What must a §103 rejection contain to be proper?

Under Graham v. John Deere and KSR, the examiner has to resolve the scope and content of the prior art, the differences between the art and the claims, and the level of ordinary skill, and then articulate a reason with a rational underpinning why a person of ordinary skill would have combined or modified the references to reach the claim. A rejection that names two references and asserts the combination would have been obvious, without saying why, is incomplete, and saying so is a legitimate response.

Should I amend or argue first?

Argue first when the rejection has a defect — a citation that does not disclose the limitation, a missing motivation, a reference that teaches away, or a combination that would not work as claimed. Amend when the art genuinely reads on the claim and the specification supports a narrower one. Amending in response to a defective rejection gives up claim scope for nothing, and every amendment is prosecution history a litigator will read later.

Why search before amending?

Because the examiner searches again after an amendment, in the same field, and an amendment chosen only to distinguish the current references often lands on a limitation the next search finds at once. A scope search of the examiner's classification before amending shows which narrowing actually clears the field and which merely trades one rejection for another — and one round of prosecution costs more than the search.

Is an examiner interview worth requesting?

Almost always, and especially after a first §103 rejection. An interview lets you learn which limitation the examiner considers the real point of novelty and which reference they are least confident in, before committing to a written response. Go in with a proposed amendment and a verified reading of the references, and the interview shortens prosecution by a round.

What are my options after a final §103 rejection?

An after-final response with amendments the examiner can enter without a new search, a request for continued examination to reopen prosecution, or an appeal to the Patent Trial and Appeal Board. An appeal is the right answer when the rejection is legally defective and the record is clean; an RCE is the right answer when the claims need work. The verified reading of the references is what tells you which.

Guides

Responding to a §103 rejection: argue, amend, or search first?

Michael W. Smith, former USPTO patent examiner · September 13, 2026

Before arguing or amending, read the cited references at the pinpoint the examiner gave and check whether they say what the rejection says they say — in my experience a meaningful share of §103 rejections rest on a citation that does not disclose the limitation it is cited for, and that is the cheapest response there is. If the citations hold, decide whether the rejection articulates a reason to combine with a rational underpinning; if it does not, argue that. Amend only after a search of the examiner's own field, so the amendment narrows to something the next search will not find.

Step one: verify the citations

An examiner writes a §103 rejection under a production quota, from a search that took hours, not days, citing the passages that looked closest on a fast read. Some of those citations are exact. Some are approximately right. And a meaningful share, in my experience on both sides of the desk, cite a column and line that does not disclose the limitation it is cited for — the reference shows the element, but not in the arrangement claimed; or shows something similar under a different word; or the examiner relied on the abstract and the body says otherwise.

So the first act of any response is the one most often skipped: read every cited reference at the pinpoint given, and write down, limitation by limitation, whether it discloses what the rejection says. A rejection with a broken citation is answered by pointing at the reference. No amendment, no claim scope surrendered, no prosecution history created. It is the cheapest possible response and it is available surprisingly often.

Step two: check the combination

If the citations hold, the rejection still has to explain why a person of ordinary skill would have put the references together. Under KSR the reason can be flexible — a known technique applied to a known device, a predictable variation, a design need — but it has to be stated, and it has to be more than the claim itself read backwards. Look for:

A stated motivation. "It would have been obvious to combine" with no reason is a conclusion, not a finding.
Teaching away. A primary reference that discourages the modification, or describes the secondary reference's approach as a problem, is your best argument and it is often sitting in the reference the examiner cited.
Operability of the combination. If bolting the secondary reference's element onto the primary reference's device would stop the primary reference from doing what it is for, the combination lacks a reasonable expectation of success.
Hindsight. A reason to combine that only exists once you have read the applicant's claim is not one a person of ordinary skill had at the time.

Step three: if you must amend, search first

When the art genuinely reads on the claim, the response is an amendment — and this is where most prosecution money is wasted. The natural move is to add the limitation that distinguishes the cited references. The examiner then searches again, in the same classification, for that limitation, and often finds it in an afternoon, because it was never the point of novelty; it was only the thing the first two references happened to lack.

The fix is to search the examiner's own field before choosing the amendment. The examiner's search notes are in the file wrapper: the classifications and databases searched. A scope search of that field, against the two or three candidate amendments, shows which one actually clears it. Amend to that one, and the next action is an allowance rather than a new §103 with a new secondary reference. One avoided round of prosecution — the response, the RCE fee, the months — costs far more than the search.

Every amendment is also a statement to a future litigator. Narrowing around a specific reference can surrender equivalents you did not mean to give up. A search-guided amendment narrows once, to the limitation that is actually novel, instead of three times to whatever the examiner cited last.

Step four: the interview

Request one. Examiners are generally glad to have an applicant who has read the references carefully, and an interview is where you learn what a written action never says: which limitation the examiner regards as the real question, which reference they are least sure of, and what amendment they would allow. Go in with the verified reading and a proposed amendment, and the interview replaces a round of written response.

After final

After a final rejection the options narrow to an after-final amendment the examiner can enter without a new search, a request for continued examination, or an appeal. Appeal when the rejection is legally defective — a broken citation, no articulated motivation — and the record is clean, because the Board reverses those. File an RCE when the claims need substantive work. The verified reading of the references is what tells you which situation you are in, which is why it is the first thing I do and the cheapest thing I sell.

What this costs here

Published on the office-action page, flat:

Verification — $750$400. Every citation in the rejection checked against the reference. About a week.
Scope check — $750$400. A search of the examiner's field against your candidate amendments, before you commit to one.
Drafted response — $1,400$800. The response written for your signature. Amendments come only in the package, with the scope check.
All three — $1,750$1,000. Amendments included.