We know that someone’s tone of voice can be a good indicator of what they are feeling in that moment. Someone who is speaking quickly, loudly, and with a high-pitched voice is likely to be feeling angry—and many of us would be able to infer this from their vocal cues. But what if these cues were more subtle? Would we still be able to understand that they were upset based on their tone of voice?
The current study examined whether listeners’ ability to identify a speaker’s emotional state depended on the level of ‘emotional intensity’ with which an emotion was conveyed. Intuitively, it makes sense that high-intensity expressions would be better recognized than low-intensity expressions. But, less is known about how fine-grained variations in vocal cues influence listeners’ accuracy of interpretation.
To answer our questions, we created a series of recordings that varied from neutral expressions to full-intensity expressions of different emotions, in 10% intervals. We used actors’ full-intensity and neutral expressions from a previous study as end points, and merged them together to create recordings that were 10% neutral and 90% angry, 20% neutral and 80% angry, 30% neutral and 70% angry, etc. We presented these recordings to listeners in increasing order of intensity, and asked them to tell us what emotion was being conveyed in each. From this, we obtained an estimate of the slope of listeners’ accuracy across increasing levels of emotional intensity, for each emotion type.
We determined that listeners’ ability to identify each emotion did not increase linearly alongside the recordings’ intensity level. Even though the acoustic characteristics of the recordings changed linearly with intensity, listeners’ accuracy did not map onto that. Further, the exact pattern of change in accuracy across intensity levels varied for different emotions. For example, anger was easy for listeners to recognize, even at low intensity levels (see blue arrow in the figure above). In contrast, listeners didn’t start to be able to identify sadness until it was expressed with more than 50% emotional intensity (see green arrow).
These results tell us that some emotions may be much harder than others to identify at low intensities. This is important, because those low-intensity versions are likely more common than full-intensity expressions in the real world. The findings of the current study point us to emotion types that may be particularly challenging to identify at low intensities, such as happiness. They also make us think twice about assuming that listeners’ accuracy is a direct reflection of linear changes in the vocal cues associated with different emotions.