Language constraints
Cold reading works by exploiting vague qualifiers.
A statement calibrated with words like “sometimes,” “often,” “tend to,” “can be,” or “may struggle” lets the reader self-confirm against any internal evidence they care to recall, and the recall is biased toward whatever the statement primes.
This is the same Forer/Barnum mechanism the Anti-Barnum threshold targets at the score-coordinate level, applied at the prose level.
Metryval’s response is a public banned list and a positive replacement rule: every personality description must instead carry a body-location, a texture, and an intensity.
The banned qualifiers are not a style choice.
They are a measurement-validity guard.
The banned list
The following words and phrases are banned in any user-facing Metryval-authored prose that describes the respondent or their pattern.
The list is enumerated rather than gestured at because enumeration is what makes it auditable.
- “sometimes”
- “often”
- “tend to” (and “tends to”)
- “can be”
- “may struggle” (and “may find”)
- “occasionally”
- “from time to time”
- “to some degree”
- “in certain situations”
- “at times”
The list is closed; additions require a documented rationale and a corresponding update to the saturation-cell linter described below.
The list is also strict; a copy variant that substitutes a synonym for the same hedging function (for example, “in moments” in place of “at times”) is rejected on the same grounds.
The somatic-specificity requirement
Every personality description must include three concrete anchors: a body location, a texture, and an intensity.
These are the specifics that vague qualifiers cannot impersonate.
- Body location. Jaw, sternum, chest, belly, spine, throat, shoulders, lower back, knees, hands, eyes, temples. The anchor is anatomical, not metaphorical.
- Texture. Knotted, brittle, fluid, crackling, luminous, dense, spacious, ropey, glassy, dull. The descriptor is sensory and concrete.
- Intensity. Faint hum, grinding force, electric jolt, low simmer, steady pull, sudden snap. The magnitude is qualified by kind, not by frequency adverb.
A compliant sentence: “A grinding tightness sits behind the sternum when an unfinished promise enters the room.”
A banned-equivalent sentence the rule rules out: “You often feel uncomfortable when there are unfinished things around.”
The first cannot be self-confirmed by every reader because it commits to a specific bodily configuration; the second can be self-confirmed by almost any reader precisely because it commits to nothing.
The somatic anchor lives in the prose register only.
It is not a clinical claim that personality maps to body region in any causal sense; the body-region-to-facet mapping is not validated by any Tier 1 or Tier 2 framework, and Metryval does not render body diagrams or score-derived somatic visualizations.
The anchor exists for the same reason poets and clinicians use concrete sensory detail: it is the textual property that resists cold-reading projection.
Why this is structural, not stylistic
Stylistic preferences vary across writers, registers, and brand voices.
Measurement-validity guards do not.
The banned list is the latter.
Three reasons compound:
The Forer mechanism. The original 1949 experiment demonstrated that nominally personalized statements composed of universal hedged claims received high self-recognition ratings even when respondents had received the same statements.
The mechanism is robust across replications.
The banned terms are precisely the linguistic tokens that enable it.
Frequency-adverb interpretation drift. Frequency adverbs (“often,” “sometimes,” “occasionally”) have interpretation ranges that span more than an order of magnitude across readers.
A statement calibrated with a frequency adverb is therefore not a measurement claim; it is a Rorschach.
The replacement rule (body location plus texture plus intensity) constrains interpretation to a narrow concrete referent.
Specific-but-wrong beats hypothetical-but-right. The product accepts the cost of specificity.
Some respondents will not recognize the somatic anchor for a given Vectyr because the anchor names a configuration that does not match theirs.
That is the measurement working correctly.
The vague-qualifier alternative would let every respondent recognize themselves in every Vectyr, which is the failure mode being structurally excluded.
How the banned list is enforced
Enforcement runs through the saturation-cell readability standard (v8.5, locked) and the L1 voice linter.
Specifically:
- Nine hard-zero gates in the saturation-cell standard reject a Vectyr-pattern artifact at compile time if any of the banned tokens appears in user-facing copy. “Hard zero” means the gate is not adjustable per Vectyr or per Chapter; a failure blocks the build, not just the render.
- The L1 voice linter pattern-matches against the closed banned-token list and a small set of synonym extensions. A failure surfaces the offending token, the source file, and the line so the author can substitute a compliant phrasing.
- The 5-layer QA stack runs the linter as one of its earliest gates so the rest of the QA cost is not spent on prose that will fail the banned-list check.
The enforcement code is part of the open-source scoring engine repository (link to follow at launch).
Anyone can inspect the linter rules and the gate definitions, and anyone can verify that a given piece of Metryval-authored prose passes them.
Scope limitations
The banned list governs Metryval-authored prose: Vectyr chapter content, Resonogram captions, share-card sentences, Pulse messages, methodology surfaces (including this one), and any other copy where the brand is the speaker.
It does not bind:
- Third-party commentary about Metryval. A journalist or reviewer is welcome to use any qualifier they find apt.
- User-generated reflection. A respondent writing in their own journal or in a share-card customization field uses their own voice; Metryval does not edit it.
- Documentation register, including this surface, when discussing the banned terms as objects of analysis rather than using them as instrument output.
- Internal engineering documentation, audit memos, and the Framework Reception Ledger entries, where precise hedged language is the correct register for documenting validity caveats.
Cross-references
- Anti-Barnum threshold — the score-coordinate-level guard that pairs with the prose-level guard documented here.
- Framework Reception Ledger — per-framework academic-reception status. The methodology citations behind the somatic-anchor rule live in the broader content-rules canon, not in the synthesized framework set.
- Methodology / Transparency — how the linter operates at authoring time, and why runtime rendering does not invoke an LLM.