How Much a Continuation Changes
Read from 128 continuing applications and their direct parents, 126 of which could be compared. Not a random sample: the children and their named parents came from a list of granted patents. Method and limits: how we read these filings.
The typical continuation in this sample barely touches its parent's description. Across 126 comparisons the median change was three words added and four removed. A minority change a great deal, up to several thousand words, and that tail is where the interesting content sits. Both facts matter, and the tail is the part that gets written about.
A continuation reuses its parent's disclosure and asks for different claims. That is the doctrine. What it looks like in practice is a question of degree, and the degree turns out to be small.
The Typical Continuation Changes Almost Nothing
| Measure | Minimum | Median | Maximum |
|---|---|---|---|
| Words added | 0 | 3 | 5,394 |
| Words removed | 0 | 4 | 11,731 |
| Changed passages | 0 | 2 | 92 |
Three words. That is the middle of this distribution, and it is the most useful single fact on this page.
It is also the empirical face of something the doctrine already says. A continuation cannot add new matter relative to its parent, so the parent's disclosure is the ceiling. The practical consequence is that a continuation is usually not a rewrite. It is the same specification, filed again, pointed at different claims.
Two comparisons could not be made and are recorded as unobserved rather than as zero. Named ancestors were not substituted for true direct parents.
Where the Rest of This Page Comes From
The passage findings below do not come from the typical case. They come from the largest-change tail, selected deliberately, and the selection matters more than the findings.
| Stage | Count |
|---|---|
| Continuing applications in the frame | 128 |
| Direct-parent description comparisons available | 126 |
| Comparisons with any changed passage | 97 |
| Selected: the 30 with the most changed words | 30 |
| Read, after one selected comparison turned out to be deletion only | 29 |
So 29 readings, every one of them from the top of the change distribution. Nothing below is representative of a continuation in general, and the median above is the context it has to be read against.
What Changes, When Something Does
In those 29, the kinds of content appearing in the changed passages:
| Kind of content | Readings containing it |
|---|---|
| Alternative ways or forms | 25 of 29 |
| Manner of performing an operation | 25 of 29 |
| Component function or role | 22 of 29 |
| Figure content description | 21 of 29 |
| Structural composition | 21 of 29 |
| Enumerated examples of a category | 21 of 29 |
| Numeric value or constraint | 21 of 29 |
| Method step | 19 of 29 |
| Definition of a term or symbol | 18 of 29 |
The two most common are alternatives and how an operation is carried out. A passage from one of the readings shows the first of those:
"the lipoplex nanoparticles encapsulated Oh-MBs are able to bind or tether to a substrate through any physical or chemical interactions, or to be added to fuse with antibodies-captured EVs or cells"
"Bind or tether", "any physical or chemical interactions", "or to be added to fuse with". That is language describing a range of ways rather than one way. Whether the parent already supported it is exactly the question this reading did not answer, and it is the question that decides whether such a passage is permissible.
A kind absent from these changed passages is not established to be absent from the documents. Several kinds can occur in one reading, so the column does not sum.
The passage is from a published US application, quoted as filed. Documents are identified by the sample's own references rather than by applicant or firm. This sample spans several technology areas, unlike the software-only corpus behind the other pages here.
Three Categories, Not One
The frame is often described as 128 continuations. By the first link recorded in each chain it is:
- 124 continuations
- 2 continuations-in-part
- 2 divisionals
These are different instruments with different rules, and a continuation-in-part is the one that may add new matter. Two of each is far too few to say anything about either, which is precisely why they are named rather than folded into the total. A continuation-in-part appearing later in a longer chain is a separate matter again and is not counted here.
For what distinguishes the three, see continuation, divisional and continuation-in-part.
What This Does Not Decide
- It does not give a base rate. The frame was drawn from granted patents and is not random, and direct-parent statuses vary.
- It makes no claim-scope finding. Claim alignments exist for this sample but they were produced by two different readers in two waves, so they are reported separately in the source and are not combined into a single figure here.
- It does not say why anything changed. A changed passage can reflect many things, including corrections and formatting.
- It says nothing about the statutory requirements for a continuation. Those are in patent continuation applications, not in a count of words.
Educational, not legal advice. Continuation strategy and copendency deadlines belong with a registered patent practitioner.