Loudness change has been recently studied for tones with linearly varying levels. The published results by different authors revealed an asymmetry in the judgment of loudness change for increasing and decreasing levels depending on the range of level. But the results and their interpretations were different between Canévet's and Neuhoff's studies. The first assumption here is that this difference was strongly dependent on the procedure used: respectively, indirect estimation versus direct estimation. On the other hand, similar results were obtained by Neuhoff (2001), Teghtsoonian et al. (2005), and, Susini et al. (2007), when a direct estimation procedure was used. Loudness change was measured in the two former studies, whereas Global loudness was measured in the latter one. Therefore, the second assumption here is that direct ratings of loudness change in Neuhoff's study are confounded with a global impression of loudness. Thus, two experiments were performed in order to test these two assumptions. In the first one, indirect and direct estimation of loudness change of increasing and decreasing levels of synthetic vowel sounds and1-kHz tones for several ramp ranges of either two sizes (15, 30 dB) were obtained by magnitude estimation respectively for two groups of 16 participants. In the second experiment, a third group of 16 participants estimate the global loudness of the same stimuli. Results of the first experiment reveal a difference between indirect and direct estimations. Indirect estimations are significantly different for ramp sizes of 15 and 30 dB, whereas direct estimations are not, and increase with end level for increasing ramps. In addition, direct estimations of loudness change show the same trend than the global loudness estimations, according to the ramp sizes and the end level. These results suggest, as assumed, that direct ratings would rather measure a global impression of loudness than a loudness change.