Human screams can convey at least six emotions

Alone in a small, padded room, Sascha Frühholz took a deep breath and unleashed an ear-splitting scream. He was there, in part, because of The Beatles.

Frühholz, a cognitive neuroscientist at the University of Oslo in Norway, couldn't get videos of the band's 1960s concerts out of his mind. As the music begins to bop, the audience viscerally reacts with joy, shrieking and screaming. There was even a name for this: Beatlemania. "These people don’t have any other way to express this overwhelming elation," he says.

While the observation may seem obvious, scientific studies on human screams have focused almost exclusively on vocalizations of anguish—and this oversight nagged at Frühholz. He and his colleagues set out to characterize the screams we utter for a range of emotions, negative and positive. By studying screams recorded in the small, padded room, the team identified six acoustically distinct scream categories: pain, anger, fear, joy, passion, and sadness. The study was published in PLOS Biology.

Unexpectedly, the researchers also found that volunteers more readily recognized—and their brains more efficiently processed—screams that were not considered alerts, including joy, passion, and sadness, compared to the screams of pain, anger, and fear. For all animal species, screams are considered a vital way to rapidly communicate danger to others nearby; why the joyful screams of this latest study seemed to invoke the strongest response remains unknown.

The study of non-verbal vocalizations in humans is relatively new, says Katarzyna Pisanski, a voice researcher at the University of Lyon who was not part of the study team. Most of the early work on humans has focused on speech and language since they’re unique in the animal world. "It’s what makes us human," she says.

But an increasing number of studies are looking at non-verbal vocalizations like screams and laughs, similar to the sounds made by other creatures. Humans express these sounds with remarkable diversity, and the function of the varying acoustic forms may hold the keys to helping us understand the evolution of human communication.

"We need to study what makes us the same to understand how we are different," Pisanski says.

Creating a scream 

Frühholz and his colleagues initially recorded their own screams while they attempted to identify the typical range of emotions that spark these intense utterances. They came up with various scenarios, like thinking of how you'd scream if your favorite soccer team won the championship, and then attempted to recreate it.

They eventually settled on six different screams they wanted to evaluate: pain, anger, fear, joy, passion, and sadness. They recruited 12 volunteers to scream with each emotion. The volunteer was primed with a description of an emotion-evoking scenario for each scream type, such as getting attacked by a stranger in a dark alley. Each person would also record a "neutral scream" for comparison, which is just an intense utterance of "ahh." They then instructed the participant to let loose in the soundproof room.

"It's not really difficult," Frühholz says of re-creating screams for different emotions. But a lot of screaming can be exhausting. "It’s the most intense vocalization that we actually can produce," he says.

One challenge with all these studies is that they must be done in a lab setting. It's unethical to cause pain or fear in study subjects, notes Pisanski. So the options to study screams are limited: they can either be acted or pulled from previous recordings like those found on YouTube.

The acted screams tend to be a bit more uniform than natural screams, but past work suggests that they’re fairly accurate, Pisanski says. “In general, given the limitations in finding genuine vocalizations, it’s the best we can do,” she says. “And people are pretty good at it.”

The team analyzed recordings of each scream by looking at 88 acoustic features, such as measurements characterizing pitch and intensity. They trained a computer algorithm on the various features that differed between screams and found it could correctly categorize screams nearly 80 percent of the time. The most accurate classification was for joy, with 89.7 percent correct classifications.

The team then studied participants listening to the recorded screams, measuring how quickly they could categorize the emotion triggering the scream by clicking an option on a computer screen. In one set of trials, they tested people's ability to select the scream type from all six emotions or neutral, and in another, the listeners only had the option of picking one of two scream types. The team also created maps of brain activity for people listening to playbacks of the screams using functional magnetic resonance imaging (fMRI).

They were interested in three particular brain systems in the fMRI scans, Frühholz explains. The first was the auditory system, which is involved in analyzing and classifying each sound. The second was the limbic system, which is involved in emotional responses, particularly during survival situations. Finally, the frontal cortex, which is involved in decision-making and helps put the sound in the broader context of a situation.

A loud surprise

The researchers unexpectedly found that listeners could most quickly recognize the non-alert screams, and in particular, joy. They more slowly recognized screams from negative emotions, including pain, fear, and anger. Similar patterns also held for fMRI analysis, which showed non-alert screams sparked greater activity in listener’s brains compared to the alert screams. Exactly why, however, remains uncertain.

The finding runs counter to the believed evolutionary function of a scream as a way to readily convey danger to anyone in earshot. "It's surprising," Pisanski says, adding that she's unsure what might drive the result.

In the last two decades, Frühholz says, the view of the brain as a "threat detector" has become increasingly common among scientists. But the new study hints this idea might not be the case for screams.

"It will make us think about screams in a more nuanced way," says Adeen Flinker, assistant professor of neurology at New York University, who was not part of the new study. In a 2015 study, Flinker and his colleagues identified a harsh, high-pitched sound variation, known as roughness, as a key in boosting a listener's ability to quickly detect sounds intending to alert, including not only screams, but also artificial alerts like whistles.

The new study identified the feature in both negative and positive screams, though the roughness is weaker in positive screams, Flinker notes. But even with this roughness, participants didn't recognize and process negative compared to positive screams as readily. While this new find doesn't necessarily eliminate the significance of roughness in provoking a response to sounds of alert, "it complicates things," he says.

It’s possible the listener's environment can affect how each scream is perceived, Flinker notes. If listeners imagine standing in a dark alley before hearing a scream, that might influence how a scream is interpreted regardless of the screamer's emotion.

In a perhaps a less-surprising result, the new study also found that positive screams were the ones most frequently misidentified as alert screams. Such a mistaken identification of the emotion behind a scream, it seems, would be beneficial to humans through time. As Pisanski says, "better safe than sorry."

More research will help scientists further break down the human response to different types of screams. While a scream may seem a far cry from everyday words, studying such nuances in vocalizations and what these non-verbal sounds communicate to others is important to tracing language to its roots, Pisanski says.

"To understand the evolution of human vocal communication and ultimately how we came to speak," she says, "we really need to understand all of these differences."