I wrote a few direct responses, but they all felt patronizing. Apologies for the length, but I'm hoping to avoid that by showing my math, as it were:
About 7 years ago, I started to wonder about a cluster of concepts that I've yet to successfully explain to anyone without getting a confused look...
Roughly, it's a curiosity about what (if anything) is in the joints and gaps and shadows around the more salient features of our existence. An example is like: how (if at all) does having systems for getting rid of unpleasant things (human waste, trash, unwanted products, corpses, people who are disabled/invalid/elderly, criminals...) shape our thinking and understanding?
Privately, I think about this as unraveling: pulling on loose threads anywhere I catch myself or others taking something for granted.
A consequence of picking around in these gaps and cracks is that I've grown frustrated with most of the discourse, research, and subsequent popular-press reporting that has any overlap with a human/social system. The problem is pretty ubiquitous. The kindest diagnosis I can give is that there may be a big hole in our language(s)/grammar.
So, the methodology of the 2003 study (https://link.springer.com/content/pdf/10.3758%2FBF03196134.p...) cited first in the paragraph you link at Wikipedia documents its sample: "The participants were 60 men and 60 women who were each paid the equivalent of $4.50 U.S. Between 20 and 30 years of age (mean age, 22.2 years), all were right-handed students enrolled in either undergraduate or graduate programs (in social science, French, or administration) at the Université de Montréal."
Likewise, here's the first sentence of the discussion: "The present results revealed relatively high accuracy levels overall. Although men did not supply significantly more correct answers than did women across the three conditions, they needed, on average, less time to carry out the mental rotation of the structures illustrated in Vandenberg and Kuse’s (1978) items."
I want to draw a few threads out of this:
1. A more honest summary of the research is something like: "60 male and 60 female college students at a single university in Montreal, between the ages of 20 and 30, when studied in the early 2000s, representing 3 majors, selected by an unspecified strategy which involved compensation of $4.50 and then broken into 3 subgroups of 20 and given a similar mental rotation task in a different format, performed the mental rotations with similarly high levels of accuracy, though the selected male students groups did so around 33% faster." This is an improvement, but it still doesn't well-reflect how low our confidence still is that it measured anything accurate (let alone valid) about statistically indistinguishable groups ~now, let alone our roughly total inability to answer whether this relationship is fixed and durable across meaningful timescales.
2. There are reasonable alternatives to the headline treatment of the result. Since there was no statistically significant difference in the accuracy of their results--only the response time--the narrative about men's superior performance this task merits scrutiny. Is the difference explained by differences in confidence? Cautiousness? Familiarity with similar tasks? The study makes a halfhearted attempt to address cautiousness by asking for a self-report on whether they double-checked their answers, and on how difficult they thought the task was. Women consistently rated the problems as more difficult (but answered as accurately). Why did they rate it as more difficult?
3. The study discusses two main reported rotation strategies in the literature--holistic and analytic--and that the holistic strategies tend to be faster than the analytic strategies. It also notes that "The literature thus suggests that, in the standard visual presentation condition, the holistic strategies preferred by men are more efficient than the analytic strategies preferred by women." I haven't read the whole study, but I don't see any attempt to tease out to what extent response time differences persist or disappear when controlling for strategy. I see no discussion of or attempt to investigate why men and women are using different strategies. But, the fact that the primary difference is response time, and a response-time difference is noted between strategies, and strategy differences are noted between the genders, and how cultural/experiential and non-intrinsic strategy selection appears as a variable, I'm a bit slack-jawed that this isn't the headline result.
4. I see no examination or breakdown of demographics/activities that might matter. Keep in mind that while they covered 120 students in total, each individual test group is 20 students with no discussion of how they were selected. These groups are small enough to be vulnerable to weird samples even if they did draw statistically random samples. How many of these students played sports? Which one? How recently? At what level? What about adjacent recreational activities they might not consider a sport, such as shooting, skating, skateboarding, bowling, rock-climbing, kayaking, mountain-biking, yoga, running, etc. How many practice some sort of art or craft? Do they play video games? Pilot any sort of vehicle? Does controlling for any of these affect the observed gender difference?
Moving beyond the study itself:
5. If the response-time difference is durable and intrinsic, it's not obvious what if any bearing a response-time difference the measured task has on how well men and women can perform an engineering job. They might not even be meaningfully correlated. Is an engineer's performance strongly bottlenecked by how quickly they rotate objects, or is this delta largely leveled in practice by other factors (e.g., my typing speed is rarely the thing that limits how much code I write in a day).
6. Even if such a speed difference does have a bearing on how well men and women perform an engineering job, there's no basis for assuming the magnitude of the speed performance difference would equal the magnitude of the job performance difference (which could be larger, smaller).