People are often better at comparing fractions when the larger fraction has the larger rather than the smaller natural number components. However, there is conflicting evidence about whether this “natural number bias” occurs for complex fraction comparisons (e.g., 23/52 vs. 11/19). It is also unclear whether using benchmarks such as 1/2 or 1/4 enhances performance and reduces the bias (e.g., 11/19 > 1/2 and 23/52 < 1/2, hence 11/19 > 23/52). We asked 107 adults to solve complex fraction comparisons that did or did not afford using benchmarks, and we assessed response time and accuracy. We found a reverse bias (i.e., smaller components—larger fraction) that was greater among participants with lower mathematics experience. Fractions' proximity to 0 or 1 facilitated performance and decreased bias; effects of other benchmarks were nonsignificant. These results challenge the generality of the natural number bias in fraction comparison and highlight its variability.

当较大的分数具有较大的自然数成分而不是较小的自然数成分时，人们通常会比较分数。但是，关于复杂分数比较是否会发生这种“自然数偏差”（例如23/52与11/19）存在矛盾的证据。还不清楚使用基准测试（例如1/2还是1/4）是否会提高性能并降低偏差（例如11/19> 1/2和23/52 <1/2，因此11/19> 23/52） 。我们要求107位成年人解决使用基准测试无法或无法承受的复杂分数比较，并评估了响应时间和准确性。我们发现，在数学经验较低的参与者中，存在较大的反向偏差（即，分量较小，分数较大）。分数接近0或1有助于提高性能并减少偏差；其他基准的影响不显着。这些结果挑战了分数比较中自然数偏差的普遍性，并突出了其可变性。