In recent theories of narrative, there is a controversy about the validity of the "person" distinction in the construction of typologies of narrators. It seems to me that this controversy is rooted in different approaches to the act of narration: the mimetic versus the nonmimetic (or what I call, for lack of a better term, narrative) approaches. Those theorists who have a structural model of narrative—the Genettian model1 (Genette, Bal, Rimmon-Kenan)—do away with the traditional typology, which distinguishes between narrators according to the criterion of person, that is, first- as opposed to third-person narrators. Conversely, the theorists who continue to hold onto a mimetic approach (Stanzel, Cohn) maintain the validity of this criterion.
My purpose is to explain the mimetic and narrative biases, respectively, [End Page 523] in the Cohn/Stanzel and the Genettian theories.2 I will claim that, on the one hand, the mimetic bias creates contradictions for the Cohn/Stanzel model, whereas, on the other, despite the narrative (non-mimetic) orientation of the Genettian model, the force of mimesis is so compelling that this model is, unwittingly, subverted by mimetic considerations, which create contradictions within it. In effect it seems that no theory of narrative is able to override completely the mimetic language game (a term I borrowed from Ron, henceforth MLG). The metalanguage, like many of the texts that are the objects of its study, takes part in MLG.
Before launching into my discussion, I would like to explain briefly my use of the term mimetic, because it has such a range of meanings. I use Ron's formulation:
Literary mimesis does not aim at truth, . . . it aims at conveying an impression . . . that it is a semblance of true discourse. A mimetic text is, in a broad sense, like a set of instructions for constructing a fictional world. This "world" would consist of representations not essentially different from those a reader may make himself of the "real" world in any major respect other than his being able to characterize them as fictional. Let us call this practice the Mimetic Language Game.(18)
In the two poetics I have chosen to discuss, the mimetic and non-mimetic positions are most evident in the controversy over the criteria for a typology of narrators.
Both Cohn and Stanzel argue for the traditional typology. Stanzel's typology of narrative situations is flexible and has gradations. Cohn basically accepts Stanzel's position but, contrary to him, claims that
no text can be placed on the boundary separating first- and third-person narration [on the figural-first person side]: for the simple reason that the grammatical difference pertaining between persons is not relative but absolute. . . . The boundary between persons . . . is both real and absolute: no gradation is possible between "I" and "he." One can therefore not conceive of first- and third-person pronouns being applied to the same character without at the same time attributing the voices that utter the different pronominal references to different speakers.
The Genettian model, on the other hand, claims that the person distinction is not valid because it is based on two different criteria. The person criterion stresses
variation in the element of the narrative situation that is in fact invariant—to wit, the presence (explicit or implicit) of the "person" of the narrator. This presence is invariant because the narrator can be in his narrative (like every subject of an enunciating in his enunciated statement) only in the "first person. . . ."
Or stated simply, "I" refers to the narrator, but "he" to the character. A narrator referring to himself, however, would use "I" even in so-called third-person narration. Conversely, in so-called first-person narration other characters would be referred to in the third person. [End Page 524]
These opposing positions result in two typologies: the Cohn/Stanzel circular typology of narrative situations,3 which is comprised of first-person, third-person authorial, and third-person figural narrative situations (with in-between gradations that are not pertinent to my discussion); and the Genettian model, which has four criteria for its typology: narrative level, degree of...