They are, in a sense, passive recipients of stimulation that has been especially designed to elicit information about cognitive algorithms located exclusively in their brains. Embodied cognition, we shall see, resists the idea that cognition is solipsistic, and so rejects the idea that subjects are passive receivers of stimulation.
If you only study cognition while people are cut off from their environment, it's reasonable to conclude that it only happens in the brain. Sternberg's tests in a lab function differently than say the grocery list example.