Behavioral inhibition (BI) is an early-appearing temperament characterized by strong reactions to novelty. BI shows a good deal of stability over childhood and significantly increases the risk for later diagnosis of social anxiety disorder (SAD). Despite these general patterns, many children with high BI do not go on to develop clinical, or even subclinical, anxiety problems. Therefore, understanding the cognitive and neural bases of individual differences in developmental risk and resilience is of great importance. The present review is focused on the relation of BI to two types of information processing: automatic (novelty detection, attention biases to threat, and incentive processing) and controlled (attention shifting and inhibitory control). We propose three hypothetical models (Top-Down Model of Control; Risk Potentiation Model of Control; and Overgeneralized Control Model) linking these processes to variability in developmental outcomes for BI children. We argue that early BI is associated with an early bias to quickly and preferentially process information associated with motivationally salient cues. When this bias is strong and stable across development, the risk for SAD is increased. Later in development, children with a history of BI tend to display normative levels of performance on controlled attention tasks, but they demonstrate exaggerated neural responses in order to do so, which may further potentiate risk for anxiety-related problems. We conclude by discussing the reviewed studies with reference to the hypothetical models and make suggestions regarding future research and implications for treatment.