Why has the neuter gender disappeared from almost all the modern Romance languages? It was completely common in Latin.
And when exactly did this happen? Did it happen in Latin itself, or only after Latin split into multiple languages and the phenomenon occurred in each of the languages separately and independently?
I've read that even in Latin, we see some variability in the declension of words as neuter or masculine. Sometimes the use of the masculine where neuter would be expected is attributed to "personification". So the fall of the neuter seems to have been a long and at least somewhat gradual process. I will update this post if I find more detailed information.
There are some reasons to think of the neuter gender in Latin as being less robustly distinguished formally and semantically than the masculine and feminine genders, although this might just be hindsight.
I would guess that form was more important than semantics, since in some Germanic languages we can see a different tendency to go from three to two genders by merging the former masculine and feminine into a "common" gender and maintaining a distinction between "common" and neuter gender.
As Arnaud Fournet said, neuter gender nouns in Latin were declined the same way as masculine nouns in most grammatical cases (genitive, dative, ablative). The neuter-masculine distinction was found in the following areas:
In the nominative and accusative plural, neuters had a characteristic -a ending. Because -a was also used as a feminine singular ending, Latin neuter plurals in -a could end up as feminine nouns in Romance languages, as mentioned in Fournet's answer. I know of two ways this happened.
As Fournet said, neuter plural nouns ending in -a could be reinterpreted as feminine singular nouns.
Alternatively, -a survives as a plural ending on nouns in some Romance languages, such as Italian; the examples given on this page are il dito, il uovo, il lenzuolo with the plurals le dita, le uova, le lenzuola. You can see that plural -a nouns in Italian take feminine plural agreement.
In the nominative singular, masculine nouns mostly ended in -s while neuter nouns mostly did not. In most Romance languages, Latin nominative singular forms were either mostly lost or heavily remodeled in a way that caused the distinction between masculine and neuter nominative singular nouns to become less relevant. Sound changes like the loss of word-final -m, or often of -s, and the addition or loss of word-final -e after consonants played a role here.
In the accusative singular, masculine and neuter forms were identical already in Classical Latin in the second declension, both ending in -um. In the third declension, some masculine and neuter accusative singular forms differed only in the presence of word-final -m, such as third-declension adjectives ending in -alem m. vs. -ale n. With the loss of word-final -m, those forms would have merged as -ale.
Certain Romance languages have a distinction for certain words between a masculine singular ending -u and what is called a "neuter" singular ending in -o; I think it's worth mentioning here that my understanding is that this is for the most part not not a survival of the Latin masculine/neuter distinction, but an innovation.*
The masculine and feminine genders in Latin both had clear semantic "cores" of male and female animates, respectively. Almost all nouns denoting male animates were masculine, almost all nouns denoting female animates were feminine. Although the masculine and feminine genders also contained many nouns that did not belong to these semantic categories, the association was productive and has been maintained to the present day in Romance languages with gender (we can see the continued relevance of semantics to gender assignation most clearly when dealing with gender agreement for proper names or pronouns that refer to human beings).
The neuter gender, in contrast, had no large semantic domain that was exclusively associated with it. It included some, but by no means all, nouns that were neither semantically male nor female. The lack of a clear distinction in meaning between masculine and neuter inanimate nouns may have contributed to the eventual loss of the distinction in form.
*In the Romance languages with an -u vs. -o distinction that I mentioned above, words fall into one category or the other not on the basis of their gender in Latin, but rather based on semantic criteria, like being a mass noun vs. a count noun.
An overview of this is provided in "The Romance Collective Neuter and the Survival of the Latin Ablative" by Gordon M. Messing (Glotta, vol. 50, no. 3/4, Vandenhoeck & Ruprecht (GmbH & Co. KG), 1972, pp. 255–62, https://www.jstor.org/stable/40266241)
Multiple contradicting origins have been proposed for the -o ending; as the title indicates, Messing thinks it comes from an ablative, modifying a suggestion by Hall 1968.
Neuter in Latin differed from Masculine only for Nominative and Accusative cases. When case-endings began to collapse in early Proto-Romance, Neuter singular was reassigned to masculine and some Neuter plural ending in -a became feminine singular.
