For Paper A, draft the claims first, then write a description in which the general teaching supports the full claimed scope and specific embodiments anchor the dependent claims and future amendments. Test your draft against two questions before finishing: can a skilled person carry out the invention across the claimed scope, and do all claims share one inventive concept? If either answer is no, fix the draft itself rather than reaching for exam technique.
Assembling the specification so claims and description stay one instrument
A specification is a single disclosure: title, technical field, background, summary, description of drawings, detailed description, claims, and abstract. Each part serves the claims; none is a standalone essay.
Give each component a job. The technical field tells the reader where the invention sits so the claims are interpreted in context. The background states the problem without conceding that the claimed solution was obvious or known. The summary states the invention in terms matching the independent claims. The detailed description teaches a skilled person how to make and use the invention across the claimed scope, and the drawings, where used, are described so that reference numerals and claim language agree.
Two structural errors undermine an otherwise sound draft, and each is easy to train against. First, a background section that describes the closest prior art in a way that reads the invention onto it, which makes your own claim look anticipated on the face of your document. Second, a summary that simply reprints claim 1 verbatim without framing the problem and the technical contribution, so the reader cannot see why the claimed combination matters. Draft the summary as a prose statement of the inventive concept, then check that its wording matches the claim rather than copies it.
Drafting the independent claim: elements, relationships, and antecedent basis
An independent claim should recite the minimum feature set that distinguishes the invention from the prior art given in the facts, expressed as structural or method relationships, with clean antecedent basis throughout.
Build the claim from the facts you are given: identify the prior art the question supplies, list the features that take the invention beyond it, and recite those features with the connections between them rather than outcomes. Prefer 'comprising' for open-ended coverage. Avoid result-oriented words such as 'fast' or 'efficient' unless the specification defines how such a result is achieved or measured, because an undefined result term invites a clarity objection and gives you nothing concrete to defend.
Watch antecedent basis: introduce an element with 'a' or 'an' on first mention and refer to it as 'the' or 'said' afterwards, so every 'the' in the claim points to a single, identifiable antecedent. A claim that says 'the housing' without ever introducing a housing is defective on its own terms. Then build the dependent claims as a ladder: each adds one feature from the description, ordered so that the narrower claims progressively capture the specific embodiments you described. This ladder is what you fall back on if the broad claim runs into trouble later.
Making the description support the claimed scope, not just one embodiment
Support means the description teaches the claimed scope as a scope; sufficiency means a skilled person can carry it out across that scope. One worked example does not support a broad genus on its own.
If a claim uses a general term, the description must show the general teaching: state the class of materials, conditions, or structures that work, explain why they work, and then give concrete examples within the class. This general teaching must be on the file before filing, because adding a new generalisation afterwards is treated as new matter. Before you finalise any description, audit each claim term and ask where the specification teaches its breadth, and describe reasonable alternatives and neighbouring ranges while you still can.
Worked scenario: you draft the claim 'a handle formed of a resilient polymer', but the description tells the reader only about one specific blend of two named polymers with a fixed ratio. The mistake is treating the single blend as adequate teaching for the genus. The better decision is to describe, at filing, the property that makes a polymer suitable here, list several workable polymer families, and present the named blend as a preferred embodiment. Why it matters: with only the blend disclosed, the claim is exposed to a challenge that the broader scope is neither supported nor enabling, and the post-filing amendment route is closed by the added-matter bar, so the claim may have to be cut back to the blend you never intended to be the whole invention.
Unity of invention and choosing between one application and a divisional
Claims share one inventive concept when they correspond technically and share the same special technical features. Where two inventions sit in one draft, plan a divisional rather than forcing them together.
Test unity by asking what technical problem each independent claim solves and what feature answers it. Claims to a product, the method of making that product, and an apparatus for that method commonly share the inventive concept when the same distinguishing feature does the work in each. Where that shared feature exists, keep the claims in one application. Where it does not, the honest move is to recognise two inventions at the drafting stage and decide which belongs in the parent.
Worked scenario: your fact pattern contains a novel valve and, almost as an afterthought, an unrelated coating process used in the client's factory. The mistake is filing both in one application to save effort, hoping the examiner will not object. The better decision is to draft the valve application properly and either leave the coating process out or prepare it as a divisional so it keeps a related filing framework, confirming current divisional timing rules and requirements directly with IPOS before relying on any deadline. Why it matters: a unity objection mid-prosecution forces a choice under time pressure, while a deliberate structure at filing preserves both inventions on your terms.
Drafting decisions that spend or preserve amendment room
Amendments generally cannot add new matter, so every drafting choice either preserves or spends your future flexibility. Write alternatives, options, and ranges into the description while you control the text.
Think of the filing draft as an ammunition store for later prosecution. Describe optional features, alternative embodiments, and useful ranges around any value you might recite, because once filed, the text is frozen as the source of anything you may later add. Claim wording matters just as much: closed wording, single point values, and definitional statements such as 'the invention is' each constrain how the scope can move later, and a phrase you cannot support cannot be argued into the claim.
Use the table below as a pre-filing checklist for the most consequential wording choices. Trace each row against your own draft: for every claim term, confirm the description gives you at least one narrowing path and that you have not locked the scope with wording you would want to reopen.
TABLE_PLACEHOLDER
| Drafting choice | Effect at filing | Consequence for later amendment |
|---|---|---|
| Broad genus term in the claim, species listed in the description | Wide initial scope with examples anchoring it | The species provide supported narrowing paths if the genus is challenged |
| 'Comprising' in claims | Open-ended: extra elements do not avoid the claim | Easy to narrow by importing described features; further broadening is generally not available |
| 'Consisting of' in claims | Closed set of elements, narrower scope from the start | The closed wording itself cannot be undone without new matter, so widening is effectively foreclosed |
| A single point value in the claim (one temperature, one ratio) | Precise but narrow scope | Neighbouring ranges written into the description keep fallback positions alive for narrowing or refinement |
| Optional features placed only in the description, not the claims | Claims stay clear and free of ambiguous optionality | Dependent claims or amendments can import the options later if prosecution requires it |
A reverse-engineering drill that trains spec-to-claim coherence
Pick a granted Singapore patent, read only its claims, then write the summary and one independent claim you would expect, and compare your version against the actual specification.
Run the drill in three passes. Pass one: open a granted patent through the IPOS Digital Hub or the patent register search, read the claims alone, and sketch the inventive concept and the broadest scope you think they cover. Pass two: write a five-sentence summary and a candidate independent claim from that sketch. Pass three: read the full specification and compare. This reverses the usual reading order and forces you to notice how professional drafts generalise a teaching and layer dependent claims.
Expected observations: professional independent claims usually recite fewer features than you predicted, and the general teaching sits in specific paragraphs you can locate. Self-check rubric: award yourself a point each if (1) your predicted broadest scope lands within one feature of the real claim 1; (2) your summary identifies the same technical problem as the specification's; (3) you can point to the paragraph that supports each claim term; and (4) you can find two dependent-claim features in the description that you had not predicted. Four out of four on two different patents is a strong learning milestone; three indicates you are close and shows you exactly which link in the chain you are still missing.
An adaptable preparation sequence and concrete readiness checks
Sequence preparation in four phases: claim anatomy drills, full specification reverse-engineering, timed drafting from a fresh fact pattern, then objection-response practice. Measure readiness by observable outputs, not hours logged.
Phase one, claim anatomy: take three granted patents and, for each, mark the antecedent basis, the 'comprising' boundaries, and the dependency ladder. Phase two, the reverse-engineering drill from the previous section, on two different technologies so you practise generalisation rather than memorising one domain. If your available weeks are short, compress phases one and two into one week by drilling a single patent per session; if you have longer, repeat each phase on new material rather than re-reading old notes.
Phase three: obtain or construct a plain-language fact pattern describing an invention with prior art context, and draft a complete specification against a self-set clock, claims first. Phase four: take your phase-three draft and write the narrowing amendment you would file if the independent claim were objected to, which immediately reveals whether your description contains the fallback material you need. Administrative matters such as exam registration, formats, and dates are published by IPOS; check the current official pages for those details rather than relying on secondary summaries.
Readiness checks you can score objectively: (1) you can draft an independent claim with correct antecedent basis in one pass; (2) for any claim term you write, you can name the paragraph that supports it; (3) your description contains at least two narrowing paths per independent claim; (4) you can test a claim set for a shared inventive concept and state the divisional consequence if it fails; and (5) you can produce a coherent abstract and summary in the final minutes of a timed draft. If any check fails, return to the phase that trains it rather than rereading the whole syllabus.
- Readiness check 1: one-pass independent claim with clean antecedent basis
- Readiness check 2: every claim term maps to a supporting paragraph
- Readiness check 3: at least two supported narrowing paths per independent claim
- Readiness check 4: unity tested, with a stated divisional plan if it fails
- Readiness check 5: abstract and summary produced under time pressure
References and further reading
Use these references to explore the concepts and check the latest information from the relevant organizations.