Maths anxiety is not a metaphor for disliking maths. It is a measurable affective state, it is distinguishable from general test anxiety, and its effect on performance operates through an identifiable cognitive mechanism.
The mechanism
Solving a mathematical problem requires holding intermediate results in working memory while operating on them. Working memory has strictly limited capacity.
Anxiety generates intrusive thoughts — about failure, about being watched, about not being a maths person — and those thoughts occupy the same limited capacity.
The prediction that follows is specific: anxiety should degrade performance most on problems that make heavy demands on working memory, and least on problems that can be solved by retrieval. That is precisely what the studies find. Simple recall of memorised facts is largely unaffected; multi-step problems requiring the maintenance of intermediate values are substantially affected.
This is important because it means the performance deficit is real and immediate rather than a matter of accumulated missed learning. A student can know the material and still perform badly under the anxiety, which then supplies evidence for the belief that produced it.
The gender finding, carefully stated
Self-reported maths anxiety is higher among girls and women in most measurements, across many countries.
Two cautions apply. Self-report may be affected by differing willingness to report anxiety, and studies using physiological measures find smaller gaps than questionnaires do. And the anxiety gap is considerably larger than the performance gap, which in most recent international assessments is small or absent.
So the honest version is: girls report more anxiety about mathematics than boys while performing comparably. The anxiety is not tracking ability. It is tracking something else.
Where it comes from
The transmission research is the most striking part of this literature.
A widely cited study found that primary school teachers' own maths anxiety predicted their female students' maths achievement by the end of the year — and did not predict their male students' achievement. The mediating factor was the students' endorsement of the stereotype that boys are better at maths. The effect appeared specifically in girls taught by more anxious teachers.
A related line of work found that parents' maths anxiety predicted their children's maths learning, but only among parents who frequently helped with maths homework. Anxious parents who did not help produced no such effect.
That conditional structure is the key detail. The transmission requires contact. It is not a genetic or general environmental effect; it is something conveyed in the specific interaction of helping.
What is being conveyed
Probably not the anxiety itself so much as a set of statements that accompany it.
"I was never any good at maths either" is offered as reassurance. What it communicates is that mathematical ability is a fixed personal property, that it runs in families, and that the child has permission to conclude she does not have it.
The same sentence is essentially never said about reading. Adults who struggle with reading do not usually tell children so as a comfort, because it is not socially acceptable to be bad at reading. It is entirely acceptable to be bad at maths, and that acceptability is the vector.
The primary teaching workforce problem
Primary teaching is heavily female in most countries, and primary teacher training in many systems requires relatively little mathematics. Studies measuring maths anxiety in trainee primary teachers find elevated levels compared with other undergraduate populations.
Combine those with the transmission finding and the structure of the problem is clear: a workforce with elevated anxiety, in daily contact with children, passing it preferentially to the students who match them.
This is not a criticism of primary teachers, who did not design the training requirements. It identifies a specific intervention point with unusually high leverage.
What reduces it
Interventions targeting teacher and parent anxiety directly, rather than student anxiety, follow logically from the transmission findings, and several programmes built on this principle have reported gains.
Structured parental involvement has been tested experimentally. A study providing families with short maths stories to read together found improvements in children's maths achievement, with the largest effects among children of the most maths-anxious parents. Giving anxious parents a script appears to work better than telling them to help.
Expressive writing before examinations — a brief structured exercise in writing about one's worries — has been shown in several studies to reduce the performance cost of anxiety, plausibly by clearing working memory. Effect sizes vary and replication has been mixed, but the intervention costs ten minutes.
Changing what adults say is free. Replacing "I was never good at maths" with "this bit is hard, let's work it out" removes the specific transmission without requiring anyone to become less anxious.
Why this is among the more hopeful findings
Because it identifies something learned rather than something fixed, it identifies a specific transmission route, and it identifies interventions that are cheap.
Very little in education research has that combination. Most of what is known about attainment gaps concerns factors that are expensive or slow to change. This one is largely a matter of what adults say to children while helping with homework, which is unusually tractable for a problem of this size.