Self-efficacy is a student’s belief in their capacity to produce a desired outcome through effort (Bandura, 1997). Extant research has demonstrated that self-efficacy is a powerful motivational antecedent of science achievement and is considered an important motivational factor to promote in the science classroom (Britner, 2008). However, social cognitive theory posits that negative emotional states, such as anxiety, are likely to thwart the positive effects of self-efficacy on achievement (Bandura, 1997). While researchers have largely examined anxiety as a negative predictor of self-efficacy, recently researchers have argued that anxiety and self-efficacy are likely to co-occur, such that as anxiety increases, the benefits of self-efficacy on achievement decrease (i.e., negatively moderated; Gala & Wood, 2012). This suggests that in order to optimise student outcomes in science, it may be important to take a “dual intervention approach”: concurrently promoting self-efficacy and reducing anxiety in science. Thus, the present investigation tests support for this approach by examining if self-efficacy and anxiety significantly interact and if this impacts science achievement. Additionally, self-efficacy and anxiety have also been shown to vary at the classroom-level, and both play a role in motivational classroom climate (Martin et al., 2012). Classroom-level anxiety and self-efficacy have also been shown to impact classroom-level achievement (Martin et al., 2012). Thus, there is a need to examine if science classrooms can be classified by these concurrent experiences of anxiety and self-efficacy, and if moderating effects of anxiety on self-efficacy impact classroom-level science achievement. Doing so would indicate if a “dual intervention approach” is also appropriate at the classroom-level. Thus, the present investigation examines (a) the extent to which science self-efficacy and anxiety predict science achievement at the student- and classroom-level, (b) if science self-efficacy and anxiety significantly interact, and (c) the extent to which this interaction has significant effects on science achievement at the student- and classroom-level. Data were collected from N = 1,075 secondary school students clustered in N = 99 science classrooms. Multi-level structural equation modelling demonstrated that, at the student- and classroom-level, science self-efficacy significantly positively predicted science achievement, whereas the relationship between science anxiety and science achievement was non-significant. At the student-level, but not the classroom-level, science anxiety significantly moderated the effects of science self-efficacy on science achievement. Plotting showed that as self-efficacy increased, students with higher anxiety had lower science achievement. These results provide support for a “dual intervention approach” at the student- but not classroom-level.