title: “L1 French eeg amplitude analyses” author: “Coughlin” date: “September 29, 2015” output: html_document

L1 French: EEG amplitude analyses

Data acquisition

This file contains eeg data collected from right-handed native French speakers (n=30), living in Canada. They participated in a masked-priming lexical decision task. Testing took approximately 45 minutes with short breaks.

A given trial in the experiment followed this timeline

Fixation (for blinks) 1000ms
Blank screen 1000ms
Mask (##########) 500 ms
Prime 50 ms
Mask 20 ms
Target 600 ms
Blank screen (for response) 1000 ms

Participants wore an electrode cap from ANT with 64 embedded channels. Left and right mastoid electrodes were embedded in the cap. Each electrode’s cable is shielded individually to guard against noise. Impedance was kept below 5 Kohms. There were no electrodes placed on the face. Only 23 electodes (including both mastoids) were recorded. Left (A1) mastoid was used as reference elecrode during recording.

Data processing

Data were processed offline in the following order, with the given parameters

All scalp electrodes were re-referenced to A1 & A2
Filter with a bandpass of 0.3 Hz to 30 Hz
Artifact rejection done automatically by standard deviation surpassing 30 Hz, in a 200ms moving-window. Rejections were then manually checked.
Averaging for each participant was for a time-window of -570 to 1100 ms; a rejection window of -570 to 700 ms; and a baseline of -470 to -270 ms. (note: rejection window was originally same length as averaging window, but an unacceptable amount of data was lost on the right-side due to blinks)

Analysing the EEG amplitude data

Set wd and load in all the .txt files

setwd("C:/Users/Katie/Desktop/Research/Dissertation/STUDIES/Studies 3 and 4 EEG/EEG Native French/Amplitude data")

data<-read.table("kate-output-500600.txt",header=TRUE)

dim(data)

## [1] 31920     5

head(data)

##   subj cond chan        win     mean
## 1 KC01 fidr  Fp1 +100..+301  3.13820
## 2 KC01 fidr  Fp1 +301..+500 -0.72904
## 3 KC01 fidr  Fp1 +500..+600 -1.84420
## 4 KC01 fidr  Fp1 +176..+275  2.17400
## 5 KC01 fidr  Fp1 +350..+449 -1.99730
## 6 KC01 fidr  Fp1 +449..+551 -1.14940

str(data)

## 'data.frame':    31920 obs. of  5 variables:
##  $ subj: Factor w/ 30 levels "KC01","KC02",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ cond: Factor w/ 8 levels "fidr","fidu",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ chan: Factor w/ 19 levels "C3","C4","Cz",..: 8 8 8 8 8 8 8 9 9 9 ...
##  $ win : Factor w/ 7 levels "+100..+301","+176..+275",..: 1 4 7 2 5 6 3 1 4 7 ...
##  $ mean: num  3.138 -0.729 -1.844 2.174 -1.997 ...

In the original dataframe condition and relatedness are encoded in the same column. The below code creates new columns for just condition and just relatedness

library(qdap)
cond<-as.character(unique(data[,2]))
conditions<-c("id","id","morph","morph", "orth", "orth", "sem","sem")
match_cond<-cbind(cond,conditions)
relateds<-c("rel","unrel","rel","unrel", "rel", "unrel", "rel","unrel")
match_rel<-cbind(cond,relateds)

data$Condition<-lookup(data$cond,match_cond)
data$Related<-lookup(data$cond,match_rel)
data$Condition<-as.factor(as.character(data$Condition))
data$Related<-as.factor(as.character(data$Related))
head(data,15)

##    subj cond chan        win     mean Condition Related
## 1  KC01 fidr  Fp1 +100..+301  3.13820        id     rel
## 2  KC01 fidr  Fp1 +301..+500 -0.72904        id     rel
## 3  KC01 fidr  Fp1 +500..+600 -1.84420        id     rel
## 4  KC01 fidr  Fp1 +176..+275  2.17400        id     rel
## 5  KC01 fidr  Fp1 +350..+449 -1.99730        id     rel
## 6  KC01 fidr  Fp1 +449..+551 -1.14940        id     rel
## 7  KC01 fidr  Fp1 +176..+301  1.77600        id     rel
## 8  KC01 fidr  Fp2 +100..+301  3.97990        id     rel
## 9  KC01 fidr  Fp2 +301..+500  0.70052        id     rel
## 10 KC01 fidr  Fp2 +500..+600 -0.54733        id     rel
## 11 KC01 fidr  Fp2 +176..+275  2.75780        id     rel
## 12 KC01 fidr  Fp2 +350..+449 -0.55682        id     rel
## 13 KC01 fidr  Fp2 +449..+551  0.51648        id     rel
## 14 KC01 fidr  Fp2 +176..+301  2.22200        id     rel
## 15 KC01 fidr   F3 +100..+301  3.00200        id     rel

Remove participants 4, 15,and 27 are removed because the number of trials rejected during averaging was excessive (e.g., 15 of 36 in more than 1 condition. Rejection was equal across conditions).

data<-data[data$subj!= "KC04"  &data$subj!= "KC15"   &data$subj!="KC27" ,]
data$subj<-factor(data$subj)

Relevel to make the unrelated condition the baseline

data$Related<-relevel(data$Related,ref="unrel")

Create subset of data just for midline electrodes (Fz,Cz,Pz) and for lateral electrodes. The dataframe for lateral electrodes (F3/4, C3/4, P3/4) then gets a variable labelling the hemisphere

midline<-data[data$chan=="Fz"| data$chan=="Cz" | data$chan=="Pz",]
midline$chan<-factor(midline$chan)

lateral<-data[data$chan=="F3"| data$chan=="F4"| data$chan=="C3" | data$chan=="C4" | data$chan=="P3"  | data$chan=="P4",]
lateral$chan<-factor(lateral$chan)

hem<-as.character(unique(lateral[,3]))
sides<-c("left","right","left","right", "left", "right")
match_sides<-cbind(hem,sides)
lateral$hemisphere<-lookup(lateral$chan,match_sides)
head(lateral)

##    subj cond chan        win     mean Condition Related hemisphere
## 15 KC01 fidr   F3 +100..+301  3.00200        id     rel       left
## 16 KC01 fidr   F3 +301..+500 -1.38630        id     rel       left
## 17 KC01 fidr   F3 +500..+600 -1.23540        id     rel       left
## 18 KC01 fidr   F3 +176..+275  1.56620        id     rel       left
## 19 KC01 fidr   F3 +350..+449 -2.46650        id     rel       left
## 20 KC01 fidr   F3 +449..+551 -0.77718        id     rel       left

Create time windows for midline and lateral (separate)

Averageing was done in the following time-windows. These time windows correspond to the peaks found in the waveform plots

100 to 300 ms
300 to 500 ms
500 to 600 ms

mid100<-midline[midline$win=="+100..+301",]
mid300<-midline[midline$win=="+301..+500",]
mid500<-midline[midline$win=="+500..+600",]

lat100<-lateral[lateral$win=="+100..+301",]
lat300<-lateral[lateral$win=="+301..+500",]
lat500<-lateral[lateral$win=="+500..+600",]

Linear mixed-effects models

Separate models were used for each of the 3 time windows, for each of the sets of electrodes (midline and lateral), for a total of 6 global analyses
The dependent variable is the mean amplitude, and the fixed independent variables are condition and relatedness (and their interaction terms). For midline models, electrode was included in the model. For the lateral models, hemisphere was included. All models had subject as a random effect.
Models were first fit with all fixed effects and interaction terms. Interaction terms were removed and the model was tested with an anova to determine if the fit is significantly improved by including the interaction term.
The best-fit model is the only that is printed for a given analysis.

Timewindow 100-300; midline electrodes

library(lme4)
library(lmerTest)
midline100_1<-lmer(mean~ Condition * Related * chan +(1|subj),mid100)
midline100_2<-lmer(mean~ Condition * Related * chan - Condition:Related:chan +(1|subj),mid100)
midline100_3<-lmer(mean~ Condition * Related * chan - Condition:Related:chan-chan:Related +(1|subj),mid100)
midline100_4<-lmer(mean~ Condition * Related * chan - Condition:Related:chan-chan:Condition +(1|subj),mid100)
midline100_5<-lmer(mean~ Condition * Related + chan + (1|subj),mid100)

summary(midline100_5)

## Linear mixed model fit by REML t-tests use Satterthwaite approximations
##   to degrees of freedom [lmerMod]
## Formula: mean ~ Condition * Related + chan + (1 | subj)
##    Data: mid100
## 
## REML criterion at convergence: 2735.9
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.2858 -0.6560  0.0166  0.6016  4.4950 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  subj     (Intercept) 5.392    2.322   
##  Residual             3.425    1.851   
## Number of obs: 648, groups:  subj, 27
## 
## Fixed effects:
##                            Estimate Std. Error        df t value Pr(>|t|)
## (Intercept)                 1.46127    0.50254  39.40000   2.908 0.005954
## Conditionmorph             -0.41910    0.29079 612.00000  -1.441 0.150026
## Conditionorth              -0.39695    0.29079 612.00000  -1.365 0.172728
## Conditionsem               -0.01838    0.29079 612.00000  -0.063 0.949623
## Relatedrel                  0.92100    0.29079 612.00000   3.167 0.001615
## chanFz                      1.56326    0.17807 612.00000   8.779  < 2e-16
## chanPz                     -0.50007    0.17807 612.00000  -2.808 0.005140
## Conditionmorph:Relatedrel  -0.21572    0.41124 612.00000  -0.525 0.600084
## Conditionorth:Relatedrel   -0.61098    0.41124 612.00000  -1.486 0.137868
## Conditionsem:Relatedrel    -1.57399    0.41124 612.00000  -3.827 0.000143
##                              
## (Intercept)               ** 
## Conditionmorph               
## Conditionorth                
## Conditionsem                 
## Relatedrel                ** 
## chanFz                    ***
## chanPz                    ** 
## Conditionmorph:Relatedrel    
## Conditionorth:Relatedrel     
## Conditionsem:Relatedrel   ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) Cndtnm Cndtnr Cndtns Rltdrl chanFz chanPz Cndtnm:R
## Conditnmrph -0.289                                                   
## Conditinrth -0.289  0.500                                            
## Conditionsm -0.289  0.500  0.500                                     
## Relatedrel  -0.289  0.500  0.500  0.500                              
## chanFz      -0.177  0.000  0.000  0.000  0.000                       
## chanPz      -0.177  0.000  0.000  0.000  0.000  0.500                
## Cndtnmrph:R  0.205 -0.707 -0.354 -0.354 -0.707  0.000  0.000         
## Cndtnrth:Rl  0.205 -0.354 -0.707 -0.354 -0.707  0.000  0.000  0.500  
## Cndtnsm:Rlt  0.205 -0.354 -0.354 -0.707 -0.707  0.000  0.000  0.500  
##             Cndtnr:R
## Conditnmrph         
## Conditinrth         
## Conditionsm         
## Relatedrel          
## chanFz              
## chanPz              
## Cndtnmrph:R         
## Cndtnrth:Rl         
## Cndtnsm:Rlt  0.500

The best model does not include any interaction terms for the electrode
There is an effect of relatedness (in ID); no interaction of morph x related and no interaction of orth x related. But there is an interaction of sem x related.

Timewindow 300-500; midline electrodes

midline300_1<-lmer(mean~ Condition * Related * chan +(1|subj),mid300)
midline300_2<-lmer(mean~ Condition * Related * chan - Condition:Related:chan +(1|subj),mid300)
midline300_3<-lmer(mean~ Condition * Related * chan - Condition:Related:chan-chan:Related +(1|subj),mid300)
midline300_4<-lmer(mean~ Condition * Related * chan - Condition:Related:chan-chan:Condition +(1|subj),mid300)
midline300_5<-lmer(mean~ Condition * Related + chan + (1|subj),mid300)

summary(midline300_5)

## Linear mixed model fit by REML t-tests use Satterthwaite approximations
##   to degrees of freedom [lmerMod]
## Formula: mean ~ Condition * Related + chan + (1 | subj)
##    Data: mid300
## 
## REML criterion at convergence: 2971.2
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -5.2223 -0.6280 -0.0053  0.6465  3.0226 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  subj     (Intercept) 10.127   3.182   
##  Residual              4.899   2.213   
## Number of obs: 648, groups:  subj, 27
## 
## Fixed effects:
##                            Estimate Std. Error        df t value Pr(>|t|)
## (Intercept)                 1.94442    0.67132  36.00000   2.896 0.006381
## Conditionmorph             -0.09196    0.34780 612.00000  -0.264 0.791555
## Conditionorth              -0.50089    0.34780 612.00000  -1.440 0.150331
## Conditionsem               -0.58529    0.34780 612.00000  -1.683 0.092916
## Relatedrel                  1.59678    0.34780 612.00000   4.591 5.35e-06
## chanFz                     -0.70613    0.21298 612.00000  -3.315 0.000969
## chanPz                      2.33104    0.21298 612.00000  10.945  < 2e-16
## Conditionmorph:Relatedrel  -0.27916    0.49186 612.00000  -0.568 0.570539
## Conditionorth:Relatedrel   -1.11518    0.49186 612.00000  -2.267 0.023721
## Conditionsem:Relatedrel    -0.97735    0.49186 612.00000  -1.987 0.047363
##                              
## (Intercept)               ** 
## Conditionmorph               
## Conditionorth                
## Conditionsem              .  
## Relatedrel                ***
## chanFz                    ***
## chanPz                    ***
## Conditionmorph:Relatedrel    
## Conditionorth:Relatedrel  *  
## Conditionsem:Relatedrel   *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) Cndtnm Cndtnr Cndtns Rltdrl chanFz chanPz Cndtnm:R
## Conditnmrph -0.259                                                   
## Conditinrth -0.259  0.500                                            
## Conditionsm -0.259  0.500  0.500                                     
## Relatedrel  -0.259  0.500  0.500  0.500                              
## chanFz      -0.159  0.000  0.000  0.000  0.000                       
## chanPz      -0.159  0.000  0.000  0.000  0.000  0.500                
## Cndtnmrph:R  0.183 -0.707 -0.354 -0.354 -0.707  0.000  0.000         
## Cndtnrth:Rl  0.183 -0.354 -0.707 -0.354 -0.707  0.000  0.000  0.500  
## Cndtnsm:Rlt  0.183 -0.354 -0.354 -0.707 -0.707  0.000  0.000  0.500  
##             Cndtnr:R
## Conditnmrph         
## Conditinrth         
## Conditionsm         
## Relatedrel          
## chanFz              
## chanPz              
## Cndtnmrph:R         
## Cndtnrth:Rl         
## Cndtnsm:Rlt  0.500

The best model does not include any interaction terms for the electrode
There is an effect of relatedness (in ID), and no interaction of morph x related. But there is an interaction of orth x related and sem x related.

Timewindow 500-600; midline electrodes

midline500_1<-lmer(mean~ Condition * Related * chan +(1|subj),mid500)
midline500_2<-lmer(mean~ Condition * Related * chan - Condition:Related:chan +(1|subj),mid500)
midline500_3<-lmer(mean~ Condition * Related * chan - Condition:Related:chan-chan:Related +(1|subj),mid500)
midline500_4<-lmer(mean~ Condition * Related * chan - Condition:Related:chan-chan:Condition +(1|subj),mid500)
midline500_5<-lmer(mean~ Condition * Related + chan + (1|subj),mid500)

summary(midline500_5)

## Linear mixed model fit by REML t-tests use Satterthwaite approximations
##   to degrees of freedom [lmerMod]
## Formula: mean ~ Condition * Related + chan + (1 | subj)
##    Data: mid500
## 
## REML criterion at convergence: 3049.1
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -4.2167 -0.6411  0.0236  0.6197  3.0881 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  subj     (Intercept) 7.616    2.760   
##  Residual             5.629    2.373   
## Number of obs: 648, groups:  subj, 27
## 
## Fixed effects:
##                             Estimate Std. Error         df t value
## (Intercept)                 6.003786   0.607404  41.700000   9.884
## Conditionmorph             -0.058753   0.372811 612.000000  -0.158
## Conditionorth              -0.657656   0.372811 612.000000  -1.764
## Conditionsem               -0.128335   0.372811 612.000000  -0.344
## Relatedrel                 -0.468632   0.372811 612.000000  -1.257
## chanFz                     -2.764886   0.228299 612.000000 -12.111
## chanPz                      1.721252   0.228299 612.000000   7.539
## Conditionmorph:Relatedrel   0.035821   0.527235 612.000000   0.068
## Conditionorth:Relatedrel   -0.165936   0.527235 612.000000  -0.315
## Conditionsem:Relatedrel    -0.002989   0.527235 612.000000  -0.006
##                           Pr(>|t|)    
## (Intercept)               1.70e-12 ***
## Conditionmorph              0.8748    
## Conditionorth               0.0782 .  
## Conditionsem                0.7308    
## Relatedrel                  0.2092    
## chanFz                     < 2e-16 ***
## chanPz                    1.71e-13 ***
## Conditionmorph:Relatedrel   0.9459    
## Conditionorth:Relatedrel    0.7531    
## Conditionsem:Relatedrel     0.9955    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) Cndtnm Cndtnr Cndtns Rltdrl chanFz chanPz Cndtnm:R
## Conditnmrph -0.307                                                   
## Conditinrth -0.307  0.500                                            
## Conditionsm -0.307  0.500  0.500                                     
## Relatedrel  -0.307  0.500  0.500  0.500                              
## chanFz      -0.188  0.000  0.000  0.000  0.000                       
## chanPz      -0.188  0.000  0.000  0.000  0.000  0.500                
## Cndtnmrph:R  0.217 -0.707 -0.354 -0.354 -0.707  0.000  0.000         
## Cndtnrth:Rl  0.217 -0.354 -0.707 -0.354 -0.707  0.000  0.000  0.500  
## Cndtnsm:Rlt  0.217 -0.354 -0.354 -0.707 -0.707  0.000  0.000  0.500  
##             Cndtnr:R
## Conditnmrph         
## Conditinrth         
## Conditionsm         
## Relatedrel          
## chanFz              
## chanPz              
## Cndtnmrph:R         
## Cndtnrth:Rl         
## Cndtnsm:Rlt  0.500

The best model does not include any interaction terms for the electrode
There is no effect of relatedness (in ID) and no interaction with any other condition x relatedness. There is a marginal effect of condition for orth

Timewindow 100-300; lateral electrodes

library(lme4)
library(lmerTest)
lateral100_1<-lmer(mean~ Condition * Related * hemisphere +(1|subj),lat100)
lateral100_2<-lmer(mean~ Condition * Related * hemisphere-Condition:Related:hemisphere +(1|subj),lat100)
lateral100_3<-lmer(mean~ Condition * Related * hemisphere-Condition:Related:hemisphere -hemisphere:Related +(1|subj),lat100)
lateral100_4<-lmer(mean~ Condition * Related * hemisphere-Condition:Related:hemisphere -hemisphere:Condition +(1|subj),lat100)
lateral100_5<-lmer(mean~ Condition * Related + hemisphere +(1|subj),lat100)
lateral100_6<-lmer(mean~ Condition * Related +(1|subj),lat100)

summary(lateral100_5)

## Linear mixed model fit by REML t-tests use Satterthwaite approximations
##   to degrees of freedom [lmerMod]
## Formula: mean ~ Condition * Related + hemisphere + (1 | subj)
##    Data: lat100
## 
## REML criterion at convergence: 5997.7
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.1475 -0.6751 -0.0167  0.6282  4.4049 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  subj     (Intercept) 5.079    2.254   
##  Residual             5.525    2.351   
## Number of obs: 1296, groups:  subj, 27
## 
## Fixed effects:
##                             Estimate Std. Error         df t value
## (Intercept)                  1.62019    0.47591   36.00000   3.404
## Conditionmorph              -0.42958    0.26117 1261.00000  -1.645
## Conditionorth               -0.24350    0.26117 1261.00000  -0.932
## Conditionsem                -0.08420    0.26117 1261.00000  -0.322
## Relatedrel                   0.57523    0.26117 1261.00000   2.203
## hemisphereright             -0.32640    0.13058 1261.00000  -2.500
## Conditionmorph:Relatedrel    0.07358    0.36935 1261.00000   0.199
## Conditionorth:Relatedrel    -0.55051    0.36935 1261.00000  -1.491
## Conditionsem:Relatedrel     -1.14263    0.36935 1261.00000  -3.094
##                           Pr(>|t|)   
## (Intercept)                0.00164 **
## Conditionmorph             0.10025   
## Conditionorth              0.35134   
## Conditionsem               0.74722   
## Relatedrel                 0.02781 * 
## hemisphereright            0.01256 * 
## Conditionmorph:Relatedrel  0.84213   
## Conditionorth:Relatedrel   0.13634   
## Conditionsem:Relatedrel    0.00202 **
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) Cndtnm Cndtnr Cndtns Rltdrl hmsphr Cndtnm:R Cndtnr:R
## Conditnmrph -0.274                                                     
## Conditinrth -0.274  0.500                                              
## Conditionsm -0.274  0.500  0.500                                       
## Relatedrel  -0.274  0.500  0.500  0.500                                
## hemsphrrght -0.137  0.000  0.000  0.000  0.000                         
## Cndtnmrph:R  0.194 -0.707 -0.354 -0.354 -0.707  0.000                  
## Cndtnrth:Rl  0.194 -0.354 -0.707 -0.354 -0.707  0.000  0.500           
## Cndtnsm:Rlt  0.194 -0.354 -0.354 -0.707 -0.707  0.000  0.500    0.500

The best model includes hemisphere just as main effect
The results show an effect of relatedness (in iD), and an interaction of sem x related, but no interaction of morph x related or orth x related. There is also an effect of hemisphere (in ID)

Timewindow 300-500; lateral electrodes

lateral300_1<-lmer(mean~ Condition * Related * hemisphere +(1|subj),lat300)
lateral300_2<-lmer(mean~ Condition * Related * hemisphere-Condition:Related:hemisphere +(1|subj),lat300)
lateral300_3<-lmer(mean~ Condition * Related * hemisphere-Condition:Related:hemisphere -hemisphere:Related +(1|subj),lat300)
lateral300_4<-lmer(mean~ Condition * Related * hemisphere-Condition:Related:hemisphere -hemisphere:Condition +(1|subj),lat300)
lateral300_5<-lmer(mean~ Condition * Related + hemisphere +(1|subj),lat300)
lateral300_6<-lmer(mean~ Condition * Related +(1|subj),lat300)

summary(lateral300_5)

## Linear mixed model fit by REML t-tests use Satterthwaite approximations
##   to degrees of freedom [lmerMod]
## Formula: mean ~ Condition * Related + hemisphere + (1 | subj)
##    Data: lat300
## 
## REML criterion at convergence: 5896.8
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -4.2649 -0.6330 -0.0242  0.5735  3.9490 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  subj     (Intercept) 7.305    2.703   
##  Residual             5.063    2.250   
## Number of obs: 1296, groups:  subj, 27
## 
## Fixed effects:
##                            Estimate Std. Error        df t value Pr(>|t|)
## (Intercept)                  2.2001     0.5529   32.2000   3.979 0.000368
## Conditionmorph              -0.2083     0.2500 1261.0000  -0.833 0.404946
## Conditionorth               -0.4691     0.2500 1261.0000  -1.876 0.060832
## Conditionsem                -0.5635     0.2500 1261.0000  -2.254 0.024361
## Relatedrel                   1.2176     0.2500 1261.0000   4.870 1.25e-06
## hemisphereright              0.6382     0.1250 1261.0000   5.106 3.80e-07
## Conditionmorph:Relatedrel    0.0382     0.3536 1261.0000   0.108 0.913988
## Conditionorth:Relatedrel    -0.8901     0.3536 1261.0000  -2.518 0.011940
## Conditionsem:Relatedrel     -0.7552     0.3536 1261.0000  -2.136 0.032866
##                              
## (Intercept)               ***
## Conditionmorph               
## Conditionorth             .  
## Conditionsem              *  
## Relatedrel                ***
## hemisphereright           ***
## Conditionmorph:Relatedrel    
## Conditionorth:Relatedrel  *  
## Conditionsem:Relatedrel   *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) Cndtnm Cndtnr Cndtns Rltdrl hmsphr Cndtnm:R Cndtnr:R
## Conditnmrph -0.226                                                     
## Conditinrth -0.226  0.500                                              
## Conditionsm -0.226  0.500  0.500                                       
## Relatedrel  -0.226  0.500  0.500  0.500                                
## hemsphrrght -0.113  0.000  0.000  0.000  0.000                         
## Cndtnmrph:R  0.160 -0.707 -0.354 -0.354 -0.707  0.000                  
## Cndtnrth:Rl  0.160 -0.354 -0.707 -0.354 -0.707  0.000  0.500           
## Cndtnsm:Rlt  0.160 -0.354 -0.354 -0.707 -0.707  0.000  0.500    0.500

The best model includes the effect of hemisphere, but no interactions.
The results show an effect of relatedness (in ID), and an interaction of orth x related and sem x related, but no interaction of morph x related. There is also an effect of hemisphere, and an effect of condition for sem. There is also a marginal effect of condition for orth.

Timewindow 500-600; lateral electrodes

lateral500_1<-lmer(mean~ Condition * Related * hemisphere +(1|subj),lat500)
lateral500_2<-lmer(mean~ Condition * Related * hemisphere-Condition:Related:hemisphere +(1|subj),lat500)
lateral500_3<-lmer(mean~ Condition * Related * hemisphere-Condition:Related:hemisphere -hemisphere:Related +(1|subj),lat500)
lateral500_4<-lmer(mean~ Condition * Related * hemisphere-Condition:Related:hemisphere -hemisphere:Condition +(1|subj),lat500)
lateral500_5<-lmer(mean~ Condition * Related + hemisphere +(1|subj),lat500)
lateral500_6<-lmer(mean~ Condition * Related +(1|subj),lat500)
summary(lateral500_5)

## Linear mixed model fit by REML t-tests use Satterthwaite approximations
##   to degrees of freedom [lmerMod]
## Formula: mean ~ Condition * Related + hemisphere + (1 | subj)
##    Data: lat500
## 
## REML criterion at convergence: 6245.3
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -3.6103 -0.6363 -0.0165  0.6458  3.0096 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  subj     (Intercept) 7.379    2.717   
##  Residual             6.672    2.583   
## Number of obs: 1296, groups:  subj, 27
## 
## Fixed effects:
##                             Estimate Std. Error         df t value
## (Intercept)                  4.86920    0.56537   34.20000   8.612
## Conditionmorph              -0.29187    0.28701 1261.00000  -1.017
## Conditionorth               -0.63836    0.28701 1261.00000  -2.224
## Conditionsem                -0.15844    0.28701 1261.00000  -0.552
## Relatedrel                  -0.58423    0.28701 1261.00000  -2.036
## hemisphereright              0.26462    0.14350 1261.00000   1.844
## Conditionmorph:Relatedrel    0.43958    0.40589 1261.00000   1.083
## Conditionorth:Relatedrel    -0.07224    0.40589 1261.00000  -0.178
## Conditionsem:Relatedrel      0.27039    0.40589 1261.00000   0.666
##                           Pr(>|t|)    
## (Intercept)               4.35e-10 ***
## Conditionmorph              0.3094    
## Conditionorth               0.0263 *  
## Conditionsem                0.5810    
## Relatedrel                  0.0420 *  
## hemisphereright             0.0654 .  
## Conditionmorph:Relatedrel   0.2790    
## Conditionorth:Relatedrel    0.8588    
## Conditionsem:Relatedrel     0.5054    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) Cndtnm Cndtnr Cndtns Rltdrl hmsphr Cndtnm:R Cndtnr:R
## Conditnmrph -0.254                                                     
## Conditinrth -0.254  0.500                                              
## Conditionsm -0.254  0.500  0.500                                       
## Relatedrel  -0.254  0.500  0.500  0.500                                
## hemsphrrght -0.127  0.000  0.000  0.000  0.000                         
## Cndtnmrph:R  0.179 -0.707 -0.354 -0.354 -0.707  0.000                  
## Cndtnrth:Rl  0.179 -0.354 -0.707 -0.354 -0.707  0.000  0.500           
## Cndtnsm:Rlt  0.179 -0.354 -0.354 -0.707 -0.707  0.000  0.500    0.500

The best model includes hemisphere as a main effect, but no interaction term of hemisphere
The results show an effect of relatedness (in ID), but no interaction of any condition x related. There is a marginal effect of hemisphere, and a significant effect of the orthographic condition

Summary of time-window global analyses

Early time-windows (100-300 ms): In the midline and lateral electrodes there is an attenuation of the early negativity when the prime-target pairs have orthographic overlap. That is, the pattern of relatedness found in the ID condition does not differ from the pattern of relatedness found in the morphological or orthographic conditions.
In the middle time-window (300-500 ms): In the middle and lateral electrodes there is an attenuation of the negativity when the prime-target pairs have morphological overlap. The effect of relatedness found in the ID condition is also found in the morphological condition, but it is not found in the orthographic condition,.
In the late time-window (500-600 ms): In the lateral electrodes there is an effect of relatedness, but no interaction of condition x relatedness for any condition. In the midline electrodes there is no effect of relatedness, and no interactions of condition x relatedness for any condition. Interestingly, there is an effect of orthographic condition in this time-window (marginal at midline, significant in lateral). This indicates an overal difference of mean in the orth condition compared to the ID condition. Looking at the difference waves in the orthographic condition, and the topographic plot, it is noticeable that the orth related condition has a positivity in this time-window (relative to its unrelated prime), whereas other conditions’ related primes elicit a more negative waveform. The effect of relatedness in the late time-window for the orth condition will be further investigated (below).
Semantic overlap between prime-target pairs does not modulate the effect of relatedness in any time-window, suggesting the masked-priming paradigm successfully inhibited semantic priming

Orthography in the late time-window

Create dataframes with summaries of mean amplitudes

orth_late_mid<-mid500[mid500$Condition=="orth",]
orth_late_lat<-lat500[lat500$Condition=="orth",]

id_late_mid<-mid500[mid500$Condition=="id",]
id_late_lat<-lat500[lat500$Condition=="id",]

morph_late_mid<-mid500[mid500$Condition=="morph",]
morph_late_lat<-lat500[lat500$Condition=="morph",]

sem_late_mid<-mid500[mid500$Condition=="sem",]
sem_late_lat<-lat500[lat500$Condition=="sem",]

Summarize means

library(plyr)


means <- ddply(mid500, c("Condition", "Related"), summarize,
          
               mean = mean(mean)
              
)

means

##   Condition Related     mean
## 1        id   unrel 5.655908
## 2        id     rel 5.187276
## 3     morph   unrel 5.597155
## 4     morph     rel 5.164344
## 5      orth   unrel 4.998251
## 6      orth     rel 4.363683
## 7       sem   unrel 5.527573
## 8       sem     rel 5.055952

If you look at the means for the other conditions in this time-window you see that the mean for orth-unrelated is lower than the other unrelateds, and the orth-related is also lower than the other relateds.
So this late effect may just be a consequence of the lexical properties of the orthographic targets (independent of priming)

check effect of condition for just unrelated items

Late time-window, lateral electrodes

unrelated<-lat500[lat500$Related=="unrel",]
unr1<-lmer(mean~Condition+hemisphere+(1|subj),unrelated)
summary(unr1)

## Linear mixed model fit by REML t-tests use Satterthwaite approximations
##   to degrees of freedom [lmerMod]
## Formula: mean ~ Condition + hemisphere + (1 | subj)
##    Data: unrelated
## 
## REML criterion at convergence: 3119.1
## 
## Scaled residuals: 
##      Min       1Q   Median       3Q      Max 
## -2.82794 -0.66867 -0.01765  0.64355  2.59451 
## 
## Random effects:
##  Groups   Name        Variance Std.Dev.
##  subj     (Intercept) 8.261    2.874   
##  Residual             6.248    2.500   
## Number of obs: 648, groups:  subj, 27
## 
## Fixed effects:
##                 Estimate Std. Error       df t value Pr(>|t|)    
## (Intercept)       4.8663     0.5951  32.7000   8.177 2.06e-09 ***
## Conditionmorph   -0.2919     0.2777 617.0000  -1.051   0.2937    
## Conditionorth    -0.6384     0.2777 617.0000  -2.298   0.0219 *  
## Conditionsem     -0.1584     0.2777 617.0000  -0.570   0.5686    
## hemisphereright   0.2704     0.1964 617.0000   1.377   0.1691    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##             (Intr) Cndtnm Cndtnr Cndtns
## Conditnmrph -0.233                     
## Conditinrth -0.233  0.500              
## Conditionsm -0.233  0.500  0.500       
## hemsphrrght -0.165  0.000  0.000  0.000

Independent of priming, the orth targets seem to elicit this late reduced positivity.
Not yet sure what this means.