Neuroscience Letters

The affective meaning of words can be quantified statistically by the "semantic differential technique". We studied a total of 55 Chinese adults in two experiments: first, 210 nouns were rated by 32 subjects. Factor analysis on these data yielded three independent semantic dimensions. Semantically unique words were used in electrophysiological experiments in another group of 23 healthy right-handed adults. Words of similar physical appearance belonging to different semantic classes were presented visually in random order. The EEG was recorded in 32 channels. Evoked activity was computed for each semantic class. Significant differences in electrical brain activation between semantic word classes were seen as early as 80 ms after stimulus onset confirming related findings in German subjects. These results illustrate similar early neural activation in subject groups of different language and culture. Neuroscience Letters, 2003, 341, 45-48 2 W. Skrandies & M.J. Chiu, Dimensions of Affective Semantic Meaning of Chinese Language It has been shown that the recording of electrical brain activity with high time resolution elicited by linguistic material yields sensitive neuronal indicators of human language processing [10, 17]. In an earlier report from our laboratory we demonstrated that semantic meaning of words influences already very early stages of processing as reflected by changes of evoked potential fields [13]. This is in contrast to studies concerned with recordings of brain activity during language processing that centered on the so-called N400 component which occurs at long latency after the presentation of a word which unexpectedly closes a sentence [6], and relates to the comprehension of sentences or jokes [1, 3]. According to the pioneering work of Osgood et al. [9], also in Hongkong Chinese the affective meaning of words can be quantified by the "semantic differential technique" that results in statistically defined independent dimensions where every word is uniquely located on the three dimensions evaluation (E, "good – bad"), potency (P, "strong – weak"), and activity (A, "active – passive"). Brain potentials elicited during reading of single German words show systematic influences of meaning, and these effects occur as early as 100 ms [13]. In the present study we extend such experiments to Chinese adults. We are concerned with the question of whether the semantic dimensions are similar in German and Chinese subjects, and whether early electrophysiological effects can also be observed with different language material in subjects with a vastly different cultural background. A total of 55 Chinese adults participated in two experiments: first, 210 nouns were rated on twelve adjective scales of opposite descriptors by 32 subjects (mean age 29.3; SD 9.6 years; 27 females, 5 males). All words consisted of two characters and had a comparable frequency of occurrence in the Chinese language [18]. Principal component analysis on the questionnaires extracted three varimax-rotated components which corresponded to the E, P, and A dimensions of Osgood et al. [8, 9]. Component scores of each word were computed, and words that scored highly positive (or negative) on one component (>0.70) but had only low scores on the other two dimensions were selected (i.e., stimuli were clearly associated with only one of the semantic dimensions). This resulted in six semantic word classes which were used subsequently. For electrophysiological experiments, list were prepared that contained a total of 1440 words stemming from the 6 classes in randomized order. Twenty-three different healthy volunteers (mean age 29.8; SD 6.0 years; 13 females, 10 males) with Chinese as native language were studied. All subjects were right-handed [5]. The EEG was recorded in 32 channels simultaneously from an array of electrodes located at the following Neuroscience Letters, 2003, 341, 45-48 3 W. Skrandies & M.J. Chiu, Dimensions of Affective Semantic Meaning of Chinese Language positions of the International 10/20 System: Fp1, Fp2, Fpz, F4, C4, P4, O2, F8, T8, TP8, F3, C3, P3, O1, F7, T7, TP7, Fz, Cz, CPz, Pz, POz, FC2, C2, CP2, FC6, CP6, FC1, C1, CP1, FC5, CP5 (see Figure 1). Linked ears served as recording reference; for analysis all signals were referred to the computed average reference. Data acquisition was preformed with a NK Neurofax EEG 1000 system (bandpass: 0.1 – 60 Hz; sampling rate: 200 Hz). For computing ERPs, the class of each stimulus presented was coded by a trigger signal. Stimuli were presented according to the randomized lists described above sequentially on a computer monitor for 1 s each. Viewing distance was 1 m, all stimuli were of high contrast (> 95%, 43 lx mean luminosity) and subtended 1.43° x 2.86°. Size was similar for all word classes. The subjects' task was to visualize the words, and recall was tested at the end of the experiment by a questionnaire. The performance on this task was routinely checked in order to verify that the subjects had actively observed and processed the stimuli. In addition, all subjects filled in semantic differential questionnaires in order to assure reliability of the stimulus material. The EEG was written continuously to harddisk, and was visually inspected offline. Eye movement and blink artifacts were detected at the three most frontal electrodes (located at Fp1, Fp2, and Fpz) that were not included in the analyses, leaving a mean of 141 segments per stimulus class. Artifact-free EEG following stimulus onset was averaged over 1000 ms according to stimulus class, and was digitally filtered (1 – 30 Hz). This resulted in six sets of average evoked potentials for each subject. Evoked components were identified topographically at peak times of the Global Field Power curve (GFP, [7, 12]), and we will report on components that occurred early after stimulus onset. At individual component latency, the topography of the field distributions was determined for each subject and experimental condition. Component latencies, field strength, and the topographic features of the evoked distributions were compared between stimulus conditions. The significance of the influence of the factors word class (E, P or A) and polarity (+ or –) was determined by ANOVAs with repeated measurements. Interactions between word class and polarity will be referred to as word dimension. Varimax-rotated principal components computed on the questionnaire data revealed three components which accounted for about 60% of the variance (64.1% in the questionnaire group; 62.5% in the EEG group). These components could be clearly identified as Osgood’s factors Evaluation (E), Potency (P), and Activity (A). Factorial similarity exceeded 0.82 for each factor indicating a very high reliability of the dimensions revealed by the semantic differential. Neuroscience Letters, 2003, 341, 45-48 2 W. Skrandies & M.J. Chiu, Dimensions of Affective Semantic Meaning of Chinese Language N A sequence of potential fields evoked by words between 70 and 895 ms after the stimulus is illustrated in Figure 1. For i c s r d T f i o w t F A i f We analyzed how latency, field strength, and scalp topography (positive or negative centroid locations in the anterior – posterior or in the left – right direction) of the components were influenced by the semantic dimensions. In the following we will concentrate on effects at small latency that suggest top-down influences on the processing of incoming meaningful information. We note that the potential distributions were similar in all conditions, and their features were modulated by semantic meaning. The earliest component which was reminiscent of the classical P100 appeared between 80 and 130 ms, and its topographical Figure 1 Potential fields elicited by Chinese words between 70 and 895 ms after stimulus onset. Maps are shown every 25 ms. Recordings from 29 electrodes as indicated by head scheme. Lines are in steps of 0.6 μV, dark areas are negative with respect to the average reference. Mean data of 23 subjects. euroscience Letters, 2003, 341, 45-48 llustration we display the grand means omputed over all word classes. There occur everal components starting around 95 ms as eflected by different topographical istribution patterns that occur sequentially. he strength and topography of the evoked ields changes over time. Steps of nformation processing are reflected by the ccurrence of components that are associated ith high field strength and stable opography [7, 12]. This is evident in igure 1 around 95, 170, or 270 ms latency. ll subjects’ individual components were dentified at the peak latency of maximal ield strength. Figure 2 Effects of semantic meaning on a component occurring between 80 and 130 ms latency. A Scalp topography as seen from front, back, right and left side of the head. Lines are in steps of 0.6 μV, dark areas are negative with respect to the average reference. B Significant effects of word polarity on component latency. 3 distribution is displayed in Figure 2. Word polarity had a significant influence on component latency (mean latency for positive W. Skrandies & M.J. Chiu, Dimensions of Affective Semantic Meaning of Chinese Language words: 108.2 ms; negative words: 105.0 ms; F(1,22) = 4.31, p<0.05) indicating that the characteristics “good”, “strong”, and “active” are followed by larger latencies than “bad”, “weak”, and “passive” (see Figure 2). The strength of the same component (reflected by GFP) showed a tendency to be influenced by word dimension (F(2,44) = 2.81; p<0.070).