Key study: Haber and Levin (2001)

Token invariant objects come in one standard size, for example a bowling ball or a basketball.
Token variant objects come in many sizes, for example a Christmas tree.
Experiment 1: a large group of male psychology students were given a questionnaire listing familiar objects and estimated their heights from memory, without seeing them.
Results: they were very accurate for token invariant objects, but less accurate as the size range of an object increased (token variant).
Conclusion: size can be judged from memory and prior knowledge, not just from seeing.

Participants: 9 male volunteers checked for normal vision.
Procedure: a large grassy field was divided into areas containing 15 familiar token invariant objects, 15 familiar token variant objects, and 15 unfamiliar cardboard shapes of unknown size, at different distances. Participants estimated the size and distance of the objects they could see. They also gave size estimates from memory.
Results: sizes of familiar objects were estimated about as accurately when seen as from memory; seeing them did not make size estimates better. Distance estimates were less accurate for token variant objects than token invariant ones. Unfamiliar shapes were still estimated fairly well but tended to be overestimated.
Conclusion: perceiving size relies heavily on knowing what the object is (prior knowledge). Distance is judged from cues in the environment, and when distance cues are poor, known size helps judge distance.

Strengths: Experiment 2 took place outdoors in a real field, so it was more realistic than a lab. Objects were placed in a standardised way, so the study could be replicated. Using two experiments gave converging evidence.
Weaknesses: small sample in Experiment 2 (9 people), all male students, so it may not generalise to women or other groups. Estimates in feet and inches depend on people's skill at using units, not just perception. It is hard to know which cues each participant used.