We show that the essential meaning conveyed by individual words is best represented by a compass-like plane described by interrelated differentials of bad-good, weak-powerful, gentle-aggressive, and safe-dangerous, joined with a third dimension of structured-unstructured (GPADS).

We uncover a linguistic ‘safety bias’ by examining how words are used in large-scale, diverse corpora.

We find the power-danger-structure framework to be naturally aligned with token usage in real corpora as well as with seemingly disparate frameworks.

We construct and demonstrate the use of an ‘ousiometer’ for measuring time series of essential meaning.

From work emerging through the middle of the 20th century, the essence of meaning has become generally accepted as being well captured by the three orthogonal dimensions of evaluation, potency, and activation (EPA). Recast in psychology as valence, arousal, and dominance (VAD), these essential dimensions have become the cornerstone of sentiment analysis across many fields.

By re-examining first types and then tokens for the English language, and through the use of automatically annotated histograms—‘ousiograms’—we find here that:

1. The essence of meaning conveyed by words is not aligned with VAD but is best described by a power-danger-structure (PDS) orthogonal framework spanned by the semantic differentials of {weak ⇔ powerful}, {safe ⇔ dangerous}, and {structured ⇔ unstructured}.
2. The primary plane of the PDS framework is consistent with a circumplex-type model with the intercardinal axes forming a goodness-aggression-structure (GAS) framework with axes {bad ⇔ good} and {gentle ⇔ aggressive}, and both frameworks can be combined as GPADS.
3. Analysis of a disparate collection of large-scale English language corpora—literature, news, Wikipedia, talk radio, and social media—shows that natural language exhibits a systematic bias toward safe, low-danger words.
4. The Pollyanna principle's positivity bias in communication is, in fact, a one-dimensional projection of an underlying safety bias.

We demonstrate remarkable agreement between the PDS framework and (1) the circumplex model of affect; and (2) a data-driven determination of archetypes in fiction stories whose primary dimensions are {fools ⇔ heros}, {angels ⇔ demons}, and {traditionlists ⇔ adventurers}. Finally, we use our findings to construct a prototype ‘ousiometer’, a distant-reading instrument that measures essential meaning in large-scale texts.