SingleWord CFS
Hugh Rabagliati
29 August 2016
Background
This is an attempt to conduct an English-language replication of Yang and Yeh (2011, Consc. & Cogn). That paper showed that neutral words break suppression more quickly than negative words.
In our replication, participants saw 300 words, in a 2*2 design that crossed Emotional Content (Negative vs Neutral words) and Length (Short vs Long words). Emotional ratings were taken from the Florida Affective Norms for English Words.
We have a smaller number of subjects than our other experiments (28 total, 26 after exclusions), because participants saw many more critical trials than most comparable experiments.
sense.pop <- read_data("./data/")
sense.pop$length <- "Short"
sense.pop[sense.pop$info %in% c("long negative","long neutral"),]$length <- "Long"
sense.pop <- subset(sense.pop, rt != "None")
sense.pop$rt <- as.numeric(sense.pop$rt)
# Remove outlier subjects by Accuracy and RT (mean acc must be > 0.9, mean rt must be < 3sd above group mean)
Acc <- summaryBy(match. + rt ~ SubjNo, data = subset(sense.pop), keep.names = T)
sense.pop <- subset(sense.pop, SubjNo %in% Acc[Acc$match. > 0.9,]$SubjNo)
sense.pop <- subset(sense.pop, SubjNo %in% Acc[Acc$rt < (mean(Acc$rt) + (3*sd(Acc$rt))),]$SubjNo)
# Remove incorrect trials
sense.pop <- subset(sense.pop, match. == 1)
# Remove RTs < 200ms
sense.pop <- subset(sense.pop, rt > 0.2)
sense.pop$rt.log <- log(sense.pop$rt)Plot RT distribution
As seen below, there is considerable skew in the raw RT data (upper plot), which somewhat remains after log transforming the RTs (lower plot). 
Plot Summaries
These bar plots suggest that the effect of Length is considerably greater than the effect of Emotional Content. 
Analyze data
Statistical analyses show a robust effect of Length, and no effect of Emotional Content. This holds for both the raw data (top analysis) and the log transformed data (bottom analysis).
## Linear mixed model fit by REML ['lmerMod']
## Formula: rt ~ prime_semantics * length + (1 + prime_semantics * length |
## SubjNo)
## Data: sense.pop
##
## REML criterion at convergence: 19040.6
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.5022 -0.3872 -0.1683 0.1196 7.9498
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## SubjNo (Intercept) 0.301967 0.5495
## prime_semanticsneutral 0.008318 0.0912 0.59
## lengthShort 0.024211 0.1556 0.26 0.06
## prime_semanticsneutral:lengthShort 0.016148 0.1271 -0.29 -0.84
## Residual 0.741309 0.8610
##
##
##
##
## -0.44
##
## Number of obs: 7431, groups: SubjNo, 26
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 1.28371 0.10961 11.712
## prime_semanticsneutral 0.04460 0.03343 1.334
## lengthShort 0.17688 0.04166 4.246
## prime_semanticsneutral:lengthShort -0.01061 0.04715 -0.225
##
## Correlation of Fixed Effects:
## (Intr) prm_sm lngthS
## prm_smntcsn 0.203
## lengthShort 0.101 0.310
## prm_smntc:S -0.072 -0.744 -0.578## Linear mixed model fit by REML ['lmerMod']
## Formula:
## rt.log ~ prime_semantics * length + (1 + prime_semantics * length |
## SubjNo)
## Data: sense.pop
##
## REML criterion at convergence: 7360.1
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.8875 -0.6100 -0.1794 0.4133 5.7987
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## SubjNo (Intercept) 0.118029 0.34355
## prime_semanticsneutral 0.002264 0.04758 0.17
## lengthShort 0.003328 0.05769 0.05 -0.31
## prime_semanticsneutral:lengthShort 0.004323 0.06575 -0.11 -0.98
## Residual 0.153453 0.39173
##
##
##
##
## 0.13
##
## Number of obs: 7431, groups: SubjNo, 26
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 0.0957248 0.0679880 1.408
## prime_semanticsneutral 0.0175018 0.0158849 1.102
## lengthShort 0.1097062 0.0171520 6.396
## prime_semanticsneutral:lengthShort -0.0002744 0.0223148 -0.012
##
## Correlation of Fixed Effects:
## (Intr) prm_sm lngthS
## prm_smntcsn 0.023
## lengthShort -0.038 0.181
## prm_smntc:S -0.006 -0.799 -0.383