To address the lack of effective tools for evaluating perceived grouping in interfaces, we introduced a multiplex cognitive network approach. This method, grounded in theory and empirical evidence, predicts perceived grouping by representing and analyzing the similarity and proximity of interface elements, such as through Euclidean distances and color-coding for proximity and similarity networks, respectively. Pearson correlation of these networks provided a measure of network coherence, serving as an indicator of perceived grouping. Through user studies on list-form and grid-form mobile interfaces, involving quick browsing and subjective evaluation tasks, we found that network coherence significantly enhances our understanding of interface grouping, and further validated by perceived visual complexity. Specifically, network coherence explained an additional 18% to 23% of the variance in perceived ease of grouping and visual complexity for list-form interfaces, respectively, and 8% to 16% for grid-form interfaces, beyond what was accounted for by control variables. Our findings suggest that network coherence is a robust indicator for evaluating perceived grouping in interface design.