.l . 111.1111 ‘11.. . .I“. 1. 1 . 1 '1.» 9.1. 1 I x 1-1 or .11 : : .1 51.35....utw1un.” .436 f. 1 t 3.5.1.. ;. 4 A \ 1. i v 4 B gin-.9.“ 1 J WRWWWL‘Hquu :1 14.13331... . .. . . 91.1.... “R. .4. . a: #9141. .131... . . .1.. .. . ‘1 ~ I 1 . ’ I 11. I ‘ kw .. . m .. . . 1.. an 4.! «mummn‘W WM - Ava: 00% 1b)! 1H1 3.0411145 \{I .. 1... ...... . t 111 111 .31; .L.11!-1....111.h1.1111 1......1 .11“ 1111111nrd111| an..fl.-..Iu VIJ‘E"! .1 \l.-h111‘|:.l‘r 114.14 -UVIL‘NV11111 ‘1... . .. .Wv...»-B\.11|11.11- £401.:- . . 11111LW. 1.111113111 1““; 1.1.11 111111.111..I111un. -1 1.1. 1.1.1.11“. 1. 1hfll1‘11: 1111131 .1111...” 1 .2 1.11 .1 1 ~11.111.111ll11..|11.1 .1 11 £3311 Q 6 .1‘1‘11‘111111 in}. 1 . . . . . . I I a» .1 : 1|! I ‘1 - I11 I. .11 1 11.11.- 9 JD! .1 111' II.‘ 1 . . . 1.. an (11.1. 1 1. I 1 .1 1 . . .. 2 - . .. 11 1a...“ .111 111 11. 1 . 1 0| ~ 4 Q. 1 .l ;.l.o...1111 . 1 - . . . - 1 .11 111 on. .0 1. c. ..1. 1. .1. C .1. .. ‘ I. .1. 1‘ 11. 1 910.1, 411‘. 1 . 1 x 1 31.1. . .1. m. P)1>51vdl¢lu\ -. .- . r ,1 .. $1 . -CG 1..- mm... ..... R 11.21. 1 .1 . tar-n1 .inL. .Tvnius Laura 1: WW .. s. 1R I {W1 .0 I! .10 v 1 IN! “HIS A. clan-3 - ¢ 1. . 1 1111.1 “13.1.“. .151 . 5.11.151”; 1.11.“. 1.13. w. 1. 1.1.11 .1 8.? 111.1H1J1wtumr .. - .1 411.1%..- - 1-11.11. 1+ . n V l I . 11% fiafiiflufihfi 1.. . a , )0. 4' v ah §1Htl I V w ,L-xo .8! 2.. nfi» v.11.» .1 . 1 o Wu)‘l1 - 1. .492"? .L . 1 ..:. . . 1 . . 1111.51.11”.- .Kflmunggéwimp 1». 1511...?“ .L...|. 11.. Mung . . . .u..|l.111.uuw1fifi.:-WW.TJMVW~_ 111; 1 . 11? ... ‘5 ‘51-.\ ‘ 14 . .u-zavx1....-.W1J$1u(r1. r1 jinn... . .r. . .Lfinoflfiéthhfl-ui} .1 . «1101.1114-vfi11u15 a . . 1 J 1 . - . .. 1-H afilfl. . . 1- . . ... 41rd. .1111. 43.1.1411? . .. i 1-... . .1141 fihuflnhah€unav LN“. . . . aw. . 1.- .. 1 . .. . . 1.18.1... hen-“1% 1 . . . . “$mequ . - 1....... . 15L. . . - . . 1.11.1.1 :1 -1. . . .. . 11.1 ‘11..)Th-l11bk3111hQ . 1115‘}. .\ 11!. ‘1 I”... ”go-h: 11 o . v ,. . w . .1 . . .1. . . u. . . . 1 1 .1 111 nV‘nflL Ik‘vfi- 51 11 -o.. .1.1. 2.111. .1 . 1 . 1 - o t l .. I 1 1 - 1 1| 1 1 “ 1n VII "'1“. v . .11 . u. .1. 1. .1 .1... -. 0|..101AO 1..”1 . .Jmmh‘t. mafia-«H .ruuhuwmfl fl. 1 .. - . .. ..., -.- .. 111 .wLMnmrI-ul-un 11 1141.11. 111 1 1M“? ,4 11 to.» 1 . 1.11 .2-1111'11 .1...“ 1‘1“. . .. . . : .. . . . .. - _ ..11|11111.1411|1-1 h- i, '1. Ign . I '11. - I '1‘ ‘ - H 11 1-11 1:1 11' 111 1-111 11.311 111 111' 11 1. 1 .1111 1. l .0 n V 1|: I In" 1 . 1 1 . I l 0 ‘1.” “A. 41 11 .0? .11 .11 1.117 1111' . . . n-fl-MNDnl-k- . I .2 11L1Q1N11111Hr1111- «I. . . 1 11111111 11111111111 filfi 41111“! 7. 1 1.1)).- 11-.. v 11. 11 1101-1 31‘ n it?! 7111519. 1101111111111I 11 1 .I‘ 1 II n "In ,1 .1\ 11111.11 1 (l1 1.1.11.1 1.1 1111.811 ,1. 111111? - 421114111kfl: .151111. .11 £1.91! 1 .- .. 1.11.1111 .1111 . . 1 1. 1 1.1.1. 1. 11 1 1111A £11 61. 1.113.710 .1111: .1. .f.-111..111 111,111.11. 11.51511 .1W1h111111.“ 1.1113 11.1111. 1.. all 11......1- .11Nu1111u1l ..-.fl1.1.u.lh....: .1411 111111111 11111 i“; 11111111131114 1111h-1nzl-DV11 1111111...11411..1¢111 1%)11hcxdopuol $.- 1 1...: .......11111.11.1M1|...1..1111IIIIHUI11 1 .. 1111141... 1.3111111111111111.111h1111 11.1011111154111-Kasu .. 11111-{114111ww11qii .-.11 II! 11111111. 11111111- . 9‘ .1 .1 1 1 11.1v111R1za111‘11111l1... ) 111.... .11113-111191540. 111 1112-1111 . . .1 - 1|- ..1I111. . 111111.11 - 11:1.E1R11 1 1.11111... 11.1 1......l1.. .1...» .dnfikn1-113n1K-111L . . -11 .11.|111111111111 11 ”MW 1 1.11 1 111110.111: 1 1. .1. .111 1 11..h4§hu.-1_:.1111 -.. ...... 1.... .1110-“31 1.1.11IW11111 .1. .1 1H 1 1 .11 1|. 1‘1. 1 lvw 11‘ 011211111111I‘.fi¢1..13. $1.1 . .1 .. 1 . . .1 £10111. 1 111u|||||1 “I; a 1%.: . I: 1911’)! .011.111I..1.}L 1.. . .1 .1. 11111111 111 “1h“: 11111.1 11% 1 .1 1 1L - 111111 ..-. 1,1.I1-1 .1... 1 {11111.11 . . 1c .. . 111211.11 . 11 . 1- 1. Illa.“ 11,. 1 . .111 91.111. . 11111. . l.-I....I111. 11111111111- 1... ~D11.1 1. 1 I 111nuh1|1 .1111 11.11311 11.111 v1 1 1 11.1.1111- (I111 1 | 1 31.11111 1W11 “1111,11..1l1 . 1.1m.“ 1.1111 1111 111.1 .‘011 1.1111 .11..11' ('1 1" 11-.uflcL- .1 .. 11.1.11. 11.11.111.111 FEM-.1 11 1111 1.111111 .1. ...-.11..H1..111 .11 -HH1 “1 -.11.HH.. f1. .1-1.|111.1 ”In .61 1) 11 1W1 1| 1 -1111-1 111. 111 111.11 1... d1 UJJ!”11.I1J. .11.!111. 1 1.111 .2 -1 111111 1111 . : . . .1 1.11 III . L. n- 11111314114 1|. 1111111111111 1.1.1.1311] 1|.I'I1 .1111||11'|11 1 1 1”» I 111111.1111' ..:N.1P..-1. 11111.1 ”.111. ..1 .11...111-1-11 1.1 1111111111.} 1. 1. 111.11 H.11-t11 1.. 011-.I11... 1 .1. In [.11 JAG-1.11.11. -11.- - . 1. -.1 -1 Mm... 114M .1" 11 1H1 . . 111111111 11.11 1.1. I llfli‘n 11 1 - XII“. - . {H.111 ..1 1. 1.111.011.“ . .. 11111-1. 1 1111 11.1.11 11, ..11-1-.1..1.1..-1o-.111H11..-1.1111“11a ..11.I.11.w1!1 1. 11-11. 111, 1 111 11.1 11 . 111 - 11! 2.1 1.11- .1111 . ,.1 11.11111 1.1111. MI: 1 .1 - “3.111111 1.11 1. 1. 1 . 111111 1111 11nd 1.111111 11“ V . I. 11 1 1-1“ I. 1. 1.1....- 1 . | I)!!! \'M . 1 1h 1l1 .14..111\.l . . 1- .111. “111111 .1411- .» .11. . llll‘ 1 - 1 - H111 1 . 1 1111 1-1111 I1 111” 11 ‘11. 1111111111- 1111 . 11 1111 :11 l 11' 1 1 1 11' \ 1 11.11"« “1 111,511 11.2 J '0 Ito-11.1.5. 11 1. .1111. 2. . 1.....1fi1 THESIS r50: 5'“ ~4,‘ . ‘ r‘ ~ .Q”- : I-b..“~;‘ '- . «$3.33 : l . ' f' ' ' .' ' '. .. ..-- 8'“ ‘. .1 . _, t_w,.1,,.; J 3 but“ ‘ ‘4‘ t» w u) «9‘ This is to certify that the dissertation entitled the effect of locus of control on performance with man-machine dialogues presented by Richard L. Hartley has been accepted towards fulfillment of the requirements for PhD . degree in Education A, / I 1/1; 4, ML— / ‘1 Major professor Date May 18, 1983 MSU is an Affirmative Action/Equal Opportunity Institution 0- 12771 MSU LIBRARIES .-_— RETURNING MATERIALS: Riace in book drop to remove this checkout from your record. FINES will be charged if book is returned after the date stamped below. ROOM USE 033M -' K n. ‘4,- PE ill“ 5’ .3 / .01) over the explicit transfer of control (test and jump). However, syntactic errors were significantly few- er for the test and jump group. Sime, et al., also mea- sured the number of additional attempts after an initial error and found that the 'error lifetimes' were greater for the test and jump. This study gives ample evidence of the superiority of nested conditional statements over the use of the GOTO to jump over program statements. In the 1977 study, Sime, Green, and Guest included an- other notation termed the repeated predicate form. By embedding the actions within the conditional statements semantic and syntactic errors were both reduced signif- icantly (p < .01, .05 respectively) and the error lifetime was .09 compared to 1.06 for the jump and test and 1.60 for the nested notation (p < .02). Unfortunately, the statement 22 formats chosen differed significantly between the nested and the repeated predicate. In the nested each action was preced- ed by the word BEGIN and terminated by the word END. In the repeated predicate BEGIN and END were not used. This differ- ence in syntax alone could have caused significant differ- ences in results (Shneiderman, 1981). Practice Effect. Several studies (Green, 1977; Sime, 1973; Shneiderman, 1976; Winer, 1971) have shown large prac- tice effects, that is, marked improvement in performance with each session. Many times the practice effect was so great as to overshadow the effects of the experiment being measured. In Sheil's review (1981) of Greeley's 1977 compar- ison of structured and unstructured languages, Sheil states “whereas the effect of the different languages is to change the mean reaction time by amounts which range from 4 to 15 percent, the effect of a single session of practice ranges from 13 to 27 percent" (p. 106). Whenever conducting an experiment with experienced and inexperienced subjects one should be cognizant of such prac- tice effects. The research of structure versus unstructured program- ming, aside from the conditional, has yielded mixed results. The results obtained do not adequately support the benefit of structured programming. This is not to say that such a benefit does not exist. Sheils states "the evidence sug- gests only that deliberately chaotic control structure de- grades performance”. One reason cited for inclusive results 23 is that structured programming is "a discipline, a way of thinking" (Sheil, 1981, p. 107) and hence is simply not easy to simulate in an experimental situation. Another problem with experiments in structured programming is that simply prohibiting "the use of a program feature without providing any motivation or alternative strategies . . . is hardly likely to produce much else other than resentment” (Sheil, 1981, p. 107). In his summary of research on programming notation Sheils states, "given the small sizes of and inconsistencies among the reported effects, it is not even clear that nota- tion is a major factor in the difficulty of programming . . . it comes as a shock how little empirical evidence there is for their importance . . . many of these effects tend to disappear with practice or experience. This raises some doubt as to whether these results reflect stable differences between notations or merely learning effects and other transients that would not be significant factors in actual programming performance” (1981, p. 108). Flowcharting. There is a lack of evidence as to the value of flowcharting prior to writing programs. Shneiderman conducted ”five successive experiments, with dif- ferent tasks and measures (and) failed to reveal any reli- able advantage of flowchart use” (Sheil, 1981, p. 109). However, the experiments themselves may have caused the in- conclusive results. Two of the experiments suffered from 'ceiling effects'. Both the control and the experimental 24 groups achieved scores near the maximum leaving little room for measured differences. In other experiments Sheils points out that the lack of results might be explained by the ”choice of materials, language, and/or participants” (1981, p. 109). Indenting (Prettyprinting). Instructors frequently encourage students to make their programs neat and perhaps even conform to a particular style. For example, in COBOL programming instruction common practices are to limit stu- dents to one statement per line, to require all Data Division entries of the same level to be in the same column, and to have the word PICTURE begin in the same column throughout the Data Division. Within the Procedure Division it is common practice to have IF-THEN-ELSE sentences indent- ed for clarity as illustrated in Figure 2. Sheil noted that research by Weisman (1974) concerning indentation and 'pretty printing' revealed “positive self- evaluations but no performance improvements for either mod- ifying, hand simulating, or answering questions about indented versus unindented versions of programs“ (Sheil, 1974, p.109). Another technique widely recommended in COBOL programming is the use of paragraph numbers. In COBOL a paragraph name (similar to a statement number in other lan- guages) may be composed of letters and/or numbers up to 18 characters in length according to ANSI 74 standards. Most COBOL users tend to create meaningful self-documenting program names. Some COBOL users go a step farther and 25 DATA DIVISION. 01 COUNTERS. 02 MINOR-COUNTER PICTURE 89(5) VALUE 0. 02 SENIOR-COUNTER PICTURE 89(5) VALUE 0. 02 OTHER-COUNTER PICTURE 89(5) VALUE 0. PROCEDURE DIVISION. IF AGE < 18 ADD 1 TO MINOR-COUNTER ELSE IF AGE > 65 ADD 1 TO SENIOR-COUNTER ELSE ADD 1 TO OTHER-COUNTER. Figure 2. IF-THEN-ELSE construct. assign a three or four digit prefix to each paragraph name such that the numbers are in sequence throughout the pro- gram. The premise for the use of the number prefixes is that such paragraph names will be easier to find when de- bugging or modifying someone else's program. No research has been conducted on the benefit of the 'paragraph name with number prefix' method. 26 Comments. It is generally considered a good programming practice to insert comment lines into a program for the purpose of providing documentation to future readers of the program. Studies that have been conducted (Weisman, 1974; Shneiderman, 1977; and Sheppard, Curtis, Milliman, & Love, 1979) have put forth mixed conclusions as to the benefit of comments. Sheppard, et al. found that comments had no effect on the accuracy of the ability to modify programs (1979), while Shneiderman found that programs were easier to modify with high level comments present (1977). The choice of language, the subject matter of the pro- gramming effort, and the clarity of the code itself could have a great effect on the benefit of comments. For exam- ple, in COBOL one can write nearly self-documenting program statements and paragraph names thereby negating the benefit of comments. In some BASIC languages (e.g., CDC BASIC), variable names are restricted to two characters and statements are identified by five digit numbers. Hence, in some BASIC languages, comments might prove very useful. A further point on comments relates to the real-world programmer. It is very common for a programmer to be re- quired to modify a program unfamiliar to him and frequently involving an unfamiliar application. The type of comments useful to the professional programmer might be of an entire- ly different type than would be thought of by the inexperi- enced programmer. For example, in COBOL the reference to a copied table might be commented in such a manner as to 27 define the source of the table and its organization which could reduce the time needed to modify the program by provid- ing a more complete understanding of the data processed. As Sheil states, "it is equally clear that a comment is only useful if it tells the reader something she either does not already know or cannot infer immediately from the code” (1981, p. 111). Variable Naming. There have been several studies of variable naming conventions (Weissman, 1974; Shneiderman, 1980; Sheppard, et al., 1979). These studies have attempted to ascertain whether mnemonic variable names were an aid in debugging programs. The experiments were inconclusive in that groups performed no better at debugging programs with mnemonic names than with programs without mnemonic names. The only improvement cited was in self-evaluation by the participants. It should be noted again that experiments typically involve small programming tasks. In such tasks mnemonic names could well offer little benefit over 'meaningless' names, as the subjects could remember a limited number of variable names without memory aids. Database Query Language "A query language is a special-purpose language for con- structing queries to retrieve information from a database of information stored in the computer” (Reisner, 1981, p. 14). The intent of a query language is to provide a tool for the 28 retrieval of data that is easy to use. The alternative is to rely on professional programmers to develop programs when- ever information is needed by users in a form not already available. Many computer manufacturers and independent software vendors now offer query languages for the many databases available. Each software firm claims ease of use as a major attribute of their product. Martin offers a 'two-day test' (previously discussed) to determine if a product is really user-oriented as opposed to an orientation more favorable to professional programmers (Martin, 1982). Ease-of-Use. The main problem encountered in query language studies has been the measurement of ease-of—use. Reisner offers several tasks that have been used in evaluating query languages (Figure 3). The kinds of tests used to measure ease-of—use vary widely. Some studies use only a single test, others have used several. Reisner summarizes the most common approaches: 1. Final exams of learning. These tests how easy a query language is to learn; they are given at the end of teaching. 2. Immediate comprehension. These help identify why particular learning problems occur. They are given during teaching, immediately after some function has been taught, to determine whether subjects can use the function, given that they know it is the one to use. 3. Reviews. These help identify why particular learning problems occur. They are given during teaching and cover functions taught up Task Query writing Query reading Query interpretation Question Comprehension Memorization Problem solving 29 Description Users are given a question stated in English and required to write a query in the given query language Users are given a query written in the query language and asked to write a translation into English. Users are given a query in the query language and a printed database with data filled in. They are asked to find the data asked for by the query. Users are given an English question and a printed database and are asked to find the data asked for. Users are asked to memorize and reproduce a database. Users are given a problem and a database and are asked to generate questions in English that would solve the problem. The questions should be answerable from the database. Figure 3. Query Language Tasks (Reisner, 1981, p.16). 30 to that time. They require that subjects know which function to use. 4. Productivity. These are tests of query lan- guage use by 'skilled' users. They test how well the language can be used after some pre- determined level of learning has been attained. 5. Retention. These test how easy a query lan- guage is to remember: how well it can be used by people who have been away from it for a period of time. 6. Relearning. These test how easy a query lan- guage is to relearn by users who have been away from it for a period of time and have forgotten some of it (1981, p. 17). Some of the basic differences in query languages are their syntactic form, procedurality, and the underlying data model. Syntactic Form. Syntactic form is the manner in which queries are constructed. Figure 4 illustrates queries in four different query languages to find the names of em- ployees in department 50. While some studies (Greenblatt & Waxman, 1978; Reisner, 1975) have compared two languages of different syntactic form, their conclusions on ease-of—use should not be attrib- uted purely to the syntactic form of the languages. The studies to date have not been sufficiently controlled to un- equivocally state, for example, that a two dimensional syn- tax as used in QBE is easier to use than a linear syntax as used in SQL. 31 Language Example SQL SELECT NAME FROM EMP WHERE DEPTNO = 50 QBE EMP NAME DEPTNO p.Brown 50 SQUARE EMP ('50') NAME DEPTNO TABLET FORM DEPTFIFTY FROM NAME, DEPTNO OF EMP KEEP ROWS WHERE DEPTNO = 50 PRINT NAME Figure 4. Syntactic Form in Four Query Languages. (Reisner 1981, p.14). Procedurality. Following is a review of a study con- ducted by Charles Welty (1979) to determine if difficult que- ries to a data base are written more easily in a procedural or in a non-procedural query language. According to Welty, ". . . a language is procedural if it specifies a step-by-step method for achieving a result. Non-procedural languages describe the desired result without specifying how it is to be achieved. (The idea is compara- ble to the difference between constructive and nonconstruc- tive existence proofs in mathematics)“ (1979, p. 16). Further background on the formation of the definitions is provided by Codd (1971). A procedural query is based on relational algebra. Through relational algebra distinct or- dered steps are defined. "An operation on a relation or re- lations always yields another relation" (Welty, 1979, p.16). 32 A relational calculus query as defined by Codd, "de- scribes the elements of the desired relation. The query is purely descriptive, containing no method for achieving the desired relation“ (Welty, 1979, p. 16). Query languages are meant to be used by persons who are not computer professionals, i.e., professional programmers. The use of such a language is ancillary to the users main work and hence the language must be easy to use or it will not be utilized. While papers have been presented (Codd, 1971; Date, 1977) on the superiority of non-procedural languages over procedural languages, the papers have not been the result of research but reflect the opinions of the authors. Welty hypothesized that ”people more often write diffi- cult queries correctly using a procedural query language than they do using a non-procedural query language“ (1979, p. 15). The languages chosen to test the hypothesis were SQL and TABLET. Both languages use a relational model as op- posed to a network or hierarchical model. Both are relation- ally complete. Both have similar language levels as measured by the Halstead method (Halstead, 1977). And both languages utilize the same terminal equipment. Their differ- ence lies in that SQL is a non-procedural or descriptive language while TABLET is a procedural or constructive lang- uage. The experiment participants were attracted by the offering of a one credit course offered by the Accounting 33 Department of the University of Massachusetts at Amherst. The instructional material for the two languages was prepared by using the SQL training manual and then rewriting the TABLET manual to be similar in format and content. The representation in the 72 participants was primarily business undergraduates. However, the background of the individuals varied considerably as can be seen from Table l. The group tended toward Freshmen and Sophomores which would imply very little business exposure. Hence, being a business undergraduate only indicates a business interest not a business background or exposure through coursework. Nearly half of each experimental group had previous pro- gramming experience. Sixty-five of the sample had calculus while the remaining 17 had at least a pre-calculus back- ground. One might question whether such a mathematical and programming background might be representative of actual us- ers of query languages in a business organization. The fa- miliarity with symbols displays knowledge that might be typ- ical of business school graduates if they were required to take a business statistics course. The subjects were divided into four groups. Two groups learned SQL, two learned TABLET. For each language two groups were identified as experienced in computers (having taken a course in FORTRAN or BASIC) or inexperienced. Instruction was via the prepared manuals. At the class sessions, a question and answer period was conducted and then a quiz administered. Any question asked in one class 34 Table 1 Subjects of SQL and TABLET Study Class SQL Senior 5 Junior 5 Sophomore l4 Freshmen 15 Major Business Admin 17 Accounting 5 Fashion Mktg 1 Marketing 1 Other 11 Computer Experience None 17 BASIC 10 FORTRAN 5 Other 3 Math Background Calculus 28 Pre-Calculus 7 Familiarity with Math Symbols > 34 < 34 = 34 u 25 n 26 TABLET N NO‘U‘I-b ...; ml-‘NQQ 27 10 36 36 27 27 Source - Welty, 1979, p.366 35 was covered in the other classes. The experiment was pre- tested at another college. Success in each language was measured by a paper and pencil final exam and a retention exam given three weeks lat- er. The tests were scored by counting errors. The test re- sponses were graded as essentially correct or incorrect ac- cording to a classification of errors developed by Reisner (1976). As query languages are designed for the casual user, the retention test is most significant. According to Welty the results of the retention test support the hypothesis that subjects using a procedural language (TABLET) would have significantly more correct responses than those using a non-procedural language (SQL). However, the tests were com- prised of “easy" and ”hard” queries which did not show the same results. No difference was shown between the languages for the easy queries. The procedural language (TABLET) showed significant advantage over the non-procedural lan- guage (SQL) for the hard queries. The results were signif- icant at the .05 level. Welty also stated that he found that experienced sub- jects performed better than inexperienced with a procedural language (p > .005) but there was no significant difference when using the non-procedural language. Experience differ- ence was based on programming exposure to either FORTRAN or BASIC. However a complete description of the backgrounds by group was not provided. The difference could be attributed 36 to factors other than programming experience, for example, mathematical background (1979). It is reasonable to conclude from this study that proce- durality in a language does enhance its ease of use and the ease of learning with difficult or hard to write queries. It is also fair to conclude that subjects with a semantic reference structure built into their long term memory will perform better when using a procedural language than those who have not developed such a structure via a programming language. As most business schools offer, and frequently re- quire, an introductory data processing course which requires minimal study of FORTRAN or BASIC, future employees of bus- iness organizations should be somewhat prepared for proced- ural query languages. The background necessary to enhance learning of a non-procedural language is yet to be deter- mined. Data Models. ”In most query languages the user is as- sumed to have a conceptual View of how the data are stored in the computer. Three well known data models are the rela- tional model in which data are assumed to be stored in the form of tables; the hierarchical model, in which data are assumed to be stored in the form of tree structures; and the network model, in which data are assumed to be stored in the form of general graph structures” (Reisner, 1981, p.15). Lochovsky and Tsichrizis conducted studies comparing the three data models (Lochovsky & Tsichritzis, 1977; Lochovsky, 1978). The subjects were computer science and 37 business administration students. They were classified as "more experienced users“ if they had six months or more of programming experience, otherwise they were classified as ”less experienced users“. The students were given instruction in the APL language and then given user manuals and programming problems to study for a week. The subjects were given tests containing query writing tasks. They were then given the task of de- bugging and running the examination queries. Another set of queries was given after the subjects had worked with the on-line system. The results of the experiments clearly showed the rela- tional model to be superior. For the less experienced us- ers, the relational model was significantly better p < .01). Experienced users did significantly better with all three models (p < .05) than the inexperienced users. The exper- ienced user, however, did better with the relational model only before their on-line experience. Two problems exist with these studies. First, the mod- els used were not products in commercial use but were devel- oped for the purpose of the experiment, hence the results can only suggest that other software products incorporating these models might yield the same results. Second, as Lochovsky points out, 'it is difficult to attri- bute the differences in performance to either the data model or the data language, since they were not separated" (Lochovsky, 1978, p. 22). 38 Brosey and Shneiderman (1978) conducted a study to de- termine if the data model alone caused differences in ease- of-use. Their study included only the relational and hier- archical models. The subjects were undergraduate students grouped by experience in programming. The beginners had two or three terms of programming. The advanced group had six terms of programming. The researchers used a question compre- hension, a memorization, and a problem-solving task. On the comprehension task the hierarchical model was easier for be- ginners to use (significant at p < .05) but not for the ad- vanced group. Schneiderman, in a review of his own experi- ment concludes that ”although the relational model is . . . possibly a convenient notation in general, there exist cir- cumstances in which the tree model is easier to use" (1980, p. 167). Reisner offers a word of caution in using the tasks of question comprehension and memorization. These tasks are less related to real-world tasks than are query writing tasks (1981). Other studies of how people organize data show that peo— ple do have structures like the data models previously dis- cussed. Reisner noted that in the studies by Durding, Becker, and Gould subjects “were able to organize words into these structures based on the semantics of words, and had difficulty on a task that required them to use words in inap- propriate structures” (Reisner, 1981, p. 22). Broadbent and Broadbent (1978) studied database 39 structures in a non-database query mode and found individual differences dependent upon educational background. Affective Dimensions of Man-Machine Interfaces This section focuses on general problems in man-machine interfaces, particularly those in the affective domain of psychology. The discussion excludes hardware factors such as keyboard or video display design, as well as excluding software topics such as menu selection or command languages. Attitude and anxiety. Users attitudes can dramatical- ly affect their performance. According to studies by Walter and O'Neil “novices with negative attitudes towards compu- ters learned editing tasks more slowly and made more er- rors”. They also suggest that anxiety (fear of failure) "may reduce short-term memory capacity and inhibit perfor- mance" (Shneiderman, 1979, p. 225). Anxiety can be caused by the unknown as in a timesharing environment. The novice user lacks complete knowledge and hence may fear the loss of files (invisible to the user in some remote location) or pos- sibly concern for destroying the computer system which he is attempting to interact with. Shneiderman suggests that "every attempt should be made to make the user at ease without being patronizing or too ob- vious” and that "the user will feel best if the instructions are lucid, in familiar terms and easy to follow". ”Diagnos- tic messages should be understandable, non-threatening and low-key . . . avoid meaningless, condemning messages such 40 as "SYNTAX ERROR" where a constructive informative message can be displayed such as 'UNMATCHED RIGHT PARENTHESIS'. He suggests that "constructive messages and positive reinforcement produce faster learning and increase user acceptance” (Shneiderman, 1980, p. 226). Control. Individuals may be classified by their de- sire to control or be controlled (external versus internal locus of control) (Rotter, 1966). Shneiderman asserts that individuals desire to be in control, and that ”with re- spect to computers, the desire for control apparently in- creases with experience” (Shneiderman, 1980, p. 226). This is an untested hypothesis. Users may resent messages which imply that the computer is in charge, for example, the au- thoritarian phrase "ENTER NEXT COMMAND” compared to the ser- vile ”READY FOR NEXT COMMAND”. The previous example is representative of changes made by the Library of Congress in their interactive systems. The Equitable Life Assurance Society has the following set of guidelines for developing interactive systems. ”Nothing can contribute more to satisfactory system performance that the conviction on the part of the terminal operators that they are in control of the system and not the system in control of them. Equally, nothing can be more damaging to satisfactory system oper- ation, regardless of how well all other as- pects of the implementation have been handled, than the operator's conviction that the terminal and thus the system are in con- trol, have 'a mind of their own', or are tug- ging against the operator's wishes”. (Shneiderman, 1980, p. 227) 41 Word processing systems have gained widespread accep- tance perhaps in large part because they give users a sense of control. Word processors are basically micro or mini com- puter systems with their own disk drives and printers. Compared to a time sharing system with remote (and invisible to the user) storage facilities and frequently their print- ers. Shneiderman perceives that users have a greater sense of control with the micro and mini systems and, hence, greater satisfaction for similar tasks. Shneiderman does not speak to the issue of whether us- ers, in fact, wish to be in control. His assumption is that all users will desire to be in control and will wish greater control as they gain experience with a given system. While the desire to be in control may be considered an admirable, or at least a desirable quality, in users there is ample ev- idence that individuals do differ in their desire to control (Shneiderman, 1980). A problem with the casual observation of human behavior is that it is easy to accept the commonplace as the norm for behavior. Certainly, many users of computer systems may ver- balize their desire for more control; for example, the de- sire for shortcuts, as they gain experience. But what of the individual who remains frustrated and anxious even after considerable exposure to computer systems? Perhaps some peo- ple should be served by man-machine dialogues that meet their need to be directed or guided. Closure. Closure is the completion of a task leading 42 to relief, in essence when our limited short-term memory is relieved of information that is no longer needed (Shneiderman, 1980). Such relief is often experienced while working with text editors. As one modifies a program or text with an editor there can be considerable anxiety up un- til the editing session ends with an EXIT or SAVE command. The implication here is that users might be more comfortable with multiple small tasks than with larger singular tasks. Some software vendors now offer the ability to save modified text before ending the editing session, as is the case with XEDIT distributed by the Control Data Corporation (1981). The only research on closure in man-machine dialogues is that on text editors; and even that research offers only con- jectures about closure as observed in experiments on other factors. Response time. Response time is the moment from which a user submits a command or request until the moment the on-line computer system responds (Dock & Essick, 1981). It is not uncommon to hear computer system users ask for faster response time when working with interactive systems. Shneiderman suggests that "a more informed view is that the acceptable response time is a function of the command type” (Shneiderman, 1980, p. 232). That is, certain commands such as those for response light pens, should be very quick while the response to seemingly complex queries should be longer. Miller has shown however that variability in response time actually causes poorer performance and lower user 43 satisfaction (1968). Users may well prefer a system with a consistent three second response time to one which varies from one to five seconds, even though the average might be the same (Shneiderman, 1979). Shneiderman conjectures that "by eliminating the variance in response time, service is perceived to be more reliable and one source of anxiety can be reduced" (1980, p. 232). In one experimental study involving the modification of five parameters with light pens the findings showed that the subjects performance improved as response time improved (Goodman & Spence, 1978). However, in another study of sub- jects performing calculations on numeric arrays, subject per- formance increased as response time was slowed (Grossberg, Wiesen, & Yntema, 1976). In this latter case, subjects changed their work habits and became more cautious. The sub- jects, however, took fewer steps to solution of the problems and frequently completed the tasks in the same time. There seems to be evidence that users work at the speed of the computer system or at least attempt to. In some cases the decision-making time is short, as in data entry, and the response time needs to be as fast as the user. In other cases where the decision making time might be longer, a fast response time could cause anxiety and mistakes. Possibly the combination of proneness to anxiety and inter- nal versus external locus of control could be related to per- formance of tasks under various response times. If, in fact, people function better with different response times, 44 perhaps computer systems (like many electronic games) should have variable response times either selected by the user or assigned by the computer system according to the number of errors the system detects. Time-sharing versus Batch Processing. Time-sharing is "the term used to describe a central processing unit that is shared by several users, usually with the use of termi- nals“ (Stern & Stern, 1982, p. 627). Time-sharing environ- ments are on-line environments, that is the ”utilization of data processing equipment that is directly under the control of the main central processing unit” (Stern & Stern, 1982, p. 624). On-line implies a users ability to interact direct- ly with a computer. Batch processing is ”the processing of data in groups or batches, as opposed to the immediate processing of data" (Stern & Stern, 1982, p. 617). Batch processing can be char- acterized in an academic setting by students keypunching pro- grams, submitting the programs to a computer center, and then picking up the output after some period of time. The period of time from submission of a job to receipt of output is termed turnaround time. Thrnaround time might range from a few minutes to a few days. As time-sharing systems came into active use and users were given a choice between batch and time-sharing several studies appeared on the subject (Schatzoff, Tsao, & Wiig, 1967; Gold, 1969; Sackman, 1970a; Sackman, 1970b; Boillot, 1974; Hansen, 1976). One viewpoint was that waiting for 45 batch output was "annoying, disruptive, and time-consuming". Another viewpoint is that time-sharing "encouraged sloppy and hasty programming, which in turn led to more errors and poorer quality work" (Shneiderman, 1980, p. 232). Studies by Schatzoff, Tsao and Wiig (1967) and by Gold (1969) indicated a higher cost for time-sharing (50 percent increase) and a greater elapsed time for batch jobs (50 per— cent longer) with no difference in total computer time used. More compilations of programs were observed in the time-sharing mode which might indicate that individuals were not as thorough in checking their work before compiling their programs. Gold states that ”the user's attitude appears to be one of the variables which may influence the user's immediate be- havior and usage of computer systems" (1969, p. 255). In a review of time-sharing versus batch processing ex- periments, Shneiderman summarizes, 'In all the experimental results, the influence of individual differences apparently played a major role. The high variance in performance and conflicting anecdotal evidence suggests that unmeasured fac- tors such as personality may influence preference and perfor- mance" (Shneiderman, 1980, p. 233-4). Lee and Shneiderman conducted studies into locus of con- trol and assertiveness regarding batch versus time-sharing preference (1978). “Locus of control focuses on the percep- tion individuals have of their influence over events. Internally controlled individuals perceive an event as 46 contingent upon their own action, whereas externally con- trolled people perceive a reinforcement for an action as being more a result of luck, chance, or fate; under the con- trol of other powerful people; or unpredictable” (Rotter, 1966, p. 1). Assertive behavior "allows an individual expression in a manner that fully communicates his personal desires with- out infringing upon the right of others". (Winship & Kelly, 1976, p. 215). Weinburg conjectures that "humble program- mers perform better in batch environments and assertive ones will be more likely to shine on-line'. (Weinberg, 1971, p. 235). Subjects in the study by Lee and Shneiderman (1978) were professional programmers who had available and worked in both a batch environment and a time-sharing environment utilizing Control Data Corporation equipment. The subjects completed questionnaires to ascertain their assertiveness, locus of control, and preference for batch or time-sharing. The groups by preference did not differ on either personal- ity dimension; but, when grouped by internal locus/high as- sertive and external locus/low assertive, there were signif- icant differences in mean preference scores. Shneiderman suggests further study with a wider variety of programming environments. Text Editor Usage. A text editor is a software prod- uct that enables users to add, delete, or change lines of text. The text is typically a program but could, also, be 47 data or a written document. Text editors have a command lan- guage including instructions to move forward and backward in a file of text and providing an ability to print lines of the text and move text lines from one location to another within the text file. Many of the characteristics of text editors are found in word processing packages. Walther and O'Neil (1974) conducted an experiment using both an inflexible text editor and a flexible text editor. The flexible version, according to Shneiderman, “permitted abbreviations, default values, user declaration of synonyms, a variety of delimiters, and other features“ (Shneiderman, 1980, p. 236). Other variables included attitude towards computers and anxiety, experience with on-line systems, and type of terminal (cathode ray tube versus hardcopy termi- nal). The subjects were evaluated in terms of errors made and time to completion of a task. Experienced users worked faster with the flexible version, but inexperienced us- ers were overwhelmed by the flexible ver- sion . . . inexperienced users made fewer errors and worked faster with the inflexible version ... hardcopy terminal users worked faster and made fewer errors suggesting that the feedback from the hardcopy terminal may facilitate performance . . . those with negative attitudes made more errors. (Shneiderman, 1980, p. 236). Sondheimer (1979) conducted experiments with profession- al programmers focusing on five features chosen for addition to an existing text editor. Sondheimer concluded that the 48 results of the experiment seemed to indicate the persistence of individual usage habits. The implication from this exper- iment is that text editing is a skill which once learned is difficult to change. Experiments in the use of text editors by highly trained individuals have been conducted by Card, Moran and Newell (1978). Their experiments were restricted in scope to present a cognitive model based on 'goals, operators, methods, and selection rules'. The model is meant to repre- sent the performance of expert users. The experimental re- sults cannot be generalized beyond the scope of the experi- ment. That is, the "user must be an expert, the task must be a routine unit task; the method must be specified in de- tail; and the performance must be error-free". The results indicate a speed advantage of display editors over line edi- tors and that there are ”speed and accuracy advantages of a mouse for selection text, when compared with a joystick, step keys, or text keys” (Shneiderman, 1980, p. 238). As word-processing systems closely approximate text ed- itors it should be a worthwhile endeavor to replicate the previously discussed experiments with current word-process- ing software. It should be noted that these experiments did not explore characteristics of the subjects (cognitive or af- fective dimensions) but only classified subjects as expert users. Shneiderman offers the following guidelines in develop- ing on-line systems; they may well be considered in 49 experimental designs for research involving such systems, - do not violate the bounds of human performance imposed by short term memory capacity - design interactions in a modular fashion to facilitate closure - be sensitive to user anxiety and desire for con- trol - provide novice users with a sense of accomp- lishment, but avoid patronizing comments - consider response time requirements - accept the personality and cognitive style diff- erences among individuals and do not attempt to make everyone behave as you do - make error messages constructive and give guid- ance for using the system in a courteous non- threatening way - give users control over what kind of and how much information they wish at every point in the interaction - have HELP facilities available for every command (Shneiderman, 1979, p. 243). CHAPTER III RESEARCH DESIGN AND METHODOLOGY This chapter presents the design of the study including the sample, the instruments used, the treatment applied, the hypotheses, and the statistical analysis used. The study was conducted over a two month period. During the first month two dialogues were administered to students of one instructor, David Wilson, hereafter referred to as Instructor A. During the second month the dialogues were administered to the students of Richard Hartley, hereaf- ter referred to as Instructor B. Each instructor had two sections of the same course, ”Introduction to Data Processing”. The first section of each instructor was given Dialogue I (the system-directed software) first. The second section of each instructor was given Dialogue II (the user- directed software) first. The Rotter Locus of Control in- strument was administered prior to the dialogue treatments to determine whether each individual had an Internal Locus of Control or an External Locus of Control. Sample The sample for the study included undergraduate stu- dents enrolled in sections of ISA 221, "Introduction to Data Processing” at Central Michigan University. This course is required of all students on the Business Administration 50 51 curriculum and is generally the student's first academic exposure to computer systems. The course is open to students with at least sophomore standing. All students had instruction on the IBM PC and an as- signment to write a BASIC program, hence all students had the same level of exposure to the equipment used in the study. (Knowledge of the BASIC language was not necessary but familiarity with the operation of the IBM configuration was relevant.) Students are assigned to specific sections of ISA 221 by a computer assisted scheduling system. Each student indi- cates their course preferences through a pre-registration process. The course requests are processed in an order de- termined by the student's classification (freshman, soph- omore, etc.) and number of earned credits. The scheduling program is designed to balance the number of students in each section of a multi-section course. Hence, there is no predetermination by the students as to the section in which they will be placed. The course ”Introduction to Data Processing“ is 16 weeks long and covers three primary topics: hardware, sys- tems analysis, and programming. The students are viewed as end users of computer systems as opposed to computer professionals. This study utilized an instrument to measure locus of control, two sets of programs to simulate user-directed and system-directed dialogues, and a set of independent 52 questions to ascertain the subject's opinions about the di- alogue treatments. Locus of Control Measure To measure locus of control, Rotter's Internal-External Locus of Control scale was used (See Appendix A for a copy of the scale.) The scale is a 29 item, forced-choice questionaire including six filler items. Each of the other 23 items offer a choice between an internal and external be- lief statement (Rotter, 1966). An internal consistency analysis for reliability yield- ed r = .70 for males and females. Test-retest reliability coefficients computed after one month were r = .60 for males and r = .83 for females (Ritchie, 1970). Correlations between Rotter's Scale and the Marlowe-Crowne Social Desirability Scale (Crowne, 1964) range from -.07 to -.35. Dialogues The dialogues presented to the subjects were developed such that the same task was utilized in both treatments. The task was the entry of payroll data into a computer sys- tem. The subjects were provided handouts (See Appendix B for a copy of the hand-outs provided to the students) repre- senting a weekly payroll worksheet on which was recorded the employee number, name, dependents, hours worked, hours sick, and hours on vacation during the week. The subjects were informed that the first ten records had already been entered 53 into the system but errors were made as indicated by the cir- cled items. The employees below the double line had not yet been entered. The subjects were instructed to correct the errors previously made and add the data not already on file. When the corrections and additions were completed the sub- jects were to produce a listing of the data and the accom- panying totals. If the totals were correct the task was com- plete. If the totals were not correct the students were to correct any errors until correct totals were obtained. The two dialogues differed only in the level of control the user had over the task. In order to ensure that they differed only in terms of control, two steps were taken af- ter the first set of dialogues were developed. In the first step four pyschologists from the Psychology Department at Central Michigan University used the two dialogues and then provided their independent suggestions for changes. After modification the dialogues were administered to a group of 55 students. After using both dialogues the students were asked to submit written evaluations of each dialogue includ- ing a discussion of the degree of control they felt they had with each dialogue. The dialogues were then modified and again reviewed with two of the psychologists. Dialogue I. Dialogue I represented the system- di- rected environment. The student was given the opportunity to add, delete, or change records but only in that order. The students were first presented with a screen display 54 which asked the student to supply the data for a new record, as shown in Appendix C. When all the additions had been en- tered the student entered the word STOP to end the addition routine. The system then automatically sorted the file. While sorting, a message appeared on the screen informing the student that the payroll file was being sorted by em- ployee number. This step was necessary as the student might have added records out of sequence. Later routines that searched the file by employee number required the file to be in employee number order. Once the file was sorted the student was asked if there were any records to be deleted. The student would either en- ter an employee number or the word STOP. The student needed this ability to delete records that were inadvertently en- tered twice, or to delete employees entered with an incor- rect employee number. Next the change routine was presented to the student. The student entered the employee number of the record to be changed and was presented with the current contents of the record. The system then asked the user, one field at a time, if a change was needed. If the student responded with a Y answer the existing data were displayed on the bottom of the screen together with a place to enter the new data. When the student had completed all changes the student typed the word STOP rather than an employee number. Once the student ended the change routine the system au- tomatically printed a listing of the file. While listing 55 the file the system calculated the totals and displayed on the screen messages indicating whether each total was cor- rect or incorrect. If any total was incorrect the system again presented the add, delete, and change routines, and then produced a new listing. The sequence was always the same regardless of what the student needed to do next. Dialogue II. The second dialogue was written in a manner that gave the user control over the data entry and modification process. This dialogue provided a command driv- en interaction. The student specified one of seven commands defined by the first screen display. (The HELP command was available to display this first screen again to remind the students of available commands.) With this dialogue the stu- dent specified only those routines needed and in the order desired by the student. For example, the student might have desired to change records first, then add the remaining re- cords. Or the student may have listed the file before beginning. The ADD routine was the same as the first dialogue ex- cept that the student was allowed to “backup.“ That is, if the student discovered that an item previously entered was in error, the student could go back to that field and reen- ter the data. (In the first dialogue it was necessary to en- ter all the data for an employee and then reply with an N to the question "Is the record correct (Y/N)?'. Once the neg- ative response was entered, it was necessary to enter the en- tire record again.) The word STOP was used in both routines 56 to terminate the addition routine. The DELETE, CHANGE, and LIST commands in the user- directed dialogue provided additional control over the pro- cess by allowing the user to specify which records were to be processed. With each of these commands the student was asked "Enter mode: (A)ll, (O)ne, (R)ange, (K)ey -'. The stu- dent entered the appropriate letter; A, O, R, or K. The A option implied that all records would be affected by the op- eration, that is, all records would be deleted, listed, or presented for changes. The 0 option allowed the student to refer to records by their position in the file. After spec- ifying O, the student was asked for the record number. The number entered indicated the position of the record in the file, for example a 3 referred to the third record. The R option allowed the student to specify the range of relative record numbers to process. The student was asked for the lower and upper bounds. For example, if the range was from 2 to 6, then the second through the sixth records were pro- cessed. This was particularly advantageous for the change routine as only the second through the sixth records of the assignment needed to be changed. The last option, K, a1- 1owed the student to specify the employee number, or key, of a specific record. Only the record corresponding to the key specified was processed. The DELETE command caused the deletion routine to be invoked. The user was asked which mode of processing was re- quired and the corresponding records were deleted. If the 57 key mode was selected the student was informed that the file must be in key order. If records were not in sequence the SORT command was used by the student to arrange the records in employee number order. The CHANGE command allowed the student to change any fields within a record by entering the name of the field. After each field was changed the student could enter STOP to end the changes for the record or could continue to change other fields by specifying additional field names. The change routine could be used to process specific records or groups of records as described previously. The LIST command provided the student with the ability to display records on the screen or the printer. The format of the display was ei- ther a block format that resembled the display used in the ADD and CHANGE routines, or the format could have been de- fined by the student. The two format options were presented by the question - (B)lock or (F)orm Listing? If the F option was selected the student was asked for the fields to be included on the listing, whether totals of the columns were to be prepared, and what the title of the re- port should be. When the listing was complete the student was prompted for the next command. Once the student had determined that the totals were correct, the student entered the command END. 58 Opinion Questions At the end of each of the two dialogues, eight ques- tions were presented to the student one at a time to deter- mine the students' feelings towards the dialogue just used (See Appendix D for a copy of the questions). The first three questions asked the student to indicate their feelings concerning the ease of use of the ADD rou- tine, the CHANGE routine, and the overall system, respective- ly. The responses ranged from "easy to use“, to “very difficult to use". The second three questions were used to assess the anxiety of the student while using the system and were asked relative to the same three categories. The valid responses were "very comfortable", ”comfortable", “neutral", ”frustrated", and “very frustrated“. The seventh question asked the student to assess their own typing proficiency on a scale of 1 to 5 where a '1' indicated the minimum profi- ciency and a ”5” indicated a proficiency above fifty words per minute. The eighth question asked the student to eval- uate the software in terms of ”user-friendliness” ranging from ”user-friendly" to ”not at all user-friendly. The ques- tions were forced choice in that the system continued to ask for a response until the student entered a valid response. Administration of the Experiment TWO weeks before the use of either dialogue the Rotter Locus of Control instrument was administered in the class- room. As the locus of control instrument was not 59 characteristic of the subject matter of the course, an ex- planation was given such that the students would not relate the control issue to the dialogue treatment. The rationale for administering the questionaire was that the Information Systems and Analysis department was in the process of devel- oping a personality profile of the students typically en- rolled in the course. The two week time period between ad- ministering the questionaire and the treatments served to further disassociate the personality aspect from the dia- logue assignments. The equipment used was the IBM PC (personal computer) with a Microline Okidata printer. Each computer had 64K mem- ory, two disc drives and a monochrome monitor. There were eight computer systems available, all within the Grawn Hall Computer Lab at Central Michigan University. The students were introduced to the dialogue experiment as a class assignment that would give them an opportunity to assess two forms of man-machine interactions that they, or their future employees, would likely encounter. Completion of the exercise, including the written assessment, constitut- ed a regular assignment on a credit/no-credit basis. The di- alogues were presented to the students via a set of over- heads that showed each screen display that the students would encounter (See Appendix C for a capy of the overheads used). Questions posed by the students were recorded so that the same information could be conveyed to all groups. At the time of the presentation the handout titled 60 ”MAN-MACHINE DIALOGUE ASSESSMENT" was given to all groups together with the procedures for one of the dialogues. One section of each instructor began with the system-directed dialogue; the other section began with the user-directed dialogue. The dialogue instructions were labelled DIALOGUE I and DIALOGUE II respectively. No mention was made of the terms system-directed or user-directed. When a student finished one dialogue the printout, with totals, was handed in, and the student was given the instruc- tional handout for the opposite dialogue. When the second dialogue was completed the student turned in the correspond- ing printout, the written assessment and preference, and the diskette used by the student. Recorded on each student's diskette was the number of steps taken, the elapsed time from beginning to end of the task, and the responses to the eight questions. Following the treatments the aforementioned data to- gether with the sex of the subjects (obtained from registra- tion data) were assembled via a number of programs to create a composite file of all the data. The final data file recorded on the student's diskette contained the following items: Social Security Number Preference (l = Dialogue I, 2 = Dialogue II) Sex (1 = Male, 2 = Female) Locus of Control (Score of 0 to 30) 61 Steps for each dialogue (The summation of the number of times the ADD routine was invoked, the number of times the CHANGE routine was invoked, and the number of times the ENTER key was pressed) Responses to Questions 1 through 8 for each dialogue (Values were 1 through 5). Hypotheses this The hypotheses which were the basis for analysis of study are as follows: 1. The internal group will prefer the user- directed software more often than the exter- nal group; the external group will prefer the system-directed software more often than the internal group. 2. The mean times for completion of the task will differ between internal and external 10- cus groups when using the system-directed software. 3. The mean number of steps taken for comple- tion of the task will differ between inter- nal and external groups when using the system-directed software. 4. The internal group's mean score for time to completion will be less than the external group's mean scores for the user-directed software. 5. The internal group's mean number of steps will be less than the external group's mean number of steps for the user-directed software. 62 Statistical Analysis The first hypothesis was tested using the Chi-square statistic and a significance level of .05. The SPSS subpro— gram CROSSTABS was used to produce a table of locus of control (internal, external) by preference (Dialogue I - system-directed, Dialogue 2 - user—directed). The second hypothesis was tested by the two-tailed t- test for independent samples using a significance level of .05. The variables involved were locus of control and time to completion for each dialogue. The third hypothesis was also tested by the t-test for independent samples using a significance level of .05. The mean number of steps was established by adding the number of times the add, delete, and change routines were invoked to the number of items entered. The test was a comparison of the mean number of steps taken when using Dialogue I, the system-directed software. The fourth hypothesis was tested in the same manner as the second hypotheses using the t-test for independent sam- ples. The mean time scores of Dialogue II for the internal locus subjects was compared to the scores for the external locus subjects. The fifth hypotheses was also tested using the t-test for independent samples. A .05 significance level was used to reject the hypothesis. The variables were the mean number of steps for internal locus subjects compared with the mean number of steps for the external locus subjects in 63 completing the assignment for Dialogue II, the user-directed software. CHAPTER IV ANALYSIS OF DATA This chapter provides an analysis of the data collected during the study. As outlined in Chapter III, the sample consisted of four sections of the course "Introduction to Data Processing”. The sections were further divided on the locus of control variable. The median score on the Rotter locus of control instrument was used to classify the sub- jects as having an internal locus of control (0-10) or an external locus of control (ll-30), hereafter referred to as internal and external, respectively. Table 2 reflects the composition of each section in terms of the locus of control attribute. Table 2 Number of Study Participants Instructor A Instructor B Total S/U U/S SZU U/S Locus of Control Internal 22 22 26 21 91 External 26 24 30 34 114 Total 48 46 56 55 205 Note. 5/0 = System-directed dialogue was used before the user-directed dialogue. User-directed dialogue was used before the system-directed dialogue. U/S 64 65 The locus of control instrument was completed by 205 students; of those 183 also completed the dialogue assign- ments. One student completed the dialogue assignments who did not complete the locus of control instrument. That stu- dent was included in all analyses except those involving the locus of control variable. The performance data were collected via diskettes sup- plied by the students. Apparent inconsistencies in the number of students included in each analysis resulted from students not typing "END" at the end of the second dialogue. Without "END”, the computer software did not write the question file or the performance file to the student's diskette for the second dialogue. Hence, there are differences in the number of students reported for each dialogue. While it was possible to have the students repeat the second dialogue, it was the judgement of the researcher that misleading performance data might result because of the addi- tional practice obtained by the student. Therefore, the students with incomplete data for the second dialogue were not asked to repeat the assignment. Hypothesis Tests The first hypothesis stated: The internal group will prefer the user- directed software more often than the ex- ternal group; the external group will prefer the system-directed software more often than the internal group. This hypothesis was tested using a Chi-square test of 66 independence. The SPSS procedure CROSSTABS was used to pro- duce a 2 X 2 table with one dimension being locus of control (internal, external) and the other dimension being prefer- ence (system-directed, user-directed) as shown in Table 3. The test resulted in a Chi-square value of .66 which was not significant (p = .41). Clearly there was no support for the hypothesis. Table 3 Locus of Control by Preference Preference System User Locus of Control Directed Directed Total Internal 42 42 84 External 56 42 98 Total 98 84 182 Note. Chi-square = .66 Significance = .41 The second hypothesis was stated as The mean times for completion of the task will differ between internal and extern- al groups when using the system-directed software. The second hypothesis was tested with a two-tailed, independ- ent t-test. The independent variable was locus of control and the dependent variable was time to completion for the system-directed dialogue. The completion times reflected the number of minutes that elapsed from the time the student 67 started the dialogue until the correct answers were ob- tained. The times recorded did not include the time neces- sary to answer the questions at the end of the dialogue exercise. Table 4 presents the means and standard devi- ations for the internal and external groups. Table 4 Completion Times for the System-Directed Dialogue Task Number of Mean Locus of Control Cases Times SD Internal 88 29.92 14.19 External 101 27.22 9.02 A test for homogeneity of variance yielded an F value of 2.48 with a probability of .000. Therefore a t-test was per- formed using a separate variance estimate for determining the standard error term. The 5 value was 1.53 (p = .128). Again the hypothesis was not supported. The third hypothesis was: The mean number of steps taken for comple- tion of the task will differ between inter- nal and external groups when using the system-directed software. The third hypothesis was also tested using a t-test for independent samples. The independent variable was again locus of control. The dependent variable was the number of 68 steps used to complete the task using the system-directed software. The number of steps was calculated by adding the number of times the ENTER key was used to the number of times the ADD, DELETE, CHANGE, and LIST routines were invoked. The means and standard deviations are shown in Table 5. The independent t-test yielded a p value of —l.12 (p = .266). The hypothesis was not supported. Table 5 Steps Taken to Complete the System-Directed Task Number of Mean Number Locus of Control Cases of Steps SD Internal 88 184.69 19.08 External 100 188.17 23.10 The fourth hypothesis stated: The internal group's mean score for time to completion will be less than the external group's mean scores for the user directed software. The fourth hypothesis was tested with a one-tailed, inde- pendent t-test. The completion times recorded for the user-directed dialogue for the internal students were compared to the completion times for the external students. Relevant means and standard deviations are given in Table 6. The p value for the difference between means was 1.03 (p = .153). The fourth hypothesis was not supported. 69 Table 6 Completion Times for the User-Directed Task Number of Mean Locus of Control Cases Time SD Internal 77 34.75 15.11 External 88 32.36 14.70 The fifth hypothesis was: The internal group's mean number of steps will be less than the external group's mean number of steps for the user-directed software. The fifth hypothesis was also tested using a one-tailed t-test for independent samples. The dependent variable was the number of steps needed to complete the task with the user-directed software. The number of steps was again calcu- lated by adding the number of times the ENTER key was used to the number of times the ADD, DELETE, CHANGE and LIST rou- tines were invoked. Means and standard deviations are given in Table 7. The analysis yielded a p of .68 (p = .249). Hence the fifth hypothesis could not be supported. Having failed to support any of the hypotheses a poss- ible explanation was sought. One possible explanation was that the locus of control groups were not significantly different from each other because they were split at the median. 70 Table 7 Number of Steps for the User—Directed Task Number of Mean Locus of Control Cases Steps SD Internal 73 126.58 34.45 External 84 122.75 35.35 The groups were redefined to include only the upper and lower 27 percent, as has been suggested by Kelly (1939). The test for each of the hypotheses was repeated using the newly defined, smaller groups. The results are shown in Tables 8 and 9. With a Chi-square of .115 (p = .734) and obtained E's ranging from -l.38 to +1.08 (p ranged from .172 to .364), all of the tests again proved non- significant. Even with a greater differentiation between internals and externals, none of the hypotheses was supported. Table 8 Locus of Control by Preference Using Upper and Lower 27% Locus of Preference Control System-Directed User-Directed Total Internal 19 23 42 External 25 24 49 Total 44 47 91 Note. Chi-square = .115 Significance = .734 71 away emaaabuosu umou ooawmuloco u m ”ouoz . . . Hm.~v hh.>~H oe accumuxm ANN as me me Ho.o¢ o¢.mma hm HaaumucH xmna wouomuwaluwmp How mmmum . . . i «H.Hm No.5wa Hm accumuxm «AH em as am H me.ma G¢.~mfl me HmaumuaH xmme omuomnwaiewummm MOM mmmum . . . Hm.na mm.¢m mv accumuxm Beam so om mm Gm.ha o~.Gm oe HmcumuaH xmwa owuomnwolummo new mafia . . . om.m om.mm mm accumuxm seem so mm as H mn.aa -.m~ me HmaumucH xmma wouomnfialsouwwm HON mafia musmmmz woonauomumm m up u am one: mwmno mo nonsdz m wanes wmsouw Houucoo mo msooq Amnmv umzoqlummms Mom wUGMEMOMHmm 72 Supplementary Analyses Order Effect. T-tests were done to determine if the order of the treatments had any effects on the performance by the subjects. Performance was measured in terms Of time to completion and in terms of the number of steps necessary to achieve the correct totals on the data entry task. The subjects were classified by the dialogue with which they began the experiment. the comparison of performance measures. ranging from -1.07 to +1.52 (2 it was obvious that there were The order of the treatment did performance. The preference for either affected by the order in which tered. A 2 X 2 table with one Table 10 shows the results of With E values ranging from .131 to .747), no significant differences. not have an effect on dialogue could also have been the treatments were adminis- dimension being the order of the treatment (system-directed/user-directed, user- directed/system-directed) and the other dimension being pref- erence (system-directed dialogue, user-directed dialogue) was constructed (Table 11). A Chi-square test revealed no support for preference being dependent upon the order of the treatments. Instructor Effect. Next the data were analyzed to de- termine if the classroom instructor had an effect on the performance of the students with either treatment. 73 .xmmu omuomuflolemumwm may cucumn tmEHOMHom mm3 xmmu touooufioiummo u m\D .xmmu touomuwciuoms wnu mHOan ooEMOmuom mus xmmu touowufiousoumwm u D\m .muoz . . . Hm.wm Hm.mmH on m\D has so mma mm om.¢m mm.mma em a\m xmwe owuooufloiummo uOM mmmum . . . m>.ma mm.vma mm m\D Gma oo mma om H m¢.- w~.mma HOH o\m xmma wouooufloisoumwm new mmmum . . . I mw.ma om.vm mm m\: mmm on mma no H o~.ma mm.~m om D\m xmme touomufioiummo How mafia . . . wc.HH vo.hm mm m\D and so mma «m H ma.~H Ho.a~ «OH o\m xmme omuomuwoiEmummm now mafia mupmmmz mocmEMOMHOm Q up H am one: mommo mo umnEdz OH wanna QUCMEHOWHQQ CO HOQHO UCOEMOHB MO UUQMMW 74 Table 11 Effect of Treatment Order on Preference Order of Treatment Preference S/U U/S Total System-Directed 52 47 99 User-Directed 48 36 84 Total 100 83 183 Note. S/U = System-directed dialogue was used first. U/S = User-directed dialogue was used first. Chi-square value of .226 Significance = .633 The students were grouped by the instructor for the sec- tion in which the students were enrolled. Time to comple- tion of each task and the number of steps taken to complete each task by instructor group, constituted the performance data shown in Table 12. With §_values from .65 TO 1.60 (p from .113 to .519) it was apparent that there was no support for the hypothesis that the instructor of the course had influenced the performance of the students. The effect of the course instructor on the students' preference was also analyzed. The analysis was a 2 X 2 Chi- square with one dimension being the course instructor (Instructor A, Instructor B) and the other dimension the student's preference (system-directed, user-directed). The data are given in Table 13. The test yielded a Chi-square of .000 (p = 1.0), thus there was no support for the hypothesis that preference was 75 . . . mm.mm wm.o~a mm m nouosuumcH xmme pmuowufialumms How mmmum . . . ha.m~ mm.mma mm m uouonuumcH eom so man he mm.o~ me.smfl as e uouosuumcH gums omuomuwaleoumhm now mmmum . . . mm.ma Hm.mm mm m nouosuumCH mam so MGH me m¢.mH mm.¢m «A a nouosnumcH xmma wmuomufioiummo Mom mafia . . . mm.HH hm.>m pm m uouoauumcH ham so ama mm Hm.HH Am.m~ «a a nouosuumcH xmme omuomufioiewumwm now mane nuance: mocmEHOMHom a up u am one: mommo mo umnEdz OUCMEHONHOQ CO HOUUSHUmCH QwHDOU MO HUMMMW NH manna 76 dependent upon the instructor of the course. Table 13 Effect of Course Instructor on Preference Course Instructor Preference Instructor A Instructor B Total System-Directed 45 54 99 User-Directed 39 45 84 Total 84 99 183 Note. Chi-square = .0 Significance = 1.0 Sex of the Subjects. The effect of the subjects' sex (male, female) upon their preference for either dialogue was analyzed using a Chi-square statistic. The 2 X 2 table was constructed with sex as one dimension and preference as the other dimension. As can be seen the results from Table 14 indicated that the sex of the preference. Tab Effect of Sex individual did not relate to 1e 14 on Preference Sex Preference Male Female Total System-Directed 49 50 99 User-Directed 40 44 84 Total 89 94 183 Note. Chi-square = .010 Significance = .916 77 The performance of the subjects was also analyzed by sex (male, female) to determine if either sex varied signif- icantly in terms of time to completion of each task or in terms of the number of steps used to complete each task. Table 15 gives the means, standard deviations, and the results of the analysis. It was found that the females differed significantly from the males in terms of time for completion of the system-directed task (5 = 2.96, p = .004) and they were significantly faster in terms of time to completion of the user-oriented task (p = 2.77, p = .006). The females did not differ significantly from the males in terms of the number of steps to complete either dialogue task. As there was a significant difference in time between the male and female groups but no significant difference in the number of steps used to complete each task, it was hy- pothesized that one group possessed better typing skills than the other. The hypothesis was tested by comparing males and females on the typing level reported via question seven (Appendix D) at the end of each dialogue. The means and standard deviations are shown in Table 16. A one-tail t-test for independent samples produced a significant statis- tic (p = .000). The t-value indicated females reported having better typing ability than did males. An examination of reported typing ability therefore verified that there was a significant difference between males and females. At the end of each dialogue task seven other questions 78 . . . hw.vm mv.m~H mm madame new so mma as ¢~.mm vs.m~H ms was: xmwe tmuomufioiummo MOM mmmum . . .I ma.mm vv.mma hm mansmm «am so was so ma.om mm.mma mm mam: xmme omuomufloiemummm Ham mmmum . . . m~.ma mm.om em wamswm woo oo mma mm m Hm.ma mo.mm am can: xmme oouomufloiummo now mafia . . . vv.oa oo.mm hm mamswm «co co mma mm m h¢.ma Hm.om vm was: xmma omuomufialemumhm mom mafia musmmmz mocmEMOMHmm a mo u am one: mmwmo mo nonadz OOCMEHOMHOQ CO NOW MO UUQHMW ma OHQMB 79 mH.H mm.m mm mmamemm ooo oo mna mm mi mo.H mw.~ mm moan: Amiav xmme tmuoouflolummo may NO one ecu um Ho>mq mcfimwa pwuuoowm . . . .I mH.H mm.m mm mwamewm ooo oo mma mm o mo.H mw.m em moan: AmIHV some pouomuMQIEmumwm on» mo too on» an Ho>mq mowmwe pmuuommm oupmmmz mocmEMOMHmm m up u om mmcommom mommo no one: uwnEnz Hawxm mcflmme omuuommm 0» Row mo mesmCOwumHmm ma wanna 80 were asked in addition to the question on typing ability (See Appendix D). The questions were included in the experiment to explore student perceptions of the terms "ease of use”, "anxiety", and “user-friendliness“ as they relate to man-machine dialogues. As the dialogues of this study only differed in the control dimension, responses to the questions could be useful in defining the aforementioned terms as they relate to software dialogues. Ease of Useggpestions. The first three questions were asked to determine the difficulty of using each dialogue. It was predicted that internal individuals would find the user-directed software easier to use than the system-directed software. It was also thought that the external group would find the system-directed software easier to use than the user-directed software. The ease of use questions were analyzed by a one-tailed t-test for independent samples. As is shown in Table 17, there were no significant findings for the ease of use questions asked after the system-directed task (5 = -.84 to .07, p from .201 to .471). However the internal group rated the ADD command of the user-directed software as easier to use more often than did the external group (p = -1.67, p = .049). There was not a significant difference in ease of use responses for the CHANGE command or for the overall user-directed system between the internal and external groups (p = -l.51 to -l.22, p = .066 to .111). The responses to the ease of use questions were also 81 . . . I moo.H HH.~ mm Hmcuouxm HHH oo aha mm a moo.a mm.a om accumucH msmoHMMQ pmuomuflaIummD . . .I was. Hm.H Hoa accumuxm owe oo baa oH cam. om.H mm HmcuwucH osmonfla pmuomuflalemummm Afimummm HHnInm>ov m cofiummso I om: mo omcm . . . I mmo.a mo.~ mm accumuxm moo oo Ana am A sea. om.a om accumucH osmoHMMQ omuomufialuomo . . . mam. om.H Hoa Hmcuouxm are oo baa no «mm. Hm.H mm HmcuoucH osmonfia owuoouflaIEoummm Amcfludom wmcmnuv N cofiummso I on: no wmwm . . . I mom. mm.a mm accumuxm momonflo vmuomufiqummD . . .I man. mm.a Hoa accumuxm How oo hma vm mvm. mv.a mm HocuoucH osmonfia cwuomuflolemummm chfiusom come a Goduwoso I own mo omnm musmmmz moannomuwm Q up p cm uncommom mommo no one: nonesz own mo mmmm ou Houucou mo msooq mo QMLmCOLDMHom ha mance 82 analyzed in terms of the stated preference for either the system-directed or user-directed software. A one-tailed t-test for independent samples was used to compare the preference groups on their responses to each question. Table 18 shows the means and standard deviations. In each test the results were significant (p ranged from -2.38 to 5.67, p ranged from .000 to .008). Anxiety Questions. Questions 4, 5, and 6 were asked of the students to record their level of anxiety in relation to the ADD routine, the CHANGE routine, and the overall sys- tem. The responses to each question were analyzed by preference group with a one-tailed t-test for independent samples as shown in Table 19. There was not a significant relationship between preference and the responses to the anxiety question for the ADD routine of the system-directed dialogue (p = -l.55, p = .061). However, there was a significant difference in the responses to each of the other anxiety questions (5 = -2.17 to 5.19, p = .000 to .017). The hypothesis that preference was related to the level of perceived frustration was supported in all but one instance -- the effect, upon preference, of the ADD routine of the system-directed dialogue. The same anxiety questions were then evaluated in rela- tion to locus of control. It was expected that the internal group would rate the system-directed software as more frus- trating than would the external group; similarly it was hypothesized that the external group would rate the 83 . . . moo. mm.H mo Hmcuwuxm .ooo ow moa no m evo.a mm.N No HmcuwucH msoonfio owuomufiquomo . . . I omm. mo.a Hm accumuxm moo mm ova mo N omm. m¢.H mo HmcuwucH momoHMNQ omuooufloIEmummm Aemumwm HHnIuo>oo m coNumoso I om: mo mmmm . . . ooh. mm.a mp Hmcuouxm ooo om mod ma v owo.a oH.N No HmcuoDCH momonfia pmuomuonummD . . . I NHo.H no.a Ho accumuxm ooo mm omH av N ooo. vv.a mo ancuoucH osmonfia omuomufiaIEwummm chNusom mmcmnoo N codummso I mm: no down . . . ooh. ov.a mm accumuxm Hoo om voa NH m Hmo. oo.a No ancuoucH mooonNo omuoouonummD . . . I Hmm. oo.a Ho accumuxm Noo oo ooa mm N oNh. mm.H om HmcuoucH osmonfio oouomuNoIEmuwmm Amcfiuoom oooo H coflummoo I on: no omnm musmmmz mocmSHOMAmm o mo p cm wmcoommm mommo no can: nonsdz ma magma mm: mo ommm on mocwummmum mo oflnmcoNumHom 84 cofiummso mumwxcm on» cmn3 poms ocfion was umcu momoHnwo on» on muowmu momonfia H .ooxmm magma mp3 .wuoz . . . mmo. om.a mo omuomufiqummD cmuummmum ooo as «ea ma m Ham. em.~ ma cmuomuLaIsmumsm emnummmum momonNa pmuomufiquomD . . . I mom. oo.a Hm owuomufiqummD owuummoum mHo oo ooa ha N wNo. Nh.a om wouomuwaIEmumam pmuuommum osmoamfia UmuomufiaIEmuwwm ismummm Hamum>ov m cofiummao I annexed . . . ooo. mm.a mo wouomnNQIummD wouuomoum ooo me mma me a mmo.a ~m.~ mm cmuomunaIEmumsm emuuoumum woooHMNQ omuomuonummD . . . I ham. mo.N Ho omuomufiolumms omuuomoum oHo oo moa NH N Ham. oo.H om owuomnonamumwm pmuuommum womoamfia pwuowuonawumam Aocfiudom mmcmnuo m codummoo I wumfixcm . . . Non. mm.a mo omuomnNoInmmD twuuommum moo Go «Ga mm m Hmm. mm.a mm emuomnnousmumsm emuummmum mdooHMNO pwuomuNQIummD . . . I Now. oo.a Hm omuooufiqummD omuummmum Hes oo maa mm a mop. mm.H mm omuomuoousmummm omuummmum msmonHo cmuomuonEmummm chfiusom oodv v aofiummso I >uo«xc¢ a up p cm mmcommmm mommo wo macaw mocmummmum wuwfixn< nonEdz wsmoammNo com: A mumfixcd ou wocmumwmum mo ownmcofiunamm ma OHQMB 85 user-directed software as more frustrating than would the internal group. The data to test the hypotheses are shown in Table 20. Neither hypothesis was supported (3 = -l.27 to -.08, p = .102 to .468). User Friendly Questions. Question number 8 asked the student to rate each dialogue in terms of user-friendliness. It was expected that the system-directed dialogue would receive a higher user-friendly rating (lower score) by those preferring that dialogue than it would from the group preferring the user-directed dialogue. Table 21 illustrates that the difference in response scores between the preference groups was not significant for the system-directed dialogue (p = -l.52, p = .065). However when the same question was asked at the end of the user-directed dialogue the results were significant (p = 3.68, p = .000). That is, those who preferred the user-directed dialogue gave the dialogue a better user-friendly rating than those who preferred the system-directed dialogue. Finally the internal/external groups were compared on their responses to the question of user-friendliness (Table 22). There was no significant difference between the responses to the question asked at the end of the system-directed dialogue (E = -.73, p = .232) but there was a significant difference (5 = 1.69, p = .047) between the groups on the user-friendly question when asked at the end of the user-directed dialogue. 86 mp3 coHumoso wumecm on» .cwxwm mEHu on» um poms osmOHmHt on» on mummmn momonHo H .muoz . . . I ooo. oH.N mm Hocuouxm osmonHn pmuomuHqume . . .I one. mo.H HoH Hmcuwuxm ooo oo ooH oo Hmo. mm.H om HmcumucH msmonHa omuomanIEmumwm HEmumhm Hchm>oo o coHummso I aumecm . . . I HNo.H mH.N mo Hmcumuxm NoH oo HNH NN H moo. mo.H om HmcuoHCH mnmonHo touomuHqummD . . .I who. wo.H HoH Hmcumuxm mom oo ooH mN ooo. mo.H om HmcumucH osmonHo pmuomuHoIEoummm HocHunom mmcano m coHummpo I wumec< . . . I Noo. mo.H mo Hmcumuxm hoH oo HNH mm H omo. No.H ow anumuCH womonHa pouomuHqume . . .I How. mm.H HoH Hmcuouxm moH oo hoH no vmo. No.H mo HmcumuaH momonHa omuoouHaIEmumam AmcHusom come v coHuwoso I mumec< m mo p cm mmcoommm women no Houucoo mo msooq wumecd uwnEdz womoHnHQ cums H wumecm ou Houucov mo mDUOH mo oHnmcoHHMHmm ON OHQMB 87 ooo. hm.NwH mw.m me. mm.H mo weaveHHDIuemD Ueuuemeum ooo. no.N No oeuoeuHoIEeummm oeuuemeum esmoHeHo weuoeuHquemD . . . I «on. mm.H Hm oeueeuHquemD peuuemeum moo oo msH mm H has. mG.H mm emuomuHaIamnmmm vehememum eomoHeHo oeuoeuHQIEeumMm m coHuweso I wHoceHuquema m up p cm encommem memeo mo msouw eoaeuemenm one: Henadz esmoHeHo mwQCMHCCQHthume Cu OOGQHGMOHQ MO QHSMCOflUMHmm HN eHQMB 88 . . . I omo. om.H mm Hecueuxm ooo oo HoH mm H woo. m>.H om HmcueHCH esmoHeHa oeuoeuHQIuemD . . .I ooo. Ho.H HoH Hecueuxm NmN oo an mo New. mm.H mm HecueucH eDOOHeHo oeuoeuHoIEeummm o :oHumeso I wcheHuquemD m, up u am emcommem memee mo Houucou mo mpooq one: HenEdz eDmOHeHQ wmecHHoceHum news on Houucoo mo msooq mo oHnwcoHueHem NN eHneB CHAPTER V CONCLUSIONS AND RECOMMENDATIONS The theory on which this study was based was that per- formance in a man-machine interaction is dependent upon 1) an individual's cognitive abilities, 2) an individual's short term memory capacity, and 3) the degree of anxiety pro- duced by the interaction. This study focused on anxiety in the man-machine interaction. It has been suggested by the literature that individ- uals with different personality attributes will respond differently to the same computer environment. That is, if the software is not congruent with the personality of the individual, anxiety will result. It has also been suggested that one characteristic of computer software, control, might well be a factor in causing anxiety in man-machine dia- logues. In order to test the aforementioned performance theory, that is, that anxiety in a man-computer interaction has an effect on performance, a situation was constructed to induce anxiety using the locus of control characteristic. Two computer software dialogues were developed such that one dialogue offered limited control to the user and one dialogue gave the user considerable control over the interaction. Hence two computer environments were created to satisfy two different groups of users. The users were 89 90 then classified as internal (those who desire control) and external (those who do not desire control), based on scores from the Rotter Locus of Control Scale. Performance was measured in terms of the time to comple- tion of each task and in terms of the number of steps taken to complete each task. It was found that there were no sig- nificant differences in performance between internal and external groups when the groups were defined by the median locus of control score. The locus of control groups were then redefined to include only the extreme upper and lower 27 percent and the tests conducted again. Even with only the most internal and most external individuals included there was no evidence to support the theory that performance varies because of anxiety as produced by personality differ- ences. As the results did not support the theory several ques- tions need to be answered at this juncture. Were the study instruments sound? Were the performance measurements valid? Were the tasks appropriate? And finally, was the study a valid test of the theory or is the theory wrong? Locus of Control The locus of control scale by Rotter was designed to measure generalized expectancies of individuals. If the instrument is an accurate reflection of locus of control, as the literature suggests, then it must be concluded that locus of control is not a factor in situations where 91 individuals interact with computers. However, it is poss- ible that those who are generally internal in other situations are in fact not as internal when working with computers or that those who are external in the general sense are not as external when working with computers. That is, the locus of control measure may not be situation specif- ic. What is needed is an instrument that can accurately assess locus of control in specific situations. It is also possible that the desire for control in a man-machine interaction is a function of the user's familiar- ity with the task and the software employed. Individuals, whether they be internal or external in a general sense, may well react differently when working with computers. In order to pursue this thought it would be necessary to pro- vide users with software that allows them to select the amount of control they desire and to observe their choices over a period of time. What is important to note is that the generalized measure of locus of control did not material- ize as relevant to performance in the use of either the limited control (system-directed) or the control (user- directed) dialogue. Performance Measures Performance was measured in two dimensions, time and steps to completion of the task. The time recorded was the number of minutes that elapsed from the time the exercise was started until correct totals for the task were achieved. 92 The time was recorded by the computer software and could not be altered by the user. As can be seen from the analysis, females completed the assignments in less time than males. As the completion time can be affected in this type of study by typing ability subjects should be tested for typing pro- ficiency prior to the experimental treatment to control for this variable. While the difference in mean completion times for males and females has been attributed to typing ability, this is not the only conclusion that is possible. Other differences between males and females should be explored to determine if there are other important differences that could affect time to completion. Typing ability could also have an impact on performance other than the obvious ability to enter data at a faster rate. It is possible that limited typing ability might cause frustration in a man-machine interaction. If such frustration exists comprehension of the syntactic and seman- tic requirements of the dialogue could be affected and hence performance could be influenced as well. Performance was also measured in terms of the number of steps needed to complete the task. The number of steps required was the sum of the number of times each routine was used and the number of times the enter key was pressed. The enter key was used after any data was entered at the key- board. If a field was changed or added the enter key was pressed, hence the amount of activity within a routine was 93 measured. The step measurement was therefore an accurate reflection of the total amount of activity but was not use- ful for measuring activity within each routine. It is suggested that in future studies the activity in each rou- tine should be measured independently so that more detailed analysis would be possible. With the system-directed dialogue the user did not have alternative ways to complete the task, hence the number of steps was strictly a function of the number of errors made by the student. With the user-directed dialogue the student had options within the change and list routines. For exam- ple, if the student elected to change a range of records rather than specific records, several fewer steps were needed. The step count for the user-directed dialogue did not differentiate between steps taken because of errors and steps taken because of the approach to solving the problem. Future studies should measure performance in a manner that would allow analysis of both variables. Such analysis would have been mandatory if the purpose of this study had been to analyze the method used to complete the task. Although the performance measures could be improved, they did adequately measure performance for the purposes of this study. They did not contribute to the negative findings of this study. Dialogue Routines The system-directed software was developed such that the user had very limited control over the task to be 94 completed, whereas the user-directed software was designed to give the user considerable control. The degree to which each dialogue actually conveyed a sense of control to the student could have had a profound effect on the results of the study. Although the pilot study and evaluations by the psychologists indicated that the dialogues were effective in isolating control, verbal responses from participants raise some questions relative to the change routine. As stated previously, in the system-directed dialogue the student could only change a specific employee record and after speci- fying that record the student had to answer yes or no to each field displayed on the screen. The presentation of the material on the screen in this dialogue was more pleasing to many students than the method of presentation in the user- directed dialogue. While the format of the screen did not affect the amount of control the student actually had, it could have had an affect on their attitude toward the soft- ware and hence their performance as well as their prefer- ence. Also, had the preference question been subdivided into several questions, more could have been learned about the specific routines employed and hence the overall prefer- ence for either dialogue. In the change routine of the user-directed dialogue the student had several options including the ability to change a range of records (the most efficient method) or of chang- ing a specific record. Regardless of the choice made the student next specified fields to be changed by typing the 95 name of the field to be changed. Even though the user- directed dialogue enabled the student to change data more quickly many students expressed a dislike for typing in field names. Software that would allow the same degree of control but would perhaps accept shorter abbreviations or codes representing the field names would alleviate this crit- icism. For example, VAC might be used in place of VACHRS, or REG might be used in place of REGHRS. It is not likely that the specification of the whole field name adversely affected the results. Treatment Task The task performed using each dialogue involved the add- ing, changing, deleting, sorting, and listing of records. Such activities are common to a data entry operation, and are usually performed by clerical personnel. The task was well suited to this study because it was easily learned and did not introduce extraneous variables that a more complex task might involve. The task itself however raises several concerns. As the task was rather elementary in terms of the skills required there might have been inadequate motivation to perform at one's capacity. This is a particular concern with the user-directed task where alternative methods of com- pleting the task were possible. There was no reward provided for superior performance except the satisfaction of completing the task. Had the assignment grade taken into account the number of steps to completion, performance might 96 have improved. With the system-directed dialogue the only way to reduce the number of steps would be to make fewer mis- takes. With the user-directed dialogue either fewer errors or a better choice of options would have yielded fewer steps to completion. Increased incentive for improving performance would have in turn increased the possible anxiety to a level that might have affected the overall performance and hence the outcome of the study. Self-Report Questions After each dialogue seven questions were asked relating to ease of use, anxiety, and the user-friendliness of the software. The questions were intended to capture the atti- tude of the subjects towards each dialogue at the time the subjects were using the computer software. The questions were analyzed in terms of dialogue prefer- ence and locus of control. While there were significant findings relative to locus of control the reader is cautioned that the validity of the internal, external classi- fications in the man-computer interaction might well be questioned. Ease of Use. There were three ease of use questions for each dialogue relating to the ADD routine, the CHANGE routine, and the overall system. There was no significant relationship between locus of control and ease of use for the systems-directed dialogue. There also was no signifi- cant relationship between locus of control and ease of use 97 of the CHANGE routine or the overall system for the user— oriented dialogue. However, the external locus of control group rated the ADD routine of the user-directed dialogue easier to use than did the internal group. The ADD routines in the two dia- logues were identical except for two features. First, the ADD routine could be invoked at any time in the user-direct- ed dialogue. Second, the user-directed dialogue allowed the user to back up one field at a time to correct mistakes made during data entry. It is suggested that these features are more important to internal individuals than to external indi- viduals. The responses to each ease of use question were also compared by preference group. In all six cases there were significant differences between the group responses. It may be concluded that in the selection of dialogues that differ only in the control dimension, ease of use will be a factor in individual preference. Anxiety and User-Friendliness. Five of the six anx- iety questions posed at the end of each dialogue proved to be related to preference. Based on this relationship it can be said that the level of anxiety was sufficient in the study to cause the students to consider this dimension of the software in their preference. However, there was no relationship between locus of control and the responses to any of the anxiety questions. The user-friendly question was also asked at the end of 98 each dialogue. The responses for the system-directed dia- logue were not found to differ significantly between preference groups or between locus of control groups. How- ever, the responses to the user-friendly question asked at the end of the user-directed dialogue did differ signifi- cantly for both the preference groups and the locus of control groups. The internal group rated the user-directed software as user-friendly more often than did the external group. Likewise, the group that preferred the user-directed dialogue rated that dialogue as user-friendly more often than did the group that preferred the system-directed dialogue. It can be stated that software which gives individuals control over the man-machine interaction (dialogue) will be considered as user-friendly by internal locus of control in- dividuals. Similarly external individuals will tend to rate the same software as not being as user-friendly. Effect of Order and Instructor The subjects were classified according to which dialogue they used first to determine if the order of the treatments had an effect upon their performance or prefer- ence. There was no relationship between the order in which the dialogues were used and the performance or preference of the group. The subjects were also grouped according to course instructor. Performance did not vary significantly between 99 the two groups, nor was there a relationship between prefer- ence and instructor. It can therefore be concluded that the instructors did not have a significant impact on the stu- dents' preference or performance. Performance Theory It is concluded that the performance theory was not sup- ported because of the personality trait chosen to test the theory. Either the locus of control instrument used does not properly classify students for their locus of control when working with computers, or if the classification is cor- rect, the lack of congruence between the individual and the software is not sufficient to test the theory. It is not possible, based on this study, to ascertain which is true. It is suggested that possibly both are true and that to prop- erly test the theory other approaches should be examined. Such approaches should utilize stronger measures of individ- ual differences, preferably in the cognitive domain. Recommendations for Further Study The theory that performance is affected by anxiety was clearly not supported by this study. As the personality trait, locus of control, did not produce sufficient anxiety to support the theory, other approaches are suggested. First, anxiety in man—machine interactions requires definition. In this study anxiety was treated as an incon- gruence between personality traits and computer dialogues. 100 While this definition may be adequate for an exploratory study, further work should begin by refining the definition. It is also possible that anxiety manifests itself in forms that are measurable. The development of a working defi- nition and the development of an instrument to measure anxiety in a man-machine interaction is needed. Second, this study did not attempt to measure anxiety but only to induce anxiety in one situation and remove anxi- ety in the other. It is suggested that anxiety exists on a continuum and that performance at different levels of anxi- ety will be a curvilinear relationship. That is, at certain points anxiety might well enhance performance, at other points it may diminish performance. Third, personality differences were not sufficient to produce anxiety at a level that affects performance, hence other facets of the man-machine interaction need to be manip- ulated to test the theory. The other facets to be studied might be in the form of motor skills, such as typing, or might be cognitive differences such as reading comprehens- ion. The advantage of using such individual characteristics outside the affective domain is that there are readily avail- able instruments for measuring these traits; instruments which have greater validity than is possible with those available for measuring affective traits. Fourth, typing proficiency affected performance in this study but did not affect preference. What is not known is the impact of motor skills, such as typing, on comprehension 101 in a man-computer interaction. Studies need to be conducted to determine if typing ability might influence the rate at which computer users learn the syntactic and semantic struc- tures of man-computer dialogues. Fifth, motivational factors should be established in studies on performance to insure that students are working to their capacity when completing the exercise. Sixth, a longitudinal study would perhaps reveal more about the desire of individuals for control over man-machine dialogues. While it is known that novice users do not seek control, it is not known at what level of exposure to man- machine dialogues users begin to desire control. Another experiment might offer several levels of control within a single software product. (A similar method is used in video games where the user can select their level of pro- ficiency from novice to expert.) With such software the users' selection of control level could be recorded over a series of treatments to ascertain if locus of control was relevant to desire for control in computer software when a choice was available. APPENDIX A 102 APPENDIX A ROTTER LOCUS OF CONTROL SCALE Student Number This is a questionnaire to find out the way in which certain important events in our society affect different people. Each item consists of a pair of alternatives lettered a or b. Please select the one statement of each pair (and only one) which you more strongly believe to be the case as far as you're concerned. Be sure to select the one you actually believe to be more true rather than the one you think you should choose or the one you would like to be true. This is a measure of personal belief: obviously there are no right or wrong answers. Please answer these items carefully but do not spend too much time on any one item. Be sure to find an answer for every choice. In some instances you may discover that you believe both statements or neither one. In such cases, be sure to select the ppg you most strongly believe to be the case as far as you're concerned. Also try to respond to each item independently when making your choice; do not be influenced by your previous choices. Circle the apprOpriate answer for ypp. l a. Children get into trouble because their parents punish them too much. b. The trouble with most children nowadays is that their parents are too easy with them. 2 a. Many of the unhappy things in peOple's lives are partly due to bad luck. b. PeOple's misfortunes result from the mistakes they make. 3 a. One of the major reasons why we have wars is because people don't take enough interest in politics. b. There will always be wars, no matter how hard peOple try to prevent them. 4 a. In the long run peOple get the respect they deserve in this world. b. Unfortunately, an individual's worth often passes unrecognized no matter how hard he tries. S a. The idea that teachers are unfair to students is non-sense. b. Most students don't realize the extent to which their grades are influenced by accidental happenings. 6 a. Without the right breaks one cannot be an effective leader. b. Capable people who fail to become leaders have not taken advantage of their Opportunities. a. b. b. 10 ll 12 13 14 15 l6 17 18 103 No matter how hard you try some pe0ple just don't like you. PeOple who can't get others to like them don't understand how to get along with others. Heredity plays the major role in determining one's personality. It is one's experiences in life which determine what they're like. I have often found that what is going to happen will happen. Trusting to fate has never turned out as well for me as making a decision to take a definite course of action. In the case of the well prepared student there is rarely if ever such a thing as an unfair test. Many times exam questions tend to be so unrelated to course work that studying is really useless. Becoming a success is a matter of hard work, luck has little or nothing to do with it. Getting a good job depends mainly on being in the right place at the right time. The average citizen can have an influence in government decisions. This world is run by the few people in power, and there is not much the little guy can do about it. When I make plans, I am almost certain that I can make them work. It is not always wise to plan too far ahead because many things turn out to be a matter of good or bad fortune anyhow. There are certain people who are just no good. There is some good in everybody. In my case getting what I want has little or nothing to do with luck. Many times we might just as well decide what to do by flipping a coin. Who gets to be the boss often depends on who was lucky enough to be in the right place first. Getting peOple to do the right thing depends upon ability, luck has little or nothing to do with it. As far as world affairs are concerned, most of us are the victims of forces we can neither understand, nor control. By taking an active part in political and social affairs the peOple can control world events. Most people don't realize the extent to which their lives are controlled by accidental happenings. There realy is no such thing as "luck". 19 20 21 22 23 24 25 26 27 28 29 a. b. 104 One should always be willing to admit mistakes. It is usually best to cover up one's mistakes. It is hard to know whether or not a person really likes you. How many friends you have depends upon how nice a person you are. In the long run the bad things that happen to us are balanced by the good ones. Most misfortunes are the result of lack of ability, ignorance, laziness, or all three. With enough effort we can wipe out political corruption. It is difficult for peOple to have much control over the things politicians do in office. Sometimes I can't understand how teachers arrive at the grades they give. There is a direct connection between how hard I study and the grades I get. A good leader expects peOple to decide for themselves what they should do. A good leader makes it clear to everybody what their jobs are. Many times I feel that I have little influence over the things that happen to me. It is hmpossible for me to believe that chance or luck plays an important role in my life. PeOple are lonely because they don't try to be friendly. There's not much use in trying too hard to please people, if they like you, they like you. There is too much emphasis on athletics in high school. Team sports are an excellent way to build character. What happens to me is my own doing. Sometimes I feel that I don't have enough control over the direction my life is taking. Most of the time I can't understand why politicians behave the way they do. In the long run the peOple are responsible for bad government on a national as well as on a local level. APPENDIX B 105 MAN-MACHINE DIALOGUE ASSESSMENT The purpose of this assignment is to give you exposure to two different man-machine dialogues that are freQuently seen in software today. As students of business, you will become the future managers of various business functions. In that capacity, you and your staff will interact with computers more than any previous generation has. It is therefore important that you are aware of alternatives available in man-computer dialogues. This assignment consists of two exercises which will be handed out separately. Both exercises involve the same task, the data entry function in a payroll application on the IBM PC microcomputer. You will be provided with a page of data representing a typical Departmental Time Report and with Operating Procedures for each exercise. (The individual exercises are labelled DIALOGUE I and DIALOGUE 11.) You will start with whichever dialogue is distributed to you first. Only after the first exercise is handed in should you do the other exercise. On each Departmental Time Report a double line marks where a previous operator stopped. That is, the first half of the data has been entered already. The circled items above the double line indicate that the particular item was entered incorrectly and must be changed. The data below the double line must be entered by you. The exercises differ only in the data used and the manner in which the data is manipulated. Bach exercise permits you to add, delete, change, or list the payroll data. The specific procedures are outlined on separate pages labled ngrating Procedures. Each exercise is complete when you have tots s at match the totals on the bottom of the Departmental Time Reports. As you complete each exercise evaluate the experience in your own words. The main difference in the two exercises is the dialogue used. Focus your comments on the dialogue. There are no right or wrong answers, only valid opinions. When you have finished the first exercise hand in your printout and pick up the second assignment. ‘ When you have finished both exercises complete the final part the evaluation indicating our rso prefergnce. At this time hand in your a stte, the eve net on form, and your second printout. 106 DIALOGUE EVALUAQION FORM Student 0 Section 0 Please provide your own observations on using each dialogue. List any comments or criticisms as well as suggestions for improvements. DIALOGUE I DIALOGUE II ..........OOOQODOOQQOOOOQO0......QOOOOCOOOOOOOQ......ICOQOOOOOOOQQOQO Complete this portion after you have had an opportunity to use both dialogues. Which dialogue would 122 personally prefer? Please circle either a or b. a) DIALOGUE I b) DIALOGUE II Why? 107 DISKETTE NUMBERS 1 THROUGH 8 SSN Section Operating Procedures - DIALOGUE I The programs you need for this task are available from the Grawn Lab personnel. Present this sheet when asking for the diskette. If the 13! PC is not on insert the program diskette in drive A and your diskette in drive B. Turn on the computer. (If the computer is already on, insert the diskettes as above, and reboot the system by pressing CTRL, ALT, and DBL.) The program will ask you for your social security number, name, and section number. The following steps outline what you will be doing with this program. 1. You will be asked to add any new records not already entered. After each record entered you will be asked 'Is IT CORRECT (Y/U)‘. If you reply with a Y the record will be written to the diskette and you will be prompted for the next record. If you reply with an N the record will not be written to the diskette, and you will be prompted to enter the data again. when you have added all the records you may end the ADD routine by typing the word STOP when prompted for employee number. (Complete all entries by pressing the EITBR key.) 2. The system will next sort the file by employee number. This step is necessary when records have not been entered in employee number order. 3. The program will next process any changes you need to make. Enter the -ployee number of the record you wish to change. The record will be displayed on the screen with a prompt message 'CRANGS (Y/R)‘ beside each field. Enter a Y if you wish to change the corresponding field. c. The program will next list the file and provide you with totals of each numeric field. If there are any errors the program will tell you which totals are erroneous and will begin again asking for any additions to the file (in case you forgot to enter a record). If you have no additions, respond with 'STOP' when asked for employee number. 5. The process will continue as outlined in steps 1 through 4 until the totals are correct. 6. Once you have achieved the correct totals you will be presented with questions to answer. Please be as accurate as possible when answering the questions. ' 7. Return the program diskette to the lab personnel. If you this is your first exercise, hand in your printout and pick up the last exercise. If you have completed both exercises, hand in your printout, your evaluation of the two dialogues, and your diskette. 108 Acme Manufacturing Co. Departmental Time Report DATA FOR DIALOGUE I t.iit.........fiiflitfififltifiii.........tifliflfififlfifl....‘ltfiififitfitflfifitififlfififi....It. an" 107000 OLA" 557 01' W 1010 1001.5, 09010 3 1 ® 0 m 1L 0 1015 gncxos. DAVID 5 3 00 2 0.95 0 (:L__ 1020 011312, uses Q) (2 (D 1025 cums, 01031101. 0 3 00 0 (533 0 0 1030 czanxa, STANLEY 1 1 (:2) 0 CID Q a 1035 00510101110, 13an 0 3 10 5 7.05 0 10 1000 0131003101, .3710 3 1 00 7 9.35 0 0 1005 al.2110610, assa g 3 30 o 0.75 g 0 1050 PARNDRICH. IVA! 6 A; 40 9 9.E0 0 O 1055 many nonsm- 0 2 F 30 0 0.70 0' 0 1000 car-ruse, came; 5 1 00 0 7.95 0 0 1005 gusmucz, go: 2 2 30 3 7.05 g 0 1070 900000111, onvzu. 3 3 00 0 5.05 0 0 1075 aavzms, 31m 9 3 45 0 0.05 A 0 1000 0111311110, 01.31010 3 g 0 0 7.95 0 00 1005 3001-, 30101.1»: 0 _1 00 0 0.05 0 0 1090 110mm“, 0m 3 1 30 0 0.95 0 2 1095 Jmczmm. u. 1 1 11 1 9.05 1 20 1100 run-1r, 001mm 0 3 31— 0 7.9g 0 0 1105 111.1011, noun: 0 g 00 0 0.0; 0 0 1110 Lyme, 0mg 0 1: 00 0' 94; 0 0 1115 mgr. scams 3 1 32 0 - 7.95 ' 0 0 W 2 L '40 33.4.1 . <1 o 1125 gmrr. 0m: 5 2 32 0 0.95 0 0 1130 mm rum! 0 3 - 00 0 3.3 0 -W"-—"_" . Tsfi 110'— 1(39 DISKETTE NUMBERS 9 THROUGH 16 SS“ Section Operating Procedures - DIALOGUE II The programs you will need for this task are available from the Grawn Lab personnel. Present this sheet when asking for the diskette. Insert the program diskette in drive A and~your diskette in drive 3. If the IBM is not on, turn it on now. If it is already on, reboot the system by pressing CTRL, ALT, and DEL. The program will prompt you for your social security number, name, and your section number. with this series of programs you will be asked for 'commands’. The valid commands are ADD, CEANGE, DELETE, SORT, LIST, HELP, and END. Each command has options as outlined below. The commands may be used in any order. ADD. Adds a new record to the file. options: Shift“ - Pressing the shift and the ‘ symbol found over the '6' key will allow you to 'back-up' when entering new data. You may use this key sequence repeatedly to back up more than one field. The ADD routine is terminated by typing the word STOP instead of an employee number. . DELETE Delete a record from the file. options: A - Delete all records. ' O - Delete gag record. for example, the first, or the fifth. R - Delete a range of records, such as the third through the sixth. K - Delete a specific record b! providing the record key, that is, t e employee number. 11(3 CEANGE Change fields within record(s). SORT- LIST HELP options: A - The program will present all records 0 - The program will present a specific record within the file. A 3 indicates the third record, etc. R - A range of records will be presented for changing. Por example, you could specify changing the second through the eight records in the file. A - A specific record is presented for changing. The record is identified by its key, i.e., employee number. As each record is presented for your changes the program will ask for the field to be changed. After each change is made you may either change another field or end the changes for the record by typing the word STOP. The valid field names are presented on the screen with the current data. Sort the file to employee number (EHPI). options: A - Sort to ascending order (lowest to highest) D - Sort to descending order (highest to lowest) List record(s) on the screen or the printer. options: Screen or Printer E - Block mode: this mode is the same format as you see when adding or changing. P - Porm listing: in this mode you specify fields you would like on the output. You need not include the fields DEP and SRIPT if you want the data to fit on one line. When you have entered all the fields you need, press the ENTER key. TOTAL uuusnxc FIELDS (Y/N) - Answer with a 'r' t: have totals printed under each numeric f e d. TITLE - an a propriate title for the for e list ng. Provides a list of the valid commands. You must type END when you arrive at the correct ,totals, otherwise the remainder of this exercise will not be completed and it will be necessary to redo the entire exercise. ' 111 Acme Manufacturing Co. Departmental Time Report DATA FOR DIALOGUE II itt.Oitfiflfliflfitfiiiifiiitflififlti.itIit.................fiififififiiflfiitflfliit...it.it. 0000 0000 o_0_L gm- 5130 or am szcx vac 1010 ABLE. Jog! 3 1 35 9 0.07 1 15 0 1015 nexus, DAVID 5 3 £0) 2 m 0 0 1020 0mm, 00003 g 2 35 1 9.50 0 (0) 1025 cnzws ' 0100101. (ID 3 KID (7) 63—5—03 (0) (0) 1030 CZARKA srnnnrr 1 1 33 0 (57355 0 0 1035 000100010, 0001' 0 3 (:0) 0 C7753} CE (Q 1000ggprsowsxx. JAN 3 1 - 37 0 50.9 0 0 1005 0121003. ABBA g 3 39 0 0.00 1 0 1050 rnannaxcn. xvgggvgr 0 g 0 0 9.09 0 0 1055 PRARY nonrnr 0 2 37 0 9.09 0 0 1000 gggrggo, snags; 5 1 0 0 7.07 7 0 1005 chasrrwzcz, 000 2 g_3 30 3 7.30 3 0 1070 onoossxx, ORVAL 3 3 30 0 0.50 0 0 1075 0.01.3.5. sans 9 3 52 15 0.50 0 0 1000 EILLHAN, GLENN 3 2 25 0 5.00 15 0 1005 0000, wrnrrau 0 1 0 0 7.95 0 00 1090 inggzurn, 0010 3 1, 33 0 3.05 0 .5 1095 annczrusrr. AL 1 1 22 g 5.09 1 21 1100 0000002, 00101000 0 3 23 0 0.05 0 0 1105 101.100, 0000111 0 i 39 7 5.35 0 0 1110 errrs, 00000 0 g; 00 0 9.05 0 0 1115 gages. 0000003 3 1 23 0 5.97 0 0 1120 nosrnssx, MARE 7 g!g__ 0 A; 7.30 0 0 W 5 ,3' 31 3 i1”, 7 ° W! 4 3 30310 ' ...: o 0 TOTALS - . 3 ' ’ 000 99 100.53 09 100 APPENDIX C 112 APPENDIX C OVERHEADS USED TO ILLUSTRATE ALL SCREEN DISPLAYS A>DATE Current Date is Tue 1-01-1980 Enter new date: 2-6-83 Screen 1 - Dialogue I and II PRESS 'CAPS LOCK' KEY - THEN ENTER PRESS 'NUM LOCK' KEY -THEN ENTER Screen 2 - Dialogue I and II SOCIAL SECURITY NUMBER (NO HYPHENS): 364500717 LAST FIRST NAME (NO COMMAS):? DOE JOHN SECTION NUMBER :? 5558 IS THE ABOVE CORRECT (Y/N)? Y Screen 3 - Dialogue I and II 113 ACME MANUFACTURING CO. WEEKLY PAYROLL - ADDITIONS If you make an error while entering data, note the error on paper and continue entering the record. EMPLOYEE NUMBER NAME (DOE, JOHN) DEPENDENTS (0-9) SHIFT (1,2,0R 3) REGULAR HOURS OVERTIME HOURS PAY RATE SICK HOURS VACATION HOURS Screen 4 - Dialogue I 114 THE SYSTEM IS NOW SORTING RECORDS TO EMPLOYEE NUMBER Screen 5 - Dialogue I ACME MANUFACTURING CO. WEEKLY PAYROLL - DELECTIONS Check your listing for nay invalid employee numbers. Enter the invalid number to delete the record, or enter 'STOP'. ENTER EMPLOYEE NUMBER: :----: Screen 6 - Dialogue I 115 ACME MANUFACTURING CO. WEEKLY PAYROLL - CORRECTIONS Enter employee number of record to change, or the word STOP Employee # :1015: BACKUS, DAVID *********************************************************** Name : ABLE JOHN Change (Y/N) :N: Dependents : 3 Change (Y/N) :N: Shift : 1 Change (Y/N) :N: Regular hours : 2.00 Change (Y/N) :N: Overtime hours: 6.00 Change (Y/N) :N: Pay rate : 5.69 Change (Y/N) :Y: Sick hours : 12.00 Vacation hours: 7.00 Current Data: 12.00 Change to : ? Screen 7 - Dialogue I 116 THE THE THE THE THE ACME MANUFACTURING CO. WEEKLY PAYROLL - VERIFICATION TOTAL REGULAR PAY IS NOT CORRECT. TOTAL OVERTIME PAY IS NOT CORRECT. PAY RATE HASH TOTAL IS INCORRECT. TOTAL SICK HOURS IS NOT CORRECT. TOTAL VACATION HOUR IS INVALID. When the printer has finished you may add, change, or delete as necessary to correct the errors. Screen 8 Dialogue I 117 ACME MANUFACTURING Payroll Data Entry Dialogue II The valid commands are : ADD CHANGE DELETE LIST SORT END Note that the CHANGE, DELETE, and LIST require that you specify whether all records, a specific record, or a range of records are to be used. FILE - EXPERZ KEY FIELD - EMP# # OF RECORDS - 10 COMMAND OR END - : ---------- : Command Screen - Dialogue II 118 FILE - EXPERZ KEY FIELD - EMP# # OF REOCRDS - 10 COMMAND -:ADD ------- : ( 11 ) EMP# < 4 >:1060: NAME < 15 >GATTUSO, JANICE: DEP < l >:5: SHIFT < l >:2: REGHRS :? 0 OTHRS :? 4 RATE :? 7.67 SICKHRS :? 7 VACHRS :? 6 Are data entries correct (Y or N)? Y Add Screen - Dialogue II FILE - EXPERZ KEY FIELD - EMP# # OF RECORDS - ll COMMAND -:DELETE----: Enter mode: A)ll, O)ne, R)ange, K)ey - K *** KEY FILE MUST BE SORTED *** Enter key value 1060: Delete Screen - Dialogue II 119 FILE - EXPERZ KEY FIELD - EMP# # OF RECORDS - ll COMMAND -:CHANGE----: Enter mode: A)ll, O)ne, R)ange, K)ey - K *** KEY FILE MUST BE SORTED *** Enter key value 1060: ( 11 ) EMP# :1060 NAME :GATTUSO, JANICE DEP :5 SHIFT :2 REGHRS : 40 OTHRS : 4 RATE : 7.67 SICKHRS : 7 VACHRS : 6 Enter name of field - REGHRS REGHRS :? Change Screen - Dialogue II 120 FILE - EXPER2 KEY FIELD - EMP# # OF RECORDS - ll COMMAND -:LIST ------ : Listing on printer (Y or N)? Y Enter mode: A)ll, O)ne, R)ange, K)ey - A B)lock or F)orm listing? F EMP# NAME DEP SHIFT REGHRS OTHRS RATE SICKHRS VACHRS Select fields you wish for form Enter name of field - EMP# Enter name of field NAME Enter name of field REGHRS Enter name of field OTHRS Enter name of field RATE Enter name of field SICKHRS Enter name of field VACHRS Enter name of field STOP Total numeric fields (Y or N)? Y Forms title - LIST OF DATA USED IN DIALOGUE II List Screen - Dialogue II APPENDIX D APPENDIX D QUESTIONS PRESENTED AT THE END OF EACH DIALOGUE SESSION How would you rate the operations allowing additions to the file? a) b) c) d) e) easy to use moderately easy to use neutral difficult to use very difficult to use How would you rate the operation allowing changes to the file? a) b) c) d) e) easy to use moderately easy to use neutral difficult to use very difficult to use How would you rate this program overall in terms of ease of use? a) b) c) d) e) easy to use moderately easy to use neutral difficult to use very difficult to use How did you feel when using the add routine? a) b) c) d) e) very comfortable comfortable neutral frustrated very frustrated How did you feel when using the change routine? a) b) c) d) e) very comfortable comfortable neutral frustrated very frustrated How did you feel about the overall system? a) b) c) d) e) very comfortable comfortable neutral frustrated very frustrated 121 122 7. How fast do you type? a) b) c) d) e) hunt and peck know where to place hands but have to look at keyboard frequently 10 to 30 words per minute 30 to 50 words per minute above 50 words per minute 8. How would you rate this system? a) b) c) d) e) user-friendly almost user-friendly neutral not very user-friendly not at all user-friendly SELECTED BIBLIOGRAPHY SELECTED BIBLIOGRAPHY Boillot, M. Computer communication modes and their effect on student attitudes towards programming. Nova University thesis, April 1974. (ERIC Reproduction Service No. ED 098 957. Broadbent, D. E. & Broadbent, M. H. P. "The allocation of descriptor terms by individuals in a simulated retrieval system.” Ergonomics, 1978, 21, 343-354. Brosey, M. & Schneiderman, B. “Two experimental comparisons of relational and hierarchical database models.“ International Journal of Man-Machine Studies, 1978, 10, 625-637. Card, S. & Moran, T. ”The keystroke level model of user performance time with interactive systems.‘I Communications of the ACM, July 1980, 2;, 7, 396-410. Codd, E. F. "A data base sublanguage founded on the relational calculus.“ Proceedings, ACM SIGFIDET, Workshop on Data Description, Access, and Control, 1971, pp. 35-68. Conrades, G. H. ”Information Processing - Dollars to Cents". Information Processing January 1982, l, 1’ 166-16d. Crowne, D. P. & Marlowe, D. The Approval Motive. Wiley, New York, 1964. Control Data Corporation XEDIT VERSION 3 REFERENCE MANUAL, 60455730, Publications and Graphics Division, ARH219, St. Paul, 1981. Date, C. J. An Introduction to Database Systems (2nd ed.). Addison-Wesley, 2nd edition, Reading, Massachusetts, 1977. Dehning, W., Essig, H., & Maass, S., ”The Adaptation of Virtual Man-Computer Interfaces to User Requirements in Dialogs.“ Lecture Notes in Computer Science, 1981, 110, Springer-Verlag, Berlin. Dijkstra, E. W. ”GO TO statement considered harmful." Communications of the ACM, 1968, 11, 147-148. Dock, V. & Essick, E. Principles of Business Data Processing with MIS (4th ed.). Science Research Associates, Chicago, Ill., 1982, pp. 205, 350, 473. 123 124 Gold, M. M. "Time-Sharing and batch processing: An experimental comparison of their values in a problem-solving situation." Communications of the ACM, May 1969, 13, 5, 249-259. Goodman, T. & Spence, R. I'The effect of system response time on interactive computer aided problem solving.“ ACM SIGGRAPH '78, Conference Proceedings, 1978, pp. 100-104. Gore, M. & Stubbe, J. Elements of Systems Analysis (2nd ed.). Wm. C. Brown, Dubuque, Iowa, 1979, p. 255. Green, T. R. G. ”Conditional program statements and their comprehensibility to professional programmers." Journal of Occupational Psychology, 1977, 50, 93—109. Greenblatt, D. & Waxman, J. 'A study of three database query languages." In B. Shneiderman (Ed.), Databases: Improving Usability and Responsiveness, Academic Press, New York, 1978, pp. 77-97. Grossberg, M., Wiesen, R. A., & Yntema, D. B. ”An experiment on problem solving with delayed computer responses." IEEE Transactions on Systems, Man, and Cybernetics, SMC-6, March 1976, 3, 219-222. Halstead, M. “Elements of Software Science.“ Operating and Programming Systems Series, Elsevier Computer Science Library, New York, 1977. Hansen, J. V. “Man-machine communications: An experimental analysis of heuristic problem-solving under on-line and batch-processing conditions.“ IEEE Transactions on Systems, Man and Cybernetics, November 1976, 6, 11, 746-752. Kelly, T. L., "The Selection of Upper and Lower Groups for the Validation of Test Items.“ J. Educ Pschol., January 1939, pp l7-24. Lee, J. & Schneiderman, B. “Personality and programming: Time-sharing vs. batch processing.“ Proceedings of the ACM National Conference, 1978, pp. 561-569. Lochovsky, F. H. & Tsichritzis, D. C. ”User performance considerations in DBMS selection.“ Proceedings of the ACM SIGMOD, 1977, pp. 128-134. Lochovsky, F. H. Data base management system user performance. Ph.D. dissertation, Univ. Toronto, Canada, 1978. 125 Manuel, T. ”Japan Maps Computer Domination." BYTE, May 1982, pp. 140-144. Martin, J. Applications without programmers. Prentice-Hall, Inc., Englewood Cliffs, N.J., 1982. Miller, L.A., & Thomas, J.C., “Behavioral Issues in the Use of Interactive Systems.” Proceedings of the 6th Informatik Symposium of IBM Germany on Interactive Systems, Lecture Notes in Computer Science, 1976, 42, Springer-Verlag, Berlin, pp 193-215. Miller, R. B. ”Response time in man-computer conversational transactions.“ Proceedings Spring Joint Computer Conference 1968, 33, AFIPS Press, Montvale, New Jersey, 267-277. Rathus, S. "A 30-item Schedule for Assessing Assertive Behavior." Behavior Therapy, 1973, 4, 398-406. Reisner, P. "Use of psychological experimentation as an aid to development of a query language." IEEE Transactions on Software Engineering, SE-3, 1977, 3, 218-229. Reisner, P. ”Human Factors Studies of Database Query Languages: A Survey and Assessment." Computing Surve 3, March 1981, 13, 1, 13-31. Reisner, P. Boyce, R. F., & Chamberlin, D. D. “Human factors evaluation of two data base query languages: SQUARE and SEQUEL.‘ Proceedings of the National Computer Conference, AFIPS Press, Montvale, New Jersey, 1975. Ritchie, M. "Human Factors in the Long Run." Human Factors April 1970, $3, 153-157. Rotter, J. B. "Generalized Expectancies for Internal vs. External Control of Reinforcement.” Psychological Monographs, 1966, pg, 1, 1-28. Sackman, H. “Experimental analysis of man-computer problem-solving.” Human Factors, 1970, 12, 187-201. (a) Sackman, H. Man-Computer Problem Solving, Auerbach Publishers Inc., Princeton, New Jersey, 1970. (b) Schatzoff, M., Tsao, R., & Wiig, R. "An experimental comparison of time-sharing and batch processing." Communications of the ACM, May 1967, $9, 5, 261-265. 126 Sheil, B. A. ”The Psychological Study of Programming.” Computer Surveys, March 1981, 13, 1, 101-120. Sheppard, S. B., Curtis, B., Milliman, P., Borst, M. A., & Love, T. “First year results from a research program on human factors in software engineering.” Proceedings of the National Computer Conference, 1979, 48, AFIPS Press, Montvale, New Jersey, 73-79. Shneiderman, B. “Exploratory Experiments in Programmer Behavior." International Journal of Computer and Information Sciences, June 1976, 5, 2, 123-143. Shneiderman, B. "Human Factors Experiments for Developing Quality Software.“ State of the Art Report on Software Reliability, 1977, (Berkshire, England: Infotech). Shneiderman, B. "Human Factors Experiments in Designing Interactive Systems." COMPUTER, December 1979, pp. 21-38. Shneiderman, B. Software Psychology - Human Factors in Computer and Information Systems. Winthrop Publishers, Inc., Cambridge, Massachusetts, 1980. Sime, M. E., Green, T. R. G., & Guest, D. J. “Psychological evaluation of two conditional constructions used in computer languages." International Journal of Man-Machine Studies, 1973, 5, 1, 105-113, 123-143. Sime, M. E., Green, T. R. G., & Guest, D. J. “Scope marking in computer conditionals - a psychological evaluation." International Journal of Man-Machine Studies, 1977, 2, 107-118. Sondheimer, N. 'On the fate of software enhancements." Proceedings of the National Computer Conference, 1979, 42, AFIPS Press, Montvale, New Jersey. Stern, R. A. & Stern N. An Introduction to Computers and Information Processing, 1982, John Wiley & Sons, New York. Walther, C. E. & O'Neil Jr., H. F. "On-line user-computer interface - the effects of interface flexibility, terminal type, and experience on performance.“ Proceedings of the National Computer Conference, 1974, 43, AFIPS Press, Montvale, New Jersey. Weinberg, G. M. The Psychology of Computer Programming, Van Nostrand Reinhold, New York, 1971. 127 Weissman, L. ”Psychological complexity of computer programs: An experimental methodology.“ ACM SIGPLAN Notices, 1974, 2. Welty, C. "A Comparison of a Procedural and a Nonprocedural Query Language: Syntactic Metrics." Ph.D. Dissertation, University of Massachusetts, 1979, University Microfilms International, Ann Arbor, Mich. Winer, B. J. Statistical Principles in Experimental Design, McGraw-Hill, New York, 1971. Winship, B. J. & Kelly, J. D. “A verbal response model of assertiveness." Journal of Counseling Psychology, 1976, 23, 3, 215-220.