# How to spot misleading graphs in scientific papers ## Quick Answer - Examine every axis for truncation, non-zero baselines, and inconsistent scaling before trusting any visual trend in a research figure. - Cross-check figure claims against the paper's raw data, methods section, and reporting guidelines to verify the graph tells the same story as the numbers. - Remember that a graph can be technically accurate yet still mislead through cherry-picked time windows, inappropriate chart types, or omitted error bars. ## Why Graph Integrity Matters in Scientific Reading Scientific papers communicate findings through a combination of text, tables, and figures. Graphs serve a special role because they compress large amounts of numerical information into a visual format that readers can process quickly. This compression is also the source of vulnerability. A figure that appears convincing at a glance may contain design choices that exaggerate differences, hide variability, or imply relationships the data do not support. For students learning to read research papers, for laboratory staff evaluating protocols, for animal owners assessing veterinary studies, and for life-science professionals making evidence-based decisions, the ability to critically evaluate figures is a core skill. The stakes are practical. A misleading graph in a paper about feed additives could lead a farmer to adopt an ineffective product. A misleading graph in a clinical study could lead a researcher to build an experiment on a false foundation. A misleading graph in a policy review could shape regulations that affect animal welfare. The [Committee on Publication Ethics Core Practices](https://publicationethics.org/core-practices) describe the responsibilities of journals and authors in maintaining research integrity. These practices cover data handling, peer review, and misconduct procedures. They establish that authors have an obligation to represent their findings honestly. Readers, however, cannot assume every published figure meets that standard. Active evaluation is necessary. This article provides a practical framework for spotting misleading graphs. It covers the common visual manipulation tactics, explains how to check a figure against the underlying evidence, and offers a checklist you can apply to any graph in any scientific paper. ## At a Glance: Common Graph Manipulation Tactics | Tactic | What It Looks Like | Why It Misleads | How to Detect It | |--------|-------------------|-----------------|------------------| | Truncated Y-axis | Axis starts at a value other than zero, often near the data range | Exaggerates small differences between groups | Check the axis minimum and maximum values against the data range | | Cherry-picked time window | Graph shows only a favorable subset of the full study period | Hides trends that contradict the authors' conclusion | Compare the figure's time range with the methods description | | Inappropriate chart type | Bar chart for continuous data, pie chart for many categories, line graph for categorical data | Implies relationships or comparisons the data cannot support | Ask whether the chart type matches the variable type and research question | | Missing error bars | Points or bars shown without measures of variability | Makes differences appear more reliable than they are | Look for standard deviation, standard error, or confidence intervals | | Inconsistent scaling | Different axis ranges across panels or figures | Prevents direct visual comparison between groups or time points | Check whether all panels use the same axis limits | | Dual axes with different scales | Two series plotted on left and right axes | Creates false visual correlations between unrelated variables | Examine both axis scales and consider whether the relationship is meaningful | | Omitted data points | Outliers or conflicting observations removed without explanation | Produces a cleaner but inaccurate picture | Compare the figure with supplementary data or raw data files | | 3D effects and perspective distortion | Angled bars, shadows, or rotated charts | Distorts the visual size of data values | Prefer simple two-dimensional charts and check the actual numeric labels | ## Core Principles for Evaluating Figures ### Understand the Research Question First Before you evaluate a graph, identify the research question the paper is trying to answer. A figure that looks misleading may actually be appropriate for a specific analytical purpose. Conversely, a figure that looks clean may be answering a different question than the one stated in the abstract. Read the methods section before studying the figures. The methods should describe the study design, the variables measured, the statistical tests used, and the software employed for analysis. The [EQUATOR Network](https://www.equator-network.org/) provides reporting guidelines for many study types. These guidelines tell authors what information they must include so that readers can assess the validity of the research. When a paper follows an appropriate reporting guideline, the figures should be interpretable in the context of a complete methods description. ### Check the Axis Mathematics The most common graph manipulation involves the axes. A truncated y-axis starts at a value close to the data range instead of at zero. This makes small differences between groups appear large. For example, a bar chart showing survival rates of 90 percent and 92 percent will look dramatic if the y-axis starts at 88 percent. The same data plotted with a zero baseline will show that the two groups are nearly identical. Not every graph must start at zero. Line graphs showing change over time, for example, often use a non-zero baseline to show meaningful variation. The key question is whether the axis choice obscures or exaggerates the true magnitude of the effect. Check the axis labels, the tick marks, and the range. Ask whether the visual impression matches the numeric difference. ### Look for the Variability A graph that shows only means or medians without any measure of spread is incomplete. Error bars, confidence intervals, or shaded regions indicate how much variation exists within each group. Without this information, you cannot tell whether the observed difference is meaningful or whether it falls within normal variation. When error bars are present, check what they represent. Standard deviation describes the spread of the data. Standard error describes the precision of the mean estimate. Confidence intervals describe the range within which the true value likely falls. These measures answer different questions. A figure that labels its error bars clearly allows you to interpret the results correctly. A figure that omits this information forces you to rely on the authors' interpretation. ### Examine the Data Coverage Graphs can mislead by showing only part of the data. A study that collected data over two years might present a figure showing only the final three months. A study that measured multiple outcomes might present only the outcomes that showed significant differences. This practice, sometimes called cherry-picking, produces figures that are technically accurate but substantively misleading. Compare the figure with the methods description. The methods should state the full study period, the complete set of outcomes measured, and the criteria for including or excluding data points. If the figure covers a shorter period or a narrower set of outcomes than the methods describe, ask why. The [NIH Data Management and Sharing Policy](https://sharing.nih.gov/data-management-and-sharing-policy) describes expectations for making research data available. When data are shared, you can check the figure against the full dataset. ## Practical Workflow for Evaluating Any Figure ### Step 1: Read the Caption and Methods Start with the figure caption. It should state what the figure shows, what the axes represent, what the symbols mean, and what statistical tests were applied. Then read the relevant section of the methods to understand how the data were collected and analyzed. ### Step 2: Examine the Axes Check the minimum and maximum values on both axes. Note whether the axis starts at zero. Note whether the scale is linear or logarithmic. Note whether all panels in a multi-panel figure use the same scale. Record any axis choices that seem unusual. ### Step 3: Assess the Chart Type Ask whether the chart type is appropriate for the data. Bar charts are appropriate for comparing categorical groups. Line graphs are appropriate for showing trends over time. Scatter plots are appropriate for showing relationships between two continuous variables. A chart type that does not match the data structure can create a false impression. ### Step 4: Look for Variability Indicators Check for error bars, confidence intervals, or shaded regions. If they are absent, note that the figure does not show variability. If they are present, check the caption or methods to confirm what they represent. ### Step 5: Compare the Figure with the Text Read the results section and compare the numeric values reported in the text with the values shown in the figure. They should match. If the text reports a difference that the figure does not show, or if the figure shows a difference the text does not report, investigate further. ### Step 6: Check for Data Availability Look for statements about data availability. Many journals now require authors to state whether data are available and where they can be accessed. The [NIH Grants and Funding](https://grants.nih.gov/) pages describe the expectations for data sharing in NIH-funded research. When raw data are available, you can verify the figure against the source numbers. ### Step 7: Apply Reporting Guidelines Check whether the paper follows an appropriate reporting guideline from the [EQUATOR Network](https://www.equator-network.org/). These guidelines specify what information must be reported for different study types. A paper that follows the relevant guideline is more likely to provide the information needed to evaluate its figures. ## Options and Tradeoffs in Graph Design ### Bar Charts versus Dot Plots Bar charts are common in scientific papers, but they have limitations. The bars draw attention to the height of each column, which can exaggerate differences when the y-axis does not start at zero. Dot plots show each data point individually, which provides more information about the distribution. When you see a bar chart, check whether the authors also provide the underlying data points or a measure of variability. ### Line Graphs with Interpolated Points Line graphs connect data points with lines. This is appropriate when the variable is continuous and measurements were taken at regular intervals. It is less appropriate when the data are categorical or when the intervals between measurements are uneven. A line graph can imply a smooth trend where none exists. Check the methods to see how often measurements were taken and whether the line represents a statistical model or a simple connection of points. ### Logarithmic Scales Logarithmic scales are useful for data that span several orders of magnitude. They can also be used to compress variation and make differences appear smaller. When you see a logarithmic scale, note it and consider whether the choice is appropriate for the data. The same data plotted on a linear scale may look very different. ### Multi-Panel Figures Multi-panel figures allow authors to show several related graphs in one image. They also create opportunities for misleading comparisons. If different panels use different axis scales, the visual comparison between panels is distorted. Check whether all panels use the same scale before comparing them visually. ## Records and Measurements for Figure Evaluation ### What to Record When you evaluate a figure, keep a record of your observations. This record will help you compare figures across papers and track patterns in how data are presented. Record the following items: - The paper citation and figure number - The research question stated in the abstract - The chart type used - The axis ranges and whether they start at zero - Whether error bars or confidence intervals are shown - What the error bars represent according to the caption - Whether the figure matches the numeric values in the text - Whether the data are available for verification - Any concerns you identified and the questions you would ask the authors ### How to Use the Record Use this record when you compare multiple papers on the same topic. A paper that shows complete data with appropriate variability measures is more trustworthy than one that shows only favorable subsets. A paper that follows reporting guidelines and shares its data allows you to verify its figures. A paper that does neither requires more caution in interpretation. ### Limitations of Your Assessment Your assessment of a figure is limited by the information available to you. If the paper does not report the full methods, you cannot fully evaluate the figure. If the data are not available, you cannot verify the numbers. If the paper does not follow a reporting guideline, you cannot be sure the authors included all relevant information. These limitations are reasons for caution, not reasons to dismiss a paper entirely. ## Common Failure Patterns in Figure Interpretation ### Pattern 1: Trusting the Visual Impression Readers often trust the visual impression of a graph without checking the numbers. A graph with a truncated axis creates a visual impression of a large difference. The numeric labels on the axis tell a different story. The failure occurs when the reader accepts the visual impression without reading the axis values. ### Pattern 2: Ignoring the Error Bars Readers sometimes focus on the height of the bars or the position of the points and ignore the error bars. This leads to overconfidence in the reliability of the observed differences. The error bars are essential information. When they overlap substantially between groups, the difference may not be statistically meaningful. ### Pattern 3: Comparing Panels with Different Scales Multi-panel figures invite visual comparison between panels. When the panels use different scales, the visual comparison is misleading. A small difference in one panel may look large because the scale is narrow, while a large difference in another panel may look small because the scale is wide. ### Pattern 4: Assuming the Figure Matches the Text Readers sometimes assume that the figure and the text tell the same story. This is not always true. The text may report a subset of the results shown in the figure, or the figure may show a subset of the results described in the text. Comparing the two is an essential step. ### Pattern 5: Overlooking the Methods The methods section contains the information needed to interpret the figures. Readers who skip the methods cannot fully evaluate the figures. The methods describe the study design, the measurement procedures, the statistical analysis, and the criteria for including or excluding data. Without this information, figure evaluation is incomplete. ## Welfare and Safety Context for Animal-Related Research For readers who work with animals, the evaluation of graphs has direct welfare implications. Decisions about feed formulations, housing conditions, veterinary treatments, and breeding programs are often based on published research. A misleading graph can lead to decisions that harm animal welfare or waste resources. When evaluating figures in animal-related research, pay particular attention to the following points: - The number of animals in each group and whether the figure shows individual data points or only group summaries - The measures of variability and whether they reflect the range of outcomes across animals - The duration of the study and whether the figure covers the full period - The outcomes measured and whether the figure shows all outcomes or only selected ones - The statistical methods used and whether they are appropriate for the study design The [National Library of Medicine Research Methods Resources](https://www.ncbi.nlm.nih.gov/books) provide access to authoritative texts on research methodology. These resources can help you understand the statistical concepts needed to evaluate figures in animal research. ## Professional Escalation Criteria ### When to Consult a Statistician You should consider consulting a statistician when a figure is central to a decision you are making and you cannot fully evaluate it yourself. Specific situations include: - The figure uses a statistical method you do not understand - The figure shows complex relationships that are difficult to interpret - The paper does not provide enough information to evaluate the figure - The figure conflicts with the text or with other figures in the paper - The decision you are making has significant consequences, such as a treatment choice or a policy recommendation ### When to Contact the Authors Contacting the authors is appropriate when you have specific questions about a figure that the paper does not answer. The [ORCID for Researchers](https://info.orcid.org/researchers) pages describe how researchers maintain their professional records and can be contacted. When you contact an author, be specific about your question. Ask about the axis choices, the error bars, the data coverage, or any other aspect of the figure you cannot evaluate from the paper alone. ### When to Report Concerns If you believe a figure is intentionally misleading or if you have evidence of data fabrication or manipulation, you should consider reporting your concerns. The [Committee on Publication Ethics Core Practices](https://publicationethics.org/core-practices) describe the procedures journals follow for investigating allegations of misconduct. You can contact the journal editor with your concerns. Provide specific evidence, including the figure in question, the relevant sections of the paper, and your analysis of why the figure is misleading. ## Limitations of This Guidance This guidance helps you evaluate figures critically, but it has limitations. You cannot always determine whether a graph is misleading from the figure alone. Some design choices are matters of convention instead of manipulation. Some figures that appear unusual are appropriate for the data and the research question. Your assessment should be based on the full context of the paper, not on the figure in isolation. You also cannot assume that a figure that passes your checks is completely accurate. Authors may make errors in data entry, analysis, or figure preparation. Peer review catches some errors but not all. The [Committee on Publication Ethics Core Practices](https://publicationethics.org/core-practices) describe the responsibilities of journals in maintaining research integrity, but no system is perfect. Finally, your ability to evaluate figures depends on your statistical knowledge. If you are new to reading scientific papers, you may need to build your skills gradually. The [National Library of Medicine Research Methods Resources](https://www.ncbi.nlm.nih.gov/books) provide a starting point for learning about research methodology and statistical concepts. ## Frequently Asked Questions ### How can I tell if a graph is misleading just by looking at it? You cannot reliably tell from looking alone. You must check the axis values, the error bars, the chart type, and the methods description. A graph that looks dramatic may be misleading, and a graph that looks modest may be accurate. The visual impression is never sufficient for evaluation. ### What is the most common way graphs are manipulated in scientific papers? Truncated axes are the most common manipulation. The y-axis starts at a value near the data range instead of zero, which exaggerates small differences between groups. Always check the axis minimum and maximum values before interpreting a graph. ### Do all graphs need to start at zero? No. Line graphs showing change over time often use non-zero baselines to show meaningful variation. The key question is whether the axis choice obscures or exaggerates the true magnitude of the effect. A zero baseline is not always required, but the axis choice should be justified. ### What should I do if the figure does not match the text in a paper? Compare the figure and the text carefully. Note the specific discrepancies. If the text reports a difference that the figure does not show, or if the figure shows a difference the text does not report, contact the authors for clarification. If you suspect misconduct, report your concerns to the journal editor. ### How important are error bars in evaluating a graph? Error bars are essential for understanding variability. Without them, you cannot tell whether the observed differences are meaningful or whether they fall within normal variation. Check what the error bars represent, whether standard deviation, standard error, or confidence intervals, because these measures answer different questions. ### Can a graph be technically accurate and still misleading? Yes. A graph can show accurate data points and still mislead through design choices. Cherry-picking a favorable time window, using an inappropriate chart type, or omitting error bars can all produce a graph that is technically accurate but substantively misleading. The figure must be evaluated in the context of the full paper. ### Where can I find the raw data to verify a figure? Check the data availability statement in the paper. Many journals require authors to state whether data are available and where they can be accessed. The [NIH Data Management and Sharing Policy](https://sharing.nih.gov/data-management-and-sharing-policy) describes expectations for data sharing in NIH-funded research. When data are available, you can verify the figure against the source numbers. ### What should I do if I find a misleading graph in a published paper? First, document your observations. Note the figure number, the specific concerns, and your analysis. Then consider contacting the authors for clarification. If you believe the graph is intentionally misleading or if you have evidence of misconduct, report your concerns to the journal editor following the procedures described in the [Committee on Publication Ethics Core Practices](https://publicationethics.org/core-practices). ## Using the Evidence | Source | Best use in this topic | Important limitation | |---|---|---| | [Research Methods Resources](https://www.ncbi.nlm.nih.gov/books) | official guidance | Check the linked page for current local requirements | | [EQUATOR Network](https://www.equator-network.org/) | official guidance | Check the linked page for current local requirements | | [Core Practices](https://publicationethics.org/core-practices) | official guidance | Check the linked page for current local requirements | ## Related Articles - [How to Present Data in Research Papers: Methods, Tables, and Figures](/blog/guides/how-to-present-data-in-research-papers-methods-tables-and-figures) - [Data Visualization for Biological Research: Choosing Figures That Preserve Meaning](/blog/guides/data-visualization-for-biological-research-choosing-figures-that-preserve-meanin) - [Scientific Data Provenance For Reproducible Research](/blog/guides/scientific-data-provenance-for-reproducible-research) - [How to Cite Scientific Papers: A Guide to Citation Formats](/blog/guides/how-to-cite-scientific-papers-a-guide-to-citation-formats) - [DNA Drawing: How to Represent Structure Without Misleading Readers](/blog/guides/dna-drawing) ## References and Further Reading - [Research Methods Resources](https://www.ncbi.nlm.nih.gov/books). National Library of Medicine. - [EQUATOR Network](https://www.equator-network.org/). EQUATOR Network. - [Core Practices](https://publicationethics.org/core-practices). Committee on Publication Ethics. - [NIH Grants and Funding](https://grants.nih.gov/). National Institutes of Health. - [ORCID for Researchers](https://info.orcid.org/researchers). ORCID. - [Data Management and Sharing Policy](https://sharing.nih.gov/data-management-and-sharing-policy). National Institutes of Health. - [NCBI Literature Resources](https://www.ncbi.nlm.nih.gov/). National Center for Biotechnology Information. - [PubMed](https://pubmed.ncbi.nlm.nih.gov/). National Library of Medicine. > This article is educational and does not replace institutional policy, professional advice, or applicable safety and regulatory requirements.