A number of studies have used functional MRI to see what our brain looks like as we recall pleasant memories, watch scary movies or listen to sad music. Scientists have even had some success telling which of these stimuli a subject is experiencing by looking at his or her scans. But does this mean it is possible to tell what emotions we are experiencing in the absence of prompts, as we let our mind wander naturally? That is a difficult question to answer, in part because psychologists disagree about how emotions should be defined. Nevertheless, some scientists are trying to tackle it.

In a study reported in the June 2016 issue of Cerebral Cortex, Heini Saarimäki of Aalto University in Finland and her colleagues observed volunteers in a brain scanner who were being prompted to recall memories they associated with words drawn from six emotional categories or to reflect on a movie clip selected to provoke certain emotions. The participants also completed a questionnaire about how closely linked different emotions were—rating, for instance, whether “anxiety” is closer to “fear” than to “happiness.” The researchers found that pattern-recognition software could detect which category of emotion a person had been prompted with. In addition, the more closely he or she linked words in the questionnaire, the more his or her brain scans for those emotions resembled one another.

Another study, published in September 2016 in PLOS Biology by Kevin LaBar of Duke University and his colleagues, attempted to match brain scans of people lying idle in a scanner to seven predefined patterns associated with specific emotions provoked in an earlier study. The researchers found they could predict the subjects' self-reported emotions from the scans about 75 percent of the time.

Not everyone agrees, however, that studying emotions this way—as averages of many people's brains while they undergo a stimulus—makes sense. Psychology professor Lisa Feldman Barrett of Northeastern University and author of How Emotions Are Made (Houghton Mifflin Harcourt, 2017), who was not involved in either study, says that so far no one has clearly demonstrated that patterns taken from one study can be used to recognize the same emotion in another group of people provoked by a different stimulus. Such brain patterns, Barrett says, are just statistical summaries, not unique signatures that exist only when someone has a certain experience. And one person's emotions may not look the same in a brain scan as another person's. “Maybe you have five [different] patterns for anger, maybe I have seven, maybe somebody else has two,” Barrett adds. “Maybe they overlap, maybe they don't.”

Going forward, we are likely to see diverse perspectives on what emotion is and how to study it. “For now,” Saarimäki says wryly, “I think we are still safer if you just ask people how they are feeling, rather than trying to read their brain.