brycewang-stanford/abstract-writing-guide

Abstract Writing Guide

A skill for writing effective research abstracts that clearly communicate your study's purpose, methods, results, and significance. Covers structured and unstructured formats, word count optimization, and common pitfalls.

Abstract Structures

Structured Abstract (IMRaD)

Most journals in science and medicine require a structured abstract:

BACKGROUND/OBJECTIVE: 1-2 sentences on the problem and study aim
METHODS: 2-3 sentences on study design, participants, and analysis
RESULTS: 2-4 sentences with key quantitative findings
CONCLUSION: 1-2 sentences on implications and significance

Total: typically 150-300 words (check journal requirements)

Unstructured Abstract

Common in social sciences and humanities:

Sentence 1: Context and problem statement
Sentence 2: Research gap or objective
Sentence 3: Methods overview
Sentence 4-5: Key findings
Sentence 6: Implications and significance

Total: typically 150-250 words

Writing Process

Step-by-Step Framework

def generate_abstract_outline(study_info: dict) -> dict:
    """
    Generate an abstract outline from study information.

    Args:
        study_info: Dict with keys: 'problem', 'gap', 'objective',
                    'design', 'sample', 'methods', 'key_findings',
                    'implications', 'word_limit'
    Returns:
        Structured abstract outline with estimated word counts
    """
    word_limit = study_info.get('word_limit', 250)

    # Allocate words proportionally
    sections = {
        'background': {
            'allocation_pct': 0.15,
            'content': [study_info['problem'], study_info['gap']],
            'template': "Despite {problem}, {gap}. This study aimed to {objective}."
        },
        'methods': {
            'allocation_pct': 0.25,
            'content': [study_info['design'], study_info['sample'], study_info['methods']],
            'template': "A {design} was conducted with {sample}. {methods}."
        },
        'results': {
            'allocation_pct': 0.35,
            'content': study_info['key_findings'],
            'template': "Results showed {findings}."
        },
        'conclusion': {
            'allocation_pct': 0.25,
            'content': study_info['implications'],
            'template': "These findings suggest {implications}."
        }
    }

    for section in sections.values():
        section['target_words'] = round(word_limit * section['allocation_pct'])

    return {
        'sections': sections,
        'word_limit': word_limit,
        'tips': [
            'Write the Results section first (most important)',
            'Use past tense for methods and results',
            'Avoid abbreviations unless space-saving is critical',
            'Include at least one quantitative result with effect size',
            'Do not include citations in the abstract'
        ]
    }

Example: Well-Written Abstract

Background: Chronic low back pain affects 23% of the global adult population,
yet optimal exercise interventions remain unclear. This systematic review and
meta-analysis aimed to compare the effectiveness of yoga, Pilates, and
general exercise for reducing pain and disability in chronic low back pain.

Methods: We searched PubMed, CINAHL, Cochrane Library, and PEDro from
inception to December 2025. Randomized controlled trials comparing yoga,
Pilates, or general exercise to usual care were included. Pain (VAS/NRS)
and disability (ODI/RMDQ) were pooled using random-effects meta-analysis.

Results: Thirty-seven RCTs (N = 3,421) met inclusion criteria. All exercise
types significantly reduced pain compared to usual care: yoga (SMD = -0.73,
95% CI [-0.94, -0.52]), Pilates (SMD = -0.62, 95% CI [-0.81, -0.43]),
and general exercise (SMD = -0.48, 95% CI [-0.63, -0.33]). Yoga showed
significantly greater pain reduction than general exercise (p = 0.03)
but not Pilates (p = 0.28). Heterogeneity was moderate (I-squared = 48-61%).

Conclusion: Yoga, Pilates, and general exercise all effectively reduce
chronic low back pain, with yoga showing a modest advantage over general
exercise. Clinicians should consider patient preferences when recommending
exercise modalities.

Title Writing

Title Construction Principles

def evaluate_title(title: str) -> dict:
    """Evaluate a paper title against best practices."""
    checks = {}

    # Length check
    word_count = len(title.split())
    checks['word_count'] = word_count
    checks['length_ok'] = 10 <= word_count <= 20

    # Key content checks
    checks['has_colon'] = ':' in title  # Structured titles often use colons
    checks['starts_with_article'] = title.lower().startswith(('a ', 'an ', 'the '))
    checks['has_abbreviation'] = any(w.isupper() and len(w) > 1 for w in title.split())

    # Title type classification
    if '?' in title:
        checks['type'] = 'question'
    elif ':' in title:
        checks['type'] = 'compound'
    elif title.split()[0].endswith('ing'):
        checks['type'] = 'gerund'
    else:
        checks['type'] = 'declarative'

    checks['recommendations'] = []
    if word_count > 20:
        checks['recommendations'].append('Consider shortening -- aim for 10-15 words')
    if word_count < 8:
        checks['recommendations'].append('Consider adding specificity')
    if checks['starts_with_article']:
        checks['recommendations'].append('Avoid starting with articles (A/An/The)')
    if checks['has_abbreviation']:
        checks['recommendations'].append('Spell out abbreviations in titles')

    return checks

Title Patterns by Discipline

| Pattern | Example | Common In | |---------|---------|-----------| | Method: Finding | "Deep Learning for Protein Structure: A Comparative Study" | CS, Engineering | | Declarative | "Mindfulness Reduces Cortisol in College Students" | Psychology, Medicine | | Question | "Does Remote Work Improve Productivity?" | Social Sciences | | Descriptive | "Gene Expression Patterns in Early-Stage Lung Cancer" | Biology, Medicine |

Common Mistakes to Avoid

1. Vague conclusions: "More research is needed" -- instead, state specific implications
2. Missing numbers: Always include at least one quantitative finding
3. New information: Never introduce information not in the paper
4. Excessive background: Readers know the field -- get to your contribution quickly
5. Passive voice overuse: Use active voice where possible ("We found" not "It was found")
6. Promising more than delivering: The abstract must accurately represent the paper's scope