Joyce Berg, John Dickhaut, and Kevin McCabe in "Trust, Reciprocity, and Social History," Games and Economic Behavior, , 1995, pp. 122-142 present an experiment that studies trust and reciprocity in an investment setting. The research question are: "Is trust a primitive in economic models of behavior?" and "What factors increase (or decrease) the lieklihood of trust in economic transactions?"

The one-shot investment game was run under double blind social distance protocol guaranteeing complete anonymity of subjects' decisions. Such procedures eliminate mechanisms which could sustain investments without trust - reputation from repeated interactions, contractual precommitments, and potential punishment threats.

32 pairs of subjects participated in the experiment. Out of 32 first movers, 30 sent positive amount and only 2 sent zero, sending $5.16 on average. Out of 30 subjects who sent positive amounts five sent the whole endowment of $10, one sent $8, three sent $7, five sent $6, six sent $5, two sent $4, four sent $3, two sent $2, and two as well sent $1.

Out of 28 players 2 who received amounts greater than $1 from their paired players 1, twelve returned $0 or $1. However, 12 out of the same 28 returned more than their paired player sent what resulted in positive net returns. Investments of $5 had an average payback of $7.17, while invsetments of $10 had an average payback of $10.20. On average all 32 subjects returned $15.48, resulting in average payback of $4.56.

Based on the results of the no history treatment, one can speculate whether the subjects' decisions are influenced by social norms. According to Coleman (1990) the norms could be defined as a right by others to control an individual's action that can be achieved through sanctions or rewards. Norms may also exist if sanctions or rewards have been internalized by the individual. The internalization of norms is more likely an individual identifies himself with a particular group. "In the no history experiment subjects were all University of Minnesota undergraduates. From this viewpoint, social history provides common information about the use of trust within a group; such a history may increase social identity and reinforce an individual's predisposition towards trust. By providing social history in a double-blind, one-shot setting, we focus on the internalization of social norms, as opposed to other potential mechanisms for reciprocity such as reputation building." (pp. 132)

In the social history treatment wach subjects instructions included a report summarizing the decisions of the previous 32 subjects who participated in the no history investment game. "If, in the no history treatment, subjects in the room A (first movers) failed to appreciate the dominant strategy of the room B decisions (second movers), then providing a social histiry to room A makes subjects more aware of the existence of subjects in room B who do not reciprocate. This could result in a loss of trust, resulting in lower amounts sent. (...) Alternatively, room A subjects may focus on the positive net returns for the $5 and $10 levels. This focus could result in an increase in trust, resulting in more decisions to send either $5 or $10(...). For room B subjects, social history may make keeping money more acceptable (...). Alternatively, social history may stimulate reciprocity when $5 and $10 are sent (...)."(pp. 133)

In the social history treatment the first movers sent on average $5.36 resulting in an average payback of $6.46. Only 3 out of 28 first movers in the social history treatment sent zero and $5 and $10 were now sent half of the time. However, this increase was not statistically significant. On the other hand, out of 24 second movers who were sent amount greater than $1, six returned $0 or $1 to their paired players. 13 of the same 24 subjects reurned more than their paired player sent them, resulting in positive net returns. The investment of $5 had an average payback of $7.14 and the investments of $10 had an average payback of $13.17. The statistical analysis of data reveals that the increase in correlation between amounts sent and payback decisions when social history is provided was significant. The test comparing paybacks in the no history investment game to social history paybacks is significant at the p = 0.1 level. There was also a shift in average return from negative $0.50 in the no history treatment to a positive $1.10 in the social history treatment. The average amount sent increased only by $0.20, so most of this change is attributed to a higher average oaybacj of $1.80. The data provide a strong rejection of the subgame perfect prediction of self-regarding preferences model for the first movers and a mixed evidence for the second movers.