Psychology 121, Lecture 14

Evolution of Objective Testing

by Hal S. Kopeikin, Ph.D. © 2000


 
 

Two Traditions in Measurement of Typical Performance

1. Structured or Objective Tests.

  • Grew from the psychometric tradition prevalent in American universities
  • Clearly defined structure (standardized content, administration, scoring)
  • Scientific, traditional concerns for reliability and validity
  • Emphasis on precise, efficient focus on defined variables (IQ, diagnosis, aptitude, etc.)
    •

    2. Projective Tests

  • Emerged from European clinical contexts
  • Descendants of Gestalt perceptual psychology and Psychoanalytic traditions
  • Vague stimuli evoke wide range of possible responses
  • Tests are seen as one more way of collecting information for synthesis, integration, analysis
  • Emphasis is on obtaining rich, unrestricted sample of mental processes
    •

    The Evolution of Objective Tests

    The earliest typical performance tests were logically keyed, with items selected for face validity.

    Assumptions behind this approach included

    1. People have an accurate view of themselves and can convey that view
    2. People will portray themselves honestly, objectively, and accurately
    3. People will interpret items consistently, i.e. each item will mean the same thing to everyone
  • By 1940s each assumption had been challenged and objective testing was viewed skeptically
    •
  • Empirical (criterion-related) keying and factor analysis were proposed to avoid these weaknesses.
  • As the drawbacks of these approaches also became evident, test makers integrated the three approaches to compensate for the weaknesses of each.
    •
  • The development of the SII and MMPI both reflect the shift to empirical keying, and combining that method with theoretical and logical keying.
    •  For most of class today we will discuss the MMPI and SII as examples of the evolution in objective testing. Other tests will be mentioned briefly, as time permits. See the text for summaries of the Mooney Problem Checklists, CPI, 16PF, EPPS, JPI, Myers-Briggs, MCMI-II, and NEO Personality Inventory.

    Minnesota Multiphasic Personality Inventory (MMPI) & MMPI-2

    Composed of

    1. Eight (8) empirically developed clinical scales contrasting "normals" vs. Hypocondriacs, Depressives, Hysterics, Psychopathic Deviates, Paranoids, Psychasthenics, Schizophrenics, & Hypomanics. Table 15-1, p 415 for more details. Two other scales, Masculine-Feminine, and Introversion-Extroversion was included for information about personal style that was deemed important, albeit not diagnostic per se.
    2. Three validity scales to detect possible distortions in test-taking/self presentation. These were the L (lie), F (infrequent response), and K scales. These are described on page 416 (cf. Table 15-2). There was also a ? scale, which simply told how many items were left blank. On the MMPI-2, two major new validity scales were added VRIN (variable response inconsistency true), and TRIN (true response inconsistency true), as described on page 420.

    The Evolution of MMPI Interpretation

  • People who scored highly on one scale often scored highly on others. This meant the scales were intercorrelated. In fact, many items are scored on more than one scale. Consequently, the MMPI did well discriminating between "normal" and "mentally ill" test-takers, but not nearly as well diagnosing the latter. Thus, the original purpose of the MMPI, and approach to interpretation, basically failed. Alternative approaches appeared.
    •
  • Interpretation by scale elevation was more successful, but potentially misleading (see Table 15-3, p. 418). Because the scales were quite heterogeneous, it was risky to describe the nature of people with particular scale elevations.
  • One solution was to use combinations of scales ("code types"). Interpretation was based on patterns of scale elevations rather than scales in isolation. For example, elevations of the 2,7,and 9 meant the person was likely to have an agitated, anxious depression rather than a flat, withdrawn type.
  • Another, complimentary, and often better approach was factor-analytic subscales. Scales were divided into more homogeneous subscales, and the absolute elevations on those was more interpretable.
    •

    The Validity Scales were a major hit

    While no test is unfakable, the MMPI made tremendous progress in detecting factors which might distort responses. People still can "fake good" or "fake bad," but extreme efforts are typically identifiable and scores are adjusted for smaller distortions. Two kinds to distortions psychologists worry about are
  • Response Sets : Attempts at impression management (fake good, fake bad, social desirability, etc.) The effect of a response set depends on the apparent meaning of the item.
  • Response Styles: General response tendencies independent of item content (acquiesence, extremity, yeah saying/naysaying, random responding, etc.)
    •