We are commonly advised to turn the toes in (medially rotate) in order to emphasise the lateral head, and to turn the toes out (laterally rotate) in order to emphasise the medial head. And this notion is mostly (1, 2) but not entirely (3) supported by a limited body of research. That is, medial and lateral rotation of the ankle does appear to alter the degree of activation of the lateral and medial heads, respectively, but to what degree this is true, and whether other factors are at play warrants further investigation.
Due to the position of the calcaneus relative to the orientation of the lateral and medial heads, we might further expect that plantar flexion from a rolled-outward (supinated) and rolled-inward (pronated) position would affect their activation. This question has not specifically been investigated, but one study on the effect of subtalar joint position suggests that it may not result in any significant difference.
It should be understood, above all, that the medial and lateral heads of the gastrocnemius work as a functional pair, their co-contraction producing two force vectors whose sum is a single vertical or near-vertical vector transmitted through the Achilles' tendon. And that is combined with the single-joint soleus muscle to form the functional triad of the triceps surae. Its primary function is to plantar flex the ankle, to control the rate of dorsi flexion, and to stabilise the joint in the other planes whilst doing so. Thus, the degree of control we have on muscle activation without plantar or dorsi flexion is limited to our simulating joint instability.
The structure of the triceps surae, especially considering the combination of single- and multi-joint heads, suggests not functional differences in line of pull—the three heads all converge on the same tendon—but rather functional differences in activation at different angles of plantar or dorsi flexion and, in the case of the gastrocnemius, angles of knee flexion. This theory is supported by one study, which concluded that “the greater the dorsi flexion angle, the stronger the bias of force to the [medial gastrocnemius] compared with [the lateral gastrocnemius].”
Thus, in order to train both lateral and medial heads of the gastrocnemius, we should both vary the medial/lateral rotation of our feet, and ensure that we train our exercises through the full range of motion—full dorsi flexion to full plantar flexion. We can further medially rotate our ankles and work through ranges of greater plantar flexion to focus on the lateral head of the gastrocnemius.