An important component of arbovirus surveillance is monitoring the vector for presence of the pathogen. Intervention and preventive programs need early detection of arboviral activity in mosquito populations. In this report, we examine the factors affecting the probability of detection of mosquito infections. Since arbovirus infection rates in mosquito populations are very low, observations of zero-infected mosquito samples are common. Using statistical models, we describe methods to estimate the probability of detection and upper bounds of confidence intervals of mosquito infection rates as measures of confidence for observations of zero infection. Our results show that detection of low levels of mosquito infections requires large samples (greater than 1,600 individuals) for a high probability (0.8) of detection. Due to focal transmission of arboviruses, grouping samples over different sampling sites and times is often inappropriate for detection of mosquito infection. We emphasize sample size as a key determinant in detection of mosquito infections and recommend intensified entomologic surveys at sentinel sites to detect arboviral activity.