chap 11

maturation

People naturally change over time, independent of any treatment. Spontaneous improvement (or decline) is an alternative explanation for pre-to-post change.

Example: Boys at a summer camp became calmer after a low-sugar diet was introduced. But boys typically 'settle in' to a new environment over time regardless of diet. Without a comparison group of equally rowdy campers who kept eating sugar, you cannot separate the two explanations.

history threat

An external event happens around the same time as the treatment and affects everyone in the group.

Example: A dorm reduced electricity usage during a 'Go Green' campaign. But the season also changed from summer to fall. Did the campaign cause the reduction, or did cooler temperatures? A comparison dorm with no campaign would clarify.

regression to the mean

When a group is selected because of an extreme score, random factors that caused that extreme score won't repeat exactly, so scores naturally drift toward the mean.

Example: Depressed women who seek therapy do so when they're feeling especially bad. Some of their extreme scores were partly caused by temporary factors (a bad week, a recent loss). At posttest, those factors may not be as severe, so scores improve even without effective therapy.

attrition

People drop out of studies. When the people who leave are systematically different from those who stay, the final sample is biased.

Example: If the three most severely depressed women drop out of a therapy study, the posttest average automatically drops — not because therapy worked but because the highest scores are simply gone

testing threat

Taking a test changes participants. A second testing improves scores through practice, or worsens them through fatigue.

Example: Students score higher on a reading test the second time simply because they've seen the format before, not because of any educational program.

instrumentation

The measuring tool itself changes between pretest and posttest, making scores incomparable.

Example: Coders observing children's aggression become more lenient over weeks of coding. What looks like a decrease in aggression at posttest is really just a looser standard of measurement.

ceiling or floor effects

All groups score at the extreme high end (ceiling) or low end (floor) of the DV, leaving no room for between-group differences.

Example: Asking 5-year-olds to identify the first letter of their name is so easy that everyone gets it right. The reading test has a ceiling effect — all groups score 100% and any real differences disappear.

error variability

the spread or dispersion of unexplained deviations (residuals) around a true, expected, or predicted value. It measures the uncertainty and random noise that remains after accounting for known variables