SDS Section 2: Hazards identification

The verdict, and everything that follows from it

Section 2 does not describe the product. It passes judgment on it, and then prints the consequences. Subsection 2.1 states which hazard classes and categories the product falls into; 2.2 converts that judgment into the exact symbols and wording that must appear on the container; 2.3 records what the classification machinery could not capture. It is the only place on the sheet where the hazards are stated as a decision rather than as data.

The order is load-bearing, and it runs one way only. “Based on the classification, provide the appropriate label elements”, says the UN text; “Based on the classification, at least the following elements appearing on the label”, says REACH Annex II. Nothing in 2.2 is a choice. A pictogram that no class in 2.1 produces, or a class in 2.1 with no matching element in 2.2, means the two halves were written at different times and one is out of date.

2.1 — the classification, and what it has to agree with

A class paired with a category, never an adjective

A classification is not “harmful” or “corrosive”. It is a hazard class paired with a category, and the pairing is what selects the label. Annex 4 works the examples itself: “flammable liquid Category 1”, “skin corrosive, Category 1A”. The class names the kind of harm; the category is the band the evidence places the product in. The hazard classes and categories are the fixed vocabulary — a sheet that invents its own is not classifying anything.

Where a class is differentiated and the differentiation produces different hazard statements, that differentiation has to survive into section 2. Acute toxicity is the standard case: Annex 4 spells out acute oral, acute dermal and acute inhalation toxicity Category 1 as three separate statements, not one, and where more than one category applies within a differentiated class, all of them are communicated. “Acute toxicity Category 1” on its own does not say by which route — and therefore does not determine which H statement belongs on the label.

The sheet has to agree with what the supplier told the authorities

This requirement is peculiar to the EU and has the sharpest teeth. Where a supplier notified a substance to the classification and labelling inventory under Article 40 of CLP, or supplied that information in a REACH registration, subsection 2.1 requires the classification on the sheet to be the same as the one notified or registered.

The consequence is that section 2 is checkable from outside the company: one version of the classification sits in a public database, the other in a customer’s hand, and where they diverge nobody need argue about the science to establish that something is wrong.

REACH also demands the negative case be stated. If a mixture does not meet the criteria for classification, that must be said clearly — a blank subsection reads identically to an unfinished one.

The paragraph aimed at somebody else

The last requirement of 2.1 is easy to miss because it does not look like a rule: the most important adverse physical, human health and environmental effects must be listed in accordance with sections 9 to 12, in such a way as to allow non-experts to identify the hazards.

Two things follow. The content is drawn from later sections — physical and chemical properties, toxicological information, ecological information — so it summarises evidence that appears further down rather than making an independent claim. And it carries a readability standard, which almost nothing else in Annex II does: a paragraph written in classification shorthand satisfies the cross-reference and fails the audience test.

2.2 — label elements are derived, never chosen

Subsection 2.2 asks as a minimum for hazard pictogram(s), signal word(s), hazard statement(s) and precautionary statement(s) as they appear on the label. It then adds a requirement that is frequently skipped: the applicable label elements under Article 25(1) to (6) and Article 32(6) of CLP. Four elements are the floor, not the ceiling.

Both texts make a printing concession here. REACH allows a graphical reproduction of the full pictogram in black and white, or of the symbol only, in place of the colour pictogram; Annex 4 goes further and permits the name of the symbol — “flame”, “skull and crossbones”. The concession exists because sheets are photocopied and faxed. It also means a monochrome symbol carries full weight, and says nothing about what the physical container label looks like.

The governing phrase is appearing on the label. Section 2 mirrors a real label, so the precautionary statements listed here are the ones actually selected for that container — not the whole menu the classification makes available. A section 2 carrying forty P-statements mirrors nothing; it transcribes a lookup table.

2.3 — the subsection that is not about classification

PBT, vPvB and endocrine disruption

Since Regulation (EU) 2020/878, 2.3 carries a declaration that has nothing to do with the label: PBT and vPvB status against Annex XIII, inclusion in the Article 59(1) list for endocrine-disrupting properties, and identification as an endocrine disruptor under Regulation (EU) 2017/2100 or 2018/605 — and for a mixture, all of that for each such substance at 0,1 % by weight or more.

None of these produce a pictogram. They are regulatory status facts, and they sit here because a purchaser needs to know them before the substance enters a product, not after.

Hazards that produce no classification at all

The second half of 2.3 collects the hazards the criteria do not reach — air contaminants formed during processing, dustiness, sub-threshold explosive properties, dust explosion, cross-sensitisation, suffocation, freezing, odour or taste potency, harm to soil-dwelling organisms. For dust, REACH supplies the wording itself — “May form explosible dust-air mixture if dispersed” — and Annex 4 points competent authorities at the standardised phrases in A11.2.7.3.

Almost every item on that list is a hazard of use, not a property of the substance in the drum. That is what 2.3 is for — and why it must not become a soft landing for a hazard that does have a classification.

The three items the US text asks for and the EU does not

Appendix D numbers section 2 as five lettered items, not three subsections — and three of them have no EU counterpart.

Item (d) requires hazards not otherwise classified to be described where the classification process identified them. HNOC is a defined term: § 1910.1200(c) makes it an adverse physical or health effect found in the scientific evidence that meets no hazard class the standard addresses — and expressly not an effect that merely falls below a class cut-off, or sits in a GHS category OSHA never adopted. That exclusion is the interesting part, because the EU’s 2.3 asks for precisely the sub-threshold cases the US definition rules out.

Item (e) requires an arithmetic statement: where an ingredient with unknown acute toxicity is used in a mixture at ≥ 1 % and the mixture was not classified on testing of the mixture as a whole, the sheet must state that X % of the mixture consists of ingredients of unknown acute toxicity. It declares the quality of the evidence rather than the hazard, and Annex II has nothing like it.

Item (c) is newer. Paragraph (d)(1)(ii) requires the classification to take in chemical reaction products associated with known or reasonably anticipated uses, and Appendix D separates them out: item (a) is the classification excluding (d)(1)(ii), item (c) reports what (d)(1)(ii) produced. A US sheet therefore distinguishes the hazards of the product as shipped from the hazards of what it becomes in foreseeable use.

How section 2 connects to the rest of the sheet

• Subsection 2.1 hands ingredient information straight to subsection 3.2: the classification is the conclusion, section 3 is the evidence, and a classification that cannot be traced there is unverifiable.
• The “most important adverse effects” paragraph is drawn from sections 9 to 12, so a change in toxicological or ecological information obliges a rewrite here too.
• Abbreviated codes in 2.2 are permissible only because section 16 carries the full text of every classification and hazard statement. The two are a single mechanism.
• The classification decides which protection section 8 has to specify — a skin corrosive and a skin irritant do not get the same glove.
• The dust and air-contaminant entries in 2.3 reappear as handling conditions in section 7 and as decomposition behaviour in section 10.
• Section 14 classifies the product under a separate body of transport law. Nothing in section 2 may be read across into it, and nothing in it changes section 2.

What a good section 2 looks like

Three tests, answerable by reading the sheet alone:

1. Does every label element trace back to a class? Take each pictogram, the signal word and each statement in 2.2, and find the class in 2.1 that generates it. Anything with no parent, and any class with no child, is a defect. The inspector and the label constructor run that check mechanically.
2. Is 2.3 answered rather than left? PBT, vPvB and endocrine-disruption status are questions with three honest answers — meets the criteria, does not meet them, or not assessed. “None known” is not one of them.
3. Could a non-expert repeat the hazards back to you? That is the standard 2.1 sets. A summary that only makes sense to someone already holding the classification criteria satisfies the cross-reference to sections 9 to 12 and fails the requirement written beside it. Compare with section 2 on a real substance page to see how short it can be and still pass.