{
  "id": "hick-hyman-law",
  "title": "Hick–Hyman Law",
  "lang": "en",
  "source_lang": "en",
  "summary": "The Hick–Hyman law describes how average choice reaction time increases approximately with the information uncertainty of the possible responses.",
  "category": "society",
  "subcategory": "psychology",
  "type": "concept",
  "aliases": [
    "Hick's law",
    "Hick–Hyman choice reaction time law"
  ],
  "generated_at": "2026-11-06",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/hick-hyman-law/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/hick-hyman-law/",
      "json": "https://aiwiki.page/zh/hick-hyman-law.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "The Hick–Hyman law is an empirical relationship in cognitive psychology stating that, in many choice-reaction tasks, the average time required to produce an appropriate response increases approximately linearly with the information uncertainty of the alternatives. When the alternatives are equally likely, this means that reaction time increases approximately with the logarithm of their number, rather than directly with the number itself. Named after psychologists William Edmund Hick and Ray Hyman, the law connects experimental studies of human performance with information theory and informs research in human–computer interaction and human factors."
    },
    {
      "heading": "Experimental basis",
      "text": "A typical experiment presents one stimulus from a known set and requires a corresponding response. For example, several lights may each be assigned to a different button; when a light appears, the participant presses its assigned button. Researchers measure reaction time from stimulus onset to a specified response event, such as the initial button press.\n\nIn a simple reaction task, only one response is possible. In a choice-reaction task, the participant must identify the stimulus and select the appropriate response. Increasing the number of possible stimulus–response alternatives generally increases average reaction time, although the size and form of this increase depend on the task.\n\nThe law describes an average relationship across repeated trials. It does not predict an identical response time on every trial or imply that the number of alternatives alone determines performance. Stimulus discriminability, response mapping, practice, preparation, and accuracy requirements also matter."
    },
    {
      "heading": "Mathematical formulation",
      "text": "Equally likely alternatives\n\nFor \\(N\\) equally probable alternatives, a common information-theoretic formulation is\n\n\\[\\overline{T}=a+b\\log_2 N,\\]\n\nwhere:\n\n- \\(\\overline{T}\\) is mean choice reaction time;\n- \\(a\\) is a fitted intercept representing time not explained by the uncertainty term;\n- \\(b\\) is a fitted slope, commonly expressed in milliseconds per bit;\n- \\(N\\) is the number of alternatives.\n\nThus, increasing the number of equally likely alternatives from two to four adds one bit of uncertainty, as does increasing it from four to eight. The model predicts the same increase in average reaction time for either doubling, provided the fitted parameters remain applicable.\n\nAnother widely used version is\n\n\\[\\overline{T}=a+b\\log_2(N+1).\\]\n\nThe two forms are related but not identical, particularly for small numbers of alternatives. The formula and experimental convention must therefore be specified when comparing fitted results. Neither form supplies universal numerical values for \\(a\\) or \\(b\\).\n\nUnequal probabilities\n\nWhen alternatives differ in likelihood, their number alone does not capture the uncertainty of the task. An important generalization uses Shannon entropy:\n\n\\[H=-\\sum_{i=1}^{N}p_i\\log_2 p_i,\\]\n\nwhere \\(p_i\\) is the probability of alternative \\(i\\), and the probabilities sum to one. The corresponding relationship is\n\n\\[\\overline{T}=a+bH.\\]\n\nFor equally probable alternatives, \\(p_i=1/N\\), so \\(H=\\log_2 N\\). If some alternatives occur much more frequently than others, entropy is lower than in an equally probable set of the same size. When participants learn and use these probabilities, average response time may consequently be shorter.\n\nThis is primarily a relationship between average reaction time and average uncertainty. It should not automatically be interpreted as an exact prediction of the reaction time for each individual alternative from its information value, \\(-\\log_2 p_i\\)."
    },
    {
      "heading": "Historical development",
      "text": "Research on the duration of mental processes predates the law. Nineteenth-century reaction-time experiments, notably those of Franciscus Donders, compared tasks involving different demands for stimulus discrimination and response selection.\n\nHick’s 1952 paper, On the Rate of Gain of Information, examined the relationship between choice reaction time and information. Hyman’s 1953 paper, Stimulus Information as a Determinant of Reaction Time, investigated how stimulus information, including differences in occurrence probabilities, affected reaction time. These studies helped establish an information-theoretic approach to human performance during a period when concepts from communications engineering were becoming influential in psychology.\n\nTheir significance was not merely the observation that more alternatives can slow responses. The central contribution was a quantitative relationship between response latency and the uncertainty resolved by selecting a response."
    },
    {
      "heading": "Interpretation",
      "text": "The law is often explained by treating response selection as a process of resolving uncertainty. If additional bits of uncertainty require additional processing time, reaction time will increase linearly with information.\n\nA related illustration imagines repeatedly dividing an equally probable set of alternatives into two groups. The number of binary divisions needed to identify one alternative grows as \\(\\log_2 N\\). This makes the logarithmic relationship intuitive, but it does not demonstrate that the brain literally conducts a serial binary search.\n\nThe law is an empirical regularity rather than a complete theory of decision mechanisms. Different processing models can produce similar relationships. Likewise, interpreting the reciprocal of the slope as an information-processing rate is useful within a fitted task, but does not establish a fixed, general-purpose bandwidth for human thought."
    },
    {
      "heading": "Applications",
      "text": "Interface and control design\n\nThe law provides a framework for considering response-selection demands in menus, control panels, command systems, and other interfaces. Clearly distinguished alternatives and understandable stimulus–response mappings can make selection easier.\n\nHowever, the common design slogan “fewer choices are always faster” goes beyond the law. Selecting a command in an interface may involve reading, visual search, remembering its location, navigating categories, and moving a pointer. These components need not follow the same logarithmic relationship.\n\nDividing a large menu into several smaller menus may reduce the alternatives at each step while adding navigation steps. Under an idealized entropy model, dividing an equally probable choice into balanced stages does not necessarily reduce total information uncertainty. Any practical benefit or cost depends on organization, labeling, familiarity, and interaction overhead.\n\nHuman-performance research\n\nResearchers use the law to study how people respond to uncertainty under controlled conditions. Manipulating alternative counts and occurrence probabilities helps distinguish information-related effects from other influences on performance.\n\nThe relationship is also relevant to the design and evaluation of operational controls, where operators must recognize signals and choose corresponding actions. Such applications require task-specific measurement: coefficients obtained from a laboratory button-press experiment cannot simply be transferred to complex real-world decisions.\n\nRelationship to movement time\n\nThe Hick–Hyman law is often discussed alongside Fitts's law, which models the time needed to move to a target as a function of movement distance and target size. The two address different aspects of performance: choosing a response and executing a targeted movement.\n\nIn an interaction such as selecting a menu item, both processes may matter. Their contributions are not necessarily independent, however, and simply adding predictions from the two laws is not always justified."
    },
    {
      "heading": "Limitations and qualifications",
      "text": "Several conditions constrain the law’s interpretation:\n\n- Stimulus–response compatibility. A natural correspondence, such as a left-hand signal requiring a left-hand response, can make selection easier than an arbitrary mapping. Strong compatibility may substantially reduce the effect of alternative count.\n- Practice and expertise. Repeated performance can change response strategies, strengthen associations, and alter both the intercept and slope.\n- Sequential dependencies. Repetitions and predictable sequences influence preparation. The overall frequency of each alternative may therefore be insufficient to describe the uncertainty experienced on a particular trial.\n- Speed–accuracy trade-off. Participants can respond more quickly by accepting more errors. Comparisons of reaction time require attention to accuracy and task instructions.\n- Perceptual and search demands. If identifying the stimulus requires extensive search or difficult discrimination, these demands may dominate the measured time.\n- Complex decisions. Choosing among careers, purchases, or policy proposals involves preferences, evaluation, and information gathering, not simply selecting a predefined response to a signal.\n\nThe law is also distinct from choice overload, which concerns possible effects of large choice sets on motivation, satisfaction, or willingness to decide. A logarithmic increase in choice reaction time does not by itself establish that larger assortments reduce satisfaction or prevent decisions."
    }
  ],
  "links": [
    "cognitive-psychology",
    "information-theory",
    "human-computer-interaction",
    "information-entropy",
    "probability",
    "psychology"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}