An inventor who has built one device naturally describes one invention. The application covers the machine, the way it is made, the way it is used, and the consumable it takes. To the inventor that is a single coherent thing. To an examining authority it can be four separate searches, and only one of them has been paid for.
That mismatch produces an objection that surprises applicants more than any other, because it is not an attack on patentability at all. It is an administrative demand to divide, and the way it is answered determines how many patents the project ends up with, what each of them costs, and how long any of them last.
Basis · Why the Split Happens
One Search Fee Buys One Search
The rule underneath the letter is economic before it is legal. Examination fees are set on the assumption that one application requires one search of one technical field. When an application in fact requires several — different classification areas, different databases, different bodies of prior art — the office either does unpaid work or asks the applicant to choose. Every major system chose the second option, and the letter that results is a restriction requirement.
Two different tests are used to justify it. The first asks whether the claimed inventions are independent and distinct: whether each could stand alone, be used separately, and would require a materially different search. The second, common where unity of invention is the governing concept, asks whether all the claims share a single general inventive concept — one common technical feature that is itself novel and contributes to the inventive step. The tests sound similar and diverge in practice: a set of claims can be entirely unitary under the second and still be split under the first.
The recurring split lines are predictable enough to design around. An apparatus and a method of making that apparatus are frequently treated as distinct. So are an apparatus and a method of using it. So are a combination and one of its subcombinations, where the subcombination has utility on its own. So are alternative chemical species claimed under a broad genus, and so are mechanically unrelated embodiments bolted into one specification because they were invented in the same month. A drone body, a battery-swap dock and a flight-planning routine described in one filing are three searches wearing one cover sheet, whatever the product roadmap says.
Response · Election
Electing, Traversing, and the Short Clock on Both
The letter identifies groups of claims and asks the applicant to elect one group for examination. The period is short — one to two months is typical where the letter raises nothing else, sometimes extendable on a fee — and the election is mandatory. Failing to reply does not preserve the claims; it abandons the application.
Two ways of replying exist and the difference matters. An election without traverse accepts the division and moves on. An election with traverse, or a formal disagreement filed alongside it, nominates a group for examination while preserving an argument that the split was wrong. The second costs almost nothing extra and keeps a review route open, because in many systems the correctness of a division can be challenged on request or reviewed later; an applicant who simply accepted it has usually given that up.
Non-elected claims are withdrawn from consideration rather than rejected, which is the crucial distinction. They have not been refused, and the subject matter is not lost — it survives in the specification and can be pursued in a separate filing. Some systems also provide for rejoinder: once the elected apparatus claim is allowed, method claims that were withdrawn but are fully covered by the allowed apparatus can be brought back into the same application without a new filing. That mechanism is easy to miss and worth asking about explicitly, because it converts a second application into a free amendment.
What should be elected is a commercial question dressed as a procedural one. The correct answer is almost never the broadest group; it is the group covering what a competitor would have to do to compete. A method-of-use claim that reads on the customer rather than the manufacturer is difficult to enforce, however elegant it looks, and electing it because it feels bigger is a common and expensive error.
Non-elected claims are not refused. They are parked — and parking has a fee, a deadline and a clock that never resets.
The one distinction to hold on toDivisionals · The Second Filing
What a Divisional Keeps and What It Costs
The parked subject matter is pursued through a divisional application: a new filing, carved from the same specification, claiming the material the parent was not examined on. Its defining feature is that it retains the parent's filing and priority dates. Nothing published in the intervening years becomes prior art against it, which is why the mechanism exists at all.
Its defining cost is that the term is measured from the same original date. A divisional filed in year four of a parent's life is not a fresh twenty-year right; it is a right with roughly sixteen years left to run, entering the examination process from the back of the line and typically taking two or three of those remaining years to grant. The later the carve-out, the worse that arithmetic gets.
Everything else about it is a duplicate. A divisional pays its own filing, search and examination fees, generates its own drafting and correspondence bills, and — once granted — carries its own annual renewal fees for the rest of the term. Across three territories, a family of one parent and three divisionals is twelve national files to prosecute and twelve renewal schedules to fund. Independent inventors who accept a split without modelling this discover the real number in year six, and it is routinely several times what the original filing cost.
Timing is unforgiving in a second way. A divisional must generally be filed while the parent is still pending — in most systems, on or before the day the parent grants, with some offices imposing a fixed window measured from the objection instead. Once the parent issues and no divisional is on file, the family is closed permanently. That deadline lands in the same brief window as the grant fee, which is why the notice of allowance is the most consequential piece of post-filing correspondence there is, a point made at length in the account of what follows a filing.
Hazard · Double Patenting
The Trap Waiting Inside the Second Application
Splitting an application creates a hazard that did not exist while everything sat in one file: two related patents whose claims are too close together. Where the divisional's claims are not patentably distinct from the parent's, a double-patenting objection follows, and the remedies differ sharply between systems. Some allow the applicant to file a formal undertaking tying the two patents together in ownership and expiry, curing the objection at the cost of permanently linking them. Others prohibit the same subject matter being claimed twice outright, leaving amendment as the only route.
There is an added irony worth planning for. In several systems an applicant who was ordered to divide is protected from a later double-patenting attack on the very ground the office itself imposed — but that protection usually depends on the divisional keeping to the boundaries the office drew. Redrafting the carved-out claims to creep back across the line can forfeit the shelter and reopen the objection. The discipline is to file the divisional along the fault line the examiner identified, not along the one the applicant would have preferred.
Claim-count fees compound the same problem. Most offices charge per claim above a threshold — commonly ten or fifteen total, three or four independent — with the surcharge repeating in each divisional. A ninety-claim specification split three ways can generate excess-claim fees in every one of the resulting files. Trimming to the claims that will actually be litigated or licensed is cheaper before the split than after.
Divide along the line the examiner drew. Divide along a better line and the objection you escaped comes back with interest.
On drafting the carved-out claimsPrevention · Drafting Ahead
Deciding the Split Before Anyone Else Does
A restriction requirement is far easier to prevent than to answer, and the prevention happens at drafting. Where all claims can be tied to one novel technical feature stated in the independent claim and echoed through the dependent ones, unity is straightforward to argue and the objection often never issues. Where a specification is assembled by stapling three loosely related developments together to save a filing fee, the split is close to guaranteed and the fee saving is deferred rather than made.
The practical sequence is to decide, before filing, which single claim the project actually needs, and to build the application around it — filing the peripheral material as separate applications on their own timetable if it merits protection at all. That forces a hard commercial judgement early, which is uncomfortable and considerably cheaper than making it in year four under a deadline. Inventor accounts of long product programmes tend to describe the same lesson from experience: the interview with an invention-services executive on how independent inventors are guided through the process and the reported development story behind the MixAid product both turn on narrowing to one commercially load-bearing idea rather than protecting everything at once.
Field matters as well. Applications in areas where a single product genuinely spans a device, a process and a material — energy systems and environmental technology are the usual examples, as this survey of independent inventors working on green technology illustrates — attract division more often, and their applicants should budget for at least one divisional from the outset rather than treating it as a surprise.
None of this makes the letter bad news. A division confirms that the specification contains more than one patentable idea, which is a better problem than the alternative. It also arrives with a schedule attached: the elected claims rejoin the ordinary flow of examination described in the piece on how long applications wait and what moves them, while the carved-out material starts that wait again from the beginning. Where the elected group then stalls against a stubborn objection, the next decision is whether to keep amending or to take the argument above the examiner.
Being told you have too many inventions is a good problem, billed as a bad one.