3.3 Writing the Methods Section¶
What you will learn on this page
- Practical tips for describing procedures and experimental settings in academic English
- Prompts for turning technical content into clear natural language
- A typical Methods subsection structure and ready-to-use templates for participant descriptions
- Common vague expressions and how to make them specific
- Citation conventions for software and tools
- Statistical reporting formats (APA 7th style)
- How to keep Methods consistent with tables and figures
- Cautions when using AI for qualitative data analysis
- How to write a Data Availability Statement
How this page fits in
The role of the Methods section (ensuring reproducibility) and typical move patterns are explained in
2.1 The IMRaD Structure and
2.2 Checking Conventional Expressions with Corpora.
This page assumes that foundation and focuses on a hands-on workflow for converting your procedure notes into publishable academic English.
Core principles of Methods writing¶
The single most important principle of a Methods section is to provide enough detail that another researcher could reproduce your study.
In practice, this means including:
- Concrete numbers: sample size, duration, number of sessions/trials, time on task
- Procedural sequence: what happened, in what order, and under what conditions
- Tools and version information: software names, versions, parameter settings
- Ethics information: approval body, approval number if applicable, consent procedures
The tense is primarily past tense, since you are describing procedures that have already been carried out (→ 2.3 Tense rules by section).
Methods is not a sales pitch
Methods should be factual and concrete. Avoid persuasive wording such as "this innovative method," and avoid interpretation. Interpretation belongs in Discussion.
Typical Methods subsection structure¶
Methods is typically organized into the following subsections. Names may vary by field or journal, but the information to include is largely the same.
| Subsection | What to describe | Notes |
|---|---|---|
| Participants | Sample size, age, gender, background, selection criteria, exclusion criteria | State ethics approval explicitly |
| Materials / Instruments | Questionnaires, tests, software, equipment used | Include version numbers and reliability coefficients |
| Procedure | Data collection steps in chronological order | Enough detail for reproducibility |
| Data Analysis | Analysis methods and software used | Also describe how assumptions were checked |
For computer science and information processing
In CS-related papers, the Methods section is often called Proposed Method or System Design. In such cases, the focus is on algorithm descriptions and system architecture diagrams, but the principle of writing with reproducible detail remains the same.
Participant description template¶
Participant descriptions are largely formulaic, so using a template is efficient.
Participant description template (English)
A total of [N] participants ([N] males, [N] females; M age = [X] years,
SD = [X]) took part in this study. They were [affiliation/attribute].
Participants were recruited through [method].
The inclusion criteria were [criteria]. [N] participants were excluded
because [reason]. This study was approved by [ethics review body]
(approval number: [number]). All participants provided written
informed consent prior to data collection.
Prompt: convert participant information to English
Turning technical content into clear English (AI prompts)¶
Researchers often record their experimental procedures as notes or bullet points. Converting these into natural English paragraphs is one of the tasks where AI is most effective.
Prompt: bullet points → Methods paragraph
Convert the procedure notes below into natural academic English
as a Methods section following IMRaD structure.
- Use past tense and mostly passive voice
- Include enough detail for reproducibility
- Do not change any numbers or conditions
- Do not add any new information
[Paste bullet-point procedure notes here]
Prompt: clarify existing Methods description
Common vague expressions and how to fix them¶
Methods written by Japanese-speaking authors often contain the following kinds of vague expressions. The criterion for judgment is: could another researcher reproduce the study from this description alone?
Before / After example¶
Before (vague):
Several participants were excluded. The test was administered and participants were given enough time. Data were analyzed statistically using the software.
After (specific):
Three participants were excluded due to incomplete responses. The test was administered individually in a quiet room during regular class hours. Participants were given 30 minutes to complete the task. Data were analyzed using a paired-samples t-test in R (version 4.3.1; R Core Team, 2023).
Common vague patterns and how to improve them¶
| Vague expression | Why it is problematic | Better alternative |
|---|---|---|
| Several participants were excluded. | Number is unknown | Three participants were excluded due to incomplete responses. |
| The test was administered. | When, where, and how are unclear | The test was administered individually in a quiet room during regular class hours. |
| Data were analyzed statistically. | Which method is unknown | Data were analyzed using a paired-samples t-test. |
| The software was used for analysis. | Software name is missing | R (version 4.3.1; R Core Team, 2023) was used for all statistical analyses. |
| Participants were given enough time. | "enough" is subjective | Participants were given 30 minutes to complete the task. |
| A questionnaire was used. | Number of items and scale are unknown | A 20-item questionnaire using a 6-point Likert scale was used. |
Prompt: detect vague expressions in Methods
Identify vague expressions in the Methods section below.
The criterion for "vague" is: could another researcher reproduce the study
from this description alone?
For each issue:
(1) Quote the relevant phrase
(2) Explain why it is vague (what information is missing)
(3) Suggest what should be made specific (use [to be added: __] format)
[Paste Methods text here]
Software and tool citation conventions¶
Methods requires accurate citation of the software and tools you used. Formats vary by field and journal, but the following patterns are common.
In-text citation patterns¶
| Software | Example |
|---|---|
| R | All analyses were conducted using R (version 4.3.1; R Core Team, 2023). |
| R package | Linear mixed-effects models were fitted using the lme4 package (version 1.1-35; Bates et al., 2015) in R. |
| Python | Data preprocessing was performed using Python (version 3.11; Van Rossum & Drake, 2023). |
| SPSS | Statistical analyses were performed using IBM SPSS Statistics (version 29.0; IBM Corp., 2023). |
| Excel | Descriptive statistics were calculated using Microsoft Excel (version 16.0). |
| Questionnaire/test | Vocabulary size was measured using the Vocabulary Size Test (Nation & Beglar, 2007). |
Citation notes¶
- Always include version numbers: Results can differ depending on the software version
- Include in References as well: Software cited in the text must also appear in the reference list
- Include URLs: For open-source software, including the download URL is helpful
Prompt: check software citation format
Check whether the software and tools mentioned in the Methods section below
are cited appropriately.
Checklist:
(1) Is the version number specified?
(2) Is developer/author information included?
(3) Should a corresponding entry be added to References? (If so, suggest the format)
[Paste Methods text here]
If you used generative AI in a limited way (for example, language editing), follow your venue's disclosure rules (see 1.3 Guidelines from Publishers and Educational Institutions).
Statistical reporting format (APA 7th)¶
Different fields have different reporting formats, but in the social sciences, the APA (American Psychological Association) 7th edition format is widely used.
Reporting formats for common statistics¶
| Analysis | Reporting format | Example |
|---|---|---|
| t-test | t(df) = value, p = value, d = value | t(28) = 2.45, p = .021, d = 0.65 |
| ANOVA | F(df1, df2) = value, p = value, η² = value | F(2, 57) = 4.12, p = .022, η² = .13 |
| Correlation | r(df) = value, p = value | r(48) = .42, p < .001 |
| Chi-square test | χ²(df, N = value) = value, p = value | χ²(2, N = 120) = 8.34, p = .015 |
Reporting notes¶
- Do not place a zero before the decimal point in p values (✓ p = .021, ✗ p = 0.021)
- When p < .001, write p < .001 rather than reporting the exact value
- Always report effect sizes (d, η², r, etc.)
- Use italics for statistical symbols (t, F, p, r, N, etc.)
- In APA style, Greek letters are not italicized (e.g., χ², η², α, β, ω).
- Round test statistics to two decimal places
Prompt: check statistical reporting format
Check the statistical reports in the Results section below
against APA 7th edition format.
Checklist:
(1) Statistical symbols in italics (represented as *symbol* in text)
(2) p-value format (no leading zero, handling of p < .001)
(3) Whether effect sizes are reported
(4) Degrees of freedom included
(5) Consistent number of decimal places
Point out any issues and show the correct format.
[Paste Results section here]
Keep Methods consistent with figures and tables¶
Methods descriptions and the figures and tables in Results must be consistent. You can also use AI to check for consistency.
Compare the Methods description and the figure/table captions below
and check for inconsistencies.
- Variable name discrepancies
- Match between analysis methods and graph types
- Consistency in participant numbers
Methods:
[Paste Methods here]
Figure/table captions:
[Paste caption list here]
Referring to figures and tables
Adding instructions such as "include explanations linked to Figure 1" in your prompt improves consistency between figures/tables and the main text. However, you need to show the AI the actual figure/table content, either as text description or as an image input.
Figure and table caption conventions¶
Figure and table captions also have writing conventions.
| Element | Figure | Table |
|---|---|---|
| Number placement | Below the figure (Figure 1) | Above the table (Table 1) |
| Title | In italics, concise | In italics, concise |
| Notes | Below the figure with Note. | Below the table with Note. |
Prompt: generate figure/table captions
Create APA 7th edition-compliant captions (title + notes)
for the following figure/table.
Conditions:
- Figure/Table number: [number]
- Title should express the content concisely (one sentence or less)
- Explain abbreviations in notes if applicable
- Significance notation: *p < .05, **p < .01, ***p < .001
Figure/table content:
[Describe the figure/table content here]
Generating description samples from figures and tables¶
Since generative AI accepts images as well as text input, you can show AI a screenshot of your Figure or Table directly and have it generate sample descriptions for how to refer to it in the main text. This approach is especially useful when you are unsure how to write about a particular figure or table.
Prompt: generate description samples from a figure/table
The attached figure (Figure 1) shows the results of [brief description of the experiment].
Write 2–3 sample sentences for referring to this figure in a Results section.
- Use past tense
- Include a reference to the figure number (e.g., "As shown in Figure 1, ...")
- Use specific numbers as far as they can be read from the figure
- Limit to factual reporting; do not include interpretation or discussion
Tips for showing figures/tables to AI
- Provide not just the image but also context such as the research purpose and variable descriptions — this produces more appropriate descriptions
- Treat the generated samples as reference for how to write, and always verify the accuracy of numbers and interpretations yourself
- For complex tables, describing the table structure in text in addition to the image improves accuracy
Using AI for qualitative data analysis: cautions¶
The use of generative AI for coding and classifying open-ended response data is increasing. However, at present, the accuracy of automated coding by generative AI is not sufficient.
Mizumoto & Teng (2025) classified learners' open-ended responses using multiple LLMs (GPT-4o, GPT-o1, GPT-o3 mini, Llama, Gemini, Claude, DeepSeek) and reported that Cohen's Kappa coefficients generally fell in the range of 0.4–0.7, indicating that AI coding did not fully agree with human coding.
Guidelines for using AI coding
- A practical approach is a two-stage workflow where generative AI serves as a "first-pass candidate annotator" and humans perform the final review
- Using AI for "material generation" purposes (such as generating roleplay scenarios or draft test items) is effective
- When using generative AI for coding purposes in research, describe the procedure and its limitations explicitly in the Methods section
For theme extraction from qualitative data, Attention (introduced in 3.2) can also be useful. It offers theme analysis, sentiment analysis, and summarization features, making it helpful for exploratory analysis of open-ended data. However, as noted above, final classification judgments must be reviewed by humans.
Data Availability Statement¶
Many journals now require a Data Availability Statement describing whether and where data are available. This is placed in the Methods section or at the end of the paper.
Common patterns¶
| Situation | Example |
|---|---|
| Data made public | The dataset generated during the current study is available in the [repository name] repository, [URL or DOI]. |
| Available upon reasonable request | The data that support the findings of this study are available from the corresponding author upon reasonable request. |
| Restricted for ethical reasons | The data are not publicly available due to privacy/ethical restrictions but are available from the corresponding author on reasonable request. |
| Using existing public data | This study used publicly available data from [data source name] ([URL or DOI]). |
Prompt: create a Data Availability Statement
Based on the research overview below, create a Data Availability Statement
in English.
Conditions:
- Target journal: [journal name]
- Data availability: [public / available upon request / restricted]
- Reason: [applicable reason]
- Repository (if public): [repository name]
[Paste research overview here]
Data repositories
General-purpose repositories available across disciplines include Zenodo, figshare, and OSF (Open Science Framework). If your field has a discipline-specific repository (for example, IRIS for linguistics), consider using that as well.
For details on reproducible research practices (sharing data and code via repositories, integrating analysis code and results with R Markdown, and creating synthetic data when raw data cannot be shared), see In'nami et al. (2022).
Methods self-check checklist¶
Once you have finished writing your Methods section, check it against the following list. This checklist is specific to Methods (for a pre-submission checklist covering the entire paper, see 3.5 Revision Techniques).
- Participant numbers and attributes are described concretely
- Ethics approval number and informed consent are described
- Version information for materials and tools is included
- Software citations are also included in References
- Procedure is described in chronological order with enough detail for reproducibility
- Analysis methods are described specifically with software names stated
- Past tense is used consistently throughout
- Variable names match those reported in Results figures and tables
- No vague expressions remain (several, some, enough, various, etc.)
- Data Availability Statement is included (if required by the journal)
Prompt: comprehensive Methods check
For Results and Discussion writing, continue to:
3.4 Writing Data Analysis, Results, and Discussion