To theoretically explore the behavior of instrinsic noise in one- and two-enhancer models, we used the formalism of Sanchez et al., 2011; Sánchez and Kondev, 2008. As described in Materials and methods, the predicted CVs from these models are estimates for intrinsic noise. (A) We plot the mean expression level versus CV for the five different enhancer models and one set of parameters, k = l = 1, p=1, γ = 0.1. The single enhancer model (dark purple) drives the highest CV, indicating that, under the assumptions of our models, adding an additional enhancer generally lowers intrinsic noise. Except for XOR model (yellow), all other models produce more mRNA than the single enhancer model. The other colors are: blue, OR model; green, additive model; brown, synergistic model. (B) Here, we plot the CV as a function of l, the rate of promoter-enhancer dissociation, for the five models above and vary l from 0.1 to 10 on a logarithmic scale with k = 1, p=1, γ = 0.1. With the exception of the XOR model with high l, the single enhancer model drives a higher CV than the models with two enhancers for the same value of l. These results show that, under the simplifying assumptions made in these models, the addition of a second enhancer generally lowers the predicted intrinsic noise. In our experimental data (Figure 6), we only observe a significant decrease in interallele noise for the shadow enhancer pair compared to the single distal or single proximal enhancer. Duplications of either the proximal or distal enhancer do not have significantly lower noise than their respective single enhancer constructs. Therefore, we expect that the simple addition of an identical enhancer likely does not fulfill the simplifying parameter assumptions used here and suggests that further investigation is needed to understand the complexity of the relationship between interallele noise and the numbers of enhancers controlling a promoter.