# Communication Scale

> Locate validated communication measurement scales, assess their reliability and validity evidence, and plan cross-language or cross-cultural adaptation with CFA and measurement invariance. Use for scale selection, psychometric review, questionnaire items, 量表选择/信效度/测量不变性/跨文化改编/量表改编. Do not use for construct definition, which routes to communication-construct.

- Skill: `haochengw372-hash/communication-scale` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add haochengw372-hash/communication-scale`
- Raw SKILL.md: https://api.skillmd.com/api/skills/haochengw372-hash/communication-scale/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: haochengw372-hash (https://skillmd.com/u/haochengw372-hash)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/haochengw372-hash/communication-scale

---


# Communication Scale

## Overview

This Skill runs the Measurement/Scale Gate: it prevents treating a convenient Likert translation as a validated scale and stops cross-group comparisons before invariance is established.

## Locate-first protocol

Before writing items, locate:

1. scale compendiums and measurement reviews in communication;
2. validation papers in Communication Methods and Measures and allied journals;
3. the original scale paper and published appendices with full items;
4. existing validated translations for the target language/culture.

For each candidate record: source journal, validation samples, reliability (coefficient alpha and omega where reported), validity evidence (content, construct, discriminant, criterion), original response format, scoring, and original context/population/language.

## Scale-sheet format

```text
Construct:
Scale name and citation:
Items (verbatim, with original numbering):
Response format and anchors:
Scoring (reverse-coding, averaging/summing, subscales):
Reliability evidence (alpha/omega, samples):
Validity evidence:
Original context (population, language, medium):
Adaptation status (original / translated-unvalidated / validated adaptation):
Decision:
```

## Adaptation protocol

For cross-language or cross-cultural use, do not treat a back-translation as validation. Run, in order:

1. translation by a qualified translator;
2. independent back-translation;
3. expert review of discrepancies (construct equivalence, item equivalence);
4. cognitive pretest with target-population respondents;
5. pilot study for item behavior;
6. confirmatory factor analysis of the intended structure;
7. measurement invariance: configural, metric, and scalar (at minimum metric before comparing relations, scalar before comparing means);
8. reliability estimation on the new sample.

Any item modification (wording, format, short form, platform context) invalidates the original validity evidence and triggers revalidation.

## Decision rules

- Use the existing scale when it fits the construct definition and the target population.
- Use an existing validated translation when one exists and its invariance evidence is documented.
- Adapt only with the full protocol; never present the adaptation as the original scale's validity.
- If no scale fits the construct definition, return to the construct gate: this is a measurement-development project, not a scale-selection task.

## Common anti-patterns

- Reporting alpha as the only psychometric evidence.
- Citing a scale's validity after shortening it or changing response anchors.
- Claiming cross-cultural validity from a single WEIRD validation sample.
- Comparing group means without scalar invariance.
- Writing "adapted from X" while using items the source never validated.

See [references/scale-adaptation.md](references/scale-adaptation.md) for the step-by-step adaptation and invariance checklist.