As wearables, such as smartwatches and data gloves, become ubiquitous in our daily lives, it is possible to capture users' behaviors as they go, and then to authenticate them using biometrics. A variety of behaviors have been exploited to authenticate users in a plethora of scenarios. However, there is still a lack of systematic investigations on what types of behaviors present better performance for authentication, and therefore, limiting our understanding of the custom design of behavioral ''passwords'' to fit people's needs. This paper is designed to help answer this question. First, we design a collection of behaviors, which is used to explore the effect of different types of behaviors on authentication. This collection involves hand and finger behaviors since they are common and natural human actions; it also reasonably covers both basic movements and complex behaviors, two major categories of behaviors. We evaluated authentication performance along the dimensions of behaviors, users (sensitivity) and classification models (independence). A valuable finding shows that finger-issued movements present almost equivalent performance with complex behaviors. Therefore, it suggests applying them to the wearable and mobile devices as an alternative to complex behaviors, and could considerably reduce requirements for computational resources compared to complex behaviors. Furthermore, they extend the applications of behavioral authentication to a wider array of scenarios, like performing behavioral authentication such as moving fingers in one's pocket, which is common for wearable and mobile applications.