Although human infants demonstrate early competence to retain visual information, memory capacities during infancy remain largely undocumented. In three experiments, we used a Visual Paired Comparison (VPC) task to examine abilities to encode identity (Experiment 1) and spatial properties (Experiments 2a and 2b) of unfamiliar complex visual patterns during the first year of life. In the first experiment, 6- and 9-month-old infants were familiarized with visual arrays composed of four abstract patterns arranged in a square configuration. Recognition memory was evaluated by presenting infants with the familiarized array paired with a novel array composed of four new patterns. The second couple of experiments aimed to examine infant ability to encode the spatial relationships between each pattern of the array (e.g., where is A in the square configuration). The 6-, 9- and 12-month-old infants were tested on a spatial version of the VPC task, in which the novel array was composed of the same patterns than the familiarized array but arranged differently within the square configuration. Results indicated that infants retained the identity of the patterns but not their specific spatial relationships within the square configuration (i.e., allocentric location of the patterns), suggesting either an immaturity of the processes involved in object-to-location binding, or the inappropriateness of unfamiliar complex objects to reveal such early allocentric abilities.