NEWS

Oct 06 - The MSCEIT is now available in a number of languages (other than English).

Sep 06 - The revised youth version of the MSCEIT is in data collection. I hope that this assessment will spur renewed interest in developing emotional literacy programs in schools by demonstrating the impact these emotion skills have on students - academically and socially.

Aug 05 - MSCEIT-YV

we are in data collection with the youth version of the MSCEIT (ages 10 through 17).

UPDATES

July 8, 2004 - MSCEIT Manual Reference Corrected

We have been alerted by an astute reader of a citation error in the MSCEIT Manual.  Caruso and Wolfe (2001) recently published a list of careers and ratings describing the amount of emotional intelligence that various careers are generally believed to require.”  The reference for this list is Yate, M. (1997). CareerSmarts: Jobs With a Future. New York: Ballantine. Please note that the list is simply the result of a job analysis of various occupations, not an empirically-derived list of occupations and associated MSCEIT scores.

October 29, 2003 - Clarification of MSCEIT Manual

This revision clarifies the data and text reported in Chapter 6 of the MSCEIT Manual

Chapter 6: Reliability of the MSCEIT

Reliabilities were estimated on the basis of that portion of the standardization sample who took the MSCEIT Version 2.0 in its final form (N = 2112).  The first 2888 participants had taken an earlier, research version of the test (the MSCEIT RV1.1), which included additional, now-discarded, items.  This final-form version was regarded as optimal for providing an accurate reliability estimate.  Correlations among tasks and branches (discussed below) are based on the full standardization sample as they were expected to be more robust across forms of the test. 

The MSCEIT Version 2.0 reliabilities are shown in Table 6.1.  Reliabilities for the total, area, and branch scores were computed using split-half analyses as the items were heterogeneous.  Reliabilities for the eight, individual task scores were computed as internal-consistency (alpha) reliabilities.

Using general consensus scoring, the MSCEIT has a full scale reliability of .93, with area reliabilities of .90 (experiential) and .88 (strategic).  Branch score reliabilities range from .79  to .91.  That is, it is a highly reliable test at the Branch, Area, and Total score levels.  Brackett & Mayer (2001) found a test-retest reliability for the full-scale MSCEIT V2.0 of r = .86, over a 3-week period with an N of 60.  Expert scoring was fairly comparable, with a full-scale test reliability of .91 and branch scores ranging from .76 to .90.  These reliabilities were only trivially changed when adding in the excluded 2888 participants (e.g., the general consensus-scored reliability for the total test was somewhat higher, at .93).

The MSCEIT subtasks are less reliable than the branch, area, and total scores, and test interpretation should focus on the Branch, Area, and Total score levels at which the test was designed to be used, except in the case of clear, statistically significant, and dramatic differences in task scores.  That said, at the task level, the alpha coefficients are comparable to those on tests such as the WAIS-R.  (Note that reliabilities for this same data set are reported in Mayer, Salovey, Caruso, & Sitarenios, 2003; several task reliabilities vary by .01 or so due to the use of different computer platforms and software for the two analyses).  

References:

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

Mayer, J. D., Salovey, P., Caruso, D. R., & Sitarenios, G. (2003).  Measuring emotional intelligence with the MSCEIT V2.0.  Emotion, 3, 97-105. 

July 3 2003 - Confirmatory Factor Analyses

If you are testing a four-factor CFA solution, you must set one additional parameter constraint in model estimations: We held the within-area covariances, e.g., those between Perceiving and Using, and between Understanding and Managing to be equal (but allowed them to take on any value).  Such an arrangement should take care of the apparent tendency of LISREL, AMOS, and other programs to set the covariance between Perceiving and Using at a level near 1.0.  Adding that modest constraint to the parameters leads to a solution that better reflects the r = .40 to .50 correlation that is typically found in data sets.

June 22 2003 - Bias Scores

The commercial version of the MSCEIT yields a bias score, as does the research scoring.  MHS found that the original release of the research scoring reversed these bias scores.  This was corrected as of July 1, 2002.  If you have an older data set, and you wish to use these scores, ask MHS to re-score your data set. 

June 20 2003 - Updated Norms

As with traditional IQ tests, updating the MSCEIT norms leads to over- or under-estimates of emotional intelligence.  If you are comparing your data to MSCEIT norms, you can either adjust your scores upward or get the data set re-scored by MHS.  We recommend that you not use MSCEIT standard (IQ-type) scores in your research work so that you can compare data sets over time. 

June 9 2003 - CRITICAL INFO ON COMPUTING MSCEIT RELIABILITY

Several researchers have reported obtaining lower reliabilities for MSCEIT ability and task scores than expected.  We have found a few procedural problems that have resulted in these findings.  First, as we note, while task reliabilities are computed with coefficient alpha, the four ability scores and total score are computed using split-half reliabilities, not alphas, as there is a great deal of item heterogeneity.  Second, some researchers have not used the Spearman-Brown correction.  Third, if you use SPSS and other stat packages, these procedures take the first half of items and compare them against the second half.  This is not going to work for the MSCEIT, since there are several different item types.  We used an odd-even split half method -- which may require you to compute split-halves without use of the SPSS Reliability function. 

How To Compute MSCEIT Reliability

Use the method below to compute reliability for Total, Area (Experiential and Strategic), and Branch (Perceiving, Using, Understanding, Managing) level scores

Example for Perceiving Emotions

Step 1: Make a composite (additive) score of all the odd items for Faces and Pictures together
Step 2: Make a composite (additive) score of all the even items for Faces and Pictures together
Step 3.  Intercorrelate the two composite scores.  This represents the reliability of a scale of half the length as the actual scale.
Step 4.  Apply the Spearman Brown correction to obtain the estimated reliability. The Spearman Brown correction states that: The reliability for the scale under study = two times the reliability of the half scale (which is represented by the intercorrelation of the two composite scales, as calculated above), divided by 1 plus the reliability of the half scale.

June 27, 2002 - MSCEIT MANUAL REVISION

Tables listing the means and SD's for the Sensations and Facilitation tasks for MSCEIT V 2.0 are reversed!
 

May 22 2002 - MSCEIT MANUAL REVISION

Pages 35 and 36 of the manual make reference to a "MSCEIT RV1.0."  That should have read MSCEIT RV1.1, as in the rest of the manual.

For those interested, the MSCEIT RV1.0 was a preliminary version of the MSCEIT RV1.1.  The difference?  One face was changed for the MSCEIT RV1.1.  Only about 30 participants ever took the MSCEIT RV1.0.  The mention in the manual is not intended to be to the RV1.0; it is just a copyediting problem.