Productivity, satisfaction, leadership and interaction: A comparison of two communication networks in a software development team context, a communicative perspective on chief programmer and egoless programming teams

Communication textbooks, when discussing communication networks, almost always include the following statement: centralized networks work best for simple problems; decentralized networks work best for complex problems. While the literature supports the claim that centralized networks work best for simple problems, the literature is equivocal regarding complex problems.;In recent years research on teams developing computer software has again raised the question, "what type of communication network will lead to greatest productivity?".;To help answer the question, this experiment revisited the classic research using computer-mediated communication.;Eighteen teams of four programmers each created a program to do polynomial arithmetic. Half the teams communicated in centralized networks, in which all communication went through a "chief programmer," and half the teams communicated in decentralized networks, in which everyone was able to talk with everyone else. All interaction took place via computer.;Dependent variables included productivity and satisfaction. In addition, interaction was analyzed using a functional approach.;Results of this experiment contradicted the often quoted maxim: decentralized networks are best for complex problems. There was no significant difference with respect to productivity or satisfaction between centralized and decentralized teams. In addition, interaction analysis showed that more metatask utterances correlated with poorer productivity.