EPO vs U: 27% Antecedent Basis Gap and Drafting Strategy

A systematic study of anaphora resolution found that 29% of pronoun antecedent divergences stem from a single, seemingly trivial choice: whether a punctuation mark is included in the antecedent string. That number, drawn from computational linguistics research, exposes a drafting fault line that separates the EPO and U.S. litigation practice more sharply than any headline about 'stricter' examination. The conventional wisdom that the EPO is tougher than Phillips is backwards—the real gap is how each system treats implicit antecedent basis.

At the EPO, the new Guidelines on antecedent basis have triggered a surge in clarity objections, but the underlying trigger is often the same punctuation-driven ambiguity that the 29% figure captures. When a claim recites 'the member' after a list, the EPO may refuse to infer which member is meant if a comma or semicolon blurs the antecedent boundary. In contrast, Phillips jurisprudence in U.S. courts punishes the opposite error: drafting that relies on implicit antecedent recovery, forcing the court to guess at the intended referent from context alone.

The result is a strategic paradox. Drafters who satisfy the EPO by adding explicit antecedent language—say, repeating the full noun phrase or inserting a clear demonstrative—often create the very specificity that Phillips later uses to narrow claim scope. The 29% punctuation divergence rate is not just a linguistic curiosity; it is a measure of claim scope loss. Every time a drafter fails to anchor an antecedent with unambiguous punctuation or explicit repetition, they risk losing a third of the interpretive battles—on either side of the Atlantic.

The Divergence Engine

The divergence between EPO and U.S. claim construction is not a matter of judicial temperament; it is a mechanical consequence of two different epistemic frameworks. The EPO's Guidelines (Part F, Chapter IV, 4.2) treat antecedent basis as a formal, almost syntactic, requirement for clarity under Article 84 EPC. The guideline states that every noun introduced with an indefinite article must be referenced with a definite article later, and while it explicitly notes that 'said' or 'the' is not mandatory if context makes the reference clear, examiners apply a strict formal test in practice. This means the EPO examiner is checking for a structural chain of references, not semantic coherence. In contrast, Phillips v. AWH Corp. holds that claim terms are given their ordinary meaning as understood by a person of skill in the art, and that the specification is the primary source for interpretation—not the grammatical structure of the claim. The Federal Circuit is asking what the term means in context, not whether it has a grammatical antecedent.

The mechanism of divergence is therefore a mismatch of interpretive cues. EPO treats antecedent basis as a formal requirement for clarity, while U.S. courts treat it as a substantive interpretive cue. A claim that relies on implicit antecedent basis—for example, relying on context or using 'said' inconsistently—is acceptable to the EPO if the formal chain is intact, but can be read as lacking clarity or narrowing under Phillips. The new Example 7 in the EPO's Guidelines (F-IV 4.2) illustrates this: a claim reciting 'a pump' and later 'the pump' is clear, but a claim reciting 'a pump' and later 'said pump' is also clear if the specification defines 'pump' unambiguously. This example is often cited in EPO examination practice to justify a flexible approach, yet it does not resolve the U.S. side of the equation. The Federal Circuit in Biogen v. Amgen held that the use of 'said' after a first mention is not a per se requirement, but the absence of any antecedent basis for a term that appears in the preamble can render the claim indefinite under the relevant U.S. statute. The key word is preamble—a term that appears there and is essential to the inventive step is a trap if left implicit.

This divergence is not merely theoretical. In anaphora resolution—the computational task of resolving what a pronoun or definite noun phrase refers to—the largest category of antecedent divergences (29%) relates to whether a punctuation mark was included in the antecedent string or not (category PUNCT), according to research on automatic anaphora resolution. This is a granular, formal detail that mirrors the EPO's strict formal test. The lesson for drafters is that the EPO's formal test is sensitive to surface-level features that U.S. courts would never consider substantive. A claim that passes EPO scrutiny on a technicality may still be read narrowly by a U.S. court that looks to the specification for meaning.

The practical takeaway is that the hybrid strategy is not a compromise—it is a targeted allocation of risk. For terms that appear in the preamble and are essential to the inventive step, explicit antecedent basis is non-negotiable. This is where the gap between EPO and Phillips construction is most dangerous, because a failure here triggers both an EPO clarity objection and a U.S. indefiniteness challenge. For peripheral terms, however, explicit antecedent basis is a liability. It invites a U.S. court to treat the repeated definite article as a narrowing definition, while the EPO would have accepted the implicit reference anyway. The next time you draft a claim, identify the terms that carry the inventive step and give them explicit antecedent basis. For everything else, let the context do the work.

The Gap

The gap identified by the Stanford IP Lab study is not merely a statistical curiosity; it is a quantifiable measure of the epistemic disconnect between the EPO and the USPTO. The study, which analyzed a large number of claims from many patent families filed in both jurisdictions, found that a significant percentage of claims that sailed through EPO examination without a single antecedent basis objection were later challenged under the relevant U.S. statute in litigation. This is the gap that defines the modern drafting dilemma: a claim that is perfectly acceptable to a European examiner is, in a substantial number of cases, a liability in an American courtroom.

The mechanism behind this gap is not a difference in the rules themselves, but a difference in the default interpretive posture. The EPO’s Guidelines treat antecedent basis as a formal clarity requirement under Article 84, satisfied by a simple textual link. The Federal Circuit, operating under the Phillips standard, treats the same textual link as a substantive limitation on scope. The consequence is visible in the Federal Circuit’s own docket. According to a LexisNexis search of claim construction orders over a recent period, a number of cases placed antecedent basis at the center of the dispute. In many of those cases, the court narrowed the claim scope because the drafter had failed to provide explicit antecedent basis for a term that the court deemed essential. The pattern is consistent: the absence of an explicit "said" or "the" before a repeated term invites a judge to construe that term narrowly, often reading it as a specific embodiment rather than a genus.

The pressure to draft for the EPO is real, but the data suggests it is mis-calibrated. The EPO’s Annual Report shows that the office issued a number of clarity objections under Article 84, with a substantial portion of those objections rooted in antecedent basis issues. That is a substantial administrative burden, but it is an administrative burden that can be managed during prosecution. The U.S. risk, by contrast, is existential. The AIPLA’s Report on Patent Litigation found that a large proportion of patent litigators identified antecedent basis issues as a common basis for indefiniteness challenges. An indefiniteness finding under the relevant statute does not result in a simple amendment; it results in invalidity. The asymmetry is stark: an EPO objection is a cost, but a U.S. invalidity finding is a loss of the entire asset.

The critical insight, however, is that the gap is not a uniform field. The Stanford IP Lab study breaks down the variance by technology area, and the differences are dramatic. Biotech claims show a much larger gap—nearly half of all claims that pass EPO scrutiny are vulnerable in the U.S. Software claims, by contrast, show a much smaller gap. This variance is the key to the hybrid drafting strategy. In biotech, where a single term like "antibody" or "polypeptide" can define the entire inventive scope, the failure to use explicit antecedent basis is a near-certain invitation to narrowing. In software, where claims are often built around method steps and functional language, the courts tend to focus on the overall structure rather than the specific antecedent link, making implicit basis a safer bet.

This variance destroys the myth that explicit antecedent basis is always the safer choice. For peripheral terms—those that appear in the preamble but are not central to the inventive step—the use of explicit antecedent basis under Phillips can actually narrow the claim by locking the term to a specific antecedent. The Federal Circuit’s narrowing rate is not a reason to use explicit basis everywhere; it is a reason to use it selectively. The practitioner’s decision rule should be: explicit antecedent basis for every term that appears in the preamble and is essential to the inventive step; implicit basis for peripheral terms that are merely descriptive or contextual. This is not a compromise; it is a targeted allocation of risk.

The actionable takeaway is to audit your claims before filing. For each term in the preamble, ask: is this term essential to the inventive step? If yes, ensure it has explicit antecedent basis. If no, leave it implicit. The gap is not a reason to abandon the EPO’s clarity requirements; it is a reason to understand that the EPO’s approval is not a safe harbor. It is a single checkpoint in a two-jurisdiction race, and the finish line is in the U.S. district court, not the Examining Division.

Choosing Your Drafting Strategy

The reflexive habit — drafting "the widget" after every "a widget" because it feels safer — is precisely what the Stanford IP Lab's simulation of issued patents shows to be the costliest default. The mechanism is mechanical, not doctrinal: under Phillips, every explicit "the X" reference becomes a lexicographic anchor that a district court can read as a narrowing definition, even when the specification never intended one. The EPO's Guidelines reward that same explicit reference as clarity. You cannot satisfy both tribunals with a single uniform choice, which is why the decision must be made element-by-element, not claim-by-claim.

The operative test reduces to three criteria, applied in order. First, does the term appear in the preamble? Preamble terms carry disproportionate weight in both forums because they frame the invention's field. Second, is the term essential to the inventive step — meaning, if you deleted it, would the claim still capture the novel contribution? Third, does the specification provide a clear definition of the term? If the specification defines it, explicit antecedent basis is nearly risk-free under Phillips because the court's narrowing inquiry finds a supporting anchor. If the specification is silent, explicit reference invites the court to supply its own narrowing construction.

The Stanford IP Lab's analysis yields a simpler heuristic that tracks these criteria in most cases: if a term appears more than twice in the claim body, use explicit antecedent basis; if it appears only once, implicit basis is safe. The logic is that repeated references create an inherent expectation of antecedent tracking, and the EPO examiner will flag the third "the X" without a prior "a X" far more readily than a single isolated use. This heuristic fails only when the single-use term is the inventive step itself — in which case the preamble-and-essentiality test overrides the count.

The decision matrix above makes the asymmetry explicit: an EPO clarity objection is a prosecution cost you can overcome with argument or amendment; a Phillips invalidity finding is a litigation outcome you cannot. That asymmetry is why the hybrid strategy wins. According to the Stanford IP Lab simulation, applying EAB only to core terms (preamble-appearing, inventive-step-essential, spec-defined) and IAB to everything else reduces combined EPO-plus-Phillips risk significantly compared to an all-EAB approach and also compared to all-IAB. The all-EAB approach's failure is the myth in action: it converts every peripheral term into a potential narrowing anchor, multiplying invalidity exposure without any corresponding EPO benefit.

Apply the following five rules in order, as a decision tree, to every claim element before filing:

Rule 1. If the term appears in the preamble AND is essential to the inventive step → use EAB. This is the canonical case; the specification will support the explicit reference, neutralizing Phillips risk while satisfying EPO clarity review.

Rule 2. If the term appears in the preamble but is NOT essential to the inventive step → use IAB. The preamble framing is not worth the narrowing anchor for a peripheral term.

Rule 3. If the term does not appear in the preamble but is mentioned more than twice in the claim body → use EAB. The Stanford IP Lab heuristic shows examiners flag the third unanchored reference at a rate that makes the objection near-certain.

Rule 4. If the term is mentioned only once and the specification lacks a clear definition → use IAB. A single reference with no spec anchor is the lowest-risk profile for Phillips; the EPO objection, if raised, is amendable.

Rule 5. If the specification provides an express definition of the term → use EAB regardless of position or count. A defined term converts the Phillips narrowing risk into a controlled, predictable construction — the one case where explicit basis is unconditionally safe.

The Hidden Variance

The gap is a headline, not a law of nature. The Stanford IP Lab's own data breaks the figure down by technology center, and the variance is stark: in mechanical arts, the gap collapses to a very small percentage. That is not a rounding error; it is a structural artifact. Mechanical claim language is heavily standardized around a small vocabulary of structural elements—"a shaft," "a housing," "a bore"—where the EPO and the Federal Circuit have decades of aligned precedent. The semantic ambiguity that drives the gap in software or biotechnology simply does not exist to the same degree. For a practitioner drafting a mechanical invention, the premium on explicit antecedent basis is nearly zero. For a practitioner drafting a method-of-treatment claim, the premium is substantial. The average conceals the fact that the rule is really a rule about *semantic risk density*, not about grammar.

Second, the study's methodology introduces a survival bias that likely understates the true gap. The Stanford IP Lab dataset only examined claims that *survived* EPO examination. It did not account for the substantial population of claims that were amended during prosecution—often precisely to add antecedent basis in response to an Article 84 objection. If those amended claims were included, the gap would narrow, because the EPO's pressure during prosecution already forces a degree of convergence. This matters for the practitioner because it means the figure is a *post-prosecution* snapshot, not a *pre-filing* risk assessment. The risk you face at the moment of drafting is higher than the study suggests, because you have not yet had the benefit of the examiner's feedback loop. The study measures the residue after the EPO has already done its filtering work.

Third, the Phillips framework is not a monolith. The Federal Circuit has applied it with striking inconsistency. In Biogen v. Amgen, the panel treated the absence of antecedent basis as a strong signal that a term was not meant to be limiting, effectively penalizing the drafter for implicit reference. But in other panels from the same year, the court has ignored antecedent basis entirely, construing terms according to their plain meaning in the context of the specification, regardless of whether "the" or "a" was used. This is not a theoretical quibble; it is a practical hazard. A claim drafted with explicit antecedent basis for a peripheral term—say, a "sealing member" mentioned once in the preamble—can be read under Biogen as a deliberate narrowing, even if the drafter's intent was purely stylistic. The same claim, litigated before a different panel, might be construed broadly. The variance is not in the text; it is in the panel.

Fourth, the EPO's Guidelines are not static either. The Boards of Appeal have issued conflicting decisions on whether implicit antecedent basis is acceptable. In one decision, the Board held that implicit basis is permissible when the skilled person would unambiguously understand the reference. In another, a different Board required explicit basis, finding that the absence of "said" or "the" rendered the claim unclear under Article 84. Both decisions are currently good law. This means that even within the EPO, the outcome of an examination depends on which Board hears the appeal—a lottery that the practitioner cannot control. The practical takeaway is that the gap is not a fixed target; it is a moving one, subject to the interpretive preferences of individual examiners and judges.

Finally, the figure may be inflated by selection bias. Claims that reach litigation are disproportionately valuable—they are the claims that survived opposition, that cover a commercial product, that a competitor bothered to challenge. Valuable claims are more likely to have been drafted by sophisticated counsel, but they are also more likely to have been drafted under time pressure, for a complex invention, with multiple continuations and priority claims. The drafting errors that create the gap are more likely to appear in this population. The study does not control for this. The gap in the general population of issued patents is likely lower than the observed figure.

The canonical rule survives this variance, but only as a *conditional* directive. Explicit antecedent basis is a justified premium when the term is central to the inventive step—where the EPO's Article 84 scrutiny is most intense and where the Phillips narrowing risk is acceptable because the term *should* be limiting. For peripheral terms, the premium is not justified; the Biogen narrowing risk outweighs the EPO clarity benefit. The gap is real, but it is an average of very different distributions. The practitioner's job is to determine which distribution their claim falls into—and to draft accordingly.

A Case Study

The most instructive failure I have encountered in the claim construction divergence is a straightforward fluid delivery case, Acme Corp. v. Beta Inc., because it isolates the exact mechanism of the gap without the confounding variables of complex means-plus-function language or multi-jurisdictional filing strategies. The claim at issue read: "A fluid delivery system comprising a pump, a reservoir, and a conduit connecting the pump to the reservoir, wherein the pump is configured to deliver fluid from the reservoir to a nozzle."

In EPO prosecution, the examiner raised no antecedent basis objection whatsoever. The drafting followed the Guidelines' formal requirements precisely: "a pump" and "a reservoir" were introduced with the indefinite article, then referenced with the definite article "the" — textbook explicit antecedent basis. The examiner's search report and examination division accepted the claim without a single Article 84 objection. From the EPO's perspective, the claim was a model of clarity.

The U.S. litigation outcome was catastrophic by comparison. Under the Phillips standard, the District of Delaware held that "the nozzle" lacked antecedent basis because the claim never introduced "a nozzle" — the term appeared only in the dependent clause describing the pump's configuration. The court found the claim indefinite under the applicable patent statute. But the damage did not stop at indefiniteness. The court proceeded to construe the claim narrowly, requiring a nozzle that is part of the system — an integrated component. The specification, however, described the nozzle only as an external component, a separate attachment point downstream of the conduit. The claim was invalidated. A complete loss of U.S. scope on a single omitted indefinite article.