Skip to main content
Social Sci LibreTexts

5.1: Naturalistic Designs and Causal Inferences

  • Page ID
    10344
  • \( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \) \( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)\(\newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\) \( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\) \( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\) \( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\) \( \newcommand{\Span}{\mathrm{span}}\) \(\newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\) \( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\) \( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\) \( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\) \( \newcommand{\Span}{\mathrm{span}}\)\(\newcommand{\AA}{\unicode[.8,0]{x212B}}\)

    So we had a good visit with laboratory settings and experimental designs, and although we will meet up with them again at the end of this section (when we arrive at “converging operations”), for now let’s just focus on their limitations in providing detailed process-oriented causal accounts involving conditions we can’t randomly assign and potential causal factors we can’t manipulate. In this chapter, we consider the potential utility of field settings and alternative naturalistic designs. What we appreciate about naturalistic field studies is obvious already from the limitations we encountered with lab and experimental designs: In naturalistic studies we can examine the effects of potential causes that we can’t manipulate directly (e.g., maltreatment, ability tracking, peer rejection). Moreover, we can watch these processes operate in their actual multi-level contexts, and we can follow them for months and years.

    But after we sow our wild oats, freely roaming this interesting space and documenting some of the fascinating pathways our participants are taking, we will want to turn from description to explanation. And as soon as we do, we will start to feel homesick for our familiar labs and experiments. In the field, the causes we want to analyze come in annoying piled up clumps of conditions, whereas in our experiments, we had the capacity to create treatment and control groups that we exposed to carefully distinguished and calibrated causal strands. And in the field, these causal clumps land on particular people, sometimes in a “rich get richer” or other complex pattern, whereas in our experiments, they were randomly assigned to groups of people who are the same on everything else. So in naturalistic studies, any causal exposure is always confounded with participants’ pre-existing differences.

    One broad way to characterize the difference between lab experiments and naturalistic studies in the field, is that in lab experiments researchers have front row seats and very clear lines of sight on a phenomenon, but we can never be positive if it is exactly the one that we are interested in observing; whereas in naturalistic field studies, researchers can be certain that the actual phenomenon we want to understand is happening right in front of us, but it is all happening so fast and in so many directions (and maybe we are in the balcony and seated behind a pole) that we can never be sure of exactly what we are seeing. In other words, to get causal information out of naturalistic field studies takes quite a bit of work—conceptual, inferential, and empirical.

    Wait a minute. Aren’t experiments the only way to show causality?

    Yes, experiments can provide important evidence of causal processes. But let’s consider the kinds of causal evidence that can be provided by naturalistic studies.

    Wait another minute. Are we talking about correlational studies? Because we know for a fact that “correlation does not prove causation.”

    Right, it is correct that correlation by itself does not prove causation. But let’s take a minute to understand why this is true, and then to see whether there are some things that researchers can do to improve the designs of their studies so that naturalistic studies, using more than correlations, can provide evidence about causes. Because, remember—correlation may not prove causation, but causal processes do generate correlations- causes create effects and so effects covary with their causes. In fact, this covariation is one of the defining condition of causality (see box). As a result, correlations (or covariation or contingencies, however you want to label them) may be the smoke that leads us to our causal fires. The problem is that many things besides causation lead to correlations, and so we have to work hard to decipher the causal evidence among all the other kinds of covariation information we are examining.

    John Stuart Mill (1843) on causality. To establish causality, three basic conditions must be met: 1. The presumed effect(s) must covary with their presumed cause(s); 2. The presumed cause(s) must precede their effects in time; and 3. All other plausible alternative explanations for the effect must be excluded.