Deleterious effects of intermittent interruptions on the task performance of knowledge workers: A laboratory investigation.

Contemporary businesses compete in a highly reactive marketplace that demands a new breed of sophisticated knowledge workers. Managing these valuable people requires understanding the effects of their work environment on productivity. Frequent interruptions are an integral part of the knowledge worker's day. In this laboratory study, two attributes of interruption, frequency and length, are examined in conjunction with two levels of task complexity. A 2 x 2 x 2 factorial design with a control group for each level of task complexity resulted in ten unique treatments. Using personal computers, 122 student subjects from undergraduate courses in production management, took a practice, multiple-choice examination (the primary task) over course material. All subjects were allowed exactly 45 minutes to work on the primary task. Interruption episodes of trivia questions were generated at random intervals. Subjects were interrupted either two or six times with interruption lengths of either 30 seconds or 120 seconds. Performance on the primary task is measured using a point scoring system. A post experiment questionnaire was used to validate experiment manipulations. Using a multiple regression approach to analysis of covariance, approximately 70% of variance in performance is explained. Final models include a covariate for prior classroom performance. Four significant conclusions emerged from the analysis: (1) In the high complexity version of the primary task, short interruptions result in an average performance reduction of 44% relative to control subjects. (2) For the low complexity task, long interruptions result in an 11% average performance improvement over control subjects. (3) Performance in long interruption treatments is significantly better than performance under short interruption treatments for both levels of task complexity. (4) No consistent effects for frequency of interruption occur at the levels used in the study