"Long before it's in the papers"
January 27, 2015


Scientists report reading emotions from brain activity

June 19, 2013
Courtesy of Carnegie Mellon University
and World Science staff

Sci­en­tists say that based on brain ac­ti­vity, they’ve iden­ti­fied with some ac­cu­ra­cy which emo­tion a per­son is feel­ing—and used the re­sults to bet­ter learn how the brain or­ga­nizes emo­tions.

In a study pub­lished June 19 in the jour­nal PLoS One on­line, they com­bined brain scan­ning with com­put­er learn­ing tech­niques to meas­ure the rel­e­vant brain sig­nals.

Areas of en­hanced blood flow as­socia­ted with va­rious em­otions, shown by a com­puter mod­el. (Courtesy Carnegie Mellon U.) 

The find­ings pro­vide the first re­li­a­ble pro­cess to an­a­lyze emo­tions, the re­search­ers claim. A lack of re­li­a­ble ways to as­sess these had ham­pered past re­search, they said—mostly be­cause peo­ple can be re­luc­tant to hon­estly re­port their feel­ings, but al­so be­cause some emo­tions are un­con­scious.

The find­ings, from Car­ne­gie Mel­lon Uni­vers­ity in Pitts­burgh, build on pre­vi­ous work by Mar­cel Just and Tom M. Mitch­ell of the same school. They had used si­m­i­lar tech­niques to find out peo­ple’s thoughts of con­crete ob­jects, a pro­cess of­ten dubbed “mind read­ing.”

The new work could be used to as­sess some­one’s “e­mo­tional re­sponse to al­most any kind of stim­u­lus, for ex­am­ple, a flag, a brand name or a po­lit­i­cal can­di­date,” said Car­ne­gie Mel­lon’s Ka­rim Kas­sam, lead au­thor of the new stu­dy.

One prob­lem was to find a way to re­peat­edly and re­liably evoke dif­fer­ent emo­tions in the par­ti­ci­pants. Tra­di­tion­al meth­ods, such as show­ing sub­jects emo­tion-inducing film clips, would have worked poorly be­cause the clips’ im­pact fades with re­peat­ed dis­play, the in­ves­ti­ga­tors said. 

To solve this, they brought in ac­tors from the uni­vers­ity’s dra­ma school and asked them to work them­selves in­to dif­fer­ent emo­tional states re­peat­edly while ly­ing in a brain scan­ner—a func­tion­al Mag­net­ic Res­o­nance Im­ag­ing ma­chine, which meas­ures blood flow in the brain. 

The re­sults from this and oth­er tests were then fed in­to a pro­gram de­signed to use the in­forma­t­ion to “learn” the sig­na­tures of the var­i­ous emo­tions. 

When new peo­ple were placed in the scan­ner, the com­put­er was found to guess their emo­tions with a 71 per­cent “rank ac­cu­ra­cy.” This means when the com­put­er pro­duced a list of its guesses from most to least like­ly, the right an­swer did­n’t al­ways top the list, but on av­er­age it was near the top—the 71st per­cen­tile.

The study shows that de­spite ob­vi­ous dif­fer­ences be­tween peo­ple’s psy­chol­o­gy, our brains tend to en­code emo­tions “in re­markably si­m­i­lar ways,” said Aman­da Mark­ey, a grad­u­ate stu­dent in­volved in the re­search.

A sur­pris­ing find­ing, the sci­en­tists said, was that al­most the same ac­cu­ra­cy was achiev­a­ble even when the com­put­er read pat­terns in only one part of the brain. “This sug­gests that emo­tion sig­na­tures aren’t lim­it­ed to spe­cif­ic brain re­gions, such as the amyg­da­la, but pro­duce char­ac­ter­is­tic pat­terns through­out a num­ber of brain re­gions,” said Vlad­i­mir Cher­kass­ky, a re­search pro­grammer at the uni­vers­ity in­volved in the work.

The re­search team al­so found that the com­put­er was best at iden­ti­fy­ing hap­pi­ness, worst at iden­ti­fy­ing en­vy, and rarely con­fused pos­i­tive and neg­a­tive emo­tions—sug­gest­ing these have dis­tinct sig­na­tures. It was al­so least likely to con­fuse lust with oth­er emo­tions, sug­gest­ing lust cre­ates brain ac­ti­vity dis­tinct from oth­er feel­ings.

“We found that three main or­gan­iz­ing fac­tors un­der­pinned the emo­tion neu­ral [brain] sig­na­tures,” said Just. These are the emo­tion’s pos­i­ti­vity or nega­ti­vity; its in­tens­ity; and “its so­cial­ity—in­volve­ment or non-involvement of anoth­er per­son. This is how emo­tions are or­gan­ized in the brain.”

* * *

Send us a comment on this story, or send it to a friend

Sign up for

On Home Page         


  • St­ar found to have lit­tle plan­ets over twice as old as our own

  • “Kind­ness curricu­lum” may bo­ost suc­cess in pre­schoolers


  • Smart­er mice with a “hum­anized” gene?

  • Was black­mail essen­tial for marr­iage to evolve?

  • Plu­to has even cold­er “twin” of sim­ilar size, studies find

  • Could simple an­ger have taught people to coop­erate?


  • F­rog said to de­scribe its home through song

  • Even r­ats will lend a help­ing paw: study

  • D­rug may undo aging-assoc­iated brain changes in ani­mals

Scientists say that based on brain activity, they’ve identified with some accuracy which emotion a person is feeling—and used the results to better learn how the brain organizes emotions. In a study published June 19 in the journal PLoS One online, they combined brain scanning with computer learning techniques to measure the relevant brain signals. The findings provide the first reliable process to analyze emotions, the researchers claim. A lack of reliable ways to assess these had hampered past research, they said—mostly because people can be reluctant to honestly report their feelings, but also because some emotions are unconscious. The findings, from Carnegie Mellon University in Pittsburgh, build on previous work by Marcel Just and Tom M. Mitchell of the same school. They had used similar techniques to find out people’s thoughts of concrete objects, a process often dubbed “mind reading.” The new work could be used to assess someone’s “emotional response to almost any kind of stimulus, for example, a flag, a brand name or a political candidate,” said Carnegie Mellon’s Karim Kassam, lead author of the new study. One problem was to find a way to repeatedly and reliably evoke different emotions in the participants. Traditional methods, such as showing subjects emotion-inducing film clips, would work poorly because the clips’ impact fades with repeated display, the investigators said. To solve this, they brought in actors from the university’s drama school and asked them to work themselves into different emotional states repeatedly while lying in a brain scanner—a functional Magnetic Resonance Imaging Machine, which measures blood flow in the brain. The results from this and other tests were then fed into a program designed to use the information to “learn” to tell apart the signatures of the various emotions. When new people were placed in the scanner, the computer was found to guess their emotions with a 71% “rank accuracy.” This means when the computer produced a list of its guesses from most to least likely, the right answer didn’t always top the list, but on average it was near the top—the 71st percentile. The study shows that despite obvious differences between people’s psychology, our brains tend to encode emotions “in remarkably similar ways,” said Amanda Markey, a graduate student involved in the research. A surprising finding, the scientists said, was that almost the same accuracy was achievable even when the computer read patterns in only one part of the brain. “This suggests that emotion signatures aren’t limited to specific brain regions, such as the amygdala, but produce characteristic patterns throughout a number of brain regions,” said Vladimir Cherkassky, a research programmer at the university involved in the work. The research team also found that the computer was best at identifying happiness, worst at identifying envy, and rarely confused positive and negative emotions—suggesting these have distinct signatures. It was also least likely to confuse lust with other emotions, suggesting lust creates brain activity distinct from other feelings. “We found that three main organizing factors underpinned the emotion neural [brain] signatures,” said Just. These are the emotion’s positivity or negativity; its intensity; and “its sociality — involvement or non-involvement of another person. This is how emotions are organized in the brain.”