Abstract: Since the concept of dynamic capabilities was first introduced by Teece and his collaborators (1997) many studies have been conducted to further clarify the concept and its underpinnings. Zollo and Winter (2002) defined a dynamic capability as “a learned and stable pattern of collective activity through which the organization systematically generates and modifies its operating routines in pursuit of improved effectiveness”. This definition distinguishes dynamic capabilities from operational capabilities, but suggests that, like operational capabilities, dynamic capabilities consist of patterned organizational behaviors, i.e. organizational behavioral routines. The question is that if so what the difference between them is. The present paper proposed that the stable pattern in and the determinant of dynamic capabilities were not behavioral routines but cognitive representations. Specifically speaking, on the one hand proper cognitive representations contributed to the improvement of dynamic capabilities; on the other hand behavioral routines without proper cognitive representations impeded the development of dynamic capabilities.
In order to test the propositions a game called “Transform The Target” with 9 cards for 3 players (the original game) and its variant (the variant game) were introduced as experimental tasks. An experimental design with 2 (with or without cognitive representations) × 2 (original game or variant game) between-subject factors was adopted. A total of 240 students participated in the experiment, every 3 students formed a team to play two rounds of the game, each of which consisted of 20 hands and must be completed within 25 minutes. According to whether they were informed of the problem space or not and whether they were asked to play the original or variant game in the first round, the 80 teams were divided into 4 blocks and were asked to play their respective games. In the second round all teams were asked to play the variant game. The average move time per hand and the strategy or path in the problem space they took were used as the indicators to measure behavioral routines. The rate of transfer from the capabilities learned from the original game in the first round to the variant game in the second round was used as the indicator to measure dynamic capabilities. The analyses of the whole processes of some typical hands of the games played by some typical teams were also used to gather the detailed information of behavioral routines. The numerical data were analyzed by the use of repeated measures of ANOVA.
The results showed that while proper cognitive representations contributed to the improvement of dynamic capabilities, behavioral routines without proper cognitive representations impeded the development of dynamic capabilities. To put it briefly, the determinant of dynamic capabilities was cognitive representations but not behavioral routines. These results marked a breakthrough in the study of dynamic capabilities, wherein the orthodox view was that the stable pattern in and the determinant of dynamic capabilities were behavioral routines. This breakthrough opened up new areas for the academic studies of dynamic capabilities, for example, the studies of the nature, function, structure and measurement of dynamic capabilities. And it also provided new thinking for the firms and other organizations in the real world to develop their dynamic capabilities.

Key words: dynamic capabilities, cognitive representations, behavioral routines, collective problem solving, three-player TTT game