Do Kinds Exist?
Genre theory is a two-thousand-year argument about whether literary kinds are real things, useful conventions or convenient fictions.
The classical divisions
Plato, in Republic III, divides by manner of speaking: pure narrative, where the poet speaks in their own voice; mimesis, where the poet speaks as the characters, as in tragedy and comedy; and the mixed mode of epic, which does both. Three modes, and the division is about the position of the speaking voice rather than about subject matter.
Aristotle's Poetics cuts by means (rhythm, language, harmony), objects (better or worse people) and manner (narrated or enacted). The familiar triad of lyric, epic and dramatic is not in fact classical: it is assembled much later, by the German Romantics and Hegel, and given a philosophical scaffolding — subjective, objective and their synthesis — that its classical sources never had.
The nominalist objection
Benedetto Croce made the sharpest attack: only individual works exist. Genres are abstractions imposed afterwards, and no work was ever improved by conforming to one. Every real work exceeds its category, and when we praise a work we praise what is singular in it.
The objection is strong and yet incomplete, because genre demonstrably does something. A reader who begins a detective novel expects a solution; the same reader beginning an elegy expects grief and consolation. Those expectations are not imposed by critics after the fact, they are the conditions under which reading proceeds at all.
Family resemblance
Alastair Fowler's Kinds of Literature (1982) resolves this with Wittgenstein. Genres are not classes defined by necessary and sufficient conditions, but types held together by family resemblance: a network of overlapping similarities, where every member shares many features with others but no single feature is shared by all.
This is testable, and the result is often surprising. Take five works of a genre and six candidate features; you may well find each work carrying four of the six and yet no feature present in every one. The count of features common to all members can be zero, while the family is perfectly real. Genres are also historical: they are born, change and die, and a genre is better described by its exemplars than by its rules.
Genre as a contract
Todorov distinguishes historical genres, those a period actually recognised and named, from theoretical ones a critic constructs. Jauss makes genre a horizon of expectations the work confirms or disappoints. Derrida, in "The Law of Genre", observes that a text always participates in genres without belonging to one, since the mark of genre is itself never simply inside the work.
Bakhtin's addition is the most consequential: the novel is the genre that will not sit in the system. It is the only major genre still developing, it absorbs and parodies all the others, it works in the unfinished present rather than the closed past of epic, and it is inherently polyglossic: many voices, none of them final. A system of genres built for antiquity necessarily fails on it.
Common pitfall: treating genre as a set of rules a work either obeys or breaks. Almost every work of consequence is a modification of its kind, and the modification is legible only against the expectation, so a work "breaking" a genre depends on that genre for its meaning, and is not outside it.