The neuron fires strongly on document‐structure markers such as section headings or metadata tokens (e.g. “Abstracts,” “Reference,” “Posted,” “edited,” large numeric values in specs), i.e. it identifies section titles and other header/metadata elements rather than ordinary text.