What if teachers could see an idea evolving in a child's mind even before the child knew it? New theories of learning along with new research methods are transforming the way we design and evaluate educational activities. In particular, eye-tracking instruments enable analysts to monitor the evolution of sensorimotor action schemes students develop in solving movement problems engineered to ground mathematical concepts. I will explain the design rationale of the Mathematics Imagery Trainer project, based in embodiment theory, and then present clinical and eye-tracking data evidencing 'attentional anchors,' dynamical perceptual forms children create spontaneously out of selected environmental features to facilitate the enactment of challenging bimanual movements; the children then become conscious of the attentional anchors as things that, through cultural mediation, take on mathematical meanings. As such, we offer empirical support for historical claims from Piaget, Vygotsky, and Varela pertaining to cognitive growth processes. In turn, we have proof-of-concept data to argue for the utility of formative multimodal assessment in mathematics teachers' instructional practice.