Recent theories in numerical cognition suggest that humans are equipped with a mental system that supports the representation and processing of symbolic and nonsymbolic magnitudes, called the Approximate Number System (ANS). Prior research also suggests that the acuity of the ANS can predict individuals’ mathematical ability. However, results from research within the field has proven to be inconsistent with one another which raises questions about the reliability and validity of methods used to measure the ANS. The present study attempts to replicate the results found in studies suggesting that ANS acuity correlates with mathematical ability. The study also investigates the reliability and validity of different task that have been used to measure the ANS, and also presents a new method of measuring the ANS with an adaptive method. The results show that two tasks correlate significantly with mathematical ability, and multiple regression analyses show that ANS acuity can predict mathematical ability when controlling for general intelligence. Furthermore, the results also further highlight the issue of methodological flaws in previous studies.