In lieu of an abstract, here is a brief excerpt of the content:

  • Tennyson's Phantom Ballads
  • Ewan Jones (bio)

In Surfing Sncertainty, the philosopher Andy Clark puts forward a bold speculative theory of mental operation, based on the computational principle of predictive processing. The idea has roots in the nineteenth century: the German psychophysiologist Hermann von Helmholtz was the first methodically to postulate a notion of unconscious inference (unbewusster Schluss), according to which subjects actively model a sensory world to which they possess necessarily limited access. Subjects, in this model, do not passively absorb sense data, but continually and unconsciously test their encounters with the world against working predictions: when a mismatch occurs, error correction sets in. This process has underpinned recent developments in neural networks, which—unlike earlier computational models that were trained, by humans, to achieve prespecified tasks—themselves learn and deploy probabilistic learning models.1

Clark's effort to synthesize recent work in computation, robotics, and linguistics intervenes in debates that have deep philosophical roots. Predictive processing is opposed to any reductionist or behavioralist account that sees human interaction as a merely passive response to stimuli; rather, subjects actively produce the sensory flows to which they are subject—making perception, in Clark's memorable phrasing, a continuous exercise in "controlled hallucination" (p. 14). This self-actuating nature of cognition might smack of the more absolute forms of idealism, associated, for instance, with Fichte. Yet just as Helmholtz's physiological research represented a profoundly modified form of Kantianism, so Clark's endeavor differs from such precursors through its robust anti-representationalism. The mental models through which subjects encounter the world are not static or fixed categories. Rather, they are both produced and continually updated by cognitive and sensorimotor activity. Concepts do not merely subsume sensuous particulars; both are rather locked into a process of reciprocal generation, or "bootstrapping."

While clear difficulties remain as to how to move from such sensory processing to higher-order mental contents such as desire or memory, I find Clark's speculative proposal terrifically exciting, in large part because it squares with my intuitive sense of how cognition operates as a dynamic interaction between top-down concepts and bottom-up sensory input. It provides a technical vocabulary for describing what happens to our minds and bodies when, [End Page 201] having read only a very small string of language—say, four words: "Tyger, tyger, burning bright"—we immediately begin to model rhythmical patterns that do not only anticipate but also actively produce the subsequent information that we encounter. Unconscious inference resists the bad choice between an internalist phenomenology, according to which such experiences would be merely private, and an extrinsic account that would take mode or genre only as social givens. In fact, modes and genres are simply embodied predictions that have a tendency to forget the fact.

Clark suggests such aesthetic applications only in passing, as when he remarks on the human tendency to "hear" a predicted yet absent acoustic stimulus, a phenomenon "every bit as perceptually striking and salient as the inclusion of an unexpected note" (p. 45). I want to develop such ideas in the form of one specific material object, in line with the focus of this special issue: the object in question being Alfred Tennyson's Poems in two volumes (1842). In the analysis that follows, I build up from the local instance of an unheard note to reflect more broadly upon the notion of ballad—taking the latter to mean, in this case, not a stable literary category, but rather an unconscious inference that subjects come to make, by way of their minds and bodies. Such large questions prove inseparable from the evaluative and moral dimensions of Tennyson's composite work.

I focus on the 1842 Poems as a whole, because—to continue to extend the form of thinking that I briefly sketched above—it represents an ideal training-set for unconscious inferences to emerge and to be deployed. It does so not only to the extent that a collection contains more information than a single lyric poem: any long poem would accomplish this goal equally well. Rather, I am interested by the way in which the collection offers a number of discrete iterated works, which we can successively combine...