The TTO-method is often used to value health states, but it is susceptible to several biases and methodological difficulties. One of these is a violation of procedural invariance, which means that the way a TTO-question is framed, i.e. either by fixing the period in imperfect health or that in perfect health, can have a substantial effect on the elicited value of a health state. There are four important sources of discrepancy of the two procedures: loss aversion, maximum endurable time, scale compatibility and discounting. In this article, we present the results of a new test of procedural invariance in which we avoided or corrected for two of these sources (discounting and maximum endurable time). Our results indicate that while correcting for discounting does diminish the difference between the two TTO-procedures, a large and significant violation of procedural invariance remains. Loss aversion is probably the main determinant of the remainder of this difference.