H567/02 Section A - Piliavin et al. (1969): Subway Samaritan

Helping and the research question

Piliavin, Rodin and Piliavin studied an emergency in which passengers could see both the person in need and other potential helpers.

Earlier laboratory work found diffusion of responsibility: when several people were believed to be present, each could feel less personally responsible for acting. Some earlier emergencies were heard rather than seen. The subway study asked whether this pattern would occur in a face-to-face, public setting where escape was difficult. It also investigated apparent illness versus drunkenness, victim race and the effect of a helping model.

The researchers were responding to contemporary concerns about bystander inaction after Kitty Genovese's murder. The often-repeated account that exactly 38 witnesses all watched and did nothing is not needed to establish the research question and should not be treated as a verified description of that event.

Its theme is Responses to people in need. It belongs to social psychology because helping is examined in relation to the perceived victim, other bystanders, social expectations and opportunities to leave, rather than assumed to be only a stable altruistic personality.

Field design and passengers

The researchers staged 103 collapses on New York subway express trains and observed unsuspecting passengers.

About 4,450 men and women travelled in the sampled cars between 11 a.m. and 3 p.m. on weekdays from 15 April to 26 June 1968. This was an opportunity sample: people happened to use that route at those times. A typical car had about 43 passengers; the area around the collapse averaged 8.5. The paper described the usual racial composition as approximately 45% Black and 55% White. These historical observer categories do not describe all aspects of ethnicity or identity.

This was a field experiment for manipulated victim condition and programmed model conditions, with observation used to measure helping. Apparent victim race also varied, but only one of the four male victims was Black, confounding race with that actor's individual characteristics. Naturally occurring group size was measured, not experimentally assigned. Different train journeys supplied the conditions; the researchers could not guarantee that a regular passenger never encountered more than one trial.

Measured outcomes included whether spontaneous help occurred, its latency, the helpers' observed characteristics, number of helpers, movement away from the critical area and passenger comments. The trial, rather than each of the approximately 4,450 passengers, is the denominator for results such as “62 of 65 cane trials received spontaneous help”.

Collapse procedure and model

Four student teams used a consistent collapse sequence during the roughly 7½-minute nonstop journey between 59th and 125th Streets.

Each team contained a male victim, a male model and two female observers. They boarded through different doors. The victim stood near a pole in the designated end section, staggered and collapsed about 70 seconds after departure, then lay on his back looking upwards. Observers sat outside that critical area and recorded events unobtrusively.

Victims were men aged 26–35, identically dressed in a jacket and old trousers without a tie. On 65 trials the victim carried a black cane and appeared sober. On 38 he smelled of alcohol and carried a bottle in a brown bag. Every victim performed both conditions. Condition numbers were unequal because teams' schedules, a procedural departure by one team and a student strike interrupted the planned balance.

Models were four White men aged 24–29. A model could stand in the critical or adjacent area and intervene early, about 70 seconds after collapse, or late, about 150 seconds after collapse. No-model trials were also planned. Programmed assistance involved raising the victim to a sitting position and remaining with him. If nobody helped by the end of a journey, the model helped the victim stand before the team left.

One observer recorded people in the critical area and helpers; the other recorded the adjacent area and helping latency. Both noted comments. The model's actual intervention depended on whether passengers had already helped. Consequently a planned model condition does not mean that a model actually acted on every trial. Once a model intervenes, later helping cannot reveal when a passenger would have helped without that model; this complicates interpretation of latency differences.

Helping findings and denominators

The apparently ill victim received more spontaneous help than the apparently drunk victim.

The first percentage is $\frac{62}{65}\times100\approx95.4\%$; the second is $\frac{19}{38}\times100=50.0\%$. These are trial rates, not percentages of all passengers who helped. On non-model trials the median first-help latency was five seconds for cane victims and 109 seconds for drunk victims; the latter calculation assigned 400 seconds to nonresponses. These latency summaries apply to that subset, not every trial.