We present new feedback methods as part of Computer-Assisted Pronunciation Learning (CAPL) system. For the speech sounds that pronunciation errors are pointed out, we isolate the central frames of their corresponding segments using the HMM segmenter and trigger the feedback modules. For vowels, we plot the formant frequency on the vowel diagrams mapped onto the open/close position on the jaw and front/back on the tongue, whereas for consonants we classify the frames into the categorical articulation on the place and manner of articulation in order to see whether or not the learners' articulation is correct. Finally, we show examples of learners' improved pronunciation through the system feedback.