How Courts Read Contracts: Ambiguity, Extrinsic Evidence, and the Canons of Construction

Summary. Two courts reading the same contract can reach opposite results, and the reason is almost never the words. It is the threshold question of whether the court will look outside the document at all. This article explains that threshold, the two competing traditions answering it differently, and everything following from it: the parol evidence rule and what an integration clause actually accomplishes, the interpretive hierarchy running from express terms through course of performance and course of dealing to usage of trade, the canons of construction with honest assessments of what each is worth, and the drafting practices that keep a dispute from turning on a doctrine.

A contract for the sale of "chicken" turns out to mean different things to the buyer and the seller. The buyer wanted young broiling chickens. The seller shipped stewing chickens — older, tougher, cheaper birds that are unambiguously chicken.

Judge Friendly's opinion in Frigaliment Importing Co. v. B.N.S. International Sales Corp., 190 F. Supp. 116 (S.D.N.Y. 1960), begins: "The issue is, what is chicken?" It then works through the contract language, the negotiations, trade usage, the market price (which was below the cost of broilers, suggesting the seller's reading), government regulations defining "chicken" to include stewing birds, and the parties' conduct — and holds that the buyer, who bore the burden of proving its narrower meaning, failed to carry it.

That case is the whole subject. A word that seemed obvious to each side meant something different to the other; the resolution required looking at everything around the contract; and the outcome turned on who had to prove what.

This article is about how courts actually do that.

Part I: The threshold question, and the two traditions

Everything depends on whether a court will consider evidence outside the document.

The traditional or "four corners" approach. The court first determines, from the document alone, whether the language is ambiguous. If it is not, the plain meaning governs and extrinsic evidence is inadmissible to vary it. If it is ambiguous, extrinsic evidence comes in to resolve the ambiguity.

New York is the leading exponent. In W.W.W. Associates, Inc. v. Giancontieri, 77 N.Y.2d 157 (1990), the Court of Appeals refused to consider evidence that a reciprocal cancellation clause was intended for one party's benefit only, holding that "when parties set down their agreement in a clear, complete document, their writing should as a rule be enforced according to its terms," and that extrinsic evidence is not admissible to create an ambiguity in an otherwise unambiguous writing.

The contextual approach. The court considers extrinsic evidence at the outset, to determine whether the language is "reasonably susceptible" to the meaning a party urges — because words have no meaning apart from the circumstances in which they were used.

California is the leading exponent. In Pacific Gas & Electric Co. v. G.W. Thomas Drayage & Rigging Co., 69 Cal. 2d 33 (1968), Justice Traynor rejected the view that a court can determine a contract's meaning from the instrument alone: "The test of admissibility of extrinsic evidence to explain the meaning of a written instrument is not whether it appears to the court to be plain and unambiguous on its face, but whether the offered evidence is relevant to prove a meaning to which the language of the instrument is reasonably susceptible."

Why this matters more than any canon. In a four-corners jurisdiction, a party with strong evidence of what the parties actually meant may never get to present it. In a contextual jurisdiction, a party with clear text may face a trial about what the text meant. The choice-of-law clause therefore does interpretive work that most drafters never consider.

And the honest criticism of the contextual approach, made memorably in Trident Center v. Connecticut General Life Insurance Co., 847 F.2d 564 (9th Cir. 1988): applying California law, the Ninth Circuit was compelled to allow extrinsic evidence contradicting language it called "the most unambiguous in the English language," and observed that under Pacific Gas "no contract, however clear, can be enforced without a trial" — a result the court plainly disliked and was bound to reach.

Most jurisdictions sit between these poles, and the practical question in any dispute is where the forum's law falls and how the specific judge applies it.

Part II: Ambiguity — what it is and what it is not

A contract is ambiguous when it is reasonably susceptible to more than one meaning. It is not ambiguous merely because the parties disagree about what it means, which they always do by the time anyone is reading it in litigation.

Patent ambiguity appears on the face of the document — an internal inconsistency, a term defined two ways, a schedule that contradicts the body.

Latent ambiguity appears only when the language is applied to facts. The classic illustration is Raffles v. Wichelhaus, the "Peerless" case: a contract for cotton to arrive on the ship Peerless from Bombay, where there were two ships named Peerless sailing from Bombay months apart. Nothing in the document reveals the problem; the world does.

Why the distinction matters: extrinsic evidence is generally admissible to demonstrate a latent ambiguity even in jurisdictions that will not admit it to create one, because the ambiguity arises from the facts rather than from the words.

Ambiguity is a question of law for the court in most jurisdictions, and its resolution — once ambiguity is found — is generally a question of fact. That allocation decides whether a case can be resolved on summary judgment, which is why the ambiguity fight is usually the whole motion. See Summary Judgment Under Rule 56.

Delaware's formulation is useful because so many commercial agreements are governed by its law. In GMG Capital Investments, LLC v. Athenian Venture Partners I, L.P., 36 A.3d 776 (Del. 2012), the Delaware Supreme Court held that where a contract is ambiguous, summary judgment is improper if the moving party's reading is not the only reasonable one — and that extrinsic evidence must then be considered. The court also reiterated that a contract is not ambiguous simply because the parties disagree.

Part III: The parol evidence rule

The parol evidence rule is not a rule of evidence. It is a rule of substantive contract law: where the parties have adopted a writing as the final expression of their agreement, prior or contemporaneous agreements are discharged to the extent they contradict it.

The analysis has two steps:

Step one: is the writing integrated, and how completely?

Not integrated — the parol evidence rule does not apply.
Partially integrated — final as to the terms it contains; consistent additional terms may be proved.
Completely integrated — the exclusive statement of all terms; even consistent additional terms are excluded.

Step two: what is offered, and for what purpose?

Excluded: prior or contemporaneous agreements that contradict an integrated writing; and, in a completely integrated writing, consistent additional terms.

Not excluded — and this list is where most of the practice lies:

Evidence to interpret the meaning of terms (subject to the jurisdiction's threshold rule discussed above).
Evidence of a subsequent modification — the rule reaches only prior and contemporaneous agreements.
Evidence of fraud, duress, mistake, illegality, or unconscionability.
Evidence that the contract was subject to a condition precedent to its effectiveness.
Evidence supporting reformation for mutual mistake or scrivener's error.
Evidence of a collateral agreement supported by separate consideration.
Evidence of course of performance, course of dealing, and usage of trade under the UCC, which may explain or supplement even a completely integrated writing.

The merger or integration clause — "This Agreement constitutes the entire agreement between the parties and supersedes all prior negotiations, representations, and agreements" — is evidence of complete integration and is given substantial weight. But it is not conclusive in every jurisdiction, and it does not bar fraud claims in many, which is why sophisticated agreements add a non-reliance clause: an affirmative representation by each party that it has not relied on any statement outside the agreement. That formulation is materially more effective against a fraudulent inducement claim than a merger clause alone, and Delaware in particular has enforced it.

Part IV: The interpretive hierarchy

Where the agreement is silent or unclear, the sources of meaning are ranked.

Under the UCC — § 1-303 and § 2-202 — the express terms, course of performance, course of dealing, and usage of trade are to be construed as consistent with each other whenever reasonable; and when that is unreasonable, the order of priority is:

Express terms
Course of performance — the parties' conduct under this agreement, repeated with knowledge and without objection
Course of dealing — the parties' conduct in prior transactions with each other
Usage of trade — practices so regularly observed in a place, vocation, or trade as to justify an expectation that they will be observed in the transaction

Why the ranking is worth memorizing. Course of performance beats course of dealing beats trade usage — but all three can supplement and explain express terms, and course of performance is also relevant to waiver and modification. A party that accepted nonconforming performance eleven times without objection is in a very different position than one that objected the first time.

Outside the UCC, the Restatement (Second) of Contracts adopts a similar hierarchy at §§ 202 and 203, together with the general rules of interpretation at § 200 and following: the whole writing is interpreted together; technical terms get technical meaning; and an interpretation giving a reasonable, lawful, and effective meaning to all terms is preferred to one leaving a part unreasonable, unlawful, or of no effect.

Practical consequence for both drafters and litigators. In a commercial dispute, the first discovery request should be for the parties' communications about performance — invoices, delivery records, acceptance and rejection notices, and the correspondence around them. That record is frequently more probative of meaning than the contract, and it is where the case is won. See Mastering Document Discovery.

Part V: The canons, honestly assessed

Courts recite canons. They rarely decide cases on them alone. Here is what each is actually worth.

Give effect to all terms; avoid surplusage. Strong. Courts genuinely resist readings that render a provision meaningless, and "your interpretation makes Section 7 do nothing" is among the most effective interpretive arguments available.

The specific controls the general. Strong and frequently dispositive. A detailed provision addressing a subject prevails over a general clause that also touches it.

Read the contract as a whole. Strong as a framing principle; it structures the argument rather than deciding it.

Contra proferentem — construe ambiguity against the drafter. Moderate, and highly context-dependent. It applies with real force to adhesion contracts, consumer agreements, and insurance policies, where courts routinely invoke it. It applies weakly or not at all between sophisticated parties who both had counsel and both edited the document — and many commercial agreements now expressly waive it: "The parties have participated jointly in the negotiation and drafting of this Agreement, and any rule of construction that ambiguities are resolved against the drafting party shall not apply."

Ejusdem generis — a general term following specific ones is limited to things of the same kind. Strong in the settings where it matters, and it is the reason force majeure catch-alls fail. "Fire, flood, earthquake, war, or other cause beyond the party's control" does not reach a market collapse, because a market collapse is not of the same kind as the enumerated events. The drafting response is to say so expressly: "whether or not similar to the foregoing." See Force Majeure, Impracticability, and Frustration of Purpose.

Noscitur a sociis — a word is known by its associates. Moderate; a cousin of ejusdem generis.

Expressio unius — the expression of one thing excludes others. Weak to moderate, and frequently overused. It works where the list plainly purports to be exhaustive and fails where it does not. The drafting response is "including, but not limited to," or, better, "including without limitation."

The last antecedent rule — a qualifying phrase modifies only the immediately preceding term. Weak, mechanical, and regularly displaced by the series-qualifier principle: where a modifier follows a parallel series, it may apply to the whole series, particularly where a determiner is not repeated. Both canons appear in the same opinions on opposite sides. The drafting response is to break the sentence into subparts rather than to rely on a comma.

Handwritten over typed; typed over printed. Narrow but reliable, in the rare case with a marked-up form.

Words over numerals. Narrow and reliable.

Interpretation against forfeiture. Moderate; courts prefer readings that avoid a disproportionate penalty.

Public policy and lawfulness. A preferred reading is one that is lawful and enforceable.

The honest summary. Canons are tiebreakers and rhetoric. They almost never beat: text that plainly says something; a reading that makes another provision meaningless; or a course of performance in which the parties themselves demonstrated what they meant. Argue those three first, and use the canons to close.

Part VI: When interpretation is not the answer

Sometimes the problem is not what the contract means but whether it says what the parties agreed.

Reformation corrects a writing that, because of mutual mistake or scrivener's error, does not reflect the parties' actual agreement. It requires clear and convincing evidence, and it is the correct remedy where the wrong number, the wrong party name, or the wrong legal description was typed.

Mutual mistake as to a basic assumption may support avoidance where it materially affects the exchange and the adversely affected party did not bear the risk.

Unilateral mistake rarely supports relief, unless the other party knew or had reason to know of it, or enforcement would be unconscionable.

Misunderstanding — the Peerless problem — means no contract was formed at all where the parties attached materially different meanings and neither knew or had reason to know of the other's meaning.

Gap filling is different from interpretation. Where the parties simply did not address a subject, courts supply a term: the UCC's provisions on price, delivery, payment, and warranties; the implied covenant of good faith and fair dealing; and, in employment and other relational contexts, various default rules. The implied covenant does not override express terms in most jurisdictions — it constrains discretion the contract confers, rather than adding obligations the contract omitted.

Unconscionability and public policy can defeat a term the words plainly reach, particularly in consumer and adhesion contexts.

Part VII: Drafting so that none of this matters

Every doctrine above exists because someone wrote something unclearly. The drafting responses:

Define the terms that carry weight. A definitions section is not decoration. If "Net Revenue," "Confidential Information," "Change of Control," or "Material Adverse Effect" matters, define it — and then use the defined term consistently, capitalized, everywhere.

Say what the parties intended the words to mean, where a term is unusual: "For the avoidance of doubt, 'chicken' means young broiling chickens of not more than 2½ pounds."

Use "including without limitation" rather than bare "including," and add "whether or not similar to the foregoing" to any catch-all you intend to be broad.

Break sentences into numbered subparts rather than relying on punctuation to allocate modifiers. If a qualifier applies to all three items, say so in a lead-in.

State the order of precedence among the agreement, exhibits, schedules, statements of work, and purchase orders — and then actually follow it, because a purchase order with conflicting terms sent after the agreement is the origin of a great many disputes. See The Battle of the Forms Under UCC Section 2-207.

Include a merger clause, and a non-reliance clause, and understand the difference.

Address contra proferentem with a joint-drafting provision if that is what you want.

Choose governing law deliberately, knowing that the choice determines whether extrinsic evidence comes in at the threshold. See Choice of Law, Forum Selection, and Where Your Dispute Will Be Decided.

Require written modification and no oral waiver — and then follow it in practice, because a course of performance inconsistent with the clause can waive the clause itself.

Read the document once for consistency after every negotiation round. Most patent ambiguities are introduced in the last two exchanges, when a defined term is changed in one place and not another.

Part VIII: A worked example

The contract. A supply agreement: "Supplier shall deliver 10,000 units per month. Buyer shall purchase all of its requirements for the Product from Supplier. This Agreement may be terminated by either party upon 90 days' notice for material breach that remains uncured after 30 days' written notice."

The dispute. Buyer's requirements drop to 4,000 units per month after it loses a customer. Supplier sues, arguing Buyer must take 10,000. Buyer argues it must buy only its actual requirements, which are 4,000.

The interpretive analysis:

1. Is there an ambiguity? Two provisions arguably conflict: a fixed quantity and a requirements obligation. That is a patent ambiguity on the face of the document, and it exists in any jurisdiction.

2. Give effect to all terms. A reading that makes "10,000 units per month" meaningless is disfavored; so is one that makes "all of its requirements" meaningless. The reading that reconciles both — 10,000 as a minimum, requirements above that exclusively from Supplier — gives each provision work to do, and it is the strongest textual argument.

3. Specific over general. If one provision is in a detailed "Quantity" section and the other in a general "Purchase Obligations" recital, the specific one controls.

4. Course of performance. For eighteen months Buyer ordered between 9,000 and 12,000 and paid for what it ordered. When it ordered 9,000, Supplier invoiced 9,000 and did not object. That course of performance is powerful evidence that neither party understood 10,000 as a firm minimum — and it is admissible in both four-corners and contextual jurisdictions, because it is subsequent conduct rather than prior negotiation.

5. Course of dealing and trade usage. Prior contracts between these parties, and industry practice on whether "requirements" contracts carry minimums.

6. Good faith limits on requirements. Even on Buyer's reading, a requirements obligation is measured by good faith, and a buyer may not reduce its requirements disproportionately for the purpose of escaping the contract. If Buyer's reduction reflects a genuine loss of business, it is in good faith; if it reflects a decision to source elsewhere, it is not.

7. Extrinsic evidence. In a contextual jurisdiction, the negotiation record comes in from the start. In a four-corners jurisdiction, it comes in only after the court finds the ambiguity — which, given the patent conflict, it will.

8. Contra proferentem. Supplier drafted the form. Between two commercial parties both represented by counsel, the canon carries modest weight, and it is the last argument rather than the first.

The likely outcome. The course of performance dominates. Eighteen months of invoicing without objection is what the parties themselves said the contract meant, and courts credit that over both parties' litigation positions.

The drafting fix, in one sentence: "Buyer shall purchase a minimum of 10,000 units per month and shall purchase all of its requirements for the Product in excess of that minimum exclusively from Supplier." Nine seconds of drafting, and no dispute.

Part IX: Interpreting the clauses that get litigated most

Certain provisions generate a disproportionate share of interpretive disputes. Each has a known failure pattern.

Indemnification. The recurring questions: does it cover first-party losses or only third-party claims; does it reach the indemnitee's own negligence (which most states require to be stated in clear and unmistakable language, sometimes in conspicuous type); is it triggered by a "claim" or by a "loss"; and does the duty to defend exist separately from the duty to indemnify, arising earlier and on a broader standard. Drafting fix: say which losses, say whether the indemnitee's own negligence is covered, and address defense expressly. See Indemnification and Limitation of Liability.

Limitation of liability. Does the consequential damages waiver survive a breach of the limited remedy? Does the cap apply to indemnity obligations, to confidentiality breaches, to gross negligence? Is the cap an aggregate or per-claim? Drafting fix: state the exclusions from the cap expressly, and address what happens if the exclusive remedy fails of its essential purpose.

"Best efforts," "commercially reasonable efforts," and "reasonable efforts." Courts do not apply a settled hierarchy, and several have held there is no meaningful difference among them absent a definition. Drafting fix: define the standard by reference to conduct — "shall devote at least [X] personnel," "shall spend not less than $[Y]," "shall take the actions listed on Schedule ___."

Termination for convenience versus for cause. Notice periods, cure periods, the survival of obligations after termination, and the effect on work in progress and prepaid amounts. Drafting fix: a survival clause listing the sections that survive, by number.

Material Adverse Effect. Extensively litigated in acquisitions, with a well-developed body of Delaware law requiring a durationally significant, company-specific effect and with carve-outs for general economic and industry conditions. Drafting fix: define it with carve-outs and, where possible, a quantitative threshold.

Exclusivity and requirements. As in the worked example above — reconcile fixed quantities with requirements obligations, and remember the good faith limit under UCC § 2-306.

Confidentiality. Duration, the definition of Confidential Information, the standard exceptions (independently developed, publicly available, rightfully received), residual knowledge clauses, and the treatment of compelled disclosure. See Drafting a Nondisclosure Agreement Checklist.

Assignment and change of control. Whether "assignment" reaches a merger or a change of control is jurisdiction-specific and frequently outcome-determinative in an acquisition. Drafting fix: say whether a change of control constitutes an assignment.

Notice provisions. Method, address, when deemed effective, and — the recurring failure — whether email counts. A termination notice sent by email under a clause requiring certified mail is frequently ineffective, and the consequence is another full term.

Dispute resolution. Scope ("arising out of or relating to" is broader than "arising under"), carve-outs for injunctive relief, arbitration rules and seat, and class waivers. See Selecting and Drafting an Arbitration Clause.

Part X: How to litigate an interpretation dispute

The doctrine is only useful when organized into an argument. The sequence that works:

1. Start with the text, and make the strongest textual argument first. Which reading gives every provision work to do? Which reading makes a provision meaningless? That argument is available in every jurisdiction, it does not depend on the ambiguity threshold, and it persuades judges more than any canon.

2. Determine the forum's threshold rule. Four corners or contextual? This governs your motion strategy: in a four-corners jurisdiction, a clear-text party moves for judgment on the pleadings or summary judgment and resists any evidentiary detour; in a contextual jurisdiction, the same party must be prepared for a trial on meaning and should build the extrinsic record anyway.

3. Find the course of performance. Before anything else in discovery: invoices, purchase orders, delivery records, acceptance and rejection notices, payment records, and the correspondence around them. What the parties did is the most persuasive evidence of what they meant, it is admissible everywhere, and it frequently decides the case.

4. Then course of dealing and trade usage. Prior contracts between these parties; industry practice, proved by testimony from participants or by a trade association's published standards.

5. Then the negotiation record, subject to the parol evidence rule and to the forum's threshold. Look for drafts showing a term that was proposed and removed — that is among the most powerful interpretive facts available, because it shows the parties considered and rejected the reading your opponent now urges.

6. Then the canons, as reinforcement.

7. Consider whether interpretation is the right frame at all. If the writing does not reflect the actual agreement, the claim is reformation, and the evidentiary standard and the remedy are different. If the parties never agreed, it is misunderstanding and there may be no contract.

8. Frame the ambiguity question with the procedural posture in mind. Ambiguity is generally a question of law; its resolution is generally a question of fact. A party seeking summary judgment must therefore argue no ambiguity; a party resisting it must argue ambiguity plus a reasonable competing reading supported by evidence. Choosing the wrong frame loses winnable motions.

Part XI: Frequently asked questions

"The contract says one thing but we all knew it meant another." Whether you can prove that depends on the forum's threshold rule, on whether the writing is integrated, and on what kind of evidence you have. Course of performance is your best evidence because it is subsequent conduct, not prior negotiation, and it is admissible even where the parol evidence rule bars the negotiation record.

"There's an integration clause. Are we stuck?" It is strong evidence of complete integration, and it will exclude prior and contemporaneous agreements. It generally does not bar: evidence of subsequent modification; evidence of fraud (in many jurisdictions); evidence of a condition precedent to effectiveness; reformation for mutual mistake; or course of performance and trade usage under the UCC. A non-reliance clause is materially stronger against fraud claims than a merger clause alone.

"Two provisions conflict. Which wins?" Work through, in order: does one provision expressly govern the conflict (an order-of-precedence clause)? Is one specific and the other general? Does one reading make a provision meaningless? What did the parties actually do? Only then reach the canons.

"Can we fix a typo after signing?" By amendment, if both agree. If one does not, the remedy is reformation for mutual mistake or scrivener's error, on clear and convincing evidence — and the evidence is usually the drafts showing what both sides believed they were signing.

"The other side is relying on an email from before signing." Object under the parol evidence rule if the writing is integrated and the email contradicts it. But note the exceptions, and note that in a contextual jurisdiction the email may come in to show what the words were reasonably susceptible of meaning.

"Does 'best efforts' actually require more than 'reasonable efforts'?" Courts are inconsistent, and several have said there is no meaningful difference absent a definition. If the standard matters, define it by reference to specific conduct.

"Our purchase order has different terms than the master agreement." Classic problem. The master agreement should contain an order-of-precedence clause, and the purchase order should incorporate rather than supersede it. Where the documents were exchanged as offer and acceptance, UCC § 2-207 governs which terms are in the contract. See The Battle of the Forms and Purchase Order and Terms of Sale Review Checklist.

"We've been accepting late deliveries for two years. Have we waived the deadline?" Possibly, and this is exactly what course of performance does. A no-waiver clause helps but is not conclusive — in several jurisdictions a course of performance inconsistent with a no-waiver clause can waive the clause. The fix is to object in writing, each time, and to reserve rights expressly.

"Which governing law should we choose?" Among many considerations, one that drafters rarely weigh: the choice determines whether extrinsic evidence comes in at the threshold. If you value textual predictability, a four-corners jurisdiction serves you. If you expect to need context, a contextual jurisdiction does. See Choice of Law, Forum Selection, and Where Your Dispute Will Be Decided.

Part XII: For non-lawyers reading a contract

Most contracts are read by the people who will perform them rather than by lawyers, and a few habits catch most of the problems.

Read the definitions first. Capitalized terms mean exactly what the definitions section says, and nothing else. A contract about "Products" where "Products" is defined to exclude the thing you are buying is a contract about nothing.

Read the obligations backwards. Start with what happens if something goes wrong — termination, remedies, limitation of liability, indemnity, dispute resolution — because that is what the document is actually for. The parts describing what everyone intends to do are the parts nobody argues about.

Find every number and every date, and check that each is right. Quantities, prices, percentages, notice periods, cure periods, terms, and renewal deadlines. Most disputes involve one of them.

Ask what a provision does. If you cannot say what work a sentence performs, either it performs none — in which case a court may strain to give it meaning you did not intend — or you have not understood it.

Look for the word "not." Negatives, double negatives, and exceptions to exceptions are where meaning inverts and where readers skim.

Check that the exhibits exist, that they are attached, and that the body's references to them are accurate. "As set forth on Schedule B" is a problem when there is no Schedule B.

Notice what the contract does not say. Silence is a term. If the agreement says nothing about who owns the work product, who bears shipping costs, what happens on a change of control, or what notice terminates it, the law's default rule applies — and the default rule may not be what you assumed.

Watch the notice clause. Then, when the time comes, use the method it requires. More rights are lost to a termination notice sent by email under a clause requiring certified mail than to any interpretive doctrine in this article.

And write down what you do. The single most persuasive evidence of what a contract means is what the parties did under it — invoices, deliveries, acceptances, objections, and the emails around them. Keeping that record costs nothing and, in a dispute, is worth more than the contract.

Part XIII: When the "contract" is really five documents

Most commercial relationships are not governed by one document. There is a master agreement, a statement of work, an exhibit of specifications, a purchase order that the buyer's system generated automatically, an invoice with terms printed on the back, and a click-through the vendor's portal made someone accept before granting access. Courts read this pile as a single agreement where the parties intended one, and the interpretive fights follow predictable lines.

Incorporation by reference. A document referred to in the agreement becomes part of it if the reference is clear and the incorporated document is identified with reasonable certainty. "The Vendor's standard support policy" is not reasonably certain if the vendor can rewrite it tomorrow; "the Support Policy attached as Exhibit C" is. The practical question courts ask is whether the party to be bound had a reasonable opportunity to know what it was agreeing to. A URL that resolves to a page the drafter may change unilaterally is the weakest form of incorporation there is, and if the incorporated terms are onerous — an arbitration clause, a liability cap, a fee-shifting provision — the weakness of the reference is often the whole case.

Order of precedence. When the documents conflict and the agreement says nothing about which controls, courts apply the ordinary tools: the specific governs the general, the later governs the earlier when the later is a valid amendment, and a negotiated term governs a form term. That last principle is codified in the UCC's battle-of-the-forms provisions and reflected in the common law's hostility to preprinted boilerplate that contradicts what the parties actually discussed. But relying on those defaults is a gamble, because each of them can be argued both directions. An express precedence clause — the agreement controls over the SOW, the SOW controls over the exhibits, and no purchase order term has any effect regardless of what it says or when it is issued — converts a fact question into a reading exercise.

Amendments and the writing requirement. Nearly every commercial contract says it may be amended only in a signed writing. Nearly every long-running commercial relationship then proceeds to modify itself informally: a different delivery schedule agreed by email, a price change accepted by paying the higher invoice, a scope expansion handled by nobody objecting. Most jurisdictions hold that a no-oral-modification clause can itself be waived by conduct, which means the clause protects you less than it appears to. The UCC narrows this — § 2-209(2) makes a no-oral-modification clause enforceable between merchants — but even there, an attempted modification can operate as a waiver. The lesson for the operator is not that the clause is useless; it is that the clause only works if the organization behaves consistently with it. Object in writing when the other side deviates. Paper the changes you accept. A file full of unpapered accommodations is a file that will be read, three years later, as course of performance establishing the meaning you never agreed to.

Primary authority and further reading

Restatement (Second) of Contracts §§ 200–223, particularly § 201 (whose meaning prevails), § 202 (rules in aid of interpretation), § 203 (standards of preference), § 209–216 (integrated agreements and the parol evidence rule), § 221 (usage supplementing an agreement), and § 222–223 (usage of trade and course of dealing).
UCC § 1-303 (course of performance, course of dealing, and usage of trade), § 2-202 (final written expression and parol or extrinsic evidence), § 2-306 (output and requirements contracts and the good faith limit), and § 2-207 (additional terms).
Pacific Gas & Electric Co. v. G.W. Thomas Drayage & Rigging Co., 69 Cal. 2d 33 (1968) — the contextual approach.
W.W.W. Associates, Inc. v. Giancontieri, 77 N.Y.2d 157 (1990) — the four corners approach.
Trident Center v. Connecticut General Life Insurance Co., 847 F.2d 564 (9th Cir. 1988) — the critique of the contextual approach.
Frigaliment Importing Co. v. B.N.S. International Sales Corp., 190 F. Supp. 116 (S.D.N.Y. 1960) — burden of proof on meaning, and the full range of interpretive sources.
GMG Capital Investments, LLC v. Athenian Venture Partners I, L.P., 36 A.3d 776 (Del. 2012) — ambiguity, summary judgment, and extrinsic evidence under Delaware law.
Raffles v. Wichelhaus (the Peerless case) — latent ambiguity and misunderstanding.
Corbin on Contracts and Williston on Contracts, which represent the contextual and formalist traditions respectively and whose disagreement is the disagreement described above.