Observer Effect

The term observer effect generally refers to the possibility that an act of observation may affect the properties of what is observed. However, depending on the context and the mechanisms involved, it may indicate effects of a very different nature. Observer effects are a threat to validity in much of educational research. After providing examples of observer effects to illustrate its meaning, this entry examines how to limit observer effect. The entry also considers common misconceptions about observer effect.

Imagine measuring the pressure of an automobile tire. When connecting the manometer, it is easy to let out some air, so that the measured pressure may not correspond to the pressure of the tire before the act of the measurement was initiated. Similarly, imagine that we want to know the temperature of a liquid and that to do so we use a thermometer. Because the latter has its own temperature, when it is immersed in the liquid, it can change the liquid’s temperature, so again the observed value may not correspond to the temperature of the liquid before the measurement.

Effects of this kind occur in many domains of physics and are usually named probe effects. They also occur in other domains, like for the so-called heisenbug in computer programming (the term is a pun on the name of physicist Werner Heisenberg), denoting a software bug that can alter its behavior, or even disappear, when one attempts to probe it. In social science, the term observer’s paradox (coined by the American linguist William Labov) is used to refer to situations in which the presence of the observer can alter the results of the observation. For instance, in sociolinguistics, when a researcher attempts to gather data on natural speech, the researcher may alter the way of speaking of the interviewed (which may become more formal) by the researcher’s mere presence. Other designations denote variants of the observer’s paradox, like the Hawthorne effect or the experimental demand effect, always describing situations in which the behavior of persons may be altered in ways that are usually not intended by the experimenters, just because they are monitored or placed in a specific experimental context.