Claim Construction and Markman Hearings: How Courts Decide What a Patent Actually Covers

Summary. Why the fight over a handful of words usually decides the whole case.

The sentence that decides the case

There is a moment in nearly every patent case when a lawyer realizes the entire dispute has collapsed into a single word.

Sometimes it is a preposition. Sometimes it is a modifier nobody noticed during prosecution — substantially, adjacent, a plurality of. Sometimes it is a word that seemed perfectly clear until a competitor built something that satisfied one reading and not the other. Whatever it is, the case now turns on what that word means, and the person who will decide is a federal district judge who may never have seen a patent in this technology before.

That proceeding is called claim construction, and the hearing at which it happens is called a Markman hearing, after the Supreme Court case that put it in a judge's hands. Practitioners sometimes call it "the case within the case," which undersells it. It is usually just the case. Studies of patent litigation consistently find that a large fraction of cases settle or resolve on summary judgment within weeks of a claim construction order, because once the words are fixed, the answer to "does this product infringe?" often becomes arithmetic.

Understanding claim construction is therefore not a specialty within patent law. It is the center of it.

What a patent claim actually is

Start with the object itself, because a great deal of confusion comes from thinking a patent protects an invention in some loose, general sense. It does not. A patent protects exactly what its claims say it protects, and nothing else.

A United States patent has three functional parts:

The specification — the written description of the invention, including drawings. It teaches a person skilled in the field how to make and use the invention. 35 U.S.C. § 112 requires that the specification contain a written description sufficient to enable that person to practice the full scope of what is claimed.
The claims — numbered sentences at the end, each of which defines the legal boundary of the monopoly. Section 112(b) requires claims "particularly pointing out and distinctly claiming the subject matter which the inventor . . . regards as the invention."
The prosecution history — the paper trail of the negotiation between the applicant and the patent examiner, including rejections, amendments, and arguments.

The standard metaphor is that claims are like the metes and bounds of a deed: the specification tells you what the property looks like, but the claims tell you where the fence line runs. The metaphor is imperfect — surveyors have instruments, and words do not — but it captures the essential point. Infringement is measured against the claims, not against the described embodiment, not against the commercial product, and not against what the inventor meant to invent.

A claim is a single sentence with a peculiar grammar:

1. An apparatus for filtering fluid, comprising: a housing defining an inlet and an outlet; a filter element disposed within the housing; and a bypass valve coupled to the housing and configured to open when a pressure differential across the filter element exceeds a threshold.

The opening phrase is the preamble. The transitional word — here, comprising — is a term of art meaning "including at least," so an accused device with additional elements still infringes. (Consisting of, by contrast, is closed: extra elements defeat infringement.) The indented phrases are limitations or elements, and here is the rule that governs everything: every limitation must be present in the accused product. Miss one and there is no literal infringement, no matter how similar the rest is. Practitioners call this the all-elements rule, and it is why defendants hunt for the single limitation that is missing rather than arguing the general dissimilarity of the products.

That structure explains why the construction of one word can end a case. If bypass valve is construed to require a mechanically actuated valve, and the accused product uses a solenoid controlled by firmware, the case may be over. If it is construed to mean any structure that permits flow around the filter element, the case proceeds to a jury.

Why a judge decides, not a jury

For most of the twentieth century, courts treated claim meaning as an issue that could go to the jury along with everything else. That produced exactly what one would expect: inconsistent outcomes, unreviewable verdicts, and patents that meant different things in different courthouses.

In Markman v. Westview Instruments, Inc., 517 U.S. 370 (1996), the Supreme Court held unanimously that the construction of patent claims is a matter for the court, not the jury, and that this allocation does not violate the Seventh Amendment right to jury trial. The Court's reasoning was partly historical and partly functional. Historically, it found no clear practice of submitting claim meaning to eighteenth-century juries. Functionally — and this is the passage that shaped the next thirty years — the Court emphasized the value of uniformity:

"[T]reating interpretive issues as purely legal will promote (though it will not guarantee) intrajurisdictional certainty through the application of stare decisis on those questions not yet subject to interjurisdictional uniformity."

In other words: a patent should mean the same thing on Monday in Delaware as it means on Friday in Texas, and the only way to get there is to have judges decide, with appellate review to keep them aligned. The exclusive appellate jurisdiction of the Federal Circuit over patent appeals, conferred by 28 U.S.C. § 1295, is the other half of that design.

Markman did not tell district courts when to construe claims or how to run the proceeding. It created a task and left the procedure to the districts, which is why claim construction practice varies so much from courthouse to courthouse even though the substantive law is national.

Teva and the standard of review

For nearly two decades after Markman, the Federal Circuit reviewed all aspects of claim construction de novo — that is, with no deference at all to the district judge. A trial judge could hold a three-day hearing with live expert testimony, write a seventy-page opinion, and have the appellate court substitute its own reading without a word of deference. Reversal rates were, unsurprisingly, high.

Teva Pharmaceuticals USA, Inc. v. Sandoz, Inc., 574 U.S. 318 (2015) changed that, though less dramatically than either side hoped. The Court held that claim construction remains a legal question reviewed de novo, but that any subsidiary factual findings about extrinsic evidence — what a term meant to skilled artisans at the time, what a scientific convention was, what an expert credibly testified — are reviewed for clear error under Federal Rule of Civil Procedure 52(a)(6).

The practical consequence is a two-track record. When a district court construes a term using only the patent, its prosecution history, and dictionaries, the review is de novo and the appellate court is free to disagree. When the district court resolves a genuine dispute about underlying science by crediting one expert over another, that finding gets deference.

This creates a real strategic choice, and it is one of the few places where the standard of review can be engineered:

A party confident in the district judge wants the record to contain contested extrinsic evidence and wants the court to make explicit factual findings, so that the ruling travels to the Federal Circuit wearing armor.
A party expecting to lose below wants the construction to rest purely on intrinsic evidence, so the Federal Circuit reviews it fresh.

Sophisticated litigants brief this deliberately. Ask any experienced patent trial lawyer why they submitted a fifty-page expert declaration on a term that seemed obvious, and the honest answer is often: to build a clear-error shield.

The evidence hierarchy: what the judge is allowed to read

The governing framework comes from the Federal Circuit's en banc decision in Phillips v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005), which remains the single most-cited authority in claim construction briefing. It establishes an ordered hierarchy.

Tier one: the claims themselves

Courts start with the words, read as a person of ordinary skill in the art would have understood them at the time of the invention — not today, and not as a layperson would. That hypothetical person, universally abbreviated POSITA, is a legal construct with a defined education and experience level, and the parties fight about their qualifications because a more sophisticated POSITA reads terms differently than a less sophisticated one.

Two structural canons do real work here:

Claim differentiation. Where an independent claim uses a general term and a dependent claim adds a limitation, the general term presumptively does not include that limitation — otherwise the dependent claim would be redundant. If claim 1 says "a fastener" and claim 3 says "the fastener of claim 1, wherein the fastener is a threaded screw," then "fastener" in claim 1 is presumptively broader than screws.

Consistent usage. The same term is presumed to mean the same thing throughout the patent, and different terms are presumed to mean different things.

Tier two: the specification

Phillips famously called the specification "the single best guide to the meaning of a disputed term" and "usually dispositive." It is here that most cases are actually decided, and here that the central tension of claim construction lives:

The court must read the claims in light of the specification, without reading limitations from the specification into the claims.

That sentence is easy to state and genuinely hard to apply. Every claim construction dispute of any interest is a fight about which side of that line a proposed limitation falls on. The patentee says the accused construction improperly imports an embodiment; the accused infringer says the specification defines the invention that narrowly.

Two escape hatches let the specification control outright:

Lexicography. A patentee may act as their own dictionary, defining a term differently from ordinary usage — but only by doing so with reasonable clarity and precision. A definition that appears as "as used herein, module means . . ." will govern. A definition buried in a passing description usually will not.

Disavowal. A patentee who clearly and unmistakably disclaims scope is held to it. Language like "the present invention requires," "in all embodiments," "unlike prior art devices, the invention never . . ." is dangerous drafting, and defense counsel comb specifications for exactly these phrases. The standard is demanding — the Federal Circuit repeatedly says disavowal must be "clear and unmistakable" — but when it is met, it is dispositive.

Tier three: the prosecution history

The file wrapper shows what the applicant told the Patent Office to get the patent allowed. Because the public is entitled to rely on those statements, arguments made to distinguish prior art can narrow claim scope. This is prosecution disclaimer, and it is the most common way a patentee loses a term it thought it owned.

The pattern is familiar. An examiner rejects claim 1 over a prior art reference. The applicant amends and argues: "Applicant's invention, unlike Smith, requires the sensor be positioned downstream of the valve." Eight years later the patentee sues a company whose sensor sits upstream and argues that position does not matter. It is too late. The prosecution history is "intrinsic evidence" of the highest order, and the applicant's own words control.

The prosecution history of related applications — continuations, parents, foreign counterparts in some circumstances — may also be consulted, which is why sophisticated portfolio management treats every office action response as a potential future exhibit.

Tier four: extrinsic evidence

Dictionaries, treatises, learned articles, and expert testimony sit at the bottom. Courts may consider them, but Phillips was explicitly a corrective against an earlier line of cases that had elevated dictionaries above the specification. Extrinsic evidence is "less reliable" and may not be used to contradict an intrinsic-evidence meaning.

Expert testimony matters most for two things: establishing who the POSITA is, and explaining the underlying technology so the judge can read the patent intelligently. It matters least when offered as pure advocacy — an expert's opinion that a term "means" what the client needs it to mean carries little weight, and judges say so in orders with some regularity.

Indefiniteness: when no construction is possible

Sometimes the answer is that the term cannot be construed at all. Section 112(b) requires claims that "particularly point out and distinctly claim." A claim that fails that standard is indefinite and therefore invalid.

The modern test comes from Nautilus, Inc. v. Biosig Instruments, Inc., 572 U.S. 898 (2014), which rejected the Federal Circuit's older, patentee-friendly "insolubly ambiguous" standard and replaced it with this:

"[A] patent is invalid for indefiniteness if its claims, read in light of the specification delineating the patent, and the prosecution history, fail to inform, with reasonable certainty, those skilled in the art about the scope of the invention."

Reasonable certainty is the operative phrase. It permits some imprecision — the Court acknowledged that "absolute precision is unattainable" — while demanding enough clarity that competitors can determine what is off limits.

Indefiniteness fights cluster around a few recurring categories:

Terms of degree — substantially, about, close to, sufficient. These survive when the specification supplies an objective baseline for measurement and fail when it does not.
Subjective terms — aesthetically pleasing, user friendly, high quality. These usually fail, because the scope depends on who is looking.
Terms with no antecedent basis — a claim referring to "the controller" when no controller was previously introduced.
Mixed statutory class claims — a claim that covers both an apparatus and a method of using it, so that a competitor cannot tell whether making or using triggers infringement.
Unmeasurable parameters — a numeric limitation where the specification discloses no measurement technique and different accepted techniques give different answers.

Because invalidity must be proven by clear and convincing evidence — the standard confirmed in Microsoft Corp. v. i4i Ltd. Partnership, 564 U.S. 91 (2011) under the presumption of validity in 35 U.S.C. § 282 — indefiniteness is a hard win. But it is a complete win when it lands, which is why defendants raise it on nearly every term of degree.

Means-plus-function claiming

Section 112(f) permits a claim element to be expressed as "a means . . . for performing a specified function" without reciting the structure. The bargain is strict: such a claim is construed to cover only the structure disclosed in the specification for performing that function, plus equivalents.

Two rules make this a live battlefield:

Using the word "means" creates a rebuttable presumption that § 112(f) applies. Omitting it creates the opposite presumption — but the Federal Circuit's Williamson v. Citrix decision (en banc, 2015) made that second presumption much easier to overcome. Nonce words like module, mechanism, element, device, unit, and component now frequently trigger means-plus-function treatment even without "means."
If the specification discloses no corresponding structure, the claim is indefinite. For computer-implemented functions, the Federal Circuit generally requires disclosure of an algorithm, not merely a general-purpose processor. A specification that says "a processing module configured to determine eligibility" without describing how is a frequent casualty.

This is why software patent claims drafted with functional language are so vulnerable, and why accused infringers routinely argue that innocuous-sounding claim terms are secretly means-plus-function limitations with no supporting structure.

The doctrine of equivalents and its cage

Literal infringement requires every limitation to be present as construed. The doctrine of equivalents permits a finding of infringement where an accused element is not literally within a limitation but differs only insubstantially — the classic formulation from Graver Tank & Mfg. Co. v. Linde Air Products Co., 339 U.S. 605 (1950) asks whether it "performs substantially the same function in substantially the same way to obtain the same result."

Warner-Jenkinson Co. v. Hilton Davis Chemical Co., 520 U.S. 17 (1997) preserved the doctrine but disciplined it, holding that equivalence must be assessed element by element, not for the invention as a whole — a limitation that prevents the doctrine from swallowing the all-elements rule.

Then Festo Corp. v. Shoketsu Kinzoku Kogyo Kabushiki Co., 535 U.S. 722 (2002) built the cage. When a claim is narrowed by amendment during prosecution for reasons related to patentability, prosecution history estoppel presumptively bars the patentee from recapturing the surrendered territory through equivalents. The Court rejected an absolute bar in favor of a rebuttable presumption, which the patentee can overcome by showing the equivalent was unforeseeable at the time, that the rationale for amendment bore only a tangential relation to the equivalent, or some other reason why the applicant could not reasonably have been expected to describe it.

In practice, Festo means the doctrine of equivalents is available mostly for claims that sailed through prosecution untouched — an increasingly rare species. Claim construction and equivalents are therefore joined at the hip: every amendment that saved a claim from an examiner also fenced in its future reach.

How a Markman proceeding actually runs

There is no national procedural rule for claim construction. What exists instead is a patchwork of local patent rules, adopted first in the Northern District of California and now used in some form in most patent-heavy districts. The mechanics are broadly similar and roughly track the general discovery framework of Federal Rule of Civil Procedure 26.

A representative sequence:

Stage	Typical timing	What happens
Infringement contentions	14–45 days after scheduling conference	Patentee identifies asserted claims and charts them against accused products
Invalidity contentions	45–60 days later	Defendant identifies prior art and invalidity theories, including indefiniteness
Exchange of proposed terms	~2 weeks later	Each side lists claim terms it believes need construction
Exchange of preliminary constructions and evidence	~3 weeks later	Proposed constructions plus supporting intrinsic and extrinsic evidence, including expert declarations
Meet and confer	1–2 weeks	Parties narrow the list; courts increasingly cap the number of terms
Joint claim construction statement	~30 days before briefing	Filed with the court: agreed constructions, disputed terms, each side's position
Opening, responsive, reply briefs	4–8 weeks	Sometimes simultaneous, sometimes sequential
Technology tutorial	Before or at the hearing	Neutral-ish education for the judge; sometimes live, sometimes video
Markman hearing	Varies enormously	Argument, occasionally live expert testimony
Order	Weeks to many months	Constructions issue, often with reasoning

Two structural features deserve emphasis.

The number of terms is contested, and courts are impatient. Parties who identify twenty-five disputed terms will usually be told to pick ten, or five. Many districts now impose caps by standing order. This forces a genuine triage problem, discussed below.

The technology tutorial is underrated. A judge who does not understand how the technology works cannot construe the claims sensibly, and the tutorial is the only part of the proceeding without an opposing narrative running in parallel. Good tutorials teach the field, not the case. Bad tutorials are argument in a lab coat, and judges notice.

A worked example

Corvidae Instruments holds U.S. Patent No. 9,XXX,XXX covering a portable water-quality analyzer. Claim 1 recites, in relevant part:

a sample chamber; a light source positioned to direct light through the sample chamber; a detector spaced apart from the light source and positioned to receive light that has passed through the sample chamber; and a processor configured to determine turbidity from an output of the detector.

Halcyon Analytics sells a competing device in which the light source and detector are mounted on a single integrated circuit substrate, separated by roughly two millimeters and an opaque partition.

Corvidae sues. Halcyon's non-infringement position rests entirely on one phrase: spaced apart.

Halcyon's proposed construction: "physically separated by a distance sufficient that the light source and detector are discrete components not sharing a common substrate."

Corvidae's proposed construction: "not in contact with one another."

Where does each side look?

Halcyon goes to the specification. Every figure shows the light source and detector on opposite walls of the chamber. The background section criticizes prior art devices in which "co-located optical components introduce crosstalk that degrades measurement accuracy." The summary states that "the invention overcomes this limitation by positioning the detector remotely from the source." Halcyon also finds a prosecution history gem: to overcome a rejection over a reference disclosing an integrated optical package, the applicant argued that the claimed detector is "spaced apart from — and therefore optically isolated from — the source, unlike the monolithic assembly of Tanaka."

Corvidae goes to claim differentiation. Dependent claim 6 recites "wherein the detector is positioned on a wall of the sample chamber opposite the light source." If claim 1 already required opposing walls, claim 6 would add nothing. Corvidae also argues that the specification's criticism of crosstalk describes a problem, not a claim limitation, and that the figures are embodiments, not definitions.

How a court is likely to resolve it. Corvidae's claim differentiation argument is real and would probably defeat a construction requiring opposite walls. But claim differentiation is a presumption, not a rule, and it cannot overcome clear prosecution disclaimer. The statement distinguishing Tanaka's "monolithic assembly" is the problem. Having told the examiner that the claimed arrangement is not monolithic in order to secure allowance, Corvidae cannot now argue that a monolithic assembly with a two-millimeter gap satisfies the limitation.

A likely construction: "positioned apart from the light source such that the two are not part of a single integrated optical assembly." Under it, Halcyon almost certainly wins summary judgment of non-infringement, and the doctrine of equivalents is unavailable because Festo estoppel attaches to the amendment that added the phrase.

The lesson is not that Corvidae had a bad patent. It is that the case was decided in 2019, in an office action response, by a prosecutor who was trying to solve a different problem.

Which terms to fight over

Because courts limit the number of disputed terms, choosing them is a genuine strategic exercise. Some working principles:

Construe terms that are case-dispositive, not terms that are merely contested. The question is not "do we disagree about this word?" but "if the court adopts our construction, what happens?" A term where a win produces summary judgment is worth a slot. A term where a win produces a marginally better jury argument is not.

Understand that a construction can win the infringement case and lose the validity case. This is the fundamental tension of patent litigation. A patentee who obtains a broad construction may find that the same construction reads on prior art the defendant has been holding. Defendants exploit this deliberately by proposing constructions that are broad enough to be invalidating and then arguing invalidity if the patentee accepts. Every proposed construction should be stress-tested against the prior art before it is filed.

Do not ask for "plain and ordinary meaning" when you need a fight resolved. The Federal Circuit's decision in O2 Micro International Ltd. v. Beyond Innovation Technology Co., 521 F.3d 1351 (Fed. Cir. 2008), holds that a district court cannot discharge its duty by assigning "plain and ordinary meaning" when the parties have a genuine dispute about scope. Doing so improperly leaves a legal question to the jury. A party that genuinely wants no construction should say why the term is clear; a party that wants to preserve an argument for trial should be careful, because an unresolved dispute is reversible error and reversible error is a retrial nobody budgeted for.

Watch for terms that trigger § 112(f). If the accused infringer can convert a functional term into a means-plus-function limitation, the scope collapses to the disclosed structure — or the claim dies for lack of any.

Consider the preamble. Preambles are generally not limiting unless they give "life, meaning, and vitality" to the claim — but they are limiting when they supply antecedent basis for a later term, or when the applicant relied on them during prosecution to distinguish prior art. A preamble that reads "A method of treating diabetes, comprising . . ." may be the only thing separating the claim from prior art teaching the same chemical steps.

After the order

A claim construction order is not a final judgment and generally cannot be appealed on its own. That produces a familiar sequence.

The losing party looks for a path to judgment. The classic maneuver is a stipulated judgment of non-infringement: the patentee concedes that under the court's construction it cannot prove infringement, judgment enters for the defendant, and the patentee appeals the construction with a clean record. This is efficient and common. It requires care — a stipulation that concedes too much can waive arguments on appeal, and a stipulation that concedes too little may not create an appealable final judgment.

Alternatively, the parties may seek certification under 28 U.S.C. § 1292(b) for interlocutory appeal of a controlling question of law. The Federal Circuit accepts these sparingly. Historically it has been reluctant to take claim construction piecemeal, reasoning that the full record aids review.

Or the case proceeds. Expert reports are rewritten to conform to the constructions. Summary judgment motions follow. Damages theories are revised, because a narrower construction often means a smaller royalty base. In parallel proceedings at the Patent Trial and Appeal Board — inter partes review under 35 U.S.C. § 311 — the same terms are construed under the same Phillips standard, and the interaction between the two forums is governed in part by 35 U.S.C. § 315. A district court construction may be persuasive at the Board and vice versa, and inconsistent positions across forums are a gift to opposing counsel.

Reconsideration is possible but rare. Courts sometimes revisit constructions as the record develops, particularly where later evidence reveals that a construction is unworkable. But the practical window narrows quickly, and a party that sat on an argument will be told it was waived.

What claim construction cannot fix

It is worth being clear about the limits.

Claim construction cannot rescue a claim that is invalid under 35 U.S.C. § 101 for claiming an abstract idea — the framework from Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014) operates on the claim as construed but is not itself a construction exercise. It cannot cure a lack of novelty under § 102 or obviousness under § 103 where the prior art teaches the claimed combination on any reasonable reading. And it cannot supply a written description that is not there. Amgen Inc. v. Sanofi, 598 U.S. 594 (2023) is the recent reminder: a claim to an entire genus defined by function must be enabled across its full scope, and no construction makes an unenabled claim enforceable.

What claim construction can do is determine, with unusual finality for something that happens eighteen months into a case, who is going to win.

What good claim construction briefing looks like

Judges read a great many claim construction briefs, and the ones that persuade share recognizable habits.

They lead with the technology, not the argument. A brief that opens with two paragraphs explaining what the device does and why the disputed term matters is easier to follow than one that opens with a string cite to Phillips. The judge already knows Phillips.

They quote the intrinsic evidence at length, in context. The single most common error in claim construction briefing is the fragmentary quotation — five words from column 4 that appear to support a construction and that collapse when the surrounding sentence is read. Opposing counsel will restore the context, and the judge will remember which side did the trimming. Quote the full passage even when the full passage is imperfect, and explain why it still helps.

They engage claim differentiation honestly. If the dependent claims cut against your construction, say so and explain why the presumption is overcome here. A brief that ignores an obvious structural argument reads as either careless or evasive.

They tell the court what happens next. Judges are managing dockets. A construction section that ends with "under this construction, the accused product lacks the limitation and summary judgment will follow" is more useful than one that ends with a citation. Some judges dislike being told the consequences, on the theory that construction should be technology-driven rather than outcome-driven; know your judge. But most appreciate understanding what is at stake.

They propose constructions that read like constructions. A proposed construction should be a substitute definition that could be dropped into the claim and read grammatically. Constructions that are really arguments — "does not cover the accused product's arrangement" — get rejected, and rightly.

Recurring interpretive fights, and how they usually come out

A handful of patterns account for a large share of disputes. Knowing the base rates is useful even though every case is different.

"A" or "an" in an open claim. Presumptively means "one or more," not "exactly one," when used with the transitional word comprising. Overcoming that presumption requires clear intrinsic evidence — usually a specification that consistently describes a single instance and distinguishes prior art having several.

Numeric ranges and endpoints. "Between 5 and 10" and "from 5 to 10" are usually construed as inclusive of the endpoints. "At least about 5" imports the about problem, which under Nautilus requires the specification to supply enough guidance to make the boundary reasonably certain.

Order of steps in a method claim. Steps are not presumed to occur in the order written unless logic or the specification requires it. If step (c) uses a product created in step (b), sequence is required by grammar and logic. Otherwise the patentee usually keeps the flexibility — a point that matters enormously for divided-infringement arguments.

Functional language in an apparatus claim. "Configured to" is generally construed to require actual capability, not mere theoretical capability after modification. A device that could perform the function only after reprogramming usually does not infringe. But "adapted to" and "capable of" have drifted in meaning across cases, and the specification's usage tends to control.

The whereby and wherein clauses. A wherein clause that adds a substantive requirement is limiting. A whereby clause that merely states the result of previously recited steps usually is not. The line is fuzzy, and the drafter's intent, as revealed by the prosecution history, often decides it.

Trademark-like or coined terms. When an applicant invents a word, the specification is the only dictionary. These terms are either well-defined and enforceable or undefined and indefinite; there is rarely a middle ground.

Special contexts

Design patents. Design patent claims are pictures, not sentences, and the Federal Circuit generally discourages elaborate verbal constructions of them. Courts do construe designs to identify what is ornamental versus functional and to describe the scope of any broken-line disclaimers, but the Egyptian Goddess framework asks whether an ordinary observer would find the designs substantially the same, which is not a claim construction inquiry in the ordinary sense.

Chemical and pharmaceutical claims. Markush groups ("selected from the group consisting of A, B, and C"), product-by-process claims, and claims defined by physical parameters generate distinctive fights. Product-by-process claims are construed as products — the process steps do not limit infringement, though they may limit validity, an asymmetry that surprises people every time.

Standard-essential patents. Where a patent is asserted as essential to a technical standard, the standard document itself is sometimes offered as extrinsic evidence of what terms meant to the relevant technical community. Courts vary on how much weight to give it, and the patentee's own essentiality declarations to the standards body can function much like prosecution statements.

Claims construed in earlier litigation. A prior construction is not automatically binding on a new defendant, who was not a party and had no opportunity to litigate. But issue preclusion may bind the patentee, who did litigate and lose, and district courts give substantial weight to prior constructions of the same term in the interest of the uniformity Markman sought. Asserting a patent that has been construed narrowly is a decision to be made with eyes open.

Frequently asked questions

Is a Markman hearing required? No. Courts must construe disputed terms, but nothing requires a live hearing. Some judges decide on the papers. Others hold multi-day proceedings with expert testimony. Ask early, because the answer changes how you build the record.

When does claim construction happen? It varies enormously — anywhere from six months to two years into a case. Some judges construe early to promote settlement; others wait until summary judgment so the record is complete. Districts with local patent rules tend to be earlier and more predictable.

Can the court construe a term the parties did not dispute? Yes. Courts occasionally construe terms sua sponte, particularly when a party's argument depends on an unstated construction. This is a genuine risk of leaving a term unaddressed.

What standard applies at the PTAB? Since 2018, the Board applies the same Phillips standard used in district court, replacing the older "broadest reasonable interpretation" standard for most proceedings. That alignment makes cross-forum consistency both easier to achieve and more important to maintain.

Does the jury ever hear about claim construction? Yes — the constructions are given to the jury as instructions, and experts must testify consistently with them. What the jury does not hear is the argument about what the constructions should have been.

Can a party appeal a claim construction it agreed to? Generally no. Agreed constructions are binding, and stipulating to a construction to streamline a hearing forfeits the right to complain about it later. Agree only to constructions you can live with at trial and on appeal.

How much does claim construction cost? It is one of the most expensive discrete phases of a patent case. Between contentions, expert declarations, tutorials, briefing, and the hearing, seven-figure totals are ordinary in a case with several patents and many terms. This is a large part of why patent cases settle before the order issues.