It seems pretty unlikely given the number of experimental participants in the pilot that we’ll be able to look at anything statistically: we collected data from a total of 14 participants from 3 conditions. The 3 conditions we looked at (from the larger set) were Condition 1 (Sound Symbolic + Congruent Category Labels: 6 participants), Condition 4 (Sound Symbolic + Congruent Individual Labels: 3 participants), and Condition 10 (No Label: 5 participants).

So lets aggregate this data and take a look at it

#First we need to add Block back into the data, which necessitates splitting into Training and Testing

KTest <- subset(KPilot, Trial.Type == "Testing")
KTrain <- subset(KPilot, Trial.Type == "Training")

KTrain$Block <- ceiling((KTrain$TrialNum/16))   #Puts block numbers in
KTest$Block <- ceiling((KTest$TrialNum/24)) 

KPilot <- rbind(KTrain, KTest)

#Now we should be able to aggregate the data the way we want

KPilotAgg <- aggregate(RespCorr ~ Condition + Trial.Type + Block + ID, data= KPilot, mean)
KPilotRTAgg <- aggregate(RT ~ Condition + Trial.Type + Block + ID, data= KPilot, mean)

KPilotAgg$RT <- KPilotRTAgg$RT
KPilotAgg$Block <- as.factor(KPilotAgg$Block)
KPilotAgg$Condition <- as.factor(KPilotAgg$Condition)

#Rearrange and Labels Factors

KPilotAgg$Trial.Type <- factor(KPilotAgg$Trial.Type, 
                              levels = c("Training", "Testing"))

KPilotAgg$Condition <- factor(KPilotAgg$Condition, 
                              levels = c(1,4,10), 
                              labels = c("Sound Symbolic - Category", "Sound Symbolic - Individual", "No Label"))

#Plot that shit

ggplot(data = KPilotAgg, aes(x=Block, y= RespCorr)) +
  geom_jitter(aes(y=RespCorr, color = Condition ), width = 0.2, height = 0) +
  stat_summary(fun.y = mean, geom= "smooth", aes(colour = Condition, group = Condition)) +
  labs(x="Block", y="Proportion Correct") +
  ggtitle("Stekic et al. Pilot Data") +
  theme_alan() +
  scale_y_continuous(limits = c(0.25,1), breaks = c(0.25,0.5,0.75,1), expand = c(0, 0)) +
  facet_grid(~ Trial.Type, scales = "free")

We can see here that the pilot data that we have really isn’t great - there is a lot of individual variation in performance, which is why I plotted all the individual data - and that leads to it being very difficult to draw any conclusions about what to expect from the actual data

What about for RTs instead?

ggplot(data = KPilotAgg, aes(x=Block, y= RT)) +
  geom_jitter(aes(y=RT, color = Condition ), width = 0.2, height = 0) +
  stat_summary(fun.y = mean, geom= "smooth", aes(colour = Condition, group = Condition)) +
  labs(x="Block", y="Proportion Correct") +
  ggtitle("Stekic et al. Pilot Data") +
  theme_alan() +
 # scale_y_continuous(limits = c(0.25,1), breaks = c(0.25,0.5,0.75,1), expand = c(0, 0)) +
  facet_grid(~ Trial.Type, scales = "free")

The RT data has a bit more consistency going on but still isn’t perfect, and again the individual and even by-block variation is high- e.g. we have one participant who for some reason took over 8 seconds per trial on average in Block 7 of Training - some of this would be cleaned up by typical procedures for throwing out long trials (they likely were distracted by something else), but this isn’t perfect by any means.