Skip to content

USPTO guide

How to respond to a 35 U.S.C. 112 rejection

By , Co-founder and CEO, PatentAssist.ai
Updated

Section 112 rejections are about your own document, not the prior art. Either the specification does not support what the claims now say, or the claims do not say it clearly enough. That makes them the most fixable rejections in an office action, and also the ones that can quietly cost you the most scope if you answer them badly. This guide covers each subsection the examiner may cite, what the examiner has to establish, and how to respond. For deadlines and the reply format, see the parent guide on responding to a USPTO office action.

Quick answer

  • 112(a) written description: show possession with paragraph-level citations, or narrow to what is described (MPEP 2163).
  • 112(a) enablement: the full scope must be enabled (Amgen v. Sanofi, 2023), judged on the Wands factors (MPEP 2164.01(a)).
  • 112(b): fix by amendment in most cases; the USPTO's threshold is lower than a court's (Packard during examination, Nautilus after grant).
  • 112(f): "module for" and similar placeholders can trigger it (Williamson); software needs a disclosed algorithm.
  • 112(d): a dependent claim must include every limitation of its parent and add one.

The five 112 rejections at a glance

35 U.S.C. 112 has six subsections. Office actions cite four of them, and 112(a) hides two distinct requirements.

Types of 35 U.S.C. 112 rejections
Rejection What it means Usual fix
Written description 112(a); MPEP 2161.01, 2163 The specification does not show the inventor had possession of what is now claimed. Point to the passages that describe each limitation, or narrow the claim to what is described.
Enablement 112(a); MPEP 2164 A skilled person could not make and use the full scope of the claim without undue experimentation. Argue the Wands factors with evidence, or narrow the claim to the enabled scope.
Indefiniteness 112(b); MPEP 2171-2173 The claim's boundaries are unclear: missing antecedent basis, a relative term without a standard, an example inside the claim. Usually an amendment. Sometimes an explanation from the specification of how a skilled reader understands the term.
Means-plus-function 112(f); MPEP 2181-2185 Not a rejection by itself: the examiner is construing a functional term as covering only the structure disclosed. A 112(b) or 112(a) rejection follows if that structure is missing. Recite structure, show the term names a structure, or identify the corresponding structure or algorithm in the specification.
Improper dependent claim 112(d); MPEP 608.01(n) A dependent claim omits a limitation of its parent or adds nothing to it. Rewrite it in independent form, correct the dependency, or cancel it.

112(a) written description

The specification must reasonably convey to a skilled person that the inventor had possession of the claimed invention as of the filing date (Ariad v. Eli Lilly, Fed. Cir. 2010, en banc, MPEP 2163). The description as filed is presumed adequate. The examiner has the initial burden of presenting, by a preponderance of evidence, why a skilled person would not recognize the claimed invention in the disclosure, and must make express findings of fact (MPEP 2163.04).

The rejection shows up in three situations. The first is an amendment that adds language not in the application as filed. That is new matter, barred by 35 U.S.C. 132(a), and claims containing it are rejected under 112(a) (MPEP 2163.06). The second is a genus claim, especially one defined by function, supported by a few examples. The specification needs a representative number of species or structural features common to the genus, and when the genus varies a lot, the examples must reflect that variety. The third is software claimed by function. MPEP 2161.01 says the algorithm or steps for performing the computer function must be described in enough detail that a skilled person would understand how the inventor intended the function to be performed. Restating the function is not enough.

The response is a table, not an essay. For each limitation the examiner questions, give the paragraph, figure or original claim that supports it. Exact words are not required; equivalent description is. If the support really is missing, the choice is to cancel the language, replace it with what is described, or file a continuation-in-part, which gives the new matter a later effective filing date.

Illustrative support table (hypothetical)
Claim 1 limitation Support in the application as filed
"a polymer layer 20 to 60 micrometers thick" [0031] ("between 20 and 60 µm"); Table 2, Examples 1-4 (22, 35, 48 and 60 µm)
"bonded without adhesive" [0040] (heat lamination, "no adhesive layer is required"); FIG. 5; original claim 7

112(a) enablement after Amgen

In Amgen Inc. v. Sanofi, 598 U.S. 594 (2023), the Supreme Court held that claims to a genus of antibodies defined by what they do (bind particular residues of PCSK9 and block it from binding a cholesterol receptor) were not enabled. The Court said a specification does not always have to describe how to make and use every embodiment, and a reasonable amount of experimentation is allowed. But "[i]f a patent claims an entire class of processes, machines, manufactures, or compositions of matter, the patent's specification must enable a person skilled in the art to make and use the entire class." The more you claim, the more you must enable (MPEP 2164.01).

The USPTO's enablement guidelines of January 10, 2024 (89 FR 1563) say examiners keep using the In re Wands factors, in every technology, and the Federal Circuit has said it sees no meaningful difference between Wands' "undue experimentation" and Amgen's "unreasonable experimentation" (Baxalta v. Genentech, 2023, quoted in MPEP 2164.01). The factors in MPEP 2164.01(a):

  1. The breadth of the claims
  2. The nature of the invention
  3. The state of the prior art
  4. The level of one of ordinary skill
  5. The level of predictability in the art
  6. The amount of direction provided by the inventor
  7. The existence of working examples
  8. The quantity of experimentation needed to make or use the invention based on the content of the disclosure

The examiner has the initial burden to establish a reasonable basis to question enablement, and must explain why it doubts any statement in the disclosure, backed by evidence or reasoning (MPEP 2164.04). A rejection that lists the Wands factors without findings on them is weak. Your answer should go factor by factor: what the specification teaches (direction, examples), what skilled people already knew, and why the experimentation needed to cover the whole claim is routine in this field. Where the claim is functional and the art unpredictable, as in Amgen, consider narrowing to the structures you actually made, or adding the "general quality" running through the class that the Court said can enable a genus from a few examples, if your specification describes one.

Illustrative argument: enablement
The Office action asserts that claim 1 is not enabled across the claimed thickness range but does not make findings on the Wands factors or explain why the disclosure's statements are doubted. MPEP 2164.04. The claimed range is bounded (20 to 60 µm), the specification gives the extrusion conditions for the full range ([0033]-[0036]) and four working examples spanning it (Table 2), and the declaration of Dr. R. Expert under 37 CFR 1.132 explains that setting die gap and line speed to reach a target film thickness was routine before the effective filing date. Any experimentation required is routine, not undue.

A Rule 132 declaration is often the strongest tool here. Once a prima facie case is made, the burden falls on you to show, with proofs where necessary, that a skilled person could make and use the invention using the disclosure as a guide (MPEP 716.09). The declaration can explain what was known; it cannot supply something the specification should have contained and skilled people did not know.

112(b) indefiniteness

After grant, a court asks whether the claim, read with the specification and prosecution history, informs skilled people of the scope of the invention "with reasonable certainty" (Nautilus v. Biosig, 572 U.S. 898 (2014)). The USPTO uses a different, lower threshold during examination. Under In re Packard (Fed. Cir. 2014), a claim is indefinite when it contains words or phrases whose meaning is unclear, after applying the broadest reasonable interpretation (MPEP 2173.02 and 2173.05(e)). The reason given in the MPEP is that applicants can amend during prosecution, and should. Breadth is not indefiniteness: a broad but clearly bounded claim passes 112(b) (MPEP 2173.04).

Common indefiniteness issues
Issue Example MPEP
Lack of antecedent basis "the lever" with no earlier "a lever", or two levers and then "said lever" 2173.05(e)
Relative terms "thin", "about", "substantially" with no standard in the specification for measuring them 2173.05(b)
Examples and preferences in a claim "a fiber such as glass fiber" where it is unclear whether glass fiber is required 2173.05(d)
Functional language without boundaries "a coating that prevents corrosion" with no structure or test that tells what does or does not 2173.05(g)
Unclear 112(f) structure "a control module configured to optimize flow" with no algorithm disclosed 2181, subsection II.B

Not every relative term is a problem. MPEP 2173.05(b) says terms of degree are acceptable if a skilled person would understand what is claimed in light of the specification, and "such as" in a claim does not by itself make it indefinite (2173.05(d)). If the specification gives a standard (a test, a range, a comparison), argue it. If it does not, amend.

Illustrative 112(b) amendment (hypothetical)

1. (Currently amended) A filter comprising:
a housing having an inlet and an outlet; and
a thin membrane having a thickness of 20 to 60 micrometers disposed between the inlet and the outlet,
wherein the a support grid holds the membrane , such as a stainless steel grid.

21. (New) The filter of claim 1, wherein the support grid is stainless steel.

Replaces a relative term with the range the specification gives, adds antecedent basis for the grid, and moves the example into a dependent claim where it limits clearly.

112(f) means-plus-function after Williamson

Under 112(f), an element expressed as a means or step for performing a function is construed to cover only the corresponding structure, material or acts described in the specification and their equivalents. Since Williamson v. Citrix Online (Fed. Cir. 2015, en banc), the absence of the word "means" only creates a rebuttable presumption that 112(f) does not apply. The presumption is overcome when the term fails to recite sufficiently definite structure, or recites function without enough structure to perform it.

MPEP 2181 applies a three-prong test. 112(f) applies if (A) the limitation uses "means", "step" or a generic placeholder; (B) the placeholder is modified by functional language, often linked by "for", "configured to" or "so that"; and (C) it is not modified by sufficient structure, material or acts to perform the function. The MPEP lists placeholders that may invoke it: "mechanism for", "module for", "device for", "unit for", "component for", "element for", "member for", "apparatus for", "machine for" and "system for". The examiner should state expressly in the action when a limitation is being interpreted under 112(f).

A 112(f) interpretation is not itself a rejection, but it has consequences. For a computer-implemented function, the specification must disclose an algorithm for performing it, or the claim is indefinite under 112(b) (MPEP 2181, subsection II.B). The corresponding structure is the special purpose computer programmed with that algorithm, not a general purpose processor. The same gap usually supports a written description rejection as well (MPEP 2161.01 and 2185).

You have three responses. Argue the term is understood by skilled people as the name for a structure (a "detent", a "baffle", a "compiler"), citing evidence. Amend to recite the structure or the algorithm steps. Or accept the interpretation and identify, in the reply, the structure or algorithm in the specification that corresponds to each function. Accepting it narrows the claim to that structure and its equivalents, which may be acceptable if the disclosed algorithm is the invention.

Illustrative 112(f) amendment (hypothetical)

10. (Currently amended) A system comprising:
a flow sensor; and
a control module configured to optimize pump speed based on the sensor output a processor and memory storing instructions that, when executed by the processor, cause the processor to: compute a moving average of the flow sensor output over a window of N samples; compare the moving average with a setpoint; and change a pump drive frequency in steps of no more than 1 Hz per second until the moving average is within a tolerance band of the setpoint.

The steps must come from the specification. Reciting a processor executing specific steps supplies the structure prong (C) looks for.

112(d) dependent claims

A dependent claim must refer to an earlier claim and specify a further limitation, and it incorporates every limitation of the claim it refers to. The test in MPEP 608.01(n), subsection III is whether the claim includes every limitation of its parent and adds one more. A claim that drops or replaces an element of claim 1 is not a proper dependent claim, however narrow it is otherwise. Since Pfizer v. Ranbaxy (Fed. Cir. 2006), examiners reject such claims under 112(d) rather than merely objecting to them.

Two nuances help. A dependent claim may be in a different statutory class: a method of making the product of claim 1 in a particular way can properly depend from it. But a product claim that depends from a method-of-making claim is proper only if the product cannot be made another way. A dependent claim also does not fail 112(d) just because the significance of its added limitation is questionable. The fix is mechanical: rewrite the claim in independent form with all the elements it needs, correct the claim it refers to, or cancel it.

How to read the rejection, and argue or amend

  1. Identify the subsection. "112(a)" alone is ambiguous: is it written description, enablement, or both? Answer each separately.
  2. Find the exact words. The examiner should identify the term or limitation at issue. If the action is vague, ask in an interview which words are the problem.
  3. Check the interpretation. Has the examiner applied 112(f) silently, or read a term more broadly than the specification supports?
  4. Weigh the cost of amending. For 112(b) and 112(d), amending is usually cheap. For 112(a), amending means narrowing, so first check whether the support is already there.
  5. Coordinate with the art. A 112 fix that narrows the claim also changes the 102 and 103 picture. See the guides to 102 and 103 rejections, and for eligibility, 101.

Common mistakes

  • Citing the specification in bulk. "Support can be found throughout the specification" does not answer a written description rejection. Give paragraph numbers for each limitation.
  • Adding claim language from memory of the invention rather than from the text. If the words or their equivalent are not in the application as filed, the amendment is new matter (35 U.S.C. 132(a), MPEP 2163.06) and draws a fresh 112(a) rejection.
  • Arguing that a broad claim is fine because breadth is not indefiniteness, when the examiner's real point is enablement or written description. Breadth is not a 112(b) problem (MPEP 2173.04), but it can be a 112(a) one.
  • Writing "module configured to" or "unit for" throughout software claims without disclosing an algorithm. If 112(f) applies, the claim is limited to the disclosed algorithm, or indefinite if there is none.
  • Treating a 112(b) rejection as a reason to argue when a two-word amendment would do. Indefiniteness rejections are the cheapest to fix by amendment, and the cleaner record helps in later litigation.
  • Fixing a 112(d) problem by deleting the dependent claim's limitation instead of correcting the dependency. A dependent claim must include every limitation of the claim it refers to.

Most 112 rejections are drafting problems that surface in examination. Clear antecedent basis, numbers instead of adjectives, and an algorithm behind every software function avoid them. See how to write patent claims.

Where PatentAssist helps

PatentAssist.ai does not prepare USPTO office action responses. Its office action feature supports Indian First Examination Reports only, in beta. Where it helps with 112 is before filing: it drafts USPTO-format claims and specifications from your disclosure, with the claim terms described in the specification, so written description, antecedent basis and fallback positions are in place before an examiner looks. Its free prior-art search across USPTO, EPO and Indian records helps you decide how broad a genus you can realistically claim.

More USPTO guides

All guides, by patent office

35 U.S.C. 112 rejections: frequently asked questions

What is a 35 U.S.C. 112 rejection?

It is a rejection based on the disclosure or the form of the claims rather than the prior art. 112(a) requires a written description of the invention and an enabling disclosure; 112(b) requires claims that particularly point out and distinctly claim the invention; 112(d) sets the rules for dependent claims; and 112(f) governs claims written as a means or step for performing a function.

What is the difference between written description and enablement?

Written description asks whether the specification shows the inventor actually possessed what is claimed (Ariad v. Eli Lilly, MPEP 2163). Enablement asks whether a skilled person could make and use the full scope of the claim without undue experimentation (MPEP 2164). A claim can fail one and pass the other, so they are separate rejections with separate answers.

How did Amgen v. Sanofi change enablement?

In 2023 the Supreme Court held that Amgen's functionally defined antibody genus claims were not enabled. If a patent claims an entire class, the specification must enable a skilled person to make and use the entire class, though a reasonable amount of experimentation is allowed. The USPTO's January 10, 2024 guidelines say the Wands factors still apply, in every technology, and MPEP 2164 now reflects that.

How do I overcome an indefiniteness rejection under 112(b)?

Most are fixed by amendment: add antecedent basis, replace a relative term with a number or a standard from the specification, or move examples out of the claim. If the term is clear to a skilled person, explain why with citations to the specification or evidence. During examination the USPTO applies the Packard standard, under which a claim is indefinite if it contains words or phrases whose meaning is unclear (MPEP 2173.05(e)).

When does 112(f) apply to a claim without the word "means"?

When the term is a generic placeholder such as "module for", "unit for" or "mechanism for", modified by functional language, and not modified by enough structure to perform the function. That is the three-prong test in MPEP 2181, based on Williamson v. Citrix (Fed. Cir. 2015, en banc). Without "means" there is a rebuttable presumption that 112(f) does not apply, which is overcome when the term does not recite sufficiently definite structure.

What is a 112(d) rejection?

A rejection of a dependent claim that does not further limit the claim it refers to or does not include all of that claim's limitations. MPEP 608.01(n) says examiners should reject such claims under 112(d) rather than object to them. The fix is usually to rewrite the claim in independent form, correct the dependency, or cancel it.

Can a declaration help with a 112 rejection?

For enablement, yes: a Rule 132 declaration can show what a skilled person knew at the filing date and that the experimentation needed was routine (MPEP 716.09). It cannot add to the specification. Information read into the specification must have been known to skilled people at the time. For written description and indefiniteness, evidence on how skilled readers understand a term can help, but an amendment is usually faster.

Sources: 35 U.S.C. 112 (full text of each subsection in our annotated section 112 page); 37 CFR 1.132; MPEP 608.01(n), 716.09, 2161, 2163, 2164, 2171, 2173, 2181 and 2185 (Ninth Edition, Revision 01.2024); USPTO memo of January 10, 2024 announcing the enablement guidelines, 89 FR 1563. Checked October 26, 2026. This guide is general information, not legal advice. Have a registered patent attorney or agent review your reply.