Piliavan - 'Subway Samaritan'

• background

  • murder of Kitty Genovese
  • bystander behaviour: an individual is less likely to to help in a situation if other people are present - pluralistic ignorance and diffusion of responsibility
• aims

  • to investigate bystander behaviour in a more natural setting - on a subway train
  • to see if characteristics of the victim would affect the extent to which help was forthcoming (drunk/cane, black/white)
• sample

  • opportunity sample of passengers who were travelling on a New York subway on a weekday
  • didn't volunteer so couldn't give informed consent
  • total = about 4,450 men and women
  • mean number of people per car was 43, mean number of people in 'critical area' was 8.5
• method

  • confederates played the role of victim and models
  • all general studies students from Columbia Uni, aged between 24 and 35
  • 4 teams of 4 students: male victim, male model, and 2 female observers
• procedure

  each trial

  • consisted of a series of 103 separate trials
  • the team boarded the train using different doors and varying the location
  • after about 70 seconds the victim staggered forward and collapsed and remained lying on his back until receiving help
  • train doesn't stop for 7 1/2 minutes
  • if no help was forthcoming when the train stopped the model helped the victim to his feet
• procedure

  victim

  • 4 male victims dressed identically
  • 38 trials = drunk, 65 = sober with a black cane

  model

  • 4 white males aged 24-29 played the models

  4 possibilities:

  • critical area, early model (70 secs)
  • critical area, late model (150 secs)
  • adjacent area, early model (70 secs)
  • adjacent area, late model (150 secs)
  • there was also a 'no-model' condition
  • choice of condition picked using random number table
• procedure

  measures

  on each trial one observer noted:

  • race, sex, location of every person seated or standing in the critical area
  • total number of persons in carriage
  • total number who helped the victim
  • race, sex, location of every helper

  the second observer noted:

  • race, sex, location of every person in the adjacent area
  • time when help was first offered

  both observers noted any comments made by the nearby passengers and also tried to elicit comments from passengers sitting near them
• results

  drunk or using cane

  • reported cane victim received spontaneous help 95% of the time (62/65)
  • drunk victim was spontaneously helped 50% of the time
  • help offered more quickly in the cane condition than the drunk condition
  • 87% of cane victims before model acted, whereas only 17% of drunk victims were helped before model
  • median latency (how long it took before help) for no model trials (no model condition) - 5 seconds = cane victim, 109 seconds = drunk trials
• results

  race and gender

  • black victim received less help and less quickly
  • slight 'same-race' effect
  • males more likely to help than females - 90% of first helpers were males - all male victims and models

  modelling

  • model intervening early (after 70 seconds) had slightly more effect than late model

  comments

  • more made in the drunk condition
  • most comments given in first 70 seconds
• results

  group size

  • more likely to help if in immediate vicinity of victim - appears to be reverse of the 'diffusion of responsibility' effect
  • could be explained in two ways:

  1. could be as original effect produced in lab and could work differently in field experiment
  2. the fact that potential helpers could see the victim may have reduced the tendency to diffuse responsibility
• research methods

  strengths

  • conducted in a natural environment using field experiment
  • labs create demand characteristics as previous research showed

  weaknesses

  • extraneous variables are difficult to control in a field experiment
  • may have been other factors that influenced the likelihood of helping - stressed by pressures of urban environment
• sample

  strengths

  • very large sample over period of several months - presumably contained wide cross-section of people
  • board target population
  • opportunity sampling wasn't restricted to the people in the area at one time

  weaknesses

  • all people sampling were comfortable in an urban environment - homeless
  • people used to offering/ignoring help
• ethnocentrism

  • target population represents an individualist culture
  • people in collectivist cultures might be more willing to help more quickly
• reliability

  • standardised procedures - different teams needed the same procedures (e.g. victim always dressed identically)
  • no inter-rater reliability - as only one observer was recording set things (e.g. where everyone was in critical area) and results weren't checked
• data

  • majority was quantitative
  • easy to compare but could oversimplify the the factors that appear to affect helping behaviour - simplistic account of bystander behaviour
  • qualitative was the comments made - personal insight into feelings and why they don't help
• ethics

  • difficult to obtain informed consent - behaviour would become unnatural
  • possible psychological harm

  1. distressed seeing someone collapse
  2. not helping might have creating anxiety
  3. felt conflict between helping and not helping

  • not given right to withdraw
  • not debriefed - left believing study was real