A field-theoretic taxonomy of emotions

Paper 5 · Pødenphant Lund (2026f) · Read on Zenodo

Fear is easy to put into words. Saudade, the Portuguese sense of missing something that may never come back, is not. This paper proposes why: emotions do not arise as fixed categories, but as particular paths through a few moving parts of the same system. Six parts generate about 45 distinct feelings. And the same architecture explains why one feeling slides straight into language while another does not, and why two camps that have quarrelled for decades (Ekman and Plutchik on one side, Barrett on the other) are both right.

What is this about?

Emotion theory has long been split into two camps. On one side: the basic-emotions tradition (Ekman, Plutchik, Panksepp), which says there are discrete natural categories (anger, fear, joy) with dedicated brain circuitry. On the other side: the constructed emotion tradition (Barrett, Russell), which says the categories are constructed social labels on a continuous underlying state of affect.

Both are right, just about different levels. One architecture explains why both are right at the same time, and why they appear to contradict each other.

Six moving parts

Emotional experience is not one phenomenon. It is the product of six distinct components, each of which does a particular piece of work:

#	Component	Values	What it does
1	Quality of the friction	strength, distribution, rhythm	What you actually experience of the underlying computation
2	Fields	Safety, Meaning, Capability, Effort	Four computational domains from BFT
3	Layers	biological → emotional → inner → cognitive → external	Gradient from substrate signal to narrative feeling
4	Valence	positive, negative, unresolved	Direction of the friction
5	Dynamic effects	expectation, surprise, prediction error	Modulation over time
6	Configurations	cross-field, ambivalence, mis-configuration, unresolved	Composite states when several components play together

Each component is conceptually distinct and can be mapped (provisionally) to neural systems.

Emotions vs feelings

An important distinction (after Damasio): emotions arise as substrate signals; feelings arise as interpretive integrations of those signals. These are not two different phenomena on the same level, but two positions on a gradient of integration depth.

• Deep-layer substrate signals with dedicated circuitry (Panksepp's territory)
• Mid-layer integrative transitions
• Surface-layer feelings with narrative interpretation (Barrett's area)

The feeling dictionary: 50 feelings and their substrate-paths

The framework does not generate only the four or five basic emotions. It generates the whole folk-psychology lexicon of feeling: ~50 named feelings, each with a predicted substrate-path expressed as fields (Safety/Meaning/Capability/Effort), valence (positive/negative), race outcome, and layer position. The table is hypothesis-generating, not theory-testing: each row is a prediction that lesion or imaging studies ought to be able to confirm or refute. The value is taxonomic clarity. The framework's commitments become explicit and falsifiable for each individual label.

The Layer column: Deep = close to substrate signals (Level 2-dominated); Mid = mid-integration (Inner layer); Surface = narrative-interpretive (Level 3-dominated); N/A = non-emotion (drive, motivation, or field-absence).

A pattern in the table: higher-position labels are systematically harder to verbalise than lower-position labels. Fear is easier to put into words than saudade; calm is easier than the particular sense of sitting in late afternoon light and remembering someone you have lost. The framework explains this structurally: deep Level 2 signals are almost one-dimensional and verbalise cleanly, whereas surface-level cross-field configurations are high-dimensional and resist clean projection into language.

What comes out of it

• ~45 distinct feeling-labels with specific substrate-paths (Feeling Dictionary, §3.10)
• Psychiatric conditions are predicted as selective field × layer dysfunctions: the same diagnosis can arise from different substrate-paths
• A developmental sequence in which emotion categories come online as the underlying field × layer matures. Meaning develops in two phases (social/relational in childhood, existential in adolescence)
• Cross-substrate predictions for language models: functional prediction signals are present; field-emotional-layer signals and interpretive integration are absent, confirmed across 15 LLM architectures
• A resolution of the Panksepp–Barrett disagreement: substrate realism (the neural substrates exist and are dissociable) does not entail category realism (folk-emotion labels carve them at natural joints). Both describe different positions on the same gradient.

What falsifies the framework?

Three classes of falsifying evidence, each with a quantitative threshold:

• F1 Neural dissociation: Cohen's d < 0.3 in a meta-analysis of ≥20 studies
• F2 Evolutionary parallel: single-species disconfirmation
• F3 Race temporal structure: drift-diffusion fits underperform gate models, BIC > 10 across ≥10 datasets

Where does it fit in?

The framework is compatible with Damasio's somatic markers, Lazarus's core relational themes, Scherer's appraisal components, LeDoux's survival circuits, Adolphs and Anderson's emotion primitives, and culturally specific labels (amae, saudade, liget) as instances of the architecture.

What I don't know

This is an architecture, not a proof. The table of about 45 feelings is hypothesis-generating: each row is a guess at the substrate-path behind it, and it is lesion and imaging studies that have to decide whether the guess holds. I have made the framework's commitments explicit and falsifiable, and I have set three concrete thresholds for when it is wrong. But I have not run the studies that would settle the matter.

And when I say that language models show parts of the pattern, I mean the functional prediction signals. They do not have the field-emotional layers or the interpretive integration a human has. It is a parallel worth taking seriously, not a claim that a model feels anything.

Related papers

• Paper 0 — BFT — the biological primer; it supplies the four fields and five regulating layers the taxonomy uses
• Paper 1 — Friction Theory — the substrate-independent foundation; race mechanisms and the BFT ⊂ FT relation

You will find the full technical detail in the technical version: Paper 5 (English technical). The full paper is on Zenodo: DOI 10.5281/zenodo.20058825.