Textbook of Usability

Introduction

The preceding chapters described the perceptual, cognitive, and motor systems in isolation. The Model Human Processor (MHP), developed by Stuart Card, Thomas Moran, and Allen Newell in their landmark 1983 book The Psychology of Human-Computer Interaction Card, 1983, integrates these systems into a single engineering model that can predict human performance on interactive tasks Card, 1986. The MHP is not intended as a complete model of the human mind. It is an engineering approximation — precise enough to make useful quantitative predictions, simple enough to be applied by designers without extensive psychology training John, 2003. It treats the human as an information-processing system with measurable parameters: processing rates, memory capacities, and decay times.

Architecture of the MHP

The MHP consists of three interacting subsystems, each with its own processor and memory.

The Perceptual System

The perceptual system receives sensory input and delivers it to working memory. It has its own buffer memories (iconic memory for vision, echoic memory for hearing) and a processing cycle time. Perceptual processor cycle time (τp): ~100 milliseconds (range: 50–200 ms) This is the time required for the perceptual system to process one "frame" of input. It sets the minimum time to detect a visual or auditory stimulus.

The Cognitive System

The cognitive system receives information from the perceptual system and working memory, applies stored knowledge from long-term memory, and generates decisions that are passed to the motor system. Cognitive processor cycle time (τc): ~70 milliseconds (range: 25–170 ms) Working memory capacity (μWM): approximately 4 ± 1 chunks for novel information Cowan, 2001. Card, Moran, and Newell's original 1983 figure was ~3 chunks (range 2.5–4.1) Card, 1983; Cowan's subsequent review converged on a slightly higher estimate, and this textbook uses the 4 ± 1 figure consistently. Working memory decay time (δWM): ~7 seconds for a single chunk (range: 5–226 s depending on context and number of chunks) The cognitive cycle time of ~70 ms represents the minimum time for a single cognitive operation — recognising a pattern, making a comparison, or selecting a response.

The Motor System

The motor system receives commands from the cognitive system and executes physical actions. Motor processor cycle time (τm): ~70 milliseconds (range: 30–100 ms) This is the time to initiate and execute a single motor action, such as a keystroke or a mouse click. Complex movements (reaching to a target, as modelled by Fitts's Law) involve multiple motor cycles.

Principles of Operation

Card, Moran, and Newell Card, 1983 identified several principles that govern how the MHP's subsystems interact.

The Recognize-Act Cycle

The cognitive processor operates in a recognize-act cycle: it recognises the current situation (by matching working memory contents against long-term memory patterns) and then selects an action. Each cycle takes approximately τc = 70 ms. Complex decisions require multiple cycles.

The Rationality Principle

"A person acts so as to attain their goals through rational action, given the structure of the task and their inputs of information and bounded by limitations on their knowledge and processing ability." This principle acknowledges that human behaviour is goal-directed and approximately rational, but constrained by the parameters of the MHP. Users do not make random errors; they make predictable errors that arise from specific bottlenecks in perception, memory, or processing.

The Variable Perceptual Processor Rate Principle

The perceptual processor speeds up when stimulus intensity increases and slows down when stimuli are faint or ambiguous. This is why bright, high-contrast displays produce faster reaction times than dim, low-contrast ones.

The Uncertainty Principle (Hick's Law)

Decision time is a function of the uncertainty in the choice. With n equally probable alternatives. Tchoice = τc × log2(n + 1) This is the MHP's formulation of Hick's Law, expressed directly in terms of cognitive cycle times.

Using the MHP for Prediction

The power of the MHP lies in its ability to generate quantitative predictions by tracing information through the three subsystems.

Simple Reaction Time

A simple reaction time task (press a button when a light appears) involves:

1. Perceptual processing: τp ≈ 100 ms
2. Cognitive processing (detect and decide): τc ≈ 70 ms
3. Motor execution: τm ≈ 70 ms Total: ~240 ms This prediction is consistent with empirical simple reaction time data, which typically ranges from 200–300 ms.

Choice Reaction Time

For a two-choice task (press left for stimulus A, right for stimulus B):

1. Perceptual processing: τp ≈ 100 ms
2. Cognitive processing: τc × log2(2 + 1) ≈ 70 × 1.58 ≈ 111 ms
3. Motor execution: τm ≈ 70 ms Total: ~281 ms

The Power Law of Practice

The MHP also accounts for the effect of practice. As users repeat a task, performance improves according to the power law of practice. Tn = T1 × n^(-α) where Tn is the time on the nth trial, T1 is the time on the first trial, and α is a learning rate parameter (typically 0.2–0.6). This law predicts the steep initial improvement and gradual plateau that characterise skill acquisition.

Strengths of the MHP

The MHP's primary strength is that it provides approximate quantitative predictions without requiring human subjects testing. A designer can estimate task times, compare alternative designs, and identify bottlenecks using nothing more than the MHP parameters and a task analysis. The model is also modular: the three subsystems can be analysed independently. If a design change affects only the perceptual demands (changing the size or contrast of a display element), only the perceptual subsystem's parameters are relevant. If the change affects decision complexity, only the cognitive subsystem's parameters matter. The MHP provides a common vocabulary for discussing human performance. Rather than vague claims that a design is "more intuitive" or "easier to use," the MHP allows designers to specify which subsystem is affected and by how much.

Limitations of the MHP

The MHP treats the human as an information processor in isolation, without accounting for:

• Emotion and motivation: The MHP has no parameters for stress, fatigue, frustration, or engagement, all of which significantly affect performance.
• Social context: Interactions that involve communication, coordination, or social pressure fall outside the model.
• Learning and adaptation: While the power law of practice provides a general curve, the MHP does not model the cognitive restructuring that accompanies learning (the shift from declarative to procedural knowledge).
• Individual differences: The MHP provides "typical" parameter values with wide ranges. Individual variation in processing speed, memory capacity, and motor skill can cause actual performance to deviate substantially from predictions.
• Error: The MHP predicts correct performance times but does not model error rates, error types, or error recovery — all of which are central to usability.

The MHP and Modern HCI

Despite its limitations, the MHP remains influential because it established the principle that human performance is predictable and quantifiable. This principle underlies the more detailed models covered in the next chapter (GOMS and the Keystroke-Level Model), as well as modern approaches to computational cognitive modelling [Newell, 1992; Anderson, 2007]. The MHP is best used as a first-pass analysis tool: a way to estimate whether a design is in the right performance ballpark before investing in more detailed modelling or user testing. It answers the question "approximately how long will this take?" — not "will users like this?" or "will users make errors?"