Analogical Leveling in Paradigms

Sound change often makes one paradigm uneven; analogy can make it more regular again. Speakers reshape a form after other forms they perceive as related, and analogical leveling is the special case in which competing stem shapes inside one paradigm are reduced. Classical Latin preserves both the results of old leveling and paradigms that were never fully leveled.

Analogy is a historical explanation, not a productive permission to “fix” a form. Unlabelled forms below belong to neutral Classical usage. Conservative or early alternatives are labelled (archaic), and historically technical claims are (academic).

Sound change creates alternation; analogy reorganizes it

Rhotacism changed older s to r between vowels. It therefore produced a paradigm like nominative honōs beside genitive honōris: final s remained in the nominative, while the old stem consonant stood between vowels in the oblique forms and became r. Later, speakers extended the oblique stem honōr- into the nominative, creating honor. The final vowel was shortened before final r, so the outcome is honor, not honōr.

Both nominatives occur in Classical authors. Honōs is conservative—Cicero strongly preferred it—while honor gained ground and became the ordinary later form. This coexistence matters: analogy is gradual social change, not an instantaneous mechanical rule.

A proportion, not a new ending

(academic) Historical grammarians often represent analogy as a proportion: if most cells use honōr-, the isolated nominative is remodeled to match. The resulting -r is not a nominative ending. The inflectional relation remains nominative honor, genitive honōris, accusative honōrem, and so forth.

Fronting honōrem changes emphasis, not case. An English-speaking reader who sees the first noun and calls it the subject has ignored accusative -em. The analogy that created honor explains the stem, but the visible ending still performs the parsing work.

The same type of leveling produced nominative arbor beside older arbōs “tree.” Arbōs is (archaic) in a Classical prose context; arbor, arboris is the practical paradigm to learn.

It would be wrong to infer that every nominative in historical -s became -r. Analogy spreads form by form and lexeme by lexeme. It can stop before an entire class has been remodeled.

Leveling is never guaranteed

Many inherited alternations survived because speakers continued to learn them as stable parts of common words. Neuter corpus, corporis keeps s in nominative and accusative singular but r in the oblique stem. Masculine cinis, cineris “ash” likewise preserves different visible shapes. Latin did not generalize corpor or ciner as their Classical nominatives.

Why one word levels and another resists is not always reducible to a neat rule. Frequency, inflectional class, the transparency of an ending, and competition between variants all matter, and the surviving evidence is incomplete. This is a genuine limit: memorize the attested dictionary pair rather than inventing a historical prediction.

Analogical endings: a related but distinct development

Analogy can also spread an ending from one inflectional class into another. This is often called analogical extension rather than strict stem leveling, because the ending—not an alternate stem—is being copied. The distinction helps keep the analysis honest.

Third-declension i-stems inherited accusative plural -īs. During the Classical period, -ēs, the form familiar from consonant stems and identical with the nominative plural, increasingly competed with it. Both multōs cīvīs and multōs cīvēs can mean “many citizens” as an accusative plural. The adjective multōs removes the case ambiguity.

The ending -īs is especially characteristic of Republican and Augustan usage and is common in verse as well as prose; -ēs becomes increasingly general. Neither form changes the lexical stem cīv-. Full distributions and noun-class diagnostics belong on the third-declension i-stems page.

The genitive plural cīvium remains a useful sign of the i-stem class. One analogical ending does not necessarily replace every distinctive form of the paradigm.

Latin and English regularization

English analogy works in the same way. A child who says goed has extended the productive pattern walked, played, opened to an irregular verb. That individual form is nonstandard, but many changes of that kind have become established over the history of English. Latin honor is not a learner's mistake: it is the successful conventional outcome of a community-wide remodeling.

This comparison also shows why “more regular” does not mean “newer in every context.” Conservative honōs and innovative honor overlapped. The task of a Classical reader is to recognize both where the evidence supports both, while producing the conventional form appropriate to the chosen author and period.

A reliable diagnostic

When a paradigm shows competing shapes:

1. Establish the attested nominative, genitive, and relevant variant from a dictionary.
2. Separate phonetic change from analogical remodeling.
3. Decide whether the stem was leveled (honōs → honor) or an ending was extended (cīvīs → cīvēs).
4. Label the chronology and register of a variant.
5. Parse the actual ending before using word order or translation.

Here nominative honor is the subject, dative ducī is the recipient, and ablative cīvibus follows ā as the personal agent. English order happens to place “an honor” after the verb, but Latin morphology has already established every relation.

Common Mistakes

1. Calling analogy a sound law

Rhotacism created the oblique r; analogy extended it into the nominative.

2. Leveling a paradigm that Latin did not level

3. Rejecting an attested analogical variant

4. Letting English order override the ending

Key Takeaways

Analogical leveling reduces alternation by extending one stem through a paradigm; analogical extension can instead spread an ending between classes. Both are historical, lexeme-by-lexeme developments. Recognize established alternatives such as honōs/honor and cīvīs/cīvēs, but never regularize an unattested form on your own.