How your response to a crisis may have less to do with your character and more to do with the people standing next to you, writes forensic engineer Sean Brady.

The stolen beer

In 1968, the Nu-Way Beverage Center, a discount beer store in Suffern, New York, was the victim of 96 robberies over a two-week period, all of which followed an almost identical pattern.

Two men, Paul Bonnarigo and Malcolm Ross, either singly or in pairs, would enter the store and ask the cashier, "What is the most expensive imported beer that you carry?". The cashier would reply, "Lowenbrau, I'll go back and check how much we have", then he'd disappear into the rear of the store. With the cashier gone, one of the men would pick up a case of beer, say to no one in particular, "They'll never miss this," then walk outside, put the beer in their car, and drive off.

When the cashier returned to the front of the store, things got interesting. He would walk back in, stand behind the counter, and say nothing. At this point, only a small percentage of customers would spontaneously report the robbery: 20%.

If a customer said nothing, the cashier would let one minute elapse, then casually enquire, "Hey, what happened to that man (those men) who was (were) here? Did you see him (them) leave?" When prompted, only 51% of the remaining customers reported the theft, with the other 49% remaining silent.

By now, you probably realise this was an experiment. The two men were students from Columbia University and were working with the cashier. The experiment was designed to investigate how people, in this case the customers in the store, would respond in a crisis.

The experiment showed that one factor, more than any other, influenced whether the customers reported the theft. It didn't matter if there were one or two robbers. Nor was the customer's gender significant – women and men were equally likely to report. What mattered was the number of customers in the store. And while you may think that a higher number of people increased the likelihood of at least one of them reporting, the opposite was true: a higher number decreased the likelihood of a report – a real-world demonstration of what psychologists refer to as the ‘bystander effect’.

The smoke-filled room

The bystander effect describes how people fail to intervene in a crisis due to the presence of others. It was most famously illustrated in the late 1960s by an experiment conducted by Bibb Latané and John Darley.

Students living in campus residences at Columbia University – the subjects – were invited to participate in a discussion about the problems involved with life at an urban university. When they arrived for the interview, they were either directed by a secretary or guided by signs to a waiting room. Inside was a sign requesting that they fill in a preliminary questionnaire. Some subjects completed the questionnaire alone, while others did so in groups of three.

The experiment began when the students had completed two pages of the questionnaire. At this point, an experimenter, who could observe the room through a one-way glass window, began to introduce harmless 'smoke' – it was titanium dioxide – into the room via a vent. The experimenters then observed how the subjects behaved when they noticed this fine-textured, but clearly visible, stream of whitish smoke, and they measured how long it took subjects to leave the room and report it.

It would turn out that the time taken for somebody to report the smoke was highly dependent on both the number and reaction of the other subjects in the room.

When one person was alone in the room, they reacted very much as you'd expect. Shortly after the smoke was introduced, the subject would glance up from their questionnaire, see it, then have a slight but distinct startled reaction.

They'd then go through a short period of indecision. Some would return to their questionnaire, only to glance up again. Most subjects would get up and walk to the vent to inspect the smoke.

Some would sniff it, some waved their hands through it. There would be another period of indecision. Then most subjects would leave the room, find somebody, and report the smoke.

Nobody panicked, and most subjects reported that there was 'something strange going on in the room, there was some sort of smoke'. The median time for a report was two minutes, and 75% of subjects had reported the smoke before the experiment was terminated at the six-minute mark.

Twenty-five per cent, six people, didn't make a report. These six subjects typically said in interviews afterwards that they didn't report it because they thought the situation was non-dangerous or was part of the experiment. Latané and Darley concluded that the reason these subjects didn't report the smoke wasn't because they were apathetic, but because they defined the situation as non-serious – thus avoiding the moral conflict of having to decide whether to take action or not.

The results were quite different when there was more than one person in the room. Eight groups of three subjects were tested, and the time taken for the first subject of each group to report the smoke was recorded.

Before we explore what happened, it is worth setting out what we would have expected to happen from a rational perspective: one person in the room reported the smoke 75% of the time; so with three people – each with a 75% probability of reporting – the odds of at least one person reporting should have been more than 98% for each test. But this didn't happen: with three subjects in the room, the odds of the smoke being reported dropped from 75% to 38%.

Of the 24 subjects tested across eight groups, only one person reported the smoke within four minutes of noticing it, by which stage the room had become very unpleasant. Overall, only three people reported the smoke, each taking considerably longer than the median time of two minutes for a single subject in the room.

The experimenters then repeated the three-subject-in-a-room experiment, but added a twist. They placed one real subject (known as the naive subject) in the room, along with two subjects who were in on the experiment (known as the two passive confederates).

The tests began, and these two confederates avoided conversation with the naive subject and worked on their questionnaires. When smoke entered the room, they glanced at it briefly, didn't comment, shrugged their shoulders, and returned to filling in their questionnaires.

They occasionally waved away the smoke, and if the naive subject addressed them about it, they acted indifferently and said, "I dunno". Of the 10 groups tested, only one of the 10 naive subjects reported the smoke, ie, the odds of reporting had dropped to 10%. The other nine naive subjects stayed in the room, worked on their questionnaire, waved away the fumes from around their faces, coughed, rubbed their eyes, and even opened windows, but none reported the smoke.

This drop in reports is startling. When there was a single subject alone in the room, 75% of them reported the smoke. These odds dropped to 38% when there were three subjects, and then they dropped further to 10% for a single naive subject surrounded by two passive confederates acting as if everything was fine.

The odds dropped due to two factors: when the number of subjects in the room increased from one to three; and when the naive subject observed the two passive confederates exhibiting no concern.

Based on this and other experiments, Latané and Darley proposed that two distinct psychological concepts can explain the subjects' behaviour. One is diffusion of responsibility, the other is social influence (or social definition).

Diffusion of responsibility

The first of these two concepts relates to the number of bystanders in the crisis. When you are the only person there, it is obvious that you are the only one who can act.

Not only are you solely responsible for acting, you are also solely responsible for any inaction. You will have to live with your decision not to offer assistance.

But when other people are there, the onus to act is not unique to you or any one individual. The responsibility to act is diffused, as is the blame for any inaction on your part.

Latané and Darley suggest that you will tend to feel that help is probably on the way, or that other bystanders are probably better qualified to assist, or that it's better not to get involved. Once others are there, it's less obvious that you should be the one to intrude on the situation.

Interestingly, this holds true even if you can't confirm whether somebody else is actually intervening and providing assistance. In another experiment conducted by Latané and Darley, a naive subject was placed in the somewhat artificial situation where they could not see other bystanders – all of whom were confederates.

Each person was in a separate room, with a microphone and headphones, and the naive subject knew how many other bystanders were involved. Then one confederate faked having a medical fit. The naive subject could hear the fit over their headphones, but they couldn’t hear whether any of the other bystanders were getting help for the confederate having the medical fit.

When the naive subject was the only bystander, they sought help for the victim in 85% of cases. But when there were four other bystanders in the group, this percentage dropped to 31% – despite the naive subject being unable to ascertain whether any of the other bystanders were providing any assistance.

Latané and Darley stress that the subject’s response was not apathetic: at the conclusion of the experiment, these subjects showed signs of nervousness and heightened emotion, and they asked if the person having the fit was okay and being looked after. They were clearly concerned about the victim, despite making no report. Latané and Darley describe this response as an avoidance-avoidance conflict: they felt guilty because they were not helping the victim, but they were also afraid of embarrassing themselves by reporting the fit and ruining the experiment.

The concept of diffusion of responsibility illustrates that the more bystanders there are to a crisis – regardless of whether or not the bystanders can confirm if somebody else is providing assistance – the less likely they are to intervene.

Social influence

Social influence is the second concept that contributes to the bystander effect, and it relates to how other bystanders are reacting.

Latané and Darley say that most crises are initially ambiguous – for example, is the smoke coming from the vent a serious concern? Before acting, you need to determine if the situation is indeed a genuine crisis – but because there is ambiguity, this can be difficult.

As a proxy, you will look to the reactions of the others around you. If they appear unconcerned, you'll likely feel less concerned, and you'll assess the situation as non-serious. But Latané and Darley also say that people try to remain calm when confronted with an ambiguous situation – they are all too aware that the other bystanders are watching them, and they won't want to appear like they're overreacting. And you'll try to do the same. The problem is that you and the other bystanders will likely misinterpret the apparent 'calm' of others and decide there is no crisis.

A fascinating aspect of social influence was highlighted when subjects in the smoke-filled room experiment were asked the question: did the reactions of the other bystanders influence you? They gave a surprising answer. They all said no. They seemed unaware that their decision-making process had been influenced by others.

Closure

If you were standing in the Nu-Way Beverage Centre in Suffern, New York, on one of the 96 occasions it was robbed in 1968, you'd probably like to think that you would have reported the theft to the cashier. But these bystander effect experiments suggest that your likelihood of doing so is shaped more by the situation – the number of people around you and their reactions – rather than by your individual character. A failure to respond is less about apathy and more about psychology.

This finding is not only personally confronting, but it upends assumptions about how people respond in a crisis. Having more people around makes intervention less, as opposed to more, likely because the responsibility to act is diffused among the group.

And people will look to those around them for cues on how to respond, rather than making a rational assessment. Latané and Darley note a powerful parallel: just as a crowd can whip itself up into destructive action like a riot, it can also be paralysed into collective inaction. The danger isn't that good people fail to act, but that we wrongly assume they always will.

This article was originally published in the December 2025 issue of OHS Professional Magazine, available here: https://www.bradyheywood.com.au/wp-content/uploads/2025/12/OHSProfessional-Dec2025-p36-41.pdf 

Author: Dr Sean Brady, managing director, Brady Heywood. Email: sbrady@bradyheywood.com.au