The neuron fires on very high-frequency/common tokens (e.g. “of,” “an,” “I’m,” “more,” “professors,” closing quotes, etc.), effectively detecting the most frequent subword units.