2.2 Results
Behavioral measures. Valence ratings of all Mandarin target words
by all participants categorized by emotion type are illustrated inFigure 1 . Neutral words (mean=3.2, SD=0.6) were rated lower
than positive words (mean=4.3, SD=0.9 ; β = 1.05, se = 0.1229, t = 8.5,
p<.001 ), and higher than negative words (mean=1.6, SD=0.8; β
= 1.58, se = 0.1275, t = 12.38, p<.001 ). The re-leveled model
with “negative” as intercept revealed a significant difference between
negative and positive words (β = 2.63, se = 0.1499, t = 17.54,
p<.001). We also calculated the percentage of items that were
rated per participant within the expected range for their predesignated
category: negative (1 or 2), neutral (3) or positive (4 or 5). This
included 80.9% of items across all participants, with less Neutral
words (M=73.8%, sd=17.4%) than Positive words (M=83.0%, sd=10.5%; β
=.092, se =.039, t = 2.321, p<.05) or Negative words
(M=85.9%, sd=13.6%; β =.121, se =.039, t = 3.041, p<.001),
while no difference was found between Positive and Negative words (β
=-.028, se =.039, t = -0.720, p=.475).
ERP analyses. Analyses included all trails free from artefact,
for items that were rated within the expected range for their
predesignated category: negative (1 or 2), neutral (3) or positive (4 or
5). The percentage of trials lost due to artefact was 16% for the
Neutral words, 14% for the Positive words and 16% for Negative words,
and to ascertain whether the valence of target words induced differences
in the ERP trace, we conducted a two-tailed permutation t-test (1000
random partitions) between Neutral and Negative target words, and
between Neutral and Positive target words for all selected electrodes
(midline: Fz, FCz, Cz, CPz, Pz, frontal central : FC1, FC3, FC5, FC2,
FC4, FC6, central-parietal: C1, C3, C5, C2, C4, C6, CP1, CP3, CP5, CP2,
CP4, CP6, and parietal: P1, P3, P5, P2, P4, P6), with time points of 5
ms across the entire epoch (0 to 1200 ms after stimulus onset). Only
differences that persisted for 10 ms or more were considered
statistically significant. The results revealed no significant
differences between Neutral and Positive, or between Neutral and
Negative target words at any selected electrode. Figure 2 shows
the average response across all 20 participants for 9 central partiel
cites; none of the permutation tests are revealed due to there being no
significant effects.
To further examine possible differences induced by Valence, we modeled
the data using linear mixed effect regression, with the LmerTest package
(Kuznetsova & Christensen, 2017) in R (R Core Team, 2017). The models,
summarized in Table 2 , included the treatment coded fixed
factor Valence (Neutral, Negative, Positive), with random intercept for
Item and Participant. A random slope of Valence for Participant was
included. Models were performed independently at 3 ROIs for the mean
voltage amplitude at the four time windows post stimulus onset,
associated with the P200 (150-300 msec), N400 (300-500 msec), LPC
(500-800 msec) and later component (800-1200 msec). No significant
differences were found in any ROI, at any time window (cf. Table
2 ).