Verbal terms have become increasingly convenient and prevalent nowadays as they can be easily used by experts (decision makers) for decision preferences and reveal the same affectivity in terms of number for decision making. Therefore, the verbal models have been applied frequently to decision systems. This paper has proposed available conditions and properties of aggregation function for verbal models from the perspective of social behavior (i.e. averaged location and polar evaluators of decision makers in group decision making). A multi-person verbal model is then established and aggregation function is also aggregated directly by verbal terms according to the above conditions and properties. The proposed model focuses on behavior of decision makers and proposes a systematic process to express the impact of these behaviors on the weight of verbal terms, which cannot be analyzed by any statistical methods.