Fear is easy to put into words; saudade is not. This paper proposes that emotions are not fixed categories but paths through a few moving parts of one system: six moving parts that generate about 45 distinct feelings, each with its own substrate-path. Abstract. Existing emotion-taxonomies either stipulate discrete natural kinds (Ekman, Plutchik, Panksepp) or deny them in favour of constructed categories built on core-affect primitives (Barrett, Russell). Neither view generates a closed taxonomy with substrate-grounded mechanism. We propose an integrative architecture grounded in Friction Theory (FT; Pødenphant Lund 2026b) as instantiated in biological organisms by Behavioural Friction Theory (BFT; Pødenphant Lund 2026a). The framework specifies six moving parts: (1) friction's quality (strength, distribution, rhythm); (2) fields (Safety, Meaning, Ability, Effort); (3) layers (Biological → Emotional → Inner → Cognitive → External); (4) valence — read out from a substrate-mechanic of distributed reactance with hierarchical aggregation across the Friction Matrix (§3.0a); (5) dynamic effects (anticipation, surprise, prediction-error); (6) configurations (cross-field, ambivalence, misconfiguration, unresolved). The framework generates ~45 feeling-labels with specific substrate-paths, predicts psychiatric conditions as selective field × layer dysfunctions, predicts a developmental sequence in which emotion-categories reach functional maturity in a specific order with Meaning developing in two phases, makes specific cross-substrate predictions for large language models, and gives the Panksepp–Barrett scope-separation a generative field × layer mechanism (building on prior coexistence proposals). Falsifying evidence specified at three classes (F1 neural-dissociation, F2 maturation-ordering, F3 race-temporal-structure) plus three distributed-reactance conditions (F-DR1 single-locus, F-DR2 substrate-inconsistent, F-DR3 field-specific dissociation). Companion papers in the series: Paper 0 (BFT): 10.5281/zenodo.19462499 (concept) Paper 1 (FT generalised): 10.5281/zenodo.20012654 (concept) Paper 2B (ICL as WM, FT as LTM): 10.5281/zenodo.20145218 (concept) — empirical anchor for §3.0a distributed-reactance. Paper 14 (Logic as Reactance): 10.5281/zenodo.20217712 (concept) — substrate-mechanic foundation for §3.0a. Paper 27 (Friction as a Measurement of the Model): 10.5281/zenodo.20679636 — the measured density-vs-length dissociation behind the §3.5b aversion prediction. Paper 28 (Dread without a Dreader): 10.5281/zenodo.20690747 — the death-anxiety case of the non-terminating race. Concept-DOI 10.5281/zenodo.20058825 auto-points to the latest version. Series position. Paper 5 in the friction-theory paper series. v1.8 changes vs v1.7 (a substantive revision sharpening the valence account and extending the taxonomy, after external feedback received from two independent parallel reviewers): Valence substrate-mechanic (§3.0a) sharpened: positive valence beyond the resolved-default calm is a graded landscape, not a single signal — phasic reward as a rate-coded positive (the friction-drop a terminating race delivers at commit), the matched-optimum as a distinct steady positive (joy, via the experiential cascade), and hedonic liking as a dopamine-independent substrate; aligned conceptually with valence as the rate of change of free-energy (offered as a hypothesis, not an identity). Liking as processing-fluency (§3.5b) and its negative mirror: effortful-but-unresolvable processing (sustained Effort-friction that never commits) read as displeasure, with a falsifiable density prediction — pre-reading aversiveness tracks summed density, not length. Fear vs anxiety re-interpreted as terminating vs non-terminating race; hedonic adaptation and gratitude as absence-gradient dynamics. §4.3b attachment styles as Safety-field race-resolution clusters; §3.0b experiential pre-emption cascade; §3.8 motivation corrected to drive-toward-optimal-friction. Prior-art correction: the basic-emotion / constructionism coexistence is credited to prior proposals (Panksepp 2007; Barrett, Lindquist TenHouten 2021); the paper's contribution is the generative field × layer Friction Matrix, not the reconciliation as such. v1.9 changelog (June 2026). v1.9 refines the positive-affect account so that the inverted-U is the signature of a capacity-meeting-demand match, not a property of a field: Ability carries it intrinsically (for Ability, raw value coincides with the capacity–demand match), while Safety and Meaning manifest the same match-U derivatively when their load is met by a finite capacity — Safety via arousal (the Yerkes–Dodson curve; thrill) and Meaning via decoding-fluency (the aesthetic curve; beauty) — with Effort monotonic. The corresponding falsifiable prediction (§4.5c) is re-stated as a pre-registrable load-versus-capacity keying test with named mediators and predefined nulls. Two tentative, single-model cross-substrate notes are added and explicitly marked non-load-bearing: negative valence as a candidate readout of stored processing-cost (§3.0a), and a dissociation between an installable value-bias (comprehension-level content-gating, fine-tunable) and the felt emotion-machinery, which remains absent — function-install and affective machinery are orthogonal axes (§4.5/§4.5b). Companion-citation currency was updated to concept DOIs. The v1.9 additions went through external dual hostile peer-review.
Tomas Pødenphant Lund (Sun,) studied this question.