The relative lengths of two rows of objects systematically affected the time to make judgments of the relative number of objects. The time to decide which row contained more objects generally increased with the amount of incongruity between length and number (e.g., reaction time was long when the numerically larger row was the shorter of the two). In addition, the time to decide that the rows contained an equal number of objects increased with the disparity in length between the rows. These results suggest that length information was used in the numerical comparison process. A model that uses an internal transformation of row length is proposed to account for the results.A frequent mental process in everyday life consists of comparing two quantities and deciding whether they are equal in number, or if not, which of the quantities is greater. Such comparisons are usually made in the presence of a variety of irrelevant stimulus attributes. For instance, the physical extent of a group of objects (e.g., how much space it occupies) does not affect the number of objects. Nevertheless, several studies (cited below) have shown that irrelevant dimensions such as physical extent often affect performance in numerical comparison tasks. The experiments that follow will investigate the specific problem of how the time to compare two rows of objects is affected by the relative lengths of the rows. The results suggest that subjects can use length information to make the numerical comparisons quickly and do not have to actually count the objects in each row. A model of this comparison process will be proposed to account for the results.