Reading focus
Syllabus: How Do We Know What a Psychological Test Measures?
Keep reliability, item structure, and construct interpretation separate when working through the examples.
Why you’re reading it. Recognize why a single coefficient cannot carry the whole measurement argument.
A question to consider. The Persistence Scale reaches alpha = .90 after several similar items are added. What can you conclude, and what remains unknown?
Why this reading
A familiar coefficient can acquire more authority than its assumptions justify. Sijtsma separates reliability from dimensional structure and demonstrates why alpha alone cannot tell us how many attributes a set of items reflects.
Before you read
In classical test theory, an observed score combines a true-score component and error. The true score is an expected score under specified repetitions, not automatically the psychological attribute itself. A covariance describes how two variables vary together.
Read for
- Distinguish alpha, reliability, and unidimensionality throughout the examples.
- Use the contrasting covariance structures to ask what information the same alpha can conceal.
- Notice the limits of estimating repeatability from one administration.
Consider afterwards
- Two scales have identical alpha values. What further evidence would you request before claiming that either measures one coherent attribute?
Editorial context
Read the title as an invitation to examine assumptions, not an instruction to reject every use of alpha. Sijtsma and Pfadt’s later discussion defends the usefulness of lower bounds and revisits correlated errors. Choosing another coefficient does not remove the need for a measurement model.