-
Notifications
You must be signed in to change notification settings - Fork 1
/
1_exp.qmd
1487 lines (1279 loc) · 67.3 KB
/
1_exp.qmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
---
title: "Experiment 1"
subtitle: "**Associating Pronouns with a Person**"
toc-title: "Experiment 1: Associating Pronouns with a Person"
---
```{r}
#| label: exp1-setup
#| include: false
library(tidyverse) # data wrangling
library(magrittr)
library(sjmisc)
options(dplyr.group.inform = FALSE, dplyr.summarise.inform = FALSE)
library(lme4) # stats
library(lmerTest)
library(buildmer)
library(brms)
library(insight) # model results
library(broom.mixed)
library(kableExtra) # tables
library(sjPlot)
library(patchwork) # plots
library(RColorBrewer)
library(ggsignif)
library(ggh4x)
library(ggdist)
library(ggtext)
source("resources/data-functions/exp1_load_data.R") # setting up data
source("resources/formatting/printing.R") # model results in text
source("resources/formatting/aesthetics.R") # plot and table themes
```
[![](resources/icons/preregistered.svg){title="Preregistration" width="30"}](https://osf.io/cmkw5) [![](resources/icons/open-materials.svg){title="Materials" width="30"}](https://github.com/bethanyhgardner/dissertation/blob/main/materials/exp1) [![](resources/icons/open-data.svg){title="Data" width="30"}](https://github.com/bethanyhgardner/dissertation/blob/main/data) [![](resources/icons/file-code-fill.svg){title="Analysis Code" width="30"}](https://github.com/bethanyhgardner/dissertation/blob/main/1_exp.qmd)
<br>
## Motivation
In addition to learning that the grammatical representation of *they* allows it to corefer with specific singular referents, learning to use they/them pronouns may require a change in how speakers access pronouns during language production. Extending existing models of pronoun production from [grammatical gender](0_introduction.qmd#def-grammatical-gender "grammatical gender") to [social gender](0_introduction.qmd#def-social-gender "social gender") [@ackerman2019; @mcconnell-ginet2014] would predict that pronouns are accessed based on morphosyntactic gender marking associated with a person's name [e.g., @schmitt1999] or based on semantic/conceptual features of a person [e.g., @anton-mendez2010]. In order to use they/them for a person instead of the expected he/him or she/her, speakers may instead need to recall episodic information about the person's stated pronouns or which pronouns other speakers use to refer to them. It is clear from other contexts, such as referring to pets with *he* and *she*, that people can learn which gender-marked pronouns to use in contexts with few gender cues based on name or appearance. This suggests that learning [specific](0_introduction.qmd#def-specific "specific singular they") singular *they* should be feasible, but may involve retrieving a person's pronouns from memory, rather than inferring them based on cues such as their name.
The first experiment investigated how people learn to associate pronouns with a person when someone's pronouns cannot be inferred from the gender association of their first name. Participants read a series of vignettes which introduced 12 characters, each of whom was associated with a name, pronouns (he/him, she/her, or they/them), a pet, and a job. Memory for pronouns was tested in a multiple-choice recognition task, and production of pronouns was tested in a written sentence completion task.
## Experiment 1A
```{r}
#| label: exp1a-load-data
exp1a_d_all <- exp1_load_data_all("a") # all memory questions
exp1a_d <- exp1_load_data_pronoun("a") # pronoun Qs with 1 row per char
summary(exp1a_d)
contrasts(exp1a_d$Pronoun)
```
### Methods
The design and analysis plan were [preregistered](https://osf.io/cmkw5 "Experiment 1 Preregistration") on the Open Science Framework. [Materials](https://github.com/bethanyhgardner/dissertation/tree/main/materials/exp1 "Experiment 1 Materials"), de-identified [data](https://github.com/bethanyhgardner/dissertation/blob/main/data "Experiment 1 Data"), and [analysis code](https://github.com/bethanyhgardner/dissertation/blob/main/exp1.qmd "Source Code") are available at this dissertation's [Github repository](https://github.com/bethanyhgardner/dissertation "Github repository").
#### Participants
```{r}
#| label: exp1a-n-participants
# Age
exp1a_n_age <- read.csv("data/exp1a_data.csv") %>%
select(SubjID, SubjAge) %>%
unique() %>%
summarise(mean = mean(SubjAge), sd = sd(SubjAge)) %>%
round(2) %>%
format(n.small = 2)
exp1a_n_age
# Gender
exp1a_n_gender <- read.csv("data/exp1a_data.csv") %>%
group_by(SubjGender) %>%
summarise(n = n_distinct(SubjID)) %>%
mutate(Text = str_c(as.character(n), " ", SubjGender)) %>%
pull(Text) %>%
str_flatten_comma()
exp1a_n_gender
# English Experience
exp1a_n_english <- read.csv("data/exp1a_data.csv") %>%
group_by(SubjEnglish) %>%
summarise(n = n_distinct(SubjID)) %>%
arrange(desc(n)) %>%
mutate(Text = str_c(as.character(n), " \"", SubjEnglish, "\"")) %>%
pull(Text) %>%
str_flatten_comma()
exp1a_n_english
```
`r n_distinct(exp1a_d$Participant)` undergraduate students from Vanderbilt University completed the study for partial course credit. The study was conducted online and took approximately 15 minutes. To characterize the participant sample, they were asked about their age (*M*~age~ = `r exp1a_n_age$mean`, *SD*~age~ = `r exp1a_n_age$sd`); gender (`r exp1a_n_gender`), and English experience (`r exp1a_n_english`). Following best practices for TGD-inclusive study design, the question about gender was a free-response box [@cameron2019; @zimman2017; @ansara2014; NASEM, -@nasem2022].
#### Materials & Procedure
Participants were introduced to 12 characters, each of whom had 4 associated facts: name, pronouns, job, and pet. There were 6 typically-masculine names (Andrew, Brian, Daniel, James, Kevin, Michael) and 6 typically-feminine names (Amanda, Emily, Jessica, Laura, Melissa, Stephanie), such that 4 characters had he/him pronouns and masculine names, 4 characters had she/her pronouns and feminine names, 2 characters had they/them pronouns and masculine names, and 2 characters had they/them pronouns and feminine names. While people who use they/them pronouns have a variety of names, the choice of gendered names here means that whether a character used the gendered pronouns associated with their name or they/them could not be predicted on the basis of the name. Thus, participants had to learn each name-pronoun pairing. Characters also had 1 of 12 jobs and 1 of 3 pets. The pet fact had the same distributional characteristics as the pronouns, but was not associated with prior expectations based on the name or with strong sociopolitical beliefs, making pet learning a reference point for comparison. Participants were randomly assigned to 1 of 3 lists. Name-pronoun pairs were counterbalanced such that each name appeared with the expected he/him or she/her in 2 lists and they/them in 1 list.
Participants read descriptions of each character in the frame *\[Name\] uses \[pronouns\]. \[Name\] works as a \[job\] and has a \[pet\]*. After a brief distractor task consisting of simple math questions, there were two tasks measuring learning of pronouns: In the [memory task]{.fw-semibold}, participants were given the character's name and answered multiple-choice questions about their pronouns, job, and pet (e.g., *What pronouns does Emily use? What is Emily's job? What kind of pet does Emily have?*). The 3 questions about each character always appeared together, the order of the 12 characters was pseudo-randomized, and the orders of the choices for each question were randomized. In the [production task]{.fw-semibold}, participants saw sentence fragments for each character (in a randomized order) that included their name and job (e.g., *After Emily got home from working as a teacher...*). Participants were asked to complete the sentence in a way that made sense to them, to measure which pronouns (if any) they used to refer to the character (e.g., *...they made dinner*). Finally, participants were shown 3 characters, 1 for each pronoun, and asked what they would say if they were introducing the character to someone. This task measured if participants used a pronoun in reference to the character or specifically mentioned which pronouns the character used, as a pilot for more open-ended prompts.
### Predictions
Given that people who use they/them pronouns report high rates of both unintentional errors and intentional [misgendering](0_introduction.qmd#def-misgendering "misgendering") [@cordoba2020; @goldberg2019; @james2016; @mclemore2018; @trevorproject2020], we predict that he/him and she/her will be remembered and produced more accurately than they/them. This outcome could be observed for one or more reasons: Participants may be unfamiliar with singular *they*, or familiar with comprehending it but unused to producing it. Singular *they* is also less frequent than *he* and *she*, and as a result may be more difficult to use, even for speakers already familiar with it. Additionally, if participants avoid using they/them as an option, instead choosing the pronouns typically associated with the character's name, accuracy for they/them would also be lower than accuracy for he/him and she/her. Alternatively, the relative novelty of they/them may improve memory, as distinctive information tends to be remembered better [@vonrestorff1933; @wallace1965]. Under this account, accuracy remembering and producing they/them would be higher than for he/him and she/her.
We hypothesize that learning to use singular *they* requires a change from inferring a person's pronoun (*he* or *she*) based on semantic/conceptual features of a person [e.g., @anton-mendez2010] or based on morphosyntactic gender associated with a person's name [e.g., @schmitt1999], and instead recalling episodic information about a person's stated pronouns. In the context of this experiment, the gender association of the character's name cannot predict whether that character uses *he/she* or *they*, meaning that the only way to consistently produce the correct pronouns is to remember the information from the character introductions. As a result, we predict that correctly remembering that a character uses they/them in the multiple-choice task should predict correctly producing *they* in the sentence completion task. Alternatively, pronoun choice in the production task may not be influenced by episodic memory for which pronouns a character uses. This would occur if, in language production, a speaker attempts to infer the character's pronouns based on their name rather than retrieving them from memory, or if a speaker chooses to not produce singular *they*. In this scenario, accuracy in the memory task would not predict accuracy in the production task.
### Results
Responses were analyzed using logistic mixed-effects regression models using *lme4* in R [@bates2015; @rcoreteam2023], with the first model analyzing memory accuracy (@tbl-exp1a-mem), the second analyzing production accuracy (@tbl-exp1a-prod), and the third analyzing a relation between memory and production accuracy (@tbl-exp1a-both). The fixed effect of pronoun was coded with orthogonal Helmert contrasts; the first contrast compared they/them to he/him and she/her, and the second contrast compared he/him to she/her. In the model that included memory accuracy as a fixed effect, it was mean-center effects coded. The maximal models included by-participant and by-item (defined as the 12 names) random intercepts and slopes for pronoun [@baayen2008; @barr2013]. The *buildmer* package [@voeten2023] was used to select the maximal converging models; all three final models included by-participant intercepts, the memory and production models also included by-participant slopes, and the memory model also included by-item intercepts.
#### Memory
```{r}
#| label: exp1a-mem-means
exp1a_r_memory_means <- exp1a_d %>%
group_by(Pronoun) %>%
summarise( # mean and SD for each pronoun
mean = mean(M_Acc),
sd = sd(M_Acc)
) %>%
add_row( # for he+she
Pronoun = "HS",
mean = exp1a_d %>% filter(Pronoun != "they/them") %>% pull(M_Acc) %>% mean,
sd = exp1a_d %>% filter(Pronoun != "they/them") %>% pull(M_Acc) %>% sd,
) %>%
add_row( # for all 3
Pronoun = "all",
mean = mean(exp1a_d$M_Acc),
sd = sd(exp1a_d$M_Acc),
) %>%
tidy_means() # add percentage, round values, fix labels
exp1a_r_memory_means
```
```{r}
#| label: exp1a-mem-model
#| cache: true
exp1a_m_memory <- buildmer(
formula = M_Acc ~ Pronoun + (Pronoun | Participant) + (Pronoun | Name),
data = exp1a_d, family = binomial,
buildmerControl(direction = "order")
)
summary(exp1a_m_memory)
exp1a_r_memory <- exp1a_m_memory@model %>% tidy_model_results()
```
For accuracy in the multiple-choice memory task (@tbl-exp1a-mem), participants responded more accurately than inaccurately across all three pronoun conditions (`r exp1a_r_memory['Intercept', 'Text']`). He/him and she/her (*M* = `r exp1a_r_memory_means['HS', 'mean']`) were remembered significantly more accurately than they/them (*M* = `r exp1a_r_memory_means['T', 'mean']`) (`r exp1a_r_memory['Pronoun=They_HeShe', 'Text']`). Errors were asymmetric: not remembering that a character used they/them was more common than incorrectly attributing they/them to a character (@fig-exp1a-mem).
| |
|---------------------------|
| **Experiment 1A: Memory** |
: Experiment 1A: Model results for the effect of Pronoun on Memory Accuracy. {#tbl-exp1a-mem .borderless}
```{r}
#| label: table-exp1a-mem
#| output: true
exp1a_tb_memory <- tab_model(
model = exp1a_m_memory@model,
transform = NULL, # show log-odds not odds ratios
show.stat = TRUE, string.stat = "z", # show z
show.ci = FALSE, # show SE instead of CI
show.se = TRUE, string.se = "SE",
show.r2 = FALSE, show.icc = FALSE, # don't make sense for logistic models
# shows intercept, p values, random effects, n group, n obs by default
digits = 3, digits.re = 3, # round to 3
dv.labels = "Memory Accuracy", # labels
pred.labels = exp1_tb_fixed_labels,
wrap.labels = 80,
CSS = table_css
)
exp1a_tb_memory$knitr %<>%
exp1_tb_random_labels() %>% # change random effects labels
drop_sigma() # drop sigma squared because it doesn't make sense for logistic
exp1a_tb_memory
```
```{r}
#| label: fig-exp1a-mem
#| fig-cap: "Experiment 1A: [A] Pronoun accuracy in the multiple-choice memory task. By-participant means are shown as points; error bars indicate 95% CIs calculated over the by-participant means. [B] Distribution of memory responses, with the correct pronoun on the X axis and the selected pronoun by color."
#| fig-asp: 0.45
#| output: true
#| cache: true
# accuracy
exp1a_p_mem_acc <- exp1a_d %>%
group_by(Participant, Pronoun) %>%
summarise(M_Acc = mean(M_Acc)) %>%
ggplot(aes(x = Pronoun, y = M_Acc, fill = Pronoun, color = Pronoun)) +
stat_summary(
fun.data = mean_cl_boot, geom = "bar",
alpha = 0.4, color = "NA"
) +
geom_point(
position = position_jitter(width = 0.35, height = 0.01, seed = 1),
size = 0.5
) +
stat_summary(
fun.data = mean_cl_boot, geom = "errorbar",
color = "black", linewidth = 0.75, width = 0.5
) +
scale_fill_brewer(palette = "Dark2") +
scale_color_brewer(palette = "Dark2") +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0.01, 0.01)) +
guides(fill = guide_none(), color = guide_none()) +
theme_classic() +
dissertation_plot_theme + # nudge Y title over from A label:
theme(axis.title.y = element_text(margin = margin(l = 4, r = 2))) +
labs(x = "Pronoun", y = "By-Participant Mean Accuracy")
# distribution
exp1a_p_mem_dist <- exp1a_d %>%
ggplot(aes(x = Pronoun, fill = M_Response)) +
geom_bar(position = "fill") +
scale_fill_brewer(palette = "Dark2") +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
theme_classic() +
dissertation_plot_theme +
theme(legend.margin = margin(l = -5)) +
labs(
x = "Correct Pronoun",
y = "Proportion of Trials",
fill = "Pronoun\nSelected"
)
# combine
exp1a_p_mem_acc + exp1a_p_mem_dist + plot_annotation(
title = "Experiment 1A: Accuracy & Distribution of Memory Responses",
tag_levels = "A",
theme = patchwork_theme
)
```
```{r}
#| label: exp1a-pets-means
# mean and sd of accuracy for pet questions
exp1a_r_pet_means <- exp1a_d_all %>%
filter(M_Type == "pet") %>%
group_by(Pronoun) %>% # mean + sd for each pronoun
summarise(
mean = mean(M_Acc),
sd = sd(M_Acc)
) %>%
add_row( # mean + sd for all 3
Pronoun = "all",
mean = exp1a_d_all %>% filter(M_Type == "pet") %>% pull(M_Acc) %>% mean,
sd = exp1a_d_all %>% filter(M_Type == "pet") %>% pull(M_Acc) %>% sd,
) %>%
tidy_means()
exp1a_r_pet_means
```
```{r}
#| label: exp1a-pets-model
#| cache: true
# take just pet and pronoun memory questions
exp1a_d_pets <- exp1_load_data_pets(exp1a_d_all)
# mean-center contrast code with pet as negative and pronoun as positive
contrasts(exp1a_d_pets$M_Type)
contrasts(exp1a_d_pets$CharPronoun)
# find random effects structure
exp1a_m_pet <- buildmer(
formula = M_Acc ~ CharPronoun * M_Type +
(M_Type * CharPronoun | Participant) +
(M_Type * CharPronoun | Name),
data = exp1a_d_pets, family = binomial,
buildmerControl(direction = "order")
)
exp1a_r_pet <- exp1a_m_pet@model %>% tidy_model_results()
# dummy code pronoun to get question type in they/them characters only
exp1a_d_pets %>% count(CharPronoun, CharPronoun_They0)
exp1a_m_pet_they <- glmer(
formula = M_Acc ~ CharPronoun_They0 * M_Type + (M_Type | Participant),
data = exp1a_d_pets, family = binomial
)
exp1a_r_pet_they <- exp1a_m_pet_they %>% tidy_model_results()
# dummy code pronoun to get question type in he/she characters only
exp1a_d_pets %>% count(CharPronoun, CharPronoun_HeShe0)
exp1a_m_pet_heshe <- glmer(
formula = M_Acc ~ CharPronoun_HeShe0 * M_Type + (M_Type | Participant),
data = exp1a_d_pets, family = binomial
)
exp1a_r_pet_heshe <- exp1a_m_pet_heshe %>% tidy_model_results()
```
```{r}
#| label: exp1a-jobs-means
exp1a_r_job <- exp1a_d_all %>%
filter(M_Type == "job") %>%
summarise(
mean = mean(M_Acc) %>% format(digits = 2, nsmall = 2),
sd = sd(M_Acc) %>% format(digits = 2, nsmall = 2)
)
exp1a_r_job
```
Recall that just as each character was associated with one of three pronouns, each character was also associated with one of three pets. For they/them characters, pronoun accuracy was compared to pet accuracy (*M* = `r exp1a_r_pet_means['T', 'mean']`)---which exhibits similar distributional characteristics---showing no significant difference (`r exp1a_r_pet_they['M_Type=Pet_Pronoun', 'Text']`). Accuracy for the characters' 12 possible jobs (*M* = `r exp1a_r_job$mean`) was not at floor, suggesting that overall, the task was not too difficult for participants (@fig-exp1a-job-pet). The pet and job questions are discussed in more detail in the appendix (@sec-supplementary-exp1a).
#### Production
```{r}
#| label: exp1a-prod-counts
exp1a_tb_prod <- table(exp1a_d$Pronoun, exp1a_d$P_Response) %>%
prop.table() %>%
addmargins() %>%
round(2)
exp1a_tb_prod
```
```{r}
#| label: exp1a-prod-means
exp1a_r_prod_means <- exp1a_d %>%
group_by(Pronoun) %>%
summarise( # mean and SD for each pronoun
mean = mean(P_Acc),
sd = sd(P_Acc)
) %>%
add_row( # for he+she
Pronoun = "HS",
mean = exp1a_d %>% filter(Pronoun != "they/them") %>% pull(P_Acc) %>% mean,
sd = exp1a_d %>% filter(Pronoun != "they/them") %>% pull(P_Acc) %>% sd,
) %>%
add_row( # for all 3
Pronoun = "all",
mean = mean(exp1a_d$P_Acc),
sd = sd(exp1a_d$P_Acc),
) %>%
tidy_means() # add percentage, round values, fix labels
exp1a_r_prod_means
```
```{r}
#| label: exp1a-prod-model
#| cache: true
exp1a_m_prod <- buildmer(
formula = P_Acc ~ Pronoun + (Pronoun | Participant) + (Pronoun | Name),
data = exp1a_d, family = binomial,
buildmerControl(direction = "order")
)
summary(exp1a_m_prod)
exp1a_r_prod <- exp1a_m_prod@model %>% tidy_model_results()
```
```{r}
#| label: exp1a-use-they
# count how many times each participant used they/them in production task
exp1a_d_they <- exp1a_d %>%
mutate(P_IsThey = ifelse(P_Response == "they/them", 1, 0)) %>%
group_by(Participant) %>%
summarise(N_They = sum(P_IsThey)) %>%
mutate(
N_They_Cat = N_They %>%
as.factor() %>%
recode(
"6" = "6+", "7" = "6+", "8" = "6+", "9" = "6+",
"10" = "6+", "11" = "6+", "12" = "6+"
)
) %>%
mutate(Dummy = "")
# proportion of participant who produced they/them at least once
exp1a_r_they <- (
(exp1a_d_they %>% filter(N_They > 0) %>% n_distinct)
/ n_distinct(exp1a_d$Participant)
)
exp1a_r_they <- round(exp1a_r_they * 100, 0)
```
Responses were coded by whether the sentence continuation used he/him, she/her, they/them, or no pronouns to refer to the character (@fig-exp1a-prod). Responses that did not include a pronoun referring to the character either used the character's name as the subject of the continuing clause (e.g., *After Emily got home from working as a teacher...Emily made dinner*), or were ungrammatical continuations with no subject (e.g., *...made dinner*). Because these responses were infrequent (`r exp1a_tb_prod['Sum', 'none']*100`% of trials) and evenly distributed between pronoun conditions, they are included in the analysis (@tbl-exp1a-prod) as incorrect responses. Participants responded more accurately than inaccurately across all three pronoun conditions (`r exp1a_r_prod['Intercept', 'Text']`). He/him and she/her (*M* = `r exp1a_r_prod_means['HS', 'mean']`) were produced significantly more accurately than they/them (*M* = `r exp1a_r_prod_means['T', 'mean']`) (`r exp1a_r_prod['Pronoun=They_HeShe', 'Text']`). They/them was incorrectly produced as he/him or she/her each about one third of the time, while he/him and she/her were each incorrectly produced as they/them about an eighth of the time. Approximately `r exp1a_r_they`% of participants produced singular *they* at least once, regardless of accuracy.
| |
|-------------------------------|
| **Experiment 1A: Production** |
: Experiment 1A: Model results for the effect of Pronoun on Production Accuracy. {#tbl-exp1a-prod .borderless}
```{r}
#| label: table-exp1a-prod
#| output: true
exp1a_tb_prod <- tab_model(
model = exp1a_m_prod@model,
transform = NULL,
show.stat = TRUE, string.stat = "z",
show.ci = FALSE, show.se = TRUE, string.se = "SE",
show.r2 = FALSE, show.icc = FALSE,
digits = 3, digits.re = 3,
dv.labels = "Production Accuracy",
pred.labels = exp1_tb_fixed_labels,
wrap.labels = 80,
CSS = table_css
)
exp1a_tb_prod$knitr %<>%
exp1_tb_random_labels() %>%
str_replace( # bug with tab_model() makes it drop random slope labels
"ρ<sub>01</sub>",
"ρ<sub>01 Pronoun (They vs He + She) | Participant</sub>"
) %>%
str_replace(
'bottom:0.1cm;"></td>',
'bottom:0.1cm;">ρ<sub>01 Pronoun (He vs She) | Participant</sub></td>'
) %>%
drop_sigma()
exp1a_tb_prod
```
```{r}
#| label: fig-exp1a-prod
#| fig-cap: "Experiment 1A: [A] Pronoun accuracy in the written sentence completion task. By-participant means are shown as points; error bars indicate 95% CIs calculated over the by-participant means. [B] Distribution of pronoun production responses, with the correct pronoun on the X axis and the produced pronoun by color. [C] Number of times each participant produced singular *they* (correct = 4), regardless of accuracy."
#| fig-asp: 1
#| output: true
#| cache: true
# accuracy
exp1a_p_prod_acc <- exp1a_d %>%
group_by(Participant, Pronoun) %>%
summarise(P_Acc = mean(P_Acc)) %>%
ggplot(aes(x = Pronoun, y = P_Acc, fill = Pronoun, color = Pronoun)) +
stat_summary(
fun.data = mean_cl_boot, geom = "bar", alpha = 0.4, color = "NA"
) +
geom_point(
position = position_jitter(width = 0.35, height = 0.01, seed = 1),
size = 0.5
) +
stat_summary(
fun.data = mean_cl_boot, geom = "errorbar",
color = "black", linewidth = 0.75, width = 0.5
) +
scale_fill_brewer(palette = "Dark2") +
scale_color_brewer(palette = "Dark2") +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0.01, 0.01)) +
guides(fill = guide_none(), color = guide_none()) +
theme_classic() +
dissertation_plot_theme +
labs(
x = " ", # not blank to make margins line up with other plots
y = "By-Participant Mean Accuracy"
)
# distribution
exp1a_p_prod_dist <- exp1a_d %>%
ggplot(aes(x = Pronoun, fill = P_Response)) +
geom_bar(position = "fill") +
scale_fill_brewer(palette = "Dark2") +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
theme_classic() +
dissertation_plot_theme +
theme(
axis.text.x = element_text(size = 11, angle = 20, vjust = 0.75),
axis.title.x = element_text(margin = margin(t = 0)),
legend.margin = margin(l = 0)
) +
labs(
x = "Correct Pronoun",
y = "Proportion of Trials",
fill = "Pronoun\nProduced"
)
# number of they/them responses per participant
exp1a_p_prod_they <- exp1a_d_they %>%
ggplot(aes(x = Dummy, fill = N_They_Cat)) +
geom_bar(position = "fill") +
scale_fill_manual(values = c("#666666", brewer.pal(6, "Purples"))) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
theme_classic() +
dissertation_plot_theme +
theme(
axis.title.x = element_text(margin = margin(t = -10)),
legend.margin = margin(l = 0)
) +
labs(
x = "Number of They/Them\nResponses per Participant",
y = "Proportion of Participants",
fill = element_blank()
)
# combine
exp1a_p_prod_acc / (exp1a_p_prod_dist | exp1a_p_prod_they) +
plot_annotation(
title = "Experiment 1A: Accuracy & Distribution of Production Responses",
tag_levels = "A",
theme = patchwork_theme
)
```
#### Memory Predicting Production
```{r}
#| label: exp1a-mp-means
exp1a_r_mp_means <- exp1a_d %>%
group_by(Pronoun, M_Acc) %>%
summarise(mean = mean(P_Acc), sd = sd(P_Acc)) %>%
tidy_means()
exp1a_r_mp_means
```
```{r}
#| label: exp1a-mp-model
#| cache: true
# memory accuracy as mean-centered factor
contrasts(exp1a_d$M_Acc_Factor)
exp1a_m_mp <- buildmer(
formula = P_Acc ~ Pronoun * M_Acc_Factor +
(1 + Pronoun | Participant) + (1 + Pronoun | Name),
data = exp1a_d, family = binomial,
buildmerControl(direction = "order")
)
summary(exp1a_m_mp)
exp1a_r_mp <- exp1a_m_mp@model %>% tidy_model_results()
```
Extending the second model to test the effects of pronoun and memory accuracy on production accuracy (@tbl-exp1a-both) showed that participants were more likely to produce the correct pronoun when they had correctly remembered it (`r exp1a_r_mp['M_Acc=Wrong_Right', 'Text']`). Memory accuracy interacted with pronoun for the comparison between they/them and he/him + she/her (`r exp1a_r_mp['Pronoun=They_HeShe:M_Acc=Wrong_Right', 'Text']`), such that memory improved production more for they/them characters than for he/him + she/her characters (@fig-exp1a-both). Remembering but not producing they/them was more common than producing it but not remembering it.
| |
|-------------------------------------------------|
| **Experiment 1A: Memory Predicting Production** |
: Experiment 1A: Model results for the effects of Pronoun and Memory Accuracy on Production Accuracy. {#tbl-exp1a-both .borderless}
```{r}
#| label: table-exp1a-both
#| output: true
exp1a_tb_mp <- tab_model(
model = exp1a_m_mp@model,
transform = NULL,
show.stat = TRUE, string.stat = "z",
show.ci = FALSE, show.se = TRUE, string.se = "SE",
show.r2 = FALSE, show.icc = FALSE,
digits = 3, digits.re = 3,
dv.labels = "Production Accuracy",
pred.labels = exp1_tb_fixed_labels,
wrap.labels = 80,
CSS = table_css
)
exp1a_tb_mp$knitr %<>% exp1_tb_random_labels() %>% drop_sigma()
exp1a_tb_mp
```
```{r}
#| label: fig-exp1a-both
#| fig-cap: "Experiment 1A: [A] Production accuracy split by memory accuracy. Error bars indicate 95% CIs calculated over trials. [B] Distribution of combined memory and production accuracy."
#| fig-asp: 0.45
#| output: true
#| cache: true
# combined accuracy
exp1a_p_compare <- exp1a_d %>%
mutate(MP_Acc =
case_when(
M_Acc == 1 & P_Acc == 1 ~ "Both Right",
M_Acc == 1 & P_Acc == 0 ~ "Memory Only",
M_Acc == 0 & P_Acc == 1 ~ "Production Only",
M_Acc == 0 & P_Acc == 0 ~ "Both Wrong"
) %>%
factor(levels = c(
"Memory Only", "Production Only", "Both Wrong", "Both Right"
))
) %>%
ggplot(aes(x = Pronoun, fill = MP_Acc)) +
geom_bar(position = "fill") +
scale_fill_manual(
values = c("pink3", "#E6AB02", "tomato3", "#367ABF"),
labels = c("Memory\nOnly", "Production\nOnly", "Both\nWrong", "Both\nRight")
) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
theme_classic() +
dissertation_plot_theme +
theme(
axis.text.x = element_text(size = 11, angle = 25, vjust = .65),
legend.box.margin = margin(t = -6),
legend.text = element_text(margin = margin(t = 3.5, b = 3.5, l = 0, r = 0)),
legend.title = element_blank(),
plot.title = element_text(
size = 11,
margin = margin(t = -5),
face = "plain"
)
) +
labs(
title = "Combined Accuracy",
x = "Pronoun",
y = "Proportion of Characters",
fill = element_blank()
)
# production accuracy split by memory accuracy
exp1a_p_split <- exp1a_d %>%
ggplot(aes(x = Pronoun, y = P_Acc, fill = Pronoun, alpha = as.factor(M_Acc))) +
stat_summary(fun.data = mean_cl_boot, geom = "bar", position = "dodge") +
stat_summary(
fun.data = mean_cl_boot, geom = "errorbar",
position = position_dodge(width = 0.9),
color = "black", linewidth = 0.5, width = 0.5) +
scale_fill_brewer(palette = "Dark2") +
scale_alpha_discrete(
range = c(0.5, 1),
labels = c("Memory\nIncorrect", "Memory\nCorrect")
) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0), limits = c(0, 1)) +
guides(
alpha = guide_legend(override.aes = theme(color = NA)),
color = guide_none(),
fill = guide_none()) +
theme_classic() +
dissertation_plot_theme +
theme(
axis.text.x = element_text(size = 11, angle = 25, vjust = .65),
legend.box.margin = margin(l = -10),
legend.text = element_text(margin = margin(t = 3, b = 3, l = 0, r = 0)),
legend.title = element_blank(),
plot.title = element_text(
size = 11,
margin = margin(t = -5),
face = "plain"
)
) +
labs(
title = "Production Accuracy \nSplit By Memory Accuracy",
x = "Pronoun",
y = "Production Accuracy",
alpha = "Memory \nAccuracy"
)
# combine
exp1a_p_split + exp1a_p_compare + plot_annotation(
title = "Experiment 1A: Memory & Production",
tag_levels = "A",
theme = theme(
plot.title = element_text(face = "bold", size = 12),
plot.margin = margin(t = 0.1, b = 0, l = 0.1, r = 0, unit = "in")
)
)
```
## Experiment 1B
In the first experiment, people learned about a set of 12 characters whose pronouns---they/them or the expected he/him or she/her---could not be predicted from their name or other cues. These results demonstrate that, after only a brief exposure, people can learn that a character uses they/them pronouns and retrieve this information from memory to produce singular *they* in reference to them. Remembering that the character used they/them was a strong predictor of accurate pronoun production, with `r exp1a_r_mp_means['T Right', 'percent']` of production trials correct when the memory trial had been correct, but only `r exp1a_r_mp_means['T Wrong', 'percent']` of production trials correct when the memory trial had been incorrect. These results are congruent with a model where learning to use singular *they* requires a change from inferring a person's pronouns (*he* or *she*) based on semantic/conceptual features of a person or based on morphosyntactic gender associated with a person's name, and instead recalling episodic information about a person's stated pronouns.
```{r}
#| label: exp1a-task-setup
# pivot back to get 1 row per task
exp1a_d_long <- exp1a_d %>%
pivot_longer(
cols = c(M_Acc, P_Acc),
names_to = "Task",
values_to = "Acc"
)
# mean-center effects code Task
exp1a_d_long$Task %<>% as.factor()
contrasts(exp1a_d_long$Task) <- cbind("=M_P" = c(-.5, .5))
contrasts(exp1a_d_long$Task)
# add dummy codes for pronoun
exp1a_d_long %<>% mutate(
Pronoun_They0 = ifelse(Pronoun == "they/them", 0, 1),
Pronoun_HeShe0 = ifelse(Pronoun != "they/them", 0, 1)
)
```
```{r}
#| label: exp1a-task-model
#| cache: true
exp1a_m_task <- buildmer(
formula = Acc ~ Task * Pronoun +
(Pronoun | Participant) + (Pronoun | Name),
data = exp1a_d_long, family = binomial,
buildmerControl(direction = "order")
)
summary(exp1a_m_task)
exp1a_r_task <- exp1a_m_task@model %>% tidy_model_results()
exp1a_m_task_they <- glmer(
formula = Acc ~ Task * Pronoun_They0 + (Pronoun | Participant),
data = exp1a_d_long, family = binomial
)
summary(exp1a_m_task_they)
exp1a_r_task_they <- exp1a_m_task_they %>% tidy_model_results()
exp1a_m_task_heshe <- glmer(
formula = Acc ~ Task * Pronoun_HeShe0 + (Pronoun | Participant),
data = exp1a_d_long, family = binomial
)
summary(exp1a_m_task_heshe)
exp1a_r_task_heshe <- exp1a_m_task_heshe %>% tidy_model_results()
```
While memory accuracy did predict production accuracy, it was not a guarantee. For they/them characters, memory was significantly more accurate than production (`r exp1a_r_task_they['Task=M_P', 'Text']`). Conversely, for he/him and she/her characters, memory was significantly less accurate than production (`r exp1a_r_task_heshe['Task=M_P', 'Text']`) (@sec-supplementary-exp1a). I interpret these findings as demonstrating that memory for which pronouns a person uses and accuracy when producing a person's pronouns are distinct but interacting processes. However, another consideration is that the order of the memory and production tasks is in play. Consider that participants only had one brief exposure to information about the characters, then solved math problems for around five minutes, then answered 36 multiple-choice questions about each character's job, pet, and pronouns. By the time participants completed the production task, they may have begun to forget which characters used they/them pronouns, instead relying more on inferences from the characters' names. This would also result in higher production accuracy for he/him + she/her and lower production accuracy for they/them, but not necessarily due to different task demands between remembering which pronoun the character uses and selecting the correct pronoun to produce. A replication study that counterbalanced task order was conducted to ensure that the pattern of results in Experiment 1A was not determined by task order. While participants in Experiment 1A completed the memory task before the production task, participants in Experiment 1B completed the production task before the memory task. In all other respects, Experiment 1B was identical.
```{r}
#| label: exp1b-load-data
exp1b_d_all <- exp1_load_data_all("b") # all memory questions
exp1b_d <- exp1_load_data_pronoun("b") # pronoun questions with 1 row per char
summary(exp1b_d)
contrasts(exp1b_d$Pronoun)
```
```{r}
#| label: exp1-join-data
# join dataframes
exp1_d <- bind_rows(
.id = "Experiment",
"1A" = exp1a_d,
"1B" = exp1b_d
)
# mean-center effects code experiment
exp1_d$Experiment %<>% as.factor()
contrasts(exp1_d$Experiment) <- cbind("=A_B" = c(-.5, .5))
contrasts(exp1_d$Experiment)
# and add pronoun effects coding back
contrasts(exp1_d$Pronoun) <- cbind(
"=They_HeShe" = c(.33, .33, -.66),
"=He_She" = c(-.50, .50, 0)
)
contrasts(exp1_d$Pronoun)
```
### Methods
```{r}
#| label: exp1b-n-participants
exp1b_n_age <- read.csv("data/exp1b_data.csv") %>%
select(SubjID, SubjAge) %>%
unique() %>%
summarise(mean = mean(SubjAge), sd = sd(SubjAge)) %>%
round(2) %>%
format(n.small = 2)
exp1b_n_age
exp1b_n_gender <- read.csv("data/exp1b_data.csv") %>%
mutate(SubjGender_Group = case_when(
str_detect(SubjGender, "female") ~ "women",
str_detect(SubjGender, "man|male") ~ "men"
)) %>%
group_by(SubjGender_Group) %>%
summarise(n = n_distinct(SubjID)) %>%
column_to_rownames("SubjGender_Group")
exp1b_n_gender
exp1b_n_english <- read.csv("data/exp1b_data.csv") %>%
mutate(SubjEnglish_Group = case_when(
str_detect(SubjEnglish, "birth") ~ "native",
str_detect(SubjEnglish, "competent") ~ "competent",
str_detect(SubjEnglish, "limited") ~ "limited"
)) %>%
group_by(SubjEnglish_Group) %>%
summarise(n = n_distinct(SubjID)) %>%
column_to_rownames("SubjEnglish_Group")
exp1b_n_english
```
`r n_distinct(exp1b_d$Participant)` Vanderbilt undergraduates completed the study for partial course credit (*M*~age~ = `r exp1b_n_age$mean`, *SD*~age~ = `r exp1b_n_age$sd`). `r exp1b_n_gender['women', 'n']` were women, and `r exp1b_n_gender['men', 'n']` were men, as reported in a free response box. `r exp1b_n_english['native', 'n']` rated their English ability as "native (learned from birth)", `r exp1b_n_english['competent', 'n']` as "fully competent, but not native", and `r exp1b_n_english['limited', 'n']` as "limited but adequate competence." The only difference from Experiment 1A was switching the order of the memory and production tasks.
### Results
Responses were analyzed using the same logistic mixed-effects model specifications as in Experiment 1A [@bates2015; @rcoreteam2023; @voeten2023], although the maximal random effects structures that converged differed slightly in this data set. @fig-exp1ab-they highlights results for the they/them characters; results for all three pronouns, parallel to the figures in Experiment 1A, are included in the appendix (@fig-exp1ab-panel1, @fig-exp1ab-panel2). To compare these results to those of Experiment 1A, an additional trio of models with the combined data included Experiment as a mean-centered fixed effect (@sec-supplementary-exp1ab).
```{r}
#| label: fig-exp1ab-they
#| fig-cap: "[A] By-participant mean memory accuracy for they/them characters, split by task order. [B] By-participant mean production accuracy for they/them characters, split by task order. [C] Production accuracy for they/them characters, split by memory accuracy and task order. Error bars indicate 95% CIs."
#| fig-asp: 1
#| output: true
#| cache: true
# memory for they/them characters
exp1_p_memory_they <- exp1_d %>%
filter(Pronoun == "they/them") %>%
group_by(Experiment, Participant, Pronoun) %>%
summarise(M_Acc_Subj = mean(M_Acc)) %>%
ggplot(aes(
x = Experiment, y = M_Acc_Subj,
color = Experiment, fill = Experiment
)) +
stat_summary(
fun.data = mean_cl_boot, geom = "bar",
position = position_dodge(width = 0.5), color = "NA", alpha = 0.4
) +
geom_point(
position = position_jitter(width = 0.4, height = 0.01, seed = 1),
size = 0.5, show.legend = FALSE
) +
stat_summary(
fun.data = mean_cl_boot, geom = "errorbar",
position = position_dodge(width = 0.9),
linewidth = 0.5, width = 0.5, color = "black"
) +
geom_signif(
comparisons = list(c("1A", "1B")), map_signif_level = TRUE,
color = "black", tip_length = 0, textsize = 3, vjust = -0.2
) +
scale_color_manual(
values = c("grey50", "#7570B3"),
labels = c("1A:\nMemory\nFirst", "1B:\nProduction\nFirst")
) +
scale_fill_manual(
values = c("grey50", "#7570B3"),
labels = c("1A:\nMemory\nFirst", "1B:\nProduction\nFirst")
) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(
expand = c(0, 0), limits = c(-0.01, 1.15),
breaks = c(0, 0.25, 0.5, 0.75, 1)
) +
theme_classic() +
dissertation_plot_theme +
theme( # text sizes/spacing for this set are a bit different
axis.text.y = element_text(size = 9),
axis.ticks.y = element_line(),
axis.title.y = element_text(size = 9),
legend.text = element_text(size = 11, margin = margin(t = 3, b = 3)),
plot.title = element_text(size = 11)
) +
guides(fill = guide_legend(byrow = TRUE)) +
labs(
title = "Memory",
x = element_blank(),
y = "Mean Accuracy By Participant",
fill = "Experiment"
)
# production for they/them characters
exp1_p_prod_they <- exp1_d %>%
filter(Pronoun == "they/them") %>%
group_by(Experiment, Participant, Pronoun) %>%
summarise(P_Acc_Subj = mean(P_Acc)) %>%
ggplot(aes(
x = Experiment, y = P_Acc_Subj,
color = Experiment, fill = Experiment
)) +
stat_summary(
fun.data = mean_cl_boot, geom = "bar",
position = position_dodge(width = 0.5), color = "NA", alpha = 0.4
) +
geom_point(
position = position_jitter(width = 0.4, height = 0.01, seed = 1),
size = 0.5, show.legend = FALSE
) +
stat_summary(
fun.data = mean_cl_boot, geom = "errorbar",
position = position_dodge(width = 0.9),
linewidth = 0.5, width = 0.5, color = "black"
) +
geom_signif(
comparisons = list(c("1A", "1B")), map_signif_level = TRUE,
color = "black", tip_length = 0, textsize = 3, vjust = -0.2
) +
scale_color_manual(
values = c("grey50", "#7570B3"),
labels = c("1A:\nMemory\nFirst", "1B:\nProduction\nFirst")
) +
scale_fill_manual(
values = c("grey50", "#7570B3"),
labels = c("1A:\nMemory\nFirst", "1B:\nProduction\nFirst")
) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(
expand = c(0, 0), limits = c(-0.01, 1.15),
breaks = c(0, 0.25, 0.5, 0.75, 1)
) +
theme_classic() +
dissertation_plot_theme +
theme( # text sizes/spacing for this set are a bit different
axis.text.y = element_text(size = 9),
axis.ticks.y = element_line(),
axis.title.y = element_text(size = 9),
legend.text = element_text(size = 11, margin = margin(t = 3, b = 3)),
plot.title = element_text(size = 11)
) +
guides(fill = guide_legend(byrow = TRUE)) +
labs(
title = "Production",
x = element_blank(),
y = "Mean Accuracy By Participant",
fill = "Experiment"
)
# production split by memory for they/them characters
exp1_p_both_they <- exp1_d %>%
filter(Pronoun == "they/them") %>%
mutate(M_Acc_Label =
case_when(
M_Acc == 0 ~ "Memory\nIncorrect",
M_Acc == 1 ~ "Memory\nCorrect"
) %>%
factor(ordered = TRUE) %>%
fct_rev()
) %>%
ggplot(aes(
x = interaction(M_Acc_Label, Experiment),
y = P_Acc,
fill = Experiment, alpha = M_Acc_Label
)) +
stat_summary(
fun.data = mean_cl_boot, geom = "bar",
position = position_dodge(), color = "NA"
) +
stat_summary(
fun.data = mean_cl_boot, geom = "errorbar",
position = position_dodge(width = 0.9), key_glyph = "rect",
color = "black", linewidth = 0.5, width = 0.5
) +
geom_signif(
xmin = c(1, 3, 1.5), xmax = c(2, 4, 3.5), y = c(0.75, 0.75, 0.9),
annotations = c("***", "***", "NS."), color = "black", alpha = 1,
tip_length = 0, textsize = 3, vjust = -0.2
) +
scale_alpha_manual(values = c(0.5, 1)) +
scale_fill_manual(values = c("grey50", "#7570B3")) +
scale_x_discrete(expand = c(0, 0), guide = "axis_nested") +
scale_y_continuous(expand = c(0, 0), limits = c(0, 1)) +
theme_classic() +
dissertation_plot_theme +
theme( # text sizes/spacing for this set are a bit different
axis.text.y = element_text(size = 9),
axis.ticks.y = element_line(),
axis.title.y = element_text(size = 9),
plot.title = element_text(size = 11)
) +
guides(alpha = guide_none(), fill = guide_none()) +
labs(
title = "Production Split By Memory",
x = element_blank(),
y = "Production Accuracy"
)
# combine
exp1_p_memory_they +
exp1_p_prod_they +
exp1_p_both_they +
guide_area() +
plot_layout(
guides = "collect",
design = "AABB
CCCD"
) +
plot_annotation(
title = "Experiments 1A & 1B: Accuracy for They/Them Characters",
theme = patchwork_theme
) +
plot_annotation(theme = theme(
plot.margin = margin(t = 10, b = -5, l = 10, r = 10)
))
```
#### Memory
```{r}
#| label: exp1b-mem-means
exp1b_r_memory_means <- exp1b_d %>%
group_by(Pronoun) %>%
summarise( # mean and SD for each pronoun
mean = mean(M_Acc),
sd = sd(M_Acc)
) %>%