Identifying gender by handwriting: you’re probably not as good at it as you think

Although linking handwriting to psychology is considered pseudoscience, there’s some merit in predicting a person’s gender based on their scrawlings.

Whether this gender impression is formed by i’s dotted with hearts or a more subtle signifier, we at Survata were interested in gauging the public’s ability to successfully identify gender based on handwriting. To do so, we gathered five handwriting samples of each gender and presented them to 3,100 respondents using our consumer survey tool, asking if they think the handwriting belongs to a man or woman (see the footnotes for details on our methodology). Overall, we found that respondents correctly identified gender by handwriting 54% of the time.

That average is misleading, though, because it is watered down by a few samples to which responses were overwhelmingly wrong. Respondents were relatively decisive on six of the ten presented samples, with at least 60% of respondents guessing one way. This decisiveness was occasionally for the wrong option, however, as shown below (to test your own gender identification abilities, view the sample on the left before looking at the gender on the right):

Gender equality

The battle of the sexes was a draw when it comes to guessing handwriting; male and female respondents, at 54%, had a virtually equal success rate. But who is better at guessing a specific gender’s writing? It seems like it takes one to know one. Men identified male handwriting successfully 64% of the time, whereas women correctly identified male handwriting writing 59% of the time. Likewise, women were more successful than men at identifying female handwriting by a margin of 49% to 45%.

The female handwriting images in our limited sample were more confounding than the male, as respondents had better luck guessing male samples by a score of 61% to 47%.  Of course, five samples per gender is not enough to generalize which gender’s handwriting is more recognizable overall.

Our take

We were surprised that respondents weren’t better at identifying the gender behind the writings, and that so many were wrong on several of the questions. The participants who submitted handwriting samples were blind to the purpose of the survey, and the samples therefore weren’t intentionally misleading. Perhaps the average person simply isn’t as good a judge as expected.

Is the public jumping to conclusions about your image? Try Survata today to find out.

Footnotes for our fellow data geeks

  1. We obtained handwriting samples from participants recruited on Craigslist.org. All handwriting samples were from participants in the San Francisco Bay Area.
  2. We did not control for geographic location, age, ethnicity, SES status, or handedness (i.e., left or right). We do not know whether those variables affect gender recognition of a handwriting sample.
  3. We interviewed 3,100 online respondents from November 22 to November 26, 2013.
  4. You can download the underlying data here.
  5. You can analyze the underlying data in Statwing here and here.
  • DecipiQQ

    The q’s and the f’s in the sentences by women are less boxy and just have a feminine feel.

  • Greg Esres

    Five samples from each gender? Can’t believe your results are meaningful.

    • disqus_yoLyfvueEe

      Life is a reading test – look at the bottom of that sample – it says 3,100 participatns. What was odd was that they chose samples that were so similar to one another. What they picked wasn’t a representative sample of the whole of variation you see in handwriting, but maybe that was the purpose – to try to purposely obscure gender (for the purposes of the chart, to see if you could accurately guess.)

    • survata

      Hi Greg, Thanks for reading. As we say in the article, “of course, five samples per gender is not enough to generalize which gender’s handwriting is more recognizable overall.” However, the 3,100 respondents we reached to judge the handwriting samples are more than enough to determine which gender is better at identifying the gender of our samples. Thanks!

      • Amber Catherine Kerr

        But 5 handwriting samples per gender is insufficient to capture the variation within and between genders. Imagine that the criteria you were measuring was “body weight as a function of gender,” rather than “handwriting style as a function of gender.” You could weigh 5 men and 5 women and then ask 3,000 (or a million) respondents to guess, based on the weights alone, whether a person was male or female. There’s so much overlap in male and female body weights that, depending on which 5 men and 5 women you chose for your samples, a clear pattern might not emerge. And yet we know that there is, overall, a very statistically significant difference between male and female body weights. You just might have to weigh more than 5 people of each gender to be able to see that clearly!

  • Johnson

    Correction: You are talking about Sex not Gender

  • Julian Miller

    asking the general public to analyze handwriting is like asking them about quantum mechanics. why not ask handwriting experts that would be far more interesting.

  • Amber Catherine Kerr

    I think that the 10 samples of handwriting that were used for this study were problematic for several reasons: (1) 10 is an extremely small sample size, about which no statistically relevant conclusions can be drawn. (2) These samples were obviously produced by people trying to write “neatly,” which obscures a lot of the individual variation in handwriting. As a teaching assistant who has graded many thousands of pages of problem sets and lab reports, I can assure you that most handwriting has much more character than these samples do!