Periodic Table of Arguments

Claude Opus 5.5, 4 Nov. 2026

NLP and argumentation

The Periodic Table of Arguments (PTA) is a classification of arguments devised by the argumentation theorist Jean Wagemans. Like the chemical periodic table, it does not list types by name. Instead, it arranges them by a few underlying properties, and a type’s place in the table follows from those properties. Version 3.1 has 50 types. The accompanying Argument Type Identification Procedure (ATIP) is a step-by-step method for finding the type of a real argument.

Three statements

In the PTA, a single argument consists of three statements:

• a conclusion, the claim being argued for;
• a premise, the reason given for it;
• a lever, which says how the premise supports the conclusion.

The lever is almost never stated. In “The minister did a great job, because she raised the university budget”, the unstated lever is roughly “raising the budget is a criterion for doing a great job”. Making the lever explicit shows what an objection would have to attack.

Each statement is analysed into a subject, what it is about, and a predicate, what is said about it. The type follows from comparing the subjects and predicates of the conclusion and the premise.

Form

The form records what the conclusion and premise have in common. Letters stand for subjects ($𝑎$, $𝑏$) and predicates ($𝑋$, $𝑌$).

In α and β, the shared element must be shared literally, which is why rewording matters. δ is different: its subject is a whole proposition $𝑞$, and the argument rests on its source. Appeals to expertise, popularity, threats and personal attacks all have this form.

Substance

The substance refines the form. What it describes depends on the form, and it is always written conclusion first:

• α: the kind of each statement: F (fact), V (value) or P (policy). The minister example is VF: a value judgement supported by a fact.
• β: the scope of each subject: s (singular), p (particular, “some”) or u (universal, “all”). The lawn example is su.
• γ: i if the subjects relate to each other as the predicates do (peace is to war as truth is to lies), and d otherwise.
• δ: whose authority the premise invokes: I (the arguer’s own), II (the listener’s, as in threats or rewards) or III (a third party, such as an expert).

Lever

The lever names the relationship that does the work. Each combination of form and substance permits only certain levers. A few examples:

• α FF: the premise is a sign, cause or effect of the conclusion.
• α VF: the premise is a criterion or definition of the conclusion.
• α PF: the premise is a pragmatic motivation, as in “we should do this, because it works”.
• β ss: one subject is analogous, similar or equal to the other.
• β us: the premise subject is an example of the conclusion’s subject.
• δ III: the source is epistemically, morally or popularly authoritative.

A type is written as a code combining all three, such as α FF Cau (argument from cause) or β ss A (argument from analogy). The full table is on the PTA website.

Identifying a type

ATIP (version 6) works through six steps:

1. Label the conclusion, premise, subjects and predicates in the text, along with markers such as “because” or “I think”.
2. Reformulate the argument as “conclusion because premise”. Resolve pronouns, add what is implied, and reword so that shared terms are literally the same.
3. Determine the form.
4. Determine the substance.
5. Choose the lever, from the options the form and substance allow.
6. Name the type.

The second step needs the most judgement. Whether two phrases count as “the same subject” is an interpretive decision, and the form depends on it.

In Quell

Quell uses the PTA to describe the structure of the arguments in its texts. The plan is for players to build arguments out of statements they have read, and for the game to recognize what kind of argument they have made. See Argument maps and Debate.

In the lab’s argument-map editor, steps 3 and 4 are mechanical. Once each statement has been reformulated and analysed, the editor derives the form and substance by rule, and the annotator chooses only the lever. The same rules would let the game type a player’s argument without a language model, provided the player builds it from statements that have already been analysed.¹

Whether 50 types is too many to teach in a game is one of the open questions about debate.

See also

• Inference Anchoring Theory and xAIF — the framework for mapping whole texts, in which PTA types label individual inferences.
• CAPNA — judging whether an argument is good, not just what type it is.
• Argument maps — how Quell’s maps combine IAT and the PTA.

References

• Wagemans, J. H. M. (2026). Argument Type Identification Procedure (ATIP), Version 6. Aligned with PTA 3.1.
• Hinton, M., & Wagemans, J. H. M. (2021). Evaluating reasoning in natural arguments: A procedural approach. Argumentation. doi:10.1007/s10503-021-09555-1.

External links

• The Periodic Table of Arguments — the official site.
• PTA 3.1, clickable — worked examples of all 50 types.
• ATIP — the identification procedure.

What links here: Argument maps, CAPNA, Constrained decoding, Cross-encoders, Dependency parsing, Inference Anchoring Theory and xAIF, Natural language inference, oAMF, Glossary, xAIF editor