Stem Alternations before Endings

A Latin stem is not always a block that remains visibly unchanged before every ending. Consonants may merge, vowels may weaken or disappear, and one lexeme may preserve more than one inherited stem shape. The result can look irregular until you separate three things: the lexical stem, the ending, and the change produced where they meet.

This page treats alternations that matter for reading Classical Latin. Unlabelled forms and examples are neutral Classical prose; older alternatives are marked (archaic) and verse-specific freedoms (literary). Macrons are pedagogical and show vowel quantity; they are not letters normally written in Roman texts.

One lexeme, more than one visible shape

An alternation is a systematic difference between related forms of one word. In rēx, rēgis “king,” the working stem rēg- is obvious in the genitive, but nominative rēx reflects the historical meeting of stem-final g and nominative -s. Latin writes the resulting sound as x. The visible nominative is therefore not built from a separate stem rēx- plus nothing.

The same boundary process appears with dux, ducis “leader”: nominative dux corresponds to stem duc-. Both noun forms have short u; the related verb dūcō has long ū. Recovering the stem consonant therefore does not authorize importing vowel quantity from a related lemma.

These rows do not constitute one rule. The two x nouns show a recoverable boundary pattern; mīles and nōmen preserve lexical stem alternations; corpus/corporis reflects historical rhotacism. The practical generalization is narrower but powerful: for a third-declension noun, the genitive normally gives better access to the oblique stem than the nominative does.

In each pair, the English noun stays unchanged apart from the possessive 's. Latin instead exposes a different part of the paradigm. The genitive is not merely a translation of “of”; it is also evidence for the stem.

Changes conditioned by a following form

Sometimes the neighboring ending or auxiliary determines which allomorph appears. The clearest high-frequency example is possum “I can,” historically built from pot- plus sum. Before a form beginning with s, t + s assimilates to ss: possum, possumus, possunt. Elsewhere the t remains: potes, potest, potestis. For Classical production, this is a fixed paradigm, not a live instruction to assimilate any t before any s.

English uses the invariant modal can except for past could. Latin distributes the same lexical meaning across forms whose internal consonant changes. Person and number must be read from the complete form, not from the first three letters.

The verb volō “want” preserves a less transparent set: volō, vīs, vult, volumus, vultis, volunt. The distributions of vol-, vī-, and vul- are inherited and lexical. They are reliable inside this paradigm, but not productive patterns that can be transferred to another verb.

Here velint is the present subjunctive, another member of the vel-/vol- set. There is no honest shortcut that derives the entire paradigm from volō by one Classical sound rule: these common forms must be learned together.

Alternation by grammatical stem

Not every different stem is caused by the immediately following ending. Latin verbs regularly select a present stem, perfect stem, and supine or participial stem according to the grammatical formation. In agō, agere, ēgī, āctum, the contrasts ag- : ēg- : āct- signal which system is being built. Once the correct stem has been selected, ordinary endings attach to it.

The time expressions support the parsing, but morphology comes first: agimus contains the present-system stem and ēgimus the perfect stem; scrībit has scrīb-, whereas scrīpsit has perfect scrīps-. English also has sing/sang/sung, but Latin principal parts feed many more forms. Review their predictive role on the principal-parts page.

A parsing procedure that works

When a form seems not to match its headword, use this order:

1. Identify a plausible ending and its feature bundle.
2. Recover the remaining stem, allowing only alternations supported by the dictionary entry or a known paradigm.
3. Check agreement, government, and the finite verb before assigning functions.
4. Use word order for focus and emphasis, not as an English-style subject–object template.

The tempting English-order reading makes “the commander” the subject. Morphology blocks it: ducem is accusative singular from dux, ducis, while mīlitēs can be nominative plural and agrees with plural sequuntur. Recovering duc- explains the word; the endings establish the syntax.

Productive Classical parsing does not mean freely generating historical reconstructions. A poet may use a contracted form (literary), and an early text may preserve a form absent from standard prose (archaic). Such evidence belongs to the history of a paradigm, but it does not license a learner to replace the attested Classical form. The pages on syncope and contraction and rhotacism separate those developments in detail.

Common Mistakes

1. Treating the nominative as an unchanged stem

The genitive shows rēg-; nominative x reflects g + s.

2. Extending a real alternation to the wrong lexeme

The pattern x can conceal either c + s or g + s; only the dictionary genitive tells you which consonant belongs to this noun.

3. Regularizing a lexical allomorph

The form vīs must be learned as part of the inherited volō paradigm.

4. Reading the first noun as subject

Stem recovery identifies ducem as a form of dux; case and agreement, not position, determine its role.

Key Takeaways

Apparent irregularity often records a regular boundary change, an inherited alternate stem, or the selection of a different principal-part stem. Use the genitive for nominal stems and principal parts for verbal stems, distinguish productive parsing patterns from lexical facts, and let endings establish syntax before English word order influences the translation.