Last edited:
[...]
- Thread starter fa21212
- Start date
- Joined
- Apr 29, 2014
- Messages
- 606
- Reaction score
- 263
Hi @fa21212 -
Great question! In a nutshell, construct validity has more to do with whether items make sense as measures of what they're supposed to measure, and internal consistency has more to with how well items cluster together in terms of having similar results.
An interesting example of this has to do with early attempts to develop intelligence tests (the Wikipedia page on construct validity mentions one of these examples that is discussed at length in Stephen Jay Gould's book The Mismeasure of Man, which is parenthetically a fascinating read). In intelligence tests developed in the early 20th century, they would ask questions like "what city are the Dodgers based out of?" Of course, the problem with this is that knowing where a given sports team play has to do with your cultural awareness of that sport, not with your innate intelligence, so that question would penalize recent immigrants unfamiliar with the culture of US sports. Imagine writing an "intelligence test" with multiple similar questions: "What city are the Dodgers based out of?" "Where do the Mets play?" "Where do the White Sox play?" etc. Responses to those questions would probably correlate pretty closely with each other, because if you know the answer to one question, you're likely to know the rest, and vice versa. So these items would have pretty good internal consistency, but they would have awful content validity, because sports trivia knowledge doesn't measure intelligence.
Hope this clarifies things!!
Great question! In a nutshell, construct validity has more to do with whether items make sense as measures of what they're supposed to measure, and internal consistency has more to with how well items cluster together in terms of having similar results.
An interesting example of this has to do with early attempts to develop intelligence tests (the Wikipedia page on construct validity mentions one of these examples that is discussed at length in Stephen Jay Gould's book The Mismeasure of Man, which is parenthetically a fascinating read). In intelligence tests developed in the early 20th century, they would ask questions like "what city are the Dodgers based out of?" Of course, the problem with this is that knowing where a given sports team play has to do with your cultural awareness of that sport, not with your innate intelligence, so that question would penalize recent immigrants unfamiliar with the culture of US sports. Imagine writing an "intelligence test" with multiple similar questions: "What city are the Dodgers based out of?" "Where do the Mets play?" "Where do the White Sox play?" etc. Responses to those questions would probably correlate pretty closely with each other, because if you know the answer to one question, you're likely to know the rest, and vice versa. So these items would have pretty good internal consistency, but they would have awful content validity, because sports trivia knowledge doesn't measure intelligence.
Hope this clarifies things!!