15.14
Cause et effet
While variables are sometimes correlated because one does cause the other, it could also be that some other factor, a confounding variable, is actuall…
Dans certains cas, la relation entre deux articles semble claire et intuitive, comme lorsqu’il y a un match de sport, les gens commandent une pizza même s’il se fait tard. Un membre du groupe jure que manger plusieurs tranches avant de se retirer pour la soirée lui donne des cauchemars.
Maintenant, la seule façon de répondre à cette question « Est-ce que manger de la pizza juste avant d’aller dormir fait faire plus de cauchemars à quelqu’un ? » est de concevoir une expérience.
Dans un type de plan expérimental, une relation de cause à effet, un chercheur peut déterminer si la manipulation d’une variable indépendante – dans ce cas, manger une pizza avant de se coucher – provoque un effet particulier – des changements dans la variable dépendante, le nombre de cauchemars qui se produisent tout au long de la nuit.
Ils pourraient affecter la moitié des participants au groupe expérimental, qui reçoit la manipulation expérimentale – l’attribution de manger trois tranches de pizza juste avant d’aller au lit – et la seconde moitié au groupe témoin, qui reçoit l’ordre de ne rien manger.
Ils peuvent également prendre certaines mesures pour contrôler les variables confusionnelles qui peuvent produire d’autres explications.
Par exemple, en répartissant au hasard les participants à différents groupes – à l’aide d’une méthode basée sur les probabilités – le chercheur peut s’assurer que les participants sont également appariés sur les facteurs de confusion potentiels, par exemple, leurs antécédents et leur prédisposition aux cauchemars, le temps qu’il leur faut pour s’endormir et même leur qualité globale de sommeil.
De plus, le chercheur pourrait mener l’expérience dans un laboratoire où encore plus de facteurs peuvent être contrôlés, comme l’environnement de sommeil et ce qu’ils regardent avant de dormir.
En fin de compte, la seule façon d’établir une causalité entre deux variables est de faire des expériences sonores !
View the full transcript and gain access to JoVE Core videos
Q1: Why can't researchers establish causality from correlational data alone?
Correlational research identifies relationships between variables but cannot prove one causes the other. A confounding variable—a third factor—may actually explain the relationship. For example, ice cream sales and crime both increase in warm weather, but temperature causes both, not ice cream causing crime. Only experiments that manipulate variables and control for alternative explanations can establish true causality.
Q2: How do experimental designs establish cause and effect relationships?
Experimental designs establish cause and effect by manipulating an independent variable and measuring its effect on a dependent variable while controlling confounds. Researchers assign participants to experimental and control groups, then use random assignment to ensure equal matching on potential confounds. Laboratory settings allow researchers to control additional factors like environment and timing, enabling sound conclusions about causality.
Q3: What is a confounding variable and why does it matter in research?
A confounding variable is an unmeasured third factor that causes systematic changes in variables of interest, creating false correlations. For instance, both cereal consumption and healthy weight may correlate with overall health consciousness rather than cereal causing weight loss. Confounding variables produce alternative explanations for observed relationships, which is why controlling for them through random assignment and laboratory conditions is essential for valid causal claims.
Q4: What are illusory correlations and how do they affect our thinking?
Illusory correlations are false beliefs that relationships exist between two things when no actual relationship exists. For example, many people believe the full moon affects human behavior, but meta-analyses show no such relationship. Illusory correlations arise from confirmation bias—seeking evidence supporting a hunch while ignoring contradictory evidence—and can lead to prejudicial attitudes and discriminatory behavior toward certain groups.
Q5: How does random assignment help control for confounding variables?
Random assignment uses probability-based methods to distribute participants equally across experimental and control groups. This ensures participants are matched on potential confounds like predisposition for nightmares, sleep latency, and sleep quality. By randomly assigning rather than allowing self-selection, researchers reduce systematic bias and increase confidence that observed effects result from the independent variable manipulation, not pre-existing differences.
Q6: Why do people mistakenly claim causation from correlational findings?
People often mistake correlation for causation because the relationship seems intuitive or because advertisements and news stories make causal claims based on correlational data. Confirmation bias leads people to accept information supporting their hunches without scrutiny. Additionally, information that comes easily to mind feels more reliable, even if limited. These cognitive shortcuts are especially problematic in media, where misleading causal claims influence public perception and behavior.
Q7: What role does laboratory control play in establishing causality?
Laboratory settings allow researchers to control numerous environmental factors beyond the independent variable, such as sleep environment, pre-sleep activities, and timing conditions. This controlled environment eliminates alternative explanations for observed effects on the dependent variable. By isolating the experimental manipulation from confounding factors, laboratory experiments provide stronger evidence for cause-and-effect relationships than field studies where many variables remain uncontrolled.