Mathematics anxiety (MA) is generally defined as a state of discomfort caused by performing mathematical tasks [1]. MA can be manifested as feelings of apprehension, dislike, tension, worry, frustration, and fear [2–4]. It is not clear what factors result in the appearance of MA. Nevertheless, potential causal factors include environmental variables (e.g., negative experiences in class, teacher characteristics), intellectual variables (e.g., the degree of abstract or logical thinking) and personality variables (e.g., self-esteem, learning style, attitude and confidence [5, 6]).

MA can develop in the early school years [5, 7] and becomes increasingly common with age [8, 9]. It is thought to affect a notable proportion of the school age population [2, 10, 11] and adults in post-secondary education [12]. Importantly, MA has several negative effects on children’s and adult’s mathematics education. For example, people who experience high levels of MA are likely to develop negative attitudes toward tasks involving mathematics, drop out of elective mathematics classes or avoid taking them altogether; in addition, those with high MA avoid pursuing careers that require quantitative skills [3, 13–15]. This can have large-scale implications. For example, only 7% of pupils in the UK take mathematics to A level, and while there are many reasons for this, many pupils give a dislike of mathematics as a reason for not continuing [16] and sometimes the dislike is very intense and ‘charged with emotion’ [ibid, p. 10].

Some have viewed MA as form of Test Anxiety (TA) [17]. Studies have shown moderate correlations between TA and MA (between .30 and .50), so they are indeed related constructs; however, measures of MA correlate more highly with one another (between .50 to .80) than with TA, which suggests that MA is a distinct construct [2, 18, 19].

Of all of the negative effects that MA has on learning and using mathematics, the relationship between MA and mathematics performance has received the most attention. Past research has shown small negative correlations between mathematics performance and MA (average correlations of -.27. and -.34 in two meta-analyses) [11, 19–24], indicating that those with high MA show poorer mathematics achievement. However, it has been argued that mathematics achievement, when measured in test situations, is always confounded with MA [2, 25]. That is, the mathematics performance of highly mathematics anxious individuals is impaired because of their “online emotional reaction to the testing situation” [2, p. 320]. Consequently, the mathematics performance of an individual with high MA may appear lower than it actually is, when measured using a test. Furthermore, time-limited testing can negatively affect the performance of high and low maths anxious individuals, but performance is not differentially affected in the two groups [26]. However, individuals with high MA can perform similarly to individuals with low MA when mathematics problems are presented in a more relaxed format [13]. Therefore, the depressed performance associated with high MA and the reported negative correlations between MA and performance may be exaggerated because of the context in which mathematics performance is measured. Nonetheless, the effect of MA on ‘online’ mathematics performance is still pertinent, as mathematics achievement, particularly in secondary and tertiary education, is measured using time-limited tests and formal examinations. Therefore, the assessment of MA in realistic test situations is highly important as these situations exert marked influence on individual career prospects and well-being.

Further research has explored the direction of the relationship between MA and performance and two major theories have been proposed. The Deficit Theory [27], claims that anxiety emerges a result of an awareness of poor mathematics performance in the past. In contrast, the Cognitive Interference Theory [28] posits that high levels of anxiety interfere with the recall of prior learning resulting in poorer performance. A meta-analysis conducted by Hembree [19] of 151 studies of MA found more evidence to support the Cognitive Interference Theory than the Deficit Theory. However, in a more recent investigation, Birgin and colleagues found that the highest unique contribution to children’s MA was from the children’s mathematics performance [29]. Similarly, in one of the few longitudinal investigations, Ma and Xu [3] found that poorer mathematics performance led to higher MA in junior and senior high school students. Together these studies lend support to the Deficit Theory. MA resulting from an awareness of prior poor performance may be related to mathematics self-efficacy beliefs as past studies have shown that maths self-efficacy is highly predictive of MA [30–32]. The findings of two recent studies that children with diagnosed mathematical disabilities show more MA [33, 34] could also indicate that poor performance leads to greater MA, though we cannot rule out the possibility that the disabilities were indeed partly caused by anxiety. Evidently the directionality of the relationship between MA and performance is open for debate and requires further research.

The relationship between gender and MA has also been studied extensively; but findings have not been consistent. There are many studies that have found significantly greater levels of MA in females than males [4, 6, 12, 13, 15, 19, 32, 35–51]. However, there are also many studies that show no gender differences in MA [3, 5, 7, 10, 29, 52–61]. There are indeed a few studies that have found higher MA levels in males than in females [62–64].

Birgin and colleagues have suggested that the lack of consistent gender effects may be because MA is not consistently defined or measured [29]. Indeed, many different MA measures have been used in past studies. The most frequently used scale is the Mathematics Anxiety Rating Scale (MARS) which has 98 items [65]. The large number of items in the scale allows the assessment of mathematics anxiety in a wide range of contexts and is therefore thought to have high construct validity. However, it requires a considerable amount of time for administration, which may make it more difficult to use with school age samples. Therefore several different shortened versions of the MARS have been developed [64, 66–69], however the psychometric properties of these shortened scales have come under scrutiny [22]. Hopko and colleagues developed a 9-item scale known as the Abbreviated Math Anxiety Scale (AMAS) which was found to have strong test-retest reliability, good internal consistency and validity [22, 41].

In comparison to the number of studies that have investigated gender differences in overall levels of MA, relatively few studies have explored whether the relationship between MA and maths performance or maths achievement differs by gender. Betz [12] found that correlations between MA and mathematics performance for University students differed according to gender and course: female psychology students showed a significant correlation between MA and mathematics achievement test scores, whereas males did not; in contrast, correlations between MA and mathematics achievement test scores emerged for both genders in students enrolled in an advanced mathematics course. Hembree’s meta-analysis revealed that females’ higher MA did not result in poorer mathematics performance and that MA was more predictive of maths performance in males [19]. Similarly, Miller and Bichsel [24] found that MA was more predictive of basic maths performance in males than in females; but MA was not more predictive of applied mathematics performance in either gender. Ma and Xu [3] also found gender differences in the relationship between MA and achievement. Specifically, they found that boys’ prior low maths achievement predicted later high MA at all grade levels, however girls’ prior low maths achievement only predicted later high MA at critical transition points during schooling (for example, transferring from middle school to secondary school). A possible explanation for the findings of a greater relationship between MA and achievement in males is that girls tend to experience MA whether or not they have any intrinsic difficulties in mathematics, whereas MA in boys is more likely to reflect initial problems in the subject. Alternatively, boys’ performance may be more negatively affected by anxiety, perhaps because it is less socially acceptable for them to communicate their anxieties, and thus they may be less likely to develop or be shown effective strategies of dealing with anxiety.

On the other hand, other studies have failed to find gender differences in the relationship between MA and performance/achievement [32, 59]; Ma [23] gives a meta-analysis of such studies.

The general pattern of results suggests that there is a relationship between MA and maths performance or achievement, but that the direction of the relationship is not clear, partly due to the fact that studies have generally been correlational rather than longitudinal. Also, different studies have used different measures of both mathematical performance and of MA, making their results hard to compare given that some measures used may have been less reliable than others.

Given the mixed results in the field it is clear that further research, utilising reliable measures of MA, is necessary to investigate gender differences in MA and the relationship between MA and performance. The current study aims to identify whether a gender difference exists in overall levels of MA in 11- to 16-year-old children, and whether the relationship between MA and mathematics performance differs for boys and girls. The current study uses a brief MA scale, the AMAS, which is appropriate for use with young children. Furthermore the current study controls for test-anxiety which is typically not controlled for in MA studies.

It was predicted that girls would report higher MA than boys. It was also predicted that there would be a negative correlation between MA and maths performance for boys and girls, and that this correlation will be stronger for boys than girls.