Changes in body circumference associated with respiratory movements were simultaneously recorded in the dog at six different levels—three costal and three abdominal.Continuous registration of respiratory movements with this method showed variations in type from time to time in the same individual.Some of these changes were spontaneous; others were elicited by deliberate introduction of new variables.A shifting from a more costal or more abdominal type towards the other respective type was frequently preceded by a momentary and sporadic stretching of the extremities.At times periodic fluctuations in respiratory movements occurred, limited to either costal or abdominal levels. These findings suggested that periodicity may be a localised phenomenon dependent on a high development of chemical sensitivity in the cord or its innervated parts.Great irregularities in the magnitude of respiratory excursions occurred mostly from changing inspiratory circumference. Under these conditions the expiratory circumference may remain surprisingly uniform.In some experiments the expiratory circumferences changed as well. At times the expiratory circumference of the chest showed gross fluctuations, while those of the abdomen remained perfectly constant. The reverse also occurred.These results may indicate, that in addition to a higher reflex mechanism there are lower localised mechanisms in the cord which serve to check the extent of respiratory movements. These mechanisms may operate through the proprioceptive fibres of individual muscles or muscle groups.The magnitude of costal and abdominal expansion in sporadic deep breaths varied out of proportion to the prevailing respiratory movements. ments. The deep breaths were, therefore, frequently of opposite type to the prevailing type. It was suggested that these momentary changes in type of breathing are due to a transitory loss of the local checking mechanisms, which may be responsible for the mode of breathing that prevails.Backward injection of sodium citrate into the femoral artery produced differential effects in costal and abdominal breathing. Intravenous injection of urethane inhibited costal breathing more than abdominal. These and other findings indicate a high degree of chemical sensitivity of the cord, important not only in controlling the degree of ventilation, but the type of breathing as well.Comparing the magnitude of increased costal circumference with the magnitude of increased abdominal circumference, 34·3 per cent. of 106 dogs were found to breathe more with the chest than with the abdomen.Using our new method of comparing costal and abdominal accommodation of air, by grossly correcting for the greater costal circumference and greater length of chest, 79 per cent. were found to be costal breathers.The mean ratio of costal to abdominal accommodation for all of the dogs studied was 2·15, indicating a decided predominance of costal breathing in the dog.Analysis of measurements on 31 female and 73 male dogs failed to show any relation between sex and mode of breathing.Prolonged experiments were frequently associated with a progressive change in type to or towards greater costal breathing. This change may be related to the accompanying augmented ventilation.The effects of double vagotomy and dorsal root section on type of breathing were uncertain.It is concluded that the central mechanism controlling respiration is extensive, and that the spinal cord is important in modifying the type of breathing.