This paper presents a cognitive model of the planning process, generalizing the theoretical architecture of the Hearsay-II system. The model assumes that planning involves the activities of various cognitive "specialists," each suggesting decisions for incorporation into the plan. These decisions include how to approach the planning problem, what knowledge is relevant, what actions to plan, specific actions to plan, and how to allocate cognitive resources. Specialists operate opportunistically, suggesting decisions when opportunities arise, rather than being systematically coordinated. The paper illustrates the model with a "thinking aloud" protocol and describes a computer simulation of the model. It contrasts the model with successive refinement models and attempts to resolve differences between the two approaches. Planning is a familiar cognitive activity that influences many aspects of daily life. It is defined as the predetermined course of action aimed at achieving a goal, the first stage of a two-stage problem-solving process. The second stage involves monitoring and guiding the execution of the plan. The paper focuses on the planning stage and aims to characterize the planning process and propose a theoretical account of it. Sacerdoti's work on planning is well-known, with NOAH implementing a successive refinement approach. NOAH formulates problems in terms of high-level goals and expands subgoals into additional subgoals. However, the authors' view of planning is different, emphasizing that planning is opportunistic, with decisions and observations suggesting opportunities for plan development. Decisions can lead to subsequent decisions at arbitrary points in the planning space, influencing decisions at higher or lower levels of abstraction. The paper presents a "thinking-aloud" protocol illustrating the kind of behavior the model must explain. It describes a planning model that allows for opportunistic planning, with specialists operating independently and suggesting decisions at different levels of abstraction. The model includes a blackboard for recording decisions and a hierarchical structure of planning levels. The model partitions the blackboard into several planes, each with different levels of abstraction, and postulates decision-making mechanisms that allow for theoretical interpretation of planning processes. The model is tested against a thinking-aloud protocol, where a subject plans a day's errands. The protocol illustrates the subject's planning process, including categorizing errands as primary or secondary, planning routes, and adjusting plans based on new information. The model explains how the subject's planning involves opportunistic decisions, mental simulation of plan execution, and the ability to revise plans based on new information. The model also accounts for the subject's ability to simulate execution of plans at different levels of abstraction and use this to guide subsequent planning. The model is contrasted with successive refinement models, and the authors attempt to resolve apparent differences between the two approaches.This paper presents a cognitive model of the planning process, generalizing the theoretical architecture of the Hearsay-II system. The model assumes that planning involves the activities of various cognitive "specialists," each suggesting decisions for incorporation into the plan. These decisions include how to approach the planning problem, what knowledge is relevant, what actions to plan, specific actions to plan, and how to allocate cognitive resources. Specialists operate opportunistically, suggesting decisions when opportunities arise, rather than being systematically coordinated. The paper illustrates the model with a "thinking aloud" protocol and describes a computer simulation of the model. It contrasts the model with successive refinement models and attempts to resolve differences between the two approaches. Planning is a familiar cognitive activity that influences many aspects of daily life. It is defined as the predetermined course of action aimed at achieving a goal, the first stage of a two-stage problem-solving process. The second stage involves monitoring and guiding the execution of the plan. The paper focuses on the planning stage and aims to characterize the planning process and propose a theoretical account of it. Sacerdoti's work on planning is well-known, with NOAH implementing a successive refinement approach. NOAH formulates problems in terms of high-level goals and expands subgoals into additional subgoals. However, the authors' view of planning is different, emphasizing that planning is opportunistic, with decisions and observations suggesting opportunities for plan development. Decisions can lead to subsequent decisions at arbitrary points in the planning space, influencing decisions at higher or lower levels of abstraction. The paper presents a "thinking-aloud" protocol illustrating the kind of behavior the model must explain. It describes a planning model that allows for opportunistic planning, with specialists operating independently and suggesting decisions at different levels of abstraction. The model includes a blackboard for recording decisions and a hierarchical structure of planning levels. The model partitions the blackboard into several planes, each with different levels of abstraction, and postulates decision-making mechanisms that allow for theoretical interpretation of planning processes. The model is tested against a thinking-aloud protocol, where a subject plans a day's errands. The protocol illustrates the subject's planning process, including categorizing errands as primary or secondary, planning routes, and adjusting plans based on new information. The model explains how the subject's planning involves opportunistic decisions, mental simulation of plan execution, and the ability to revise plans based on new information. The model also accounts for the subject's ability to simulate execution of plans at different levels of abstraction and use this to guide subsequent planning. The model is contrasted with successive refinement models, and the authors attempt to resolve apparent differences between the two approaches.