The United States exhibited wildly erratic behavior regarding the institutionalization of persons deemed deviant during the 20th century. During the first half of the century, the country institutionalized deviants in mental hospitals and asylums at extraordinarily high rates even by modern carceral standards, with peaks of about 634 and 627 persons per 100,000 adults in 1948 and 1955 respectively. Deinstitutionalization brought a radical diminution of that population, but it coincided with a sharp increase in our prison populations, which reached 600 inmates in state and federal prisons per 100,000 adults in 2000. In previous research, I analyzed these trends at the national level and explored their relationship to homicide rates in the United States. Using a Prais-Winsten regression model, I found a significant statistical relationship between the rate of aggregated institutionalization (asylums and prisons) and homicide over the period 1928 to 2000, holding constant three leading covariates of homicide (youth demographics, poverty, and unemployment). The analysis was based on national-level data and, naturally, raised the question whether the results were the product of an ecological or other error of aggregation. One key question that emerged from the previous study was whether the findings would hold at the state level. This study collects and tests state-level data and finds that, indeed, the correlations remain strong and robust using state-level panel data regressions, as well as focusing individually on the states. The study reaches the following three conclusions. First, at the national level, using the new, expanded data on mental institutions (including all institutions for those deemed mentally ill), the contrast between the mid-century and 2001 is even more pronounced: during the 1940s and 50s, the United States consistently institutionalized in mental hospitals and prisons at rates above 700 persons per 100,000 adults, reaching peaks of 778 in 1939 and 786 in 1955. The relationship between the expanded aggregated institutionalization and homicide rates over the period 1934 to 2000 is statistically significant at the national level, holding constant three leading correlates of homicide. Second, the very same relationship exists at the state level. Using state-level panel data regressions spanning the entire period from 1934 to 2001, including all 50 states, and controlling for economic, demographic, and criminal justice variables, I find a large, robust, and statistically significant relationship between aggregated institutionalization and homicide rates. The predicted relationship is not linear, but involves some slight elasticity. The findings are not sensitive to weighting by population and hold under a number of permutations, including when I aggregate jail populations as well. Third, as the panel data suggest, there are important relations when I analyze individual states. Using a Prais-Winsten regression model, the individual state analyses reveal some differences and nuance, but the overall direction of influence among the individual states shows the same striking relationship between aggregated institutionalization and homicide. I conclude from the analyses that the earlier national-level findings are not the product of an ecological fallacy or some other artifact of aggregating different state processes. They reflect a genuine and striking relationship at the state level. Naturally, the findings raise more questions than they answer. But they represent an important first step in formulating the right questions. These questions revolve then around the continuity of confinement throughout the 20th century and the issue of defining deviance. ..PAT.-Unpublished Manuscript [ABSTRACT FROM AUTHOR]