Animals exhibit rich diversity in form, including in tissues whose function depends on deforming. In fluidic locomotion, deformations contribute to aspects of performance ranging from propulsive efficiency to environmental navigation, but it is challenging to quantify and compare dynamic morphological variation. There are many possible ways to quantify shape but their correspondence to dynamic variation among distinct behaviours is unclear. We studied dynamic wing shape variation of hummingbirds hoverfeeding during a set of flight challenges in order to quantify the principal axes of shape variation (shape variables) and their relationships to conventional descriptors of wing shape like wing twist and area. The first four shape variables explained >95% of the dynamical variation in wing shape and decomposed shape changes into in- and out-of-plane components at integer multiples of the wing beat frequency. To varying degrees, these shape variables were interpretable as aspects of conventional two- and three-dimensional morphology, particularly the dominant contributions of wing twisting (~80% of shape variation) and area changes (~10% of shape variation). However, variation in wing morphology among behaviours was not predicted by each shape variable's explained variance, but instead by stroke cycle subintervals associated with specific shape changes. We identified pronounced out-of-plane deformations throughout the stroke cycle while hoverfeeding in front of a visual barrier, whereas in-plane deformations during the upstroke and pronation were especially prominent during submaximum load lifting. Our study highlights the complexity of shape changes in biological tissues during active behaviour, and how dynamic morphology contributes to agile fluidic locomotion.