If you are covering Chapter 5 or Chapter 6, you may want to assign the following article:

Brackett, M.A., & Mayer, J. D. (2003). Convergent, discriminant, and incremental validity of competing measures of emotional intelligence. Personality and Social Psychological Bulletin, 29, 1147-1158.


This article addresses an interesting topic (emotional intelligence),  is easy for your students to obtain (students who buy the book can get it by using the Infotrac® subscription that comes with ResearchDesign Explained, and is relatively easy for students to read (to make it easier, give students Table 1).



Table 1

Helping Students Understand the Article


 Tips, Comments, and Problem Areas


Incremental validity: extent to which measure helps us go beyond what we know using other measures; degree to which the measure, when combined with other measures, allows us to make more accurate predictions than if we just had the other measures

Discriminable: different from

Shared considerable variance: related; seem to be measuring the same things; affected by the same things

Inversely related: negatively correlated ; oppositely related; if a person’s score on one variable is high (e.g., scores on self-report test of emotional intelligence), that person’s score on the other variable (e.g., grades) well tend to be low.

Introduction: Beginning

2nd paragraph

unmoored: disconnected; unattached; no longer tied to

3rd paragraph

different representations of the same person”: are measuring different characteristics

4th paragraph

A key issue in measuring emotional intelligence is to define it.

Introduction: Background

Note that popular claims sometimes do not match the facts.

Introduction: MSCEIT

Note how their definition of emotional intelligence dictates that their test should have four subscales.

Factor structure: Factor analysis supports the idea that the test is measuring four different abilities (for more about factor analysis, see page 212 and pages 535-536 of Research design explained).

while ….unique variance”: the test is also measuring something that traditional intelligence tests do not measure; it is not just another traditional intelligence test.

Introduction: Eq-i

Intrapersonal EQ: knowing yourself

Unidimensional: one characteristic , component, or aspect  (a single ability)

Multidimensional: more than one characteristic; multiple components; various aspects  (multiple abilities)

Introduction: SREIT

Psychometric properties: characteristics such as reliability, validity, and factor structure that relate to a measure’s ability to measure a psychological construct.

Provisional: temporary, tentative, not  definitively established

Introduction: Comparative performance of the MSCEIT, EQ-I, and SREIT.

1st paragraph

Note that low correlations among the measures do not provide evidence of convergent validity (the measures are not all measuring the same thing; a participants’ scores on the different measures do not strongly converge on a certain emotional intelligence score or ranking). Instead, the low correlations suggest that the measures are measuring different constructs.

2nd paragraph

The paragraph presents evidence for the convergent validity of the MSCEIT

3rd and 4th paragraphs

   Each paragraph does two things: (a) presents convergent validity for the measure showing that it correlates with what it should correlate with, but (b) presents evidence suggesting that the measure may lack discriminant validity relative to several personality measures.

5th (last) paragraph

Authors more clearly state the case that EQ-i and SREIT lack discriminant validity.

Introduction: Introduction to the present study

1st paragraph

mostly independent of existing personality constructs”: will not correlate strongly with measures of other constructs; therefore, will have discriminant validity relative to other personality measures.

share considerable variance with”: will correlate strongly with measures of other constructs; therefore, will not have discriminant validity relative to other personality measures.



...exclusion of extreme outliers…” Scores were not analyzed for participants who scored extremely low (e.g., lower than 99% of the participants) or who scored extremely high (e.g., higher than 99% of the participants) .

Method: Measures of emotional intelligence

a : an indicator of a measure’s internal consistency that can range from 0 (not at all consistent; answers to one question do not correspond to answers on the next question) to 1 (answers on one question correspond perfectly with answers to the other questions). Usually, scientists expect tests to have as above .80. To learn more about as (called Cohen’s alpha), see page 104 of Research design explained.

composite of six primary (facet) scales”: scores for each trait were computed by combining the answers to six questions that made up the scale for that trait.

Method: Criterion measures

1st paragraph:

External life space criteria measures do not ask participants to report on what they feel but instead those measures ask participants to report what they did.

2nd paragraph

proxy:  substitute


Preliminary analyses

1st paragraph

The evidence suggests that the SREIT measures one factor, as its authors claimed.

2nd paragraph

They gave each participant two scores on the EQ-I: (a) one score using the revised scoring system and (b) one score using the original scoring system. The correlation between the two sets is about as close to 1 as you will find in psychology, indicating that the scoring system does not affect how participants will score.


Relations among measures

1st paragraph

zero-order correlations: simple, straightforward correlations; what you think of as correlations between pairs of variables; Pearson rs.

The authors make the case that the MSCEIT has discriminant validity, but that the other scales do not have discriminant validity.

3rd paragraph

partial correlations: correlations between two variables that factor (subtract) out associations between those two factors that appear to be due to other measured variable(s).  In this case, the zero-order correlation for MSCEIT and EQ-i was .21, but the partial correlation between those same variables when controlling for scores on psychological well-being was only .14. Similarly, the zero-order correlation for SREIT and Eq-I was .43, but the partial correlation between those two variables ( controlling for psychological well-being)  was only.09.

mostly independent: was only slightly correlated with

Possibly, EQ-i and SREIT produce similar scores because both are measuring things measured by the Psychological Well-Being scale(PWB); the MSCEIT and EQ-i may be correlated because both are measuring things measured by the Big 5 or PWB.

Results: Discriminant validity using multivariate statistics

Scores on the Big 5, as well as subscale scores on the PWB scale, could be used to predict scores on all three emotional intelligence tests. However, the Big 5 and the PWB did much better at predicting scores on the Eq-i and the SREIT than they did at predicting scores on the MSCEIT. Thus, MSCEIT seems to be measuring ability rather than personality.

Oblique rotation: a factor analysis where it is assumed that factors may correlate with each other

Eigenvalues: how much of the variability in scores  is due to the factor. The bigger the eigenvalue, the more important the factor. Eigenvalues near 1.0 indicate that the factor is not important.

Factor loadings: Like correlations between actual scores and the hypothetical factor.

Results: Predictive and incremental validity

Note that the SREIT did not predict any behaviors that you would expect to be correlated with emotional intelligence.  Note that there was little evidence of relationship between scores on the other emotional intelligence measures and behavior when simple (zero-order) correlations between the other EI measures and behavior were calculated, and there was even less evidence of a relationship when partial correlations controlling for the Big 5 and verbal SAT scores were calculated.


The article’s authors make a point made in Research design explained: you should know exactly what you want to measure before you choose a measure.

Emulate: imitate



To Chapter 5 Menu

To Chapter 6 Menu

Back to Bonus Article Menu

Back to Research Design Explained Home Page