Following the classical portfolio theory all an investor has to do for an optimal investment is to determine his risk attitude. This allows him to nd his point on the capital market line by combining a risk-free asset with the market portfolio. We investigate the following research questions in an experimental set-up: Do private investors see a relationship between risk attitude and the amount invested risky at all and do they adjust their investments if provided with dierent risk levels of the risky asset? To answer these questions we ask subjects in a between subject design to allocate a certain amount between a risky and a risk-free asset. Risky assets dier between conditions, but can be transformed into each other by combining them with the risk-free asset. We nd that mainly investors risk attitude, but also their risk perception, and the investment horizon are strong predictors for risk taking. Indeed, investors do not appear to be naïve, but they do something sensitive. Nevertheless, we observe a strong framing eect: investors choose almost the same allocation to the risky asset independently of changes in its risk-return prole thus ending up with signicantly dierent volatilities. Feedback does not mitigate the framing eect. The eect is somewhat smaller for investors with a high nancial literacy. Overall, people seem to use two mental accounts, one for the risk-free and one for the risky investment with the risk attitude determining the percentage allocation to the risky asset and not the chosen portfolio volatility. * University of Mannheim, Chair of Banking, contact: ehm@bank.bwl.uni-mannheim.de * * University of Mannheim, Chair of Banking, contact: kaufmann@bank.bwl.uni-mannheim.de * * * University of Mannheim, Chair of Banking and CEPR, London. contact: weber@bank.bwl.unimannheim.de We thank Daniel Foos, Markus Glaser, Dan Goldstein, Thomas Langer, Daniel Smith, and Sascha Steen, as well as all seminar participants at the 2nd Annual Boulder Summer Conference on Financial Decision Making, the SPUDM conference 2011, the LMU Munich, the University of Mannheim, and the University of Münster. We are grateful to the German Consumer Protection Agency (Verbraucherzentrale Bundesverband) for hosting a common press conference leading to signicant media coverage and the promotion of our on-line questionnaire and to Dominic Hiller for programming the experiment.