Research synthesis: meaning and application

You have survey results, interview notes and behavioural data on the same question, and they do not tell exactly the same story. Research synthesis is the step in which you systematically place those findings side by side and turn them into conclusions you can substantiate. That takes more than a summary per source.

Research synthesis is the bringing together and integration of findings from multiple data sources, methods or groups into conclusions that you can substantiate statement by statement. A synthesis answers a research question; a summary reports what each source says on its own.

The term has two senses. In the methodological literature, it often means systematically combining results from many studies, as in a systematic review or meta-analysis. This article uses it more broadly: for integrating findings from, say, surveys, interviews and behavioural data within one study or programme.

A survey shows how often an opinion occurs, interviews show how people explain their choice, and behavioural data show what people actually do. Synthesis turns these partial answers into one substantiated answer.

Three features distinguish a synthesis from a collection of results:

• Integration: the conclusion says something no single source could say.
• Traceability: every conclusion can be traced back to the sources and analyses it rests on.
• Transparency about uncertainty: you make visible where sources reinforce each other, where they are silent and where they clash.

Here, it is an analytical way of working, not a report chapter.

The concept has two roots. The first is synthesis across studies. The Handbook of Research Synthesis and Meta-Analysis describes research synthesis as the practice of systematically distilling and integrating data from many studies in order to draw more reliable conclusions about a research issue (Cooper, Hedges & Valentine, 2019). Systematic reviews and meta-analyses grew out of this tradition.

The second is mixed methods research, which combines quantitative and qualitative data within one study. Fetters, Curry and Creswell describe integration there at three levels: in the study design, in the methods, and in interpretation and reporting (Fetters, Curry & Creswell, 2013). The textbook by Creswell and Plano Clark sets out the corresponding research designs (Creswell & Plano Clark, 2018).

This article uses research synthesis as an umbrella term for these related practices, with the emphasis on integration within a single study.

• A summary reports per source what was found. It does not answer the research question on its own.
• A systematic review attempts to collate all the empirical evidence that fits pre-specified eligibility criteria in order to answer a specific research question. It uses explicit methods intended to minimise bias (Lasserson, Thomas & Higgins, 2024).
• A meta-analysis combines the quantitative results of several studies statistically. Whether that makes sense depends on how comparable the studies are.
• A research synthesis in the sense of this article integrates findings from multiple sources, methods or groups within one study or programme. It does not need to search exhaustively like a systematic review or pool statistically like a meta-analysis, but every conclusion must be traceable.

A synthesis within a single study therefore does not carry the evidential weight of a systematic review.

At the methods level, Fetters and colleagues distinguish four ways in which data sources become linked (Fetters, Curry & Creswell, 2013). In connecting, you link sources through sampling, such as selecting interviewees from survey respondents. In building, one source shapes how you collect the other, in merging you analyse both databases together, and in embedding you link data collection and analysis at several points.

At the level of interpretation and reporting, they name three forms: narrative integration, data transformation and the joint display. Data transformation means converting one type of data into the other, for example by counting qualitative themes. They also describe the fit of integration: the extent to which qualitative and quantitative findings cohere.

For research synthesis, that last level matters most: even sources collected separately can still be integrated explicitly at the interpretation stage.

Thomas and Harden describe thematic synthesis in three stages: coding text line by line, developing descriptive themes that stay close to the primary studies, and generating analytical themes. With these, the reviewers go beyond the primary studies and generate new interpretive constructs (Thomas & Harden, 2008). The method was developed for reviews of qualitative research, but the distinction between describing and interpreting also works within a single study.

The authors call that third stage potentially the most controversial, because it depends on the judgement of the researchers. So record which codes and sources support each analytical theme.

The following steps are an editorial way of working, not a standardised protocol.

1. Put the research question at the centre. Formulate the question and the decision the synthesis needs to support. For each source, determine which sub-question it can and cannot answer.
2. Place the sources side by side. Organise by theme or sub-question, not by source. Note each finding's source, group and method, and how large and representative that source is.
3. Look for patterns and contradictions. For each theme, mark where sources agree, complement each other, are silent or clash. A contradiction is a call for explanation, not noise.
4. Assess the strength of evidence per conclusion. State which sources each conclusion rests on, how suitable they are for this question and whether several independent sources support it.
5. State the uncertainty. Write down what you do not know, which alternative explanations remain open and what additional research would reduce the uncertainty.

A joint display is a table or figure in which you present quantitative and qualitative findings together, to reveal insights that the separate results do not provide. In an analysis of published joint displays, statistics-by-themes and side-by-side comparisons were the most common types (Guetterman, Fetters & Creswell, 2015).

A simple template, without measurement results:

The last two columns force you to formulate a substantiated judgement for each theme.

O'Cathain, Murphy and Nicholl describe the triangulation protocol. In it, you check for each theme whether findings from different components agree, partially agree, or whether there is silence or dissonance (O'Cathain, Murphy & Nicholl, 2010). Silence means that a theme arises in one dataset and not in another. The authors note that exploring an apparent discrepancy between methods may lead to a better understanding of the research question.

Moffatt and colleagues worked this out in a study in which quantitative and qualitative results diverged (Moffatt et al., 2006). Their conclusion: investigating such differences may be as illuminating as the points of similarity.

Triangulation, testing a finding with multiple sources or methods, is related and has its own article.

This example is fictional and contains no real research results.

Situation: a museum is considering a new exhibition on the history of its own city. Choice question: which format and which communication suit the visitors the museum wants to reach? Available information: a visitor survey after previous exhibitions, interviews with visitors and with people who rarely visit the museum, and ticket data with visit times and ticket types.

Approach: the researcher first formulates three sub-questions. Whom does the museum reach now? What do visitors value? What holds others back? For each sub-question, she builds a joint display with the three sources side by side.

Suppose the survey shows visitors value depth, while ticket data point to many short afternoon visits. That need not be an error in either source. Possible explanations: the survey mainly reaches visitors who already stayed longer, short visits belong to a different group, such as families, or depth means something other than time to visitors. The interviews help to choose, but it remains an interpretation.

Interpretation and next step: the synthesis states the strength of evidence per conclusion, for example "supported by three sources" or "from interviews only, still to be tested". A targeted survey question or a small test with two exhibition texts can follow.

In audience and communication research, you often combine association data, interviews, surveys and behavioural data. A synthesis helps determine which insights are robust enough for a strategic choice and which you need to test first, for example as a hypothesis for a message. See How do you measure associations?. If new findings change an existing picture of an audience, the synthesis is the basis for updating it (When and how should you update a target group profile?).

• No causality from combining. Three sources showing the same pattern make it more credible, but not yet causal; that requires an appropriate research design.
• Group level. A synthesis describes patterns in groups. It does not predict what an individual visitor, customer or employee thinks or does.
• Bias of the person synthesising. The person synthesising decides what is relevant. Confirmation bias, seeking or interpreting evidence in ways that are partial to existing beliefs or a hypothesis, has been extensively described (Nickerson, 1998). So decide in advance which question you are answering, actively look for counter-evidence and have a second person check the synthesis.
• Dependent on the sources. Weak or incomparable sources do not become stronger because you combine them.

Research synthesis turns separate findings into a substantiated answer to a research question. The core is traceability: for each conclusion, you know which sources it rests on, how strong that evidence is and what uncertainty remains. You treat contradictory findings as information about differences between groups, methods or situations.