Trust plays a key role in promoting cooperation, exchanges, and interactions among individuals and therefore it is believed to foster economic and societal development. Sender's behaviour in the popular "investment game" (Berg et al. 1995) is widely employed to measure trust among individuals, but recent economic literature has cast doubts on the accuracy of this measure of trust. These studies, however, were mostly conducted in controlled environmentsm having university students as subject polls. We played the "investment game" with 3320 rural households from 200 villages in the Adamawa region of Cameroon, recording for each participant his expectations of return on the edowment shared as first mover. In addition, participants played two additional games obtained by separating the "investment game" into two sub-games, "Triple dictator game" and "reverse triple dictator game"/ The latter two games were used to measure participants' altruism and distributional preferences. All participants were randomly assigned to two treatments with different secrecy levels to create exogenous variation in social pressure, and measure the effect of social norms on behaviour in investment game. We use behaviour observed in the sub games to test whether senders behaviour in the investment game only measures trust (and therefore a belief in someone else's trustworthiness), and whether trustworthiness in turn is a reciprocation of kindness with kindness, or unkindness with unkindness. We control for risk preferences and other demographics, and find that senders behaviour in the investment game measures mostly trust, but it is not an accurate measure of trust.