what is framework analysis in research

• Privacy Policy

Home » Framework Analysis – Method, Types and Examples

Framework Analysis – Method, Types and Examples

Table of Contents

Framework Analysis

Definition:

Framework Analysis is a qualitative research method that involves organizing and analyzing data using a predefined analytical framework. The analytical framework is a set of predetermined themes or categories that are derived from the research questions or objectives. The framework provides a structured approach to data analysis and can help to identify patterns, themes, and relationships in the data.

Steps in Framework Analysis

Here are the general steps in Framework Analysis:

Familiarization

Get familiar with the data by reading and re-reading it. This step helps you to become immersed in the data and to get a sense of its content, structure, and scope.

Identify a Coding Framework

Identify a coding framework or set of themes that will be used to analyze the data. These themes can be derived from existing literature or theories or developed based on the data itself.

Code the data by applying the coding framework to the data. This involves breaking down the data into smaller units and assigning each unit to a particular theme or category.

Chart or summarize the data by creating tables or matrices that display the distribution and frequency of each theme or category across the data set.

Mapping and interpretation

Analyze the data by examining the relationship between different themes or categories, and by exploring the implications and meanings of the findings in relation to the research question.

Verification

Verify the accuracy and validity of the findings by checking them against the original data, comparing them with other sources of data, and seeking feedback from others.

Report the findings by presenting them in a clear, concise, and organized manner. This involves summarizing the key themes, presenting supporting evidence, and providing interpretations and recommendations based on the findings.

Framework Analysis Conducting Guide

Here is a step-by-step guide to conducting framework analysis:

• Define the research question: The first step in conducting framework analysis is to clearly define the research question or objective that you want to investigate.
• Develop the analytical framework: Develop a coding framework or a set of predetermined themes or categories that are relevant to the research question. These themes or categories can be derived from existing literature or theories, or they can be developed based on the data collected.
• Data collection: Collect the data using a suitable method such as interviews, focus groups, surveys or observation.
• Familiarization: Transcribe and familiarize yourself with the data. Read through the data several times and take notes to identify any patterns, themes or issues that are emerging.
• Coding : Code the data by identifying key themes or categories and assigning each piece of information to a specific theme or category.
• Charting: Create charts or tables that display the frequency and distribution of each theme or category. This helps to summarize the data and identify patterns.
• Mapping and interpretation: Analyze the data to identify patterns, relationships, and themes. Interpret the findings in light of the research objectives and provide explanations for any significant patterns or themes that have emerged.
• Validation : Validate the findings by sharing them with others and seeking feedback. This can help to ensure that the findings are robust and reliable.
• Report writing: Write a report that summarizes the findings, includes quotes or examples from the data to support the findings and provides recommendations for future research.

Applications of Framework Analysis

Framework Analysis has a wide range of applications in research, including:

• Policy analysis: Framework Analysis can be used to analyze policies and policy documents to identify key themes, patterns, and underlying assumptions.
• Social science research: Framework Analysis is commonly used in social science research to analyze qualitative data from interviews, focus groups, and other sources.
• Health research: Framework Analysis can be used to analyze qualitative data from health research studies, such as patient and provider perspectives, to identify themes and patterns.
• Environmental research : Framework Analysis can be used to analyze qualitative data from environmental research studies to identify themes and patterns related to environmental attitudes, behaviors, and practices.
• Education research: Framework Analysis can be used to analyze qualitative data from educational research studies to identify themes and patterns related to teaching practices, student learning, and educational policies.
• Market research: Framework Analysis can be used to analyze qualitative data from market research studies to identify themes and patterns related to consumer attitudes, behaviors, and preferences.

Examples of Framework Analysis

Here are some examples of Framework Analysis in various research contexts:

• Health Research: A study on the experiences of cancer survivors might use Framework Analysis to identify themes related to the psychological, social, and physical aspects of survivorship. Themes might include coping strategies, social support, and health outcomes.
• Education Research: A study on the impact of a new teaching approach might use Framework Analysis to identify themes related to the implementation of the approach, the effectiveness of the approach, and barriers to its implementation. Themes might include teacher attitudes, student engagement, and logistical challenges.
• Environmental Research : A study on the factors that influence pro-environmental behaviors might use Framework Analysis to identify themes related to environmental attitudes, behaviors, and practices. Themes might include social norms, personal values, and perceived barriers to behavior change.
• Policy Analysis: A study on the implementation of a new policy might use Framework Analysis to identify themes related to policy development, implementation, and outcomes. Themes might include stakeholder perspectives, organizational structures, and policy effectiveness.
• Social Science Research: A study on the experiences of immigrant families might use Framework Analysis to identify themes related to the challenges and opportunities faced by immigrant families in their new country. Themes might include language barriers, cultural differences, and social support.

When to use Framework Analysis

Framework Analysis is a useful method for analyzing qualitative data when the research questions require an in-depth exploration of a particular phenomenon, concept, or experience. It is particularly useful when:

• The research involves multiple sources of qualitative data, such as interviews, focus groups, or documents, that need to be analyzed and compared.
• The research questions require a systematic and structured approach to data analysis that enables the identification of patterns, themes, and relationships in the data.
• The research involves a large and complex dataset that requires a method for organizing and synthesizing the data in a meaningful way.
• The research aims to generate new insights and understandings from the data, rather than testing pre-existing hypotheses or theories.
• The research requires a method that is transparent, replicable, and verifiable, as Framework Analysis provides a clear framework for data analysis and reporting.

Purpose of Framework Analysis

The purpose of Framework Analysis is to systematically organize and analyze qualitative data in a structured and transparent manner. The method is designed to identify patterns, themes, and relationships in the data that are relevant to the research question or objective. By using a rigorous and transparent approach to data analysis, Framework Analysis enables researchers to generate new insights and understandings from the data, and to provide a clear and structured presentation of the findings.

The method is particularly useful for analyzing large and complex qualitative datasets that require a method for organizing and synthesizing the data in a meaningful way. It can be used to explore a wide range of research questions and objectives across various fields, including health research, social science research, education research, policy analysis, and environmental research, among others.

Overall, the purpose of Framework Analysis is to provide a systematic and transparent method for analyzing qualitative data that enables researchers to generate new insights and understandings from the data in a rigorous and structured manner.

Characteristics of Framework Analysis

Some Characteristics of Framework Analysis are:

• Systematic and Structured Approach: Framework Analysis provides a systematic and structured approach to data analysis that involves a series of steps that are followed in a predetermined order.
• Transparency and Replicability: Framework Analysis emphasizes transparency and replicability, as it involves a clearly defined process for data analysis that can be applied consistently across different datasets and research questions.
• Flexibility : Framework Analysis is flexible and adaptable to a wide range of research contexts and objectives, as it can be used to analyze qualitative data from various sources and to explore different research questions.
• In-depth Exploration of the Data: Framework Analysis enables an in-depth exploration of the data, as it involves a thorough and detailed analysis of the data to identify patterns, themes, and relationships.
• Applicable to Large and Complex Datasets: Framework Analysis is particularly useful for analyzing large and complex qualitative datasets, as it provides a method for organizing and synthesizing the data in a meaningful way.
• Data-Driven: Framework Analysis is data-driven, as it focuses on the analysis and interpretation of the data rather than on pre-existing hypotheses or theories.
• Emphasis on Contextual Understanding : Framework Analysis emphasizes contextual understanding, as it involves a detailed examination of the data to identify the social, cultural, and environmental factors that may influence the phenomena under investigation.

Advantages of Framework Analysis

Some Advantages of Framework Analysis are as follows:

• Transparency : Framework Analysis provides a clear and structured approach to data analysis, which makes the process transparent and easy to follow. This ensures that the findings can be easily replicated or verified by other researchers.
• Thorough Analysis : Framework Analysis enables a thorough and detailed analysis of the data, which allows for the identification of patterns, themes, and relationships that may not be apparent through other methods.
• Contextual Understanding: Framework Analysis emphasizes the importance of understanding the context in which the data was collected, which enables a more nuanced interpretation of the findings.
• Collaborative Analysis: Framework Analysis can be used as a collaborative method for data analysis, as it allows multiple researchers to work together to analyze the data and develop a shared understanding of the findings.
• Efficient and Time-saving: Framework Analysis can be an efficient and time-saving method for analyzing qualitative data, as it provides a structured and organized approach to data analysis that can help researchers manage and synthesize large datasets.
• Comprehensive Reporting: Framework Analysis can help ensure that the research findings are comprehensive and well-reported, as the method provides a clear framework for presenting the results.

Limitations of Framework Analysis

Some Limitations of Framework Analysis are as follows:

• Subjectivity : Framework Analysis relies on the interpretation of the researchers, which can introduce subjectivity into the analysis process.
• Time-consuming : Framework Analysis can be a time-consuming method for data analysis, particularly when working with large and complex datasets.
• Limited ability to generate new theory : Framework Analysis is a deductive approach that relies on pre-existing theories and concepts to guide the analysis, which may limit the ability to generate new theoretical insights.
• Risk of oversimplification: The structured approach of Framework Analysis can lead to oversimplification of the data, as complex issues may be reduced to predefined categories or themes.
• Limited ability to capture the complexity of the data : The predefined categories or themes used in Framework Analysis may not be able to capture the full complexity of the data, particularly when dealing with nuanced or context-specific phenomena.
• Limited use with non-textual data : Framework Analysis is primarily designed for analyzing qualitative textual data and may not be as effective for analyzing non-textual data such as images, videos, or audio recordings.

About the author

Muhammad Hassan

Researcher, Academic Writer, Web developer

You may also like

Cluster Analysis – Types, Methods and Examples

Discriminant Analysis – Methods, Types and...

MANOVA (Multivariate Analysis of Variance) –...

Documentary Analysis – Methods, Applications and...

ANOVA (Analysis of variance) – Formulas, Types...

Graphical Methods – Types, Examples and Guide

Framework Analysis: Methods and Use Cases

Introduction

What is framework analysis, implementing the framework analysis methodology.

Among qualitative methods in social research, framework analysis stands out as a structured approach to analyzing qualitative data . Originally developed for applied policy analysis and multi-disciplinary health research, this method has found application in various domains due to its emphasis on transparency and systematic data analysis. As with other research methods, the objective remains to extract meaningful themes and patterns, but the framework method provides a specific roadmap for doing so.

Whether you're a seasoned researcher or someone new to the realm of qualitative methodology, understanding the nuances of framework analysis can enhance the depth and rigor of your research efforts. In this article, we will explore the methods and use cases of framework analysis, diving deep into its benefits and the analytical framework it enables researchers to develop.

Framework analysis is a systematic approach for analyzing qualitative data . Rooted in the traditions of social research relevant to policy making, it was found to be a useful tool for analysis in multi-disciplinary health research where the eventual analysis of qualitative data can identify themes and actionable insights relevant to policy outcomes.

Unlike some other qualitative analysis methods , framework analysis is explicitly focused on addressing specific research questions , making it particularly suitable for applied or policy-related qualitative research .

Purpose of framework analysis

The primary aim of framework analysis is to offer a clear and transparent process for conducting qualitative research by managing, reducing, and analyzing large datasets without losing sight of the original context. Given the vast amounts of data often generated in qualitative studies, having a systematic method to sift through this data is crucial.

By using the framework method, researchers can remain focused on their research questions while ensuring that the data collection and analysis process retains its integrity and depth.

Characteristics of framework analysis

Transparent structure: One of the distinct features of framework analysis is its emphasis on transparency . Every step in the analysis process is documented, allowing for easy scrutiny and replication by multiple researchers.

Thematic framework: Central to framework analysis is the development of a framework identifying key themes, concepts, and relationships in the data. The framework guides the subsequent stages of coding and charting.

Flexibility: While it provides a clear structure, framework analysis is also adaptable. Depending on the objectives of the study, researchers can modify the process to better suit their data and questions.

Iterative process: The process in framework analysis is not linear. As data is collected and data analysis progresses, researchers often revisit earlier stages, refining the framework or revising codes to better capture the nuances in the data.

Benefits of framework analysis

Conducting framework analysis has several advantages:

Rigorous data management: The structured approach means data is managed and analyzed with a high level of rigor, minimizing the potential influence of preconceptions.

Inclusivity: Framework analysis accommodates both a priori issues, driven by the research questions , and emergent issues that arise from the data itself.

Comparability: Given its structured nature, framework analysis allows researchers to compare and contrast data, facilitating the identification of patterns and differences.

Accessibility: By presenting data in a summarized, charted form , findings from framework analysis become more accessible and comprehensible, aiding in reporting and disseminating results.

Relevance for applied research: Given its origins in policy research and its clear focus on addressing specific research questions, framework analysis is particularly relevant for studies aiming to inform policy or practice.

Efficient, easy data analysis with ATLAS.ti

Start analyzing data quickly and more deeply with ATLAS.ti. Download a free trial today.

Successfully conducting framework analysis involves a series of structured steps. Proper implementation of framework analysis not only ensures the rigor of a qualitative analysis but also that the findings are credible and meaningful.

Familiarization with the data

Before discussing a more detailed analysis, it's paramount to understand the breadth and depth of the data at hand.

Reading and re-reading: Begin by reading textual data such as transcripts , field notes , and other data sources multiple times. This immersion allows researchers to understand participants' perspectives and grasp the overall context.

Noting preliminary ideas: As researchers familiarize themselves with the data, preliminary themes or ideas may start to emerge. Jotting these down in memos helps in forming an initial understanding and can be instrumental in the subsequent phase of developing a set of themes.

Developing a thematic framework

As is the case across nearly all types of qualitative methodology , central to framework analysis is the construction of a robust analytical framework . This structure aids in organizing and interpreting the data .

Identifying key themes: Based on the initial familiarization, it's important to identify themes that occur in the multimedia or textual data. These themes should be relevant to the research question . Researchers can begin assigning codes to specific chunks of data to capture emerging themes.

Categorizing and coding: Each identified theme can further be broken down into sub-themes or brought together under categories. At this stage, researchers can continue coding (or recoding ) their data according to these themes or categories.

Refining the framework: As the analysis progresses, the initial themes represented by your coding framework may need adjustments. It's an iterative process, where the framework can be continually refined to better fit the data.

Indexing and charting the data

Once the framework is established, the next phase involves systematically applying it to the data.

Indexing: Using the resulting coding framework , you can verify that codes have been systematically assigned to relevant portions of the data. This ensures every relevant piece of data is categorized under the appropriate theme or sub-theme.

Charting: This step involves creating charts or matrices for each theme. Data from different sources (like interviews or focus groups ) is summarized under the relevant theme. For example, a table can be created with each theme in a column and each data source in a row, and researchers can then populate the cells with relevant data extracts or notes. These charts provide a visual representation , allowing researchers to easily see patterns or discrepancies in the data.

Mapping and Interpretation: With the data systematically charted, researchers can begin to map the relationships between themes and interpret the broader implications. This step is where the true essence of the research emerges, as researchers link the patterns in the data to the broader objectives of the study.

Framework analysis is an involved process, with intentional decision-making at every step of the way. As a result, implementing structured qualitative methodologies such as framework analysis requires patience, meticulous attention to detail, and a clear understanding of the research objectives. When conducted diligently, it offers a transparent and systematic approach to analyzing qualitative data , ensuring the research not only has depth but also clarity.

Whether comparing data across multiple sources or drilling down into the nuances of individual narratives, framework analysis equips researchers with the tools needed to derive meaningful insights from their qualitative data . As more researchers across disciplines recognize its value, it stands to become an even more integral part of the research landscape.

Systematic data analysis made easy with ATLAS.ti

Powerful tools to analyze qualitative data are right at your fingertips. Download your free trial today.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List

Using framework-based synthesis for conducting reviews of qualitative studies

Mary dixon-woods.

1 Social Science Research Group, Department of Health Sciences, School of Medicine, University of Leicester, Leicester, UK

Framework analysis is a technique used for data analysis in primary qualitative research. Recent years have seen its being adapted to conduct syntheses of qualitative studies. Framework-based synthesis shows considerable promise in addressing applied policy questions. An innovation in the approach, known as 'best fit' framework synthesis, has been published in BMC Medical Research Methodology this month. It involves reviewers in choosing a conceptual model likely to be suitable for the question of the review, and using it as the basis of their initial coding framework. This framework is then modified in response to the evidence reported in the studies in the reviews, so that the final product is a revised framework that may include both modified factors and new factors that were not anticipated in the original model. 'Best fit' framework-based synthesis may be especially suitable in addressing urgent policy questions where the need for a more fully developed synthesis is balanced by the need for a quick answer.

Please see related article: http://www.biomedcentral.com/1471-2288/11/29 .

Introduction

As the appetite for more holistic overviews of research evidence has grown, the last 10-15 years have seen increasing interest and activity directed at developing methods for synthesising qualitative studies [ 1 ]. Some of these methods might be fairly described as 'new' techniques that have been developed specifically for the purpose of conducting synthesis, while others might more properly be seen as adaptations of approaches that were originally intended for primary research [ 2 ]. Framework analysis is one of the latter. Developed during the 1980s by the UK-based National Centre for Social Research, and explicitly oriented towards applied policy questions, framework analysis is a matrix-based method involving the construction of thematic categories into which data can be coded [ 3 ]. One important feature of the approach is that, unlike some other qualitative methods, it allows themes or concepts identified a priori to be specified as coding categories from the outset, and to be combined with other themes or concepts that emerge de novo by subjecting the data to inductive analysis. A practical benefit of doing this is that it enables questions or issues identified in advance by various stakeholders (such as policymakers, practitioners, or user groups) to be explicitly and systematically considered in the analysis, while also facilitating enough flexibility to detect and characterise issues that emerge from the data.

Framework analysis has become hugely popular as a way of conducting analysis of primary qualitative data, especially in areas of healthcare with policy relevance. A recent study, for example, used framework analysis in a study of mothers' interpretations of dietary recommendations [ 4 ]. Because the study had been guided by social learning theory, and because the researchers were interested in comparing dietary beliefs and behaviours across social classes, the ability of framework analysis to cope with categories specified in advance of the data collection made it a very appropriate choice of analytic strategy. A further advantage is the use of charting techniques, which help not only in enhancing the transparency of coding, but also with teamwork in relation to analysis. Many of the properties that make framework analysis an appealing option for those conducting primary research make 'framework-based synthesis' a potentially equally attractive option for those seeking to conduct a synthesis of studies. A new article by Carroll et al [ 5 ] published this month in BMC Medical Research methodology reports an interesting evolution of the framework-based synthesis approach.

Framework analysis has previously been adapted for use in synthesis of qualitative evidence by Oliver and colleagues [ 6 ], who developed a multidimensional framework for analysing public involvement in health services research. They employed an iterative process involving familiarization with the literature, gradually developing a conceptual framework based on concepts derived from the review question and the theoretical and empirical literature, applying the framework systematically to evidence from the studies included in the review, and constructing a chart for each key dimension with distilled summaries from all relevant documents. The charts were used to map the range and nature of public involvement and to find associations between themes. Several benefits of this approach were noted by Oliver et al ., including widening the scope of the review to include relevant topics identified by lay people, and the creation of data displays that could be viewed and assessed by people other than the primary analyst. Other examples of the application of the framework approach to the conduct of synthesis have now begun to appear, though it is dismaying that these and other syntheses of qualitative evidence continue to label themselves with the confusing, tautological, and inappropriate term 'meta-synthesis' [ 7 ].

In the recent BMC Medical Research Methodology manuscript [ 5 ], Carroll and colleagues develop the framework-based synthesis approach, focusing on the views of adults taking chemopreventative agents (such as aspirin or vitamins) in an effort to prevent colorectal cancer. One of the novel features of their approach is their use of a conceptual model that was used as an initial starting point for coding the evidence from 20 studies. This conceptual model was chosen because of its broad applicability to the area under review, and the authors did not engage in the more lengthy process of model specification that is often more characteristic of framework synthesis. They augmented analysis using this prespecified model with analysis that was more inductive, and ended up generating a revised conceptual model that provided a 'best fit' to the evidence reported in the studies they reviewed. The revised model included some factors that were absent from the original model, as well as adjustments to factors that had been reported in that model.

Carroll and colleagues emphasise the advantages of this kind of approach when time is short and the demand for policy-relevant evidence is urgent. It enables focusing of the research on the priorities of those commissioning the work, while still leaving some room for finding the 'best fit' in the light of what the evidence actually reports. Of course, like framework analysis for primary research [ 8 ] there are downsides of the approach too. Reviewers who have made a hefty investment in an initial conceptual model may be unconsciously motivated to recover the sunk costs of that model, and as a consequence tend to neglect evidence that presents a fundamental challenge. Putting more time into specifying the model, using a wider range of literature, and gaining the views of a wider range of stakeholders may all be important in improving the legitimacy and validity of any ensuring synthesis. There are also the usual risks of framework analysis that it can tend to suppress interpretive creativity, and thus reduce some of the vividness of insight seen in the best qualitative research. Nonetheless, as Carroll and colleagues argue, framework-based synthesis using the 'best fit' strategy is, in the right hands, likely to be a highly pragmatic and useful approach for a range of policy urgent questions.

Conclusions

Framework-based synthesis is an important advance in conducting reviews of qualitative synthesis. The 'best fit' strategy is a variant of this approach that may be very helpful when policymakers, practitioners or other decision makers need answers quickly, and are able to tolerate some ambiguity about whether the answer is the very best that could be given.

Competing interests

The author declares that they have no competing interests.

Pre-publication history

The pre-publication history for this paper can be accessed here:

http://www.biomedcentral.com/1741-7015/9/39/prepub

• Barnett-Page E, Thomas J. Methods for the synthesis of qualitative research: a critical review. BMC Med Res Methodol. 2009; 9 :59. doi: 10.1186/1471-2288-9-59. [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Dixon-Woods M, Agarwal S, Jones D, Young B, Sutton A. Synthesising qualitative and quantitative evidence: a review of possible methods. J Health Serv Res Pol. 2005; 10 :45–53. doi: 10.1258/1355819052801804. [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Ritchie J, Spencer L. In: Analyzing Qualitative Data. Bryman A, Burgess RG, editor. London: Routledge; 1994. Qualitative data analysis for applied policy research. [ Google Scholar ]
• Wood F, Robling M, Prout H, Kinnersley P, Houston H, Butler C. A question of balance: a qualitative study of mothers' interpretations of dietary recommendations. Ann Fam Med. 2010; 8 :51–57. doi: 10.1370/afm.1072. [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Carroll C, Booth A, Cooper K. A worked example of "best fit" framework synthesis: a systematic review of views concerning the taking of some potential chemopreventive agents. BMC Med Res Methodol. 2011; 11 :29. doi: 10.1186/1471-2288-11-29. [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Oliver S, Rees R, Clarke-Jones L, Milne R, Oakley AR, Gabbay J, Stein K, Buchanan P, Gyte G. A multidimensional conceptual framework for analysing public involvement in health services research. Health Exp. 2008; 11 :72–84. doi: 10.1111/j.1369-7625.2007.00476.x. [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Seymour C, Addington-Hall J, Lucassen AM, Foster CL. What facilitates or impedes family communication following genetic testing for cancer risk? A systematic review and meta-synthesis of primary qualitative research. J Genet Couns. 2010; 19 :330–342. doi: 10.1007/s10897-010-9296-y. [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Barbour R. The newfound credibility of qualitative research? Tales of technical essentialism and cooption. Qual Health Res. 2003; 13 :1019–1027. doi: 10.1177/1049732303253331. [ PubMed ] [ CrossRef ] [ Google Scholar ]

Using Framework Analysis in nursing research: a worked example

Affiliation.

• 1 School of Nursing, Midwifery & Social Work, University of Manchester, UK.
• PMID: 23517523
• DOI: 10.1111/jan.12127

Aims: To demonstrate Framework Analysis using a worked example and to illustrate how criticisms of qualitative data analysis including issues of clarity and transparency can be addressed.

Background: Critics of the analysis of qualitative data sometimes cite lack of clarity and transparency about analytical procedures; this can deter nurse researchers from undertaking qualitative studies. Framework Analysis is flexible, systematic, and rigorous, offering clarity, transparency, an audit trail, an option for theme-based and case-based analysis and for readily retrievable data. This paper offers further explanation of the process undertaken which is illustrated with a worked example.

Data source and research design: Data were collected from 31 nursing students in 2009 using semi-structured interviews.

Discussion: The data collected are not reported directly here but used as a worked example for the five steps of Framework Analysis. Suggestions are provided to guide researchers through essential steps in undertaking Framework Analysis. The benefits and limitations of Framework Analysis are discussed.

Implications for nursing: Nurses increasingly use qualitative research methods and need to use an analysis approach that offers transparency and rigour which Framework Analysis can provide. Nurse researchers may find the detailed critique of Framework Analysis presented in this paper a useful resource when designing and conducting qualitative studies.

Conclusion: Qualitative data analysis presents challenges in relation to the volume and complexity of data obtained and the need to present an 'audit trail' for those using the research findings. Framework Analysis is an appropriate, rigorous and systematic method for undertaking qualitative analysis.

Keywords: Framework Analysis; nursing; qualitative data analysis.

© 2013 Blackwell Publishing Ltd.

Publication types

• Research Support, Non-U.S. Gov't
• Nursing Research / methods*
• Nursing Research / statistics & numerical data
• Qualitative Research*
• Statistics as Topic / methods*

• Correspondence
• Open access
• Published: 18 September 2013

Using the framework method for the analysis of qualitative data in multi-disciplinary health research

• Nicola K Gale 1 ,
• Gemma Heath 2 ,
• Elaine Cameron 3 ,
• Sabina Rashid 4 &
• Sabi Redwood 2  

BMC Medical Research Methodology volume  13 , Article number:  117 ( 2013 ) Cite this article

515k Accesses

4724 Citations

82 Altmetric

Metrics details

The Framework Method is becoming an increasingly popular approach to the management and analysis of qualitative data in health research. However, there is confusion about its potential application and limitations.

The article discusses when it is appropriate to adopt the Framework Method and explains the procedure for using it in multi-disciplinary health research teams, or those that involve clinicians, patients and lay people. The stages of the method are illustrated using examples from a published study.

Used effectively, with the leadership of an experienced qualitative researcher, the Framework Method is a systematic and flexible approach to analysing qualitative data and is appropriate for use in research teams even where not all members have previous experience of conducting qualitative research.

The Framework Method for the management and analysis of qualitative data has been used since the 1980s [ 1 ]. The method originated in large-scale social policy research but is becoming an increasingly popular approach in medical and health research; however, there is some confusion about its potential application and limitations. In this article we discuss when it is appropriate to use the Framework Method and how it compares to other qualitative analysis methods. In particular, we explore how it can be used in multi-disciplinary health research teams. Multi-disciplinary and mixed methods studies are becoming increasingly commonplace in applied health research. As well as disciplines familiar with qualitative research, such as nursing, psychology and sociology, teams often include epidemiologists, health economists, management scientists and others. Furthermore, applied health research often has clinical representation and, increasingly, patient and public involvement [ 2 ]. We argue that while leadership is undoubtedly required from an experienced qualitative methodologist, non-specialists from the wider team can and should be involved in the analysis process. We then present a step-by-step guide to the application of the Framework Method, illustrated using a worked example (See Additional File 1 ) from a published study [ 3 ] to illustrate the main stages of the process. Technical terms are included in the glossary (below). Finally, we discuss the strengths and limitations of the approach.

Glossary of key terms used in the Framework Method

Analytical framework: A set of codes organised into categories that have been jointly developed by researchers involved in analysis that can be used to manage and organise the data. The framework creates a new structure for the data (rather than the full original accounts given by participants) that is helpful to summarize/reduce the data in a way that can support answering the research questions.

Analytic memo: A written investigation of a particular concept, theme or problem, reflecting on emerging issues in the data that captures the analytic process (see Additional file 1 , Section 7).

Categories: During the analysis process, codes are grouped into clusters around similar and interrelated ideas or concepts. Categories and codes are usually arranged in a tree diagram structure in the analytical framework. While categories are closely and explicitly linked to the raw data, developing categories is a way to start the process of abstraction of the data (i.e. towards the general rather than the specific or anecdotal).

Charting: Entering summarized data into the Framework Method matrix (see Additional File 1 , Section 6).

Code: A descriptive or conceptual label that is assigned to excerpts of raw data in a process called ‘coding’ (see Additional File 1 , Section 3).

Data: Qualitative data usually needs to be in textual form before analysis. These texts can either be elicited texts (written specifically for the research, such as food diaries), or extant texts (pre-existing texts, such as meeting minutes, policy documents or weblogs), or can be produced by transcribing interview or focus group data, or creating ‘field’ notes while conducting participant-observation or observing objects or social situations.

Indexing: The systematic application of codes from the agreed analytical framework to the whole dataset (see Additional File 1 , Section 5).

Matrix: A spreadsheet contains numerous cells into which summarized data are entered by codes (columns) and cases (rows) (see Additional File 1 , Section 6).

Themes: Interpretive concepts or propositions that describe or explain aspects of the data, which are the final output of the analysis of the whole dataset. Themes are articulated and developed by interrogating data categories through comparison between and within cases. Usually a number of categories would fall under each theme or sub-theme [ 3 ].

Transcript: A written verbatim (word-for-word) account of a verbal interaction, such as an interview or conversation.

The Framework Method sits within a broad family of analysis methods often termed thematic analysis or qualitative content analysis. These approaches identify commonalities and differences in qualitative data, before focusing on relationships between different parts of the data, thereby seeking to draw descriptive and/or explanatory conclusions clustered around themes. The Framework Method was developed by researchers, Jane Ritchie and Liz Spencer, from the Qualitative Research Unit at the National Centre for Social Research in the United Kingdom in the late 1980s for use in large-scale policy research [ 1 ]. It is now used widely in other areas, including health research [ 3 – 12 ]. Its defining feature is the matrix output: rows (cases), columns (codes) and ‘cells’ of summarised data, providing a structure into which the researcher can systematically reduce the data, in order to analyse it by case and by code [ 1 ]. Most often a ‘case’ is an individual interviewee, but this can be adapted to other units of analysis, such as predefined groups or organisations. While in-depth analyses of key themes can take place across the whole data set, the views of each research participant remain connected to other aspects of their account within the matrix so that the context of the individual’s views is not lost. Comparing and contrasting data is vital to qualitative analysis and the ability to compare with ease data across cases as well as within individual cases is built into the structure and process of the Framework Method.

The Framework Method provides clear steps to follow and produces highly structured outputs of summarised data. It is therefore useful where multiple researchers are working on a project, particularly in multi-disciplinary research teams were not all members have experience of qualitative data analysis, and for managing large data sets where obtaining a holistic, descriptive overview of the entire data set is desirable. However, caution is recommended before selecting the method as it is not a suitable tool for analysing all types of qualitative data or for answering all qualitative research questions, nor is it an ‘easy’ version of qualitative research for quantitative researchers. Importantly, the Framework Method cannot accommodate highly heterogeneous data, i.e. data must cover similar topics or key issues so that it is possible to categorize it. Individual interviewees may, of course, have very different views or experiences in relation to each topic, which can then be compared and contrasted. The Framework Method is most commonly used for the thematic analysis of semi-structured interview transcripts, which is what we focus on in this article, although it could, in principle, be adapted for other types of textual data [ 13 ], including documents, such as meeting minutes or diaries [ 12 ], or field notes from observations [ 10 ].

For quantitative researchers working with qualitative colleagues or when exploring qualitative research for the first time, the nature of the Framework Method is seductive because its methodical processes and ‘spreadsheet’ approach seem more closely aligned to the quantitative paradigm [ 14 ]. Although the Framework Method is a highly systematic method of categorizing and organizing what may seem like unwieldy qualitative data, it is not a panacea for problematic issues commonly associated with qualitative data analysis such as how to make analytic choices and make interpretive strategies visible and auditable. Qualitative research skills are required to appropriately interpret the matrix, and facilitate the generation of descriptions, categories, explanations and typologies. Moreover, reflexivity, rigour and quality are issues that are requisite in the Framework Method just as they are in other qualitative methods. It is therefore essential that studies using the Framework Method for analysis are overseen by an experienced qualitative researcher, though this does not preclude those new to qualitative research from contributing to the analysis as part of a wider research team.

There are a number of approaches to qualitative data analysis, including those that pay close attention to language and how it is being used in social interaction such as discourse analysis [ 15 ] and ethnomethodology [ 16 ]; those that are concerned with experience, meaning and language such as phenomenology [ 17 , 18 ] and narrative methods [ 19 ]; and those that seek to develop theory derived from data through a set of procedures and interconnected stages such as Grounded Theory [ 20 , 21 ]. Many of these approaches are associated with specific disciplines and are underpinned by philosophical ideas which shape the process of analysis [ 22 ]. The Framework Method, however, is not aligned with a particular epistemological, philosophical, or theoretical approach. Rather it is a flexible tool that can be adapted for use with many qualitative approaches that aim to generate themes.

The development of themes is a common feature of qualitative data analysis, involving the systematic search for patterns to generate full descriptions capable of shedding light on the phenomenon under investigation. In particular, many qualitative approaches use the ‘constant comparative method’ , developed as part of Grounded Theory, which involves making systematic comparisons across cases to refine each theme [ 21 , 23 ]. Unlike Grounded Theory, the Framework Method is not necessarily concerned with generating social theory, but can greatly facilitate constant comparative techniques through the review of data across the matrix.

Perhaps because the Framework Method is so obviously systematic, it has often, as other commentators have noted, been conflated with a deductive approach to qualitative analysis [ 13 , 14 ]. However, the tool itself has no allegiance to either inductive or deductive thematic analysis; where the research sits along this inductive-deductive continuum depends on the research question. A question such as, ‘Can patients give an accurate biomedical account of the onset of their cardiovascular disease?’ is essentially a yes/no question (although it may be nuanced by the extent of their account or by appropriate use of terminology) and so requires a deductive approach to both data collection and analysis (e.g. structured or semi-structured interviews and directed qualitative content analysis [ 24 ]). Similarly, a deductive approach may be taken if basing analysis on a pre-existing theory, such as behaviour change theories, for example in the case of a research question such as ‘How does the Theory of Planned Behaviour help explain GP prescribing?’ [ 11 ]. However, a research question such as, ‘How do people construct accounts of the onset of their cardiovascular disease?’ would require a more inductive approach that allows for the unexpected, and permits more socially-located responses [ 25 ] from interviewees that may include matters of cultural beliefs, habits of food preparation, concepts of ‘fate’, or links to other important events in their lives, such as grief, which cannot be predicted by the researcher in advance (e.g. an interviewee-led open ended interview and grounded theory [ 20 ]). In all these cases, it may be appropriate to use the Framework Method to manage the data. The difference would become apparent in how themes are selected: in the deductive approach, themes and codes are pre-selected based on previous literature, previous theories or the specifics of the research question; whereas in the inductive approach, themes are generated from the data though open (unrestricted) coding, followed by refinement of themes. In many cases, a combined approach is appropriate when the project has some specific issues to explore, but also aims to leave space to discover other unexpected aspects of the participants’ experience or the way they assign meaning to phenomena. In sum, the Framework Method can be adapted for use with deductive, inductive, or combined types of qualitative analysis. However, there are some research questions where analysing data by case and theme is not appropriate and so the Framework Method should be avoided. For instance, depending on the research question, life history data might be better analysed using narrative analysis [ 19 ]; recorded consultations between patients and their healthcare practitioners using conversation analysis [ 26 ]; and documentary data, such as resources for pregnant women, using discourse analysis [ 27 ].

It is not within the scope of this paper to consider study design or data collection in any depth, but before moving on to describe the Framework Method analysis process, it is worth taking a step back to consider briefly what needs to happen before analysis begins. The selection of analysis method should have been considered at the proposal stage of the research and should fit with the research questions and overall aims of the study. Many qualitative studies, particularly ones using inductive analysis, are emergent in nature; this can be a challenge and the researchers can only provide an “imaginative rehearsal” of what is to come [ 28 ]. In mixed methods studies, the role of the qualitative component within the wider goals of the project must also be considered. In the data collection stage, resources must be allocated for properly trained researchers to conduct the qualitative interviewing because it is a highly skilled activity. In some cases, a research team may decide that they would like to use lay people, patients or peers to do the interviews [ 29 – 32 ] and in this case they must be properly trained and mentored which requires time and resources. At this early stage it is also useful to consider whether the team will use Computer Assisted Qualitative Data Analysis Software (CAQDAS), which can assist with data management and analysis.

As any form of qualitative or quantitative analysis is not a purely technical process, but influenced by the characteristics of the researchers and their disciplinary paradigms, critical reflection throughout the research process is paramount, including in the design of the study, the construction or collection of data, and the analysis. All members of the team should keep a research diary, where they record reflexive notes, impressions of the data and thoughts about analysis throughout the process. Experienced qualitative researchers become more skilled at sifting through data and analysing it in a rigorous and reflexive way. They cannot be too attached to certainty, but must remain flexible and adaptive throughout the research in order to generate rich and nuanced findings that embrace and explain the complexity of real social life and can be applied to complex social issues. It is important to remember when using the Framework Method that, unlike quantitative research where data collection and data analysis are strictly sequential and mutually exclusive stages of the research process, in qualitative analysis there is, to a greater or lesser extent depending on the project, ongoing interplay between data collection, analysis, and theory development. For example, new ideas or insights from participants may suggest potentially fruitful lines of enquiry, or close analysis might reveal subtle inconsistencies in an account which require further exploration.

Procedure for analysis

Stage 1: transcription.

A good quality audio recording and, ideally, a verbatim (word for word) transcription of the interview is needed. For Framework Method analysis, it is not necessarily important to include the conventions of dialogue transcriptions which can be difficult to read (e.g. pauses or two people talking simultaneously), because the content is what is of primary interest. Transcripts should have large margins and adequate line spacing for later coding and making notes. The process of transcription is a good opportunity to become immersed in the data and is to be strongly encouraged for new researchers. However, in some projects, the decision may be made that it is a better use of resources to outsource this task to a professional transcriber.

Stage 2: Familiarisation with the interview

Becoming familiar with the whole interview using the audio recording and/or transcript and any contextual or reflective notes that were recorded by the interviewer is a vital stage in interpretation. It can also be helpful to re-listen to all or parts of the audio recording. In multi-disciplinary or large research projects, those involved in analysing the data may be different from those who conducted or transcribed the interviews, which makes this stage particularly important. One margin can be used to record any analytical notes, thoughts or impressions.

Stage 3: Coding

After familiarization, the researcher carefully reads the transcript line by line, applying a paraphrase or label (a ‘code’) that describes what they have interpreted in the passage as important. In more inductive studies, at this stage ‘open coding’ takes place, i.e. coding anything that might be relevant from as many different perspectives as possible. Codes could refer to substantive things (e.g. particular behaviours, incidents or structures), values (e.g. those that inform or underpin certain statements, such as a belief in evidence-based medicine or in patient choice), emotions (e.g. sorrow, frustration, love) and more impressionistic/methodological elements (e.g. interviewee found something difficult to explain, interviewee became emotional, interviewer felt uncomfortable) [ 33 ]. In purely deductive studies, the codes may have been pre-defined (e.g. by an existing theory, or specific areas of interest to the project) so this stage may not be strictly necessary and you could just move straight onto indexing, although it is generally helpful even if you are taking a broadly deductive approach to do some open coding on at least a few of the transcripts to ensure important aspects of the data are not missed. Coding aims to classify all of the data so that it can be compared systematically with other parts of the data set. At least two researchers (or at least one from each discipline or speciality in a multi-disciplinary research team) should independently code the first few transcripts, if feasible. Patients, public involvement representatives or clinicians can also be productively involved at this stage, because they can offer alternative viewpoints thus ensuring that one particular perspective does not dominate. It is vital in inductive coding to look out for the unexpected and not to just code in a literal, descriptive way so the involvement of people from different perspectives can aid greatly in this. As well as getting a holistic impression of what was said, coding line-by-line can often alert the researcher to consider that which may ordinarily remain invisible because it is not clearly expressed or does not ‘fit’ with the rest of the account. In this way the developing analysis is challenged; to reconcile and explain anomalies in the data can make the analysis stronger. Coding can also be done digitally using CAQDAS, which is a useful way to keep track automatically of new codes. However, some researchers prefer to do the early stages of coding with a paper and pen, and only start to use CAQDAS once they reach Stage 5 (see below).

Stage 4: Developing a working analytical framework

After coding the first few transcripts, all researchers involved should meet to compare the labels they have applied and agree on a set of codes to apply to all subsequent transcripts. Codes can be grouped together into categories (using a tree diagram if helpful), which are then clearly defined. This forms a working analytical framework. It is likely that several iterations of the analytical framework will be required before no additional codes emerge. It is always worth having an ‘other’ code under each category to avoid ignoring data that does not fit; the analytical framework is never ‘final’ until the last transcript has been coded.

Stage 5: Applying the analytical framework

The working analytical framework is then applied by indexing subsequent transcripts using the existing categories and codes. Each code is usually assigned a number or abbreviation for easy identification (and so the full names of the codes do not have to be written out each time) and written directly onto the transcripts. Computer Assisted Qualitative Data Analysis Software (CAQDAS) is particularly useful at this stage because it can speed up the process and ensures that, at later stages, data is easily retrievable. It is worth noting that unlike software for statistical analyses, which actually carries out the calculations with the correct instruction, putting the data into a qualitative analysis software package does not analyse the data; it is simply an effective way of storing and organising the data so that they are accessible for the analysis process.

Stage 6: Charting data into the framework matrix

Qualitative data are voluminous (an hour of interview can generate 15–30 pages of text) and being able to manage and summarize (reduce) data is a vital aspect of the analysis process. A spreadsheet is used to generate a matrix and the data are ‘charted’ into the matrix. Charting involves summarizing the data by category from each transcript. Good charting requires an ability to strike a balance between reducing the data on the one hand and retaining the original meanings and ‘feel’ of the interviewees’ words on the other. The chart should include references to interesting or illustrative quotations. These can be tagged automatically if you are using CAQDAS to manage your data (N-Vivo version 9 onwards has the capability to generate framework matrices), or otherwise a capital ‘Q’, an (anonymized) transcript number, page and line reference will suffice. It is helpful in multi-disciplinary teams to compare and contrast styles of summarizing in the early stages of the analysis process to ensure consistency within the team. Any abbreviations used should be agreed by the team. Once members of the team are familiar with the analytical framework and well practised at coding and charting, on average, it will take about half a day per hour-long transcript to reach this stage. In the early stages, it takes much longer.

Stage 7: Interpreting the data

It is useful throughout the research to have a separate note book or computer file to note down impressions, ideas and early interpretations of the data. It may be worth breaking off at any stage to explore an interesting idea, concept or potential theme by writing an analytic memo [ 20 , 21 ] to then discuss with other members of the research team, including lay and clinical members. Gradually, characteristics of and differences between the data are identified, perhaps generating typologies, interrogating theoretical concepts (either prior concepts or ones emerging from the data) or mapping connections between categories to explore relationships and/or causality. If the data are rich enough, the findings generated through this process can go beyond description of particular cases to explanation of, for example, reasons for the emergence of a phenomena, predicting how an organisation or other social actor is likely to instigate or respond to a situation, or identifying areas that are not functioning well within an organisation or system. It is worth noting that this stage often takes longer than anticipated and that any project plan should ensure that sufficient time is allocated to meetings and individual researcher time to conduct interpretation and writing up of findings (see Additional file 1 , Section 7).

The Framework Method has been developed and used successfully in research for over 25 years, and has recently become a popular analysis method in qualitative health research. The issue of how to assess quality in qualitative research has been highly debated [ 20 , 34 – 40 ], but ensuring rigour and transparency in analysis is a vital component. There are, of course, many ways to do this but in the Framework Method the following are helpful:

Summarizing the data during charting, as well as being a practical way to reduce the data, means that all members of a multi-disciplinary team, including lay, clinical and (quantitative) academic members can engage with the data and offer their perspectives during the analysis process without necessarily needing to read all the transcripts or be involved in the more technical parts of analysis.

Charting also ensures that researchers pay close attention to describing the data using each participant’s own subjective frames and expressions in the first instance, before moving onto interpretation.

The summarized data is kept within the wider context of each case, thereby encouraging thick description that pays attention to complex layers of meaning and understanding [ 38 ].

The matrix structure is visually straightforward and can facilitate recognition of patterns in the data by any member of the research team, including through drawing attention to contradictory data, deviant cases or empty cells.

The systematic procedure (described in this article) makes it easy to follow, even for multi-disciplinary teams and/or with large data sets.

It is flexible enough that non-interview data (such as field notes taken during the interview or reflexive considerations) can be included in the matrix.

It is not aligned with a particular epistemological viewpoint or theoretical approach and therefore can be adapted for use in inductive or deductive analysis or a combination of the two (e.g. using pre-existing theoretical constructs deductively, then revising the theory with inductive aspects; or using an inductive approach to identify themes in the data, before returning to the literature and using theories deductively to help further explain certain themes).

It is easy to identify relevant data extracts to illustrate themes and to check whether there is sufficient evidence for a proposed theme.

Finally, there is a clear audit trail from original raw data to final themes, including the illustrative quotes.

There are also a number of potential pitfalls to this approach:

The systematic approach and matrix format, as we noted in the background, is intuitively appealing to those trained quantitatively but the ‘spreadsheet’ look perhaps further increases the temptation for those without an in-depth understanding of qualitative research to attempt to quantify qualitative data (e.g. “13 out of 20 participants said X). This kind of statement is clearly meaningless because the sampling in qualitative research is not designed to be representative of a wider population, but purposive to capture diversity around a phenomenon [ 41 ].

Like all qualitative analysis methods, the Framework Method is time consuming and resource-intensive. When involving multiple stakeholders and disciplines in the analysis and interpretation of the data, as is good practice in applied health research, the time needed is extended. This time needs to be factored into the project proposal at the pre-funding stage.

There is a high training component to successfully using the method in a new multi-disciplinary team. Depending on their role in the analysis, members of the research team may have to learn how to code, index, and chart data, to think reflexively about how their identities and experience affect the analysis process, and/or they may have to learn about the methods of generalisation (i.e. analytic generalisation and transferability, rather than statistical generalisation [ 41 ]) to help to interpret legitimately the meaning and significance of the data.

While the Framework Method is amenable to the participation of non-experts in data analysis, it is critical to the successful use of the method that an experienced qualitative researcher leads the project (even if the overall lead for a large mixed methods study is a different person). The qualitative lead would ideally be joined by other researchers with at least some prior training in or experience of qualitative analysis. The responsibilities of the lead qualitative researcher are: to contribute to study design, project timelines and resource planning; to mentor junior qualitative researchers; to train clinical, lay and other (non-qualitative) academics to contribute as appropriate to the analysis process; to facilitate analysis meetings in a way that encourages critical and reflexive engagement with the data and other team members; and finally to lead the write-up of the study.

We have argued that Framework Method studies can be conducted by multi-disciplinary research teams that include, for example, healthcare professionals, psychologists, sociologists, economists, and lay people/service users. The inclusion of so many different perspectives means that decision-making in the analysis process can be very time consuming and resource-intensive. It may require extensive, reflexive and critical dialogue about how the ideas expressed by interviewees and identified in the transcript are related to pre-existing concepts and theories from each discipline, and to the real ‘problems’ in the health system that the project is addressing. This kind of team effort is, however, an excellent forum for driving forward interdisciplinary collaboration, as well as clinical and lay involvement in research, to ensure that ‘the whole is greater than the sum of the parts’, by enhancing the credibility and relevance of the findings.

The Framework Method is appropriate for thematic analysis of textual data, particularly interview transcripts, where it is important to be able to compare and contrast data by themes across many cases, while also situating each perspective in context by retaining the connection to other aspects of each individual’s account. Experienced qualitative researchers should lead and facilitate all aspects of the analysis, although the Framework Method’s systematic approach makes it suitable for involving all members of a multi-disciplinary team. An open, critical and reflexive approach from all team members is essential for rigorous qualitative analysis.

Acceptance of the complexity of real life health systems and the existence of multiple perspectives on health issues is necessary to produce high quality qualitative research. If done well, qualitative studies can shed explanatory and predictive light on important phenomena, relate constructively to quantitative parts of a larger study, and contribute to the improvement of health services and development of health policy. The Framework Method, when selected and implemented appropriately, can be a suitable tool for achieving these aims through producing credible and relevant findings.

The Framework Method is an excellent tool for supporting thematic (qualitative content) analysis because it provides a systematic model for managing and mapping the data.

The Framework Method is most suitable for analysis of interview data, where it is desirable to generate themes by making comparisons within and between cases.

The management of large data sets is facilitated by the Framework Method as its matrix form provides an intuitively structured overview of summarised data.

The clear, step-by-step process of the Framework Method makes it is suitable for interdisciplinary and collaborative projects.

The use of the method should be led and facilitated by an experienced qualitative researcher.

Ritchie J, Lewis J: Qualitative research practice: a guide for social science students and researchers. 2003, London: Sage

Google Scholar  

Ives J, Damery S, Redwod S: PPI, paradoxes and Plato: who's sailing the ship?. J Med Ethics. 2013, 39 (3): 181-185. 10.1136/medethics-2011-100150.

Article   Google Scholar  

Heath G, Cameron E, Cummins C, Greenfield S, Pattison H, Kelly D, Redwood S: Paediatric ‘care closer to home’: stake-holder views and barriers to implementation. Health Place. 2012, 18 (5): 1068-1073. 10.1016/j.healthplace.2012.05.003.

Elkington H, White P, Addington-Hall J, Higgs R, Petternari C: The last year of life of COPD: a qualitative study of symptoms and services. Respir Med. 2004, 98 (5): 439-445. 10.1016/j.rmed.2003.11.006.

Murtagh J, Dixey R, Rudolf M: A qualitative investigation into the levers and barriers to weight loss in children: opinions of obese children. Archives Dis Child. 2006, 91 (11): 920-923. 10.1136/adc.2005.085712.

Barnard M, Webster S, O’Connor W, Jones A, Donmall M: The drug treatment outcomes research study (DTORS): qualitative study. 2009, London: Home Office

Ayatollahi H, Bath PA, Goodacre S: Factors influencing the use of IT in the emergency department: a qualitative study. Health Inform J. 2010, 16 (3): 189-200. 10.1177/1460458210377480.

Sheard L, Prout H, Dowding D, Noble S, Watt I, Maraveyas A, Johnson M: Barriers to the diagnosis and treatment of venous thromboembolism in advanced cancer patients: a qualitative study. Palliative Med. 2012, 27 (2): 339-348.

Ellis J, Wagland R, Tishelman C, Williams ML, Bailey CD, Haines J, Caress A, Lorigan P, Smith JA, Booton R, et al: Considerations in developing and delivering a nonpharmacological intervention for symptom management in lung cancer: the views of patients and informal caregivers. J Pain Symptom Manag (0). 2012, 44 (6): 831-842. 10.1016/j.jpainsymman.2011.12.274.

Gale N, Sultan H: Telehealth as ‘peace of mind’: embodiment, emotions and the home as the primary health space for people with chronic obstructive pulmonary disorder. Health place. 2013, 21: 140-147.

Rashidian A, Eccles MP, Russell I: Falling on stony ground? A qualitative study of implementation of clinical guidelines’ prescribing recommendations in primary care. Health policy. 2008, 85 (2): 148-161. 10.1016/j.healthpol.2007.07.011.

Jones RK: The unsolicited diary as a qualitative research tool for advanced research capacity in the field of health and illness. Qualitative Health Res. 2000, 10 (4): 555-567. 10.1177/104973200129118543.

Pope C, Ziebland S, Mays N: Analysing qualitative data. British Med J. 2000, 320: 114-116. 10.1136/bmj.320.7227.114.

Pope C, Mays N: Critical reflections on the rise of qualitative research. British Med J. 2009, 339: 737-739.

Fairclough N: Critical discourse analysis: the critical study of language. 2010, London: Longman

Garfinkel H: Ethnomethodology’s program. Soc Psychol Quarter. 1996, 59 (1): 5-21. 10.2307/2787116.

Merleau-Ponty M: The phenomenology of perception. 1962, London: Routledge and Kegan Paul

Svenaeus F: The phenomenology of health and illness. Handbook of phenomenology and medicine. 2001, Netherlands: Springer, 87-108.

Reissmann CK: Narrative methods for the human sciences. 2008, London: Sage

Charmaz K: Constructing grounded theory: a practical guide through qualitative analysis. 2006, London: Sage

Glaser A, Strauss AL: The discovery of grounded theory. 1967, Chicago: Aldine

Crotty M: The foundations of social research: meaning and perspective in the research process. 1998, London: Sage

Boeije H: A purposeful approach to the constant comparative method in the analysis of qualitative interviews. Qual Quant. 2002, 36 (4): 391-409. 10.1023/A:1020909529486.

Hsieh H-F, Shannon SE: Three approaches to qualitative content analysis. Qual Health Res. 2005, 15 (9): 1277-1288. 10.1177/1049732305276687.

Redwood S, Gale NK, Greenfield S: ‘You give us rangoli, we give you talk’: using an art-based activity to elicit data from a seldom heard group. BMC Med Res Methodol. 2012, 12 (1): 7-10.1186/1471-2288-12-7.

Mishler EG: The struggle between the voice of medicine and the voice of the lifeworld. The sociology of health and illness: critical perspectives. Edited by: Conrad P, Kern R. 1990, New York: St Martins Press, Third

Hodges BD, Kuper A, Reeves S: Discourse analysis. British Med J. 2008, 337: 570-572. 10.1136/bmj.39370.701782.DE.

Sandelowski M, Barroso J: Writing the proposal for a qualitative research methodology project. Qual Health Res. 2003, 13 (6): 781-820. 10.1177/1049732303013006003.

Ellins J: It’s better together: involving older people in research. HSMC Newsletter Focus Serv Users Publ. 2010, 16 (1): 4-

Phillimore J, Goodson L, Hennessy D, Ergun E: Empowering Birmingham’s migrant and refugee community organisations: making a difference. 2009, York: Joseph Rowntree Foundation

Leamy M, Clough R: How older people became researchers. 2006, York: Joseph Rowntree Foundation

Glasby J, Miller R, Ellins J, Durose J, Davidson D, McIver S, Littlechild R, Tanner D, Snelling I, Spence K: Final report NIHR service delivery and organisation programme. Understanding and improving transitions of older people: a user and carer centred approach. 2012, London: The Stationery Office

Saldaña J: The coding manual for qualitative researchers. 2009, London: Sage

Lincoln YS: Emerging criteria for quality in qualitative and interpretive research. Qual Inquiry. 1995, 1 (3): 275-289. 10.1177/107780049500100301.

Mays N, Pope C: Qualitative research in health care: assessing quality in qualitative research. BMJ British Med J. 2000, 320 (7226): 50-10.1136/bmj.320.7226.50.

Seale C: Quality in qualitative research. Qual Inquiry. 1999, 5 (4): 465-478. 10.1177/107780049900500402.

Dingwall R, Murphy E, Watson P, Greatbatch D, Parker S: Catching goldfish: quality in qualitative research. J Health serv Res Policy. 1998, 3 (3): 167-172.

Popay J, Rogers A, Williams G: Rationale and standards for the systematic review of qualitative literature in health services research. Qual Health Res. 1998, 8 (3): 341-351. 10.1177/104973239800800305.

Morse JM, Barrett M, Mayan M, Olson K, Spiers J: Verification strategies for establishing reliability and validity in qualitative research. Int J Qual Methods. 2008, 1 (2): 13-22.

Smith JA: Reflecting on the development of interpretative phenomenological analysis and its contribution to qualitative research in psychology. Qual Res Psychol. 2004, 1 (1): 39-54.

Polit DF, Beck CT: Generalization in quantitative and qualitative research: Myths and strategies. Int J Nurs Studies. 2010, 47 (11): 1451-1458. 10.1016/j.ijnurstu.2010.06.004.

Pre-publication history

The pre-publication history for this paper can be accessed here: http://www.biomedcentral.com/1471-2288/13/117/prepub

Download references

Acknowledgments

All authors were funded by the National Institute for Health Research (NIHR) through the Collaborations for Leadership in Applied Health Research and Care for Birmingham and Black Country (CLAHRC-BBC) programme. The views in this publication expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.

Author information

Authors and affiliations.

Health Services Management Centre, University of Birmingham, Park House, 40 Edgbaston Park Road, Birmingham, B15 2RT, UK

Nicola K Gale

School of Health and Population Sciences, University of Birmingham, Edgbaston, Birmingham, B15 2TT, UK

Gemma Heath & Sabi Redwood

School of Life and Health Sciences, Aston University, Aston Triangle, Birmingham, B4 7ET, UK

Elaine Cameron

East and North Hertfordshire NHS Trust, Lister hospital, Coreys Mill Lane, Stevenage, SG1 4AB, UK

Sabina Rashid

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Nicola K Gale .

Additional information

Competing interests.

The authors declare that they have no competing interests.

Authors’ contributions

All authors were involved in the development of the concept of the article and drafting the article. NG wrote the first draft of the article, GH and EC prepared the text and figures related to the illustrative example, SRa did the literature search to identify if there were any similar articles currently available and contributed to drafting of the article, and SRe contributed to drafting of the article and the illustrative example. All authors read and approved the final manuscript.

Electronic supplementary material

Additional file 1: illustrative example of the use of the framework method.(docx 167 kb), authors’ original submitted files for images.

Below are the links to the authors’ original submitted files for images.

Authors’ original file for figure 1

Authors’ original file for figure 2, authors’ original file for figure 3, authors’ original file for figure 4, authors’ original file for figure 5, authors’ original file for figure 6, rights and permissions.

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( http://creativecommons.org/licenses/by/2.0 ), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and permissions

About this article

Cite this article.

Gale, N.K., Heath, G., Cameron, E. et al. Using the framework method for the analysis of qualitative data in multi-disciplinary health research. BMC Med Res Methodol 13 , 117 (2013). https://doi.org/10.1186/1471-2288-13-117

Download citation

Received : 17 December 2012

Accepted : 06 September 2013

Published : 18 September 2013

DOI : https://doi.org/10.1186/1471-2288-13-117

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Qualitative research
• Qualitative content analysis
• Multi-disciplinary research

BMC Medical Research Methodology

ISSN: 1471-2288

• General enquiries: [email protected]

Research Writing and Analysis

• NVivo Group and Study Sessions
• SPSS This link opens in a new window
• Statistical Analysis Group sessions
• Using Qualtrics
• Dissertation and Data Analysis Group Sessions
• Defense Schedule - Commons Calendar This link opens in a new window
• Research Process Flow Chart
• Research Alignment Chapter 1 This link opens in a new window
• Step 1: Seek Out Evidence
• Step 2: Explain
• Step 3: The Big Picture
• Step 4: Own It
• Step 5: Illustrate
• Annotated Bibliography
• Literature Review This link opens in a new window
• Systematic Reviews & Meta-Analyses
• How to Synthesize and Analyze
• Synthesis and Analysis Practice
• Synthesis and Analysis Group Sessions
• Problem Statement
• Purpose Statement
• Conceptual Framework
• Theoretical Framework
• Quantitative Research Questions
• Qualitative Research Questions
• Trustworthiness of Qualitative Data
• Analysis and Coding Example- Qualitative Data
• Thematic Data Analysis in Qualitative Design
• Dissertation to Journal Article This link opens in a new window
• International Journal of Online Graduate Education (IJOGE) This link opens in a new window
• Journal of Research in Innovative Teaching & Learning (JRIT&L) This link opens in a new window

Designing the Theoretical Framework

What is it.

• A foundational review of existing theories. 
• Serves as a roadmap or blueprint for developing arguments and supporting research.
• Overview of the theory that the research is based on.
• Can be made up of theories, principles, and concepts.

What does it do?

• Explains the why and how of a particular phenomenon within a particular body of literature.
• Connects the research subject with the theory.
• Specifies the study’s scope; makes it more valuable and generalizable.
• Guides further actions like framing the research questions, developing the literature review, and data collection and analyses.

What should be in it?

• Theory or theories that the researcher considers relevant for their research, principles, and concepts.

Theoretical Framework Guide

• Theoretical Framework Guide Use this guide to determine the guiding framework for your theoretical dissertation research.

Making a Theoretical Framework

How to make a theoretical framework.

• Specify research objectives.
• Note the prominent variables under the study.
• Explore and review the literature through keywords identified as prominent variables.
• Note the theories that contain these variables or the keywords.
• Review all selected theories again in the light of the study’s objectives, and the key variables identified.
• Search for alternative theoretical propositions in the literature that may challenge the ones already selected.
• Ensure that the framework aligns with the study’s objectives, problem statement, the main research question, methodology, data analysis, and the expected conclusion.
• Decide on the final framework and begin developing.

Example Framework

• Theoretical Framework Example for a Thesis or Dissertation This link offers an example theoretical framework.

Additional Framework Resources

Some additional helpful resources in constructing a theoretical framework for study:.

• https://www.scribbr.com/dissertation/theoretical-framework/
• https://www.scribbr.com/dissertation/theoretical-framework-example/
• https://www.projectguru.in/how-to-write-the-theoretical-framework-of-research/

Theoretical Framework Research

The term conceptual framework and theoretical framework are often and erroneously used interchangeably (Grant & Osanloo, 2014). A theoretical framework provides the theoretical assumptions for the larger context of a study, and is the foundation or ‘lens’ by which a study is developed. This framework helps to ground the research focus understudy within theoretical underpinnings and to frame the inquiry for data analysis and interpretation.  The application of theory in traditional theoretical research is to understand, explain, and predict phenomena (Swanson, 2013).

Casanave, C.P.,& Li,Y.(2015). Novices’ struggles with conceptual and theoretical framing in writing  dissertations and papers for publication. Publications,3 (2),104-119.doi:10.3390/publications3020104

Grant, C., & Osanloo, A. (2014). Understanding, Selecting, and Integrating a Theoretical Framework in Dissertation Research: Creating the Blueprint for Your “House. ” Administrative Issues Journal: Connecting Education, Practice, and Research, 4(2), 12–26

Swanson, R. (2013). Theory building in applied disciplines . San Francisco: Berrett-Koehler Publishers.

• << Previous: Conceptual Framework
• Next: Quantitative Research Questions >>
• Last Updated: Apr 24, 2024 11:06 AM
• URL: https://resources.nu.edu/researchtools

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 25 April 2024

A framework for the analysis of historical newsreels

• Mila Oiva   ORCID: orcid.org/0000-0002-5241-7436 1 ,
• Ksenia Mukhina 1 ,
• Vejune Zemaityte   ORCID: orcid.org/0000-0001-9714-7903 1 ,
• Andres Karjus   ORCID: orcid.org/0000-0002-2445-5072 1 , 2 ,
• Mikhail Tamm 1 ,
• Tillmann Ohm 1 ,
• Mark Mets 1 ,
• Daniel Chávez Heras   ORCID: orcid.org/0000-0002-9877-7496 3 ,
• Mar Canet Sola   ORCID: orcid.org/0000-0001-5986-3239 1 ,
• Helena Hanna Juht 4 &
• Maximilian Schich 1  

Humanities and Social Sciences Communications volume  11 , Article number:  530 ( 2024 ) Cite this article

55 Accesses

1 Altmetric

Metrics details

• Cultural and media studies

Audiovisual news is a critical cultural phenomenon that has been influencing audience worldviews for more than a hundred years. To understand historical trends in multimodal audiovisual news, we need to explore them longitudinally using large sets of data. Despite promising developments in film history, computational video analysis, and other relevant fields, current research streams have limitations related to the scope of data used, the systematism of analysis, and the modalities and elements to be studied in audiovisual material and its metadata. Simultaneously, each disciplinary approach contributes significant input to research reducing these limitations. We therefore advocate for combining the strengths of several disciplines. Here we propose a multidisciplinary framework for systematically studying large collections of historical audiovisual news to gain a coherent picture of their temporal dynamics, cultural diversity, and potential societal effects across several quantitative and qualitative dimensions of analysis. By using newsreels as an example of such complex historically formed data, we combine the context crucial to qualitative approaches with the systematicity and ability to cover large amounts of data from quantitative methods. The framework template for historical newsreels is exemplified by a case study of the “News of the Day” newsreel series produced in the Soviet Union during 1944–1992. The proposed framework enables a more nuanced analysis of longitudinal collections of audiovisual news, expanding our understanding of the dynamics of global knowledge cultures.

Similar content being viewed by others

Modeling narrative features in TV series: coding and clustering analysis

A corpus-based approach to the reception of Chinese television dramas streamed overseas

LocalView, a database of public meetings for the study of local politics and policy-making in the United States

Introduction.

Audiovisual news has affected the global knowledge landscape for over a century. As a media format, audiovisual is impactful, and news as a genre is particularly effective for forming knowledge about the world. Although what counts as news is debatable (Tworek, 2019 ), labelling a story as ‘news’ suggests that the offered content has contemporary relevance and that it provides truthful information on the surrounding world—even if we know that this is not always the case (Winston, 2018 ; Lazer et al., 2018 ). Understanding audiovisual news content is crucial because news, taking part in creating the ‘media reality’ (Morgan, 2008 ), steers our gaze to the world, affects our opinions, and shapes our identities (Imesch et al., 2016 ; Hoffmann, 2018 ; Werenskjold, 2018 ). Even if some audiences may disagree with the content (Sampaio, 2022 ), news sets the agenda for societal discussions and contributes to what we consider worth knowing.

To fully understand the functions of audiovisual news content, production, and dissemination, exploring them through large and consistent sets of data, covering a long time span, can be helpful. Examining a consistent set of data that, for example, covers all the issues of a newsreel series, gives an understanding of the variety of individual findings and contextualises them. Detecting long-term continuities, and short-term trends helps us better understand the past information culture, alongside what is perhaps specific to our time or area. However, so far, studying audiovisual news via large quantities of data or across a long time span systematically in ways that would take into account their complexity has been hampered by availability of data and integration of methods across disciplines. Here we work towards a unified approach to study audiovisual news that enables the comparison of data coming from different sources to reveal the cultural and temporal variations of the global news scene.

Newsreels were the first widely spread form of audiovisual news. Starting in France in 1909, these approximately ten-minute-long news films, shown in the weekly changing series in cinemas, informed audiences about the latest political events, innovations, sports competitions, and fashion trends. Each newsreel issue contained around five to twelve short news stories, often showing the ‘more serious’ ones first and ending with entertaining topics. Until the mid-1950s, newsreels were the main source of audiovisual news for audiences globally that also conveyed both political propaganda and commercial interests. Their production continued in some countries under state support until the 1990s (Chambers et al., 2018 ; Pozharliev and Gallo González, 2018 ; Fielding, 2009 ).

Like other audiovisual products, newsreels are multifaceted. They are multimodal combinations of moving images, sounds, music, and of spoken and written language, gestures, iconographies, and signs as deeply rooted in the surrounding societies. The meanings created are interrelated across modalities with an individual news story gaining additional meaning, depending on its embedding and temporal position in a newsreel issue. In fact, one may argue, the meaning of a newsreel issue can be understood only when looking at the contents of the other issues of the newsreel series. Therefore, in order to understand the messages and role of individual news stories in a society, it is necessary to study newsreels as a whole (Hickethier, 2016 ) and comparatively, systematically analysing larger collections through the interconnections of small-scale units. This, we argue, has to transcend the debate of a single community of practice, such as media studies or communication, which is why this paper brings forth the expertise from a broad range of research streams—including film history, computational video analysis, film studies, and the so-called New Cinema History—to study film and video in a comprehensive way.

Lately, many newsreel series have been digitised, and several national film archives as well as transnational collections, such as Europeana, the Internet Archive, and Wikimedia Commons, all increasingly provide access to newsreels in digital form. This has opened new possibilities for studying long-term patterns of audiovisual news. However, despite promising developments in various disciplines, current approaches to newsreels, as discussed below, do not allow us to fully grasp these complex cultural products. Many established research fields are relevant to the longitudinal study of historical newsreels and audiovisual news in general. However, if the current approaches from each research stream are used separately, they produce considerable gaps in the nuanced understanding of newsreels in a long temporal continuum. The available approaches are either qualitative and do not allow a systematic analysis of large-scale data, or quantitative and reveal only one aspect of the multimodal newsreel data. Table 1 summarises the related research streams and their gaps, ranging from the scope of data used, comparability of analysis, the modalities taken into account, and the elements creating meaning in audiovisual material. Simultaneously, each research stream offers a contribution that helps fill the gaps in other fields. In the following paragraphs, we elaborate further on the gaps in each of these research streams together with the beneficial contributions that they may bring.

Qualitative film history

Qualitative film historical studies on newsreels have been demonstrating the variety of production conditions, core messages, and distribution channels in a number of countries and at different times (Chambers et al., 2018 ; Garrett Cooper et al., 2018 ; Imesch et al., 2016 ). Their strength is that they take into account the interplay of multiple modalities in the film material and produce nuanced analyses of the messages they have conveyed to the audiences. Simultaneously, however, they focus on temporally restricted segments of data, use analysis methods that are hard to apply to a large quantities of data, and do not usually utilise categories that would allow systematic comparisons between different studies (Chambers, 2018 , Pozharliev and Gallo González, 2018 , Bergström and Jönsson, 2018 , Vande Winkel and Biltereyst, 2018 ; Pozdorovkin, 2012 , Veldi et al., 2019 ). The main limitation of qualitative enquiry is its incomplete ability to offer an understanding of what is prevailing and what is marginal in wider terms in the data, which makes it difficult to see the bigger picture and contextualise the findings. As van Noord ( 2022 ) notes, exploring recurring motifs or patterns in cultural data is crucial for a deeper understanding. Although an experienced qualitative scholar is usually able to point out some of the repeating patterns based on their accumulated knowledge of the field, computational methods can back that up, measure the prevalence of the pattern in the collection, and detect also other, possibly unnoticed patterns.

Computational video analysis

Computational video analysis focuses on the systematic study of large collections of data, while typically addressing a single modality rather than aggregates of contextual and temporal factors. Examples include increasingly accurate and effective methods for recognising shot and scene boundaries (Hanjalic, 2002 ; Rasheed and Shah, 2003 ), persons (Wang and Zhang, 2022 ), objects (Brasó et al., 2022 ), human poses (Broadwell and Tangherlini, 2021 ), number of individuals in a crowd (Zhang and Chan, 2022 ), events (Wan et al., 2021 ), sounds (Park et al., 2021 ), human and animal behaviour (Gulshad et al., 2023 ; Bain et al., 2021 ; Sommer et al., 2020 ) and to perform image segmentation (Hu et al., 2022 ). Different solutions for condensing audiovisual content have also been developed, either for creating video representations to enable efficient browsing (Zhao et al., 2021 ) or numerical fingerprints allowing comparisons of video content for retrieval and recommendation systems (Kordopatis-Zilos et al., 2022 ; Nazir et al., 2020 ). Deep Learning applications in computer vision have been used for various item recognition tasks in images and videos (Bhargav et al., 2019 ; Liu et al., 2020 ; Kong and Fu, 2022 ; Brissman et al., 2022 ; Kandukuri et al., 2022 ). While mainstream computational video content analysis has focused on images, other modalities, like sound, have been also gaining increased attention (Valverde et al., 2021 ; Yang et al., 2020 ; Senocak et al., 2018 ; Hasegawa and Kato, 2019 ; Hu et al., 2022 ; Ye and Kovashka, 2022 ; Sanguineti et al., 2022 ; Pérez et al., 2020 ), eventually feeding into multi-modal analysis (Mourchid et al., 2019 ; Ren et al., 2018 ). However, considering different modalities of audiovisual data, particularly within the historical focus of this paper, remains beyond mainstream in video analysis. In addition, there is a lack of discussion on how certain units of analysis, such as recognised objects or condensed forms of video content, can be credibly used to detect the ways audiovisual content creates and conveys meaning to audiences.

Computational film studies

Situated between the qualitative and quantitative study of audiovisual contents, computational film studies often combine the two approaches. This stream of literature started by using shot detection to analyse film dynamics and editing styles (Salt, 1974 ; Tsivian, 2009 ; Pustu-Iren et al., 2020 ). In addition to addressing dynamics as an important modality of audiovisual content, computational film scholars have also been combining different modalities, such as images and sound (Grósz et al., 2022 ), spoken texts (Carrive et al., 2021 ; van Noord et al., 2021 ), or shown locations (Olesen et al., 2016 ). Computational studies of newsreels more specifically have addressed the contents of news either on the level of textual descriptions of news story topics (Althaus et al., 2018 ; Althaus and Britzman, 2018 ) or at a more granular level combining different modalities by analysing the voice-over text and automatically recognising well-known individuals in the film material (Carrive et al., 2021 ).

An ongoing debate in computational film studies concerns how film creates meaning, what are the most important meaning-making units, and how they could be best extracted (Chávez Heras, 2024 ; Burghardt et al., 2020 ; Burges et al., 2021 ). A profound challenge is that many modalities of film, such as images, can be interpreted in divergent ways depending on the viewer and their context (van Noord, 2022 ; Arnold and Tilton, 2019 ; Pozdorovkin, 2012 ). Different modalities may also create juxtaposing messages (Pozharliev and Gallo González, 2018 ). David Bordwell ( 1991 ) has argued that films contain ‘cues’ on which the further comprehension and interpretation of their meaning is based. Although the spectators may have differing opinions on the profound message of a film, an important hypothesis is that they nevertheless usually agree upon what the meaning-making cues are (such as shown activities or spoken sentences). This means that the variety of “credible” interpretations of the message of the film is limited (Bordwell, 1991 ). A central premise of computational film studies is thus that it can be possible to detect these cues and reach for an aggregate meaning of films through them.

Lately, in pursuit of understanding the meanings carried by film, a number of scholars have been using recognition and annotation of pre-set categories or stylistic features, discussing whether human interpretation should be applied already at the event of recognising the items, or at a later stage of the analysis (Carrive et al., 2021 ; Bhargav et al., 2019 ; Heftberger, 2018 ; Burges et al., 2021 ; Williams and Bell, 2021 ; Hielscher, 2020 ; Cooper et al., 2021 ; Bakels et al., 2020 ; authors discussing this issue: Burghardt et al., 2020 ; Arnold et al., 2021 ; Masson et al., 2020 ). There are also scholars further problematising object recognition by stating that in addition to recognising an object we should know how it is portrayed in order to understand its meaning (Hielscher, 2020 ) and calling for more thorough thinking of which measures can be used to analyse film contents (Olesen and Kisjes, 2018 ). This discussion connects with the wider question if there are cues in film that create meaning, how to find them, how to decide what to measure, and how to make sure that what is being measured gives responses to salient research questions. Although computational studies of historical newsreels use elaborate methods (Carrive et al., 2021 ; Althaus et al., 2018 ; Althaus and Britzman, 2018 ) including more explicit discussion on the connection of the research question and the variables can be an important methodological amendment to research.

New cinema history

New Cinema History (Maltby et al., 2011 ) stresses the importance of societal and temporal context in recent studies on film production (Dang, 2020 ), circulation (Clariana-Rodagut and Hagener, 2023 ; van Oort et al., 2020 ; Verhoeven et al., 2019 ; Navitski and Poppe, 2017 ), and reception (Treveri Gennari and Sedgwick, 2022 ). The premise of this discipline is that alongside the content, the surrounding context and its change over time are crucial factors in creating the meaning of film (as also pointed out by van Noord et al., 2022 ). Focusing on the contextual factors, this research stream has dealt less with content, yet because the meaning of cultural artefacts relies on both, these aspects need to be combined to reach a more nuanced understanding of newsreels or their aggregated meaning.

Digital hermeneutics

Examining historical material adds its particularities to a study. Current digital historical research has used the concept of ‘digital hermeneutics’ to call for epistemological data aka source criticism and method criticism (Fickers et al., 2022 ; Oberbichler et al., 2022 ; Salmi, 2020 ; Paju et al., 2020 ). It is crucial to understand how the data was formed and by whom, and what kinds of activities and worldviews it reflects. Firstly, the temporal meaning change of the formally similar units has to be taken into account. For example, showing a horse in a newsreel in 1910 and 1990 most likely creates very different interpretations. Secondly, digitised data are no longer in their original format (Fickers, 2021 ), and may contain traces left by the production, storage, archiving, digitising, and acquisition processes. For instance, textual descriptions of newsreel content are often added during the digitisation of the material and thus might reflect the perceptions or diligence of the digitisers rather than the activities of the original newsreel authors (Elo, 2020 ; see also Althaus and Britzman, 2018 ). As our case study shows in Section III, heavily censored data can also offer relevant results, when interpreted with an understanding that it provides the view of the authorities. Gaps in the data can produce meaningful insights. Therefore, it is important to account for which activities and to whom the traces that are being analysed belong. Furthermore, off-the-shelf computational analysis methods are often trained on contemporary materials and may not work similarly well with historical materials without adaptation (Grósz et al., 2022 ; Bhargav et al., 2019 ; Wevers, 2021 ; Wevers and Smits, 2020 ). Finally, the quality of cultural heritage materials can vary greatly, which poses additional challenges when studying long-term developments of audiovisual news.

Towards a unified approach

To summarise, while computational video analysis customarily assumes meaning to be contained in the artefact (i.e. the video), qualitative research and New Cinema History argue that meaning only arises when the artefact comes into contact with its audience and can be perceived as having different meanings. Simultaneously, an analysis that ignores inter-subjective contingency is blind to context; an interpretive framework that ignores inter-objective dependency is blind to structure. Both the content and the context should be taken into account, and, we argue, substantial advances in audiovisual (news) studies can be made by coupling these two positions.

The analysis framework for audiovisual newsreel corpora, as outlined in this paper was co-designed within a research process that started with experimental explorations of newsreel data, while negotiating and integrating methods from a spectrum of disciplines as brought together in the CUDAN ERA Chair project for Cultural Data Analytics at Tallinn University. Oscillating between joint reflections in collaborative group work, including two three-day hackathons, and more concentrated work on individual aspects, eventually led to the proposed generalisation of multidisciplinary collaboration in a systematic research process to make sense of historical newsreels at corpus scale. Following C.P. Snow’s call regarding the necessity to bridge the so-called “two worlds” of scholarly enquiry (Snow, 2001 [1959]), our starting point was that multidisciplinary integration brings forth more than a sum of its components. The specific stages of the proposed framework, explained in more detail below, were discovered by combining the established research processes of cultural data analytics and digital history, while experimenting with different ways of integrating quantitative and qualitative approaches, including expertise that is usually found in computation and the natural sciences

The objective of the framework is to exemplify how qualitative and quantitative approaches can be successfully brought together into a joint research pipeline. Towards this purpose, we combine the strengths of qualitative film history, computational video analysis, computational film studies, and New Cinema History listed in Table 1 , while closing their mutual and common gaps. In sum, we present a framework for systematically studying large collections of historical newsreels covering several decades in the context of their temporal and cultural dynamics, diversity, and functions. We propose bringing together a comprehensive set of aspects for a nuanced understanding of newsreels as an interplay of different modalities and contextual factors. The framework includes both qualitative and quantitative research feeding into a systematic approach and ability to cover large quantities of data. The framework, which we discuss in Section II, constitutes a schematic template for research projects combining quantitative and qualitative approaches (see Fig. 1 ). In Section III, we exemplify the framework using a dataset of “News of the Day” newsreel series produced in the Soviet Union in 1944–1992. Finally, Section IV contains the discussion and concludes the article.

The newsreel framework combines qualitative and quantitative approaches into a research pipeline. It contains ( a ) pairing meaning-making units with variables , ( b ) digital data (source) and method criticism, and ( c ) combining quantitative analysis with qualitative conclusions.

Newsreel Framework

Our framework essentially centres around a workflow pipeline configuration (Oberbichler et al., 2022 ) that includes qualitative and quantitative enquiry (Fig. 1 ). There are three important stages in the pipeline: detecting and pairing the meaning-making units and variables, digital data (source) and methods criticism, plus merging and explaining analysis visualisations of different dimensions of the data (Fig. 1a–c ). The study of newsreels begins with identifying meaningful research questions and data, in relation to preceding research. Perhaps more explicit than in established qualitative approaches, we propose to identify relevant meaning-making units arising from preceding research and qualitative enquiry, and pair them with available variables at the first stage (Fig. 1a ). In the second stage, we account for different temporal layers embedded in digitised heritage data to gain a better understanding of how the variables connect with the meaning-making units and the final conclusions of the study (Fig. 1b ). After this, appropriate analysis methods are selected, keeping in mind the available variables and research questions, followed by computational analysis. In the third stage the selected variables are studied quantitatively, feeding into an examination of the resulting dimensions of analysis to jointly produce final qualitative conclusions (Fig. 1c ). This stage brings the dimensions of analysis together, critically evaluates what the findings jointly report, contextualises them, and responds to the research questions. Adding these three stages to the research pipeline ensures that newsreels are analysed systematically by considering the multidimensional nature of meaning of cultural data (Schich, 2017 ; Cassirer, 1927 ), focusing on variables relevant to the research questions, and accounting for multimodality in the final results. The framework is modular, which means that it allows selecting methods that suit the particular research question or using multiple methods comparatively, while dealing with particular meaning-making units and variables. Qualitative and quantitative enquiry are firmly intermingled and mutually dependent in this research process, as exemplified in our case study below. Importantly, different parts of the research project are continuously adjusted in relation to each other (Schich, 2017 ; Gadamer, 2013 (1960).

While meaning-making units are the elements related to human understanding of what the phenomenon under study is composed of, the variables are the metadata entries or other features of the data that can be directly analysed computationally (cf. the distinction of elements and features in GIS; Zeiler, 1999 ). Detection and pairing of the meaning-making units with the available or traceable variables (Fig. 1a , see also Fig. 2 ) improve critical evaluation of meaning and comparability identified as gaps in preceding research (see Table 1 ). Furthermore, it establishes an explicit connection between the analysed variables and the phenomenon under study, enabling critical evaluation. The preceding literature uses the term ‘cue’ both when referring to what we call here the meaning-making units and variables (e.g. Bordwell, 1991 ; Ren et al., 2018 ), which complicates differentiating between the two. The meaning-making units come from the initial idea of the study, the research question, and the preceding literature, while variables are concretely present in the data. While they arise from different roots, the successful pairing of the two concepts is crucial for a fruitful study.

a Meaning-making units selected from Supplementary Table 1 for further analysis. b Existing and enriched variables of the News of the Day data. Arrows signify data enrichment based on the original data. c Resulting dimensions of analysis that interconnect the meaning-making units and variables.

The meaning-making units are elements that make up the phenomenon under scrutiny. Examples of meaning-making units, as relevant for newsreel research and broadly agreed in literature, include images, voice-over narration, acoustic motifs, the persons, activities, or locations shown, and content topics (Supplementary Table 1 ). Contextual factors are also important, including the socio-political circumstances, other concurrently available mass-communication media, and agency-related issues, like funding and the role of audiences. Relevant meaning-making units can be identified via an extensive literature review of qualitative studies on the topic to see what elements are often suggested and by critically evaluating the gaps. Of course, they may also emerge from analysis itself, in which case the research is firmly going beyond the state of the art.

In addition to the existing feature variables, others can be added, by either manually or algorithmically enriching the data, or adding additional data sources. As Table 2 shows, the most frequent metadata entries in the largest openly available collections of digitised newsreels contain information on production year, newsreel series title, duration, and content annotations either as text or keywords. The metadata entries, together with the available newsreel videos, form the basis for extracting variables. They can be further enriched with information concerning the newsreel authors, distribution, audience reactions, etc. To obtain well-selected units for computational analysis it is crucial to critically evaluate and pair the meaning-making units necessary for responding to research questions with variables that are available or traceable via enrichment. Notably, some variables might reveal meaning-making units indirectly (e.g. the number of people working on newsreels can be indicative of funding and the societal importance of newsreels).

The second stage we propose for the analysis part of the pipeline is to incorporate digital data (source) criticism by taking into account the historical multidimensionality of heritage data, as well as the temporal change affecting the meaning-making units and variables into the study (Fig. 1b ). This stage includes qualitative historical reflection, complementing the two other stages of the framework (Fig. 1 a, c). At this stage, firstly, the researchers scrutinise how the historical traces of the data, coming from production, storage, archiving, or digitising, are present in the data, affecting which variables should be selected for further analysis. The variable can be connected with the different meaning-making units, depending on the point in time and by whom it was created. For example, if a textual description of the contents was created as a newspaper advertisement or censorship card at the time of producing the newsreels (Werenskjold, 2018 ; Althaus and Britzman, 2018 ), the variable connects to the distribution and competition within the cinema market or the political context. If it was created within the digitisation process at a later stage, it should be combined with the interpretations of the later generations of what is noticeable in the contents. Secondly, the researchers will return to this stage after completing computational analysis of the variables to weight the effect of temporal change to the analysis results. As an example, they might reflect upon whether an increasing number of cars detected is due to an explicit choice of the filmmakers, the overall increase of the amount of cars in the society, the fact that the used algorithm detects better new car models than the old ones, or some other reason. Some results may also be absent due to conscious selections in data handling. For example, as our case study in Section III demonstrates, the qualitatively observed absence of footage portraying Stalin before his death in 1953 is most likely a result of de-selection of this material from the data (Fig. 3a ). With the twofold reflections concerning the content and method dependency, this stage addresses the lack of historical contextualisation identified in the preceding literature (see Table 1 ) by proposing to take into consideration the temporal aspects of data both when selecting the variables and when performing the final analysis.

a All News of the Day issues (scatter plot): x-axis publication years, y-axis issue number; the total number of news stories per year based on textual outlines; b number of shots per issue over time; c mean shot length per issue; d shares of news story topics per year classified based on textual descriptions of newsreels using an instructable zero-shot classifier. Each news story is classified with a single class. e A UMAP projection of story embeddings, coloured by the content predictions in ( e ) and ( f ); f annual news story topic distribution averaged over years.

The third stage is that selected variables are computationally examined and visualised as different dimensions of analysis (Fig. 1c ). Evidently, the used research methods should be selected so that they respond to the research questions when applied to the available variables (for method selection and comparison cf. for e.g. Opoku et al., 2016 ; Gentles et al., 2016 ). This stage addresses the lack of multimodality identified in preceding literature (see Table 1 ), and allows to combine newsreel contents with the contexts in a more streamlined manner. These dimensions, focusing, for example, on newsreel production conditions, or visual and content dynamics of newsreels, are further combined thematically or temporally into preliminary findings. Ideally, the dimensions of analysis represent different parts of the newsreel production, content, and distribution process to reach for a more comprehensive understanding of them. The findings are merged with the wider contextual information from the preceding literature.

The approach proposed here arises from discussions within the field of Cultural Data Analytics (Arnold and Tilton, 2023 ; van Noord et al., 2022 ; van Noord, 2022 ; Manovich, 2020 ; Arnold and Tilton, 2019 ; Schich, 2017 ; CUDAN, 2020 –2024). The starting points of this multidisciplinary approach are that cultural phenomena are inherently multi-scale and vary through time and space, that the interactions of particularity and universality are important, and that the meaning of cultural phenomena lies in the multidimensional relations of entities. When reaching for a bigger picture through longitudinal exploration, the main challenge is in maintaining the multitude of the phenomenon under study and simultaneously tracking the dynamics of selected variables. In this circumstance, recognising plurality and multidimensionality is crucial for understanding cultural phenomena, and we should be careful when reducing this multitude into means or homogenous groups (van Noord et al., 2022 ; van Noord, 2022 ; Manovich, 2020 ).

The design of our newsreel framework supports maintaining the multitude of cultural data while tracking its dynamics in a manner that allows comparisons across time and datasets. The following section exemplifies the application of the proposed framework to the analysis of the “News of the Day” newsreel series, published weekly in the Soviet Union from 1944 to 1992.

Materials and Methods

The data used in the case study is a collection of 1747 issues of the Russian-language Soviet newsreel journal News of the Day digitised by Net-Film company covering the years 1944–1992. We scraped the video files of newsreels with metadata containing information on the production year, issue number, authors and brief content descriptions in Russian and English with the permission of the data provider, the Net-Film company. The data is incomplete in many ways: the collection lacks some newsreel issues; the image and audio quality of the videos is low; and the metadata is imperfect. When working with digitised historical data and analysing the results it provides, incompleteness of the data is a common feature that needs to be taken into account (Carrive et al., 2021 ). Simultaneously, as our case study shows below, systematic holes in data can reveal crucial source-critical aspects of the data, informing the whole research. It is part of a historians’ skillset to be able to work with incomplete data, and to decide how far conclusions can be drawn from it (Howell and Prevenier, 2001 ).

The methodology of our case study followed the above proposed phenomenon categorisation by defining the central meaning-making units, and organisation and enrichment of the data to receive corresponding variables. We selected the methods used for analysing the resulting variables based on the team members’ domain expertise and their evaluations on the methods that would best respond to the research question of how the world was depicted in the News of the Day and by what kinds of groups of individuals involved in newsreel production. As the more detailed description of the methods below shows, all the steps of the research process involved intermingled qualitative, quantitative, computational, and human-made processes.

Meaning-making units

The table containing the meaning-making units of newsreels (Supplementary Table 1 ) was prepared by extensive reading of the preceding qualitative literature on newsreels. Identifying meaning-making units in qualitative research was purposeful because qualitative analysis takes a more holistic view to the phenomenon under scrutiny that quantitative approaches. We collected all the meaning-making units mentioned also in passing in the studies. Because scholars use varying terminology, we homogenised and aggregated the labels of the units. In addition to giving a general view, it also helps to pinpoint groups of studies that have different emphases, for example, on more abstract motifs, or those ones emphasising the contextual and agency-related meaning-making units instead of contents.

The matrix of the most frequent variables in the largest openly accessible collections of digitised newsreels (Table 2 ) lists the most commonly used metadata fields and their presence in some of the most well-known digitised newsreel collections. For the purpose of mapping the variety and prevalence of the metadata fields, the matrix lists the metadata entries using a common description, and not the specific entry titles each individual collection uses. Different digitising and archiving projects may use different types of metadata in variable formats, which may necessitate harmonising data in projects using several collections (see also Beals and Bell, 2020 ). In addition to the listed metadata entries, many collections also contain other data. For this mapping, we did not study how well the metadata entries have been filled or the consistency of the data. We have marked with “x” those entries that already exist, and with “i” those entries that can be extracted from the data. When selecting the variables for a study, qualitative evaluation of the historical dimensions of the data is essential.

Data enrichment

We amended the data by explicating further information both from the newsreel videos and metadata. For the videos, we ran shot boundary detection analysis (SBD), extracted the middle frames of each shot, and produced a ResNet50 (He et al., 2015 ) embedding for those frames. From the textual descriptions of newsreel contents in the metadata, we identified places mentioned using Named Entity Recognition (NER), and further geocoded the recognised location by adding lat/long coordinates. We also applied automatic detection of the assumed gender of newsreel directors and other crew members based on the surnames, which are grammatically gendered in Russian (Fig. 2b ). All automated steps involved qualitative and manual validation and correction of the processed results with human expertise in the loop.

News story categories

Each News of the Day issue is split into individual stories (12,707 across the 1747 reels), which have synopsis-like descriptions in the metadata. We also corrected small numbering and consistency issues in a minority of them by hand. We then applied two types of automatic content categorisation to the stories, topic modelling and content classification. Topic modelling (often using Latent Dirichlet Allocation, a form of “soft” clustering) is a common approach in digital humanities and other fields dealing with large text collections. For topics, we use the pretrained model driven approach (Angelov, 2020 , Grootendorst, 2022 ) where texts are first embedded using a word or sentence embedding (we use fasttext; Bojanowski et al., 2017 ) and then clustered, with cluster keywords derived via grouped term-frequency inverse-document-frequency (TF-IDF) scaling. The upside of topic modelling as an explorative approach is that the topics need not be known in advance. The downside is that the clusters may be hard to interpret or even meaningless, and the number of clusters must still be defined in advance. We therefore also experimented with another classification approach.

While in the recent past classifying content or topics would have required purpose-trained supervised classifiers, the advent of instructable large language models (LLMs, such as ChatGPT) makes it possible to predict topic or class prevalence in a “zero-shot” manner. Instead of training or tuning a classifier in a supervised manner on annotated examples, generative LLMs can be simply prompted (instructed) to output relevant text, including topic tags given an input example accompanied with the prompt. The simplest example would be along the lines of “Tag this sentence as being of topic X or Y. Example: [text]”, but we find more verbose prompts with topic definitions yield more accurate results. We defined eight topics of interest based on previous qualitative literature and Soviet history: USSR politics, sports, military (defence, wars), scientific and industrial progress (includes innovation, construction projects, space and aviation), USSR economy and industry, USSR agriculture (excludes other economy topics), natural disasters, social issues and lifestyle (includes education, family, health, leisure, culture, religion topics), and a “misc” topic meant to cover everything else (for the prompts, see the Supplementary material). We tested the zero-shot classification accuracy of two models, OpenAI’s generative pre-trained transformer (GPT) models gpt-3.5-turbo-0301 and gpt-4-0301 (OpenAI, 2023 ). These achieved 88 and 84% accuracy respectively on a hand-annotated 100-story test set. We therefore applied the 3.5 model to the rest of the story synopses, as illustrated in the Results section.

Visual characteristics

We extracted 117 shots on average (ranging from 20 to 247) per newsreel video, with 126 frames (5 s) per shot on average (ranging from 4 to 4508 frames or 0.2 to 180 s). Representing each shot with one frame, the corpus consists of 205.678 frames in total. We used a pre-trained ResNet Convolutional Neural Network (CNN), to embed the extracted video frames in high-dimensional feature space. The original training set for the ResNet50 is ImageNet (Deng et al., 2009 ), a standard collection of contemporary images, and here we apply it to a collection of low-resolution mostly grayscale images. To identify clusters of visually similar frames and detect common themes across reels we projected the embedding space in 2D using common dimension reduction methods such as t-SNE (van der Maaten and Hinton, 2008 ) and UMAP (McInnes et al., 2020 ). Using the Collection Space Navigator (Ohm et al., 2023 ), an interactive open source tool for exploring image collections, was instrumental in exploring the large-scale visual data and gaining new insights to it. We also visualised all the newsreels by sequencing one frame per shot next to one another, effectively in this way creating a storyboard covering all the examined newsreels. In this part we used standard methods with known biases (see, for example Studer et al., 2019 ).

From the results of the Named Entity Recognition (NER) we extracted mentions of cities. We used Wiktionary and authors’ knowledge of Russian grammar to extract additional name-derivative words related to cities. Using this list, we counted mentions of cities in the story descriptions. We qualitatively distinguish five types of city mentions: a) city itself and city dwellers; b) organisations located in the city and named after it; c) names of a region named after the capital (for example ‘Leningrad oblast’) and organisations located there; d) toponyms named after the city which are not located there or in its vicinity including entities, treatises, and historical events (for example ‘Warsaw Pact’); e) not a mention (coincidences and homonyms). We added geo-coordinates taken from Wikipedia to the list of cities to visualise them on a map.

Crew composition

We used newsreel crew metadata to construct a directed graph of co-working relations (Verhoeven et al., 2020 ) where directors and other crew members act as nodes, and edges indicate collaboration on a newsreel issue. The edge direction is drawn from the director to all other crew members and signifies hiring and supervisory relationships. We utilised Levenshtein distance (Levenshtein, 1965 , see also Navarro, 2001 ) to detect potentially misspelt duplicate names and manually checked the need to merge nodes. The crew dataset contains information about 1251 people who worked on 1730 newsreel productions during 1954–1991 across different positions: director (1740 roles by 104 persons), cinematographer (15,145 roles by 1132 persons) and other crew (editors, sounds designers, etc.; 158 roles by 45 persons). Notably, a small portion of staff work across different roles. The dataset results in a network with 1251 unique person nodes and 15,425 person-to-person links. The first nine years of the data collection period were omitted from network analysis due to inconsistent data.

Cinematic and topic trends

The cinematic and topic trends of the News of the Day data show that newsreel production and release as measured by the number of newsreel issues appears to be stable over fifty years (Fig. 3a ) with consistent content shares dedicated to different topics (Fig. 3e ). The first and last few years (1945–1953 and 1990–1992) look somewhat different, but they have much less data than the rest of the period (Fig. 3a ). Newsreel issue numbers recorded, leased, and preserved in the sparse available data before 1954 seem to indicate that newsreels were produced more or less weekly during that period, but only a tiny fragment has been stored and/or digitised (Fig. 3a ). The absence of data before 1954 most likely relates to the ‘de-Stalinization’ of film materials after Stalin’s death in 1953, which included the confiscation of materials with excessive references to the former leader (Heftberger, 2018 ). During 1954–1986, the weekly production was stable, and newsreels were archived, kept, and later digitised systematically (apart from 1965 with missing data). From 1987, the annual number of produced newsreels decreased by half. The 1987 drop in newsreel production volumes coincides with the time of perestroika characterised by economic turbulence and the rethinking of the Soviet media ecosystem (Rodgers, 2014 ).

Topic-wise, the shares of political, economic, agricultural, and social news, classified using the zero-shot prediction approach, remained relatively stable until the mid-1980s when the social, and later political themes began to take more room of the preserved newsreels (Fig. 3f ). The trend shows an annual rhythm (Fig. 3d ), where social news topics usually increased around issue numbers 8–9, which coincided with International Women’s Day, and around issue numbers 48–52 coinciding with the New Year, both officially recognised celebrations in the Soviet Union. Also the topic of agriculture was more prominent around issues 30–40 published in August and September, which were the most important months of harvest.

With a closer look, it is possible to identify subtle changes across the observed period. Although the annual number of issues remained relatively stable during 1954–1986, the number of news stories per issue, determined based on the textual outlines in the metadata, decreased gradually during this time (Fig. 3a ). Also the number of shots in a newsreel decreased over time (Fig. 3b ), while the mean length of shots started to increase towards the end of the period (Fig. 3c ). These results show a contrary trend to the findings of scholars studying Hollywood feature films that indicate shortening shot lengths towards the end of the 20th century (Cutting et al., 2011 ). The reasons for the ‘stagnating’ Soviet newsreel dynamics should be further explored, with candidates obviously including the availability of film material of extended length, and labour cost in post production, such as cutting and composition. While we provide preliminary exploratory results here, quantitative data like these also naturally allow for the testing of specific hypotheses.

Our examination of the central frames of each shot reveals recurring visual patterns that repeat during the whole studied period (Fig. 4 ). Laying out all the frames of every issue into a storyboard shows subtle length and darkness variation of the (digitised) film material, as well as the launch of colour film in the mid-1980s (Fig. 4a right). Placing the frames in the order of year, issue, and scene number allows for comparing the recurring patterns and changes of the newsreel series. For example, the closeup of the storyboard shows that the opening title frames were customarily followed by frames showing a city scene, indicating the place of the news story. This prelude was followed by scenes depicting activities, such as leaders meeting each other (Fig. 4a left). Using the ResNet50 CNN embedding to extract visual features from the central frame of each shot allows us to examine visual similarities across reels. A UMAP projection of the embedded frames reveals aspects of these similarities at least at a coarse-grain level (Fig. 4b ). Consequently the UMAP allows for visual examination, grouping, and annotation of the most prominent image types in the collection, such as “Nature”, “Monumental gatherings”, “People in meetings”, “Closeups of people at work”, ”Industrial production”, “Title frames and other texts”, and “City views”.

a A storyboard of all newsreel issues, x-axis shot number, y-axis publication years and issue numbers in ascending order. The layout of all issues (4a right) shows the temporal variation of issue lengths and the closeup of the storyboard ( a left) visualises the first scenes of issues 6–14 from 1970. b A UMAP projection of ResNet50 embedding of all central frames of each shot with seven most prominent image clusters named by the authors as (1) “Nature”; (2) “Monumental gatherings”; (3) “People in meetings”; (4) “Closeups of people at work”; (5) ”Industrial production”; (6) “Title frames and other texts”; and (7) “City views”. We used the Collection Space Navigator (Ohm et al., 2023 ), i.e. a flexible open-source user interface, for examining the frames and to produce the figure.

City mentions

Our examination of the cities mentioned in the textual descriptions of the newsreel metadata is summarised in Fig. 5 . Spatially, it demonstrates a heavy emphasis on Europe, both within the Soviet Union and globally, while the Asian part of the Soviet Union in East of the Ural Mountains, was far less covered, matching its lower population rates (Fig. 5a, b ). Outside the Soviet Union, the Warsaw Pact socialist countries are the most frequently covered (36% of all mentions despite being 3% of world population in 1970), as well as ‘neutral’ capitalist countries such as Austria and Finland (9% of all mentions despite being less than 0.4% of world population) (Fig. 5 a, b, d). These findings match the consensus among historians studying Soviet history generally (Koivunen, 2016 ; Gilburd, 2013 ; Turoma and Waldstein, 2013 ). Timewise, the number of mentions per year trends downwards (Fig. 5c–e ), which matches the general decrease in the number of stories per year (Fig. 3a ) and is mostly due to the newsreel issues typically having fewer and longer stories in the 1970s and the 1980s than in the earlier period. It is, however, noteworthy that the number of mentions of foreign cities is shrinking even faster (Fig. 5 c, e), emphasising the decline of the fraction of stories dedicated to international events after around 1960. The temporal patterns for some cities demonstrate a variety of interesting qualitative behaviour (Fig. 5e ). Constant popularity of Leningrad/St. Petersburg seems natural in the view of its importance as the second-largest city in the USSR and the “cradle of the revolution”, the upward trend in the mentions of Minsk correlates with the rapid growth of its population in the period under consideration, and the bump in the popularity of Krasnoyarsk in the 1960s coincides with the building of the Krasnoyarsk Hydroelectric Dam, which was a topic of multiple newsreel stories. The decline of the mentions of Odesa require further historical analysis. The data for individual cities is rather sparse and noisy so extracting statistically significant information from it requires application of advanced statistical techniques and will be done in detail elsewhere.

Map showing all the cities mentioned in 1944–1992, ( a ) globally and ( b ) in Europe. The bubble size indicates the number of mentions. c Average number of mentions of top 50 cities per 1000 stories, the red line is the Soviet cities, and the blue line foreign cities. d Heatmap of city mentions per year for the top 50 most-mentioned cities (Moscow excluded due to heavy overrepresentation). e Heatmap of the top-50 most mentioned cities (Moscow excluded) per 1000 stories in the periods of 1954–1964, 1966–1976, and 1977–1992.

The analysis of newsreel production crews reveals production labour market dynamics and labour division between genders over time. Newsreel production crew numbers (Fig. 6a ) closely follow newsreel production volumes (Fig. 3a ), with ten people working on a newsreel on average. Directors who lead the productions are expectedly vastly outnumbered by other crews since newsreels contain multiple stories often shot by different cinematographers (on average nine versus a single director per newsreel). The historical labour market features several prominent directors, who lead multiple teams (as seen from high degree-centrality nodes in the director–crew network and node degree distribution in Fig. 6d–e ), and who pursue long-lasting careers (Fig. 6c ). The analysis of director gender composition reveals the existence of three distinct periods: gender equality during 1945–1959, a women director’s era during 1960–1974, and men director’s era during 1975–1992 (Fig. 6b ).

a Number of individuals working on newsreel production over time, coloured by role. b Number of individuals working as directors over time, coloured by the assumed gender (women and men). c Director career longevity for the top-20 most productive directors. d Newsreel production crew network during 1954–1991, edges drawn from directors to other crew, coloured by role. e Degree distribution for the unipartite directed newsreel production crew network, both axes in logarithm.

Merging and explaining

As we have shown above, each dimension of analysis reveals new avenues for further qualitative and quantitative enquiry. In addition, analysing similar trends, interrelated themes, or temporal sequences overarching different dimensions of analysis, including combining them in statistical modelling, may help to explain the studied phenomenon better. In our case study, interested in the worldviews portrayed in the Soviet newsreels, bringing the results from different dimensions of analysis together points out a period with emerging shifts. The most prominent temporal change, found across all dimensions of analysis, was the time of perestroika , which introduced major political and cultural changes in the Soviet Union (1985–1991). Although some of the identified changes during this period, such as the launch of colour film (Fig. 4a ), likely had little to do with the political changes, the dimensions of analysis show how profoundly the time of change was affecting different spheres of society. The number of yearly newsreel issues was cut in half, and the published issues contained far fewer news stories (1–3 stories per newsreel against the earlier number of 8–10 stories, Fig. 3a ). Simultaneously, the number of filmmakers producing newsreels was rapidly decreasing following the shrinking newsreel production (Fig. 6a, b ). It is possible that the collapsing Soviet economy and decentralising cultural policy together with the prevalence of television overran the outdated media of newsreels in the era characterised by a gradual increase of freedom of speech and press (Rodgers, 2014 ). Digging deeper via qualitative inspection, we can see that in the 1990s the newsreel contents became focused on political meetings held in Moscow, which is visible in the emphasis on political and social topics covered (Fig. 3f ), and in geographical concentration on only a few cities (Fig. 5d ). The newsreels of the perestroika were characterised by long shots of speeches (Fig. 3c ), as many newsreel issues at the time covered extensively the political discussions on the direction of the country, which provided the public in a way first-hand knowledge of who said what in the discussions. Clearly, the worldview that the News of the Day depicted to its audiences, changed in many ways.

While many of these observations are not novel to studies on Soviet history, seeing a signal pointing out the particularity of this period in all the dimensions of analysis is important. It shows that the change of policy in the mid-1980s had way more profound effects than for example, the change of leadership from Khrushchev to Brezhnev in 1964. Quantitative analysis of large amounts of data provides the necessary contextualisation emphasising the specificity of the period, which would not be possible to show in such a concrete manner by a solely qualitative study. The signal evidence furthermore becomes visible to a broader audience, beyond experts whose formation requires years of qualitative research. Additionally, harnessing the findings of the different yet complementary dimensions of analysis together reveals trends that may be interrelated. For example, the diminishing number of crew members can partially explain the decreasing number of issues and shots, and the concentration of the newsreels in only a few cities. With fewer people, it was impossible to cover a larger volume of news material from different places. Focusing only on one dimension of analysis in our case study would not have revealed this possible connection. Finally, all these findings can be enhanced by further qualitative enquiry referencing back to the historical dimensions of the data corpus used in this study and in preceding studies, as well as statistical modelling focusing on any particular questions of interest.

Discussion and conclusions

In this paper, we proposed a framework for studying historical newsreels specifically and audiovisual news more generally in large quantities, while simultaneously maintaining an understanding of the multimodality and complexity of audiovisual data and the relational way of meaning-making associated with them. Analysing newsreels using long-term and large-scale data is beneficial for our understanding of societies in question of the global information landscape, its geographical differences, and the generic features of news content. As our case study on worldviews in the News of the Day newsreel series produced weekly in the Soviet Union during 1944–1992 has demonstrated, combining different dimensions of quantitative analysis together with qualitative enquiry, helps to understand newsreel contents in a long continuum and in a more nuanced way than previously achieved. Quantitative visualisations driven by computational analysis methods help to contextualise smaller-scale qualitative analysis, simultaneously as qualitative analysis allows to explain the detected long-term changes and their nuances. Acknowledging the complexity of the data, i.e. that new quality emerges from large quantities of data, allows for a better-rounded understanding of audiovisual culture. Necessitating a range of co-authors, our approach makes an argument for multidisciplinary research and advocates studying culture by combining different methods and approaches.

The outlined framework is the first attempt to combine the different disciplinary approaches into a comprehensive study of newsreels. Weaknesses in our proposition may of course become apparent when applying it in a variety of studies, yet we argue that this too will necessitate similar multidisciplinary expertise, collaboration, and negotiation. The case study we have presented here provides a brief glimpse into the application of the framework. One limitation of our approach is that while it selects dimensions of analysis intuitively, yet based on expertise of the crowd of co-authors, it does not explore in detail the selection of analysis methods. This will be further explored in the future. In our case study, we have focused on preliminary exploratory enquiry and less on confirmatory analysis or hypothesis testing. Examining the different ways to compare a variety of datasets, coming from different sources, has not been touched upon in this article, and should be further studied to enhance transnational approaches to the study of newsreels. This article has proposed a methodological solution for studying audiovisual news, while the questions of copyright and access to comprehensive collections of audiovisual data and corresponding metadata continue to be major obstacles to further development of this field (Arnold et al., 2021 ). A further potential hurdle in scaling the approach is the necessity of access to high-performance computation infrastructure for the effective processing of large-scale audiovisual data. In sum, however, with this framework, we hope to open a discussion on how to best study audiovisual news in long-term and large-scale data.

Data availability

The data is available at the company’s website ( https://www.net-film.ru/ ). The code used for accessing the data is available at the supplementary materials.

Althaus SL, Britzman K (2018) Researching the issued content of American newsreels. In: Vande Winkel R, Chambers C, Jönsson M (eds.) Researching Newsreels: local, national and transnational case studies. Palgrave Macmillan, Switzerland, p 247–263

Chapter   Google Scholar  

Althaus SL, Usry K, Richards S, Van Thuyle B, Aron I, Huang L, Leetaru K et al. (2018) Global news broadcasting in the pre-television era: a cross-national comparative analysis of World War II Newsreel coverage. J Broadcast Electron Media 62(1):147–167. https://doi.org/10.1080/08838151.2017.1375500

Article   Google Scholar  

Angelov D (2020) Top2Vec: distributed representations of topics, arXiv, https://doi.org/10.48550/arXiv.2008.09470

Arnold T, Scagliola S, Tilton L, and Van Gorp J (2021) Introduction: special issue on audiovisual data in DH. Digit Hum Q 15 (1). http://digitalhumanities.org/dhq/vol/15/1/000541/000541.html

Arnold T, Tilton L (2019) Distant viewing: analyzing large visual corpora. Digit Scholarsh Hum 34:i3–16. https://doi.org/10.1093/llc/fqz013

Arnold T, Tilton L (2023) Distant viewing: computational exploration of digital images. MIT Press, Cambridge, MA. https://doi.org/10.7551/mitpress/14046.001.0001

Bain M, Nagrani A, Schofield D, Berdugo S, Bessa J, Owen J, Hockings KJ, et al. (2021) Automated audiovisual behavior recognition in wild primates. Sci Adv 7(46). https://doi.org/10.1126/sciadv.abi4883

Bakels J-H, Grotkopp M, Scherer T, Stratil J (2020) Matching computational analysis and human experience: performative arts and the digital humanities. Digit Hum Q 14:4

Google Scholar  

Beals M, Bell E (2020) The atlas of digitised newspapers and metadata: reports from oceanic exchanges. https://doi.org/10.6084/m9.figshare.11560059.v1

Bergström Å, Jönsson M (2018) Screening war and peace: newsreel pragmatism in neutral Sweden, September 1939 and May 1945. In: Vande Winkel R, Chambers C, Jönsson M (eds.) Researching newsreels: local, national and transnational case studies. Palgrave Macmillan, Switzerland, p 157–182

Bhargav S, van Noord N, Kamps J (2019) Deep learning as a tool for early cinema analysis. Proceedings of the 1st workshop on structuring and understanding of multimedia heritage contents. SUMAC ’19. Association for Computing Machinery, New York, NY, USA, p 61–68. https://doi.org/10.1145/3347317.3357240

Book   Google Scholar  

Bojanowski P, Grave E, Joulin A, Mikolov T (2017) Enriching word vectors with subword information. Trans Assoc Comput Linguist 5:135–146. https://doi.org/10.1162/tacl_a_00051

Bordwell D (1991) Making meaning: inference and rhetoric in the interpretation of cinema. Harvard University Press, Cambridge

Brasó G, Cetintas O, Leal-Taixé L (2022) Multi-object tracking and segmentation via neural message passing. Int J Comput Vis 130(12):3035–3053. https://doi.org/10.1007/s11263-022-01678-6

Brissman E, Johnander J, Danelljan M, and Felsberg M (2022) Recurrent graph neural networks for video instance segmentation. Int J Comput Vis. https://doi.org/10.1007/s11263-022-01703-8

Broadwell P, Tangherlini TR (2021) Comparative K-Pop choreography analysis through deep-learning pose estimation across a large video corpus. Digit Hum Q 15:1

Burges J, Armoskaite S, Fox T, Mueller D, Romphf J, Sherwood E, Ullrich M (2021) Audiovisualities out of annotation: three case studies in teaching digital annotation with mediate. Digit Hum Q 15:1

Burghardt M, Heftberger A, Pause J, Walkowski N-O, Zeppelzauer M (2020) Film and video analysis in the digital humanities—an interdisciplinary dialog. Digit Hum Q 14:4

Carrive J, Beloued A, Goetschel P, Heiden S, Laurent A, Lisena P, Mazuet F, et al. (2021) Transdisciplinary analysis of a corpus of French newsreels: the ANTRACT project. Digit Hum Q 15 (1). http://digitalhumanities.org/dhq/vol/15/1/000523/000523.html

Cassirer E (1927) Das symbolproblem und seine stellung im system der philosophie. Z für ÄEsthet Allg Kunstwiss 21:295–322

Chambers C (2018) The Irish Question: newsreels and national identity. In: Vande Winkel R, Chambers C, Jönsson M (eds.) Researching newsreels: local, national and transnational case studies. Palgrave Macmillan, Switzerland, p 265–283

Chambers Ciara, Jönsson Mats, Vande Winkel Roel (eds.) (2018) Researching newsreels: local, national and transnational case studies. Palgrave Macmillan, Switzerland

Chapman, J, Glancy M, and Harper S (2007) The new film history: sources, methods, approaches . Springer

Chávez Heras D (2024) Cinema and machine vision: artificial intelligence, aesthetics and spectatorship. Edinburgh University Press, Edinburgh

Clariana-Rodagut A, Hagener M (2023) Transnational networks of avant-garde film in the interwar period. In: Roig-Sanz D, Rotger N (eds.) Global literary studies: key concepts. De Gruyter, Berlin, p 253–277

Cooper A, Nascimento F, Francis D (2021) Exploring film language with a digital analysis tool: the case of Kinolab. Digit Hum Q 15:1

CUDAN Open Lab Seminar series 2020–2024. https://www.youtube.com/@CUDANLab/videos)

Cutting JE, Brunick KL, DeLong JE, Iricinschi C, Candan A (2011) Quicker, faster, darker: changes in Hollywood film over 75 years. I-Percept 2(6):569–576. https://doi.org/10.1068/2Fi0441aap

Dang S-M (2020) Unknowable facts and digital databases: reflections on the women film pioneers project and women in film history. Digit Hum Q 14:4

Deng J, Dong W, Socher R, Li L-J, Li K, Fei-Fei L (2009) ImageNet: a large-scale hierarchical image database. In Proceedings of the IEEE conference on computer vision and pattern recognition , 248–255. https://doi.org/10.1109/CVPR.2009.5206848

Elo K (2020) Big Data, Bad Metadata: a methodological note on the importance of good metadata in the age of digital history. In: Fridlund M, Oiva M, Paju P (eds.) Digital histories. Emergent approaches within the new digital history. Helsinki University Press, Helsinki, pp 103–111 https://doi.org/10.33134/HUP-5-6

Fickers A (2021) Authenticity: historical data integrity and the layered materiality of digital objects. In: Fickers A, Schafer V, Takats S, Zaagsma G (eds.) Digital roots. Historicizing media and communication concepts of the digital age. De Gruyter Oldenbourg, Berlin, pp 299–312 https://doi.org/10.1515/9783110740202-017

Fickers A, Tatarinov J, Heijden T (2022) Digital history and hermeneutics—between theory and practice: an introduction. In: Fickers A, Tatarinov J (eds.) Digital history and hermeneutics: between theory and practice. De Gruyter Oldenbourg, Berlin, pp 1–19 https://doi.org/10.1515/9783110723991

Fielding R (2009) Newsreels. Encyclopedia of journalism, 3 , SAGE Publications, Thousand Oaks, pp 992–994

Gadamer H-G (2013 (1960)) Truth and method . Bloomsbury Academic

Garrett Cooper Mark, Levavy SaraBeth, Melnick Ross, Williams Mark (eds.) (2018) Rediscovering U.S. newsfilm: cinema, television, and the archive. Routledge, London

Gentles SJ, Charles C, Nicholas DB, Ploeg J, McKibbon KA (2016) Reviewing the research methods literature: principles and strategies illustrated by a systematic overview of sampling in qualitative research. Syst Rev 5(1):172. https://doi.org/10.1186/s13643-016-0343-0

Article   PubMed   PubMed Central   Google Scholar  

Gilburd E (2013) The Revival of Soviet Internationalism in the Mid to Late 1950s. In: Gilburd E, Kozlov D (eds.) The Thaw: Soviet Society and Culture during the 1950s and 1960s. University of Toronto Press, Toronto, p 362–401

Grootendorst M (2022) BERTopic: neural topic modeling with a class-based TF-IDF procedure. arXiv https://doi.org/10.48550/arXiv.2203.05794

Grósz T, Kallioniemi N, Kiiskinen H, Laine K, Moisio A, Römpötti T, Virkkunen A, Salmi H, Kurimo M, Laaksonen J (2022) Tracing signs of urbanity in the finnish fiction film of the 1950s: toward a multimodal analysis of audiovisual data. In: Proceedings of the 6th digital humanities in the nordic and Baltic countries conference (DHNB 2022), 3232: 63–78. https://ceur-ws.org/Vol-3232/paper05.pdf

Gulshad S, Long T, van Noord N (2023) Hierarchical explanations for video action recognition. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, p 3703–3708. https://doi.org/10.48550/arXiv.2301.00436

Hanjalic A (2002) Shot-boundary detection: unraveled and resolved? IEEE Trans Circuits Syst Video Technol 12(2):90–105. https://doi.org/10.1109/76.988656

Hasegawa T, Kato S (2019) Dialogue mood estimation from speech sounds clusterized with speakers’ personality traits. In Proceedings of the IEEE 8th global conference on consumer electronics (GCCE) : 399–401. https://doi.org/10.1109/GCCE46687.2019.9015238

He K, Zhang X, Ren S, Sun J (2015) Deep residual learning for image recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), p. 770–778. https://doi.org/10.48550/arXiv.1512.03385

Heftberger A (2018) Digital humanities and film studies. Visualizing Dziga Vertov’s Work. Quantitative methods in the humanities and social sciences. Springer International Publishing, Switzerland, https://doi.org/10.1007/978-3-030-02864-0_4

Hickethier K (2016) The creation of cultural identity through weekly newsreels in Germany in the 1950s: as illustrated by the NEUE DEUTSCHE WOCHENSCHAU and the UFA-WOCHENSCHAU (With a Side Glance at the DEFA Weekly Newsreel DER AUGENZEUGE). In: Imesch K, Schade S, Sieber S (eds.) Constructions of cultural identities in newsreel cinema and television after 1945. Transcript Verlag, Bielefeld, p 39–54. https://doi.org/10.14361/9783839429754-003

Hielscher E (2020) The phenomenon of interwar city symphonies: a combined methodology of digital tools and traditional film analysis methods to study visual motifs and structural patterns of experimental-documentary city films. Digit Hum Q 14:4

Hoffmann K (2018) The commentary makes the difference: an analysis of the Suez War in East and West German Newsreels, 1956. In: Vande Winkel R, Chambers C, Jönsson M (eds.) Researching newsreels: local, national and transnational case studies. Palgrave Macmillan, Switzerland, p 77–92

Howell MC, Prevenier W (2001) From reliable sources: an introduction to historical methods. Cornell University Press, Ithaca

Hu X, Tang C, Chen H, Li X, Li J, Zhang Z (2022) Improving image segmentation with boundary patch refinement. Int J Comput Vis 130(11):2571–2589. https://doi.org/10.1007/s11263-022-01662-0

Imesch Kornelia, Schade Sigrid, Sieber Samuel (eds.) (2016) Constructions of cultural identities in newsreel cinema and television after 1945. Transcript Verlag, Bielefeld, 10.14361/9783839429754

Kandukuri RK, Achterhold J, Moeller M, Stueckler J (2022) Physical representation learning and parameter identification from video using differentiable physics. Int J Comput Vis 130(1):3–16. https://doi.org/10.1007/s11263-021-01493-5

Koivunen P (2016) Friends, ‘potential friends,’ and enemies: reimagining Soviet relations to the First, Second, and Third Worlds at the Moscow 1957 youth festival. In: Babiracki P, Jersild A (eds.) Socialist internationalism in the cold war. Springer International Publishing, New York, pp 219–47 https://doi.org/10.1007/978-3-319-32570-5_9

Kong Y, Fu Y (2022) Human action recognition and prediction: a survey. Int J Comput Vis 130(5):1366–1401. https://doi.org/10.1007/s11263-022-01594-9

Kordopatis-Zilos G, Tzelepis C, Papadopoulos S, Kompatsiaris I, Patras I (2022) DnS: distill-and-select for efficient and accurate video indexing and retrieval. Int J Comput Vis 130(10):2385–2407. https://doi.org/10.1007/s11263-022-01651-3

Lazer DMJ, Baum MA, Benkler Y, Berinsky AJ, Greenhill KM, Menczer F, Metzger MJ, Nyhan B, Pennycook G, Rothschild D (2018) The science of fake news. Science 359(6380):1094–1096

Article   ADS   CAS   PubMed   Google Scholar  

Levenshtein VI (1965) Binary codes capable of correcting spurious insertions and deletions of ones. Probl Inf Transm 1(1):8–17

Liu L, Ouyang W, Wang X, Fieguth P, Chen J, Liu X, Pietikäinen M (2020) Deep learning for generic object detection: a survey. Int J Comput Vis 128(2):261–318. https://doi.org/10.1007/s11263-019-01247-4

van der Maaten L, Hinton G (2008) Visualizing data using T-SNE. J Mach Learn Res 11:9

Maltby R, Biltereyst D, Meers P (eds.) (2011) Explorations in new cinema history: approaches and case studies . Wiley

Manovich L (2020) Cultural analytics. MIT Press, Cambridge, MA

Masson E, Olesen CG, Noord Nvan, Fossati G (2020) Exploring digitised moving image collections: the SEMIA project, visual analysis and the turn to abstraction. Digit Hum Q 14:4

McInnes L, Healy J, Melville J (2020) UMAP: uniform manifold approximation and projection for dimension reduction. ArXiv http://arxiv.org/abs/1802.03426

Morgan M (2008) Reality and media reality. The international encyclopedia of communication , John Wiley & Sons, New Jersey. https://doi.org/10.1002/9781405186407.wbiecr015

Mourchid Y, Renoust B, Roupin O, Văn L, Cherifi H, Hassouni ME (2019) Movienet: a movie multilayer network model using visual and textual semantic cues. Appl Netw Sci 4(1):121. https://doi.org/10.1007/s41109-019-0226-0

Navarro G (2001) A guided tour to approximate string matching. ACM Comput Surv 33(1):31–88. https://doi.org/10.1145/375360.375365

Navitski R, Poppe N (2017) Cosmopolitan film cultures in Latin America, 1896–1960. Indiana University Press, Indiana

Nazir S, Cagali T, Sadrzadeh M, Newell C (2020) Audiovisual, genre, neural and topical textual embeddings for TV programme content representation. In: Proceedings of the IEEE international symposium on multimedia (ISM) : p 197–200. https://doi.org/10.1109/ISM.2020.00041

van Noord N (2022) A survey of computational methods for iconic image analysis. Digit Scholarsh Hum 37(4):1316–1338. https://doi.org/10.1093/llc/fqac003

van Noord N, Olesen CG, Ordelman R, Noordegraaf J (2021) Automatic annotations and enrichments for audiovisual archives. ICAART 1:633–640

van Noord N, Wevers M, Blanke T, Noordegraaf J, and Worring M (2022) An analytics of culture: modeling subjectivity, scalability, contextuality, and temporality. arXiv https://doi.org/10.48550/arXiv.2211.07460

Oberbichler S, Boroş E, Doucet A, Marjanen J, Pfanzelter E, Rautiainen J, Toivonen H, Tolonen M (2022) Integrated interdisciplinary workflows for research on historical newspapers: perspectives from humanities scholars, computer scientists, and librarians. J Assoc Inf Sci Technol 73(2):225–239. https://doi.org/10.1002/asi.24565

Article   PubMed   Google Scholar  

Ohm T, Canet Solá M, Karjus A, and Schich M (2023) Collection space navigator: interactive visualization interface for multidimensional datasets. https://collection-space-navigator.github.io/

Olesen CG, Kisjes I (2018) From text mining to visual classification: rethinking computational new cinema history with Jean Desmet’s digitised business archive. TMG J Media Hist 21(2):127–145. https://doi.org/10.18146/2213-7653.2018.370

Olesen CG, Masson E, Van Gorp J, Fossati G, Noordegraaf J (2016) Data-driven research for film history: exploring the Jean Desmet collection. Mov Image: J Assoc Mov Image Arch 16(1):82–105. https://doi.org/10.5749/movingimage.16.1.0082

van Oort T, Jernudd Å, Lotze K, Pafort-Overduin C, Biltereyst D, Boter J, Dibeltulo S et al. (2020) Mapping film programming across post-war Europe (1952): arts and media. Res Data J Hum Soc Sci 5(2):109–125. https://doi.org/10.1163/24523666-00502009

OpenAI (2023) GPT-4 technical report. arXiv https://doi.org/10.48550/arXiv.2303.08774

Opoku A, Ahmed V, Akotia J (2016) Choosing an appropriate research methodology. In: Ahmed V, Opoku A, Aziz Z (eds.) Research methodology in the built environment: a selection of case studies. Routledge, London, p 32–50

Paju P, Oiva M, Fridlund M (2020) Digital and distant histories: introducing emergent approaches within the new digital history. In: Fridlund M, Oiva M, Paju P (eds.) Digital readings of history. history research in the digital era. Helsinki University Press, Helsinki, p 3–18. https://doi.org/10.33134/HUP-5

Park S, Bellur A, Han DK, Elhilali M (2021) Self-training for sound event detection in audio mixtures. In Proceedings of the ICASSP 2021–2021 IEEE international conference on acoustics, speech and signal processing (ICASSP) : 341–345. https://doi.org/10.1109/ICASSP39728.2021.9414450

Pérez AF, Sanguineti V, Morerio P, Murino V (2020) Audio-visual model distillation using acoustic images. In: Proceedings of the IEEE winter conference on applications of computer vision (WACV) : 2843–2852. https://doi.org/10.1109/WACV45572.2020.9093307

Pozdorovkin M (2012) Khronika: Soviet newsreel at the dawn of the information age . PhD thesis, Harvard University. https://dash.harvard.edu/handle/1/9823973

Pozharliev L, Gallo González D (2018) Martin Luther King’s assassination in Spain’s NO-DOs and in Bulgaria’s Kinopregledi. In: Vande Winkel R, Chambers C, Jönsson M (eds.) Researching newsreels: local, national and transnational case studies. Palgrave Macmillan, Switzerland, p 93–117

Pustu-Iren K, Sittel J, Mauer R, Bulgakowa O, Ewerth R (2020) Automated visual content analysis for film studies: current status and challenges. Digit Hum Q 14:4

Rasheed Z, Shah M (2003) Scene detection in Hollywood movies and TV shows. In: IEEE computer society conference on computer vision and pattern recognition, 2:II–343. https://doi.org/10.1109/CVPR.2003.1211489

Ravessoud C, Haver G (2016) Art exhibitions through newsreels: an avatar for identity politics (1945–1960). In: Imesch K, Schade S, Sieber S (eds.) Constructions of cultural identities in newsreel cinema and television after 1945. Transcript Verlag, Bielefeld, p 101–116. https://doi.org/10.14361/9783839429754-006

Ren H, Renoust B, Viaud M-L, Melançon G, and Satoh S (2018) Generating ‘Visual Clouds’ from multiplex networks for TV news archive query visualization. In: Proceedings of the International Conference on Content-Based Multimedia Indexing (CBMI) : 1–6. https://doi.org/10.1109/CBMI.2018.8516482

Rodgers J (2014) From Perestroika to Putin: journalism in Russia. In: Media Independence , Routledge, London, p 237–256

Roth-Ey K (2011) Moscow prime time: how the Soviet Union built the media empire that lost the cultural cold war. Cornell University Press, Ithaca

Salmi H (2020) What is digital history? Polity, Medford

Salt B (1974) Statistical style analysis of motion pictures. Film Q 28(1):13–22. https://doi.org/10.2307/1211438

Sampaio S (2022) From propaganda to attraction: reassessing the role of the public in the Newsreel Jornal Português (1938–1951). Hist J Film Radio Telev 42(3):470–492. https://doi.org/10.1080/01439685.2021.1976918

Sanguineti V, Morerio P, Del Bue A, Murino V (2022) Unsupervised synthetic acoustic image generation for audio-visual scene understanding. IEEE Trans Image Process 31:7102–7115. https://doi.org/10.1109/TIP.2022.3219228

Article   ADS   PubMed   Google Scholar  

Schich M (2017) The hermeneutic hypercycle. In: Brockman J (ed.) Know this: today’s most interesting and important scientific ideas. Harper Perennial, New York, p 561–563

Schich M (2019) Cultural analysis situs. ART-Dok eprint https://doi.org/10.11588/artdok.00006347

Senocak A, Oh T-H, Kim J, Yang M-H, Kweon IS (2018) Learning to localize sound source in visual scenes. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition : 4358–4366. https://doi.org/10.1109/CVPR.2018.00458

Snow CP (2001) The two cultures. Cambridge University Press, London (1959)

Sommer NM, Velipasalar S, Hirshfield L, Lu Y, Kakillioglu B (2020) Simultaneous and spatiotemporal detection of different levels of activity in multidimensional data. IEEE Access 8:118205–118218. https://doi.org/10.1109/ACCESS.2020.3005633

Studer L, Alberti M, Pondenkandath V, Goktepe P, Kolonko T, Fischer A, Liwicki M, and Ingold R (2019) A comprehensive study of imagenet pre-training for historical document image analysis. In: Proceedings of the international conference on document analysis and recognition (ICDAR) : 720–725. https://doi.org/10.1109/ICDAR.2019.00120

Treveri Gennari D, Sedgwick J (2022) Five Italian cities: comparative analysis of cinema types, film circulation and relative popularity in the mid-1950s. In: Sedgwick J (ed.) Towards a comparative economic history of cinema, 1930–1970. Springer International Publishing, Cham, p 249–279. https://doi.org/10.1007/978-3-031-05770-0_9

Tsivian Y (2009) Cinemetrics, part of the humanities’ cyberinfrastructure. In: Ross M, Grauer M, Freisleben B (eds.) Digital tools in media studies. analysis and research. An overview. Transcript, Bielefeld, p 93–100

Turoma Sanna, Waldstein Maxim (eds.) (2013) Empire de/centered: new spatial histories of Russia and the Soviet Union. Ashgate, Farnham

Tworek HJS (2019) News from Germany: the competition to control world communications, 1900–1945. News from Germany. Harvard University Press, Harvard, https://doi.org/10.4159/9780674240728

Valverde, FR, Valeria Hurtado J, and Valada A (2021) There is more than meets the eye: self-supervised multi-object detection and tracking with sound by distilling multimodal knowledge. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR) : 11607–11616. https://doi.org/10.1109/CVPR46437.2021.01144

Vande Winkel R, Biltereyst D (2018) Newsreel production, distribution and exhibition in Belgium, 1908-1994. In: Vande Winkel R, Chambers C, Jönsson M (eds.) Researching newsreels: local, national and transnational case studies. Palgrave Macmillan, Switzerland, p 231–246

Veldi M, Bell S, Kuhlmann F (2019) Five-year plan in four: Kolkhoz propaganda in film and documentaries in Estonia. SHS Web Conf 63:10002. https://doi.org/10.1051/shsconf/20196310002

Verhoeven D, Coate B, Zemaityte V (2019) Re-distributing gender in the global film industry: beyond #metoo and #methree. Media Ind J 6(1):135–155. https://doi.org/10.3998/mij.15031809.0006.108

Verhoeven D, Musial K, Palmer S, Taylor S, Abidi S, Zemaityte V, Simpson L (2020) Controlling for openness in the male-dominated collaborative networks of the global film industry. PLOS One 15(6):e0234460. https://doi.org/10.1371/journal.pone.0234460

Article   CAS   PubMed   PubMed Central   Google Scholar  

Wan S, Xu X, Wang T, Gu Z (2021) An intelligent video analysis method for abnormal event detection in intelligent transportation systems. IEEE Trans Intell Transport Syst 22(7):4487–4495. https://doi.org/10.1109/TITS.2020.3017505

Wang D, Zhang S (2022) Unsupervised person re-identification via multi-label classification. Int J Comput Vis 130(12):2924–2939. https://doi.org/10.1007/s11263-022-01680-y

Werenskjold R (2018) Around the world: the first Norwegian newsreel, 1930–1941. In: Vande Winkel R, Chambers C, Jönsson M (eds.) Researching newsreels: local, national and transnational case studies. Palgrave Macmillan, Switzerland, p 51–75

Wevers M (2021) Scene detection in De Boer historical photo collection. In: Proceedings of the 13th international conference on agents and artificial intelligence (ICAART 2021) , 1:601–610. https://doi.org/10.5220/0010288206010610

Wevers M, Smits T (2020) The visual digital turn: using neural networks to study historical images. Digit Scholarsh Hum 35(1):194–207. https://doi.org/10.1093/llc/fqy085

Williams M, Bell J (2021) The Media Ecology Project: Collaborative DH synergies to produce new research in visual culture history. Digit Hum Q 15:1

Winston B (2018) Wofull News from Wales: details at 11. news, newsreels, bulletins and documentaries. In: Vande Winkel R, Chambers C, Jönsson M (eds.) Researching newsreels: local, national and transnational case studies. Palgrave Macmillan, Switzerland, p 15–33

Yang K, Russell B, Salamon J (2020) Telling left from right: learning spatial correspondence of sight and sound. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR) : 9929–9938. https://doi.org/10.1109/CVPR42600.2020.00995

Ye K, Kovashka A (2022) Weakly-Supervised Action Detection Guided By Audio Narration. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops (CVPRW) : 1527–1537. https://doi.org/10.1109/CVPRW56347.2022.00159

Zeiler M (1999) Modeling our world: the ESRI guide to geodatabase design . ESRI Press

Zhang Q, Chan AB (2022) 3D crowd counting via geometric attention-guided multi-view fusion. Int J Comput Vis 130(12):3123–3139. https://doi.org/10.1007/s11263-022-01685-7

Zhao B, Gong M, Li X (2021) AudioVisual video summarization. IEEE Transactions on Neural Networks and Learning Systems : 1–8. https://doi.org/10.1109/TNNLS.2021.3119969

Download references

Acknowledgements

European Union Horizon2020 research and innovation programme ERA Chair project for Cultural Data Analytics CUDAN (Project no. 810961); National programme of the Ministry of Education and Research of the Republic of Estonia for April 2022–February 2023 (Project no. EKKD77); Estonian Research Council Public Value of Open Cultural Data (Project no. PRG1641); European Union Horizon Europe research and innovation programme CresCine—Increasing the International Competitiveness of the Film Industry in Small European Markets (Project no. 101094988).

Author information

Authors and affiliations.

Tallinn University, Tallinn, Estonia

Mila Oiva, Ksenia Mukhina, Vejune Zemaityte, Andres Karjus, Mikhail Tamm, Tillmann Ohm, Mark Mets, Mar Canet Sola & Maximilian Schich

Estonian Business School, Tallinn, Estonia

Andres Karjus

King’s College London, London, UK

Daniel Chávez Heras

University of Tartu, Tartu, Estonia

Helena Hanna Juht

You can also search for this author in PubMed   Google Scholar

Contributions

MO, KM, VZ, TO, MT, MM, AK, DCH, MCS, and MS designed research, KM and MO acquired the data, MO, KM, VZ, MT, TO, and HHJ prepared the data, MO, KM, VZ, AK, MT, TO, MM, and MS performed the analysis, and MO, VZ, and MS wrote the manuscript. MO is the first author. MS is the supervising author.

Corresponding author

Correspondence to Mila Oiva .

Ethics declarations

Ethical approval.

This article does not contain any studies with human participants performed by any of the authors.

Informed consent

Competing interests.

The authors declare no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary material, supplementary table 1, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Oiva, M., Mukhina, K., Zemaityte, V. et al. A framework for the analysis of historical newsreels. Humanit Soc Sci Commun 11 , 530 (2024). https://doi.org/10.1057/s41599-024-02886-w

Download citation

Received : 14 July 2023

Accepted : 27 February 2024

Published : 25 April 2024

DOI : https://doi.org/10.1057/s41599-024-02886-w

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

The independent source for health policy research, polling, and news.

What are the Implications of the Dobbs Ruling for Racial Disparities?

Latoya Hill , Samantha Artiga , Usha Ranji , Ivette Gomez , and Nambi Ndugga Published: Apr 24, 2024

• Issue Brief

Note: Figures 12 and 13 were updated on April 26, 2024.

Introduction

The June 2022 Supreme Court ruling in the case Dobbs v. Jackson Women’s Health Organization has significant implications for racial disparities in health and health care. The decision overturned the longstanding Constitutional right to abortion and eliminated federal standards on abortion access that had been in place for nearly 50 years in all states across the country. As a result of Dobbs , large swaths of the country lack abortion access, with a disproportionate impact on those residing in the South and Midwest.

As of April 2024, 14 states have implemented abortion bans, 11 states have placed gestational limits on abortion between 6 and 22 weeks, and 25 states and the District of Columbia provide broader access to abortions after 22 weeks gestation. (This reflects Arizona being counted in the gestational limits category, as implementation of a recently upheld Civil War-era law banning nearly all abortions in the state is still pending amid ongoing court actions.)

Pregnant women seeking abortion that reside in states that prohibit or restrict abortions either have to travel out of state or try to obtain medication abortion pills via a telehealth appointment with an out-of-state clinician, but these options are not accessible to everyone. Some women may turn to self-managed abortions, but some will not be able to obtain an abortion and have to continue a pregnancy they do not want. Additionally, there have been reports of clinicians in states with bans and early gestational limits leaving their states due to the restrictions and criminalization for care that they provide, potentially exacerbating provider shortages in some areas.

With these state-level restrictions in place, people of color residing in those states may face disproportionately greater challenges accessing abortions due to longstanding underlying social and economic inequities, which could exacerbate existing disparities in maternal and infant health. This analysis examines the implications of state restrictions on abortion coverage for racial disparities in access to care and health outcomes. It is based on KFF analysis of data from the Centers for Disease Control and Prevention (CDC), American Community Survey (ACS), Behavioral Risk Factor Surveillance Survey (BRFSS), and Survey of Household Economics and Decisionmaking (SHED) (see Methods ). Throughout this brief we refer to “women” but recognize that some individuals who have abortions do not identify as women, including transgender. Key takeaways include the following:

• Black and American Indian and Alaska Native (AIAN) women ages 18-49 are more likely than other groups to live in states with abortion bans and restrictions . About six in ten Black (60%) and AIAN (59%) women ages 18-49 living in states with abortion bans or restrictions compared with just over half (53%) of White, less than half of Hispanic (45%), and about three in ten Asian (28%) and Native Hawaiian or Pacific Islander (NHPI) (29%) women ages 18-49.
• Many groups of women of color have higher uninsured rates compared to their White counterparts, and, a cross racial and ethnic groups, uninsured rates are higher in states with abortion bans or restrictions than in those that provide broader abortion access . Among women ages 18-49, roughly a fifth of AIAN (22%) and Hispanic (21%) women are uninsured as are 14% of NHPI women and 11% of Black women compared with less than one in ten (7%) of White women. Moreover, uninsured rates for women ages 18-49 are at least twice as high in states that banned abortion compared to those in states with broader access for White (10% vs. 5%), Hispanic (33% vs. 15%), Black (14% vs. 7%), and Asian (10% vs. 5%) women, and nearly three times higher for NHPI women (29% vs. 10%).
• Women of color have more limited financial resources and transportation options than White women, making it more difficult for them to travel out-of-state for an abortion . Some may also face linguistic barriers and have immigration-related fears that create additional challenges to accessing abortions.
• The bans and restrictions on abortions may widen the already stark racial disparities in maternal health, especially since some states do not explicitly have exceptions that allow abortion services when pregnancy is jeopardizing a woman’s health . The restrictions may also contribute to growing provider shortages in some areas, as clinicians are responding to concerns about criminalization and prohibited from offering the full spectrum of pregnancy care. Moreover, abortion restrictions may have negative economic consequences on families and put pregnant people at increased risk for criminalization.

While there have been large inequities in abortion access for many years, the Dobbs ruling opened the door to widening those differences further. Black and AIAN women are more likely to live in states with abortion bans or restrictions. While data on the impact of Dobbs to date on health outcomes is limited to date, many indicators suggest that the ruling may exacerbate longstanding large disparities in maternal and infant health. The issue also has moved to the forefront of policy debates in the U.S. Sixteen percent of women voters, rising to 28% of Black women voters, say abortion is the “ most important issue ” to their vote in the 2024 presidential election.

How do Abortion Rates Vary by Race and Ethnicity?

Data on abortions by race and ethnicity are limited . The federal Abortion Surveillance System from the CDC has been providing annual national and state-level statistics on abortion for decades, based on data that is voluntarily reported by states, DC, and New York City. While most states participate, one notable exception is California, which has many protections for abortion access and is one of the most racially diverse states in the nation. Furthermore, availability of data by race and ethnicity varies among states. The most recent data in the Abortion Surveillance System, from 2021, only includes racial and ethnic data from 31 states and DC and is generally only available for White, Black, and Hispanic women. While we present the data from the Abortion Surveillance System in this brief, we recognize these limitations.

Prior to Dobbs , the abortion rate was higher among Black and Hispanic women compared to their White peers . As of 2021, the abortion rate was 28.6 per 1,000 women among Black women, compared to 12.3 per 1,000 among Hispanic women, and 6.4 per 1,000 among White women (Figure 1). Data for other racial and ethnic groups were not available. The vast majority of abortions across racial and ethnic groups are in the first trimester . Approximately eight in ten abortions among White (82%), Hispanic (82%), and Black women (80%) occur by nine weeks of pregnancy. While data on the number of abortions post- Dobbs has been released by both the #WeCount project from the Society for Family Planning and the Guttmacher Institute’s Monthly Abortion Provision Study , neither sets of data have reported demographic characteristics of abortion patients.

There are many reasons why abortion rates are higher among some women of color . As discussed below, Black, Hispanic, American Indian and Alaska Native (AIAN), and Native Hawaiian or Pacific Islander (NHPI) women have more limited access to health care, which affects their access to contraception and other sexual health services that are important for pregnancy planning. Data show that contraception use is higher among White women (69%) compared to Black (61%) and Hispanic (61%) women. Some women of color live in areas with more limited access to comprehensive contraceptive options. In addition, the health care system has a long history of racist practices targeting the sexual and reproductive health of people of color, including forced sterilization, medical experimentation, the systematic reduction of midwifery, just to name a few . Many women of color also report discrimination by providers, with reports of dismissive treatment, assumption of stereotypes, and inattention to conditions that take a disproportionate toll on women of color and certain conditions, such as uterine fibroids . These factors have contributed to medical mistrust, which some women cite as a reason that they may not access contraception. In addition, inequities across broader social and economic factors — such as income, housing, safety and education—that drive health, often referred to as social determinants of health, affect decisions related to family planning and reproductive health.

How Do State Abortion Policies Vary Across Racial and Ethnic Groups?

Overall, 16.3 million or 25% of women ages 18-49 in the US live in one of the 14 states where abortion is banned, and another 16.9 million, or 26%, live in one of the 11 states with gestational limits between 6 and 22 weeks LMP. The remaining 32.8 million, or roughly 50%, live in states that provide broader access to abortions.

White, Black, and American Indian and Alaska Native women account for larger shares of women ages 18-49 in states that have banned or limited abortion access compared to states that provide broader access to abortion . Most of the states that have banned or restricted abortion are in the South, where more than half of the Black population and roughly a third of the White (36%) and AIAN (31%) population reside. In contrast, Hispanic and Asian women make up larger shares of women ages 18-49 in states that provide broader access to abortion compared to states with abortion bans or limits. (See Appendix Table B for the racial and ethnic distribution of women ages 18-49 by state).

Six in ten of Black (60%) and AIAN (59%) women ages 18-49 live in states with abortion bans or restrictions (Figure 3) . Just over half (53%) of White women ages 18-49 live in states with bans or restrictions, while less than half of Hispanic (45%) and about three in ten Asian (28%) and NHPI (29%) women ages 18-49 live in these states. Of note, in April 2024, the Arizona State Supreme Court upheld a Civil War era law banning nearly all abortions in the state. While that law is not currently in effect, if it were to go into effect in the future, the share of AIAN women living in a state with an abortion ban would rise from about three in ten (31%) to about four in ten (41%), and the share of Hispanic women living in a state with an abortion ban would increase from 24% to 28%.

How do potential barriers to accessing abortions vary by race and ethnicity?

Variation in abortion policies by state due to the Dobbs decision will likely result in women of color facing disproportionate access barriers since they face underlying disparities in health coverage and have more limited financial resources that may make it challenging to obtain an abortion out-of-state or via telehealth.

Health Coverage

Lack of health insurance limits women’s access to a broad range of health services, including contraception and pregnancy care, and leaves them at risk for significant out of pocket expenses for care. However, having coverage does not guarantee that it includes abortion benefits. In general coverage of abortion is more limited than for many other common health services. Some states prohibit coverage of abortion in state-regulated private insurance plans, and federal law bars the use of federal dollars for abortion, including in Medicaid, the national health coverage program for low-income individuals.

AIAN, Hispanic, NHPI, and Black women between ages 18-49 have higher uninsured rates compared to their White counterparts . Among women in this age group, roughly a fifth of AIAN (22%) and Hispanic (21%) women are uninsured as are 14% of NHPI women and 11% of Black women. In contrast, less than one in ten (7%) of White women lack insurance (Figure 4). These differences in uninsured rates are driven by lower rates of private coverage among these groups. Medicaid coverage helps to narrow these differences but does not fully offset them.

Across racial and ethnic groups, uninsured rates for women ages 18-49 in states that have banned or limited abortion are higher than rates in states where abortion is available beyond 22 weeks . Overall, 16% of women ages 18-49 in states that have banned abortion are uninsured compared to 12% in states that have gestational limits on abortions less than 22 weeks and 8% in states that have broader access to abortions. Uninsured rates for women ages 18-49 are at least twice as high in states that banned abortion compared to those in states with broader access for White (10% vs. 5%), Hispanic (33% vs. 15%), Black (14% vs. 7%), and Asian (10% vs. 5%) women, and nearly three times higher for NHPI women (29% vs. 10%) (Figure 5). However, even in states where abortion is not banned, many women do not have coverage, and uninsured rates remain higher for AIAN, Hispanic, and NHPI women compared to White women.

AIAN, Black, NHPI, and Hispanic women are more likely than their White counterparts to be covered by Medicaid, which provides limited coverage for abortions . For decades, the Hyde Amendment has prohibited the use of federal funds for coverage of abortion under Medicaid, except in cases of rape, incest, or life endangerment for the pregnant person. States can choose to use state funds to pay for abortions under Medicaid in other instances. However, among the 36 states that do not ban abortion, 17 use state funds to pay for abortions beyond the Hyde limitations for Medicaid enrollees. The other 19 states and DC continue to follow the Hyde limits, meaning women in these states covered by Medicaid likely must pay out of pocket for an abortion unless they meet the narrow circumstances of the Hyde Amendment.

Social and Economic Access Barriers

Women of color have more limited financial resources and transportation options than White women, making it more difficult for them to travel out-of-state for an abortion. The median self-pay cost of obtaining an abortion exceeded $500 in 2021, but costs can vary depending on the type of abortion, location, and if an individual has coverage. Traveling out of state raises the cost of abortion due to added costs for transportation, accommodation, and childcare. Moreover, it may result in more missed work, meaning greater loss of pay. Data suggest that women of color would have more difficulty than White women affording these increased costs and may face other barriers that could prevent them from traveling to obtain an abortion and instead turning to self-managed abortions or continuing the pregnancies.

Overall, AIAN (48%), Black (43%), NHPI (41%) and Hispanic (40%) women ages 18-49 are nearly twice as likely as their White counterparts (24%) to have low incomes (below 200% of the federal poverty level or $46,060 for a family of three as of 2022) (Figure 6) . Moreover, across most racial and ethnic groups, women in states that have banned abortion are more likely to have low incomes than women in states that allow abortions beyond 22 weeks. For example, 48% of NHPI women in states that have banned abortion have low incomes compared to 38% of NHPI women in states where abortion is available after 22 weeks gestation. (See Appendix Table C for state-level data on the share of women who are low-income by race and ethnicity.)

Over half of Hispanic (57%) and Black women (58%) ages 18-49 could not cover an emergency expense of at least $500 using their current savings compared to 36% of White women in this age group (Figure 7) . (Data for this measure were not available for other racial groups.) Women who have fewer resources for an emergency expense may be more likely to seek assistance from an abortion fund , which help cover the costs of abortions for people who cannot afford them. However, abortion funds are not able to keep up with the demand and support all those seeking assistance.

Black women ages 18-49 are more likely than their White counterparts to live in a household without access to a vehicle (12% vs. 4%), and Asian and AIAN women in this age group are more likely than White women to lack vehicle access (9% and 8%, respectively, vs. 4%) (Figure 8) . Hispanic and NHPI women are also more likely than White women to lack vehicle access, although the difference is smaller (6% and 6%, respectively, vs 4%). Research shows that out-of-state travel for abortion care has risen significantly since Dobbs, but women without vehicle access may face greater challenges to traveling out of state.

Immigration-related fears make some women reluctant to travel out of state for an abortion . Among women ages 18-49, about one-third of Asian women (33%) and roughly a quarter of Hispanic (24%) and NHPI (22%) women are noncitizens, who include lawfully present and undocumented immigrants (Figure 9). Many citizen women may also live in mixed immigration status families, which may include noncitizen family members. Noncitizen women and those living in mixed immigration status families may fear that traveling out of state could put them or a family member at risk for negative impacts on their immigration status or detention or deportation, especially in states that have moved to criminalize abortions. For example, some states have enacted laws that make it illegal to “ aid or abet ” someone in obtaining an abortion while some are trying to make it illegal to take a minor across state lines to obtain an abortion.

Differences in language barriers and access to technology may also contribute to racial disparities in abortion access . Roughly a quarter of Hispanic (26%) and Asian (25%) women ages 18-49 speak English “less than very well,” as do one in ten NHPI women (10%) compared to just 1% of White women (Figure 10). This can affect their ability to find information about abortions and locate a clinic that offers abortion services. In a national KFF survey of women conducted just before the Dobbs ruling, nearly three in ten Hispanic women (29%) said if they needed an abortion, they did not know where to go or find the information, higher than other groups. Internet access is another important factor for finding information about abortion care and also for telehealth appointments, which comprise a growing share of abortion care. Among women ages 18-49, 8% of AIAN and 6% of NHPI (6%) women live in a household without internet access, compared to 2% of White women (Figure 10).

What are the Potential Implications of Abortion Restrictions on Racial Disparities in Health, Finances, and Criminal Penalties?

Stark racial disparities in maternal and infant health predate the Dobbs decision but may widen due to the new restrictions on abortions since abortion services can be a key factor in managing pregnancy complications and emergencies that can lead to poor outcomes. Data suggest that the abortion restrictions may also contribute to growing provider shortages in some areas, which may increase access challenges and have negative impacts on health. Moreover, abortion restrictions may have negative economic consequences on families and put people at increased risk for criminalization.

Maternal Health

Prior to the Dobbs ruling there were already significant racial disparities in pregnancy-related and infant mortality, which may widen due to abortion restrictions . NHPI, Black and AIAN people are more likely to die while pregnant or within a year of the end of pregnancy compared to White people (62.8, 39.9 and 32.0 per 100,000 births vs. 14.1 per 100,000 births) (Figure 11). Restrictions on access to abortions limit options to terminate pregnancies for medical reasons. While all state bans have some limited exceptions to preserve the life of pregnant women, the language of these exceptions is vague and narrow, and far fewer have health exceptions. This means that some people have been forced to remain pregnant even when the pregnancy is threatening their health , which could further widen disparities. One study estimated that a total abortion ban in the U.S. would increase the number of pregnancy-related deaths by 21% for all women and 33% among Black women.

There also are racial disparities in certain birth risks and adverse birth outcomes which may be exacerbated by the abortion restrictions . Specifically, as of 2022, higher shares of births to Hispanic, Black, AIAN and NHPI people were among those who received late or no prenatal care, or were preterm, or low birthweight, compared to White people (Figure 12). Timely prenatal care is particularly important for people with higher-risk pregnancies, yet research suggests that restrictive abortion policies may be causing people to start prenatal care later in pregnancy, which is already a concern for women of color who are more likely to experience delays in prenatal care initiation. Births among Asian people were also more likely to be low birthweight than those of White people. Moreover, while the birth rate among teens has been declining over time for all groups, the rate for Black, Hispanic, AIAN, and NHPI teens was over two times higher than the rate among White and Asian teens in 2021 (Figure 13). Research has also found that state-level abortion restrictions that were in place prior to Dobbs were associated with disproportionately higher rates of adverse birth outcomes, including preterm birth, for Black individuals, and that inequities widened as states became more restrictive.

Abortion bans and restrictions limit care for people experiencing a pregnancy loss, which some groups of women of color are at higher risk of experiencing compared to their White counterparts . Pregnancy loss, which includes miscarriage and stillbirth, is common , occurring in up to an estimated 20% of all pregnancies. Data on racial and ethnic disparities in miscarriage is limited, but research shows that the rates of fetal mortality (fetal demise following 20 weeks of gestation) are higher among Black, AIAN, and NHPI women compared to White women (Figure 14). While some miscarriages, particularly earlier in pregnancy, pass without any medical intervention, some people seek medical care to complete a miscarriage and/or because their health may worsen with the continuation of an unviable pregnancy. Almost all medications and procedures used to manage miscarriages and stillbirths are identical to those used in abortions. As a result, clinicians may hesitate to provide care even when medically indicated because of concerns they could be conflated with providing an abortion and therefore risk criminalization or penalties as a result. Since the Dobbs ruling, there have been several high-profile cases of people experiencing pregnancy losses who could not obtain timely miscarriage care due to state abortion bans, jeopardizing their health as a result. In KFF’s national survey of OBGYNs after the Dobbs decision, more than half (55%) of OBGYNs practicing in states where abortion is banned said their ability to practice within the standard of care has worsened since Dobbs .

In states where abortion is banned or severely restricted, the number of women forced to continue a pregnancy is likely to rise, with data suggesting disproportionate increases among women of color . While it is relatively early to see the impact of the Dobbs ruling on births, initial research suggests that birth rates could increase as a result. One study to date has estimated that there have been approximately 32,000 “ additional ” births as a result of the ruling, primarily concentrated in states that have banned abortions and with a disproportionate effect among people of color. A study in Texas , which had implemented a ban on abortions after six weeks gestation starting September 2021 (prior to Dobbs ), found a 2% rise in the state’s fertility rate after the law’s implementation, with the sharpest increase among Hispanic women (8%).

Provider Access and Shortages

The Dobbs decision may exacerbate health care workforce shortages, particularly among clinicians providing obstetric and gynecologic care . State-level abortion bans criminalize clinicians who provide abortion care, and this has cascading effects on other aspects of maternity care. Even prior to Dobbs , there were concerns about workforce shortages in maternity care. The estimates that more than 5 million women of reproductive age in the U.S. live in counties that have few or no obstetric providers, with the largest gaps in rural communities as well as areas with higher rates of poverty, and larger shares of Black women. Many of these areas are in states with abortion bans and gestational restrictions, and there are reports of clinicians leaving these states because they are prohibited from and criminalized for offering the full scope of services they trained for and that comport with medical standards. Abortion restrictions may also affect the pipeline of new clinicians. A few studies to date, have found declines in US medical school graduates applying to OBGYN residency positions in states with abortion bans. While all positions were filled and the changes to date have been relatively small, they could suggest that future clinicians may prefer not to practice in states that ban abortion, potentially widening existing gaps in workforce capacity.

Many OBGYNs say that the Dobbs decision has had a negative impact on racial and ethnic inequities and the broader field of maternity care . In a national KFF survey , seven in ten OBGYNs say that since the Dobbs decision, racial and ethnic inequities in maternal health (70%) as well as management of pregnancy-related medical emergencies (68%) have gotten worse. Over half think that the ability to attract new OBGYNs to the profession has worsened (55%) and 64% think the same about pregnancy-related mortality (Figure 15).

Economic Circumstances

Denying women access to abortion services has negative economic consequences . Many women who are not able to obtain abortions will have children that they hadn’t planned for and face the associated costs of raising a child. In addition to the direct costs, lack of abortion access can affect women’s longer-term educational and career opportunities. Research from the Turnaway Study , which examined the impact of an unwanted pregnancy on women’s lives, found a range of negative economic effects of abortion denials, including higher poverty rates, financial debt , and poorer credit scores among women who were not able to obtain abortions compared to women who received abortions. The study also found negative socioeconomic impacts for the children born to women who were denied abortions, which may exacerbate existing racial disparities in income. Poverty rates are already much higher among children of color than White children, and research shows children in families with lower incomes experience negative long-term outcomes, including lower earnings and income, increased use of public assistance, greater likelihood of committing crimes, and more health problems.

Criminalization

People of color may be at increased risk for criminalization in the post- Roe environment . A long history of racism in judicial policy in this country has led to disproportionately higher rates of criminalization among people of color and is likely to grow as abortion care is criminalized. Prior to the Dobbs ruling, there were already cases of women criminalized for their own miscarriages, stillbirths, or infant death, due in part to the establishment of laws that protect and prioritize “ fetal personhood .” The women charged were disproportionately women with lower incomes, Black women, and women living in southern states that have subsequently banned or greatly restricted abortion access. None of the state-level abortion bans specifically criminalize women for getting an abortion, but fetal personhood laws can conflate miscarriage and abortion. For example, in one high-profile case, Brittany Watts is an Ohio woman who faced criminal charges after she had a miscarriage at home in Fall 2023. While Ms. Watts sought medical care, other pregnant people experiencing a miscarriage or other complications may be deterred from seeking care, since treatment could be conflated with an abortion, putting their own health at risk as a result. Furthermore, many accusations of fetal harm are initiated by health care providers . State laws that penalize people who aid and abet abortion access and those that grant fetal personhood can perpetuate the culture of criminalizing pregnancy, particularly among communities of color.

• Women's Health Policy
• Racial Equity and Health Policy
• Race/Ethnicity

Also of Interest

• Key Facts on Abortion in the United States
• Abortions Later in Pregnancy in a Post-Dobbs Era
• The Hyde Amendment and Coverage for Abortion Services Under Medicaid in the Post-Roe Era
• Abortion in the United States Dashboard

Asking the better questions that unlock new answers to the working world's most complex issues.

Trending topics

AI insights

EY podcasts

EY webcasts

Operations leaders

Technology leaders

Marketing and growth leaders

Cybersecurity and privacy leaders

Risk leaders

EY Center for Board Matters

EY helps clients create long-term value for all stakeholders. Enabled by data and technology, our services and solutions provide trust through assurance and help clients transform, grow and operate.

Artificial Intelligence (AI)

Strategy, transaction and transformation consulting

Technology transformation

Tax function operations

Climate change and sustainability services

EY Ecosystems

Supply chain and operations

EY Partner Ecosystem

Explore Services

We bring together extraordinary people, like you, to build a better working world.

Experienced professionals

MBA and advanced-degree students

Student and entry level programs

Contract workers

EY-Parthenon careers

Discover how EY insights and services are helping to reframe the future of your industry.

Case studies

Energy and resources

How data analytics can strengthen supply chain performance

13-Jul-2023 Ben Williams

How Takeda harnessed the power of the metaverse for positive human impact

26-Jun-2023 Edwina Fitzmaurice

Banking and Capital Markets

How cutting back infused higher quality in transaction monitoring

11-Jul-2023 Ron V. Giammarco

At EY, our purpose is building a better working world. The insights and services we provide help to create long-term value for clients, people and society, and to build trust in the capital markets.

EY is now carbon negative

19-Sep-2022 Carmine Di Sibio

Our commitment to audit quality

13-Nov-2023 Julie A. Boland

No results have been found

Recent Searches

BEPS 2.0: as policies evolve, engagement is key

It remains to be seen whether the US will align its tax law with the OECD/G20’s global BEPS 2.0 rules. MNEs will feel the impact in 2024. Learn more.

How GenAI strategy can transform innovation

Companies considering or investing in a transformative GenAI strategy should tie generative artificial intelligence use cases to revenue, cost and expense. Learn more

Top five private equity trends for 2024

Read about the five key trends private equity firms will emphasize in 2024 as they create value

Select your location

close expand_more

Banking & Capital Markets

The bank of the future will integrate disruptive technologies with an ecosystem of partners to transform their business and achieve growth.

Disruption is creating opportunities and challenges for global banks. While the risk and regulatory protection agenda remains a major focus, banks must also address financial performance and heightened customer and investor expectations, as they reshape and optimize operational and business models to deliver sustainable returns. Innovation and business-led transformation will be critical for future growth. To remain competitive and relevant, every bank must embrace disruption and strategically build a better ecosystem — not a bigger bank.

Our worldwide team of industry-focused assurance, tax, transaction and consulting professionals integrates sector knowledge and technical experience. We work with clients to navigate digital innovation, new business models and ecosystem partnerships, helping banks become the nimble, responsive organizations that customers demand.

Five priorities for harnessing the power of GenAI in banking</p> "> Five priorities for harnessing the power of GenAI in banking

What to expect from global financial services in 2024 — Americas and EMEIA

In this webcast for Americas and EMEIA audiences, the EY Global Regulatory Network will discuss the direction of travel for regulators across key areas and how to prepare for what's coming.

Our latest thinking on Banking & Capital Markets

Impacts of Central Clearing of US Treasuries and Repo

In this webcast, panelists will discuss key themes and high-level requirements of the US Treasury and repo central clearing rules.

Can core platform modernization position a bank for future success?  

Case study: how one regional bank used core platform modernization to build a strong foundation for future profitability.

The case for a modern transaction banking platform

The evolution of corporate treasury management needs presents an opportunity for corporate banks. Learn from an industry approach.

How to transition from a tactical to strategic adoption of ISO 20022

With ISO 20022 adoption lagging amid competing global deadlines, a successful migration may hinge on changing from a tactical to a strategic mindset.

How Gen Z’s preference for digital is changing the payments landscape

EY survey shows Gen Z embraces simple, seamless payment methods. Learn more.

How can financial institutions modernize their fair-lending practices?

FIs that disregard fair banking are lagging behind FIs that enhance compliance procedures, lending models and data analytics to become more compliant. Read more.

Digital identity opportunities in financial services

Exploring the policy and regulatory trends shaping digital identity and opportunities for financial services companies in a changing payment landscape.

Explore our Banking & Capital Markets case studies

Using AI to augment pricing intelligence for banks

How an AI-powered digital tool, Smart Advisor (SA), helped one bank deliver better client service while maximizing value creation.

How a global FinTech captured growth in the SME segment

A global Fintech captured growth in an opportunistic SME segment with a differentiated, holistic strategy. Learn more in this case study.

Using AI to improve a bank’s agent effectiveness

Leveraging the power of AI and machine learning, one bank mined sales agents’ calls for performance-boosting insights. Learn more in this case study.

After cloud migration, investment bank sees potential for big dividends

A leading investment bank sought to move vital assets to the clouds by building an experienced, cross-functional team. Find out how.

How digital transformation is redesigning trade finance

Banks that adopt an agile, design-based approach to digital transformation can boost the success of their trade finance functions.

How to transform product development to outperform the competition

EY Nexus is a cloud-based platform offering access to the most advanced technologies to launch new products, businesses and services.

How EY can help

Capital Markets Services

Know how our Capital Markets consulting team can help your business grow, manage costs and meet regulatory requirements.

Consumer banking and wealth services

EY consumer banking and wealth technology solutions are designed to drive operational excellence and profitable growth. Learn more.

Corporate, Commercial and SME Banking services

Our Corporate, Commercial and SME (CCSB) Banking services team can help your business navigate through rising market expectation. Learn more.

Cost transformation

EY cost transformation teams help banks to optimize profits and fund transformation. Find out more.

Consumer lending services

Our consumer lending team can help navigate the complexities of unique lending propositions. Find out how.

EY Nexus for Banking

A transformative solution that accelerates innovation, unlocks value in your ecosystem, and powers frictionless business. Learn more.

Finance transformation

We help clients transform finance functions to be a strategic business partner for the business via value creation and controllership activities.

EY Financial Crime solutions

Our skilled teams, operational efficiencies enabled by innovative technology and flexible global delivery service centers can help you manage financial crime risk in a cost-effective, sustainable way.

Financial services risk management

Discover how EY can help the banking & capital markets, insurance, wealth & asset management and private equity sectors tackle the challenges of risk management.

IBOR transition services

EY helps global institutions prepare for the imminent transition away from Interbank Offered Rates (IBORs) to Alternate Reference Rates (ARRs). We also play a leading role in supporting regulators, trade associations and others to increase awareness and education.

Open banking services

Our open banking professionals can help your business maintain a trusted and secure open banking ecosystem while managing its risks. Learn more.

Payment services

Our payments professionals can help your business enhance innovation, drive growth and improve performance. Find out more.

Third-party risk management services

Discover how EY's Third Party Risk Management team can enable your business to make better decisions about the third parties they choose to work with.

Direct to your inbox

Stay up to date with our Editor‘s picks newsletter.

The Banking & Capital Markets team

Enjoys traveling with family, and coaching his daughters’ basketball and soccer teams. Enjoys running and playing basketball and golf.

Lee Ann Lednik

People-focused leader committed to building trust and transparency amid increasing complexity. Passionate working mom of three. Aspiring photographer. Avid sports fan.

David Kadio-Morokro

Passionate about technology, innovation, and leading EY people to solve clients’ most challenging problems.

Heidi Boyle

Passionate about helping people thrive in the workplace and creating a sense of belonging for all. Writer. Musician. Cooking enthusiast.

Seasoned financial services professional. Resides in Massachusetts with her husband and three children.

Kellen Maia de Sá

Collaborator and problem-solver with the desire to do the right thing. Leads efforts to help financial services clients with the disruption and impact of COVID-19.

Terry Cardew

Builds trust by helping banks solve business issues and stay competitive. Devoted husband. Father of six. Avid skier. NY Giants and Yankees fan. Supporter of The Fresh Air Fund and Lynne’s Kids.

• Connect with us
• Our locations
• Do Not Sell or Share My Personal Information
• Legal and privacy
• Accessibility
• Open Facebook profile
• Open X profile
• Open LinkedIn profile
• Open Youtube profile

EY refers to the global organization, and may refer to one or more, of the member firms of Ernst & Young Global Limited, each of which is a separate legal entity. Ernst & Young Global Limited, a UK company limited by guarantee, does not provide services to clients.