Behavioural Research Methods

Published in 1969, the seminal article “Design and Analysis of Research Using Time Series” by John M. Gottman, Richard McFall, and Jean Barnett remains a foundational contribution to behavioral science methodology. Amidst a research culture increasingly dominated by randomized controlled trials (RCTs), this work offers an alternative vision—one that values methodological flexibility, ecological sensitivity, and the analytical richness of temporally structured data. For researchers working in real-world contexts—psychotherapy sessions, classrooms, community interventions—this paper offers a design philosophy that is as relevant today as it was at the time of its publication.

At the heart of the authors’ argument is a call to move beyond static experimental frameworks. Traditional pretest-posttest designs, while useful for capturing whether change occurred, are limited in their ability to capture the how and when of change. Control groups, often considered the gold standard for causal inference, may not always be ethically or practically viable, especially in clinical or educational environments. The solution offered is a triangulated approach to design centered on time-series methodology. This framework combines one-group pretest-posttest designs, time-series designs, and multiple time-series comparisons to form a network of mutually reinforcing methods that mitigate different threats to internal validity.

Time-series designs, the authors argue, do far more than track change. They map the trajectory of intervention effects, expose unexpected fluctuations, and offer an empirical basis for both theory generation and causal inference. Whether the intervention is a discrete event or a prolonged program, the time-series record serves not only as a log of outcomes but as a narrative of process. The ability to embed measurement within natural settings also aligns with what the authors call an “ecological orientation,” where data are derived from, and relevant to, real-life behavior.

Critically, Gottman and colleagues challenge the prevailing skepticism around time-series analysis. They acknowledge earlier statistical hesitations, especially around autocorrelation and noise, but counter these with powerful alternatives to traditional curve-fitting techniques. The use of generating functions, such as autoregressive models and exponentially weighted moving averages (EWMA), allows researchers to model dependency, identify structure within data, and meaningfully partition signal from noise. In contrast to statistical approaches that treat temporal dependency as a nuisance, these methods position it as a source of insight.

The implications of this approach are profound, particularly for single-subject research. Rather than dismissing n=1 designs as anecdotal, the paper demonstrates how time-series methods can elevate individual trajectories to the status of generalizable scientific knowledge. Whether in therapy outcome research, behavior tracking, or intervention testing, the ability to model time-dependent processes opens new avenues for discovering patterns, testing causal structures, and predicting future behavior.

The authors also offer practical solutions for research constraints. In contexts where randomized group allocation is unfeasible or unethical, the time-lagged multiple time-series design enables robust comparisons while maintaining methodological integrity. Ethical concerns around untreated control groups are addressed by staggering intervention timing, allowing every participant to benefit while still providing a valid basis for analysis.

Even more striking is the authors’ foresight regarding feedback-based and adaptive research. They describe the time-series design as inherently dynamic, capable of incorporating ongoing results into future stages of inquiry. This positions research as a living system, responsive to change, sensitive to context, and open to iterative refinement. In this sense, their model anticipates contemporary interests in real-time data analytics, precision interventions, and personalized research designs.

In an age when digital platforms collect intensive longitudinal data from individuals, from mobile health apps to wearable sensors, the logic of this paper becomes newly relevant. The infrastructure for time-series research is now widely available; what remains is the will to use it thoughtfully. Behavioral researchers today can learn from Gottman, McFall, and Barnett not only how to analyze data more rigorously, but how to design studies that respect the complexities of human behavior as it unfolds over time.

In conclusion, Design and Analysis of Research Using Time Series stands as a classic text that deserves renewed attention. It offers a methodological framework that combines the strengths of experimental control with the realism of naturalistic observation. By integrating robust design logic with innovative statistical modeling, the authors provide a blueprint for research that is not only scientifically credible but also deeply attuned to the rhythms of real-life change.

Reference:

Gottman, J. M., McFall, R. M., & Barnett, J. T. (1969). Design and analysis of research using time series. Psychological Bulletin, 72(4), 299–306. https://doi.org/10.1037/h0028431