Geometric morphometric analyses are used to explore variation of maxillary dental arcades of Australopithecus afarensis, expanding on the work of Hanegraef and Spoor, 2025 (Morphological variation of the Australopithecus afarensis maxilla. Journal of Human Evolution, 201, 103651) by integrating evidence from a large sample of virtually reconstructed mandibles. Size and shape of maxillary dental arcades can be predicted accurately from mandibular landmarks based on strong covariation between occluding upper and lower dentitions, and a novel method was developed to correct for reduced shape variation in these predictions. As predictions are restricted to the alveolar process, morphological information about the rest of the maxilla is lost. The trade-off between a smaller sample with comprehensive morphology and a larger sample with restricted morphology is discussed. Here, we analyzed 9 original and 17 predicted A. afarensis dental arcades in the comparative context of 448 extant hominine (modern human and African ape) maxillae. This study found that (1) degrees of size and shape variation are high in A. afarensis, potentially even higher than in Gorilla species when including the predictions in the fossil sample, (2) no allometry was detected, even when expanding the A. afarensis sample with predictions, (3) size and shape do not significantly change over time when analyzing original and predicted A. afarensis dental arcades together, and (4) sexual form and shape dimorphism, but not sexual size differences, are reduced when including A. afarensis predictions in the fossil sample. Our results quantifying the range and pattern of variation of the A. afarensis maxilla provide a comparative context when assessing whether or not other Plio-Pleistocene hominin specimens are conspecific.