ta set, with categorizations beyond the binary removed, any face not categorized as a woman was categorized as a man

man

participant

articipant Proportions f

ticipants made fewer man categoriz

An i

ade categorizations b

Free Text

n figure Figure 6

of unsure and I don’t know counted as I don’t know

other category

For analysis purposes

(y-axis)

Participants

A simple v

of

{style=“color:green”;}

After being allocated to one of the three conditions,

only woman and man; 2) woman, man and other and 3) an open text box for participants to type in their resp

56 women, 47 men, and 2

as recommended by

ontrol condition,

In Study 2 we compared

allow

highly

n both conditions

Figure 3 shows that

We examined the relationship

the kinds of

s, and the

acial gender

sek)

Participants included 33 women, 35 men, and 2 participants who did not indicate gender (self-identified gender was measured using an open-ended text box, fol

part in the Stockholm Unive

the same

For testing,

s the spread of proportions of responses

an or man. This meant

We tested whether this was the case by

he multiple categories conditio

Returning to

categorization

also

Unsurprisingly given this pattern

ito

however, only illustrates the total number of categorizations across all participants

made many categorizations beyond

he multiple categories condition

by most participant

We fit the data to Bayesian mixed-effects models. In all models, facial gender (0 to 100 in seven steps) and response options (one-dimensional, two-dimensional) were included as fixed effects. Additionally, all models included varying intercepts for both participants and trials and varying slopes for facial gender. Exploratory plotting of the data suggested that the relationship between facial gender and rated gender was non was non-linear, suggesting that modelling which treated facial gender as a linear predictor would be misspecified

We used R

counted

dichotomizing the variables

so other = 1 and all other responses = 0.

recorded

ions in their condition

I don’t know

three response options

into

in a lab

across

and some participants were more categorical than others in their ratings.

again

First

orphing steps

trials

nsite

separate continua

he woman continuum and once using the man continuum. In the woman continuum, the anchors were marked not woman and woman. In the man continuum the anchors were marked not man and man.

by the researchers

similar

in

e Stockholm

the

).

other words

he evidence indicated

)

igure 8 illustrates the median and interquartile range proportion of faces categorized as women (in this data set, with categorizations beyond the binary removed, any face not categorized as a woman was categorized as a man)

expense of man categorization

seems to

he difference is so stark, we do not feel that inferential statistics add any more information, but the curious reader may find these in the supplemental materi

e Figure 6 ). Participants who only categorized faces as women or men are not represented in figure Figure 6.

illustrates how many categorizations (y-axis) beyond the binary participants made. Each bar represents how many participants (y-axis) made a certain number of categorizations (x-axis). The different colors denote the different categorizations

illustrates how many participants (x-axis) categorized how many faces (y-axis) according to the categories “other” and “don’t know” (different colors) across the two experimental conditio

responses

variations

“woman” a

odo: change order to be consistent

all participants were informed that participation was voluntary and gave written consent to participate in the study

N~free

and

2 pa

control

baseline

“other”

suggests that categorical perception was not reduced by two-dimensional response options

R = NA)

Results

were not meaningfully

test this,

Thus there are

“woman”

The pattern of scores was non-linear

We fit the data to Bayesian mixed-effects models to test the categorical effects. In a

different trials, and the order of trials was completely rand

i

one-dimensional

tilted

The morphs were made in 7 steps, from completely feminine to completely masculine

In comparison to self-identification questions where open-ended responses are seen as the most inclusive alternative (Lindqvist et al., 2020), the categorization of others benefits from response options that explicitly remind participants that not all people identify as women or men.

Participants were

experiments

carefully

impacted

A probable explanation

finding t

xternal stuff.

he results are co

We found that only multiple categories elicited beyond-binary response

Experiment 2 indicated that participants categorize beyond the binary when response options include more options than women and men only. However, the free text option did not differ from the binary option. Thus, the multiple categories condition, with its explicitly stated non-binary options seems to act as reminders to participants. Furthermore, categorization within the binary was not skewed by the addition of multiple categories or the free text option, meaning that the ratio of women and men categorizations was still about 50/50. This did not systematically affect their overall pattern of responses in terms of woman and man categorizations.

n facial gender and binary categorization (i.e. the slope of facial gender) across the conditions

We treated the binary categories condition as a neu

(Difference = 0, CI =[-0.02, 0.03], BF01= 394.93). The effect of facial femininity on woman categorizations almost was the same in the free text and binary categories (Difference > 0.001, CI =[-0.02, 0.02], BF01= 394.93).

as often in the Multiple Categories and Binary Categori

corresponding to moderate evidence

women and e

Overall rates of b

Participant Flow

outside the binary

There was also a clear differenc

In the

Grey dots represent participants who only categorized faces as women or men

to participate in the study.

Figure

summing variations

summing variations of “other” and “non-binary” and dichotomizing this new index

binary categories, free text, and multiple categorie

The experiment used a between-participants design

The experiment used a between-participants design. The two conditions were the one-dimensional (control condition), and two-dimension

In short, the stimuli comprised a multiracial set of faces morphed to vary in terms of facial gender. We defined facial gender as the degree of the female face present in the morph. In other words, a 33 face was slightly tilted toward the man, a 50 face was an even mixture and a 100 consisted only of the woman’s face. Because there were 18 pairs morphed in 7 steps, the total number of faces was 126.

facial gender as the degree of the female face present in the morph

56 women, 47 men 2 participants did not indicate gender)

Study 2 compared a control condition consisting of standard binary response options to two alternatives: a third gender option (such as ‘non-binary’ or ‘other’) and an open text box for participants to type in their respons

However,

33 and 67.

ost participants display a non-linear S-shape and this was also the pattern of the group means.

arked “not woman” and “woman”; in the “man” continuum the anchors were marked “not man” and “man”. The separate continua were presented on different trials and the order of trials

only to

These were highly correlated (R = 0.86)

multiple

Design and procedur

28, N2d = 38)

The one-dimension condition included 33 participants

N1d

“woman” and “man”.

a

pairs

useful outcome

binary gender

L. DeBruine, 2018).

All participants were informed that participation was voluntary and gave written consent to participate in the study.

one-dimensional response options and two-dimensional response options. If one-dimensional scales influence participants to think of gender as binary and opposites and two-dimensional scales don’t do this, there should be a reduced categorical e

investigate the influencing effect of one and two-dimensional response options by investigating whether participa

faces

encountering a face

Categorical effects for continuous stimuli in any domain suggests

to 80% of participants

“woman” and “man”

esearch on how people perceive and categorize the gender of others has used both dimensional scales as well as discrete categories, but in both cases almost exclusively treats gender as a binary catego

however,

Both the initial and later challenges to the gender binary in

transsexual”

for example,

“androgynous”,

later group of challenges to t

In treating gender as a psychological trait, for example,

have been encouraged to includ

see Carleton et al., 2022; Cronin et al., 2022; D’Agostino et al., 2022 for some recent example

Such measurement invisibilizes

assigned sex

Stefan Wiens

accentuated

In other words

0000-0002-8393-5316

Impact

explanation for the difference between free text and multiple categories

We found that only multiple categories elicited beyond-binary responses. Compared to binary control, neither changed the pattern of categorizations of women and men

meaning that the ratio of women and men categorizations was still about 50/50

categorize beyond the binary when response options include more options than women and me

facial femininity and woman categorizations (i.e. the slope of facial femininity

For example, does categorization of faces as non-binary systematically decrease “woman” categorization. We therefore investigated inclusive response options changed participants overall tendency to categorize women and men.

Categorizations by Participants

fit the data to Bayesian mixed-effects

were aggregated

n the free text condition, this included various variations of “other” and “non-binary”.

ummint

only

stimuli were identical

completed the study i