The accusative form in Quenya is used for the object of verbs and is unmarked in “modern” Quenya (Tarquesta). According to the Plotz letter, there was a distinct accusative form in Classical Quenya (Parmaquesta), marked with a long final vowel for singular vocalic nouns, and by the use of an -i plural for vocalic nouns in the plural:

• Singular accusative: ciryā, lassē (vs. nominative cirya, lasse)
• Plural accusative: ciryai (vs. nominative ciryar)

We don’t know how (or whether) the accusative was marked for dual or consonantal nouns (but see below).

Presumably the accusative form was also used for the object of prepositions, as it is with English object pronouns: “I saw him, I gave it to him”. It is hard to determine for certain, however, since by Tarquesta there was no distinction between the nominative and the accusative. There was a Common Eldarin (CE) objective form, which in one place Tolkien said was used for both verbal objects and the object of prepositions (PE21/67), so I think it is likely the Quenya accusative was used the same way. This is largely a moot question in Tarquesta, but is relevant if you are trying to imitate Parmaquesta grammar.

Origins of the accusative: The Common Eldarin objective mentioned above originally had an -a/-d inflection used for indirect objects (dative) and allatives (PE21/75). But this inflection gradually came to be used for direct objects in some cases:

So that the objective inflexions derived from -a, d might become merely “accusative” signs, and the “dative” require some new type of expression, or new suffix. This was the case in Q. and probably in prehistoric Beleriandic (PE21/76).

This intrusion of the originally dative/allative suffix happened principally with the addition of the suffix -a to vocalic nouns used as direct objects:

An early development was to specialize forms made with a-suffix as direct object. In that case nouns with long vocalic ending became trimoric or over-long in the final: Ulmõ (objective). The -a was, however, seldom transferred to the direct object of basic nouns (never in dissyllables such as atar) (PE21/76-77).

Based on this, it seems likely that Ancient Quenya direct-object (accusative) vocalic nouns had, after the addition of the a-suffix, over-long final vowels as noted above: nominative ✶kiryā, accusative ✶kiryã. After the later shortening of final vowels, the result would have been PQ nominative cirya, accusative ciryā, as described in Plotz (hat tip to Lokyt for presenting this theory to me).

As for consonantal nouns, I see two possibilities. First, for consonantal nouns there was a distinct CE subjective case, marked by a long vowel in the final syllable, ✶atār, nēr (vs. object forms ✶atar, ner). Phonetic pressures probably forced disyllabic forms to lose the long subjective vowel in OQ or early PQ (✶atār > PQ. atar), but the distinct subject survived into TQ for monosyllables, with the uninflected/objective form being reinterpreted as the stem: nér (ner-). The uninflected short-vowel form might have survived for time in PQ as the accusative form of monosyllables with long vowel. This paradigm was suggested to me by Lokyt on the Aglardh forums, in December 2019.

Alternately, there were primitive nouns in CE ending in a short vowel, later lost, such as ✶nată > Q. nat “thing” (VT49/30). We know that the CE suffix -a could be added to such short-vowel nouns as well:

The limitation in use was not rigid. -ā̆ forms were also made from vocalic nouns, especially from those with original short vowels, as kantă, kantā (PE21/26).

For these nouns, the short ă would be lost for the subject, but the long ā might have survived to become an accusative suffix for objects: ✶nată, natā > PQ. nat, nata. This would match the consonantal accusative paradigm Tolkien used in the late 1920s through early 1930s (see below).

Conceptual Development: In the Early Qenya Grammar of the 1920s, the accusative form was marked with a -t in the singular and -nt [-(l)int] in the plural (PE14/43, 73), though in the manuscript version Tolkien said the -nt marker was poetic, and the normal accusative plural marker was -n: -(l)in (PE14/43). In a set of Qenya Declensions charts from later in the 1920s, Tolkien decided that the accusative and base forms of the noun were the same, so that the accusative was unmarked with the exception of singular consonantal nouns where it was marked by -a (PE16/111-115).

Tolkien retained this system in the more lengthy Declension of Nouns from the early 1930s (PE21/1-41), with the caveat that vocalic declensions ending in -i or -u also added -a in the accusative (PE21/14-15). In later declension charts signs of the system described in Plotz began to emerge, with the retention of i-plurals in the accusative (PE21/50, 53), and long -í in the accusative plural of e-nouns (PE21/53). Tolkien’s notes and use of parenthesis indicate these distinct accusative forms were archaic, just as they were in Plotz.

Neo-Quenya: For purposes of Neo-Quenya, there is rarely a reason to use the accusative forms unless you are trying to construct a Parmaquesta text. If so, I’d use the Plotz accusative forms for vocalic nouns. For consonantal monosyllables, I would add -a if the noun has a short vowel, but shorten the vowel if the vowel is long: nominative nat, accusative nata; nominative nér, accusative ner. I’d assume disyllabic and longer consonantal nouns did not have a distinct accusative form, even in PQ, which aided in the eventual loss of the accusative in TQ.