Previous studies of macrolide resistance in Campylobacter were primarily focused on strains from various origins or used in vitro systems. In this study, we conducted both in vitro and in vivo experiments to examine the development, stability, and genetic basis of macrolide resistance in Campylobacter jejuni using erythromycin-resistant (Ery(r)) mutants derived from the same parent strain. Chickens inoculated with low-level Ery(r) mutants (MIC = 32 or 64 microg/ml) at 15 days old did not shed highly Ery(r) mutants (MIC > 512 microg/ml) after prolonged exposure to a low dose of tylosin. The low-level Ery resistance was not stable in vitro or in vivo in the absence of macrolide selection pressure. However, high-level Ery resistance displayed remarkable stability in vitro and in vivo. Ribosomal sequence analysis of 69 selected Ery(r) mutants showed that specific point mutations (A2074G or A2074C) occurred in all highly Ery(r) mutants. No mutations in ribosomal protein L4 were observed in any of the in vitro-selected Ery(r) mutants. However, three specific mutations in L4, G74D, G57D, and G57V, were widely found among in vivo-selected Ery(r) mutants. Insertion of three amino acids, TSH, at position 98 in ribosomal protein L22 was observed only in mutants selected in vitro. Inactivation of the CmeABC efflux pump dramatically reduced Ery MICs in Ery(r) mutants. Together, these findings suggest that multiple factors contribute to the emergence of highly Ery(r) Campylobacter in chicken, reveal resistance level-dependent stability of macrolide resistance in C. jejuni, and indicate that C. jejuni utilizes complex and different mechanisms to develop Ery resistance in vitro and in vivo.