The em unit denotes the font size of the element, except when used in the value of the font-size property, where it denotes the font size of the parent element. In this sense, in the case presented, the context is the parent element.
In the case presented, the font size of the word “toad” is thus equal to the font size of the parent of the h1. No specific value for it can be inferred for it from the data given.
When font sizes is computed, the font size of the parent of h1 has been computed when this construct will be dealt with; let’s call it s. First the font size of h1 is computed, multiplying s (the font size of the parent) by 2. Then the font size of the span element is computed, multiplying its parent’s font size by 0.5, yielding s. Theoretically, rounding errors could cause a minimal deviation in such processes, but multiplication by 2 and 0.5 can be assumed to be exact in computers.