Models of information processing agree about the assumption that separable processing stages or levels perform distinct subfunctions within the overall performance of the task, but the organization of information flow between these stages or levels is a matter of controversy. Discrete models assume that one stage transmits its output in a single distinct message to the subsequent stage. Therefore response time (RT) is the sum of the durations of all individual stages involved in a given task because no temporal overlap exists between stages. According to continuous models, however, levels transmit partial outputs such that subsequent levels begin processing as soon as the first “bits of information” are available. A subsequent level can start working while a former level continues to process information and to transmit output. Thus, different levels work in parallel. Between the extremes of fully discrete and fully continuous models of information processing, various intermediate models have been proposed in which information is transmitted from one stage to the next in a series of steps. For example, Miller (1982) proposed an intermediate model called asynchronuos discrete coding (ADC), in which separable stimulus attributes such as shape, color, and size are transmitted separately as soon as they are available. In this model, subsequent stages operate with some temporal overlap.

Many experiments have examined whether information transmission is fully discrete or fully continuous, or whether some intermediate model might hold. Most of them have investigated whether response preparation begins before the perceptual analysis of the input is complete. Preliminary response preparation would strongly contradict discrete models of information processing but would be quite consistent with both continuous and intermediate models. In fact, a number of experiments suggest that under certain circumstances response preparation can overlap with simple perceptual discrimination. Other studies, however, give evidence that sometimes response preparation occurs only after the stimulus has been fully processed.

Yet even those studies which provide evidence for preliminary response preparation do not allow for the general conclusion that continuous information transmission holds between any set of successive processing stages. Temporal overlap has been demonstrated for one particular pair of processes (simple perceptual analysis and response preparation), but that finding does not justify the conclusion that transmission of information from anyone process to the next is the same for all pairs of processes.

Interestingly, with respect to mental rotation the widely accepted theoretical assumption rests on a discrete serial model of information processing. The time in which to decide on the parity of a character increases monotonically with the angular disparity from the vertical. This finding is usually taken as evidence that the system rotates the mental representation analogously into the upright position. Theories of mental imagery have attempted to dissect this complex process into a set of functionally distinct processing stages, for example, recognition of the character, identification of its orientation, the mental rotation itself, judgment of parity, and response selection. These stages are assumed to be organized sequentially with discrete transmission of information from one stage to the next.

Other properties of the mental rotation process also make it an interesting candidate for study in the context of information transmission theories. Mental rotation is a clear-cut example of a dedicated process. For example, the effects of angular disparity and size ratio between the stimuli to be compared are additive; this suggests distinct mechanisms for mental rotation and for mental size scaling . Evidence from distinct sources suggests that the process of mental rotation might be implemented within the multimodal projection areas of the parietal cortex. In addition, mental rotation is a “central” process because it occurs at a level beyond that of the independent analysis of (for example) luminance, color, binocular disparity, or motion.

Finally, several recent studies using event-related brain potentials (ERPs) obtained fairly specific results: Presentation of a character evoked a pronounced positive deflection with a latency of about 300–600 ms (P300). The amplitude varies dependending on the amount of mental rotation actually executed by the subject: The more the character was rotated, the less positive the potential. This finding is interpreted as due to the superimposition of two brain potentials, a P300 complex with a constant amplitude in all conditions, and a parietal negativity reflecting the amount of mental rotation proper.
If mental rotation indeed is performed by a dedicated module that does little else, we must ask whether response preparation can overlap with mental rotation. Moreover, because mental rotation is a relatively slow process whose duration can be manipulated over a range of several hundred milliseconds by varying the orientation of the character, there is a sufficiently long interval, in principle, to permit preliminary response preparation to occur. If the character is identified before mental rotation begins (as suggested by imagery theorists such as Corballis, 1986; Shepard and Cooper, 1982) and if the identity of the character determines which hand should respond while parity determines whether the response is to be made (“go”) or withheld (“nogo”), then preparation of the responding hand may already have taken place before mental rotation is finished.