You are viewing a free preview of this lesson.
Subscribe to unlock all 9 lessons in this course and every other course on LearningBro.
Six years after Milgram, and against the background of a famous case in which a young woman was reportedly attacked in New York while many neighbours did nothing, Irving Piliavin and his colleagues took the psychology of helping out of the laboratory and onto a moving subway train. Their "Subway Samaritan" study is the classic study for the social-area theme of responses to people in need, and it is one of the most admired field experiments in social psychology precisely because it observed real people, in a real setting, making real decisions about whether to help a stranger who had apparently collapsed. Its findings challenged the then-dominant idea of diffusion of responsibility and introduced a cost–benefit way of thinking about helping that still frames the field today.
This lesson tells the study in the OCR "tell the story" format — background, aim, method (design, sample, procedure), results with their real figures, conclusions, and a full evaluation — before linking it to its theme, area, perspective and debates. Because Piliavin is rich in variables (victim type, race, the presence of a model, group size) it is easy to overload; the lesson keeps the core manipulations and findings sharply in focus.
| This lesson covers | OCR H567 Component 02 element | AO focus |
|---|---|---|
| Background (bystander research; diffusion of responsibility) and aim | Section A — Social; theme: responses to people in need (classic) | AO1 knowledge |
| Method: field experiment; naturalistic subway sample; the drunk vs ill (cane) manipulation; the model conditions | Section A — Core study (Piliavin) | AO1; AO2 |
| Results (helping rates; speed; effect of victim type; group size) | Section A — Core study | AO1 |
| Conclusions and the arousal:cost–reward model | Section A — Core study | AO1; AO3 |
| Evaluation: method, data, ethics, validity, reliability, sampling, ethnocentrism | Section A; Section B debates | AO3 |
| Links to theme, area, perspective and debates | Section B — Areas, perspectives, debates | AO1; AO3 |
The specification is referenced descriptively; consult the official OCR H567 specification document for its exact published wording. This lesson develops AO1 (aim, procedure, results, conclusions), AO2 (applying the cost–reward model and bystander principles to novel helping scenarios) and AO3 (evaluating the study's ethics, validity, reliability and generalisability).
Piliavin's study belongs to the wave of bystander research prompted by public concern about cases in which people in need were apparently ignored by onlookers. The leading laboratory explanation of the time, developed by Latané and Darley, was diffusion of responsibility: the idea that the more bystanders are present, the less likely (and slower) any individual is to help, because responsibility is shared out among the group and no single person feels obliged to act. Laboratory studies had supported this "bystander effect".
Piliavin and colleagues wanted to test helping in a real, natural setting rather than the laboratory, and to examine factors the lab had not — in particular the type of victim (does it matter why someone appears to need help?) and whether diffusion of responsibility would even appear when bystanders could directly see both the victim and each other. Their study is thus both a test of existing theory and an exploration of new variables.
The move from laboratory to field was itself a methodological argument. The laboratory bystander studies that had established diffusion of responsibility typically placed participants in situations where they could not see the victim or the other bystanders — for example, hearing an apparent emergency through an intercom while believing others were also listening. In such designs it is easy to assume "someone else will deal with it", because the other bystanders are abstract and unseen and the victim is not directly witnessed. Piliavin suspected that this feature — the invisibility of the emergency and of the other bystanders — might be doing much of the work in producing diffusion, and that in a real emergency, directly witnessed, the effect might weaken or vanish. A subway carriage is the ideal natural laboratory for this: the victim collapses in plain sight, the other passengers are visible to one another, and there is a defined interval between stations during which the drama unfolds. The design thus lets the researchers ask whether diffusion of responsibility is a general law of helping or an artefact of the particular, rather artificial conditions under which it had been demonstrated.
There is a further reason the study was important in its historical moment. Public discussion of bystander apathy had tended toward a bleak, almost moralising conclusion about the coldness of city life — the idea that urban dwellers had become indifferent to one another's suffering. Piliavin's field approach put this cultural narrative to an empirical test: would real New Yorkers, on a real train, actually ignore a collapsed stranger? The answer, as we shall see, was strikingly at odds with the pessimistic story, and this contribution — showing that ordinary people in a large city help far more readily than the "apathy" narrative implied — is part of why the study endures.
The broad aim was to investigate the conditions under which bystanders help a stranger in need in a real-world setting. Specifically, Piliavin et al. wanted to examine the effect of several factors on helping: the type of victim (whether the person appeared ill or drunk), the race of the victim (same or different from the bystanders), the presence and behaviour of a helping "model", and the number of bystanders (to test diffusion of responsibility). The key measures were whether help was given, how quickly, and by whom.
The study was a field experiment conducted on the New York subway, using covert, controlled observation to record behaviour. The independent variables (victim type, victim race, model condition) were manipulated by the researchers using confederates; the setting and the "sample" of bystanders were entirely natural. Because it took place in participants' real environment and they did not know they were being studied, the study has notably high ecological validity.
The participants were the passengers who happened to be travelling in the target subway carriages during the trials — an opportunity sample of the general travelling public of New York. Across the many trials, an estimated 4,450 or so men and women were present as potential helpers/bystanders, of varied ages and ethnicities, riding a particular express route. Crucially, none of them consented or knew they were participants; they were simply people on a train.
Each trial used a team of four confederates: a victim, a model (a confederate ready to help if no genuine passenger did), and two observers (female confederates who recorded the data from nearby seats). The trials ran on a stretch of the journey between stations lasting several minutes, giving passengers time to respond.
The victim (a male confederate) would stand holding a pole and then stagger and collapse to the floor of the carriage shortly after the train left a station, then lie on his back and wait. The key manipulation of victim type was staged through appearance and props:
| Victim condition | How the victim appeared |
|---|---|
| "Ill" victim | Carried a black cane and appeared sober but unwell — the classic "cane" condition. |
| "Drunk" victim | Smelled of alcohol and carried a bottle in a brown paper bag, appearing intoxicated. |
The victims also varied by race (some Black, some White) to test whether same-race or different-race victims were helped differently. The model conditions varied whether and when a confederate stepped in to help (e.g. early or late, from near or far), so the researchers could see the effect of a helping example and could ensure the victim was eventually assisted if no passenger acted.
Observers recorded, for each trial: whether the victim received help, the latency (time in seconds before the first passenger helped), the number, sex and race of helpers, the total number of passengers in the carriage and in the immediate vicinity, and passengers' spontaneous comments. Because it was covert and naturalistic, the behaviour recorded was genuine helping (or non-helping) as it spontaneously occurred.
The results were, in several respects, surprising against the prevailing bystander theory.
Helping was frequent and fast — especially for the ill victim. Contrary to the gloomy picture of unhelpful city-dwellers, the (apparently) ill victim carrying the cane was helped in the overwhelming majority of trials, and typically very quickly — often within seconds and frequently before the model had any need to intervene. Help for the ill victim was close to universal.
The drunk victim was helped less, and more slowly. The victim who appeared drunk was helped substantially less often and after a longer delay than the ill victim, though a considerable number of people still helped even the drunk victim. The clear difference between the two conditions shows that bystanders' willingness to help depends on the perceived cause of the need — an "ill" person is seen as a more deserving, lower-cost, less blameworthy target than a "drunk" one.
Diffusion of responsibility did NOT appear as predicted. The laboratory prediction was that more bystanders would mean less and slower helping. On the subway the opposite tendency was found: help was generally rapid regardless of the number of passengers, and there was no evidence that larger groups produced less helping — if anything, more potential helpers were associated with faster help. Piliavin argued this was because, unlike in the lab, bystanders on the train could see the victim (and each other) directly, so responsibility was not diffused in the same way.
Same-race helping tendency, mainly in the drunk condition. There was some tendency for passengers to help a victim of their own race more readily, and this same-race effect was more pronounced for the drunk victim (the higher-cost case) than for the ill victim, who was helped promptly by passengers regardless of race.
Who helped. The great majority of first helpers were men (which makes sense given the physical nature of assisting a collapsed person), and help often came from passengers in the immediate area of the victim.
| Finding | Result |
|---|---|
| Ill (cane) victim | Helped in the great majority of trials, usually within seconds |
| Drunk victim | Helped less often and more slowly, but still frequently |
| Effect of group size | No diffusion of responsibility; larger groups did not reduce helping |
| Race | Slight same-race helping tendency, stronger for the drunk victim |
| Sex of helpers | Overwhelmingly male first helpers |
Piliavin and colleagues concluded that helping in an emergency is governed by a cost–benefit calculation rather than by simple diffusion of responsibility, and they proposed the arousal:cost–reward model to explain it. The model has two components. First, witnessing an emergency creates unpleasant physiological arousal in the bystander, which they are motivated to reduce. Second, the bystander (largely unconsciously and rapidly) weighs the costs and rewards of the possible actions:
On this model, the ill victim is helped more than the drunk victim because helping the drunk carries higher costs (disgust, unpredictability, potential for trouble) and the drunk's plight carries lower costs of not helping (the drunk is seen as partly responsible, so bystanders feel less guilt). The absence of diffusion is explained by the visibility of the emergency: bystanders who directly witness a collapse experience strong arousal and cannot easily deny responsibility, so helping remains high even in a crowd.
The study therefore reframed helping as a rapid situational appraisal of costs and rewards, dependent on features of the victim and the setting — a distinctly social-area conclusion.
It is worth working through how the arousal:cost–reward model explains each of the study's main findings, because the ability to derive the findings from the model (rather than merely list them alongside it) is what distinguishes a strong answer. The model treats the bystander as experiencing unpleasant arousal at the sight of an emergency and as motivated to reduce that arousal in whatever way carries the best balance of costs and rewards. From this, the ill/drunk difference falls out directly: helping a drunk victim carries higher costs of helping (he may be unpredictable, unpleasant, or embarrassing to deal with) and lower costs of not helping (because he is seen as partly responsible for his own state, a bystander who walks away feels less guilt). For an ill victim carrying a cane, the calculus reverses — costs of helping are low and costs of not helping (guilt at abandoning a blameless, genuinely sick person) are high — so helping is near-universal and fast. The absence of diffusion of responsibility also follows from the model once the visibility of the emergency is taken into account: a directly witnessed collapse produces strong, immediate arousal that cannot be dampened by assuming an unseen other will act, so the pressure to help remains high regardless of how many people are present. Even the same-race effect fits: where costs are already high (the drunk victim), bystanders become more selective, and an in-group victim may carry lower perceived costs of helping.
Subscribe to continue reading
Get full access to this lesson and all 9 lessons in this course.