The hardest ground to answer is also the one written in the fewest words. An examiner cites reference A for most of your claim, reference B for the feature A lacks, and then supplies a single sentence explaining why a skilled person would have put them together. Applicants read that paragraph, see two documents they cannot deny exist, and start amending. In doing so they concede the only point that was ever in dispute.
Nothing about the documents is really being contested. Both exist, both say what they say. What is being asserted — and asserted rather than proved — is the reason for the combination. That assertion is where the argument lives, where the record is thinnest, and where a well-built reply routinely wins without surrendering a single word of claim scope.
The Burden · 01
What the Office Actually Has to Establish
Strip away local vocabulary and every examining authority is doing the same four things when it raises this ground. It must identify what the prior art discloses. It must identify how the claim differs from it. It must define the notional skilled person and the ordinary knowledge available to them at the relevant date. And it must articulate why that person, knowing only what was known then, would have arrived at the claimed combination.
In some jurisdictions this is formalised as a structured analysis: pick the closest single document, define the objective problem that the claimed difference solves, then ask not whether the skilled person could have made the modification but whether they would have. The distinction is not academic. Almost any two documents in the same field could be combined by someone with a workshop and an afternoon. The question is whether anything in the art gave them a reason to.
So the first move in answering an obviousness rejection is inventory. Write down the four elements above and check which of them the report actually contains. In a fast first action it is common to find the reason for combining reduced to a phrase — "to improve efficiency", "as a matter of design choice", "since both relate to fastening" — with no citation supporting it. A conclusion with no evidence behind it is not something you have to out-argue. It is something you point at.
The Weak Point · 02
Attack the Reason, Not the References
Four lines of attack are available against the combination itself, and they can be run together.
No motivation in the art. The stated reason appears nowhere in either document. If the benefit relied upon is not described, suggested or implied by anything the skilled person had in front of them, the reason has been supplied by the examiner rather than found in the record.
No reasonable expectation of success. Even where a person might have tried the combination, the art must give some basis for expecting it to work. In chemistry, biology and materials this is often the strongest line available.
The combination breaks something. If importing B's feature into A defeats A's stated purpose, or requires redesigning the very structure A relies on, the skilled person would not have done it. Documents that expressly warn against the approach are said to teach away, and a single sentence of that kind can end the ground.
The feature is missing from both. Combination reasoning sometimes conceals a genuine gap: one limitation is not disclosed by either document. Verify this before anything else, because it disposes of the rejection without argument.
Each line has to be tied to text. Quote the passage in A that describes its purpose, quote the passage in B that warns of the incompatibility, and let the two quotations do the work. This is also why the reading discipline matters: half of these arguments only become visible once you have read fifteen lines either side of every cited passage, which is a job best done systematically against the structure of the documents themselves before drafting a word of response.
Nobody disputes that two references can be combined. The question is whether anything in the art, at the time, gave a reason to combine them. Answer that and you have answered the rejection.
The only sentence that matters
Hindsight · 03
The Skilled Person Is a Fiction With Rules
The notional skilled person is deemed to know the whole of the art in the field, to have ordinary competence, and to have no imagination whatsoever. That last quality is the one applicants forget to invoke. This person does not experiment speculatively, does not pursue unpromising leads, and above all does not know how the story ends.
Hindsight bias is structural, not a failing of any individual examiner. Once you have read the claim, the path to it looks short — the two references seem to point at each other, the missing feature seems inevitable, the whole thing seems like tidying. The corrective is to reconstruct the state of the art without the claim in hand. If dozens of practitioners worked in this space for years using A's approach and none proposed B's feature, that silence is evidence. If the combination is so natural, the useful question in a reply is simply why nobody made it.
Number of references is a related signal. A ground built on two documents is ordinary. A ground built on four or five, each supplying one limitation, is often a sign that the invention is not in the art at all and is being assembled from fragments using the claim as the assembly instruction. Saying so plainly, and counting the documents, is a legitimate submission.
Evidence · 04
Facts Outrank Argument
Argument persuades; evidence binds. Most systems allow an applicant to file a sworn statement with supporting data, and on this ground it is often the decisive filing. Four categories carry weight.
Unexpected results. The claimed combination performs in a way the art would not have predicted — not merely better, but differently or disproportionately. This needs numbers: a side-by-side comparison against the closest document's arrangement, tested under the same conditions, with enough samples that the difference is not noise. A test of three units proves little. A test of thirty, reported with method and variance, is hard to dismiss.
Long-felt need and failure of others. If the problem was recognised in the literature for years and the documented attempts to solve it went elsewhere, the combination was evidently not obvious to the people actually trying.
Commercial performance. Useful only with a demonstrated link between the sales and the claimed feature rather than the packaging or the advertising spend. Consumer products that succeed on a specific mechanism can make that link cleanly — accounts of a kitchen tool that reached retail, including this account of a household mixing invention brought to market and a fuller description of the inventors behind it, describe exactly the kind of feature-driven adoption that supports the argument.
Copying and licensing. Competitors adopting the specific claimed arrangement shortly after disclosure is evidence about what the field considered valuable and non-obvious.
Trade-offs · 05
When to Amend Instead
Sometimes the ground is sound. If the reason for combining is genuinely supported by a passage in one of the documents, arguing costs a response round — commonly four to eight months of calendar and a fee — and ends where it started. The judgement call is whether the reasoning has a hole in it or merely an inconvenience.
When amendment is the right answer, amend once and amend deliberately. Add the limitation that captures what is actually inventive, not the first difference that clears the reference. Every added word narrows the monopoly for its full twenty-year term, and the remarks explaining the amendment become a permanent statement about scope that will be read back in any later dispute. Applicants who amend twice in a row usually end up with a claim that protects a prototype rather than an invention.
Two habits improve the odds regardless. Request an interview: twenty minutes of conversation often reveals that the examiner has read one claim term differently, and agreement reached there still has to be recorded in writing but arrives a round earlier. And keep the whole procedural picture in view — the deadlines, the four grounds and their different answers, all of which sit in the overview of how a first rejection works. Broader orientation on the route from idea to filed application, including where this ground usually appears, is set out in a general guide to turning an idea into a finished product.
Every limitation you add to escape an obviousness rejection is a limitation a competitor gets to design around for the next twenty years. Argue first. Amend once.
The cost nobody prices in
Find the sentence that supplies the reason. Ask what supports it. Reconstruct what was known before the claim existed, gather the numbers if you have them, and answer in text rather than adjectives. The documents are not the case. The sentence is.
End of report