This paper presents a physical explanation for the self-similarity observed in high-speed network traffic, based on the convergence of processes with high variability (infinite variance), supported by statistical analysis of Ethernet LAN traffic at the source level. The key result is that the superposition of many ON/OFF sources with the Noah Effect (high variability) produces aggregate traffic with the Joseph Effect (self-similarity). The Noah Effect is characterized by heavy-tailed distributions with infinite variance, and the Hurst parameter H is related to the intensity of the Noah Effect through H = (3 - α)/2, where α measures the heaviness of the tail. The study confirms that traffic data from Ethernet LANs are consistent with the Noah Effect, and provides a simple method to distinguish between traditional and self-similar traffic by checking for the presence of the Noah Effect. The results have implications for traffic modeling, simulation, and network performance analysis. The paper also discusses the robustness of the Noah Effect under different threshold values and shows that it is consistent across a wide range of time scales. The findings support the use of self-similar models for traffic analysis and highlight the importance of considering the Noah Effect in network design and performance evaluation.This paper presents a physical explanation for the self-similarity observed in high-speed network traffic, based on the convergence of processes with high variability (infinite variance), supported by statistical analysis of Ethernet LAN traffic at the source level. The key result is that the superposition of many ON/OFF sources with the Noah Effect (high variability) produces aggregate traffic with the Joseph Effect (self-similarity). The Noah Effect is characterized by heavy-tailed distributions with infinite variance, and the Hurst parameter H is related to the intensity of the Noah Effect through H = (3 - α)/2, where α measures the heaviness of the tail. The study confirms that traffic data from Ethernet LANs are consistent with the Noah Effect, and provides a simple method to distinguish between traditional and self-similar traffic by checking for the presence of the Noah Effect. The results have implications for traffic modeling, simulation, and network performance analysis. The paper also discusses the robustness of the Noah Effect under different threshold values and shows that it is consistent across a wide range of time scales. The findings support the use of self-similar models for traffic analysis and highlight the importance of considering the Noah Effect in network design and performance evaluation.