There is an urgent need for an efficient vaccine against tuberculosis. Here, we explore the potential role of carbohydrate antigens as part of a new tuberculosis vaccine. Emphasis is placed on carbohydrate-protein conjugate vaccines, using the arabinomannan portion of lipoarabinomannan, a major structural surface component of Mycobacterium tuberculosis covalently conjugated to (mycobacterial) protein antigens. Such conjugate vaccines show good protective efficacy in mice and guinea pigs in terms of prolonged survival and reduced pathology. Special attention is paid to the immunology underlying their protective capacity. Conjugate vaccines induce both cellular and humoral responses and, although antibody responses have been thought to be the main protective component, cellular responses - possibly through the CD1 pathway - are also likely to be involved.