A baby's scream will grab our attention no matter what's on the television or happening around us.

And now scientists have learnt why screams are so arresting - and it's not just because they are loud and high-pitched.

Screams activate 'fear circuits' in our brain because they have a subjective quality called 'roughness' where they quickly switch, or 'modulate', from loud to soft.
Image
© CorbisUsing an MRI scanner, researchers analysed the brains of people listening to recorded screams (Janet Leigh is shown screaming in a grab from the 1960 film Psycho). The scans revealed that the sound activates the amygdala region of the brain, which is typically associated with our fear response
This is what helps give them their jagged, jarring quality that set our nerves on edge.

The same quality of 'roughness' is also found in burglar and car alarms - suggesting engineers may have hit on the feature through trial and error.

David Poeppel, the study's co-author from New York University said: 'If you ask a person on the street what's special about screams, they'll say that they're loud or have a higher pitch.

'But there's lots of stuff that's loud and there's lots of stuff that's high pitched, so you'd want a scream to be genuinely useful in a communicative context.'

The researchers scanned the brains of people listening to recordings of screams taken from YouTube videos, films and volunteer screamers.

Using an MRI scanner, they analysed the brains of people listening to them in a sound booth at the Max Planck Institute in Frankfurt.

They noticed that screams activated parts of the brain not previously considered important in hearing.

Professor Poeppel said: 'We found that screams occupy a reserved chunk of the auditory spectrum, but we wanted to go through a whole bunch of sounds to verify that this area is unique to screams.

'In a series of experiments, we saw [that] this observation remained true when we compared screaming to singing and speaking, even across different languages.

'The only exception - and what was peculiar and cool - is that alarm signals (car alarms, house alarms, etc.) also activate the range set aside for screams.'

Normal speech has only small differences in loudness - between 4 and 5 hertz. A frequency of 1 hertz means that the sound happens once a second.
Image
© PAScreams activate these 'fear circuits' because they have a subjective quality called 'roughness' where they quickly switch - or 'modulate' - from loud to soft. Normal speech has only small differences in loudness - between 4 and 5 hertz. But screams can modulate very fast - varying between 30 and 150 Hz (illustrated)
But screams can modulate very fast, varying between 30 and 150 Hz, the researchers said.

When the team asked people to judge screams on how frightening they were, those with the highest roughness were judged the most terrifying.

Modifying the sound wave of a non-scream to be rougher can also make it scream-like.

The researchers then confirmed that increases in roughness correspond to more activation of the amygdala - the part of the brain where the fear response is activated.

Co-author Luc Arnal said: 'These findings suggest that the design of alarm signals can be further improved.
Image
© Radius Images/CorbisIn a series of experiments, the researchers said this observation remained true when we compared screaming to singing and speaking, even across different languages. The only exception was that alarm signals (smoke alarm pictured) were also found to activate the range set aside for screams
'The same way a bad smell is added to natural gas to make it easily detectable; adding roughness to alarm sounds may improve and accelerate their processing.'

The researchers plan to continue investigating human screams in the lab, particularly those of infants, to see if their screams are particularly rough.

They said they would also like to apply their analyses to animal screams to learn how much this trait is conserved across species.

The world's loudest scream was set at 129 db by teaching assistant Jill Drake in Greenwich in 2000.

The research is published in Cell Press.