The correct answer is "conducting original research to create new classified information." Derivative classification is a reproduction process: you take information that is already classified and incorporate it into a new document. Creating brand-new classified material from scratch is original classification, and it belongs to a different authorized person under a different set of rules. If you can hold that single boundary in your head, every variant of this exam question becomes easy, because every legitimate step is some form of "restate what someone already decided."

The exam phrasing matters here. Training tests built on the CDSE / ISOO derivative classification curriculum mix four real steps with one distractor, and the distractor is almost always the same idea wearing different words: "determine whether the information requires classification in the first place," "assign an original classification level based on your own judgment," or "decide if the information's unauthorized disclosure could reasonably be expected to cause damage." Each of those describes original classification. The steps that do belong share a single property β€” they all preserve a decision that has already been made by someone with original classification authority.

The Answer: Which Option Is the Exception Every Time

"Conducting original research" β€” or any phrasing that asks the derivative classifier to make a fresh classification judgment β€” is the exception. The standard derivative classification steps are: incorporate the source information, confirm the source is currently and properly classified, carry forward the classification level, carry forward the reason for classification and the declassification instructions, apply the appropriate classification markings, and follow safeguarding and dissemination controls. If an option asks you to create a classification decision, it is the wrong answer.

The confusion comes from the word "classification" appearing in both terms. Derivative classifiers make zero independent classification determinations. They are, in the language of Executive Order 13526 and its implementing directive 32 CFR Part 2001, applying classification guidance that already exists. When the exam says "assign the classification level," it means transcribe the level from the source, not choose it.

One failure case makes the distinction concrete. A contractor analyst drafting a capability brief pulls three paragraphs from a properly marked Secret//NOFORN report and one paragraph from an unclassified open-source article. She must mark the three paragraphs Secret//NOFORN, mark the fourth unclassified, and mark the overall document at the highest level present. If any single paragraph carried SCI caveats, the marking would carry those caveats too. She never decides that anything is Secret β€” she records that someone else already did.

The exam tests this because the two roles carry different legal exposure. Under 18 U.S.C. Β§ 798 and the Espionage Act provisions, mishandling derivatively classified material is prosecuted the same as mishandling originally classified material. Getting the role boundary wrong isn't a vocabulary slip; it's the difference between following guidance and inventing it.

Original Versus Derivative Classification: The Distinction That Decides the Test

Original classification is the initial determination that information requires protection in the interest of national security. Only a small set of officials β€” the President, agency heads, and those to whom they delegate in writing β€” hold original classification authority (OCA). Derivative classification is everything downstream: incorporating, paraphrasing, restating, or generating in new form information that is already classified, and marking the new material consistently with the source.

Feature Original classification Derivative classification
Who may perform it Designated OCA officials only Any cleared person trained and authorized to handle classified sources
Core act Decides whether and at what level Applies an existing decision
Authority document Original classification authority delegation Source document markings and a classification guide
Typical output A new classification decision, documented A new document carrying inherited markings
Training requirement OCA-specific Derivative classifier training, typically annual
Exam trap it generates "Determine whether information requires classification" (Correct answers) "Carry forward the level"

Read the third column carefully. Every legitimate derivative step is transitive β€” it moves a decision from a source to a new document without altering it. "Determine whether the information requires classification" belongs in column two because it is intransitive: the classifier would be originating the decision, not relaying it.

If you're studying for a test, the filter is simple. Ask whether the step adds a classification decision or carries one. Carry is derivative; add is original. That single question resolves nearly every option in this question family.

Every Legitimate Step, In Order, With the Authority Behind It

Here is the full sequence a derivative classifier actually works through. The order matters, because marking happens after review, and review happens after you identify the source.

1: Locate and review the source document or classification guide. Confirm the source exists, is legible, and bears markings you can read. A missing or illegible source is a stop condition, not a step to skip. 2: Confirm the source is currently and properly classified. A source marked Secret five years ago without a declassification date may already be declassified. Check the declassification instructions before you rely on the marking. 3: Incorporate the information into the new document, whether verbatim, paraphrased, restated, or regenerated in another form. 4: Carry forward the classification level exactly as the source states it. 5: Carry forward the reason for classification β€” the classification category code (such as 1.4(a) through 1.4(h)) β€” and the declassification instructions, including any exemption from automatic declassification under section 3.3. 6: Apply the classification markings to the portions, the document as a whole, the derivatively classified material itself, and the "Derived From" and "Declassify On" lines. 7: Apply safeguarding and dissemination controls such as NOFORN, ORCON, or FOUO controls where the source requires them. 8: Document the source on the record copy so a reviewer can trace each marking back to its origin.

Notice step 2. Almost no exam distractor includes it, yet it kills more real-world documents than any other failure. A source marked Secret in 2014 with "Declassify on: 20240101" was declassified before you incorporated it in 2026. Carrying the Secret marking forward would over-classify the new document β€” and over-classification is a violation too, not a safe default.

The "Derived From" line is the audit trail that makes the whole system work. A properly derived document says exactly which source (by document title, date, and classification guide identification number, or by the guide's specific paragraph) each marking came from. A reviewer who can't trace a marking back to a source will reject the document.

Where Does Original Classification Enter, and Who May Do It

Original classification enters at the front of the pipeline, before any derivative work exists. Someone with OCA, working from a formal classification guide, determines that specific information, if disclosed without authorization, could reasonably be expected to cause identifiable or describable damage to national security. That determination is the seed. Everything else the rest of us touch is the harvest.

The damage test is the defining feature. Original classification requires a decision under EO 13526 that the information fits one of eight categories (military plans, weapons capabilities, foreign government information, intelligence sources and methods, and so on) and that disclosure would causally produce damage in the specified amount for the level chosen. Derivative classification requires no such decision. The classifier applies the source, not the test.

Here's the edge case the exam doesn't usually test but real classifiers hit: a document is derived entirely from unclassified sources. It requires no classification markings at all β€” no "Derived From," no "Declassify On," no portion markings. A derivative classifier who feels pressure to "mark something" and forces a Secret marking onto unclassified compilation is over-classifying, which the directive prohibits. The instinct to add protection is not the same as the obligation to reproduce the source.

The Most Common Distractor: "Determine Whether the Information Requires Classification"

This is the phrase that reappears in this question family more than any other, and it is always wrong for derivative classification. The phrase "requires classification" is original-classification language: it asks whether a new decision is necessary. Derivative classification never asks that question. It asks what the source decided.

Variants you'll see and how to read them:

Distractor phrasing Why it's wrong for derivative Where it belongs
"Determine whether the information requires classification" Originates a decision Original classification
"Assign the level based on your own assessment of the damage" Substitutes your judgment for OCA's Original classification
"Decide if a reasonable expectation of damage exists" The damage test is OCA-only Original classification
"Conduct original research to generate the information" Creates new classified content Original classification
"Apply the source's markings to the new document" Carries an existing decision Derivative classification (correct)

The third row trips people up because the damage test appears in both original and derivative training. In derivative work, the damage reasoning is relayed, not performed: you reproduce the source's stated rationale, and you don't re-evaluate whether the source got it right. If you believe the source is misclassified, you flag it up the chain β€” you don't adjust the marking yourself.

One practical consequence: derivative classifiers are not supposed to downgrade or upgrade a source's markings. If the source says Secret and you believe it should be Confidential, the correct action is to leave it Secret and report the discrepancy. Downgrading on your own is a security violation even if you happen to be right.

Classification Guides, Source Documents, and the Paper Trail

The derivative classifier's authority comes from two documents: the classification guide (issued by an agency with OCA) and the source document itself. The guide tells you, at the paragraph level, what to mark and why; the source confirms the guide's application to this specific material. Most agencies maintain guides keyed to program or topic β€” the Department of Defense's Security Classification Guides and the DOE's classification guides are the canonical examples. When the guide and the source conflict, the guide typically governs, because it's the current authoritative statement.

The paper trail a trained classifier produces includes the originating agency, the source document's registry number, the guide's identification, the classification level, the category code, the declassification date or exemption, and any dissemination controls. A document with a "Derived From" line reading "Multiple Sources" is a red flag that the classifier cut corners β€” every source should be traceable.

The classification field of a source document is often abbreviated (e.g., "S//NF" for Secret//NOFORN), and the derivative document must expand the markings fully where the source does not. This detail β€” expansion to full wording β€” is a step training often skips, and it's the one that gets caught at classification review.

Markings, Declassification Schedules, and the 25-Year Rule

Every derivatively classified document carries a "Declassify On" line, and that line must come from the source. Under EO 13526, information is subject to automatic declassification 25 years after its original classification date unless the agency has invoked a section 3.3 exemption (such as information revealing intelligence sources and methods, or information that would impair national security interests). The source's declassification date governs; the derivative classifier doesn't reset the clock to the date the new document was created.

Marking line What it captures Where the value comes from
Portion markings (e.g., "(S)") Level of each paragraph The source paragraph's marking
Overall classification Highest level in the document The highest marking present
Classification authority The OCA who classified the source The source's "Classified By" line, restated
Derived From The source(s) relied upon The source document(s) or guide
Declassify On The date or exemption The source's declassification instruction
Dissemination controls NOFORN, ORCON, etc. The source's controls

Restating "Classified By" is easy to get wrong. The derivative document's "Derived From" and "Classified By" lines are different: the former names the source, the latter names the official who made the original decision. If the source's OCA line is missing, escalate rather than invent one.

Training, Refresher Cycles, and What Happens When the Rule Breaks

Derivative classifier training is required before any cleared person handles classified material in a derivative role, and refresher training runs annually in most agencies (the Intelligence Community's and DoD's derivative classifier courses are the standard vehicles). The training covers the EO 13526 framework, the derivative steps, marking mechanics, and the penalties for misclassification.

What happens when misclassification occurs? At the individual level, misclassification can trigger administrative action, loss of access, or referral for criminal investigation under 18 U.S.C. Β§ 798 and related provisions when the mishandling is willful. At the program level, agency inspectors general and the Information Security Oversight Office review classification decisions; over-classification findings are published and addressed through corrective action.

A rough failure case worth knowing: an analyst marks a document Secret//NOFORN based on a Secret//NOFORN source, but the source's declassification date had already passed. The document is technically over-classified. The fix isn't a quiet correction β€” it's a documented classification challenge, review by the OCA, and a reissued marking. That process can take weeks, which is why step 2 (confirm the source is currently classified) belongs before any marking is applied.

FAQ

What's the difference between original and derivative classification in one sentence?

Original classification is the initial decision that information needs protection, made only by a designated original classification authority; derivative classification is copying that decision forward into a new document. The test can be answered by asking whether the step adds a classification decision (original) or carries one (derivative). The distinction is defined in Executive Order 13526 and its implementing directive at 32 CFR Part 2001.

Can a derivative classifier say "I don't think this should be Secret" and mark it Confidential?

No. The derivative classifier reproduces the source's marking; adjusting a level downward is a downgrade, which requires the OCA's action, not the classifier's judgment. If you believe the source is misclassified, the proper move is to raise a classification challenge through your chain rather than mark it down. The only adjustment a derivative classifier may make is upward if a higher-level source is detected.

Is paraphrasing information considered a step in derivative classification?

Yes β€” paraphrasing and restating are explicitly derivative acts. The confusion is that the resulting text looks new, so people assume it needs a fresh classification decision. It doesn't; the classification travels with the idea, not the wording. A paraphrased paragraph that loses its portion marking is a marking error, not a new classification problem.

What is the "Derived From" line and can it say "Multiple Sources"?

"Derived From" identifies the source(s) relied upon for the document's markings. "Multiple Sources" is technically permissible only if every source is identified elsewhere in the document; standing alone, it's an incomplete audit trail and reviewers flag it. Best practice is to name each source by title and date, or by the classification guide's paragraph identifier.

Does derivative classification training expire?

Yes, typically after one year. Most agencies require annual derivative classifier training and retraining after any significant change to the classification framework. If your access has been lapsed for a period, retraining is required before you can resume handling classified material in a derivative role in most programs.

How does the "Declassify on" date work for a derivative document?

The derivative document inherits the source's declassification instruction; the classifier does not reset the 25-year clock from the new document's creation date. If the source's declassification date has already passed, the source may be declassified and should not be relied on. Check the source's date before applying the marking β€” it's the most common reason derivative documents get over-classified.

Is there an exception where a derivative classifier must conduct new research?

No. If research produces new information that isn't already in a classified source, the classification question is an original-classification issue that belongs to an OCA, not the derivative classifier. Derivative classifiers may research only to locate existing classified sources, not to generate new classified content. Mixing the two is the precise error this exam question is built to catch.