Feynman, scientific method and sociology

Water drops

Kevin Ralston, University of Edinburgh, 2020

The Scientific Method, Richard Feynman YouTube


Richard Feynman was a Nobel Prize winning physicist. He was also known as a charismatic and innovative teacher. In the extract below (from the YouTube video link above) Feynman outlined a succinct model of a scientific method to a class of learners.

A colleague recently commented to me how students in their large introductory class were sometimes able to complete that course without a sense that sociology could be an empirical pursuit. This blog illustrates how the simple research model Feynman outlines can apply in sociology. Although this may seem obvious to many, sociology is low consensus subject where empirical research is not necessarily at the forefront in the teaching or publication.

Irrespective of our own interests, as sociologists we do a disservice to our students if we limit their horizons by presenting, or making accessible, only a narrow range of sociological practice. This would bypass the role of key sociological figures such as John Goldthorpe, who argues that sociology can focus on finding empirical regularity and underlying processes. Also, Robert Merton’s middle range theory approach to relating social theory with empirical research.

A simple instantiation of how sociology can follow scientific principles, developed from the model provided by Feynman, could play a useful role in helping to demonstrate this as practice encompassed within sociology. This may help undergraduates and those new to sociology to see scientific/empirical goals as legitimate and to incorporate these into their understandings of the continuum of practice that forms contemporary sociology.

Feynman on scientific method:

Now I’m going to discuss how we would look for a new law. In general, we look for a new law by the following process.

First, we guess it.

Then we compute the consequences of the guess, to see what, if this is right, if this law we guess is right, to see what it would imply and then we compare the computation results to nature or we say compare to experiment or experience, compare it directly with observations to see if it works.

If it disagrees with experiment, it’s wrong. In that simple statement is the key to science.

It doesn’t make any difference how beautiful your guess is, it doesn’t matter how smart you are who made the guess, or what his name is … If it disagrees with experiment, it’s wrong. That’s all there is to it.

In Feynman’s model we make a guess about the world, we undertake observation to assess the credibility of that guess and we decide whether observation supports or refutes the guess.

Feynman’s model in sociology

Guess >>> observation (analysis) >>> does the analysis support the guess?


Science can begin with a guess! We make a guess about the social world, this guess is often based on theory, or previous research evidence. If we want our research to be important and exciting we look for gaps in knowledge or areas of fierce debate and we form a guess in this area. Originality and relevance score big with lecturers marking work and with reviewers reading articles.

In empirical sociology the guess often takes the form of a research question. Such as:

Are there gender differences in patterns of occupational change in the youth labour market following the Great Recession of 2008?

There are really at least two ‘guesses’ here. That there has been changes in the types of jobs young people have been doing following the great recession of 2008 and that changes were different for young men compared to young women. These are questions of considerable sociological interest. Areas such as the sociology of youth, gender inequality and shifts in social mobility are long run concerns in sociological research.

In courses on research methods the guess is often presented as a hypothesis. Formal hypotheses are rarely presented in published sociology, although many hypotheses are usually implicit in any research. There are a number of hypotheses implied in the research question above.

One hypothesis might be: the occupational position of young people has declined following the Great Recession of 2008.

A second hypothesis might be: change in occupational position following the Great Recession of 2008 has disadvantaged women.

We could find data and design a methodology to test each of these statements and either reject the hypotheses or fail to reject these hypothesis on the basis of results.

So we have guessed at what is going on in the youth labour market, we need analysis to assess our guesses.


An alternative to a science experiment in sociology is data collection and analysis. This is because our area of study is social life and the lived lives of people. It is therefore not typically possible to conduct formal experiments, or even randomised control trials.

There are practical and ethical dimensions to this. Sociological research examines complex ‘open systems’ that people inhabit. Consider the question above, on gender differences in occupational position over time. Even if it were possible to randomly select people to, for instance, apply for specific occupations, in order to assess inequality in recruitment practices, it would be ethically problematic to intervene in the lives of young people in order to do so.

To answer questions about changes by gender in the occupational position of young people, we need detailed observational data on the type of jobs young people were doing at different time points. If we want to compare what young people are doing before 2008 with those a decade later then we might need data collected from 2005 to 2018, say. If we want to be able to suggest that our findings are robust and generally applicable then these data need to be representative of the population we are interested in analysing.

Statistical methods can enable a comparison between groups (e.g. young men and women) and time points (e.g. the period prior to the 2007 crash and recession of 2008 with data collected ten years later). Statistical methods allow us to make an assessment of how likely it is that any differences or changes we see in a representative sample (such as collected by large government surveys) are reflective of differences or changes in the population as a whole.

The Labour Force Survey is a large scale representative survey that can enable the analysis of the questions posed. It has been running since the 1970s and data are collected on the UK labour four times a year, including information on people in work and the types of jobs they are doing.

Statistical methods appropriate for the analysis of this type of survey data are often applied by sociologists to examine research questions. These methods are amongst the most powerful tools that sociologists have in their toolbox. The analysis of complex survey datasets, collected over many years, can help us to understand long run social change.

Doing science

Using the Labour Force Survey we could design analyses to answer the question of whether the observed trends in our data match with our guess. If they do then we have evidence that our guess is correct. If they do not then we have evidence our guess is wrong.

In the case of our example we wanted to know about young people and whether gendering in types of jobs they do has changed in the decade following the Great Recession of 2008 (we actually did this analysis for a publication due to be published in 2020/21). In some ways this is quite a complex issue involving analysis by time points and gender, linking together data collected over a period of time.

Were we to find change between time points, and if the patterning of that change is different for men and women then this would count as evidence in support of our guesses. If not then we have evidence that our guesses are wrong and we must reassess our understanding of the phenomenon.

If this was the case, in reporting our findings, we would usually report that our data did not support our expectations. We would then typically go on to suggest why we think this might have been the case. This may well lead to discussion of the plausibility of several explanations, or alternative guesses. These alternatives might reflect and draw upon the research literature or wider theory. It may be that other research has found a similar result to the one we observe and our findings would support this, rather than our expectations.


A substantial proportion of sociology is theoretical and critical. At times this may confuse students and beguile those new to sociology into seeing sociology as potentially only encompassing these pursuits. This would be to ignore the arguments and works of key sociological figures such as John Goldthorpe, who argues that sociology can focus on finding empirical regularity and underlying processes. Also, Robert Merton’s middle range theory approach to relating social theory with empirical research.

There is a great deal of debate in the philosophy of science. For example, over whether science is about working within, and expanding, a particular paradigm (Kuhn), or constantly testing whether theory should be rejected (Popper). In Feynman’s version we start with an informed guess about the social world, undertake an analysis to assess the guess and proceed to consider whether the analysis supports the guess or not. This blog has outlined how this approach, simply outlined by Richard Feynman, can be applied in sociology. This indicates the legitimacy of approaches involving scientific observation in addressing questions on the social world. It is envisaged that this could be helpful as a first step in introducing new sociologists to the potential relevance of empirical (as well as critical or theoretical) sociology to their own sociological interests.

Acknowledgements: Thank you to Alan Marshall and Roxanne Connelly for their comments on this blog. Photo by Linus Mimietz on Unsplash


The Research Training Centre applies expertise in a diverse set of social science methods, through teaching and research. The Centre provides new insights to the key challenges in society.

More Posts