Key study: Piliavin et al. (1969) Good Samaritanism

Aim: to investigate how the type of victim (drunk or ill), the race of the victim and the number of bystanders affect helping in a real-life emergency, and whether diffusion of responsibility happens.
Sample: about 4,450 passengers on the New York subway (around 45% black and 55% white), on weekdays between April and June 1968. There were 103 trials.
Procedure: a field experiment. Teams of four students (two female observers, a male 'victim' and a male 'model') boarded a train for a 7.5-minute non-stop journey. After about 70 seconds, the victim staggered forward and collapsed.
In the cane condition he appeared sober and carried a walking stick (ill); in the drunk condition he smelled of alcohol and carried a bottle in a paper bag.
If nobody helped, the model helped after about 70 or 150 seconds. Observers recorded who helped, how fast, and the number, race and sex of passengers.

Results: the cane victim received spontaneous help on about 95% of trials; the drunk victim on about 50%.
Help usually came quickly. 90% of first helpers were male.
There was a slight tendency for same-race helping, mostly in the drunk condition.
There was no diffusion of responsibility: larger groups were not less likely to help.
Conclusion: Piliavin et al. proposed the arousal: cost-reward model: seeing an emergency causes unpleasant arousal, and people reduce it by weighing up the costs and rewards of helping. Helping a drunk person seems more costly (risk of harm or embarrassment) and less rewarding.

Strengths: a field experiment in a natural setting, so it has high ecological validity; passengers did not know they were being studied, so there were no demand characteristics. Large sample.
Weaknesses: ethical issues: passengers did not give consent, were deceived, may have been distressed, and could not be debriefed. Little control over who was in the carriage, and some passengers may have seen more than one trial. Findings from New York in 1968 may not generalise.