The research credit case law map: every doctrine, every leading case

Four decades of litigation have shaped the research credit into what examiners and courts apply today, and nearly every position the IRS takes on audit traces to a handful of decisions. The doctrine splits into five bodies of law: what counts as qualified research under the Section 41(d) four-part test; when research is "funded" by someone else and therefore excluded; how the base amount is computed and the consistency rule enforced; what substantiation courts demand; and where the line falls between a process of experimentation and routine engineering. This guide maps each doctrine to its controlling cases, notes which courts decided what, and shows how current exam positions descend from each line of authority.

The doctrine-by-case table

The table below organizes the leading research credit cases by the doctrine each one controls, with the deciding court and the proposition for which the case is most often cited.

Qualified research: the four-part test cases

Section 41(d) defines qualified research through four requirements: expenditures eligible under Section 174 (now Section 174A for domestic research), research undertaken to discover information technological in nature, an intent to develop a new or improved business component, and substantially all activities constituting a process of experimentation. The framework is explained in detail in the four-part test guide; the cases here show how it plays out in court.

Union Carbide is the foundational modern trial. Over a months-long record, the Tax Court examined dozens of projects at chemical plants and held that experiments run on production-scale equipment — during ordinary commercial production — could be qualified research where the taxpayer was testing hypotheses about process improvements. The court allowed some supply costs consumed in those trials but disallowed the vast bulk of claimed supplies, because raw materials that would have been consumed in production anyway were not costs of the research. The case remains the template for manufacturing claims: plant trials can qualify, but the incremental-cost discipline is strict.

Suder extended the framework to product development. The Tax Court walked through twelve sample projects at a phone-systems company and found eleven qualified — establishing that development need not be pioneering, only genuinely uncertain to the taxpayer, and that a structured development process can itself be the process of experimentation. But the court cut the CEO's claimed wages sharply as unreasonable compensation, a holding examiners still cite whenever founder wages dominate a claim.

Siemer Milling is the cautionary counterpart. A flour miller claimed credits for new flour formulations and process tweaks but offered almost no evidence of hypotheses, testing protocols, or evaluation of alternatives. The Tax Court denied every project, holding the taxpayer failed both the technological-in-nature and process-of-experimentation prongs — not because milling cannot involve science, but because nothing in the record showed that it did here.

Little Sandy Coal sharpened the "substantially all" requirement. The Seventh Circuit affirmed denial of credits for first-of-a-kind vessels because the taxpayer never showed that at least 80 percent of its activities, measured under Treas. Reg. §1.41-4(a)(6), constituted elements of a process of experimentation. Novelty of the finished product does not carry the burden; the taxpayer must quantify the experimentation fraction or invoke the shrink-back rule to a smaller component.

How exam positions trace to these cases

Current Information Document Requests asking taxpayers to "identify the alternatives evaluated and the method of evaluation" descend directly from Siemer Milling. The demand to quantify the percentage of project activities that constitute experimentation — now baked into Section G of Form 6765 business-component reporting — is Little Sandy Coal operationalized. Reasonableness challenges to owner wages are Suder's second holding at work.

Funded research: the risk-and-rights cases

Section 41(d)(4)(H) excludes research "funded by any grant, contract, or otherwise by another person." Treas. Reg. §1.41-4A(d) supplies the two-part gloss the courts apply: research is not funded if (1) payment is contingent on the success of the research, so the taxpayer bears the financial risk of failure, and (2) the taxpayer retains substantial rights in the results. Both prongs must be satisfied; failure of either forfeits the claim for that contract. The doctrine, and how to review contracts against it, is covered in the funded research exclusion guide.

Fairchild Industries is the fountainhead. The Federal Circuit held that a fixed-price incentive contract with the Air Force was not funding, because Fairchild was paid only for work that met contract specifications — payment was contingent on success, so Fairchild bore the risk. The decision established that the inquiry looks to who bears the cost if the research fails, judged from the contract terms at signing, not from whether payment ultimately occurred.

Geosyntec applied the framework to a portfolio of environmental engineering contracts. The Eleventh Circuit held fixed-price contracts unfunded (the firm ate any overruns) but capped cost-plus contracts funded, because the client reimbursed costs as incurred up to the cap regardless of research outcome. The case is the standard citation for the proposition that contract payment structure, not project riskiness, controls.

Populous Holdings brought the doctrine to professional services. On summary judgment, the Tax Court held an architecture firm's fixed-fee design contracts unfunded: the firm bore the risk of redesign at its own cost, and it retained substantial rights because nothing in the contracts barred it from reusing the knowledge and design know-how in future projects. Retained rights need not be exclusive or patent-like.

Dynetics shows the doctrine's granularity. The Court of Federal Claims analyzed a defense contractor's agreements contract by contract, refusing to generalize across a portfolio, and read inspection, acceptance, and rework clauses closely to decide where real economic risk sat. Most of the sampled contracts failed. The lesson: funding is decided at the level of the individual agreement, and boilerplate acceptance clauses are not a substitute for genuine contingency.

Base period and computation: Norwest and the consistency rule

Norwest Corp. v. Commissioner, 110 T.C. 454 (1998), is the leading Tax Court decision on two fronts. First, it articulated early standards for internal-use software — including a discovery-type test and a high-threshold-of-innovation requirement — that shaped two decades of regulatory drafting culminating in the current Treas. Reg. §1.41-4(c)(6) rules. Second, and more durably, it enforced the Section 41(c)(6) consistency rule: QREs in the base period must be determined on the same basis as QREs in the credit year. A taxpayer cannot apply a generous definition to the credit year and a stingy one to the base years to inflate the increment. The consistency rule survives fully intact under the alternative simplified credit, where the prior-three-year average serves as the base — an examiner who knocks out a category of credit-year QREs will also insist it be removed from the base, and vice versa.

Substantiation: the Eustace and Shami line

No doctrine generates more exam adjustments than substantiation, and two cases define its poles.

Eustace (the Tax Court decision affirmed by the Seventh Circuit in 2002, involving Applied Systems' insurance-agency software) held that ordinary software maintenance, debugging, and feature work was not a process of experimentation, and that without evidence distinguishing experimental work from routine development, no portion of the wages qualified. The Seventh Circuit's affirmance is the classic statement that software development is not automatically research.

Shami addressed executive wages. The Fifth Circuit sustained disallowance of large wage QREs claimed for two senior executives of a hair-care company because the record contained no credible evidence they personally performed or directly supervised qualified research. Title and general oversight are not qualified services under Section 41(b).

The counterweight is McFerrin, where the Fifth Circuit held that once a taxpayer establishes that qualified research occurred, the court should estimate the associated expenses under the Cohan rule rather than deny everything for imperfect records. The sequence matters: estimation rescues quantification, never qualification. Siemer Milling and Eustace failed at the qualification threshold, where Cohan cannot reach.

The practical hierarchy of substantiation outcomes looks like this, moving from full allowance to full denial.

Contemporaneous project records (Union Carbide model)75% of claim sustained

Credible testimony + sampling (Suder model)70% of claim sustained

Proof research occurred, weak cost records (McFerrin)40% of claim sustained

Assertions without experimentation evidence (Siemer)0% of claim sustained

Illustrative pattern drawn from the cases discussed; actual outcomes turn on each record.

Process of experimentation versus routine engineering

The recurring merits question across all of these cases is whether the taxpayer's work was a process of experimentation — the systematic evaluation of alternatives to resolve uncertainty — or routine engineering, adaptation, and quality control excluded under Section 41(d)(4). The line the cases draw:

• Qualifies: plant-scale trials testing process hypotheses (Union Carbide); iterative product development through structured design, build, and test cycles (Suder).
• Fails: development where no alternatives were identified or evaluated on the record (Siemer Milling); routine coding, debugging, and adaptation (Eustace); whole-project claims where the experimentation fraction was never measured (Little Sandy Coal); work whose uncertainty was aesthetic rather than technological (Leon Max, applying the Section 41(d)(3)(B) style-and-taste exclusion).

Leon Max deserves particular note as the boundary case for creative industries: a fashion designer's sample-making process failed because the uncertainties resolved — fit, drape, look — were matters of taste, not principles of engineering or physical science.

Using the map in practice

Three working rules fall out of the case law. First, build the file the courts reward: project-level records identifying the uncertainty, the alternatives, and the testing method, because Siemer Milling and Little Sandy Coal show that the absence of that record is fatal regardless of the underlying science. Second, read every research contract against Fairchild's risk-and-rights test before claiming contract research, and do it agreement by agreement as Dynetics requires. Third, keep the base honest: Norwest's consistency rule means every credit-year position has a base-period mirror. When an exam letter arrives, identifying which case the examiner's position descends from is the first step in building the defense — each doctrine has both a government-favorable and a taxpayer-favorable line, and the record you assembled determines which one applies. The individual briefs linked throughout this guide, and the IRS research credit overview, fill in the rest.

The map does not favor aggressive claims. Roughly half of the leading cases ended badly for the taxpayer, and the ones that ended well — Union Carbide, Suder, Populous — were won on records built long before the notice of deficiency. That is the real lesson of the litigation history.