Hypothesis Generation

Scientific Hypothesis Generation

Overview

Hypothesis generation is a systematic process for developing testable explanations. Formulate evidence-based hypotheses from observations, design experiments, explore competing explanations, and develop predictions. Apply this skill for scientific inquiry across domains.

When to Use This Skill

This skill should be used when:

• Developing hypotheses from observations or preliminary data
• Designing experiments to test scientific questions
• Exploring competing explanations for phenomena
• Formulating testable predictions for research
• Conducting literature-based hypothesis generation
• Planning mechanistic studies across scientific domains

Workflow

Follow this systematic process to generate robust scientific hypotheses:

1. Understand the Phenomenon

Start by clarifying the observation, question, or phenomenon that requires explanation:

• Identify the core observation or pattern that needs explanation
• Define the scope and boundaries of the phenomenon
• Note any constraints or specific contexts
• Clarify what is already known vs. what is uncertain
• Identify the relevant scientific domain(s)

2. Conduct Comprehensive Literature Search

Search existing scientific literature to ground hypotheses in current evidence. Use both PubMed (for biomedical topics) and general web search (for broader scientific domains):

For biomedical topics:

• Use WebFetch with PubMed URLs to access relevant literature
• Search for recent reviews, meta-analyses, and primary research
• Look for similar phenomena, related mechanisms, or analogous systems

For all scientific domains:

• Use WebSearch to find recent papers, preprints, and reviews
• Search for established theories, mechanisms, or frameworks
• Identify gaps in current understanding

Search strategy:

• Begin with broad searches to understand the landscape
• Narrow to specific mechanisms, pathways, or theories
• Look for contradictory findings or unresolved debates
• Consult references/literature_search_strategies.md for detailed search techniques

3. Synthesize Existing Evidence

Analyze and integrate findings from literature search:

• Summarize current understanding of the phenomenon
• Identify established mechanisms or theories that may apply
• Note conflicting evidence or alternative viewpoints
• Recognize gaps, limitations, or unanswered questions
• Identify analogies from related systems or domains

4. Generate Competing Hypotheses

Develop 3-5 distinct hypotheses that could explain the phenomenon. Each hypothesis should:

• Provide a mechanistic explanation (not just description)
• Be distinguishable from other hypotheses
• Draw on evidence from the literature synthesis
• Consider different levels of explanation (molecular, cellular, systemic, population, etc.)

Strategies for generating hypotheses:

• Apply known mechanisms from analogous systems
• Consider multiple causative pathways
• Explore different scales of explanation
• Question assumptions in existing explanations
• Combine mechanisms in novel ways

5. Evaluate Hypothesis Quality

Assess each hypothesis against established quality criteria from references/hypothesis_quality_criteria.md:

Testability: Can the hypothesis be empirically tested? Falsifiability: What observations would disprove it? Parsimony: Is it the simplest explanation that fits the evidence? Explanatory Power: How much of the phenomenon does it explain? Scope: What range of observations does it cover? Consistency: Does it align with established principles? Novelty: Does it offer new insights beyond existing explanations?

Explicitly note the strengths and weaknesses of each hypothesis.

6. Design Experimental Tests

For each viable hypothesis, propose specific experiments or studies to test it. Consult references/experimental_design_patterns.md for common approaches:

Experimental design elements:

• What would be measured or observed?
• What comparisons or controls are needed?
• What methods or techniques would be used?
• What sample sizes or statistical approaches are appropriate?
• What are potential confounds and how to address them?

Consider multiple approaches:

• Laboratory experiments (in vitro, in vivo, computational)
• Observational studies (cross-sectional, longitudinal, case-control)
• Clinical trials (if applicable)
• Natural experiments or quasi-experimental designs

7. Formulate Testable Predictions

For each hypothesis, generate specific, quantitative predictions:

• State what should be observed if the hypothesis is correct
• Specify expected direction and magnitude of effects when possible
• Identify conditions under which predictions should hold
• Distinguish predictions between competing hypotheses
• Note predictions that would falsify the hypothesis

8. Present Structured Output

Use the template in assets/hypothesis_output_template.md to present hypotheses in a clear, consistent format:

Standard structure: 1. Background & Context - Phenomenon and literature summary 2. Competing Hypotheses - Enumerated hypotheses with mechanistic explanations 3. Quality Assessment - Evaluation of each hypothesis 4. Experimental Designs - Proposed tests for each hypothesis 5. Testable Predictions - Specific, measurable predictions 6. Critical Comparisons - How to distinguish between hypotheses

Quality Standards

Ensure all generated hypotheses meet these standards:

• Evidence-based: Grounded in existing literature with citations
• Testable: Include specific, measurable predictions
• Mechanistic: Explain how/why, not just what
• Comprehensive: Consider alternative explanations
• Rigorous: Include experimental designs to test predictions

Resources

references/

• hypothesis_quality_criteria.md - Framework for evaluating hypothesis quality (testability, falsifiability, parsimony, explanatory power, scope, consistency)
• experimental_design_patterns.md - Common experimental approaches across domains (RCTs, observational studies, lab experiments, computational models)
• literature_search_strategies.md - Effective search techniques for PubMed and general scientific sources

assets/

• hypothesis_output_template.md - Structured format for presenting hypotheses consistently with all required sections

Scientific Hypothesis Generation: [Phenomenon Name]

1. Background & Context

Phenomenon Description

[Clear description of the observation, pattern, or question that requires explanation. Include:

• What was observed or what question needs answering
• The specific context or system in which it occurs
• Any relevant constraints or boundary conditions
• Why this phenomenon is interesting or important]

Current Understanding

[Synthesis of existing literature, including:

• What is already known about this phenomenon
• Established mechanisms or theories that may be relevant
• Key findings from recent research
• Gaps or limitations in current understanding
• Conflicting findings or unresolved debates

Include citations to key papers (Author et al., Year, Journal)]

Knowledge Gaps

[Specific aspects that remain unexplained or poorly understood:

• What aspects of the phenomenon lack clear explanation?
• What contradictions exist in current understanding?
• What questions remain unanswered?]

---

2. Competing Hypotheses

Hypothesis 1: [Concise Title]

Mechanistic Explanation: [Detailed explanation of the proposed mechanism. This should explain HOW and WHY the phenomenon occurs, not just describe WHAT occurs. Include:

• Specific molecular, cellular, physiological, or population-level mechanisms
• Causal chain from initial trigger to observed outcome
• Key components, pathways, or factors involved
• Scale or level of explanation (molecular, cellular, organ, organism, population)]

Supporting Evidence: [Evidence from literature that supports this hypothesis:

• Analogous mechanisms in related systems
• Direct evidence from relevant studies
• Theoretical frameworks that align with this hypothesis
• Include citations]

Key Assumptions: [Explicit statement of assumptions underlying this hypothesis:

• What must be true for this hypothesis to hold?
• What conditions or contexts does it require?]

---

Hypothesis 2: [Concise Title]

Mechanistic Explanation: [Detailed mechanistic explanation distinct from Hypothesis 1]

Supporting Evidence: [Evidence supporting this alternative explanation]

Key Assumptions: [Assumptions underlying this hypothesis]

---

Hypothesis 3: [Concise Title]

Mechanistic Explanation: [Detailed mechanistic explanation distinct from previous hypotheses]

Supporting Evidence: [Evidence supporting this explanation]

Key Assumptions: [Assumptions underlying this hypothesis]

---

[Continue for Hypothesis 4, 5, etc. if applicable]

---

3. Quality Assessment

Evaluation Against Core Criteria

Rating scale: Strong / Moderate / Weak

Detailed Evaluation

Hypothesis 1

Strengths:

• [Specific strength 1]
• [Specific strength 2]

Weaknesses:

• [Specific weakness 1]
• [Specific weakness 2]

Overall Assessment: [Brief summary of hypothesis quality and viability]

Hypothesis 2

[Similar structure]

Hypothesis 3

[Similar structure]

---

4. Experimental Designs

Testing Hypothesis 1: [Title]

Experiment 1A: [Brief title]

Design Type: [e.g., In vitro dose-response / In vivo knockout / Clinical RCT / Observational cohort / Computational model]

Objective: [What specific aspect of the hypothesis does this test?]

Methods:

• System/Model: [What system, organism, or population?]
• Intervention/Manipulation: [What is varied or manipulated?]
• Measurements: [What outcomes are measured?]
• Controls: [What control conditions?]
• Sample Size: [Estimated n, with justification if possible]
• Analysis: [Statistical or analytical approach]

Expected Timeline: [Rough estimate]

Feasibility: [High/Medium/Low, with brief justification]

Experiment 1B: [Brief title - alternative or complementary approach] [Similar structure to 1A]

---

Testing Hypothesis 2: [Title]

Experiment 2A: [Brief title] [Structure as above]

Experiment 2B: [Brief title] [Structure as above]

---

Testing Hypothesis 3: [Title]

Experiment 3A: [Brief title] [Structure as above]

---

5. Testable Predictions

Predictions from Hypothesis 1

1. Prediction 1.1: [Specific, measurable prediction]

• Conditions: [Under what conditions should this be observed?]
• Magnitude: [Expected effect size or direction, if quantifiable]
• Falsification: [What observation would falsify this prediction?]

2. Prediction 1.2: [Specific, measurable prediction]

• Conditions: [Conditions]
• Magnitude: [Expected effect]
• Falsification: [Falsifying observation]

3. Prediction 1.3: [Additional prediction]

---

Predictions from Hypothesis 2

1. Prediction 2.1: [Specific, measurable prediction]

• Conditions: [Conditions]
• Magnitude: [Expected effect]
• Falsification: [Falsifying observation]

2. Prediction 2.2: [Additional prediction]

---

Predictions from Hypothesis 3

1. Prediction 3.1: [Specific, measurable prediction]

• Conditions: [Conditions]
• Magnitude: [Expected effect]
• Falsification: [Falsifying observation]

---

6. Critical Comparisons

Distinguishing Between Hypotheses

Comparison: Hypothesis 1 vs. Hypothesis 2

Key Distinguishing Feature: [What is the fundamental difference in mechanism or prediction?]

Discriminating Experiment: [What experiment or observation would clearly favor one over the other?]

Outcome Interpretation:

• If [Result A], then Hypothesis 1 is supported
• If [Result B], then Hypothesis 2 is supported
• If [Result C], then both/neither are supported

---

Comparison: Hypothesis 1 vs. Hypothesis 3 [Similar structure]

---

Comparison: Hypothesis 2 vs. Hypothesis 3 [Similar structure]

---

Priority Experiments

Highest Priority Test: [Which experiment would most efficiently distinguish between hypotheses or most definitively test a hypothesis?]

Justification: [Why is this the highest priority? Consider informativeness, feasibility, and cost]

Secondary Priority Tests: 1. [Second most important experiment] 2. [Third most important]

---

7. Summary & Recommendations

Summary of Hypotheses

[Brief paragraph summarizing the competing hypotheses and their relationships]

Recommended Testing Sequence

Phase 1 (Initial Tests): [Which experiments should be done first? Why?]

Phase 2 (Contingent on Phase 1 results): [What follow-up experiments depend on initial results?]

Phase 3 (Validation and Extension): [How to validate findings and extend to broader contexts?]

Expected Outcomes and Implications

If Hypothesis 1 is supported: [What would this mean for the field? What new questions arise?]

If Hypothesis 2 is supported: [Implications and new questions]

If Hypothesis 3 is supported: [Implications and new questions]

If multiple hypotheses are partially supported: [How might mechanisms combine or interact?]

Open Questions

[What questions remain even after these hypotheses are tested?]

---

References

[List key papers cited in the document, formatted consistently]

1. Author1, A.B., & Author2, C.D. (Year). Title of paper. Journal Name, Volume(Issue), pages. DOI or URL

2. [Continue for all citations]

---

Notes on Using This Template

• Replace all bracketed instructions with actual content
• Not all sections are mandatory - adapt to your specific hypothesis generation task
• For simpler phenomena, 3 hypotheses may be sufficient; complex phenomena may warrant 4-5
• Experimental designs should be detailed enough to be actionable but can be refined later
• Predictions should be as specific and quantitative as possible
• The template emphasizes both generating hypotheses and planning how to test them
• Citation format can be adjusted to field-specific standards

Experimental Design Patterns

Common Approaches to Testing Scientific Hypotheses

This reference provides patterns and frameworks for designing experiments across scientific domains. Use these patterns to develop rigorous tests for generated hypotheses.

Design Selection Framework

Choose experimental approaches based on:

• Nature of hypothesis: Mechanistic, causal, correlational, descriptive
• System studied: In vitro, in vivo, computational, observational
• Feasibility: Time, cost, ethics, technical capabilities
• Evidence needed: Proof-of-concept, causal demonstration, quantitative relationship

Laboratory Experimental Designs

In Vitro Experiments

When to use: Testing molecular, cellular, or biochemical mechanisms in controlled systems.

Common patterns:

1. Dose-Response Studies

• Purpose: Establish quantitative relationship between input and effect
• Design: Test multiple concentrations/doses of intervention
• Key elements:
• Negative control (no treatment)
• Positive control (known effective treatment)
• Multiple dose levels (typically 5-8 points)
• Technical replicates (≥3 per condition)
• Appropriate statistical analysis (curve fitting, IC50/EC50 determination)

Example application: "To test if compound X inhibits enzyme Y, measure enzyme activity at 0, 1, 10, 100, 1000 nM compound X concentrations with n=3 replicates per dose."

2. Gain/Loss of Function Studies

• Purpose: Establish causal role of specific component
• Design: Add (overexpression) or remove (knockout/knockdown) component
• Key elements:
• Wild-type control
• Gain-of-function condition (overexpression, constitutive activation)
• Loss-of-function condition (knockout, knockdown, inhibition)
• Rescue experiment (restore function to loss-of-function)
• Measure downstream effects

Example application: "Test if protein X causes phenotype Y by: (1) knocking out X and observing phenotype loss, (2) overexpressing X and observing phenotype enhancement, (3) rescuing knockout with X re-expression."

3. Time-Course Studies

• Purpose: Understand temporal dynamics and sequence of events
• Design: Measure outcomes at multiple time points
• Key elements:
• Time 0 baseline
• Early time points (capture rapid changes)
• Intermediate time points
• Late time points (steady state)
• Sufficient replication at each time point

Example application: "Measure protein phosphorylation at 0, 5, 15, 30, 60, 120 minutes after stimulus to determine peak activation timing."

In Vivo Experiments

When to use: Testing hypotheses in whole organisms to assess systemic, physiological, or behavioral effects.

Common patterns:

4. Between-Subjects Designs

• Purpose: Compare different groups receiving different treatments
• Design: Randomly assign subjects to treatment groups
• Key elements:
• Random assignment to groups
• Appropriate sample size (power analysis)
• Control group (vehicle, sham, or standard treatment)
• Blinding (single or double-blind)
• Standardized conditions across groups

Example application: "Randomly assign 20 mice each to vehicle control or drug treatment groups, measure tumor size weekly for 8 weeks, with experimenters blinded to group assignment."

5. Within-Subjects (Repeated Measures) Designs

• Purpose: Each subject serves as own control, reducing inter-subject variability
• Design: Same subjects measured across multiple conditions/time points
• Key elements:
• Baseline measurements
• Counterbalancing (if order effects possible)
• Washout periods (for sequential treatments)
• Appropriate repeated-measures statistics

Example application: "Measure cognitive performance in same participants at baseline, after training intervention, and at 3-month follow-up."

6. Factorial Designs

• Purpose: Test multiple factors and their interactions simultaneously
• Design: Cross all levels of multiple independent variables
• Key elements:
• Clear main effects and interactions
• Sufficient power for interaction tests
• Full factorial or fractional factorial as appropriate

Example application: "2×2 design crossing genotype (WT vs. mutant) × treatment (vehicle vs. drug) to test whether drug effect depends on genotype."

Computational/Modeling Experiments

When to use: Testing hypotheses about complex systems, making predictions, or when physical experiments are infeasible.

7. In Silico Simulations

• Purpose: Model complex systems, test theoretical predictions
• Design: Implement computational model and vary parameters
• Key elements:
• Well-defined model with explicit assumptions
• Parameter sensitivity analysis
• Validation against known data
• Prediction generation for experimental testing

Example application: "Build agent-based model of disease spread, vary transmission rate and intervention timing, compare predictions to empirical epidemic data."

8. Bioinformatics/Meta-Analysis

• Purpose: Test hypotheses using existing datasets
• Design: Analyze large-scale data or aggregate multiple studies
• Key elements:
• Appropriate statistical corrections (multiple testing)
• Validation in independent datasets
• Control for confounds and batch effects
• Clear inclusion/exclusion criteria

Example application: "Test if gene X expression correlates with survival across 15 cancer datasets (n>5000 patients total), using Cox regression with clinical covariates."

Observational Study Designs

When Physical Manipulation is Impossible or Unethical

9. Cross-Sectional Studies

• Purpose: Examine associations at a single time point
• Design: Measure variables of interest in population at one time
• Strengths: Fast, inexpensive, can establish prevalence
• Limitations: Cannot establish temporality or causation
• Key elements:
• Representative sampling
• Standardized measurements
• Control for confounding variables
• Appropriate statistical analysis

Example application: "Survey 1000 adults to test association between diet pattern and biomarker X, controlling for age, sex, BMI, and physical activity."

10. Cohort Studies (Prospective/Longitudinal)

• Purpose: Establish temporal relationships and potentially causal associations
• Design: Follow group over time, measuring exposures and outcomes
• Strengths: Can establish temporality, calculate incidence
• Limitations: Time-consuming, expensive, subject attrition
• Key elements:
• Baseline exposure assessment
• Follow-up at defined intervals
• Minimize loss to follow-up
• Account for time-varying confounders

Example application: "Follow 5000 initially healthy individuals for 10 years, testing if baseline vitamin D levels predict cardiovascular disease incidence."

11. Case-Control Studies

• Purpose: Efficiently study rare outcomes by comparing cases to controls
• Design: Identify cases with outcome, select matched controls, compare exposures
• Strengths: Efficient for rare diseases, relatively quick
• Limitations: Recall bias, selection bias, cannot calculate incidence
• Key elements:
• Clear case definition
• Appropriate control selection (matching or statistical adjustment)
• Retrospective exposure assessment
• Control for confounding

Example application: "Compare 200 patients with rare disease X to 400 matched controls without X, testing if early-life exposure Y differs between groups."

Clinical Trial Designs

12. Randomized Controlled Trials (RCTs)

• Purpose: Gold standard for testing interventions in humans
• Design: Randomly assign participants to treatment or control
• Key elements:
• Randomization (simple, block, or stratified)
• Concealment of allocation
• Blinding (participants, providers, assessors)
• Intention-to-treat analysis
• Pre-registered protocol and analysis plan

Example application: "Double-blind RCT: randomly assign 300 patients to receive drug X or placebo for 12 weeks, measure primary outcome of symptom improvement."

13. Crossover Trials

• Purpose: Each participant receives all treatments in sequence
• Design: Participants crossed over between treatments with washout
• Strengths: Reduces inter-subject variability, requires fewer participants
• Limitations: Order effects, requires reversible conditions, longer duration
• Key elements:
• Adequate washout period
• Randomized treatment order
• Carryover effect assessment

Example application: "Crossover trial: participants receive treatment A for 4 weeks, 2-week washout, then treatment B for 4 weeks (randomized order)."

Advanced Design Considerations

Sample Size and Statistical Power

Key questions:

• What effect size is meaningful to detect?
• What statistical test will be used?
• What alpha (significance level) and beta (power) are appropriate?
• What is expected variability in the measurement?

General guidelines:

• Conduct formal power analysis before experiment
• For pilot studies, n≥10 per group minimum
• For definitive studies, aim for ≥80% power
• Account for potential attrition in longitudinal studies

Controls

Types of controls:

• Negative control: No intervention (baseline)
• Positive control: Known effective intervention (validates system)
• Vehicle control: Delivery method without active ingredient
• Sham control: Mimics intervention without active component (surgery, etc.)
• Historical control: Prior data (weakest, avoid if possible)

Blinding

Levels:

• Open-label: No blinding (acceptable for objective measures)
• Single-blind: Participants blinded (reduces placebo effects)
• Double-blind: Participants and experimenters blinded (reduces bias in assessment)
• Triple-blind: Participants, experimenters, and analysts blinded (strongest)

Replication

Technical replicates: Repeated measurements on same sample

• Reduce measurement error
• Typically 2-3 replicates sufficient

Biological replicates: Independent samples/subjects

• Address biological variability
• Critical for generalization
• Minimum: n≥3, preferably n≥5-10 per group

Experimental replicates: Repeat entire experiment

• Validate findings across time, equipment, operators
• Gold standard for confirming results

Confound Control

Strategies:

• Randomization: Distribute confounds evenly across groups
• Matching: Pair similar subjects across conditions
• Blocking: Group by confound, then randomize within blocks
• Statistical adjustment: Measure confounds and adjust in analysis
• Standardization: Keep conditions constant across groups

Selecting Appropriate Design

Decision tree:

1. Can variables be manipulated?

• Yes → Experimental design (RCT, lab experiment)
• No → Observational design (cohort, case-control, cross-sectional)

2. What is the system?

• Cells/molecules → In vitro experiments
• Whole organisms → In vivo experiments
• Humans → Clinical trials or observational studies
• Complex systems → Computational modeling

3. What is the primary goal?

• Mechanism → Gain/loss of function, dose-response
• Causation → RCT, cohort study with good controls
• Association → Cross-sectional, case-control
• Prediction → Modeling, machine learning
• Temporal dynamics → Time-course, longitudinal

4. What are the constraints?

• Time limited → Cross-sectional, in vitro
• Budget limited → Computational, observational
• Ethical concerns → Observational, in vitro
• Rare outcome → Case-control, meta-analysis

Integrating Multiple Approaches

Strong hypothesis testing often combines multiple designs:

Example: Testing if microbiome affects cognitive function 1. Observational: Cohort study showing association between microbiome composition and cognition 2. Animal model: Germ-free mice receiving microbiome transplants show cognitive changes 3. Mechanism: In vitro studies showing microbial metabolites affect neuronal function 4. Clinical trial: RCT of probiotic intervention improving cognitive scores 5. Computational: Model predicting which microbiome profiles should affect cognition

Triangulation approach:

• Each design addresses different aspects/limitations
• Convergent evidence from multiple approaches strengthens causal claims
• Start with observational/in vitro, then move to definitive causal tests

Common Pitfalls

• Insufficient sample size (underpowered)
• Lack of appropriate controls
• Confounding variables not accounted for
• Inappropriate statistical tests
• P-hacking or multiple testing without correction
• Lack of blinding when subjective assessments involved
• Failure to replicate findings
• Not pre-registering analysis plans (clinical trials)

Practical Application for Hypothesis Testing

When designing experiments to test hypotheses:

1. Match design to hypothesis specifics: Causal claims require experimental manipulation; associations can use observational designs 2. Start simple, then elaborate: Pilot with simple design, then add complexity 3. Plan controls carefully: Controls validate the system and isolate the specific effect 4. Consider feasibility: Balance ideal design with practical constraints 5. Plan for multiple experiments: Rarely does one experiment definitively test a hypothesis 6. Pre-specify analysis: Decide statistical tests before data collection 7. Build in validation: Independent replication, orthogonal methods, convergent evidence

Hypothesis Quality Criteria

Framework for Evaluating Scientific Hypotheses

Use these criteria to assess the quality and rigor of generated hypotheses. A robust hypothesis should score well across multiple dimensions.

Core Criteria

1. Testability

Definition: The hypothesis can be empirically tested through observation or experimentation.

Evaluation questions:

• Can specific experiments or observations test this hypothesis?
• Are the predicted outcomes measurable?
• Can the hypothesis be tested with current or near-future methods?
• Are there multiple independent ways to test it?

Strong testability examples:

• "Increased expression of protein X will reduce cell proliferation rate by >30%"
• "Patients receiving treatment Y will show 50% reduction in symptom Z within 4 weeks"

Weak testability examples:

• "This process is influenced by complex interactions" (vague, no specific prediction)
• "The mechanism involves quantum effects" (if no method to test quantum effects exists)

2. Falsifiability

Definition: Clear conditions or observations would disprove the hypothesis (Popperian criterion).

Evaluation questions:

• What specific observations would prove this hypothesis wrong?
• Are the falsifying conditions realistic to observe?
• Is the hypothesis stated clearly enough to be disproven?
• Can null results meaningfully falsify the hypothesis?

Strong falsifiability examples:

• "If we knock out gene X, phenotype Y will disappear" (can be falsified if phenotype persists)
• "Drug A will outperform placebo in 80% of patients" (clear falsification threshold)

Weak falsifiability examples:

• "Multiple factors contribute to the outcome" (too vague to falsify)
• "The effect may vary depending on context" (built-in escape clauses)

3. Parsimony (Occam's Razor)

Definition: Among competing hypotheses with equal explanatory power, prefer the simpler explanation.

Evaluation questions:

• Does the hypothesis invoke the minimum number of entities/mechanisms needed?
• Are all proposed elements necessary to explain the phenomenon?
• Could a simpler mechanism account for the observations?
• Does it avoid unnecessary assumptions?

Parsimony considerations:

• Simple ≠ simplistic; complexity is justified when evidence demands it
• Established mechanisms are "simpler" than novel, unproven ones
• Direct mechanisms are simpler than elaborate multi-step pathways
• One well-supported mechanism beats multiple speculative ones

4. Explanatory Power

Definition: The hypothesis accounts for a substantial portion of the observed phenomenon.

Evaluation questions:

• How much of the observed data does this hypothesis explain?
• Does it account for both typical and atypical observations?
• Can it explain related phenomena beyond the immediate observation?
• Does it resolve apparent contradictions in existing data?

Strong explanatory power indicators:

• Explains multiple independent observations
• Accounts for quantitative relationships, not just qualitative patterns
• Resolves previously puzzling findings
• Makes sense of seemingly contradictory results

Limited explanatory power indicators:

• Only explains part of the phenomenon
• Requires additional hypotheses for complete explanation
• Leaves major observations unexplained

5. Scope

Definition: The range of phenomena and contexts the hypothesis can address.

Evaluation questions:

• Does it apply only to the specific case or to broader situations?
• Can it generalize across conditions, species, or systems?
• Does it connect to larger theoretical frameworks?
• What are its boundaries and limitations?

Broader scope (generally preferable):

• Applies across multiple experimental conditions
• Generalizes to related systems or species
• Connects phenomenon to established principles

Narrower scope (acceptable if explicitly defined):

• Limited to specific conditions or contexts
• Requires different mechanisms in different settings
• Context-dependent with clear boundaries

6. Consistency with Established Knowledge

Definition: Alignment with well-supported theories, principles, and empirical findings.

Evaluation questions:

• Is it consistent with established physical, chemical, or biological principles?
• Does it align with or reasonably extend current theories?
• If contradicting established knowledge, is there strong justification?
• Does it require violating well-supported laws or findings?

Levels of consistency:

• Fully consistent: Applies established mechanisms in new context
• Mostly consistent: Extends current understanding in plausible ways
• Partially inconsistent: Contradicts some findings but has explanatory value
• Highly inconsistent: Requires rejecting well-established principles (requires exceptional evidence)

7. Novelty and Insight

Definition: The hypothesis offers new understanding beyond merely restating known facts.

Evaluation questions:

• Does it provide new mechanistic insight?
• Does it challenge assumptions or conventional wisdom?
• Does it suggest unexpected connections or relationships?
• Does it open new research directions?

Novel contributions:

• Proposes previously unconsidered mechanisms
• Reframes the problem in a productive way
• Connects disparate observations
• Suggests non-obvious testable predictions

Note: Novelty alone doesn't make a hypothesis valuable; it must also be testable, parsimonious, and explanatory.

Comparative Evaluation

When evaluating multiple competing hypotheses:

Trade-offs and Balancing

Hypotheses often involve trade-offs:

• More parsimonious but less explanatory power
• Broader scope but less testable with current methods
• Novel insights but less consistent with current knowledge

Evaluation approach:

• No hypothesis needs to be perfect on all dimensions
• Identify each hypothesis's strengths and weaknesses
• Consider which criteria are most important for the specific phenomenon
• Note which hypotheses are most immediately testable
• Identify which would be most informative if supported

Distinguishability

Key question: Can experiments distinguish between competing hypotheses?

• Identify predictions that differ between hypotheses
• Prioritize hypotheses that make distinct predictions
• Note which experiments would most efficiently narrow the field
• Consider whether hypotheses could all be partially correct

Common Pitfalls

Untestable Hypotheses

• Too vague to generate specific predictions
• Invoke unobservable or unmeasurable entities
• Require technology that doesn't exist

Unfalsifiable Hypotheses

• Built-in escape clauses ("may or may not occur")
• Post-hoc explanations that fit any outcome
• No specification of what would disprove them

Overly Complex Hypotheses

• Invoke multiple unproven mechanisms
• Add unnecessary steps or entities
• Complexity not justified by explanatory gains

Just-So Stories

• Plausible narratives without testable predictions
• Explain observations but don't predict new ones
• Impossible to distinguish from alternative stories

Practical Application

When generating hypotheses:

1. Draft initial hypotheses focusing on mechanistic explanations 2. Apply quality criteria to identify weaknesses 3. Refine hypotheses to improve testability and clarity 4. Develop specific predictions to enhance testability and falsifiability 5. Compare systematically across all criteria 6. Prioritize for testing based on distinguishability and feasibility

Remember: The goal is not a perfect hypothesis, but a set of testable, falsifiable, informative hypotheses that advance understanding of the phenomenon.

Literature Search Strategies

Comprehensive literature search grounds hypotheses in existing evidence. This reference provides efficient strategies for PubMed (biomedical) and general scientific search.

Search Strategy Framework

Three-Phase Approach

1. Broad exploration: Understand the landscape, identify key concepts 2. Focused searching: Target specific mechanisms, theories, findings 3. Citation mining: Follow references from key papers

Before Searching

Clarify: What evidence is needed? What study types are relevant? What time frame? What evidence level (mechanistic, correlational, causal)?

PubMed Search Strategies

Use PubMed for: Biomedical/life sciences, clinical studies, molecular/cellular mechanisms, disease etiology, drug research.

Key Techniques

1. Start with Reviews

• Add "review" to search terms or use Article Type filters
• Reviews synthesize literature and provide comprehensive reference lists
• Example: https://pubmed.ncbi.nlm.nih.gov/?term=wound+healing+diabetes+review

2. Use MeSH Terms

• Standardized vocabulary captures concept variations
• "Myocardial Infarction" finds "heart attack," "MI," "cardiac infarction"

3. Boolean Operators

• AND: All terms required (diabetes AND wound healing AND inflammation)
• OR: Any term ((Alzheimer OR dementia) AND gut microbiome)
• NOT: Exclude (cancer treatment NOT surgery)
• Quotes: Exact phrases ("oxidative stress")
• Wildcards: gene* finds gene, genes, genetic, genetics

4. Filters

• Publication types: Clinical Trial, Meta-Analysis, Systematic Review, RCT
• Date ranges: Recent (2-5 years) for cutting-edge, older for foundational

5. Similar Articles and Cited By

• Find one relevant paper, use "Similar articles" for related work
• "Cited by" finds newer work building on it

PubMed Search Patterns

Web Search Strategies

Use web search for: Non-biomedical sciences, interdisciplinary topics, preprints, grey literature, cross-domain analogies.

Key Techniques

1. Target Academic Sources

• site:arxiv.org (physics, CS, math, quantitative biology)
• site:biorxiv.org (biology preprints)
• site:edu (academic institutions)
• filetype:pdf (academic papers)

2. Domain-Specific Terms

• Chemistry: "mechanism," "reaction pathway," "synthesis"
• Physics: "model," "theory," "experimental validation"
• Materials: "properties," "characterization"
• Ecology: "population dynamics," "community structure"

3. Boolean Logic

• Use quotes for exact phrases: "spike protein mutation"
• Use OR for alternatives: (transmissibility OR transmission rate)
• Combine: "spike protein" AND (transmissibility OR virulence) AND mutation

4. Author/Lab Searches

• Find relevant paper → search author's other work → find lab website

Comprehensive Search Workflow

1. Reviews first: Identify key concepts, influential papers, current state 2. Focused primary research: Specific mechanisms, experimental evidence 3. Broaden with web search: Related fields, preprints, analogies 4. Citation mining: Follow references, use "cited by" 5. Iterate: Add new terms, narrow/broaden as needed

Evaluating Source Quality

Primary Research Quality

Review Quality

• Systematic reviews (highest): Pre-defined search, explicit criteria, quality assessment, meta-analysis
• Narrative reviews (variable): Expert synthesis, may have selection bias, useful for context

Time Allocation

Diminishing Returns

Stop when: Finding same papers repeatedly, new searches yield irrelevant papers, sufficient evidence gathered, multiple lines converge.

Search more when: Major gaps remain, conflicting evidence unresolved, hypothesis inconsistent with literature.

Common Pitfalls

Special Cases

Emerging topics (limited literature):

• Search analogous phenomena in related systems
• Look for preprints and conference abstracts
• Note limited evidence in hypothesis generation

Controversial topics (conflicting literature):

• Document both sides systematically
• Look for methodological differences explaining conflict
• Generate hypotheses explaining discrepancies

Interdisciplinary topics:

• Search each field's databases with field-specific terminology
• Look for bridging papers citing across fields

Integration with Hypothesis Generation

Literature informs hypotheses through:

• Established mechanisms to apply to new contexts
• Known pathways relevant to phenomenon
• Analogies from different systems
• Gaps suggesting novel mechanisms

Balance literature dependence:

• Too dependent: Hypotheses merely restate known facts
• Too independent: Hypotheses ignore relevant evidence
• Optimal: Grounded in evidence, extend understanding in novel ways, generate testable predictions