Spatial neurons not fitting a specific sub-type.
What It Does
Space.General neurons encode spatial meaning that doesn't fit cleanly into the six specific sub-types. They activate on general spatial language: prepositions with spatial meaning ('in', 'on', 'at', 'near', 'between', 'through'), spatial metaphors ('the argument fell apart', 'the theory stands up'), and mixed spatial references that combine position, direction, and containment.
How It Behaves
Space.General neurons are distributed across all layers and are the largest Space sub-type by count. They represent the generalized spatial processing capacity of language models — the ability to parse and represent spatial meaning across all its forms. They interact with all other Space sub-types and with Time neurons (spatial movement has temporal extent).
Research Example
In Gemma 2B, Space.General neurons activate strongly on spatial prepositions ('in', 'on', 'under') even in non-spatial metaphorical contexts ('under the influence', 'on the edge of collapse', 'in trouble'). The model has generalized spatial preposition processing, using Space.General neurons as a substrate for spatial metaphor comprehension across all domains.