What inner feelings do flies sing?

“I left chocolate chip cookies on the table.”

What would be your reaction to hearing that statement? If you hadn’t eaten for the past 10 hours, you might jump and run towards the cookies. If you had just eaten a buffet dinner, you might not blink an eye. In other words, our responses to the surrounding world often depend on our internal states, such as hunger, thirst, and arousal.

For us humans, behavioral states like hunger might seem easy to intuit; we can distinguish when we’re hungry from when we’re not, and we can tell others about how we’re feeling. Even then, descriptions of our internal states are highly subjective and hard to quantify—how hungry is “really hungry”? This problem is further magnified in non-human animals: we cannot directly ask a fruit fly how it’s feeling today.

Figure 1: Schematic of the hypothesis that an identical courtship situation might elicit different male songs depending on the male’s internal state.

Figure 1: Schematic of the hypothesis that an identical courtship situation might elicit different male songs depending on the male’s internal state.

Calhoun and his colleagues at Princeton University sought to tackle the challenge of identifying behavioral states in fruit flies and how they alter the flies’ responses to sensory input. In particular, they were interested in how flies’ internal state changes their courtship behavior. During courtship, the male fly sings to the female fly, and his songs are dynamically patterned in response to his sensory experience. For example, the male might alter his songs based on his distance to the female, his forward velocity, and her rotational velocity. The authors wondered if, in the same way that we react differently to the same information when we’re hungry or full, the fly had internal states that changed how such sensory cues shape song output.

To address whether flies have internal states that shape courtship, the scientists first video-and-audio-recorded hundreds of hours fly courtship. From the videos, they extracted many sensory cues such distance and velocity, and from the audio, they extracted songs. They then built a virtual male fly model that could transform the sensory cues extracted from the video into a prediction of male fly song.

This model did a decent job of predicting songs, but with room for improvement (Figure 2, middle). Perhaps a shortcoming of the model was that it didn’t account for different internal states­ that might alter how sensory cues shape song output. A model with two internal states performed a little better, and a model with three performed even better, almost perfectly predicting how the male fly would respond to particular sensory cues (Figure 2, bottom). A model with four or more states performed no better than one with three.

Figure 2: Example of a recorded fly song and its model predictions using 1 state and 3 states. Colors indicate different song modes in the recording and in the predictions.

Figure 2: Example of a recorded fly song and its model predictions using 1 state and 3 states. Colors indicate different song modes in the recording and in the predictions.

The authors concluded from this result that male flies had three internal states during courtship. They then went back to the video recordings and saw that the fly was in one of the behavioral states most often when it was “chasing” the female. It was in another state when it was “close” to the female. The remaining state basically captured all times when the fly was neither “chasing” nor “close”. In other words, the male interpreted the same sensory cues differently depending on whether he was close to the female, chasing her, or neither.

So next time you react to the news of cookies, think about how your response might depend on whether you’re “hungry” or “full”. Then remember that flies also have similar internal states that govern how they respond to their sensory input. Buzz buzz.

Read the source study here: https://www.nature.com/articles/s41593-019-0533-x

Edited by: Isabel Low

References

Calhoun, A. J., Pillow, J. W., & Murthy, M. (2019). Unsupervised identification of the internal states that shape natural behavior. Nature neuroscience22(12), 2040-2049.