Three empirical findings concerning the responses of the intermediate stage of number conservation are to be presented and discussed from the view-point of the change of dominance between perceptual and numerical-inferential cues. (A) Responses on conservation tasks given to Ss of our previous experimental education of number conception were examined for i) consis tency of conservation responses to various sub-items, ii) reliability of conservation responses with a2-week interval.Four out of 8 conservation sub-items given were as follows: After S recognized the equivalence of two collections, E transformed one of them into the prescribed configurations and asked S,“Now, which is more?” Other 4 sub-items concerend the invariance of the quantity of a set beforeand after transformations. The results showed: i) The number of elements in a set or the type of transformations had little effect on the difficulty of items, though the conservation tasks without a standard stimulus (a nontransformed collection) were the more difficult. About 80-85% of 109 5-and 6-year-old childrenmade internally consistent responses, i.e., conservation or nonconservation response to all of the4items. ii) About 70% of 50 children made 4 consistent conservation or non-conservation responses in both of the tests, administered at a two-week interval, and more than 90%, fell in the same category if the distinction was made between 4-conservationrespondents and others. (B) An experiment was carried out to examine the effect of suggestion or counter-suggestion upon responses at conservation tasks.i) Nine 5-yearold children, who had been non-conservers in the ordinary conservation situation, were given suggestions to use numerical rather than perceptual cues. For three of then, the mode of response changed completely and made conservation responses at the test one-week later.Further 2 of them showed a little fluctuation to suggestions but finally settled down at the non-conservation stage.The other 4 children showed no sign of change in the positive direction. ii) On the contrary, [8 5-year-old children, who had always made correct responses at number conservation tasks with a standard stimulus, were given counter-suggestions to emphasize perceptual differences generated by the transformation of one collection. Although one of them regressed to the non-coservation stage, the other 7 did not show any instabilities. (C) The extinction of principle of number conservation was attempted.This experiment, which copied Smedslund's procedure, was to observe children's responses to the tricky situation in which number (quantity of a collection) did not conserve with. spatial transformations.Five out of 9 5-year-old children, who had acquired conservation spontaneously, and 4 out of 7 children who had learned conservation by one of the two training curricula, resisted the attempt at extinction and often interpreted apparent non-conservation as meaning that one element was added or taken away secretly.The remaining. 7 children easily gave up the concept of conservation after 2 or 3 cheating trials and warranted nonconservation by perceptual reasons.It must be noted that 4 non-extinguished children among conservers-by-training were those who had been given the curriculum inducing number conservation by the imagination of inter-number relations, the 3 extin guised were these who had only been given repeat ed confirmation of invariance of a set by counting. Also discussed in detail were the implications of these empirical evidences to process or dynamics. of transition from perceptual-cue-dominance to numerical-inference-cue-dominance as well as appreciated effects of various training procedures.