Research Gap Examples by Field (With How to Find Your Own)

A research gap example is most useful when it shows the shape of a good gap: specific, grounded in named literature, and answerable with a study you could actually run. Abstract definitions don't transfer between fields — a methodological gap in clinical medicine looks nothing like one in machine learning. So here are worked examples across six fields, each stated the way a proposal should state it, followed by the method for producing your own.

If you haven't read it yet, the companion guide How to Find a Research Gap covers the seven gap types these examples draw on.

Medicine & public health

• "Community-based mental-health interventions show strong effects in high-income settings but have not been trialed in low-resource primary care." — a population + geographic gap: the finding exists, the setting doesn't.
• "Trials disagree on whether intermittent fasting improves insulin sensitivity, and none stratify by baseline metabolic health." — an evidence gap: the contradiction is the opening.
• "Screening tools for postpartum depression are validated in mothers but not in non-birthing partners, despite documented paternal risk." — a population gap hiding in a well-studied instrument.

Computer science & machine learning

• "Retrieval-augmented generation is benchmarked on open-domain QA; its failure modes on long, citation-dense scientific text are uncharacterised." — a methodological + data gap.
• "Fairness metrics for classifiers assume a static population; their behaviour under distribution shift over time is largely unstudied." — a theoretical gap: the framework doesn't cover the observed condition.

Education

• "Adaptive-learning platforms report gains on short-term post-tests, but their effect on retention beyond one semester is rarely measured." — a methodological gap: the outcome window is too short.
• "AI writing tutors are studied in first-language English classrooms; their effect for multilingual writers is largely unexamined." — a population gap. (Education is one of the largest gap corpora we index — see the education research gaps hub.)

Psychology

• "Replications of classic priming effects cluster in WEIRD samples; whether the effects hold in non-Western populations is unresolved." — a **population
  • evidence** gap.
• "Interventions for adolescent anxiety are evaluated at the individual level; peer-network spillover effects are theorised but not measured." — a theoretical gap awaiting a study design.

Engineering

• "Battery-degradation models assume steady-state cycling; behaviour under the irregular loads of real-world grid storage is not characterised." — a practical / application gap.
• "Digital-twin fault detection is validated on single components; system-level cascading-fault behaviour has no standard benchmark." — a data gap (no benchmark exists yet).

Biology

• "Deep-learning models for protein structure are validated on globular proteins; their accuracy on intrinsically disordered regions remains unquantified." — a methodological + data gap.
• "Microbiome-host interaction studies are dominated by gut samples; skin and respiratory niches are comparatively unmapped." — a geographic gap, where "geography" is anatomical.

What every one of these has in common

Read them again and the pattern is identical across fields:

1. A named body of work that gets close ("interventions show strong effects…", "models are validated on…").
2. A precise place it stops ("…but not in low-resource primary care").
3. An implied study — you can already picture the design that would fill it.

That three-part shape is the test. If your gap doesn't have all three, it's still a topic, not a gap.

How to find your own — the fast method

The manual route works: read recent reviews, mine the "limitations" and "future work" sections of ten papers, and look for the untested edge. It's also slow, and no single reader can hold a field's whole citation network in their head.

That coverage problem is what the free AI Research Gap Finder solves: enter a topic and it surfaces open, citable gaps distilled from the 250-million-paper OpenAlex corpus — each with the key and most-cited literature around it and a starting proposal. Treat what it returns the way you'd treat a colleague's suggestions: a shortlist to confirm against the "future work" sections it points you to. The tool does the coverage; you do the judgment.

Once you've settled on a gap and written the paper, two more free tools finish the pipeline: the Journal Finder matches your title and abstract to a ranked shortlist of venues (with predatory-journal flags), and AI Peer Review gives editor-ready feedback before you submit.

The short version

• A good research gap has a fixed shape in every field: work that gets close, a precise place it stops, and an implied study that fills it.
• Name the gap type — evidence, population, methodological, theoretical, practical, geographic, or data — and the study design usually follows.
• To cover more ground than any one reader can, start from the research gap finder and refine from there.