Recent welfare reforms have regulated clients by linking behavioral requirements to the receipt of benefits (Ethridge and Percy, 1993; Handler and Hasenfeld, 1991). Such reforms use a generic policy instrument, social regulation, that has three characteristics: (1) rules or standards prescribing responsible behavior; (2) enforcement agents and auditors to monitor and deter deviations from the rules; and (3) applied to those who violate the rules (Bardach, 1989). This article examines how welfare bureaus monitor and deter deviations from the rules. In the process, the article describes and compares two alternative mechanisms for implementing social regulation. Under the auspices of federal waivers, states linked receipt of grants through the Aid to Families with Dependent Children program (AFDC) to behaviors such as job training, school attendance, work requirements, and paternity establishment (Mead, 1986; and 1996; Besharov and Gardiner, 1996; Wiseman, 1996; Tweedie, 1995; Savner and Greenberg, 1995). Federal policy now mandates work requirements. However, many states have used their authority under the Temporary Assistance for Needy Families program (TANF) to regulate welfare clients' behavior in a variety of other ways as well. A study by the Childrens' Defense Fund (1997) reports that of the 42 states that had submitted TANF plans, 17 included a family cap, eight required some version of Learn-fare, and seven had an immunization requirement. The mechanisms used to implement such reforms require close examination because the quality of state and local administration is critical to their success (Mead, 1996; 1997). Ethridge and Percy have described an idealized model of state program administration: states require accurate and up-to-date recordkeeping (often involving multiple bureaucratic organizations), client notification, due process protections, and efficient award of incentives or application of sanctions (1993,343-346). This idealized model may be difficult to realize in practice. Accuracy, notice, due process, and efficiency may be compromised by problems such as interorganizational coordination, organizational complexity, strategic behavior by clients, welfare fraud, or goal displacement. These problems are related to local administrative practices. The context in which we examine social regulation is the implementation of a welfare reform in Maryland, the Primary Prevention Initiative (PPI). The initiative was a statewide demonstration project linking AFDC benefits to education and health requirements, implemented under a federal waiver. In particular, this article focuses upon the procedures used to verify that clients complied with behavioral requirements. The Primary Prevention Initiative is well-suited for such an examination because it used two different procedures, one based upon the bureau-client relationship and the other based upon interorganizational relationships. We compare these procedures and discuss their significance for the design, implementation, and evaluation of behavioral welfare reforms. Evidence is presented from a variety of sources including federal, state, and local government documents; direct observation of welfare case intake and redetermination; participation in and observation of management and coordination meetings within Maryland's welfare system; analysis of administrative reports about welfare bureau performance; and interviews with key state and local officials. Detailed information from a survey of local welfare bureau staff in Maryland is also presented (see the Methodology Box for additional information on methods, including survey procedures and characteristics of respondents). The Primary Prevention Initiative The objectives of the Primary Prevention Initiative were to improve the health and education of AFDC clients or their children. The initiative had two behavioral requirements: school-aged AFDC recipients were required to attend school at least 80 percent of the time and pre-school AFDC recipients were to receive immunizations and preventive health care (Norris and Bembry, 1995). …