and technologies.

The surveys primarily used HTML+ forms. Specifically, radio button (buttons that only allow one answer to be selected pe r question) were used to list available choices or corresponded to numerical ratings. We intentionally designed as few q uestions as possible that required text entry as this imposes greater demands on the end-user and decreases control over of the content of the responses.

The survey was officially announced on January 17th, 1994, on the comp.infosystems.www newsgroup. The posting sim ply stated the impetus behind the survey (i.e., to provide the community with a characterization of WWW users) and the l ocation via the Uniform Resource Locator [URL] of the survey. Subsequent notices were mailed to NCSA's What New Page maintainers and the maintainers of the WWW server at CERN due to the high visibility of these servers. Several point ers from other documents on the Web were also observed.

SURVEY QUESTIONS

The General Information Survey (11 questions) asked questions that involved no prior knowledge of WWW technologies. We m ainly asked questions such as user age, occupation, domain, geographical location, and computer platform and usage. As w ith all the other surveys, a comment box was located at the end of the survey for users to contribute whatever informati on they thought relevant.

The HTML Survey (12 questions) was intended to characterize how difficult HTML is for users to learn and use. Question a sked respondents about the number of HTML documents authored, whether or not they knew if HTML was an International Stan dards Organization (ISO) standard (HTML is not an ISO standard), how difficult/easy they found it to learn FORM and ISMA P, and if the documentation on HTML was current and easy to understand. We also asked background questions inquiring abo ut the number of years of programming experience, and the number of languages known. An additional question was added on Jan 21, 1994. It asked whether or not users had prior experience with Standard Generalized markup Language (SGML) (from which HTML is derived).

The HTTP Survey (14 questions) was primarily designed to determine the extent of knowledge people have about HTTP. Hence , respondents found questions asking about whether they knew about server scripts, modules, CGI scripts, and ISMAPs and if they had used any of these methods. We also asked if respondents operated a server (and if so, which one and what por t it listened to). Users were also asked to rate, if applicable, the existing servers at the time (CERN, Gn, NCSA, and P lexus) on a scale from one to nine.

The Mosaic Survey (5 questions) was intended to gain feedback from users of NCSA's Mosaic. This questionnaire was the sh ortest of all the surveys, (we had asked the staff at NCSA to submit additional questions). Basically, we wished to know what platforms users primarily used (this question was heavily biased - see Discussion below), what rating they would a ssign to Mosaic, how they perceived the support staff at NCSA, and how helpful they found the on-line help.

The WWW Browser/Internet Usage Survey (20 questions) sought to find out more about how and why people use their WWW brow ser. This category was broken down into five subcategories: frequency, motivation, typical activities, media type assess ment, and general usage traits. The first two categories are self-explanatory, while the others might require some expla nation. Since WWW browsers allow access to almost all Internet resources, we were interested in the degree to which thes e browsers are replacing the client software designed for each individual resource. Hence, we asked questions on browser use to access of gopher, archie, FTP, wais, etc., as well as questions on use for exploration and accessing other non-t raditional resources (e.g., weather). For the media type assessment, we wished to determine media preferences. The surve ys used radio and check box buttons for questions addressing motivation, frequency, and general usage traits, and used s calar radio buttons for the typical activities and media assessment questions.(1)

RESULTS

Overall, there were 4,853 responses to all surveys combined. Because our design did not ensure that only one surv ey was submitted per user, we developed software to identify multiple submissions. With this list, we manually went thro ugh and compared the time of submissions and if necessary, the content of each submission. That is, if two surveys were received within a fifteen minute period, the content was inspected for differences in responses. This method is both tim e consuming and vulnerable to error. In all, there were 1006 submissions received from identical machines. From this, we discarded 76 responses, or seven and a half percent from duplicate machines (or one and a half percent of the 4,777 net responses)(2). In addition, since our design accepted partially completed surveys, the number of responses gathered for each question frequently differ within surveys.

In the next section, we discuss the findings from each survey, followed by a discussion of these results.

GENERAL USAGE

There were over 1344 valid responses in this survey category, accounting for 28% of all the responses. It is important t o note that the above number represents the total number of people who completed the survey, but as previously mentioned , this does not mean that there were that number of responses for each question.

The results indicate that 56% of respondents are between the ages of 21 and 30, 94% are male, and 69% are located in Nor th America. Over 97% percent report using Mosaic as their browsing tool, 88% percent use UNIX as their platform, and 55% use single-user machines (though, as we discuss in the Shortcomings section, the latter results must be interpreted wit h care) (see Table 1).

In terms of occupation, 45% of the respondents describe themselves as professionals, and 22% as graduate students (the t wo largest categories). Thirty-six percent of the respondents claim to spend over 10 hours/week exploring the Internet, while 42% claim to spend 0 to 5 hours/week. Of interest to enterprises contemplating commercial use of the Internet, 55% of the respondents say they might be willing to pay fees for access to WWW information repositories, while 40% say they would not (see Table 1).

BROWSER USAGE

There were over 1000 valid responses to questions regarding browser use and activities (see Tables 8 and 9). A majority of users access their browsers several times daily; 41% say they spend less 5 hours per week exploring the Internet, whi le 21% claim to spend more than 10 hours per week. Users seem to be fairly evenly divided in their orientation, with 28% preferring search mode, 38% preferring text mode, and 33% preferring visual mode (see Table 9).

We surveyed users as to how often they used their WWW browser, instead of accessing specific client services (e.g., Wais , FTP, etc.), where 1 = "never" and 9 = "always." The results indicate that, overall, users show a s trong preference for using their WWW browser for general Internet exploration and instead of the standard Gopher and Wai s clients. They show a slight preference for using a Web browser for text retrieval and Archie services. They are neutra l in their preference for using WWW browsers to access publisher information, weather, and news of research activities a nd abstracts (see Table 8).

Finally, we asked users to rate their preferences of the various media supported by the Web. Users show a strong prefere nce for keyword search and embedded images. They show a preference for text and spawned images, and a slight preference for movies and sounds (see Table 8).

HTML PATTERNS

The survey on HTML was answered by 701 people. Table 2. presents the actual number of responses per question as well as the percent per category within each question. Interestingly, slightly over 50% of the users have authored over ten HTML documents, while 7% have never authored a HTML document. Just the same, while a handful (3%) incorrectly answered that HTML was an ISO standard, most respondents (66%) were unsure. Most people (77%) had over ten years of programming experi ence and knew six to ten programming languages (41%). The late-entry question on prior SGML use indicates that 91% of th e users had never used SGML before.

HTTP PATTERNS

The HTTP was the least answered survey (10% of the total responses), with only 481 people submitting responses (see Tabl e 4 for the frequency and percent information) Most users operated a server (78%), with NCSA's server being the most wid ely operated (71%), and port 80 (78%) the most listened to port. Note that the total number of responses differ for each of these questions, so care must be taken in cross-question comparisons. Even though 58% knew what server scripts are ( 41% have knowledge of CGI scripts), most have never written a CGI script or used ISMAPs. Finally, out of the people who had used the different servers (CERN, Gn, NCSA, Plexus), the highest mean rating, 7.5 (on a 0 to 9 point scale), went to NCSA's server, though no server received a rating below 5.8 (same scale).

MOSAIC PATTERNS

There were 1079 responses to the Mosaic survey. Not surprisingly, most users (93%) were using Mosaic for the UNIX platfo rm (see Table 6), due to the unavailability of HTML+ forms support in Macintosh and DOS/Windows environments. Over a thi rd reported never using the items contained in Mosaic's Help menu while roughly 30% used the Help menu on a weekly or mo nthly basis. Only 7% found the on-line help on Mosaic to be not useful, with the rest split between considering the help useful (52%) and somewhat useful (40%). The mean rating of Mosaic was 8.086 (same scale as above) with the support staf f at NCSA receiving a mean rating of 7.458.

USER PATTERNS

The survey results allowed us to identify patterns or clusters among users' preferences and activities, using the likeli hood ratio chi-square test and Cramer statistic (G2 and C). From the General Survey (see Tabl e 10), we note a strong interaction between shared machines and the occupation of the users (G2 (6) = 199.290, p < .001; C(6) = .388), with more users sharing machines in the educational environments and less in the professional environments. We also observe an interaction between shared machines and the users age (G 2(6) = 122.729, p < .001; C(6) = .306), with younger users more likely to share machines than older users. As can be expected for the above interactions, we find that occupation is associated with user age ( G2(36) = 835.563, p < .001, C(36) = .326), reflecting the younger responses from stud ents. Mild interactions exist between shared machines and willingness to pay fees, with shared machine users less willin g to pay for fees and single machine users willing to pay fees for WWW database accesses. Interestingly, we note a sligh t interaction between user occupations and fees (G2(12) = 52.816, p < .001; C(12) = .141), where students would not pay for access and professionals would.

The analyses of the HTML survey offer less insight. Here, we note that the number of years programming interacts strongl y with the number of programming languages known (G2(16) = 733.633, p < .001; C(16) = .514). This follows intuition, as users with few years programming know few languages and users with over twelve year s programming correspond to knowledge of over twelve languages known. A moderate interaction was observed between the nu mber of HTML documents authored and uncertainty of whether HTML is an ISO standard (G2(8) = 6 1.080, p < .001; C(8) = .210).

Numerous interactions exist between the questions in the HTTP survey, though, most are as expected (see Table 14). Of no te, we found a strong interaction between knowledge of ISMAPs and ISMAP use (G2(2) = 277.407, p < .001; C(2) = .774) along with knowledge of CGI scripts and CGI script use (G2(2) = 155.007, p < .001; C(2) = .582). As expected then, knowledge of server scripts relates to knowledge and use of CGI scripts and ISMAPs , with the latter having a slightly weaker association. Similarly, we observe that operating a server interacts moderate ly with knowledge and use of CGI scripts and ISMAPs.

Analysis of the Mosaic survey (see Table 13) revealed that those users who used the Help menu on a weekly basis were mor e inclined to view the on-line help on Mosaic as useful (G2(6) = 124.135, p < .001; C(6) = .244).

In terms of the Usage Survey (see Table 14), the most frequent Web users appeared to explore the Internet the most (G 2(10) = 312.501, p < .001; C(10) = .370). Although a mild interaction, we note that as the preferenc e for using Web browsers for Internet exploration increased, users appeared to prefer text-based search; as preference f or using Web browsers for Internet exploration decreased, users preferred keyword search (G2( 4) = 312.501, p <.001; C(4) = .103).

SURVEY SHORTCOMINGS AND PLANNED IMPROVEMENTS

There were several problem in our survey methods that we plan to correct in future runs. For example, we encountered a t rade-off between keeping the questions concise and removing ambiguity. In the general survey, one question asked the nam e of the domain to which the user belongs. While our intention was to classify the users based upon educational, commerc ial, military, government, etc. categories. it became apparent that the respondents, especially those not from the Unite d States, were unclear as to whether we were inquiring about their domain (i.e., DNS) or their work activity. Also, with our intention of creating easy to use, point-and-click surveys, we opted to use radio buttons instead of text-entry fie lds. While this permitted less user overhead, it constrained some user responses by forcing them to answer "other&q uot; or "somewhat," often times inappropriately. We plan to remedy these problems in future releases by workin g closely with survey experts.

We also encountered difficulties with keeping track of users across surveys. This problem results mostly from the statel ess nature of the HTTP protocol. That is, the server does not keep track of what machine accessed which documents; this is only logged to file. This meant that we were unable to take the answers submitted by a user for the General Survey an d perform analyses against questions in the other surveys. Hence, we could not test for possible interactions like the a ge of the users and user media orientation. There are several ways which we can avoid this issue in the future. One meth od would be to have each user provide a unique key upon starting the first survey, and then use this key for subsequent submissions. Hence, all responses to the surveys could be logged with a unique key for each user. This would enable the cross-question analyses we desire. This method adds additional burden to users by making them responsible for rememberin g the key for future surveys (which may be completed several days later). Another method would extensively utilize the a bility of HTTP servers to execute programs. Under this scheme, a program would produce the desired survey and attach a h idden key with the survey. Upon receiving the completed survey, the program would create another survey and, again, enco de a key within the document. As with the previous method, cross-survey question analysis is enabled, but this time, we impose a potentially severe time constraint on the user, i.e., the time to complete all the surveys. If a user decides t o stop participating in the currently returned survey, any future submissions will require complete re-entry of all surv eys, which is undesirable. Finally, HTTP could be modified to make it into a state machine, though the likelihood of thi s happening is uncertain.

Similar server related problems occur if we accept only fully completed surveys. While we gain from having consistent nu mber of responses to each question in each survey, we would have to develop software that checked each answer, and creat ed an additional document containing the questions that were not answered. The software would also have to keep track of the previous answers and, once the other answers were successfully returned, merge the results into one entry. We will most likely implement this method in future surveys.

In addition, at the time of our survey, few browsers and platforms supported the HTML+ forms extension. Thus, our result s are biased in favor of those that did support Forms (i.e., XMosaic for UNIX). Presently, more platforms and browsers s upport the extension, thus we expect to eliminate the bias with subsequent survey runs.

A final bias is due to a sampling problem, that is, the self-selected nature of survey respondents. Clearly, frequent We b users were more likely to answer our survey, and hence skew the results. We hope that future deployments of the survey will begin to attract responses from less-frequent users, and thereby allow us to characterize the changing Web user po pulation.

Given the dynamic nature of WWW use and technologies, we believe that surveys run twice a year ought to provide an optim al trade-off between maintaining respondents from survey to survey and charting the Web's growth and changes. In additio n, we hope that the WWW community will allow us to remain the sole Web surveyors in this domain. We fear that if other r esearchers clutter the Web with similar surveys, the overall utility of such surveys will be greatly diminished. In ligh t of such a request to the community, we gladly open ourself to suggestions and specific research agendas of other resea rchers.

CONCLUSION

In this paper, we reported results from a survey of World Wide Web users. As shown by the high number of survey responde nts, we demonstrated the viability of using the Web and HTML+ forms for conducting on-line surveys. In addition, our res ults served to begin to describe the range of Web users, their activities, and preferences. In the future, we plan to de ploy our survey every six months. We believe that this will be a useful means for tracking the growth and changes in Web uses and population.

REFERENCES

"A Growing Internet is Trying to Take Care of Business", New York Times, December 12, 1993, page F7.

Berners-Lee, T. (1994) Hypertext Markup Language Plus. URL:http://info.cern.ch/hypertext/WWW/MarkUp/HTMLPlus/htmlplus_1.html

Berners-Lee. T. (1993) Hypertext Transfer Protocol. Internet Engineering Task Fo rce Working Draft, 5 Nov 1993. URL:ftp://ds.internic.net/internet-drafts/draft-ietf-iiir-http-00.txt

Berners-Lee, T. (1994) Uniform Resource Locators. Internet Engineering Task Forc e Working Draft, 21 March 1994. URL:ftp://ds.internic.net/internet-drafts/draft-ietf-uri-url-03.txt

Ragget, D. (1994) Hypertext Markup Language Plus. URL:ftp://ds.internic.net/internet-drafts/draft-raggett-www-html-00.txt

AUTHOR INFORMATION

JAMES PITKOW received his B.A. in Computer Science Applications in Psychology from the University of Colorado Boulder in 1993. He is a graduate student in the College of Computing at Georgia Institute of Technology. His research interests i nclude user modelling, adaptive interfaces, and usability.

MIMI RECKER received her Ph.D. from the University of California, Berkeley, in 1992. She is currently a Research Scienti st in the College of Computing at the Georgia Institute of Technology. Her research interests include cognitive science approaches to learning, metacognition, instruction, interactive learning environments, human-computer interaction, cogni tive modelling and multimedia.

APPENDIX ONE

Table 1: Results from the General Survey 
--------------------------------------------------------------------------------------------
Question 1.       Mac       PC         Unix      VMS       Other                       Total  
Primary Plat      48 /      66 /       1181 /    33 /      7 /                         1335   
form / %          3.60      4.94       88.46     2.47      0.01                               
Question 2.       Cello     Lynx       Mosaic    Other     Samba                       Total  
Primary           1 /       26 /       1294 /    6 /       4 /                         1331   
Browser / %       0.08      1.95       97.22     0.38      0.30                               
Question 3.       No        Yes                                                        Total  
Shared            596 /     736 /                                                      1332   
Machine/ %        44.74     55.26                                                             
Question 4.       Under 5   6 to 10    11 to 20  21 to 30  31 to 40  41 to 50  50 +    Total  
Hrs in Front of   2         40         102       245       295       273       360     1344   
Computer / %      0.15      7.59       7.59      18.23     21.95     2.98      26.79          
Question 6.       Asia      Australia  Europe    North     South                       Total  
                                                 Amer.     Amer.                              
Location / %      10 /      27 /       377 /     910 /     2 /                         1326   
                  0.75      2.04       28.43     68.63     0.15                               
Question 7.       UGrad.    Admini-    Business  Faculty   Grad.     Profess-  Other   Total  
                  Student   stration                       Student   ional                    
Occupation / %    188       38 /       53 /      94 /      287 /     594 /     75 /    1329   
                  14.15     2.86       3.99      7.07      21.60     44.70     5.64           
Question 8.       Maybe     No         Yes                                             Total 
Willing to Pay    727       532        66                                              1325   
Fees / %          54.87     40.15      4.98                                                   
Question 9.       Under 20  21 - 25    26 - 30   31 - 35   36 - 40   41 - 50   50 +    Total 
Age / %           85 /      380 /      362 /     218 /     144 /     103 /     26 /    1344   
                  6.32      28.27      26.93     16.22     10.71     7.66      1.93           
Question 10.      Female    Male                                                       Total  
Gender/ %         67 /      1247 /                                                     1314   
                  5.10      94.90                                                             
--------------------------------------------------------------------------------------------
Table 2: Results from the HTML Survey 
------------------------------------------------------------------
Question 1.       None   1 to 5  6 to 10  11 to 20  Over 20  Total  
Num Docs          52 /   189 /   102 /    86 /      270 /    701    
Authored / %      7.42   26.96   14.55    12.27     38.52           
Question 2.       Yes    No      Unsure                      Total  
Is HTML ISO Stan  21 /   211 /   463 /                       701    
dard / %          3.00   30.10   66.05                              
Question 3.       None   1 to 3  4 to 6   7 to 12   Over 12  Total  
Years of Program  21 /   55 /    145 /    233 /     240 /    701    
ming / %          3.00   7.85    20.68    33.24     34.24           
Question 3.       None   1 to 3  4 to 6   7 to 12   Over 12  Total  
Num Programming   16 /   111 /   292 /    172 /     106 /    701    
Langs / %         2.28   15.83   41.65    24.11     15.12           
Question 12.      Yes    No                                  Total  
Prior SGML Use /  31 /   333 /                               364    
%                 8.52   91.48                                      
------------------------------------------------------------------
Table 3: Results from the HTML Survey 
---------------------------------------------------------------------
Question                                Mean   Std. Dev.  Total  
Question 5. < 1 - 9 >                    2.258  1.377      654    
Hours taken to learn HTML                                              
Question 6. < easy (1) - hard (9) >     2.132  1.393      651    
Overall, learning of HTML                                              
Question 7. < easy (1) - hard (9) >     3.458  1.749      295    
Learning Forms                                                         
Question 8. < easy (1) - hard (9) >     3.781  2.072      201    
Learning ISMAP                                                         
Question 10. < easy (1) - hard (9) >    3.951  2.360      551    
Finding up to the minute HTML documentation                            
Question 11. < easy (1) - hard (9) >    3.064  1.681      626    
Understanding of HTML documentation                                    
---------------------------------------------------------------------
Table 4: Results from the HTTP Survey 
---------------------------------------------------------------------
Question 1.                   Yes         No                    Total  
Operate HTTP Server / %       378 /       103 /                 481    
                              78.59       21.41                        
Question 2.                   GN or Cern  NSCA   Plexus  Other  Total  
If Yes, Which Server / %      28 /        271 /  48 /    30 /   377    
                              7.43        71.88  12.73   7.96          
Question 3.                   80          8000   8001    8080   Total  
Server Port Number / %        279 /       10 /   13 /    16 /   354    
                              78.81       2.82   3.67    4.52          
Question 4.                   Yes         No                    Total  
If no, Able to Add to DB / %  122 /       32 /                  154    
                              79.22       20.78                        
Question 9.                   Yes         No     Some           Total  
Knowledge of Script / %       260 /       66 /   130 /          476    
                              58.82       13.87  27.31                 
Question 10.                  Yes         No     Some           Total  
Knowledge of Modules / %      73 /        341 /  57 /           471    
                              15.50       72.40  12.10                 
Question 11.                  Yes         No     Some           Total  
Knowledge of CGI / %          190 /       191 /  87 /           471    
                              40.60       40.81  18.59                 
Question 12.                  Yes         No                    Total  
Written CGI scripts / %       177 /       285 /                 462    
                              38.31       61.69                        
Question 13.                  Yes         No     Some           Total  
Knowledge of ISMAP/ %         138 /       217 /  111 /          466    
                              29.61       46.57  23.82                 
Question 14.                  Yes         No                    Total  
Used ISMAP scripts / %        128 /       344 /                 472    
                              27.12       72.88                        
---------------------------------------------------------------------
Table 5: Results from the HTTP Survey 
-------------------------------------------------------------------------
Question                                   Mean   Std. Dev.  Total  
Question 5. < horrid (1) - excellent (9) >  5.815  1.854      292    
Overall Rating of server                                                   
Question 6. < horrid (1) - excellent (9) >  6.041  1.958      98     
Overall Rating of GN's server                                              
Question 7. < horrid (1) - excellent (9) >  7.453  1.239      750    
Overall Rating of NCSA's server                                            
Question 8. < horrid (1) - excellent (9) >  6.788  1.686      170    
Overall Rating of Plexus's server                                          
-------------------------------------------------------------------------
Table 6: Results from the Mosaic Survey 
-----------------------------------------------------------
Question 1.           UNIX    PC      Mac              Total  
Which Mosaic Plat     1006 /  48 /    25 /             1079   
form / %              93.23   4.45    2.32                    
Question 4.           Daily   Weekly  Monthly  Never   Total  
Freq of Use of Help   89 /    306 /   326 /    356 /   1077   
Menu / %              8.26    28.41   30.27    33.05          
Question 5.           Yes     No      Some             Total  
Is On-line Help       551 /   74 /    428 /            1053   
Useful / %            52.33   7.03    40.65                   
-----------------------------------------------------------
Table 7: Results from the Mosaic Survey 
-------------------------------------------------------------------------
Question                                   Mean   Std. Dev.  Total  
Question 2. < horrid (1) - excellent (9) >  8.086  0.881      1081   
Overall Rating of Mosaic                                                   
Question 3. < horrid (1) - excellent (9) >  7.458  1.520      334    
Rating of Support Staff @ NCSA                                             
-------------------------------------------------------------------------
Table 8: Results from the Usage Survey 
------------------------------------------------------------------------------
Question                                        Mean   Std. Dev.  Total  
Question 3. < never (1) - always (9) >            5.81  3.39       1145   
Use Browser to Retrieve Documents                                               
Question 4. < never (1) - always (9) >            7.42  3.96       1048   
Use Browser Instead of Gopher                                                   
Question 5. < never (1) - always (9) >            7.60  5.44       1007   
Use Browser Instead of Wais                                                     
Question 6. < never (1) - always (9) >            5.65  8.17       1026   
Use Browser Instead of Archie                                                   
Question 7. < never (1) - always (9) >            4.20  5.15       1116   
Use Browser to View Research                                                    
Question 8. < never (1) - always (9) >            7.30  2.23       1138   
Use Browser to Explore Internet                                                 
Question 9. < never (1) - always (9) >            4.10  6.32       1139   
Use Browser to Retrieve Weather Information                                     
Question 10. < never (1) - always (9) >           4.97  4.41       1127   
Use Browser to Access Reference Materials                                       
Question 11. < never (1) - always (9) >           4.16  4.72       1116   
Use Browser to Access Publishers, Periodicals, etc.                             
Question 12. < never (1) - always (9) >           4.33  5.13       1122   
Use Browser to Access Call for Papers, Proposals, etc.                          
Question 13. < love (1) - hate (9) >              6.62  2.57       1140   
User `s Preference of Text                                                      
Question 14. < love (1) - hate (9) >              7.30  2.31       1140   
User `s Preference of Embedded Images                                           
Question 15. < love (1) - hate (9) >              6.48  3.42       1124   
User `s Preference of Spawned Images                                            
Question 16. < love (1) - hate (9) >              5.41  4.33       1089   
User `s Preference of Sounds                                                    
Question 17. < love (1) - hate (9) >              5.83  4.19       1097   
User `s Preference of Movies                                                    
Question 18. < love (1) - hate (9) >              7.93  1.87       1130   
User `s Preference of Search by Keyword                                         
------------------------------------------------------------------------------
Table 9: Results from the Usage Survey 
-------------------------------------------------------------------------------------
Question 1.            Over 9/  5 to 8/  1 to 4 /   Few times   Once /   Once/  Total  
                       day      day      day        / week      week     month         
Frequency of Browser   235      202      481        199         23       9      1153   
Use / %                20.38    17.52    41.72      17.26       1.99     0.78          
Question 19.           Search   Text     Visual                                 Total  
Search Orientation /   319      432      372                                    1344   
%                                                                                      
Question 20.           Under 5  6 to 10  Over 10                                Total  
Hours Exploring the    484      407      252                                    1153   
Internet / %           41.98    35.30    21.86                                         
-------------------------------------------------------------------------------------

APPENDIX TWO

Table 10: Results of Chi-Square for General Survey 
----------------------------------------------------------------------------------------------
Interaction                              Chi-Square Value  Degrees of   Probability  Cramer V   
                                                           Freedom                              
Machine Shared by Hrs in Front of Cptr   43.834            6            0.000        0.183 *    
Question 3 by 4                                                                                 
Machine Shared by Domain                 139.779           6            0.000        0.324      
Question 3 by 5                                                                                 
Machine Shared by Location               24.233            4            0.000        0.135 *    
Question 3 by 6                                                                                 
Machine Shared by Occupation             199.290           6            0.000        0.388      
Question 3 by 7                                                                                 
Machine Shared by Willing to Pay Fees    30.632            2            0.000        0.152      
Question 3 by 8                                                                                 
Machine Shared by Age                    122.729           6            0.000        0.306 *    
Question 3 by 9                                                                                 
Hrs in Front of Computer by Occupation   169.460           36           0.000        0.147 **   
Question 4 by 8                                                                                 
Hours in Front of Computer by Fees       23.295            12           0.000        0.094 **   
Question 4 by 8                                                                                 
Hours in Front of Computer by Age        90.139            30           0.000        0.118      
Question 4 by 9                                                                                 
Occupation by Fees                       52.816            12           0.000        0.141      
Question 7 by 8                                                                                 
Occupation by Age                        835.564           36           0.000        0.326 **   
Question 7 by 9                                                                                 
Willing to Pay Fees by Age               26.628            12           0.009        0.101 *    
Question 8 by 9                                                                                 
----------------------------------------------------------------------------------------------
NOTES: Since there were only two responses for the under 15 category in Question 9 (Age), the responses were not include d into the Chi-Square calculations.

* Denotes a negative Kendall Tau-B result

** WARNING: More than one-fifth of the cells were sparse (Frequency < 5) significance suspect

Table 11: Results of Chi-Square for HTML Survey 
-------------------------------------------------------------------------------------------
Interaction                            Chi-Square Value  Degrees of   Probability  Cramer V  
                                                         Freedom                             
Documents Authored by ISO Standard     61.080            8            0.000        0.210     
Question 1 by 2                                                                              
Documents Authored by Years Program.   30.766            16           0.014        0.105 *   
Question 1 by 3                                                                              
Documents Authored by Languages        30.766            16           0.014        0.105 *   
Question 1 by 4                                                                              
Standard by Years Programming          19.541            8            0.012        0.119     
Question 2 by 3                                                                              
Standard by Languages                  19.722            8            0.011        0.119     
Question 2 by 4                                                                              
Years Programming by Languages         733.663           16           0.000        0.514     
Question 3 by 4                                                                              
-------------------------------------------------------------------------------------------
* Denotes a negative Kendall Tau-B result

Table 12: Results of Chi-Square for HTTP Survey 
--------------------------------------------------------------------------------------------
Interaction                             Chi-Square Value  Degrees of   Probability  Cramer V  
                                                          Freedom                             
Operate by Access to WWW Database       27.185            1            0.000        0.388     
Question 1 by 4                                                                               
Operate by Knowledge of Server Script   69.383            2            0.000        0.383     
Question 1 by 9                                                                               
Operate by Knowledge of Module          7.643             2            0.22         0.128     
Question 1 by 10                                                                            
Operate by Knowledge of CGI Script      62.132            2            0.000        0.365     
Question 1 by 11                                                                              
Operate by CGI Script Use               50.759            1            0.000        0.316     
Question 1 by 12                                                                              
Operate by Knowledge of ISMAP           44.283            2            0.000        0.309     
Question 1 by 13                                                                              
Operate by ISMAP Use                    24.838            1            0.000        0.224     
Question 1 by 14                                                                              
Server by Knowledge of Server Script    41.596            6            0.000        0.236     
Question 2 by 9                                                                              
Server by Knowledge of Module           145.472           6            0.000        0.444     
Question 2 by 10                                                                              
Server by Knowledge of CGI Script       31.442            6            0.000        0.206     
Question 2 by 11                                                                              
Server by CGI Script Use                23.121            3            0.000        0.252     
Question 2 by 12                                                                              
Server by Knowledge of ISMAP            10.919            6            0.091        not significant
Question 2 by 13                                                                        
Server by ISMAP Use                     8.859             3            0.031        0.155     
Question 2 by 14                                                                              
Server Script by Knowledge of Module    53.810            4            0.000        0.239     
Question 9 by 10                                                                              
Server Script by CGI Script             150.649           4            0.000        0.401     
Question 9 by 11                                                                              
Server Script by CGI Script Use         112.718           2            0.000        0.494     
Question 9 by 12                                                                              
Server Script by ISMAP                  97.941            4            0.000        0.325     
Question 9 by 13                                                                              
Server Script by ISMAP Use              54.503            2            0.000        0.340     
Question 9 by 14                                                                              
Know. of Module by CGI Script           47.495            4            0.000        0.226     
Question 10 by 11                                                                             
Know. of Module by CGI Script Use       31.997            2            0.000        0.264     
Question 10 by 12                                                                             
Know. of Module by ISMAP                46.291            4            0.000        0.224     
Question 10 by 13                                                                             
Know. of Module by ISMAP Use            24.985            2            0.000        0.231     
Question 10 by 14                                                                             
Know. of CGI by CGI Script Use          155.007           2            0.000        0.582     
Question 11 by 12                                                                             
Know. of CGI by ISMAP                   72.535            4            0.000        0.281     
Question 11 by 13                                                                             
Know. of CGI by ISMAP Use               40.937            2            0.000        0.296     
Question 11 by 14                                                                             
CGI Use by ISMAP                        71.809            2            0.000        0.398     
Question 12 by 13                                                                             
CGI Use by ISMAP Use                    63.903            2            0.000        0.350     
Question 12 by 14                                                                             
ISMAP by ISMAP Use                      277.407           2            0.000        0.774     
Question 13 by 14                                                                             
--------------------------------------------------------------------------------------------
Table 13: Results of Chi-Square for Mosaic Survey 
------------------------------------------------------------------------------------------
Interaction                           Chi-Square Value  Degrees of   Probability  Cramer V  
                                                        Freedom                             
Use of Help by On-Line Docs Helpful   124.135           6            0.000        0.244     
Question 4 by 5                                                                           
                                                                                            
------------------------------------------------------------------------------------------
Table 14: Results of Chi-Square for Usage Survey 
------------------------------------------------------------------------------------------
Interaction                           Chi-Square Value  Degrees of   Probability  Cramer V  
                                                        Freedom                             
Frequency of Use by Orientation       21.060            10           0.031        0.097     
Question 1 by 19                                                                            
Frequency of Use by Hours Exploring   312.501           10           0.000        0.370     
Question 1 by 20                                                                            
Orientation by Hours Exploring        23.518            4            0.000        0.103     
Question 19 by 20                                                                           
                                                                                            
------------------------------------------------------------------------------------------

Footnotes

(1)
The actual questionnaires in HTML+ form are available via: http://www.gatech.edu/pitkow/survey/survey-1-1994/index.html. Postscript and ASCII versions are accessible via anonymous ftp from: ftp.cc.gatech.edu in /pub/gvu/www/pitkow/survey/survey-1-1994. All the comments collected from the surveys are accessible via: http://www.gatech.edu/pitkow/survey/survey-1-1994/comments/.
(2)
Specifically, 29 responses from the General Survey, 5 from the HTML Survey, 27 from the HTTP Survey, 11 from the Mosaic Survey, and 4 from the WWW Browsers Usage Survey were removed from further processing.