Abstract: Most studies of routines had been based on field observation until the invention of Transform The Target, a game with 6 cards for 2 players (two-player TTT game for short) created by Cohen and Bacdayan (1994). Since then four experimental studies have been conducted to explore in laboratory the emergence and the measurement of routines in two-player TTT，the influence of procedural knowledge on the emergence of routines, and the relation between incentives, routines and performance. In this study a game with 9 cards for 3 players (three-player TTT game) was developed to investigate the emergence of routines in multilateral (not merely bilateral) interaction and the relation between knowledge (including declarative and procedural knowledge), routines and performance.
Based on this three-player TTT game an experimental design with one between-subject factor (knowledge) and one within-subject factor (time) was adopted. A total of 96 students participated in the experiment, every 3 students formed a team to play the game, so there were altogether 32 teams, in which 16 teams were informed of the problem space (called groups with schema) whereas the other 16 teams not informed (groups without schema). Every team in each group was asked to play 40 hands of the game in 45 minutes. There are 3 main strategies or paths in the problem space to successfully play the game. The difference between the adoption times of the path adopted most frequently and that of the path adopted least frequently divided by the total adoption times was one indicator of the degree of routinization. The average move time of a hand and the variation across teams in the number of moves required to complete the hand in each group were the other indicators of the degree of routinization, and the decreasing rates of them the indicators of the speed of routinization. Each hand a team successfully completed earned it 150 points, each move the team made cost it 10 points, the value of the net points divided by the time needed to complete the hand was used as the indicator of performance. Repeated measures ANOVA and linear regression were used to analyze the data.
The main results of this study were as follows: first, routines did emerge in three-player TTT game; secondly, while the degree of routines of the group with schema was lower on the whole than that of the group without schema, the increasing rate in the former is faster than that in the latter; and thirdly, the performance of the group without knowledge is higher than that of the group with knowledge, but the difference is significant at the earlier stage whereas it is not at the later stage.
The above results indicated that the problem space given to the subjects who had not played this game before was only declarative knowledge to them at the beginning of the experiment. It would take a period of time for them to memorize this declarative knowledge. At the behavioral level only through the repeated play of the game this declarative knowledge would help facilitate the formation of procedural knowledge. At the neural level this would require their hippocampal engagement and compete with the striatum for the control over learning. But when the declarative knowledge was memorized it would help the procedural learning and hence increase the level of proficiency. Generally speaking, the declarative and procedural knowledge jointly (not the latter alone) contributed to the emergence of organizational routines, and organizational routines in turn contributed to the organizational performance

Key words: collective problem solving, three-player TTT game, knowledge, routines, performance