This study aimed to compare the development and sensory channel feedback effects in number line estimation between children in grades 2 and 4 through two experiments, using both bounded number line estimation (BNLE) tasks and unbounded number line estimation (UNLE) tasks, respectively. Both Experiment 1 and Experiment 2 employed a mixed design of 2 (Grade: Grade 2, Grade 4) × 3 (Feedback Channel: visual feedback, auditory feedback, audio-visual feedback) × 2 (Scale: small, 0–100; large, 0–1000). Grade and feedback channel were between-subjects factors, while scale was a within-subjects factor. The results showed that in the BNLE tasks, children in both grades demonstrate the ability to estimate numbers within the linear representation stage, with grade 2 estimating numbers on the range of 0–100, while grade 4 extends this to the range of 0–1000. However, children in grade 2 are still undergoing the transformation from logarithmic-linear representation to the range of 0–1000 in their number estimation. In the UNLE tasks, grade 4 children showed greater linear model explanatory power within the range of 0–10 and no significant difference in the range of 10–30. Grade 2 children lacked significant linear or logarithmic representation characteristics, suggesting a potential transition phase. Auditory feedback enhanced BNLE task performance, while visual feedback improved UNLE task performance. The findings indicate that primary school children tend to shift from logarithmic to linear representation in both tasks, with higher BNLE task accuracy. Feedback proves beneficial in enhancing estimation accuracy in both types of NLE tasks, and distinctive feedback channel effects emerge in BNLE and UNLE tasks due to the varying strategies employed by children in each task.