How does one satellite invention turn into five patents? Usually through continuation applications. A continuation is a later patent application that pursues new or different claims directed to an invention already disclosed in an earlier, still-pending application by the same applicant — without adding any new technical content, and while keeping the earlier application's filing date. It is the mechanism that lets a company file once, then keep returning to the Patent Office to carve additional claim scope out of the same original disclosure, sometimes for years. For anyone counting patents to gauge a company's R&D, the continuation is the single biggest reason raw counts can mislead: a family of five patents may rest on one act of invention.

The Patent Office defines the term precisely in the Manual of Patent Examining Procedure.

"A continuation application is an application for the invention(s) disclosed in a prior-filed copending nonprovisional application, international application designating the United States, or international design application designating the United States. The disclosure presented in the continuation must not include any subject matter which would constitute new matter if submitted as an amendment to the parent application."— MPEP § 201.07, USPTO

Two phrases carry the weight. "Copending" means the continuation must be filed while the parent is still alive — pending before the Office, not abandoned or already issued. "Must not include any subject matter which would constitute new matter" means the continuation is bound to the original disclosure: it can re-slice the same technical content into different claims, but it cannot smuggle in a new improvement. The reward for staying inside those lines is priority: the continuation is entitled to the parent's earlier effective filing date, which sets the date against which prior art is measured. A continuation filed in 2026 off a 2021 parent is examined as though it were filed in 2021.

Continuation, divisional, continuation-in-part

Three members of the "continuing application" family are easy to confuse, and the distinctions are practical. A continuation, as above, pursues different claims to the same disclosed invention with no new matter. A divisional application also adds no new matter, but it arises from a restriction requirement: when an examiner determines that one application contains two or more independent and distinct inventions and requires the applicant to elect one, the non-elected inventions can be pursued in a divisional that claims that distinct subject matter. A continuation-in-part (CIP) is the outlier — it repeats substantial portions of the parent but adds new matter, so the new material gets only the CIP's later filing date while the carried-over material keeps the parent's date. In a spacecraft portfolio, you will see all three: continuations broadening or re-aiming claims on a known antenna or propulsion disclosure, divisionals splitting a multi-invention filing the examiner forced apart, and CIPs layering an improvement onto an earlier design.

The strategic uses are well understood. Applicants file continuations to keep a family pending so they can write new claims that read on a competitor's later product, to pursue broader claims after securing a narrow grant, or simply to keep options open while a market develops. Because a continuation can be filed any time the family remains copending, a single foundational disclosure can anchor a stream of grants spread across many years — each a separate patent number, each enforceable, all tracing to one priority date and one original specification.

Why this distorts filing-trend analysis

For portfolio and landscape work, continuations create a counting hazard. Filing velocity — patents or applications per assignee per year — is a standard proxy for research intensity, but it silently double-counts when a company runs heavy continuation practice. Five continuations off one parent register as five filings while representing one underlying invention. Two companies with identical genuine R&D output can show very different patent counts purely because one files aggressive continuation chains and the other does not. The honest correction is to analyze by patent family or by earliest priority date rather than by raw document count: collapse each continuation chain to its parent, and the picture shifts from "how many patents issued" to "how many distinct inventions were disclosed and when."

A related instrument is the terminal disclaimer, which often travels with continuation chains. Because every member of a family shares the parent's earliest filing date, the patents in a chain expire together at the twenty-year mark measured from that priority date — continuations do not extend the protection period beyond the original term. Where an examiner finds later claims not patentably distinct from earlier ones in the same family, the applicant may file a terminal disclaimer tying the later patent's expiration to the earlier one and agreeing they remain commonly owned. For an analyst, this is another reason to collapse a family to its priority date: a chain of five continuations is five enforceable patents that all lapse on the same day, covering scope carved from one disclosure, not five independent twenty-year monopolies. Counting them as five long-lived assets overstates both their number and their combined lifespan.

The signals also read differently. A burst of new first-filed applications in a CPC subgroup points to fresh research direction. A burst of continuations off older parents points to a company deepening claim coverage around inventions it already disclosed — defensive scope-building, or claims being shaped to track a rival's product, rather than new technical ground. Both are real strategy, but they mean different things. The discipline, when reading a spacecraft assignee's filing trend, is to ask of each filing whether it is a new disclosure or a continuation of an old one — because the MPEP's definition guarantees that the second kind adds claims, not inventions, and a count that ignores the difference reports motion where there may be none.