Distribution of Total Error Scores
The distribution of initial error scores which may be anticipated from particular populations is indicated in the following table. The unselected column shows the errors which may be expected from each percentile of an unselected group of people aged 15 to 45. The next column, 'in-plant' applicants, was drawn from 300 tests of employed personnel applying for promotion to color control positions (shader trainees, laboratory testers and laboratory technicians) in a paint manufacturing plant. The last column, experienced, is accumulated from tests on 150 personnel with three to twenty years experience in the color control laboratories of manufacturers of dyes, rugs, plastics, textiles and paints. Their job titles included shaders, matchers, mixers, inspectors, passers, dyers and titration testers.
| Percentage of group | Expected to make error scores of, or less than | ||
| Unselected | In-plant applicants | Experienced | |
| 90 pc, upper 10% | 12 | 4 | 0 |
| 80 pc, upper 20% | 20 | 8-12 | 0 |
| 70 pc, upper 30% | 28 | 16 | 4 |
| 60 pc, upper 40% | 36 | 20 | 4-8 |
| 50 pc, upper 50% | 46 | 24 | 8 |
| 40 pc, upper 60% | 60 | 28 | 12 |
| 30 pc, upper 70% | 76 | 36 | 16 |
| 20 pc, upper 80% | 96 | 44 | 24 |
| 10 pc, upper 90% | 120 | 68 | 40 |
Distribution of Total Error Scores by percentiles: First tests
These figures form the basis of the percentile scores given by the computer program.
Test-Retest Reliability
The test-retest reliabilities shown below are taken from routine tests of applicants for promotion in a paint factory. In most cases, the retests were taken within a few days of the first. An average reduction of 30% in total error scores can be seen between the first test and retest. Little average improvement occurs on the third test.
| Test | Retest | 2nd Retest | |
| Test | .82 | .67 | |
| Retest | .83 | ||
| Mean | 32.02 | 22.68 | 20.55 |
| Standard Deviation | 29.02 | 23.08 | 23.44 |
| N | 196 | 196 | 137 |
Test-Retest Reliability
These figures suggest that the second trial yields the more important score. Some of the test-retest unreliability is doubtless due to variations in experience in handling this kind of material and would be eliminated by discarding the first score. However, if it is planned to use norms based on the second trial alone, the following conditions must be met:
·every individual must be given a second test.
·the first test must be taken as seriously as the second, and not announced as practice.
·test and retest should be separated by a period of hours or days.
Usefulness of Total Error Scores
Total error scores represent the color discrimination ability of an individual at a given point in time. It is possible that color discrimination ability for that individual may improve at a later time as a result of further training and experience.