The time-aware recommendation model captures the dynamic evolution of user interests and item popularity by analyzing users’ historical behaviors, resulting in more accurate recommendations. Recent research has increasingly shifted from traditional sequential models to Graph Neural Networks to more effectively encode complex dynamic collaborative information. Despite these advances, many approaches often overlook important higher-order relations at the motif level and the spatial dynamics within graphs. In this work, we propose a novel time-aware recommendation method named MoDynRec (Motif-Enhanced Dynamic Graph Learning for Recommendations), which effectively models higher-order features and their spatiotemporal evolution by incorporating motifs into dynamic graph representation learning. The approach employs a two-fold strategy: (1) a motif-preserving structural encoder, which aggregates higher-order information from diverse motifs to retain critical structural patterns, and (2) an attention-based temporal encoder, which integrates multi-head self-attention mechanisms to capture sequential evolution patterns in the temporal dimension. This is further enhanced by dynamic graph convolution networks, which explore spatial correlations among nodes, improving adaptability to evolving relationships over time. Extensive experiments on two publicly available datasets demonstrate the effectiveness of the proposed model.