INTRODUCTION

The United States is the most technologically advanced nation in the world. However, despite this, the United States ranks only fifth in international measures of productivity per person. The incongruence, at least in part, may be explained by workplace stress (9). While eustress, a "positive stressor," supplies motivation, job satisfaction and productivity to a worker, adversary stress experiences often have negative workplace outcomes. One occupation population regarded with a high level of workplace stress is application programmers, a subset of information systems professionals (6).

Programmer stress is both extremely common and extremely problematic. Fujigaki (6) found stress present during almost all software life cycle phases. More specifically, the stress took on different forms in two different phases. During the "requirements definition and design" researchers found stress in the form of high anxiety and depression, while during the coding phase, the stress took on the form of irritability and falling morale.

Furuyama (6) found that programmers operating under stress make far more mistakes than they would under more normal circumstances. Thirty-seven percent of mistakes could have been avoided by appropriate scheduling and by eliminating unrealistic deadlines for the developers. The study concluded that 24% of all design faults were directly attributable to programmer stress and that "deep thinking, such as searching for solutions in a huge problem space...is required in the design phase... [and] is easily affected by stress, which causes imperfect investigation" (6, p. 18).

A vast number of software engineers are experiencing occupational health problems because of the heavily increased market demand of software development (4). Fujigaki and Mori (4) conducted a longitudinal study of work stress among information system professionals measuring both physiological and psychological responses to work stress. The specific work stressors included communication problems, technical difficulties and ambiguity in specification. The researchers describe an "exhaustion after overwork" phase in which cortisol levels increase not at the time of busyness, but further along an overwork continuum. This study is unique in its contribution to stress research as it considers the specific task of the software engineer and the stress cycle along a time-work continuum.

A prominent psychological equivalent in the stress response is anxiety. Anxiety is described by Spielberger (12) as an unpleasant emotional experience, and/or as a personality trait which is a relatively stable individual difference in personality. Further, state anxiety is a measurement of emotional reaction that varies from one situation to another, while trait anxiety is the amount of anxiety generally experienced by an individual.

Music, an aid to mood change, may be a valuable tool for anxiety reduction and, in turn, an aid in productiveness. Affect modification through a music stimulus has been indicated as a unique property of music therapy (14). Incorporating the affective/motivational properties in music perception into the work process could provide affective change in computer programming work.

Music may also be effective in the relaxation response when paired with other stimuli. A study by Byrnes (3) compared the effects of audio, visual, and paired audiovisual stimuli on the experience of stress. The audio condition consisted of SaintSaens' "Aquarium" from Carnival of the Animals, the video condition was a moving tropical seascape, and the third condition combined both audio and visual. Subjects' stress responses were recorded through a continuous response digital interface (CRDI) throughout each condition. A statistical significance was found between beginning and ending stress response measurements for the combined audiovisual stimulus.

Studies of workplace productivity indicate an increase of about 5% after background music systems have been installed. Performance improvement is explained firstly by mood enhancement and secondly by a masking effect. Music seems to boost enthusiasm, increase relaxation and lessen nervousness, and, as a result, the elevated mood contributes to higher productivity. Background music also masks distracting sounds such as extraneous conversations and machine sounds. The effect of the masking is that employees concentrate better and, as a result, are more effective and efficient workers (10).

Presenting extraneous environmental stimuli by way of stereo headset is one way in which work performance has indicated further improvement. One hundred and fifty employees in clerical and administrative jobs were volunteers for a study in which music was administered through personal headsets. The research indicated an average of 10% increase in productivity for the individuals using the headsets. More specifically, the productivity in the more "simple" jobs, such as data entry, had an increase of 14%. The more "complex" jobs, such as account analysis, increased on average by 6%. Oldham, (11) suggests that familiarity of music and discontinuous presentation are most effective in improving performance, turnover intentions, organization satisfaction, and mood states.

Blood and Ferriss (1) hypothesized an increase in interaction and group conversations due to background music. The findings indicated greater satisfaction with communication in a group setting with background music. The researchers also found that the modality of the music was an important component when considering the effects of background music on human interaction. For example, greater productivity and greater satisfaction with communication occurred for the groups hearing music in a major key versus a minor key.

While studies of music and productivity have focused on simple or monotonous tasks, there is little literature on the effects of music for productivity and stress reduction in a complex task. A change from a negative to a positive mood may aid in stress reduction and creative thinking that is more spontaneous. A study of responses to music through measurement of anxiety and task effectiveness in the information systems field would provide valuable research findings.

The purpose of this study was to test the effect of music listening on students' stress levels and logic and syntax error rates in a computer programming task.

Hypotheses

H1: Listening to music will not affect the logic error rate of a computer program task.

H2: Listening to music will not affect the syntax error rate of a computer program task.

H3: Listening to music will not affect the state anxiety levels of a computer science student.

METHOD

Subjects

Volunteers were 72 undergraduate students (male=58, female=14) enrolled in an "Introduction to Programming" course 60-141 at University of Windsor. The average age of the students was 24, the average year level was 1.79, with Asian (50%) and Caucasian (29%) as the predominant ethnicities. Students enrolled in the course were given full credit for the computer programming assignment whether or not they participated. Subjects were assigned to three music conditions (Control, Primer, and Periodic) by computer laboratory sections.

Materials

Anxiety levels were self-reported through a ten-item version of the Spielberger State Anxiety Inventory (13). The State Anxiety Inventory Form (STAI) included ten statements requiring a response on a four-point Likert scale (Appendix). Inventories were scored by summing the weighted scores for the ten-item questionnaire (12).

Eleven minutes of music played for both the primer and periodic music condition included the 3rd movement from Brahms' Symphony #1, and the I pini del Gianicola from Respighi's Pines of Rome. The periodic music condition, in addition to the above listed 11 minutes of music, included Haydn's Concerto for Cello and Orchestra No. 1 in C, Adagio; Sibelius' Swan of Tuonela; and Villa-lobos' Bachianas Brasileiras, No. 5.

Two computer programming tasks, designed by the professor of the programming course, were administered. The first 15-minute task required the student to locate and correct syntax errors and the second 20-minute task required the student to locate and correct logic errors in a given C language program. The computer science professor and his teaching asSiStantS evaluated students' programming achievement following completion of each task. The maximum possible score for each task was five points.

Students in the music conditions were asked to rate their enjoyment of the music on a four-point Likert scale.

Design and Procedure

The experiment took place over one week and involved participants from 11 computer laboratories. The experimental design addressed between and within comparisons of state anxiety from three conditions: control, primer and periodic, as well as between group comparisons of syntax and logic scores. The control group did not receive music, the primer condition received 11 minutes of music prior to the programming, and the periodic condition received 11 minutes of music prior to the task as well as periodically throughout the programming assignments.

At the start of the experiment students were given computer laboratory assignment instructions in regard to the upcoming programming task. Instructions were given from either the computer science professor or the teaching assistant. Immediately following the computer programming instructions all students completed the state anxiety inventory. The control group, upon completing the inventory, then began the computer task.

Subjects in the primer music condition prior to music listening received the following instructions from the researcher.

The music listening will be 11 minutes in length. You will hear the third movement of Brahms' Symphony #1 and the 3rd movement of Respighi's Pines of Rome. Simply make yourself comfortable and focus on and follow the sounds to the best of your ability.

Immediately following the 11 minutes of music listening both the primer and periodic groups completed for a second time the state anxiety inventory questionnaire. They were also asked to rate their enjoyment of the music. The control group simply continued from the initial state anxiety measurement to the computer programming task. Subjects then completed each programming task while the computer science teaching assistants and the computer science professor evaluated each student's achievement. Upon completion of the computer task, the student indicated, for the third time, his/her responses to the state anxiety inventory questionnaire.

A data analysis package, SPSS, was used to obtain descriptive statistics, ANOVA and repeated measures and analyses. The ANOVA was used to test for differences in state anxiety levels, and syntax and logic scores between groups. The repeated measures analysis of variance was used to test for within subject differences for three state anxiety levels. An alpha level of .05 was used for all statistical tests.

State Anxiety Levels

This study attempted to determine if there were any differences between control and music conditions (primer and periodic) in state anxiety levels. No statistically significant difference was found for state anxiety no. 1, the state anxiety measurement administered at the start following the initial instructions. Means and ANOVA results of the first state anxiety inventory are presented in Tables 1 and 2.

Subjects in the primer and periodic conditions received for a second time the self analysis questionnaire. They responded to the ten-item questionnaire immediately following 11 minutes of music listening. The control group did not receive music and therefore their state anxiety no. I scores served for mean comparison to the music groups' state anxiety no. 2 scores. As illustrated in Table 2, statistically significant differences in state anxiety no. 2 means were found between the groups. Further analysis using the Scheffe Post Hoc Test reveals statistically significant differences between the means of the control and primer group, and the control and periodic group (Table 3, Figure 1). As differences in state anxiety have been indicated, Hypothesis no. 3 is rejected.

Subjects in all conditions received once more the self analysis questionnaire following the completion of the computer task. As illustrated in Table 2, a statistically significant difference in state anxiety no. 3 means was found between groups. Further analysis using the Scheffe Post Hoc Test (Table 4 and Figure 2) reveals a statistically significant difference between the means of the control and the periodic group. Once again, differences in state anxiety have been indicated and therefore Hypothesis no. 3 is rejected.

A repeated measures analysis was employed to evaluate differences in state anxiety for each group over time. Pillars Trace Multivariate Test revealed a statistically significant difference within group state anxiety levels (Table 5).

Further analysis by use of a paired samples test indicated there was no statistically significant difference between the first and last state anxiety level for the control group. There were however statistically significant differences between two state anxiety levels for the primer group as indicated in Table 7. The first and last state anxiety level as indicated by pair 2 in the paired samples test (Table 7) approached statistical significance.

A statistically significant difference was indicated by Pillars Trace Multivariate Test for differences between state anxiety levels within the periodic music condition (Table 8). The paired sample test revealed a statistically significant difference between the first and second level and a near difference between the second and third state anxiety level (Table 9, Figure 3).

Enjoyment of Music

Subjects in the primer and periodic music conditions rated their enjoyment of the music on a scale from 1 (not at all) to 4 (very much). The mean for both groups was 2.96, an average closest to "moderately" on the rating scale (Table 10). Comments from the students in response to the music listening and the computer task were also recorded.

Syntax and Logic Task Achievement

Syntax and logic task achievement were evaluated by computer "experts" on a scale of 1 to 5; 5 given to those whose corrected program could be executed with complete success. While there were no statistically significant differences between the conditions for both syntax and logic task achievement (Table 11), the periodic condition scored the highest means for both tasks (Table 12).

DISCUSSION The use of music listening prior to and during computer programming tasks appears to be beneficial for decreasing anxiety for student programmers. This finding is consistent with other studies investigating the role of music in reducing stress (3, 10, 11). It appears that music, when compared to no intervention, is effective in reducing anxiety levels. Further, the paired samples test revealed that the periodic music group exhibited the least amount of anxiety throughout the computer programming task. It appears then that the length of music listening aids in anxiety reduction. This is also demonstrated by comparing the effect of the music between the primer and periodic music group. The primer music group reduced in anxiety immediately following the music listening, but by the end of the computer task the students' anxiety levels escalated beyond the initial state anxiety level. While the anxiety did not climb to the level experienced by the control group, it still increased without the continuation of the music listening.

There were no statistically significant differences found between groups in the syntax and logic scores. However, it is interesting that the periodic music group scored the highest means in both tasks. This outcome may present a pattern that would become more evident if the study was carried out over several sessions of music listening and also with a greater number of subjects.

The music chosen for this study were selections listed in a Guided Imagery and Music (GIM) program. The intent of the music used in the GIM program is to aid participants in imaging and to elicit feelings and memories (2). Bonny's philosophy in choosing music for the GIM process is based on the assumption that "there seems to be a consensus among listeners about performances which activate a deep feeling response and those which do not" (2, p. 58).

It is interesting to note that the rating of enjoyment (mean of 2.96 out of a possible 4) was the response from a student group which included an ethnicity of 50% Asian. The music was from a western classical tradition and was moderately liked by the majority of the students. The comments of the students were helpful for further insight into the music listening experience. While some students enjoyed the music listening "very much" and expressed a decrease in anxiety, others expressed being distracted, as well as needing other types of music for listening. Perhaps the use of headsets as suggested by Oldham (11) would have been helpful in meeting these individual listening needs.

The professional systems designer and programmer have a societal responsibility to make software "user friendly." As programmers gain training, they come to know what is meant by "intuitive" or "user-friendly." The expert programmer must know how the nature of a program impacts on the collaborative process as used in work groups (8). In the final stage of developing systems, the product the programmer produces for the end-user must improve a company's productivity and competitive position.

The societal responsibility to produce a "user friendly" system may be impeded by work stress, programming errors, and low job satisfaction. Finding new ways to improve motivation, productivity and creativity may be an area in which music can contribute. Glass (6) suggested that finding solutions through a sociological approach to the productivity problems may be more helpful than the purely computing-oriented research approach of the past.

While the cited literature findings address the expert programmer population, the human components of stress and productivity are also significant for the beginning programmer. The main difference between the expert and novice programmer is that, in the introductory course on computer programming, the student's concentration is "directed toward mastering the features of the programming language used, rather than on wider aspects of the programmer's craft" (9).

This study addressed, through the environmental musical stimulus, the measurement of anxiety and task performance. While environmental music studies have been applied to work settings involving monotonous tasks, this study has addressed the effect of music on a more complex task such as computer programming. The application of music psychology to the process of computer programming can provide practical contributions to academic and organizational settings.