We propose a quality of service (QoS) monitoring program for broadband access to measure the impact of proprietary network spaces. Our paper surveys other QoS policy initiatives, including those in the airline, and wireless and wireline telephone industries, to situate broadband in the context of other markets undergoing regulatory devolution. We illustrate how network architecture can create impediments to open communications, and how QoS monitoring can detect such effects. We present data from a field test of QoS-monitoring software now in development. We suggest QoS metrics to gauge whether information "walled gardens" represent a real threat for dividing the Internet into proprietary spaces. To demonstrate our proposal, we are placing our software on the computers of a sample of broadband subscribers. The software periodically conducts a battery of tests that assess the quality of connections from the subscriber's computer to various content sites. Any systematic differences in connection quality between affiliated and non-affiliated content sites would warrant research into the behavioral implications of those differences. If, however, the data shows that there are no chronic biases in connection quality, then it would be fair to conclude that the walls on the garden are low enough not to be detrimental to public communications. QoS monitoring is timely because the potential for the Internet to break into a loose network of proprietary content domains appears stronger than ever. Recent court rulings and policy statements suggest a growing trend towards relaxed scrutiny of mergers and the easing or elimination of content ownership rules. This policy environment could lead to a market with a small number of large, vertically integrated network operators, each pushing its proprietary content on subscribers. The move towards proprietary space conflicts with the open philosophy on which the Internet was founded. That alone, however, is not a reason for regulators to intervene. Policy makers require empirical evidence that proprietary barriers require a public response. Unfortunately, traditional indicators of harm to consumers from industry mergers, like measures of programming diversity on cable systems, are insufficient when Internet connectivity becomes the norm for access networks. Conventional measures of diversity make little sense if all providers offer access to the Internet. Any differences in information variety resulting from special, proprietary content would be swamped by the content of the Internet. What matters more than a binary measure of availability are the quality and equality of the connection to diverse content providers over access networks. It is unlikely that an Internet access provider would completely block the content of its competitors. Any provider wishing to steer subscribers away from nonaffiliated content would be more likely to do so by delivering that content at a slightly lower quality compared to affiliated content. The degradation of quality need not be blatant to be effective--a differential of a few milliseconds between affiliated and unaffiliated sites should suffice to condition users to abandon nonaffiliated, "slower" content sources. A QoS monitoring system would alert policymakers to the development of such scenarios.