How many replicates do you need to be a statistically sound experiment?

elizabeth loftus conducted a study to observe the reliability of eyewitnesses during the 1970s. their baseline was to define how accurate are the recollections of eyewitnesses.

her studies do not support the use of eyewitness. loftus mentions in her work that we are not required to recall some experiences, meaning that rarely an individual eyewitness testimony is trustworthy. a good example is presented in her work, where all individuals watched a video and were than asked "about how fast were the cars going when they hit/smash? ". the testimonies changed based on the word she used to make the question (hit or smash), proving the fragility.