Human adults, infants, and non-human animals are believed to be equipped with an Approximate Number System (ANS) supporting non-symbolic representations of numerical magnitudes. Recent research has questioned both the validity and reliability of tasks intended to measure acuity in the ANS. Issues with validity and reliability might be due to differences in methodology. In the present study, we compare four tasks designed to measure ANS acuity, using a within-subjects design. The tasks are compared with respect to response and presentation format effects previously studied in the psychophysics literature, but largely ignored in the ANS literature. We find a presentation format effect and show that when non-symbolic numerical stimuli are presented sequentially the magnitude of the second stimulus is overestimated. Further, the results indicate that people’s sensitivity to differentiate between non-symbolic numerosities is dependent on response format. The implications of the results to measures of ANS acuity are discussed.