Psychometric methodology · Touzani (2000) · Churchill (1979)

Validity and Reliability of Measurement Scales

Psychometric concept

Any serious psychometric assessment rests on two properties: reliability (does the instrument measure consistently?) and validity (does it measure what it claims to?). The validation process follows Churchill's paradigm: construct definition, item generation, purification via factor analysis, internal consistency testing (Cronbach), convergent and discriminant validation. This page summarizes those requirements in accessible language.

Key dimensions

01

Reliability

Consistency and stability of measurement over time and across items.

02

Validity

Alignment between the theoretical construct and its empirical measurement.

Model categories

Cronbach's alpha

Internal consistency index. Acceptable threshold: ≥ 0.70; preferred: ≥ 0.80.

Content validity

Do the items cover all facets of the construct?

Construct validity

Confirmatory factor analysis: theoretical structure verified empirically.

Convergent validity

Expected correlations with related measures.

Discriminant validity

Independence from unrelated measures.

Predictive validity

Ability to predict a future criterion (e.g., job performance).

Key takeaways

  • A non-reliable test cannot be valid — reliability is necessary but not sufficient.
  • Predictive validity is the most critical property for a selection test.
  • A high alpha but an incoherent factor structure often signals a poorly defined construct.
  • Any serious test publishes its reliability coefficients and validation studies.

Apply this framework to your teams

TalentAssessment builds custom assessments rooted in the most established psychometric models.

Try for free→