I have been thinking about geoengineering for climate modification since I worked on the committee that produced the 1992 National Academies report, Policy Implications of Greenhouse Warming. Over the years, and increasingly now, I have been puzzled by the scientific community’s attitudes toward the issue; those puzzles were raised again by Barbara Goss Levi’s story (Physics Today, August 2008, page 26).It has been customary to discuss geoengineering without offering an explicit definition. I propose the following one: Geoengineering is purposeful action intended to manipulate the environment on a very large—especially global—scale. Geoengineering is, presumably, undertaken to reverse or reduce impacts of human actions.Decreasing human emissions of carbon dioxide and other greenhouse gases is a good idea for many reasons, including climate modification; yet it is not clear to me why manipulating the CO2 content of the atmosphere is not considered geoengineering. If, using the above definition but narrowing it to the case under consideration, geoengineering includes purposeful manipulation of physical, chemical, and biological variables on the global scale for the purpose of changing global climate, then manipulation of carbon dioxide concentrations fits the definition as well as do, for example, manipulating atmospheric aerosol content to control albedo or manipulating the ocean’s iron content to increase the long-term oceanic storage of carbon. Exclusion of carbon dioxide manipulation from geoengineering has led to a double standard in considering possible negative consequences. There is legitimate concern about side effects of particulate manipulation and the like, but I have not heard much worrying about manipulating carbon dioxide. In a highly nonlinear feedback-controlled system like global climate, we would expect complex hysteresis effects: Decreasing a control variable such as greenhouse gas will not necessarily lead the climate back along some path like the one it followed when the control variable was increased. The end state of control-variable manipulation may not at all resemble the original state before the control variable was increased, nor will it necessarily be a state we want to be in. I have heard of no concern about those possibilities, which might be rate dependent, involve transient behavior not to our liking, or lead us through bifurcations into unexpected states.It seems to me we need to be concerned about possible not-so-benign effects of decreasing carbon dioxide and other greenhouse gases before we grab the control knob and turn it down. I hope someone is at least exercising suitable climate models, if relevant ones exist, to examine possible end states, paths to them, and transient effects of carbon dioxide geoengineering.© 2009 American Institute of Physics.