When cartooning without reference you have two general methods available:
1. Recall a contour you've committed to muscle memory, e.g. through blind contour studies.
2. Construct a form from primitive geometry, which can be memorized as a recipe like stating the figure's proportions in "heads of height" (e.g. Andrew Loomis's figure drawing books).
While you can adjust the proportions of contours in 2D by moving your hand with exaggeration, adjusting a form in perspective is a much more calculated thing, and artists tend to gravitate towards simpler, letter-like ratios in their proportions unintentionally as they make adjustments. Thus, unusual poses and perspectives tend to be distorted towards a kind of default, symbolic figure - and mapping them to the intended character needs reference to help the brain see how the proportions differ. There are tricks that can reduce the guesswork(e.g. Burne Hogarth Dynamic Figure Drawing) by starting with drawing a well-understood perspective and then making 90 degree camera moves to restate the pose in a different viewpoint. But a reference lets you get a contour immediately, and it avoids distortion, so comic artists will tend towards using reference for solving the really hard drawing problems - even if you have no technology to help, you could get a friend to make the pose and then draw the contour of that, then fix up whatever's necessary by adding some construction over that.
For similar reasons cartoonists will tend to draw the same kind of "symbolic" face - enough to indicate character and staging, but not measured like a portrait drawing. Depending on how they practiced faces, it could be their own face or an ideal construction. And if they have the visual library available, they can make new faces by combining the symbols differently. A good caricaturist can just look and get the contours and have 80% of the likeness in an instant, and then fill in the rest symbolically.