Using altimeter data of the Kuroshio region, dominant patterns of variability are determined by applying multichannel singular spectrum analysis. A statistically significant propagating mode of variability with a timescale close to 7 months is found. In addition, output from a high-resolution simulation of the Parallel Ocean Climate Model (POCM) is analyzed, which also displays variability on a timescale of 7 months, although not statistically significant at the 95% confidence level. To explain the physics of the apparent preference for variability at this timescale within a strongly eddying flow, we investigate the stability of barotropic North Pacific flows. Within a shallow-water model, there appear to be three different steady flow paths of the Kuroshio that all become unstable to oscillatory modes. One of these modes has a timescale of 7 months for reasonable values of the parameters and has other features in common with the 7-month statistical modes from the POCM output and from the observations. The connection between results from the stability analysis and from the analysis of the observations and POCM output leads to the conjecture that the preference for the 7-month variability is due to the existence of a large-scale barotropic instability of the mean Kuroshio path.