This paper presents a system for diagnostic reasoning based on structure and behavior, developed at the Massachusetts Institute of Technology Artificial Intelligence Laboratory. The system is designed to troubleshoot digital electronic circuits by reasoning from first principles, using knowledge of structure and behavior. It has been implemented and tested on several examples, demonstrating its effectiveness in identifying faults in complex systems. The system uses a technique called constraint suspension, which allows for the identification of components that could be responsible for observed symptoms. This approach enables the system to focus its efforts initially, yet methodically expand its focus to include a broad range of faults. The system also employs a systematic method for generating candidate components, based on the underlying assumptions of the device's structure and behavior. The paper discusses the importance of making explicit the assumptions underlying reasoning and describes a technique for systematically enumerating these assumptions. This leads to an overall strategy for troubleshooting based on the progressive relaxation of underlying assumptions. The system can focus its efforts initially, yet will methodically expand its focus to include a broad range of faults. The paper also explores the concept of adjacency, which proves to be useful in understanding why some faults are especially difficult and why multiple different representations are useful. The concept of adjacency is used to explain the difficulty in reasoning about bridges and to guide the selection and use of multiple representations. The paper also discusses the limitations of the system, including the difficulty of modeling analog devices and incomplete models. It also notes that the system is not a complete solution to the problem of troubleshooting, but rather a useful tool that can be used in conjunction with other methods. The paper concludes by comparing the system to other approaches to similar problems, noting that it offers a novel approach to troubleshooting that is based on the structure and behavior of the device. The system is able to handle a wide range of faults, including those that are not easily modeled by traditional fault models. The system is also able to handle complex devices, including those with multiple levels of organization.This paper presents a system for diagnostic reasoning based on structure and behavior, developed at the Massachusetts Institute of Technology Artificial Intelligence Laboratory. The system is designed to troubleshoot digital electronic circuits by reasoning from first principles, using knowledge of structure and behavior. It has been implemented and tested on several examples, demonstrating its effectiveness in identifying faults in complex systems. The system uses a technique called constraint suspension, which allows for the identification of components that could be responsible for observed symptoms. This approach enables the system to focus its efforts initially, yet methodically expand its focus to include a broad range of faults. The system also employs a systematic method for generating candidate components, based on the underlying assumptions of the device's structure and behavior. The paper discusses the importance of making explicit the assumptions underlying reasoning and describes a technique for systematically enumerating these assumptions. This leads to an overall strategy for troubleshooting based on the progressive relaxation of underlying assumptions. The system can focus its efforts initially, yet will methodically expand its focus to include a broad range of faults. The paper also explores the concept of adjacency, which proves to be useful in understanding why some faults are especially difficult and why multiple different representations are useful. The concept of adjacency is used to explain the difficulty in reasoning about bridges and to guide the selection and use of multiple representations. The paper also discusses the limitations of the system, including the difficulty of modeling analog devices and incomplete models. It also notes that the system is not a complete solution to the problem of troubleshooting, but rather a useful tool that can be used in conjunction with other methods. The paper concludes by comparing the system to other approaches to similar problems, noting that it offers a novel approach to troubleshooting that is based on the structure and behavior of the device. The system is able to handle a wide range of faults, including those that are not easily modeled by traditional fault models. The system is also able to handle complex devices, including those with multiple levels of organization.