Piliavin et al. (Subway Samaritans)

Knowing the theory behind Piliavin lets you explain why people helped, not just report numbers.

The bystander effect. This is the finding that the more people present at an emergency, the less likely any one person is to help. It became famous after the 1964 Kitty Genovese murder in New York, where (it was reported) many neighbours failed to act.

Diffusion of responsibility. The main explanation for the bystander effect: when many people are present, each individual feels less personal responsibility to act ('someone else will do it'), so responsibility is spread across the group. Earlier lab studies (e.g. Latané & Darley) found this — but they used artificial, often non-face-to-face emergencies. Piliavin wanted to test helping in a real, face-to-face emergency.

The arousal:cost-reward model (Piliavin's explanation). This is the model the study helped develop. It says helping works in steps:

• Seeing an emergency creates unpleasant physiological arousal.
• This arousal is unpleasant, so the bystander is motivated to reduce it.
• The bystander chooses the response that reduces arousal at the lowest net cost, weighing:
  • costs of helping (effort, danger, disgust, embarrassment), versus
  • costs of not helping (guilt, blame from others), versus
  • the rewards of helping (praise, thanks, feeling good).

Applying it to cane vs drunk. Helping the drunk victim has a higher cost of helping (disgust, possible aggression) and a lower cost of not helping (he is seen as partly to blame), so people help less. The cane victim is clearly not responsible, so the cost of not helping (guilt) is high → more help. This is the heart of an A* answer.

Psychology being investigated: helping (prosocial) behaviour and bystander behaviour.

Aim. To investigate, in a realistic field setting, what affects whether and how quickly people help a stranger in an emergency. Specifically, Piliavin et al. examined the effect of:

• the type of victim — an apparently ill victim (carrying a cane) vs an apparently drunk victim;
• the race of the victim (black or white) and whether help was same-race;
• the presence/timing of a model (a helper) on whether others then helped;
• the number of bystanders — to test whether diffusion of responsibility occurs in a real, face-to-face emergency.

Sample: approximately 4,450 men and women who happened to be travelling on a particular New York subway route on weekdays over about 2 months in 1968. They were unaware they were in a study (an opportunity sample of real passengers). About 45% were black and 55% white; on average ~43 people were in the carriage and ~8.5 in the 'critical area' near the victim.

Setting: the 8th Avenue express train, on the 7.5-minute non-stop journey between 59th Street and 125th Street — so passengers were 'trapped' together and could not easily leave.

Teams & roles: four teams of Columbia University students, each team made up of a victim, a model (potential helper) and two female observers. The victims were male, aged 26–35 (three white, one black), dressed identically.

Procedure:

1. About 70 seconds into the journey, the victim staggered forward and collapsed, lying on his back and staring at the ceiling until helped.
2. Two victim conditions (the key IV): the 'cane' victim appeared ill (sober, carrying a black cane); the 'drunk' victim smelled of alcohol and carried a bottle in a brown paper bag.
3. Model conditions (counterbalanced): if no one helped, the model intervened at a set time — early (~70s) or late (~150s), from a position adjacent to or distant from the victim.
4. The two observers covertly recorded: how many people helped, how quickly, the race/sex of helpers, the number of people in the carriage, and comments made by passengers.

• Type of victim mattered most. The cane victim was helped on ~95% (62 of 65) of trials; the drunk victim only ~50% (19 of 38) of trials.
• Help was fast and spontaneous. The cane victim usually received help within about 5 seconds — often before the model was due to intervene. Help came without the model in most cases.
• No diffusion of responsibility. Having more bystanders did not reduce helping (if anything, larger groups produced help slightly faster) — the opposite of the bystander effect found in labs.
• Helpers were mostly male. About 90% of first helpers were men (the victim was male and helping was physical).
• A slight same-race tendency, especially for the drunk victim (people were a little more likely to help a victim of their own race when help was 'costly').
• Some passengers moved away or made comments to justify not helping, especially in the drunk condition.

• People do help in real, face-to-face emergencies — often quickly and spontaneously, and diffusion of responsibility did not occur in this setting.
• The type of victim matters: an apparently ill victim (cane) is helped far more than an apparently drunk one, because the drunk raises the cost of helping (disgust) and lowers the cost of not helping (seen as partly to blame).
• Helping behaviour can be explained by the arousal:cost-reward model — bystanders feel unpleasant arousal and act to reduce it at the lowest net cost.
• A slight same-race effect (especially when costs are high) suggests in-group factors play a role in costly helping.

Strengths

• High ecological validity — a real emergency with real passengers in a natural setting (the subway), so behaviour is genuine.
• No demand characteristics — participants did not know they were being studied, so they behaved naturally (no 'please-the-experimenter' effect).
• Large sample (~4,450) of varied, ordinary people → better generalisability than a small lab sample.
• Mixed data — quantitative (frequency/speed of help) plus qualitative (passengers' comments) gives a fuller picture.

Weaknesses — ethics (a major issue)

• No informed consent and deception — passengers did not agree to take part and believed the emergency was real.
• No right to withdraw — they were 'trapped' on a non-stop train.
• Psychological harm — witnessing a collapse (especially blood in a later variation) could distress bystanders, and there was no debrief.

Other weaknesses

• Low control of extraneous variables — being a field experiment, factors like the exact passengers, crowd size and mood varied between trials (confounds).
• Observation reliability — only two observers recording a fast, chaotic scene may have missed or misjudged behaviour; some data were lost.
• Generalisability — only NYC subway passengers were studied, so findings may not generalise to other places/cultures (an ethnocentric sample), and helping a male victim may not reflect helping in general.

Issues & debates

• Individual vs situational: strongly situational — the type of victim (the situation) drove helping.
• Application to everyday life: understanding and encouraging helping (e.g. why people fail to help, training, the cost-reward of intervening).