scientific-skills/scientific-writing/references/imrad_structure.md
IMRAD (Introduction, Methods, Results, And Discussion) is the predominant organizational structure for scientific journal articles of original research. Adopted as the majority format since the 1970s, it is now the standard in medical, health, biological, chemical, engineering, and computer sciences.
The IMRAD structure mirrors the scientific method:
This logical flow makes scientific papers easier to write, read, and evaluate.
A full scientific manuscript typically includes these sections in order:
Attract readers and accurately represent the paper's content.
Provide a complete, standalone summary enabling readers to decide if the full paper is relevant to them.
⚠️ CRITICAL: Write abstracts as flowing paragraphs, NOT with labeled sections.
Most scientific papers use unstructured abstracts written as one or two cohesive paragraphs. This is the standard format for the majority of journals including Nature, Science, Cell, PNAS, and most field-specific journals.
❌ WRONG - Structured abstract with labels:
Background: Hospital-acquired infections remain a major cause of morbidity.
Methods: We conducted a 12-month before-after study...
Results: Post-intervention, surface contamination decreased by 47%...
Conclusions: UV-C disinfection significantly reduced infection rates.
✅ CORRECT - Flowing paragraph style:
Hospital-acquired infections remain a major cause of morbidity, yet optimal
disinfection strategies remain unclear. We conducted a 12-month before-after
study in a 500-bed teaching hospital to evaluate UV-C disinfection added to
standard cleaning protocols. Environmental surfaces were cultured monthly and
infection rates tracked via surveillance data. Post-intervention, surface
contamination decreased by 47% (95% CI: 38-56%, p<0.001), and catheter-associated
urinary tract infections declined from 3.2 to 1.8 per 1000 catheter-days (RR=0.56,
95% CI: 0.38-0.83, p=0.004). No adverse effects were observed. These findings
demonstrate that UV-C disinfection significantly reduces environmental contamination
and infection rates, suggesting it may be a valuable addition to hospital infection
control programs.
While written as flowing prose, the abstract should cover these elements in order:
Only use labeled sections (Background/Objective, Methods, Results, Conclusions) when:
Even for structured abstracts, write each section as complete sentences, not fragments.
Transcriptomic aging clocks offer unique advantages for assessing biological age by
capturing dynamic cellular states and acute responses to perturbations. Using the
ARCHS4 database containing uniformly processed RNA-seq data from over 1.2 million
human samples, we developed deep neural network models to predict chronological age
from gene expression profiles. Our best-performing model achieved a mean absolute
error of 4.2 years (R² = 0.89) on held-out test data, substantially outperforming
traditional machine learning approaches including elastic net regression (MAE = 6.8
years) and random forests (MAE = 5.9 years). Feature importance analysis identified
genes enriched in senescence, inflammation, and mitochondrial function pathways as
the strongest predictors. Cross-tissue validation revealed that lung and blood
samples yielded the most accurate predictions, while liver showed the highest
variance. These findings establish deep learning as a powerful approach for
transcriptomic age prediction and identify candidate biomarkers for biological
aging assessment.
Convince readers that the research addresses an important question using an appropriate approach.
Paragraph 1: The Big Picture
Paragraphs 2-3: Narrowing Down
Paragraph 4: The Gap
Final Paragraph: This Study
Provide sufficient detail for others to replicate the study and evaluate its validity.
Another expert in the field should be able to repeat your experiment exactly as you performed it.
Present the findings objectively without interpretation.
Show, don't interpret. Save interpretation for the Discussion.
Opening Paragraph
Subsequent Paragraphs
Each Finding Should Include:
Example: "Mean systolic blood pressure decreased by 12 mmHg in the intervention group compared to 3 mmHg in controls (difference: 9 mmHg, 95% CI: 4-14 mmHg, p=0.002)."
When to Use:
How to Reference:
Required Elements:
Example: "Groups differed significantly on test performance (t(48) = 3.21, p = 0.002, Cohen's d = 0.87, 95% CI: 0.34-1.40)."
By Objective:
Effect of intervention on primary outcome
Effect of intervention on secondary outcome A
Effect of intervention on secondary outcome B
By Analysis Type:
Descriptive statistics
Univariate analyses
Multivariate analyses
Chronological:
Baseline characteristics
Short-term outcomes (1 month)
Long-term outcomes (6 months)
Interpret findings, relate them to existing knowledge, acknowledge limitations, and propose future directions.
Paragraph 1: Summary of Main Findings
Paragraphs 2-4: Interpretation in Context
Paragraph 5: Strengths and Limitations
Paragraph 6: Implications
Final Paragraph: Conclusions and Future Directions
Comparing to Prior Work:
"Our finding of a 30% reduction in symptoms aligns with Smith et al. (2023), who
reported a 28% reduction using a similar intervention. However, Jones et al. (2022)
found no significant effect, possibly due to their use of a less intensive protocol."
Proposing Mechanisms:
"The observed improvement in cognitive function may result from increased cerebral
blood flow, as evidenced by the concurrent increase in functional MRI signals in the
prefrontal cortex. This interpretation is consistent with the vascular hypothesis of
cognitive enhancement."
Acknowledging Limitations:
"The cross-sectional design prevents causal inference. Additionally, the convenience
sample from a single academic medical center may limit generalizability to community
settings. Self-reported measures may introduce recall bias, though we attempted to
minimize this through structured interviews."
Provide a concise summary of key findings and their significance.
This randomized trial demonstrates that a 12-week mindfulness intervention significantly
reduces anxiety symptoms in college students, with effects persisting at 6-month follow-up.
These findings support the integration of mindfulness-based programs into university mental
health services. Given the scalability and cost-effectiveness of group-based mindfulness
training, this approach offers a promising strategy to address the growing mental health
crisis in higher education.
citation_styles.md)| Section | Verb Tense |
|---|---|
| Abstract - Background | Present (established facts) or past (prior studies) |
| Abstract - Methods | Past |
| Abstract - Results | Past |
| Abstract - Conclusions | Present |
| Introduction - General background | Present |
| Introduction - Prior studies | Past |
| Introduction - Your objectives | Present or past |
| Methods | Past (your actions), present (general procedures) |
| Results | Past |
| Discussion - Your findings | Past |
| Discussion - Interpretations | Present |
| Discussion - Prior work | Present or past |
| Conclusion | Present |
| Venue Type | Length | Structure | Methods Placement | Key Focus |
|---|---|---|---|---|
| Nature/Science | 2,000-4,500 words | Modified IMRAD | Supplement | Broad significance |
| Medical | 2,700-3,500 words | Strict IMRAD | Main text | Clinical outcomes |
| Field journals | 3,000-6,000 words | Standard IMRAD | Main text | Technical depth |
| ML conferences | 8-9 pages (~6,000 words) | Intro-Method-Experiments-Conclusion | Main text (concise) | Novel contribution |
Typical 8-page structure:
Key differences from journals:
| Venue | Intro | Methods | Results/Experiments | Discussion/Conclusion |
|---|---|---|---|---|
| Nature/Science | 10% | 15%* | 40% | 35% |
| Medical (NEJM/JAMA) | 10% | 25% | 30% | 35% |
| Field journals | 20% | 25% | 30% | 25% |
| ML conferences | 12-15% | 30-35% | 40-45% | 5-8% |
*Methods often in supplement for Nature/Science
Key medical journal features:
Key ML conference features:
| Venue | Audience | Intro Focus | Methods Detail | Results/Experiments | Discussion/Conclusion |
|---|---|---|---|---|---|
| Nature/Science | Non-specialists | Broad significance | Brief, supplement | Story-driven | Broad implications |
| Medical | Clinicians | Clinical problem | Comprehensive | Primary outcome first | Clinical relevance |
| Specialized | Experts | Field context | Full technical | By experiment | Mechanistic depth |
| ML conferences | ML researchers | Novel contribution | Reproducible | Baselines, ablations | Brief, limitations |
ML conference emphasis:
What gets checked:
Common rejection reasons:
ML conference specific evaluation:
Journal → ML conference:
ML conference → Journal:
Specialist → Broad journal:
Broad → Specialist:
All venues:
ML conferences add: