Long-term memory increase induced by Ca
2+ is dependent on de novo gene transcription. Control bees are compared with bees that received an injection of actinomycin-D (ACT-D) 3 h after conditioning. In control bees (black bars), injection of caffeine 20 min before one-trial conditioning induced a significant increase in retention performance (percentage of specific response, % SR, % individuals responding to the conditioned stimulus and not to the new odor) at 72 h, compared with bees which received a one-trial conditioning (χ2 = 10.41, P = 0.0013). This replicates the results shown in Figure 3. Bees which received three conditioning trials showed high long-term memory (LTM) performance, as usual. Injection of ACT-D (white bars) almost totally erased the promnesic effect of caffeine, so that there was no longer any difference between one-trial and one-trial plus caffeine bees. Bees of the caffeine group injected with ACT-D (n = 62) thus showed a significant decrease of SR compared with control bees (n = 61) (χ2 = 16.62, P < 0.001). No difference appears for the control one trial conditioning (χ2 = 0.24, P = 0.62; Control: n = 64; ACT-D: n = 69). LTM produced by three-trial conditioning was also utterly erased (χ2 = 40.29, P < 0.001; control: n = 63; ACT-D: n = 70).