Liljeholm & Cheng(2009) figured out that virtual sample can predict participant’s causal inference on contingency with Power-PC>0 but had no ideal about weather the virtual sample can also predict causal inference on contingency with Power-PC=0 and weather the same increment of virtual sample lead to the same increment of causal judgment. The paper conducted a experiment to probe the influence of the same increment of virtual sample on causal strength estimate on contingency with Power-PC=0 and Power-PC>0. 84 participants served in experiment adopted summer format contingency. Experiment design by 2(causal direction: generate, prevent)x2(exact sample:16, 64)x3(Power-PC: 0, .50, 1)x4(virtual sample: 12.5%, 50%, 87.5%, 100%) repeated measure, causal direction varied between subjects and other three factors varied within subjects. Three levels of virtual sample, 12.5%, 50% and 87.5%, was designed to test the influence of the same increment of virtual sample, and three levels of Power-PC, 0, .50 and 1, was designed to test participant’s causal strength estimate on contingency with Power-PC=0 and Power-PC>0. Experiment material is a booklet include 1 exercise contingency to make sure that participant understand experiment task and 24 randomly arranged experiment contingencies to obtain participant’s causal strength estimate. Participant gave their strength estimate under each contingency. Results show that: a) participant’s strength estimate change in the same way with virtual sample size for contingency with Power-PC=1 and .50, but in the opposite direction with virtual sample size for contingency with Power-PC=0; b) the strength estimate change significantly for contingency with Power-PC=.50 and 1 when virtual sample change from 12.5% to 50% or from 50% to 87.5%, but change significantly for contingency with Power-PC=0 only when virtual sample change from 50% to 87.5%, the increment of strength estimate induced by virtual sample change from 12.5% to 50% is bigger than(equal to) the increment of strength estimate induced by virtual sample change from 50% to 87.5% for contingency with Power-PC=1(.50). It seem that virtual sample play a steadily role in participant’s strength estimate, the same increment of virtual sample lead to different increment of strength estimate and different change trend of strength estimate on contingency with different Power-PC. However, these results share the same theory implication: high virtual sample lead participant’s causal inference to Power-PC model’s prediction, which been deemed to maximum likelihood estimation of strength estimate.