What Reply Remarks Contain, by the Section Cited

The companion page, what a response to an office action contains, reports what each reply did about each ground: amended, argued, both, or cancelled the claims. That measurement is per ground. This page asks a different question, which kinds of content the remarks carry, and here the counts are per reply. The reading coded kinds across each reply as a whole, and this page counts those found in its remarks. It also quoted each ground's response, but the kind counts here are per reply bundle, so none of them ties a kind to a ground.

A reply bundle is everything filed in answer to one action. It can hold more than one submission, such as a request for continued examination with its amendment.

Why the Sections Cannot Be Pulled Apart

An action often prints more than one statutory section, and one reply answers all of it. Counting the four sections this page groups by (101, 102, 103 and 112):

The family column does not sum to 22, because one family can have several replies in different rows.

The overlap between particular sections is large. Each row below is one group; each cell says how many of that group's bundles also answered an action printing the other section.

So 16 of the 25 replies in the section 101 group (11 of its 13 families) are also in the section 103 group. Whatever their remarks contain is counted twice, once under each section. A difference in counts between two groups can only come from the replies they do not share; a difference in shares can also come from the groups being different sizes.

What Appears Whatever the Section

Four kinds rank in the five most frequent for every group on this page. Each row gives the bundles and families with that kind, against the group's own totals.

This is the backbone of a reply's remarks in this sample: say which claims stand where, restate how the examiner treated them, report a position, and ask for an official step. It varies little with the section printed, from 88% to 100% of replies in every group, which is itself the main finding. Reported claim treatment is by definition a report of what the examiner did, so a kind's presence in the remarks does not mean the applicant asserted it.

One bundle, in the section 101 and section 103 groups, had no available remarks text. It carries no kinds and stays in those denominators, as it does in the source.

Where the Groups Differ

The table lists every kind that appears in at least five bundles and meets one of two rules for a section: it never appears in a bundle whose action did not print that section, or its share of bundles is at least 50 points higher with the section than without it. "Without" means the other bundles among the 47.

Bundle denominators: section 101, 25 with and 22 without; section 103, 33 with and 14 without. Family denominators: 13 with and 11 without for section 101; 18 with and 8 without for section 103. A family can sit on both sides, because its replies can answer different actions.

Information storage appears in seven bundles, and the action behind every one of them printed both section 101 and section 103, so it meets the first rule for both and cannot be assigned to either.

No kind met either rule for section 102 or section 112(b). Their widest gaps, checked across all 74 kinds, were 41 points for section 102 and 28 points for section 112(b). Section 102 has 14 bundles, 9 of which also answered section 103; at that size a single reply moves a share by about 7 points.

Read as descriptions, the section 101 rows are about characterizing what is claimed, the reasons given, examples from decisions, and the claimed system's components and stored information. The section 103 rows are about cited references and claims: which references, what they omit or do not explicitly teach, how they are used, why they would be combined, and reasoning carried from one claim to another. It is tempting to read these as the arguments each section calls for. The measurement cannot carry that reading. It records only that the passages were in the same remarks.

A Check: Section 103 on Its Own

One way to loosen the overlap is to look only at replies to actions that printed one of the four sections. 10 bundles, in 6 families, answered an action that printed section 103 and none of sections 101, 102 or 112. Four of those ten also carried grounds with no printed basis or a CFR citation.

Two of the six section 103 rows hold up: assertions that cited material omits or does not explicitly teach something, and reuse of reasoning across claims, appear in nearly every one of these replies. The other four appeared in at most two of the ten. Most of their presence in the section 103 group came from replies whose action also printed another section, so the section 103 group alone does not account for them.

The same check cannot be run for section 101. Only 4 bundles, in 3 families, answered an action that printed section 101 and none of the other three, below this page's minimum of five.

Every Kind, Every Group

Selecting the rows that differ is how a story gets built by accident, so the full composition follows. All 74 kinds of the frozen taxonomy are listed, including the 9 that appear in no reply's remarks. Each cell gives bundles, then families, against the group's own totals.

What Was Left Out, and Why

• Groups below five. A group is shown only when both its bundle and family denominators are at least five. That removes section 112(a) (5 bundles, 3 families, and the action behind every one of the 5 also printed section 101), section 112(d) (1 bundle, 1 family), printed CFR citations (3 bundles, 2 families) and printed MPEP citations (1 bundle, 1 family).
• Section 112 is shown as 112(b). 17 of the 20 bundles whose action printed any subsection of section 112 printed 112(b). The other three printed 112(a) only, which is below the minimum above.
• Grounds with no printed basis. 16 bundles, in 12 families, answered grounds where the reader recorded no statutory citation. Every one of those 16 bundles also answered a ground printing section 101, 102, 103 or 112, so each is counted in at least one section group as well as in the "All replies" column.
• Kinds in fewer than five bundles. Nine kinds meet the first rule for at least one section but appear in fewer than five bundles in all, as few as one. They are in the full table and not in the differences table.

What This Does Not Decide

• It does not recommend content for a reply to any section. It describes what 47 replies did contain.
• It does not describe the statutory sections themselves. They are named here only as the examiner's action printed them.
• Kinds were assigned by a single model reader. The source records that its checks bind each coded passage to the filed text; they do not establish that the reader's meaning is right. The kind list was built from other families of the same population and applied by the same reader.
• A kind counts if any passage of the remarks carried it, including passages coded as negative or optional. It is presence, not volume.
• Only remarks text that could be retrieved was read. Text that could not be obtained is unobserved, not absent.
• It is 47 replies from 22 families. It is not a survey of practice, and no firm or practitioner is compared with another.

Educational, not legal advice. What a reply argues is the practitioner's judgment on the specific claims and record, with deadlines attached.

If You Are Reading an Office Action Now

This page describes the content of remarks across a sample. For what each reply did about each ground, see what a response to an office action contains. For the provisions themselves and the procedure around them, our practitioner-facing site covers office actions in depth at patent-actions.com.