Reliability The reliability of a type of measurement indicates how free that measurement is
from random error.4 A reliable measurement therefore generates consistent results.
Assuming that a person’s intelligence is fairly stable over time, a reliable test of
intelligence should generate consistent results if the same person takes the test several
times. Organizations that construct intelligence tests should be able to provide (and
explain) information about the reliability of their tests. Usually, this information involves
statistics such as correlation coeffi cients. These statistics measure the degree to which two
sets of numbers are related. A higher correlation coeffi cient signifi es a stronger
relationship. At one extreme, a correlation coeffi cient of 1.0 means a perfect positive
relationship—as one set of numbers goes up, so does the other. If you took the same vision
test three days in a row, those scores would probably have nearly a perfect correlation. At
the other extreme, a correlation of 21.0 means a perfect negative correlation—when one
set of numbers goes up, the other goes down. In the middle, a correlation of 0 means there
is no correlation at all. For example, the correlation (or relationship) between weather and
intelligence would be at or near 0. A reliable test would be one for which scores by the
same person (or people with similar attributes) have a correlation close to 1.0. Reliability
answers one important question—whether you are measuring something accurately—but
ignores another question that is as important: Are you measuring something that matters?
Think about how this applies at companies that try to identify workers who will fi t in well
