Semantic Learning & Recognition Logistic Regression Analyses
Ashlie
2025-04-07
Semantic Learning Task
Data Manipulations
# Load data (1 data frame)
slid_data <- read.csv(here("Semantic Learning & ID Data","Semantic Learning & ID Data","Semantic learning and recognition data.csv"),stringsAsFactors=TRUE) # sentence length data
slid_data %<>%
mutate(Participant.ID = factor(Participant.ID),
Group = factor(Child.or.adult),
SL.Meaning.Condition = factor(SL.Meaning.Condition),
SL.Correct.Response = factor(SL.Correct.Response),
SL.Triplet.Number = as.numeric(SL.Triplet.Number),
SL.Trial.Order.by.Task = as.numeric(SL.Trial.Order.by.Task),
SL.Response = factor(SL.Response),
SL.Response.Type = factor(SL.Response.Type),
SL.Correct.Based.on.Only.SL. = as.numeric(SL.Correct.Based.on.Only.SL.),
SL.Correct.Enough.for.IT. = as.numeric(SL.Correct.Enough.for.IT.),
IT.Trial.Number = as.numeric(IT.Trial.Number),
IT.Trial.Order.by.Task = as.numeric(IT.Trial.Order.by.Task),
IT.Correct.Response = as.numeric(IT.Correct.Response),
IT.Response = as.numeric(IT.Response),
IT.Correct. = as.numeric(IT.Correct.),
Exclude. = as.numeric(Exclude.),
SL.Nonword = factor(SL.Nonword),
SL.meaning.cond.c = as.numeric(SL.Meaning.Condition) - 1.5)
str(slid_data)## 'data.frame': 4680 obs. of 36 variables:
## $ Participant.ID : Factor w/ 156 levels "049b2dc2","070821_1f_10",..: 113 143 47 93 51 131 128 155 152 145 ...
## $ Child.or.adult : Factor w/ 2 levels "A","C": 1 1 1 1 1 1 1 1 1 1 ...
## $ SL.block.or.condition : Factor w/ 31 levels "apple_block",..: 25 25 25 25 25 25 25 25 25 25 ...
## $ Task : Factor w/ 1 level "SL": 1 1 1 1 1 1 1 1 1 1 ...
## $ SL.Triplet.Number : num 30 30 30 30 16 16 16 18 18 18 ...
## $ SL.Trial.Order.by.Task : num 4 22 24 27 3 11 19 4 8 13 ...
## $ SL.Meaning.Condition : Factor w/ 2 levels "M-","M+": 2 2 2 2 2 2 2 2 2 2 ...
## $ SL.Nonword : Factor w/ 31 levels "bov","chaws",..: 21 21 21 21 18 18 18 15 15 15 ...
## $ SL.Correct.Response : Factor w/ 17 levels "apple","bed",..: 1 1 1 1 4 4 4 3 3 3 ...
## $ SL.Response : Factor w/ 263 levels ""," "," nice",..: 7 163 7 7 175 97 208 20 106 20 ...
## $ SL.Response.Category : Factor w/ 3 levels "correct word",..: 1 3 1 1 2 2 3 1 2 1 ...
## $ SL.Correct.Based.on.Only.SL. : num 1 1 1 1 0 0 0 1 0 1 ...
## $ SL.Correct.Enough.for.IT. : num 1 0 1 1 0 0 0 1 0 1 ...
## $ SL.Response.Type : Factor w/ 6 levels "correct response (no meaning)",..: 2 6 2 2 4 4 6 2 4 2 ...
## $ IT.Block : Factor w/ 31 levels "ID_apple","ID_bed",..: 31 31 31 31 31 31 31 31 31 31 ...
## $ IT.Trial.Type : Factor w/ 30 levels "recog1","recog10",..: 1 1 1 1 12 12 12 23 23 23 ...
## $ Task.1 : Factor w/ 1 level "IT": 1 1 1 1 1 1 1 1 1 1 ...
## $ IT.Trial.Number : num 1 1 1 1 2 2 2 3 3 3 ...
## $ IT.Trial.Order.by.Task : num 1 1 1 1 1 1 1 1 1 1 ...
## $ IT.Meaning.Condition : Factor w/ 2 levels "M-","M+": 2 2 2 2 2 2 2 2 2 2 ...
## $ IT.Nonword : Factor w/ 30 levels "bov","chaws",..: 20 20 20 20 17 17 17 15 15 15 ...
## $ IT.Correct.Response : num 1 1 1 1 3 3 3 2 2 2 ...
## $ IT.Response : num 3 3 1 5 1 5 4 1 5 3 ...
## $ IT.Correct. : num 0 0 1 0 0 0 0 0 0 0 ...
## $ IT.Response.Type : Factor w/ 11 levels "#N/A","correct response (M+ target word)",..: 4 4 2 9 4 9 10 11 9 10 ...
## $ Semantic.vs.Other : Factor w/ 3 levels "#N/A","other foil",..: 2 2 NA NA 2 NA 3 2 NA 3 ...
## $ Familiar.vs.Unrelated : Factor w/ 3 levels "#N/A","familiar foil",..: 2 2 NA NA 2 NA NA 3 NA NA ...
## $ Language.Status : Factor w/ 8 levels "","bilingual (English, Chinese)",..: 7 7 7 7 7 6 2 7 7 5 ...
## $ Exclude. : num 0 0 0 1 1 0 0 0 0 0 ...
## $ Exclusion.Reason : Factor w/ 18 levels "","Completed portion of SL task before restarting and completing whole study",..: 1 1 1 12 4 1 1 1 1 1 ...
## $ Age : num 26 24 26 29 25 27 25 25 28 24 ...
## $ Gender : Factor w/ 6 levels "","3","5","female",..: 4 5 4 5 5 4 4 5 5 5 ...
## $ Highest.Completed.Education.Level: Factor w/ 8 levels "","associate's degree",..: 7 2 2 2 4 3 3 5 5 7 ...
## $ Highest.Degree.Earned : Factor w/ 24 levels "","B.A","B.A.",..: 1 1 1 1 20 2 3 15 15 1 ...
## $ Group : Factor w/ 2 levels "A","C": 1 1 1 1 1 1 1 1 1 1 ...
## $ SL.meaning.cond.c : num 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 ...# Create some data subsets
incl_slid_data <- subset(slid_data, Exclude. != 1 & !grepl("bilingual", Language.Status, ignore.case = TRUE)) # only included, monolingual participants
incl_slid_data %<>%
mutate(Age = as.numeric(Age),
age.c = as.numeric(scale(Age, center = TRUE, scale = FALSE)),
Gender = factor(Gender),
gender.c = as.numeric(Gender) - 1.5)
str(incl_slid_data)## 'data.frame': 3150 obs. of 38 variables:
## $ Participant.ID : Factor w/ 156 levels "049b2dc2","070821_1f_10",..: 113 143 47 155 152 153 140 81 114 57 ...
## $ Child.or.adult : Factor w/ 2 levels "A","C": 1 1 1 1 1 1 1 1 1 1 ...
## $ SL.block.or.condition : Factor w/ 31 levels "apple_block",..: 25 25 25 25 25 25 25 25 25 25 ...
## $ Task : Factor w/ 1 level "SL": 1 1 1 1 1 1 1 1 1 1 ...
## $ SL.Triplet.Number : num 30 30 30 18 18 17 19 19 19 20 ...
## $ SL.Trial.Order.by.Task : num 4 22 24 4 8 11 3 13 24 21 ...
## $ SL.Meaning.Condition : Factor w/ 2 levels "M-","M+": 2 2 2 2 2 2 2 2 2 2 ...
## $ SL.Nonword : Factor w/ 31 levels "bov","chaws",..: 21 21 21 15 15 6 16 16 16 14 ...
## $ SL.Correct.Response : Factor w/ 17 levels "apple","bed",..: 1 1 1 3 3 14 17 17 17 15 ...
## $ SL.Response : Factor w/ 263 levels ""," "," nice",..: 7 163 7 20 106 221 225 251 225 243 ...
## $ SL.Response.Category : Factor w/ 3 levels "correct word",..: 1 3 1 1 2 1 1 1 1 1 ...
## $ SL.Correct.Based.on.Only.SL. : num 1 1 1 1 0 1 1 1 1 1 ...
## $ SL.Correct.Enough.for.IT. : num 1 0 1 1 0 1 1 1 1 1 ...
## $ SL.Response.Type : Factor w/ 6 levels "correct response (no meaning)",..: 2 6 2 2 4 2 2 2 2 2 ...
## $ IT.Block : Factor w/ 31 levels "ID_apple","ID_bed",..: 31 31 31 31 31 31 31 31 31 31 ...
## $ IT.Trial.Type : Factor w/ 30 levels "recog1","recog10",..: 1 1 1 23 23 25 26 26 26 27 ...
## $ Task.1 : Factor w/ 1 level "IT": 1 1 1 1 1 1 1 1 1 1 ...
## $ IT.Trial.Number : num 1 1 1 3 3 4 5 5 5 6 ...
## $ IT.Trial.Order.by.Task : num 1 1 1 1 1 1 1 1 1 1 ...
## $ IT.Meaning.Condition : Factor w/ 2 levels "M-","M+": 2 2 2 2 2 2 2 2 2 2 ...
## $ IT.Nonword : Factor w/ 30 levels "bov","chaws",..: 20 20 20 15 15 6 16 16 16 14 ...
## $ IT.Correct.Response : num 1 1 1 2 2 4 1 1 1 2 ...
## $ IT.Response : num 3 3 1 1 5 4 1 1 1 2 ...
## $ IT.Correct. : num 0 0 1 0 0 1 1 1 1 1 ...
## $ IT.Response.Type : Factor w/ 11 levels "#N/A","correct response (M+ target word)",..: 4 4 2 11 9 2 2 2 2 2 ...
## $ Semantic.vs.Other : Factor w/ 3 levels "#N/A","other foil",..: 2 2 NA 2 NA NA NA NA NA NA ...
## $ Familiar.vs.Unrelated : Factor w/ 3 levels "#N/A","familiar foil",..: 2 2 NA 3 NA NA NA NA NA NA ...
## $ Language.Status : Factor w/ 8 levels "","bilingual (English, Chinese)",..: 7 7 7 7 7 7 7 7 7 7 ...
## $ Exclude. : num 0 0 0 0 0 0 0 0 0 0 ...
## $ Exclusion.Reason : Factor w/ 18 levels "","Completed portion of SL task before restarting and completing whole study",..: 1 1 1 1 1 1 1 1 1 1 ...
## $ Age : num 26 24 26 25 28 23 25 24 25 27 ...
## $ Gender : Factor w/ 2 levels "female","male": 1 2 1 2 2 2 1 2 1 1 ...
## $ Highest.Completed.Education.Level: Factor w/ 8 levels "","associate's degree",..: 7 2 2 5 5 7 3 3 3 3 ...
## $ Highest.Degree.Earned : Factor w/ 24 levels "","B.A","B.A.",..: 1 1 1 15 15 1 3 9 7 13 ...
## $ Group : Factor w/ 2 levels "A","C": 1 1 1 1 1 1 1 1 1 1 ...
## $ SL.meaning.cond.c : num 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 ...
## $ age.c : num 6.71 4.71 6.71 5.71 8.71 ...
## $ gender.c : num -0.5 0.5 -0.5 0.5 0.5 0.5 -0.5 0.5 -0.5 -0.5 ...mp_slid_data <- subset(incl_slid_data, SL.Meaning.Condition == "M+") # only M+ trials
# Calculate the proportions of responses in the Semantic Learning task for each participant
# Step 1: Define valid response types per condition
valid_responses <- tribble(
~SL.Meaning.Condition, ~SL.Response.Type,
"M+", "correct response (target word)",
"M+", "incorrect response (no meaning)",
"M+", "incorrect response (no response)",
"M+", "incorrect response (wrong word)",
"M-", "correct response (no meaning)",
"M-", "incorrect response (any meaning)",
"M-", "incorrect response (no response)"
)
# Step 2: Summarise participant-level counts
participant_counts <- incl_slid_data %>%
group_by(Participant.ID, Group, SL.Meaning.Condition, SL.Response.Type) %>%
summarise(n = n(), .groups = "drop")
# Step 3: Get unique Participant × Condition combos from actual data
participant_conditions <- incl_slid_data %>%
distinct(Participant.ID, Group, SL.Meaning.Condition)
# Step 4: Create complete grid of valid combinations per participant
full_grid <- participant_conditions %>%
inner_join(valid_responses, by = "SL.Meaning.Condition")## Warning in inner_join(., valid_responses, by = "SL.Meaning.Condition"): Detected an unexpected many-to-many relationship between `x` and `y`.
## ℹ Row 1 of `x` matches multiple rows in `y`.
## ℹ Row 1 of `y` matches multiple rows in `x`.
## ℹ If a many-to-many relationship is expected, set `relationship =
## "many-to-many"` to silence this warning.# Step 5: Left join actual counts onto full grid (so missing = 0)
propsl_data <- full_grid %>%
left_join(participant_counts,
by = c("Participant.ID", "Group", "SL.Meaning.Condition", "SL.Response.Type")) %>%
mutate(n = replace_na(n, 0)) %>%
group_by(Participant.ID, Group, SL.Meaning.Condition) %>%
mutate(
total_responses = sum(n),
proportion = n / total_responses
) %>%
ungroup()
# Notes:
# SL Meaning Condition = context condition (M- = unsupportive, M+ = supportive); default reference level = unsupportive/M-
# For SL Meaning Condition, M+ = -0.5 (after centering), M- = 0.5 (after centering)
# For Gender, female = -0.5 (after centering), male = 0.5 (after center)Sanity checks
# Is each participant in only one group?
# Expected: TRUE (between-subjects design)
slid_data %>% group_by(Participant.ID) %>%
summarise(Group = n_distinct(Group)) %>%
as.data.frame() %>%
{.$Group == 1} %>%
all()## [1] TRUE# Number of included participants per experimental condition
# Expected: Children = 45, Adults = 69
incl_slid_data %>% group_by(Participant.ID) %>%
summarise(Group = first(Group)) %>%
ungroup() %>%
group_by(Group) %>%
summarise(n = n())## # A tibble: 2 × 2
## Group n
## <fct> <int>
## 1 A 60
## 2 C 45# Number of different target words (collapsed across participants) by group
# Expected: 15
mp_slid_data %>%
group_by(Group) %>%
summarise(items = n_distinct(SL.Correct.Response))## # A tibble: 2 × 2
## Group items
## <fct> <int>
## 1 A 15
## 2 C 15 #summarise(items = n_distinct(Distinct.Target))
# Number of different psuedowords (collapsed across participants) by group
# Expected: 15
mp_slid_data %>%
group_by(Group) %>%
summarise(items = n_distinct(SL.Nonword))## # A tibble: 2 × 2
## Group items
## <fct> <int>
## 1 A 15
## 2 C 15Participants
Inclusions/exclusions
slid_data %>%
distinct(Participant.ID, Group, Language.Status, Exclude., Exclusion.Reason)%>%
group_by(Exclude.,Group, Language.Status, Exclusion.Reason)%>%
summarise(n = n())%>%
mutate(total.Exclusion.Reason.n = sum(n))%>%
knitr::kable()## `summarise()` has grouped output by 'Exclude.', 'Group', 'Language.Status'. You
## can override using the `.groups` argument.| Exclude. | Group | Language.Status | Exclusion.Reason | n | total.Exclusion.Reason.n |
|---|---|---|---|---|---|
| 0 | A | bilingual (English, Chinese) | 2 | 2 | |
| 0 | A | bilingual (English, Korean) | 1 | 1 | |
| 0 | A | bilingual (English, Spanish) | 5 | 5 | |
| 0 | A | bilingual (English, Vietnamese) | 1 | 1 | |
| 0 | A | monolingual | 60 | 60 | |
| 0 | C | 45 | 45 | ||
| 1 | A | bilingual (English, French) | This participant provided the same 2 answers for almost all of the R/IT trials | 1 | 1 |
| 1 | A | monolingual | Participant didn’t seem to understand WLT (provided nonwords as responses) | 1 | 29 |
| 1 | A | monolingual | Repeated attempts by Mturk workers that should be excluded | 9 | 29 |
| 1 | A | monolingual | This participant did not complete the study | 1 | 29 |
| 1 | A | monolingual | This participant did not respond to almost any of the R/IT trials (let almost all of them time out) | 1 | 29 |
| 1 | A | monolingual | This participant did not respond to any of the R/IT trials (let them all time out) | 1 | 29 |
| 1 | A | monolingual | This participant selected “no meaning” for most of the R/IT trials | 3 | 29 |
| 1 | A | monolingual | This participant spent ten or less minutes | 10 | 29 |
| 1 | A | monolingual | This participant spent ten or less minutes and provided the same 2 answers for almost all of the answers in the WLT and most of the R/IT | 1 | 29 |
| 1 | A | monolingual | This participant spent ten or less minutes, didn’t seem to understand WLT (provided repeated, nonsensical responses), and selected “no meaning” for all of the R/IT trials | 1 | 29 |
| 1 | A | monolingual | This participant spent ten or less minutes, provided the same 2 answers for almost all of the answers in the WLT, and provided repeated answers for much of R/IT | 1 | 29 |
| 1 | A | N/A | Repeated attempts by Mturk workers that should be excluded | 1 | 11 |
| 1 | A | N/A | This participant did not fill out the demographics | 4 | 11 |
| 1 | A | N/A | This participant did not fill out the demographics and selected “no meaning” for all of the R/IT trials | 1 | 11 |
| 1 | A | N/A | This participant spent ten or less minutes | 2 | 11 |
| 1 | A | N/A | This participant spent ten or less minutes and did not fill out the demographics | 1 | 11 |
| 1 | A | N/A | This participant spent ten or less minutes, provided repeated answers for much of WLT, and selected “no meaning” for all of the R/IT trials | 1 | 11 |
| 1 | A | N/A | This was a practice run | 1 | 11 |
| 1 | C | Completed portion of SL task before restarting and completing whole study | 1 | 1 |
Age and gender demographics
# Age
incl_slid_data %>%
distinct(Participant.ID, Group, Age)%>%
group_by(Group)%>%
summarise_at('Age',
list(~mean(., na.rm=T),
~sd(., na.rm=T),
~median(., na.rm=T),
~min(., na.rm=T),
~max(., na.rm=T),
~sum(!is.na(.))))%>%
dplyr::rename("n" = "sum")%>%
dplyr::select(Group, n, mean, sd, median, min, max)%>%
mutate(total.n = sum(n))%>%
knitr::kable()| Group | n | mean | sd | median | min | max | total.n |
|---|---|---|---|---|---|---|---|
| A | 60 | 25.38333 | 1.7083213 | 25 | 22 | 29 | 105 |
| C | 45 | 11.15556 | 0.7964568 | 11 | 10 | 12 | 105 |
# Gender
incl_slid_data %>%
distinct(Participant.ID, Group, Gender)%>%
group_by(Group, Gender)%>%
summarise(n = n())%>%
mutate(total.Gender.n = sum(n))%>%
knitr::kable()## `summarise()` has grouped output by 'Group'. You can override using the
## `.groups` argument.| Group | Gender | n | total.Gender.n |
|---|---|---|---|
| A | female | 33 | 60 |
| A | male | 27 | 60 |
| C | female | 26 | 45 |
| C | male | 19 | 45 |
Descriptives
# Semantic Learning task performance
incl_slid_data_partsum <- incl_slid_data %>%
group_by(Participant.ID, Group,SL.Meaning.Condition) %>%
summarise(SL_corr_partacc = mean(SL.Correct.Enough.for.IT., na.rm = TRUE), .groups = "drop")
sl_summary_df <- incl_slid_data_partsum %>%
group_by(Group,SL.Meaning.Condition) %>%
summarise(
mean_accuracy = mean(SL_corr_partacc),
sd_accuracy = sd(SL_corr_partacc),
n = n(),
.groups = "drop"
) %>%
mutate(
label_y = mean_accuracy + sd_accuracy + 0.02, # Add small buffer
total_n = sum(n)
)
knitr::kable(sl_summary_df)| Group | SL.Meaning.Condition | mean_accuracy | sd_accuracy | n | label_y | total_n |
|---|---|---|---|---|---|---|
| A | M- | 0.9611111 | 0.0697498 | 60 | 1.0508609 | 210 |
| A | M+ | 0.7755556 | 0.1472749 | 60 | 0.9428304 | 210 |
| C | M- | 0.9570370 | 0.0767464 | 45 | 1.0537834 | 210 |
| C | M+ | 0.6459259 | 0.1757076 | 45 | 0.8416336 | 210 |
# Proportions of Semantic Learning task
propsl_data_partsum <- propsl_data %>%
group_by(Participant.ID, Group,SL.Meaning.Condition, SL.Response.Type) %>%
summarise(proportion_partacc = mean(proportion, na.rm = TRUE), .groups = "drop")
propsl_summary_df <- propsl_data_partsum %>%
group_by(Group, SL.Meaning.Condition, SL.Response.Type) %>%
summarise(
mean_proportion = mean(proportion_partacc),
sd_proportion = sd(proportion_partacc),
n = n(),
.groups = "drop"
) %>%
mutate(
label_y = mean_proportion + sd_proportion + 0.02, # Add small buffer
total_n = sum(n)
)
knitr::kable(propsl_summary_df)| Group | SL.Meaning.Condition | SL.Response.Type | mean_proportion | sd_proportion | n | label_y | total_n |
|---|---|---|---|---|---|---|---|
| A | M+ | correct response (target word) | 0.7755556 | 0.1472749 | 60 | 0.9428304 | 735 |
| A | M+ | incorrect response (no meaning) | 0.1388889 | 0.1395671 | 60 | 0.2984560 | 735 |
| A | M+ | incorrect response (no response) | 0.0000000 | 0.0000000 | 60 | 0.0200000 | 735 |
| A | M+ | incorrect response (wrong word) | 0.0855556 | 0.0650559 | 60 | 0.1706114 | 735 |
| A | M- | correct response (no meaning) | 0.9622222 | 0.0653735 | 60 | 1.0475957 | 735 |
| A | M- | incorrect response (any meaning) | 0.0377778 | 0.0653735 | 60 | 0.1231513 | 735 |
| A | M- | incorrect response (no response) | 0.0000000 | 0.0000000 | 60 | 0.0200000 | 735 |
| C | M+ | correct response (target word) | 0.7555556 | 0.1669694 | 45 | 0.9425250 | 735 |
| C | M+ | incorrect response (no meaning) | 0.1422222 | 0.1106455 | 45 | 0.2728677 | 735 |
| C | M+ | incorrect response (no response) | 0.0014815 | 0.0099381 | 45 | 0.0314196 | 735 |
| C | M+ | incorrect response (wrong word) | 0.1007407 | 0.1021212 | 45 | 0.2228619 | 735 |
| C | M- | correct response (no meaning) | 0.9570370 | 0.0767464 | 45 | 1.0537834 | 735 |
| C | M- | incorrect response (any meaning) | 0.0414815 | 0.0769217 | 45 | 0.1384032 | 735 |
| C | M- | incorrect response (no response) | 0.0014815 | 0.0099381 | 45 | 0.0314196 | 735 |
Visualizations
# Semantic Learning task performance
incl_slid_data_partsum %>%
mutate(#SL.Meaning.Condition = factor(SL.Meaning.Condition, levels = c("M-", "M+"),
#labels = c("Unsupportive Context", "Supportive Context"))),
SL.Meaning.Condition = factor(SL.Meaning.Condition, levels = c("M+","M-")) # Set the desired order of variables
)%>%
ggplot(aes(x = SL.Meaning.Condition, y = SL_corr_partacc, fill = SL.Meaning.Condition, color = SL.Meaning.Condition)) +
stat_summary(fun = mean, position = position_dodge(width = .9),
geom="bar", colour = "black") +
geom_hline(yintercept = .5, color = "grey", linetype = "dashed") +
facet_grid(~Group) +
stat_summary(fun.data = mean_sdl, fun.args = list(mult = 1),
geom = "errorbar",
position = position_dodge(width = 0.9),
size = 1, width = 0.25, color = "black") + # Standard deviation bars
geom_text(
data = sl_summary_df,
aes(
x = SL.Meaning.Condition,
y = label_y,
label = sprintf("%.2f", mean_accuracy),
group = Group
),
position = position_dodge(width = 0.9),
size = 4,
color = "black") + # Add formatted text on top of bars
# scale_fill_brewer(palette = "Dark2") +
#langcog::scale_fill_solarized()+
labs(y = "Mean Accuracy of Target Meaning/Word Identification",
x = "Group",
fill = "Meaning Condition") +
theme(axis.text.x = element_text(hjust = 0.5, angle = 0, size = 12, color = "black"), # Customize x-axis text
axis.text.y = element_text(size = 12, color = "black"),
axis.title.x = element_text(size = 14, color = "black", face = "bold", margin = margin(t = 10)), # x-axis title
axis.title.y = element_text(size = 14, color = "black", face = "bold", margin = margin(r = 10)), # y-axis title
strip.text = element_text(size = 12, color = "black", face = "bold"), # Facet labels
legend.key.size = unit(.5, 'cm'),
legend.text = element_text(size = 10),
legend.title = element_text(size = 11),
legend.margin = margin(t = 0, unit = 'cm'),
plot.title = element_text(size = 12)) +
guides(fill = FALSE, color = FALSE) + # Remove both fill and color legends
scale_y_continuous(breaks = seq(0, 1, .25))## Warning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.
## ℹ Please use `linewidth` instead.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.## Warning: The `<scale>` argument of `guides()` cannot be `FALSE`. Use "none" instead as
## of ggplot2 3.3.4.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.
# Proportions of Semantic Learning task
## M+
mp_resp_types <- c(
"correct response (target word)",
"incorrect response (no meaning)",
"incorrect response (no response)",
"incorrect response (wrong word)"
)
propsl_data_partsum %>%
filter(
SL.Meaning.Condition == "M+",
SL.Response.Type %in% mp_resp_types
) %>%
mutate(SL.Response.Type = factor(SL.Response.Type, levels = mp_resp_types)) %>%
droplevels() %>%
ggplot(aes(x = SL.Response.Type, y = proportion_partacc, fill = Group)) +
stat_summary(fun = mean, position = position_dodge(width = .9),
geom="bar", colour = "black") +
geom_hline(yintercept = .5, color = "grey", linetype = "dashed") +
facet_wrap(~ Group, scales = "free_y", drop = TRUE) + # Drop unused categories
stat_summary(fun.data = mean_sdl, fun.args = list(mult = 1),
geom = "errorbar",
position = position_dodge(width = 0.9),
size = 1, width = 0.25, color = "black") + # Standard deviation bars
scale_x_discrete(limits = mp_resp_types) + # <-- THIS IS KEY
geom_text(
data = propsl_summary_df,
aes(
x = SL.Response.Type,
y = label_y,
label = sprintf("%.2f", mean_proportion),
group = Group
),
position = position_dodge(width = 0.9),
size = 4,
color = "black"
) + # Add formatted text on top of bars
labs(title = "Proportions of M+ Semantic Learning Task Responses",
x = "Response Type",
y = "Mean Proportion of Responses",
fill = "Group"
) +
theme(
axis.text.x = element_text(hjust = 1, angle = 45, size = 12, color = "black"),
axis.text.y = element_text(size = 12, color = "black"),
axis.title.x = element_text(size = 14, color = "black", face = "bold", margin = margin(t = 10)),
axis.title.y = element_text(size = 14, color = "black", face = "bold", margin = margin(r = 10)),
strip.text = element_text(size = 12, color = "black", face = "bold"),
legend.key.size = unit(.5, 'cm'),
legend.text = element_text(size = 11),
legend.title = element_text(size = 12, face = "bold"),
legend.margin = margin(t = 0, unit = 'cm'),
plot.title = element_text(size = 14, face = "bold")
) +
scale_y_continuous(breaks = seq(0, 1, .25))## Warning: Removed 4 rows containing missing values or values outside the scale range
## (`geom_text()`).
## M-
mm_resp_types <- c(
"correct response (no meaning)",
"incorrect response (any meaning)",
"incorrect response (no response)"
)
propsl_data_partsum %>%
filter(
SL.Meaning.Condition == "M-",
SL.Response.Type %in% mm_resp_types
) %>%
mutate(SL.Response.Type = factor(SL.Response.Type, levels = mm_resp_types)) %>%
droplevels() %>%
ggplot(aes(x = SL.Response.Type, y = proportion_partacc, fill = Group)) +
stat_summary(fun = mean, position = position_dodge(width = .9),
geom="bar", colour = "black") +
geom_hline(yintercept = .5, color = "grey", linetype = "dashed") +
facet_wrap(~ Group, scales = "free_y", drop = TRUE) + # Drop unused categories
stat_summary(fun.data = mean_sdl, fun.args = list(mult = 1),
geom = "errorbar",
position = position_dodge(width = 0.9),
size = 1, width = 0.25, color = "black") + # Standard deviation bars
scale_x_discrete(limits = mm_resp_types) + # <-- THIS IS KEY
geom_text(
data = propsl_summary_df,
aes(
x = SL.Response.Type,
y = label_y,
label = sprintf("%.2f", mean_proportion),
group = Group
),
position = position_dodge(width = 0.9),
size = 4,
color = "black"
) + # Add formatted text on top of bars
labs(title = "Proportions of M- Semantic Learning Task Responses",
x = "Response Type",
y = "Mean Proportion of Responses",
fill = "Group"
) +
theme(
axis.text.x = element_text(hjust = 1, angle = 45, size = 12, color = "black"),
axis.text.y = element_text(size = 12, color = "black"),
axis.title.x = element_text(size = 14, color = "black", face = "bold", margin = margin(t = 10)),
axis.title.y = element_text(size = 14, color = "black", face = "bold", margin = margin(r = 10)),
strip.text = element_text(size = 12, color = "black", face = "bold"),
legend.key.size = unit(.5, 'cm'),
legend.text = element_text(size = 11),
legend.title = element_text(size = 12, face = "bold"),
legend.margin = margin(t = 0, unit = 'cm'),
plot.title = element_text(size = 14, face = "bold")
) +
scale_y_continuous(breaks = seq(0, 1, .25))## Warning: Removed 6 rows containing missing values or values outside the scale range
## (`geom_text()`).
Recognition Task
Data Manipulations
# Create some data subsets (from previously loaded dataset)
mp_slid_data <- subset(incl_slid_data, SL.Meaning.Condition == "M+") # only M+ trials
learned_slid_data <- subset(mp_slid_data, SL.Correct.Enough.for.IT. == 1) # only M+ trials where nonsense words' target meanings were IDed
mp_foil_data <- subset(mp_slid_data, !is.na(Semantic.vs.Other)) # only M+ trials where semantically-related, unrelated, or familiar foils were selected
learned_foil_data <- subset(learned_slid_data, !is.na(Semantic.vs.Other)) # only M+ trials where nonsense words' target meanings were IDed AND then only either a semantically-related, unrelated, or familiar foil was selected
# Calculate the proportions of responses in the Semantic Recognition task for each participant (disregarding learned/not learned status in SL task)
# Step 1: Define valid response types for IT
valid_IT_responses <- tribble(
~IT.Response.Type,
"correct response (M+ target word)",
"familiar foil",
"unrelated foil",
"no meaning",
"NA",
"semantically-related foil"
)
# Step 2: Count IT responses per participant (no SL.Meaning.Condition involved)
it_participant_counts <- mp_slid_data %>%
group_by(Participant.ID, Group, IT.Response.Type) %>%
summarise(n = n(), .groups = "drop")
# Step 3: Get participant × valid IT response type combinations
it_full_grid <- mp_slid_data %>%
distinct(Participant.ID, Group) %>%
crossing(IT.Response.Type = valid_IT_responses$IT.Response.Type) # Using predefined valid response types
# Step 4: Left join actual counts onto the grid to fill missing values
propit_data <- it_full_grid %>%
left_join(it_participant_counts,
by = c("Participant.ID", "Group", "IT.Response.Type")) %>%
mutate(n = replace_na(n, 0)) %>%
group_by(Participant.ID, Group) %>%
mutate(
total_responses = sum(n),
proportion = n / total_responses
) %>%
ungroup()
# Calculate the proportions of responses in the Semantic Recognition task for each participant for only *learned* words
# Step 1: Define valid response types for IT
valid_IT_responses <- tribble(
~IT.Response.Type,
"correct response (M+ target word)",
"familiar foil",
"unrelated foil",
"no meaning",
"NA",
"semantically-related foil"
)
# Step 2: Count IT responses per participant (no SL.Meaning.Condition involved)
learned_it_participant_counts <- learned_slid_data %>%
group_by(Participant.ID, Group, IT.Response.Type) %>%
summarise(n = n(), .groups = "drop")
# Step 3: Get participant × valid IT response type combinations
learned_it_full_grid <- learned_slid_data %>%
distinct(Participant.ID, Group) %>%
crossing(IT.Response.Type = valid_IT_responses$IT.Response.Type) # Using predefined valid response types
# Step 4: Left join actual counts onto the grid to fill missing values
learned_propit_data <- learned_it_full_grid %>%
left_join(learned_it_participant_counts,
by = c("Participant.ID", "Group", "IT.Response.Type")) %>%
mutate(n = replace_na(n, 0)) %>%
group_by(Participant.ID, Group) %>%
mutate(
total_responses = sum(n),
proportion = n / total_responses
) %>%
ungroup()
# Notes:
# SL Meaning Condition = context condition (M- = unsupportive, M+ = supportive); default reference level = unsupportive/M-
# For SL Meaning Condition, M+ = -0.5 (after centering), M- = 0.5 (after centering)
# For Gender, female = -0.5 (after centering), male = 0.5 (after center)Descriptives
# Semantic Recognition task performance (disregarding learned/not learned status in SL task)
mp_slid_data_partsum <- mp_slid_data %>%
group_by(Participant.ID, Group) %>%
summarise(IT_corr_partacc = mean(IT.Correct., na.rm = TRUE), .groups = "drop")
it_summary_df <- mp_slid_data_partsum %>%
group_by(Group) %>%
summarise(
mean_accuracy = mean(IT_corr_partacc),
sd_accuracy = sd(IT_corr_partacc),
n = n(),
.groups = "drop"
) %>%
mutate(
label_y = mean_accuracy + sd_accuracy + 0.02, # Add small buffer
total_n = sum(n)
)
knitr::kable(it_summary_df)| Group | mean_accuracy | sd_accuracy | n | label_y | total_n |
|---|---|---|---|---|---|
| A | 0.3688889 | 0.1593143 | 60 | 0.5482032 | 105 |
| C | 0.3718519 | 0.1636014 | 45 | 0.5554533 | 105 |
# Semantic Recognition task performance for *learned* words
learned_slid_data_partsum <- learned_slid_data %>%
group_by(Participant.ID, Group) %>%
summarise(IT_corr_partacc = mean(IT.Correct., na.rm = TRUE), .groups = "drop")
learned_it_summary_df <- learned_slid_data_partsum %>%
group_by(Group) %>%
summarise(
mean_accuracy = mean(IT_corr_partacc),
sd_accuracy = sd(IT_corr_partacc),
n = n(),
.groups = "drop"
) %>%
mutate(
label_y = mean_accuracy + sd_accuracy + 0.02, # Add small buffer
total_n = sum(n)
)
knitr::kable(learned_it_summary_df)| Group | mean_accuracy | sd_accuracy | n | label_y | total_n |
|---|---|---|---|---|---|
| A | 0.4271696 | 0.1734506 | 60 | 0.6206202 | 105 |
| C | 0.4393798 | 0.1939308 | 45 | 0.6533106 | 105 |
# Proportions of Semantic Recognition task (disregarding learned/not learned status in SL task)
propit_data_partsum <- propit_data %>%
group_by(Participant.ID, Group,IT.Response.Type) %>%
summarise(proportion_partacc = mean(proportion, na.rm = TRUE), .groups = "drop")
propit_summary_df <- propit_data_partsum %>%
group_by(Group, IT.Response.Type) %>%
summarise(
mean_proportion = mean(proportion_partacc),
sd_proportion = sd(proportion_partacc),
n = n(),
.groups = "drop"
) %>%
mutate(
label_y = mean_proportion + sd_proportion + 0.02, # Add small buffer
total_n = sum(n)
)
knitr::kable(propit_summary_df)| Group | IT.Response.Type | mean_proportion | sd_proportion | n | label_y | total_n |
|---|---|---|---|---|---|---|
| A | NA | 0.0000000 | 0.0000000 | 60 | 0.0200000 | 630 |
| A | correct response (M+ target word) | 0.3703175 | 0.1581543 | 60 | 0.5484717 | 630 |
| A | familiar foil | 0.2578571 | 0.1340458 | 60 | 0.4119029 | 630 |
| A | no meaning | 0.2751587 | 0.1715924 | 60 | 0.4667511 | 630 |
| A | semantically-related foil | 0.0633333 | 0.0992818 | 60 | 0.1826151 | 630 |
| A | unrelated foil | 0.0333333 | 0.0527939 | 60 | 0.1061272 | 630 |
| C | NA | 0.0000000 | 0.0000000 | 45 | 0.0200000 | 630 |
| C | correct response (M+ target word) | 0.3718519 | 0.1636014 | 45 | 0.5554533 | 630 |
| C | familiar foil | 0.2118519 | 0.1075803 | 45 | 0.3394321 | 630 |
| C | no meaning | 0.3511111 | 0.1702049 | 45 | 0.5413160 | 630 |
| C | semantically-related foil | 0.0296296 | 0.0523306 | 45 | 0.1019603 | 630 |
| C | unrelated foil | 0.0355556 | 0.0612991 | 45 | 0.1168546 | 630 |
# Proportions of Semantic Recognition task for *learned* words
learned_propit_data_partsum <- learned_propit_data %>%
group_by(Participant.ID, Group,IT.Response.Type) %>%
summarise(proportion_partacc = mean(proportion, na.rm = TRUE), .groups = "drop")
learned_propit_summary_df <- learned_propit_data_partsum %>%
group_by(Group, IT.Response.Type) %>%
summarise(
mean_proportion = mean(proportion_partacc),
sd_proportion = sd(proportion_partacc),
n = n(),
.groups = "drop"
) %>%
mutate(
label_y = mean_proportion + sd_proportion + 0.02, # Add small buffer
total_n = sum(n)
)
knitr::kable(learned_propit_summary_df)| Group | IT.Response.Type | mean_proportion | sd_proportion | n | label_y | total_n |
|---|---|---|---|---|---|---|
| A | NA | 0.0000000 | 0.0000000 | 60 | 0.0200000 | 630 |
| A | correct response (M+ target word) | 0.4287081 | 0.1719068 | 60 | 0.6206149 | 630 |
| A | familiar foil | 0.2320241 | 0.1433781 | 60 | 0.3954022 | 630 |
| A | no meaning | 0.2421080 | 0.1893168 | 60 | 0.4514248 | 630 |
| A | semantically-related foil | 0.0608621 | 0.1009820 | 60 | 0.1818442 | 630 |
| A | unrelated foil | 0.0362977 | 0.0628175 | 60 | 0.1191152 | 630 |
| C | NA | 0.0000000 | 0.0000000 | 45 | 0.0200000 | 630 |
| C | correct response (M+ target word) | 0.4393798 | 0.1939308 | 45 | 0.6533106 | 630 |
| C | familiar foil | 0.1887736 | 0.1238211 | 45 | 0.3325947 | 630 |
| C | no meaning | 0.3287325 | 0.1888939 | 45 | 0.5376264 | 630 |
| C | semantically-related foil | 0.0262722 | 0.0600091 | 45 | 0.1062813 | 630 |
| C | unrelated foil | 0.0168419 | 0.0515258 | 45 | 0.0883677 | 630 |
write.csv(mp_slid_data$IT.Correct., "check_sd.csv", row.names = FALSE)Visualizations
# Semantic Recognition task performance (disregarding learned/not learned status in SL task)
mp_slid_data_partsum %>%
#mutate(SL.Meaning.Condition = factor(SL.Meaning.Condition, levels = c("M+","M-")) # Set the desired order of variables
#)%>%
ggplot(aes(x = Group, y = IT_corr_partacc, fill = Group, color = Group)) +
stat_summary(fun = mean, position = position_dodge(width = .9),
geom="bar", colour = "black") +
geom_hline(yintercept = .5, color = "grey", linetype = "dashed") +
#facet_grid(~Group) +
stat_summary(fun.data = mean_sdl, fun.args = list(mult = 1),
geom = "errorbar",
position = position_dodge(width = 0.9),
size = 1, width = 0.25, color = "black") + # Standard deviation bars
geom_text(
data = it_summary_df,
aes(
x = Group,
y = label_y,
label = sprintf("%.2f", mean_accuracy),
group = Group
),
position = position_dodge(width = 0.9),
size = 4,
color = "black") + # Add formatted text on top of bars
# scale_fill_brewer(palette = "Dark2") +
#langcog::scale_fill_solarized()+
labs(y = "Mean Accuracy of Target Meaning/Word Identification",
x = "Group")+
#fill = "Meaning Condition") +
theme(axis.text.x = element_text(hjust = 0.5, angle = 0, size = 12, color = "black"), # Customize x-axis text
axis.text.y = element_text(size = 12, color = "black"),
axis.title.x = element_text(size = 14, color = "black", face = "bold", margin = margin(t = 10)), # x-axis title
axis.title.y = element_text(size = 14, color = "black", face = "bold", margin = margin(r = 10)), # y-axis title
strip.text = element_text(size = 12, color = "black", face = "bold"), # Facet labels
legend.key.size = unit(.5, 'cm'),
legend.text = element_text(size = 10),
legend.title = element_text(size = 11),
legend.margin = margin(t = 0, unit = 'cm'),
plot.title = element_text(size = 12)) +
guides(fill = FALSE, color = FALSE) + # Remove both fill and color legends
scale_y_continuous(breaks = seq(0, 1, .25))
# Semantic Recognition task performance for *learned* words
learned_slid_data_partsum %>%
#mutate(SL.Meaning.Condition = factor(SL.Meaning.Condition, levels = c("M+","M-")) # Set the desired order of variables
#)%>%
ggplot(aes(x = Group, y = IT_corr_partacc, fill = Group, color = Group)) +
stat_summary(fun = mean, position = position_dodge(width = .9),
geom="bar", colour = "black") +
geom_hline(yintercept = .5, color = "grey", linetype = "dashed") +
#facet_grid(~Group) +
stat_summary(fun.data = mean_sdl, fun.args = list(mult = 1),
geom = "errorbar",
position = position_dodge(width = 0.9),
size = 1, width = 0.25, color = "black") + # Standard deviation bars
geom_text(
data = learned_it_summary_df,
aes(
x = Group,
y = label_y,
label = sprintf("%.2f", mean_accuracy),
group = Group
),
position = position_dodge(width = 0.9),
size = 4,
color = "black") + # Add formatted text on top of bars
# scale_fill_brewer(palette = "Dark2") +
#langcog::scale_fill_solarized()+
labs(y = "Mean Accuracy of Target Meaning/Word Identification",
x = "Group")+
#fill = "Meaning Condition") +
theme(axis.text.x = element_text(hjust = 0.5, angle = 0, size = 12, color = "black"), # Customize x-axis text
axis.text.y = element_text(size = 12, color = "black"),
axis.title.x = element_text(size = 14, color = "black", face = "bold", margin = margin(t = 10)), # x-axis title
axis.title.y = element_text(size = 14, color = "black", face = "bold", margin = margin(r = 10)), # y-axis title
strip.text = element_text(size = 12, color = "black", face = "bold"), # Facet labels
legend.key.size = unit(.5, 'cm'),
legend.text = element_text(size = 10),
legend.title = element_text(size = 11),
legend.margin = margin(t = 0, unit = 'cm'),
plot.title = element_text(size = 12)) +
guides(fill = FALSE, color = FALSE) + # Remove both fill and color legends
scale_y_continuous(breaks = seq(0, 1, .25))
# Proportions of Semantic Recognition task (disregarding learned/not learned status in SL task)
propit_data_partsum %>%
mutate(IT.Response.Type = factor(IT.Response.Type, levels = c("correct response (M+ target word)","semantically-related foil", "familiar foil", "unrelated foil", "no meaning")) # Set the desired order of variables
)%>%
filter(IT.Response.Type != "NA") %>% # Removes both NA and character "NA"
droplevels() %>%
ggplot(aes(x = IT.Response.Type, y = proportion_partacc, fill = Group)) +
stat_summary(fun = mean, position = position_dodge(width = .9),
geom="bar", colour = "black") +
geom_hline(yintercept = .5, color = "grey", linetype = "dashed") +
facet_grid(~Group) +
stat_summary(fun.data = mean_sdl, fun.args = list(mult = 1),
geom = "errorbar",
position = position_dodge(width = 0.9),
size = 1, width = 0.25, color = "black") + # Standard deviation bars
geom_text(
data = propit_summary_df %>%
filter(IT.Response.Type != "NA") %>%
droplevels(),
aes(
x = IT.Response.Type,
y = label_y,
label = sprintf("%.2f", mean_proportion),
group = Group
),
position = position_dodge(width = 0.9),
size = 4,
color = "black") + # Add formatted text on top of bars
labs(#title = "Proportions of Semantic Recognition Task Responses",
x = "Response Type",
y = "Mean Proportion of Responses",
fill = "Group") +
scale_x_discrete(labels = c(
"correct response (M+ target word)" = "correct response\n(M+ target word)",
"semantically-related foil" = "semantically-related foil",
"familiar foil" = "familiar foil",
"unrelated foil" = "unrelated foil",
"no meaning" = "no meaning"
))+
theme(axis.text.x = element_text(hjust = 1, angle = 45, size = 12, color = "black"), # Customize x-axis text
axis.text.y = element_text(size = 12, color = "black"),
axis.title.x = element_text(size = 14, color = "black", face = "bold", margin = margin(t = 10)), # x-axis title
axis.title.y = element_text(size = 14, color = "black", face = "bold", margin = margin(r = 10)), # y-axis title
strip.text = element_text(size = 12, color = "black", face = "bold"), # Facet labels
legend.key.size = unit(.5, 'cm'),
legend.text = element_text(size = 11),
legend.title = element_text(size = 12, face = "bold"),
legend.margin = margin(t = 0, unit = 'cm'),
plot.title = element_text(size = 12),
plot.margin = margin(t = 10, r = 20, b = 20, l = 10, unit = "pt"))+
#guides(fill = FALSE, color = FALSE) + # Remove both fill and color legends
scale_y_continuous(breaks = seq(0, 1, .25))
# Proportions of Semantic Recognition task for *learned* words
learned_propit_data_partsum %>%
mutate(IT.Response.Type = factor(IT.Response.Type, levels = c("correct response (M+ target word)","semantically-related foil", "familiar foil", "unrelated foil", "no meaning")) # Set the desired order of variables
)%>%
filter(IT.Response.Type != "NA") %>% # Removes both NA and character "NA"
droplevels() %>%
ggplot(aes(x = IT.Response.Type, y = proportion_partacc, fill = Group)) +
stat_summary(fun = mean, position = position_dodge(width = .9),
geom="bar", colour = "black") +
geom_hline(yintercept = .5, color = "grey", linetype = "dashed") +
facet_grid(~Group) +
stat_summary(fun.data = mean_sdl, fun.args = list(mult = 1),
geom = "errorbar",
position = position_dodge(width = 0.9),
size = 1, width = 0.25, color = "black") + # Standard deviation bars
geom_text(
data = learned_propit_summary_df %>%
filter(IT.Response.Type != "NA") %>%
droplevels(),
aes(
x = IT.Response.Type,
y = label_y,
label = sprintf("%.2f", mean_proportion),
group = Group
),
position = position_dodge(width = 0.9),
size = 4,
color = "black") + # Add formatted text on top of bars
labs(#title = "Proportions of Semantic Recognition Task Responses",
x = "Response Type",
y = "Mean Proportion of Responses",
fill = "Group") +
scale_x_discrete(labels = c(
"correct response (M+ target word)" = "correct response\n(M+ target word)",
"semantically-related foil" = "semantically-related foil",
"familiar foil" = "familiar foil",
"unrelated foil" = "unrelated foil",
"no meaning" = "no meaning"
))+
theme(axis.text.x = element_text(hjust = 1, angle = 45, size = 12, color = "black"), # Customize x-axis text
axis.text.y = element_text(size = 12, color = "black"),
axis.title.x = element_text(size = 14, color = "black", face = "bold", margin = margin(t = 10)), # x-axis title
axis.title.y = element_text(size = 14, color = "black", face = "bold", margin = margin(r = 10)), # y-axis title
strip.text = element_text(size = 12, color = "black", face = "bold"), # Facet labels
legend.key.size = unit(.5, 'cm'),
legend.text = element_text(size = 11),
legend.title = element_text(size = 12, face = "bold"),
legend.margin = margin(t = 0, unit = 'cm'),
plot.title = element_text(size = 12),
plot.margin = margin(t = 10, r = 20, b = 20, l = 10, unit = "pt"))+
#guides(fill = FALSE, color = FALSE) + # Remove both fill and color legends
scale_y_continuous(breaks = seq(0, 1, .25))
# Proportions of Semantic Recognition task for child data only (disregarding learned/not learned status in SL task)
kidpropit_data_plot <- propit_data_partsum %>%
mutate(IT.Response.Type = factor(IT.Response.Type, levels = c("correct response (M+ target word)","semantically-related foil", "familiar foil", "unrelated foil", "no meaning")) # Set the desired order of variables
)%>%
filter(IT.Response.Type != "NA" & Group == "C") %>% # Removes both NA and character "NA"
droplevels() %>%
ggplot(aes(x = IT.Response.Type, y = proportion_partacc, fill = Group)) +
stat_summary(fun = mean, position = position_dodge(width = .9),
geom="bar", colour = "black", fill = "#619CFF") +
geom_hline(yintercept = .5, color = "grey", linetype = "dashed") +
facet_grid(~Group) +
stat_summary(fun.data = mean_sdl, fun.args = list(mult = 1),
geom = "errorbar",
position = position_dodge(width = 0.9),
size = 1, width = 0.25, color = "black") + # Standard deviation bars
geom_text(
data = propit_summary_df %>%
filter(IT.Response.Type != "NA" & Group == "C") %>%
droplevels(),
aes(
x = IT.Response.Type,
y = label_y,
label = sprintf("%.2f", mean_proportion),
group = Group
),
position = position_dodge(width = 0.9),
size = 4,
color = "black") + # Add formatted text on top of bars
labs(#title = "Proportions of Semantic Recognition Task Responses",
x = "Semantic Recognition Response Type",
y = "Mean Proportion of Responses",
fill = "Group") +
scale_x_discrete(labels = c(
"correct response (M+ target word)" = "correct response\n(target word)",
"semantically-related foil" = "semantically-related\nfoil",
"familiar foil" = "familiar foil",
"unrelated foil" = "unrelated foil",
"no meaning" = "red X\n(no meaning)"
))+
theme(axis.text.x = element_text(hjust = .5, angle = 0, size = 12, color = "black"), # Customize x-axis text
axis.text.y = element_text(size = 12, color = "black"),
axis.title.x = element_text(size = 14, color = "black", face = "bold", margin = margin(t = 10)), # x-axis title
axis.title.y = element_text(size = 14, color = "black", face = "bold", margin = margin(r = 10)), # y-axis title
strip.text = element_text(size = 12, color = "black", face = "bold"), # Facet labels
legend.key.size = unit(.5, 'cm'),
legend.text = element_text(size = 11),
legend.title = element_text(size = 12, face = "bold"),
legend.margin = margin(t = 0, unit = 'cm'),
legend.position = "none",
plot.title = element_text(size = 12),
plot.margin = margin(t = 10, r = 20, b = 20, l = 10, unit = "pt"))+
#guides(fill = FALSE, color = FALSE) + # Remove both fill and color legends
scale_y_continuous(breaks = seq(0, 1, .25))
ggplot2::ggsave(filename = "ID Task Responses - Kids.jpeg",
plot = kidpropit_data_plot,
device = "jpeg",
dpi = 300,
width = 12,
height = 8,
units = "in",
limitsize = FALSE)
# Proportions of Semantic Recognition task for child data only for *learned* words
learned_kidpropit_data_plot <- learned_propit_data_partsum %>%
mutate(IT.Response.Type = factor(IT.Response.Type, levels = c("correct response (M+ target word)","semantically-related foil", "familiar foil", "unrelated foil", "no meaning")) # Set the desired order of variables
)%>%
filter(IT.Response.Type != "NA" & Group == "C") %>% # Removes both NA and character "NA"
droplevels() %>%
ggplot(aes(x = IT.Response.Type, y = proportion_partacc, fill = Group)) +
stat_summary(fun = mean, position = position_dodge(width = .9),
geom="bar", colour = "black", fill = "#619CFF") +
geom_hline(yintercept = .5, color = "grey", linetype = "dashed") +
facet_grid(~Group) +
stat_summary(fun.data = mean_sdl, fun.args = list(mult = 1),
geom = "errorbar",
position = position_dodge(width = 0.9),
size = 1, width = 0.25, color = "black") + # Standard deviation bars
geom_text(
data = learned_propit_summary_df %>%
filter(IT.Response.Type != "NA" & Group == "C") %>%
droplevels(),
aes(
x = IT.Response.Type,
y = label_y,
label = sprintf("%.2f", mean_proportion),
group = Group
),
position = position_dodge(width = 0.9),
size = 4,
color = "black") + # Add formatted text on top of bars
labs(#title = "Proportions of Semantic Recognition Task Responses",
x = "Semantic Recognition Response Type",
y = "Mean Proportion of Responses",
fill = "Group") +
scale_x_discrete(labels = c(
"correct response (M+ target word)" = "correct response\n(target word)",
"semantically-related foil" = "semantically related\nfoil",
"familiar foil" = "familiar foil",
"unrelated foil" = "unrelated foil",
"no meaning" = "red X\n(no meaning)"
))+
theme(axis.text.x = element_text(hjust = .5, angle = 0, size = 12, color = "black"), # Customize x-axis text
axis.text.y = element_text(size = 12, color = "black"),
axis.title.x = element_text(size = 14, color = "black", face = "bold", margin = margin(t = 10)), # x-axis title
axis.title.y = element_text(size = 14, color = "black", face = "bold", margin = margin(r = 10)), # y-axis title
strip.text = element_text(size = 12, color = "black", face = "bold"), # Facet labels
legend.key.size = unit(.5, 'cm'),
legend.text = element_text(size = 11),
legend.title = element_text(size = 12, face = "bold"),
legend.margin = margin(t = 0, unit = 'cm'),
legend.position = "none",
plot.title = element_text(size = 12),
plot.margin = margin(t = 10, r = 20, b = 20, l = 10, unit = "pt"))+
#guides(fill = FALSE, color = FALSE) + # Remove both fill and color legends
scale_y_continuous(breaks = seq(0, 1, .25))
ggplot2::ggsave(filename = "ID Task Responses - Kids.jpeg",
plot = learned_kidpropit_data_plot,
device = "jpeg",
dpi = 300,
width = 12,
height = 8,
units = "in",
limitsize = FALSE)
# Proportions of Semantic Recognition task for adult data only (disregarding learned/not learned status in SL task)
adultpropit_data_plot <- propit_data_partsum %>%
mutate(IT.Response.Type = factor(IT.Response.Type, levels = c("correct response (M+ target word)","semantically-related foil", "familiar foil", "unrelated foil", "no meaning")) # Set the desired order of variables
)%>%
filter(IT.Response.Type != "NA" & Group == "A") %>% # Removes both NA and character "NA"
droplevels() %>%
ggplot(aes(x = IT.Response.Type, y = proportion_partacc, fill = Group)) +
stat_summary(fun = mean, position = position_dodge(width = .9),
geom="bar", colour = "black") +
geom_hline(yintercept = .5, color = "grey", linetype = "dashed") +
facet_grid(~Group) +
stat_summary(fun.data = mean_sdl, fun.args = list(mult = 1),
geom = "errorbar",
position = position_dodge(width = 0.9),
size = 1, width = 0.25, color = "black") + # Standard deviation bars
geom_text(
data = propit_summary_df %>%
filter(IT.Response.Type != "NA" & Group == "A") %>%
droplevels(),
aes(
x = IT.Response.Type,
y = label_y,
label = sprintf("%.2f", mean_proportion),
group = Group
),
position = position_dodge(width = 0.9),
size = 4,
color = "black") + # Add formatted text on top of bars
labs(#title = "Proportions of Semantic Recognition Task Responses",
x = "Semantic Recognition Response Type",
y = "Mean Proportion of Responses",
fill = "Group") +
scale_x_discrete(labels = c(
"correct response (M+ target word)" = "correct response\n(target word)",
"semantically-related foil" = "semantically-related\nfoil",
"familiar foil" = "familiar foil",
"unrelated foil" = "unrelated foil",
"no meaning" = "red X\n(no meaning)"
))+
theme(axis.text.x = element_text(hjust = .5, angle = 0, size = 12, color = "black"), # Customize x-axis text
axis.text.y = element_text(size = 12, color = "black"),
axis.title.x = element_text(size = 14, color = "black", face = "bold", margin = margin(t = 10)), # x-axis title
axis.title.y = element_text(size = 14, color = "black", face = "bold", margin = margin(r = 10)), # y-axis title
strip.text = element_text(size = 12, color = "black", face = "bold"), # Facet labels
legend.key.size = unit(.5, 'cm'),
legend.text = element_text(size = 11),
legend.title = element_text(size = 12, face = "bold"),
legend.margin = margin(t = 0, unit = 'cm'),
legend.position = "none",
plot.title = element_text(size = 12),
plot.margin = margin(t = 10, r = 20, b = 20, l = 10, unit = "pt"))+
#guides(fill = FALSE, color = FALSE) + # Remove both fill and color legends
scale_y_continuous(breaks = seq(0, 1, .25))
ggplot2::ggsave(filename = "ID Task Responses - Adults.jpeg",
plot = adultpropit_data_plot,
device = "jpeg",
dpi = 300,
width = 12,
height = 8,
units = "in",
limitsize = FALSE)
# Proportions of Semantic Recognition task for adult data only for *learned* words
learned_adultpropit_data_plot <- learned_propit_data_partsum %>%
mutate(IT.Response.Type = factor(IT.Response.Type, levels = c("correct response (M+ target word)","semantically-related foil", "familiar foil", "unrelated foil", "no meaning")) # Set the desired order of variables
)%>%
filter(IT.Response.Type != "NA" & Group == "A") %>% # Removes both NA and character "NA"
droplevels() %>%
ggplot(aes(x = IT.Response.Type, y = proportion_partacc, fill = Group)) +
stat_summary(fun = mean, position = position_dodge(width = .9),
geom="bar", colour = "black") +
geom_hline(yintercept = .5, color = "grey", linetype = "dashed") +
facet_grid(~Group) +
stat_summary(fun.data = mean_sdl, fun.args = list(mult = 1),
geom = "errorbar",
position = position_dodge(width = 0.9),
size = 1, width = 0.25, color = "black") + # Standard deviation bars
geom_text(
data = learned_propit_summary_df %>%
filter(IT.Response.Type != "NA" & Group == "A") %>%
droplevels(),
aes(
x = IT.Response.Type,
y = label_y,
label = sprintf("%.2f", mean_proportion),
group = Group
),
position = position_dodge(width = 0.9),
size = 4,
color = "black") + # Add formatted text on top of bars
labs(#title = "Proportions of Semantic Recognition Task Responses",
x = "Semantic Recognition Response Type",
y = "Mean Proportion of Responses",
fill = "Group") +
scale_x_discrete(labels = c(
"correct response (M+ target word)" = "correct response\n(target word)",
"semantically-related foil" = "semantically related\nfoil",
"familiar foil" = "familiar foil",
"unrelated foil" = "unrelated foil",
"no meaning" = "red X\n(no meaning)"
))+
theme(axis.text.x = element_text(hjust = .5, angle = 0, size = 12, color = "black"), # Customize x-axis text
axis.text.y = element_text(size = 12, color = "black"),
axis.title.x = element_text(size = 14, color = "black", face = "bold", margin = margin(t = 10)), # x-axis title
axis.title.y = element_text(size = 14, color = "black", face = "bold", margin = margin(r = 10)), # y-axis title
strip.text = element_text(size = 12, color = "black", face = "bold"), # Facet labels
legend.key.size = unit(.5, 'cm'),
legend.text = element_text(size = 11),
legend.title = element_text(size = 12, face = "bold"),
legend.margin = margin(t = 0, unit = 'cm'),
legend.position = "none",
plot.title = element_text(size = 12),
plot.margin = margin(t = 10, r = 20, b = 20, l = 10, unit = "pt"))+
#guides(fill = FALSE, color = FALSE) + # Remove both fill and color legends
scale_y_continuous(breaks = seq(0, 1, .25))
ggplot2::ggsave(filename = "ID Task Responses - Adults.jpeg",
plot = learned_adultpropit_data_plot,
device = "jpeg",
dpi = 300,
width = 12,
height = 8,
units = "in",
limitsize = FALSE)Analyses
RQ1: When an individual correctly identifies a new word’s meaning in the Semantic Learning task, do they correctly identify the word’s meaning in the Semantic Recognition task?
Children
Results:
- intercept not significant, so child participants who failed to
identify the meaning of the word in the Semantic Learning task were not
significantly more likely to correctly or incorrectly identify the image
in the Semantic Recognition task
- 2.91 times more likely to correctly identify the image in the Semantic
Recognition task if they had correctly identified the word’s meaning in
the Semantic Learning task, compared to if they didn’t
Child participants who formed a correct semantic representation during learning were significantly more likely to get the Semantic Recognition trial correct—almost 3 times higher odds, even after accounting for age, gender, and subject/item variability.
# # maximal model
# rq1.c.m1 <- glmer(IT.Correct. ~ SL.Correct.Enough.for.IT. + age.c + gender.c +(1 + SL.Correct.Enough.for.IT. |Participant.ID) +(1 + SL.Correct.Enough.for.IT. |SL.Correct.Response)+(1+ SL.Correct.Enough.for.IT.|SL.Trial.Order.by.Task)+(1+ SL.Correct.Enough.for.IT.|IT.Trial.Order.by.Task), data=subset(mp_slid_data,Group == "C"), family = 'binomial', control=glmerControl(optimizer="bobyqa", optCtrl=list(maxfun=2e4))) # isSingular warning, so checking importance of each component to decide how to simplify model
#
# summary(rePCA(rq1.c.m1)) # PC 2 for Semantic Learning task trial order (SL.Trial.Order.by.Task), Semantic Recognition task trial order (IT.Trial.Order.by.Task), and participant account for negligible proportions of variance
#
# # model with above random slopes removed
rq1.c.m2 <- glmer(IT.Correct. ~ SL.Correct.Enough.for.IT. + age.c + gender.c +(1|Participant.ID)+ (1 + SL.Correct.Enough.for.IT. |SL.Correct.Response)+(1|SL.Trial.Order.by.Task)+(1|IT.Trial.Order.by.Task), data=subset(mp_slid_data,Group == "C"), family = 'binomial', control=glmerControl(optimizer="bobyqa", optCtrl=list(maxfun=2e4))) # converges!
# final model results
tab_model(rq1.c.m2, show.se = TRUE,show.ci = FALSE,show.stat = TRUE, show.df = TRUE)| IT.Correct. | |||||
|---|---|---|---|---|---|
| Predictors | Odds Ratios | std. Error | Statistic | p | df |
| (Intercept) | 0.23 | 0.27 | -1.27 | 0.205 | Inf |
| SL Correct Enough for IT | 2.91 | 0.80 | 3.89 | <0.001 | Inf |
| age c | 0.98 | 0.13 | -0.13 | 0.900 | Inf |
| gender c | 1.17 | 0.26 | 0.71 | 0.478 | Inf |
| Random Effects | |||||
| σ2 | 3.29 | ||||
| τ00 Participant.ID | 0.17 | ||||
| τ00 IT.Trial.Order.by.Task | 0.03 | ||||
| τ00 SL.Trial.Order.by.Task | 0.08 | ||||
| τ00 SL.Correct.Response | 0.12 | ||||
| τ11 SL.Correct.Response.SL.Correct.Enough.for.IT. | 0.42 | ||||
| ρ01 SL.Correct.Response | -0.68 | ||||
| ICC | 0.13 | ||||
| N Participant.ID | 45 | ||||
| N SL.Correct.Response | 15 | ||||
| N SL.Trial.Order.by.Task | 30 | ||||
| N IT.Trial.Order.by.Task | 31 | ||||
| Observations | 675 | ||||
| Marginal R2 / Conditional R2 | 0.064 / 0.182 | ||||
summary(rq1.c.m2)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: IT.Correct. ~ SL.Correct.Enough.for.IT. + age.c + gender.c +
## (1 | Participant.ID) + (1 + SL.Correct.Enough.for.IT. | SL.Correct.Response) +
## (1 | SL.Trial.Order.by.Task) + (1 | IT.Trial.Order.by.Task)
## Data: subset(mp_slid_data, Group == "C")
## Control: glmerControl(optimizer = "bobyqa", optCtrl = list(maxfun = 20000))
##
## AIC BIC logLik deviance df.resid
## 863.5 908.6 -421.7 843.5 665
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.4958 -0.7278 -0.5010 0.9408 2.2953
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## Participant.ID (Intercept) 0.17282 0.4157
## IT.Trial.Order.by.Task (Intercept) 0.02539 0.1594
## SL.Trial.Order.by.Task (Intercept) 0.07857 0.2803
## SL.Correct.Response (Intercept) 0.11698 0.3420
## SL.Correct.Enough.for.IT. 0.42399 0.6511 -0.68
## Number of obs: 675, groups:
## Participant.ID, 45; IT.Trial.Order.by.Task, 31; SL.Trial.Order.by.Task, 30; SL.Correct.Response, 15
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.45000 1.14441 -1.267 0.205
## SL.Correct.Enough.for.IT. 1.06959 0.27469 3.894 9.87e-05 ***
## age.c -0.01725 0.13686 -0.126 0.900
## gender.c 0.15553 0.21934 0.709 0.478
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) SL.C.E age.c
## SL.C.E..IT. -0.182
## age.c 0.981 -0.043
## gender.c 0.154 0.084 0.159Adults
Results:
- intercept is significant, so adult participants who failed to identify
the meaning of the word in the Semantic Learning task were significantly
more likely to not correctly identify the image in the Semantic
Recognition task (i.e., pick a foil/say no meaning)
- 4.8 times more likely to correctly identify the image in the Semantic
Recognition task if they had correctly identified the word’s meaning in
the Semantic Learning task, compared to if they didn’t
Adult participants who formed a correct semantic representation during learning were significantly more likely to get the Semantic Recognition trial correct—about 4.8 times higher odds, even after accounting for age, gender, and subject/item variability.
Comparing patterns of results across the two participant groups, while Semantic Learning task performance still supports Semantic Recognition task performance in children, it has a weaker effect in children than in adults.
# # maximal model
#rq1.a.m1 <- glmer(IT.Correct. ~ SL.Correct.Enough.for.IT. + age.c + gender.c + (1 + SL.Correct.Enough.for.IT. |Participant.ID)+(1 + SL.Correct.Enough.for.IT. |SL.Correct.Response)+(1+ SL.Correct.Enough.for.IT.|SL.Trial.Order.by.Task)+(1+ SL.Correct.Enough.for.IT.|IT.Trial.Order.by.Task), data=subset(mp_slid_data,Group == "A"), family = 'binomial', control=glmerControl(optimizer="bobyqa", optCtrl=list(maxfun=2e4))) # isSingular warning, so checking importance of each component to decide how to simplify model
#
# summary(rePCA(rq1.a.m1)) # all PC 2s for Semantic Learning task accuracy (SL.Correct.Enough.for.IT.) account for negligible proportions of variance
#
# # model with all random slopes removed
# rq1.a.m2 <- glmer(IT.Correct. ~ SL.Correct.Enough.for.IT. + age.c + gender.c + (1 |Participant.ID)+(1 |SL.Correct.Response)+(1|SL.Trial.Order.by.Task)+(1|IT.Trial.Order.by.Task), data=subset(mp_slid_data,Group == "A"), family = 'binomial', control=glmerControl(optimizer="bobyqa", optCtrl=list(maxfun=2e4))) # isSingular warning, so checking importance of each component to decide how to further simplify model
#
# summary(rePCA(rq1.a.m2)) # random intercept for Semantic Learning task order (SL.Trial.Order.by.Task) and Semantic Recognition task trial order (IT.Trial.Order.by.Task) accounts for a negligible proportion of variance
#
# # model with random intercept for Semantic Learning task order (SL.Trial.Order.by.Task) and Semantic Recognition task trial order (IT.Trial.Order.by.Task) removed
rq1.a.m3 <- glmer(IT.Correct. ~ SL.Correct.Enough.for.IT. + age.c + gender.c + (1 |Participant.ID)+(1 |SL.Correct.Response), data=subset(mp_slid_data,Group == "A"), family = 'binomial', control=glmerControl(optimizer="bobyqa", optCtrl=list(maxfun=2e4))) # converges!
# final model results
tab_model(rq1.a.m3, show.se = TRUE,show.ci = FALSE,show.stat = TRUE, show.df = TRUE)| IT.Correct. | |||||
|---|---|---|---|---|---|
| Predictors | Odds Ratios | std. Error | Statistic | p | df |
| (Intercept) | 0.15 | 0.06 | -4.64 | <0.001 | Inf |
| SL Correct Enough for IT | 4.80 | 1.08 | 6.96 | <0.001 | Inf |
| age c | 1.00 | 0.05 | 0.05 | 0.957 | Inf |
| gender c | 1.24 | 0.22 | 1.20 | 0.231 | Inf |
| Random Effects | |||||
| σ2 | 3.29 | ||||
| τ00 Participant.ID | 0.14 | ||||
| τ00 SL.Correct.Response | 0.14 | ||||
| ICC | 0.08 | ||||
| N Participant.ID | 60 | ||||
| N SL.Correct.Response | 15 | ||||
| Observations | 900 | ||||
| Marginal R2 / Conditional R2 | 0.106 / 0.176 | ||||
summary(rq1.a.m3)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: IT.Correct. ~ SL.Correct.Enough.for.IT. + age.c + gender.c +
## (1 | Participant.ID) + (1 | SL.Correct.Response)
## Data: subset(mp_slid_data, Group == "A")
## Control: glmerControl(optimizer = "bobyqa", optCtrl = list(maxfun = 20000))
##
## AIC BIC logLik deviance df.resid
## 1117.8 1146.6 -552.9 1105.8 894
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.4151 -0.7930 -0.4219 0.9933 3.6762
##
## Random effects:
## Groups Name Variance Std.Dev.
## Participant.ID (Intercept) 0.1394 0.3734
## SL.Correct.Response (Intercept) 0.1396 0.3737
## Number of obs: 900, groups: Participant.ID, 60; SL.Correct.Response, 15
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.866224 0.401821 -4.644 3.41e-06 ***
## SL.Correct.Enough.for.IT. 1.568281 0.225440 6.957 3.49e-12 ***
## age.c 0.002893 0.053101 0.054 0.957
## gender.c 0.214073 0.178667 1.198 0.231
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) SL.C.E age.c
## SL.C.E..IT. -0.501
## age.c -0.812 0.021
## gender.c -0.098 0.102 0.089