To remain coherent with these premises, the theory should make it possible to maintain a clear distinction between diegetic and narrative: diegesis, in the sense of a reality established by inference starting from the data of perception, should remain separate from narration, or how such reality is presented on screen. Moreover, the objective contents of the narration (what is represented, the signified) should maintain its primacy with respect to the modalities of the secondary and subjective narration (representation, the signifier). This distinction, deriving directly from the propositional conception of film language, connects up with a topological interpretation of the diegetic/nondiegetic relationship, which presupposes a juxtaposition of specific places (Ricoeur [1984] 1985: 179). A coherent theoretical account of the concept should clearly maintain the separation between the two dislocations. In addition, the topology of every possible film sequence with respect to diegesis should tend to be definable in only one manner: if diegesis is an objective moment or specific place in the reality of the film, the diegetic aspects must on principle remain distinguishable from the nondiegetic aspects. Each aspect of each single narrative sequence can thus find itself, at a certain moment, on only one side of the divide separating two realities which are complementary but ontologically distinct. There cannot be moments of indecision, but only moments of discontinuity or shift. Each single element can of course change its position with respect to diegesis, and may indeed follow quite complex trajectories; but the instant of this shift must always be identifiable or at least imaginable.

These implications highlight a series of problems which are difficult to resolve. In terms of the diegetic/narrative distinction, it is impossible to be coherent in maintaining it as a distinction between objective contents and subjective modalities of the narration: in the case of fiction everything clearly points to the fact that they coincide. Even in a film which is as classically historical and narrative as Stanley Kubrick’s Barry Lyndon (1975), it is clear that the story being narrated can only be a function of the narration, or rather of one of the narrations: whether the first person narrative of the protagonist of Thackeray’s novel on which the film is based, the third person narrative of the unseen narrator in the film, or again the narrative confronting us on screen, comprising images, dialogue and music. Nothing points to an objective entity which exists somewhere outside the narration which could be referred to as ‘the world of Barry Lyndon’ or ‘the story of Barry Lyndon’. All we have as a primary feature is the narration, while the narrated contents – which are in fact the product of inference – prove to be quite secondary, as well as varying from one viewing to the next. Any claim to the contrary would mean recognising the possibility of an objective (objectively described) contents with respect to which every possible human narration would only be one of the many subjective modes of narrating the reality. This is clearly untenable: the narrated contents do not exist except in their own peculiar form, as they are mediated by narrative viewpoints. This viewpoint is not an accidental reality of the contents of the narration; on the contrary, there can be no narration without a viewpoint. Thus if narration is a subjective construction, diegesis is no less subjective; indeed it is all the more so, being the result of a (subjective) act of inference based on a construction which has itself been mediated from a subjective viewpoint.

Even maintaining the diegetic/nondiegetic distinction as a topological one is not without its difficulties. In the first place there are plenty of cases when the distinction cannot be made with respect to the world of the narrated story, above all when it is a question of the audio dimension and of diegetic or nondiegetic music. This is true in part because music does not easily lend itself to the propositional conception of film language. Music neither offers nor represents images or models of situations; signified and signifier tend to overlap, when they are not actually indistinguishable. Thus the attempt to distinguish between diegetic and nondiegetic music comes down to identifying the source and accounting for its dislocation. This is not without interest, but it does not throw light on the (contextual) sense of music in the audiovisual structure. There are other difficulties too. As has frequently been pointed out even by those who uphold the objectivity of diegesis (Gorbman 1980; Percheron 1980; Bordwell 1985; Johnson 1989; Brown 1994), the function of music is broadly independent of its topology. Whether it is diegetic or not, this function, and the very meaning of music, depend above all on its ability to introduce itself into the audiovisual structure at the narrative level. The music can fulfil varying functions according to whether we are considering the perspective of the spectator or of the various characters (so far as these can be inferred), each of whom represents a different viewpoint (Norden 2007). The inescapable incoherence of the theory of diegesis is seen above all in the fact that scholars so often relegate music to ambiguous zones of the overall topology. The observation that music enjoys great liberty in shifting with respect to diegesis (Gorbman 1980: 196) should not lead to the substantial difficulties that force analysis to postulate broad zones where topology is either uncertain or cannot be decided. By definition, analysis is based on the theoretical reduction of complexity; if the decision concerning the topology of music were merely the result of analytical observation, it would always be possible to describe the shift without any difficulty in terms of diegesis. The difficulty cannot be attributed to the peculiarities of a specific situation on screen: such cases are so common and significant that we have to ask ourselves about the adequacy of the basic concept. The proposal to conserve the diegetic/nondiegetic divide while at the same time making it more fluid, resorting to the theoretical expedient of describing the shift from one state to another as a process (Stilwell 2007: 184-185), appears to elude a genuine conceptualisation of diegesis, focusing instead on the question of continuity over time. For the purposes of analysis, the discrete conception of temporality is abandoned, while in fact this is fundamental to the concept of diegesis. It implies a world which is coherent with logical and spatiotemporal parameters, involving an articulation based on precise events; moreover, the audiovisual structure always enables a discrete analytical observation.

In general, zones of topological indeterminacy – which may even require the introduction of such a term as “transdiegetic” (Taylor 2007) – cannot readily be reconciled with the representative perspective presupposed by the concept of diegesis. In ontological terms this perspective separates the act of representation (signifier) from what is represented (signified), enabling a clear distinction between diegetic and nondiegetic. Far from dialectically reinforcing the difference between two places – as the emphasis on crossing the border does (Stilwell 2007: 184) – the postulation of an uncertain topology calls into question the utility of the distinction between positions which are antithetical by definition, even if they are complementary. On the contrary, the attempt to relativise the analytical centrality of the diegetic/nondiegetic distinction by introducing it into a complex network of conceptual oppositions or relations does have at least one merit. It reinterprets situations on screen which are often considered topologically ambiguous as the outcome of strategies based on the possibility of freely articulating in narrative terms the relationship between audio elements whose position with respect to diegesis is perfectly clear (Smith 2009). In this sense the relativisation of the distinction leads automatically to a more scrupulous use.