Reasoning after a terrorist attack

On Thursday the 7th of July 2005, suicide bombers set off bombs in central London. There were four bombs: three on underground trains and one on a bus. The resulting explosions killed 52 and injured 770. Blanchette, Richards, Melnyk, and Lavda (2007) investigated the effects of these attacks on how people reason, using participants from London (UK), some of whom actually witnessed the attacks or were in the same area of city, and from Manchester and London in Canada, progressively further from the affected area.

Rapid gist of the reasoning task

The authors used syllogisms as their reasoning task, selecting two forms of problem. One is classically true in some, but not all, models:

Some A are B
Some B are C, therefore
Some A are C

Here the Some A are B and the Some B are C are the premises which you’re supposed to assume are true and the Some A are C is a conclusion.

The other form they use is classically true in all models:

Some B are A
All B are C, therefore
Some A are C

The A‘s, B‘s and C‘s were instantiated with three types of content: about terrorism (e.g. suicide bombers), generally emotional content (e.g. paedophiles, leukaemia), and neutral (e.g. reading romance novels), and the conclusion of the syllogism was either believable or unbelievable. That gives a total of 12 syllogisms from crossing all the content types, believability, and whether the syllogism was classical valid or invalid. A couple of examples:

(1) Some Muslims are terrorists, some terrorists are suicide bombers, therefore some suicide bombers are Muslim.

To enter the classical mindset, imagine a world where all you know about suicide bombers and terrorists is the two premises above. It’s certainly possible that in that world there’d be no Muslim suicide bombers. The smallest example is one where there are two terrorists. One is a Muslim and a terrorist but not a suicide bomber. The other is a terrorist and a suicide bomber but not Muslim. So, classically, you’re supposed to say that this argument is invalid. The authors hypothesised that for this kind of problem, people in London would be more likely to be influenced by their beliefs and so say that the argument was valid.

(2) Some terrorist attacks are murders, all terrorist attacks are moral acts, therefore some murders are moral acts.

This is an example of an unbelievable conclusion which, with a classical interpretation, does follow from the premises (classically). If some terrorist attacks are murders then there’s (at least) one that’s a murder. But that murder is also a moral act by the second premise, therefore some murders are moral acts. You may doubt the truth of the premises however.

The first experiment—one week after the attacks

The participants (around 30 in London, 30 in Manchester, and 15 in Canada) were all asked questions about their emotional state. The Londoners (UK) showed slightly more fear than the others (mean score of 55 versus 48 for Manchester and 37 for Canada; scores ranged from 1-100). Counterintuitively, perhaps, London showed least despair and most positive emotion. The authors explain this by arguing that the Londoners had more opportunities for “active coping”, e.g. by helping the authorities with various tasks during the aftermath of the attacks.

For the reasoning part, each participant saw 4 problems, chosen at random from the 12 under the constraints that 2 were valid, and 2 unbelievable. They were asked to decide if the syllogism was valid or not. The syllogisms were scored according to whether the answer was consistent with classical logic.

In general, people were most accurate on neutral problems, followed by terrorism content, followed by general emotional. There were no differences found between the three locations in terms of overall accuracy, nor was an interaction found between content type and location.

Now an interesting result: focusing on the terrorism content and incongruent problems (believable and invalid or unbelievable and valid), participants from London were most accurate with respect to classical logic (75%), followed by Manchester (68.4%), followed by Canada (27.3%). This is not at all what the authors expected—they thought that people from London would be more likely to show effects of belief bias.

The participants were asked to predict on a scale from 1-100 how likely it was that there’d be a similar attack in Britain within one month, one year, and five years. This led to another interesting result: those who were not classical in their responses to the terrorist-incongruent conjectures gave a higher probability that there’d be an attack in the following month (a mean of 34.5 versus 19.0).

The second experiment—six months later

In each location around half of the previously recruited participants took part. The stimuli were the same, except an unseen set of materials were given to each participant. Focusing on the reasoning results, again the groups differed only for the terrorism content and incongruent problems. Londoners was most accurate (83.3%), Canada this time came second (50%), then Manchester (38.5%).

Did the result relating risk estimates to logicality still hold? We don’t know as the analysis is not reported.

Interesting stuff

I don’t want to give my spin on how to interpret this all just yet—stay tuned. But as you may guess, I’d want to argue (following, e.g., Stenning and van Lambalgen) that a broader notion of “logical” would be more useful.


Blanchette I., Richards, A., Melnyk, L., and Lavda A. (2007). Reasoning About Emotional Contents Following Shocking Terrorist Attacks: A Tale of Three Cities. Journal of Experimental Psychology: Applied, 13(1), 47-56.