Weakness - Milgram's S lacked internal validity
Orne and Holland suggest participants guessed the electric shocks were fake. So Milgram was not testing what he intended to test (i.e. obedience).
However, Sheridan and King's participants gave real shocks to a puppy; 54% of males and 100% of females deilvered what they thought was a fatal shock.
So the obedience in Milgram's study might be geneuine, 70% of Milgram's participants believed the shocks were genuine.
Strength - Milgram's S has good external validity
Milgram argued that the lab-based relationhsip between experimenter and particiapnt relfected wider real-life authoirty relationships.
Hofling et al found that levels of obedience in nurses on a hospital ward to unjustified demands by doctors were very high (21 out of 22 nurses obeyed).
Therefore the processes of obedience in Milgram's study can be generalised.
Strength - Replications support Milgram's S
In a French documentary contestants in a reality TV game show were paid to give (fake) electric shocks - when ordered by the presented - to other participants (actors).
80% gave the maximum 450 volts to an apparently unconscious man. Their behaviour was like that of Milgram's participants, e.g. many signs of anxiety.
This supports Milgram's original conclusions about obedience to authority and shows that his findings were not just a on-off.
Weakness - Alternative explanation is SIT
Obedience is about group identification. Milgram's participants identified with the experimenter (the science of the study). When obedience levels fell, the participants identified more with the victim.
Haslam and Reicher suggest the first three 'prods' are appeals fro help with science ('experiment requires you continue'). Only the 4th prod demands obedience. Every time this was used, the participant quit.
The participants did not give shocks due to obedience, but due to thier identificaiton with the experimenter as a scientist (as explained by social idenitity theory).
Weakness - Milgram's S has ethical issues
Baumrind cirticised Milgram's deceptions. Participants believed the allocation of roles was randomly assigned, but it was fixed.
The most significant decpetion was that participants believed the electric shocks were real. Baumrind objected because deception is a betrayal of trust that damages the reputation of psychologists and their research.
Deception of pariticipants may also make them less likely to volunteer for future research.