15.1 EEG Correlates of Preferences for Flickering Lights

effective durations of alpha rhythms, measured at occipital electrodes O1 and O2, were longer for the more preferred visual stimuli. In the preferred conditions, alpha rhythms persist longer, with higher temporal coherence, and extend over wider spatial regions of the cerebral cortex.

As reported by Lindsay (1952), who considered the relation of brain activity and behavioral states, the presence of alpha rhythms in the EEG of a human subject corresponds well to mental states associated with relaxation and free creative thought. The term alpha refers to the frequency band between about 8 and 13 Hz. The differentiation of basic emotions (i.e., intention, anxiety, aggression, sadness, and joy) by means of EEG-coherences has been discussed extensively (Hinrichs and Machleidt, 1992). Intention, aggression, and joy are mainly characterized by an increase in alpha-coherence, whereas a decrease is seen for anxiety and sorrow. In Chapter 4, we have discussed the method for using the ACF to analyze brain waves to examine the relationship between brain activities and the scale value of subjective preference as an overall impression of the sound field. We analyzed the effective duration of the normalized ACF envelope (τe) of the alpha waves when temporal factors such as the initial time delay gap between the direct sound and the first reflection ( t1) and the subsequent reverberation time (Tsub) were varied. Results showed that the τe of the alpha waves is longer only in the left cerebral hemisphere for the preferred conditions of these temporal factors. The relationship between subjective preference and the ACF of the alpha waves in response to the tempo of a noise burst, for example, has been investigated (Chen et al., 1997). Results showed that the τe of the alpha waves is longer only in the left cerebral hemisphere for the preferred tempo of a noise burst. Petsche (1996) analyzed EEG changes caused by mental processes of a higher order by using coherence analysis. Acts of creative thinking, whether verbal, visual, or musical, were characterized by a more increased coherence between occipital and frontopolar electrode sites than were other mental tasks. Results were interpreted as showing a stronger involvement of the long cortico-cortical fiber systems in creative tasks.

In this section we examine whether the scale value of subjective preference of visual stimuli reflects the temporal information in EEG in the left or right cerebral hemisphere (Soeta et al., 2002a). First, the PCT was performed for flickering light sources of varied period and mean luminance. From results of scaling the value of subjective preference, the most preferred and relatively less preferred light sources were selected as paired stimuli for brain wave recordings. Then, relationships between the scale value of subjective preference and the factors extracted from the ACF in the alpha waves were examined.

The light source was a 7-mm-diameter green light-emitting diode (LED), and was viewed by the subject at the distance of 0.6m from it in dark surroundings. The LED stimulus field was spatially uniform, and its size corresponded with 0.67◦ of the visual angle. The luminance of the stimulus is given by

where L0 is the mean luminance, m is modulation (relative) amplitude fixed at 1.0, and f is the temporal frequency of the stimulus. The period T = 1/f was set at 0.4, 0.8,

1.6, or 2.4 s, and mean luminance was set at 7.5, 30, and 120 cd/m2. The duration of the stimuli was fixed at 5 s. Ten 23to 25-year-old subjects participated in the experiment. All had normal or corrected-to-normal vision. They adapted to the dark and looked at the LED stimulus seated in a dark room with a comfortable thermal environment. The PCT was conducted for each subject by having each subject compare 66 pairs per session and having 10 sessions. The subject was asked which stimulus they preferred to watch. The scale value of subjective preference of each subject, which is regarded as the linear psychological distance between light sources, was obtained (Ando and Singh, 1996; Ando, 1998).

The average scale values of preference obtained from the 10 subjects are shown in Fig. 15.1. The most preferred period, [T]p, for all subjects were estimated by fitting a suitable polynomial curve to a graph on which the scale values were plotted. The value of [T]p for all of the subjects was 1.27 s at a mean luminance of 7.5 cd/m2, 1.49 s at a mean luminance of 30 cd/m2, and 1.76 s at a mean luminance of 120 cd/m2. The most preferred period, i.e. peak of the scale value, shifted gradually toward longer periods as the mean luminance was increased (Fig. 15.2). This may imply that the most preferred condition corresponds to a constant total amount of excitation from the physical environment. Too much or too little excitation may be less preferred.

Fig. 15.1 Averaged scale values of subjective preference as a function of the flicker period. Different symbols indicate different mean luminance. ,

7.5 cd/m2; ◦ , 30 cd/m2; , 120 cd/m2. Solid curve is expressed by Equation (15.2)

Similar to the above, the preference evaluation curve may commonly be

where α and β are the weighting coefficient and a constant, respectively, and x = log10T – log10[T]p. Values of α and β were obtained by using the quasi-Newton numerical method, respectively, and were approximately 4.90 and 1.56, respectively. It is interesting that the value of β ≈ 3/2 was consistent with other preference judgments performed including those for the sound and visual fields. When the hori-

Fig. 15.2 The most preferred flicker periods [T]p obtained by 10 subjects as a function of the mean luminance

zontal axis of Fig. 15.1 is normalized by the most preferred period [T]p, then all results may be reduced by a single curve as shown in Fig. 15.2. Without loss of any generality, the scale value can be adjusted to zero at the preferred condition, so that scale values for different [T]p values of mean luminance had similar tendencies as shown in Fig. 15.3. Therefore, the preference evaluation curve can be calculated by Equation (15.2) with β = 3/2.

Effects of the period and mean luminance on the scale value of preference were examined for all 10 subjects using the two-way analysis of variance (ANOVA) method. The results clearly indicated that effects of the period were significant

Fig. 15.3 Scale values of preference as a function of the flicker period normalized by the most preferred periods [T]p. Different symbols indicate different mean luminance. , 7.5 cd/m2; ◦ , 30 cd/m2; , 120 cd/m2. Solid curve is expressed by Equation (15.2)

(p < 0.01). The period and mean luminance were independent influences on the subjective preference judgment.

Next, the same subjects that were used in the preference tests participated in EEG recordings. The EEG was recorded under three conditions: (1) period varied and mean luminance fixed; (2) period fixed and mean luminance varied; (3) both period and mean luminance varied. To find a significant effect of subjective preference on an EEG, the most preferred flickering light and the relatively less preferred flickering light were selected as paired stimuli. The paired stimuli were set for each subject according to their individual preferences. The subject watched the most and the least preferred flickering lights alternatively. A series of EEG was recorded 3 times for each subject, and each series consisted of 10 stimuli pairs.

The EEG was recorded from the left and right cerebral hemispheres of subjects using silver electrodes (7 mm diameter) at scalp locations T3, T4, T5, T6, O1, and O2, as shown in Fig. 15.4 (10–20 International Electrode Placement System). A reference electrode was attached to the earlobe of a subject. A ground electrode was placed on the forehead of a subject. The recorded data were filtered with a digital band-pass filter with cutoff frequencies of 8 and 13 Hz (alpha-wave range).

Fig. 15.4 Top view of a subject’s head and electrode positions on the scalp for EEG recordings (10–20 International Electrode Placement System)

An example of a measured ACF is shown in Fig. 15.5a. The ACF may be characterized by four variables (see Sections 2.2 and 5.2). Figure 15.5b shows the absolute value in the logarithmic form as a function of the delay time. To find the degree of ACF envelope decay, the effective duration, τe, is determined. As shown in Fig. 15.5b, the straight-line regression of the ACF can be drawn by using only the initial declining portion, 0 dB > 10 log |φ(τ)| > –5 dB. In most case, the envelope decay of the initial part of the ACF may fit a straight line. The value of τ1

Fig. 15.5 (a) An example of a normalized autocorrelation function ACF of an EEG alpha-band signal (8–13 Hz) showing definitions of the delay time of the first peak τ 1 and its amplitude φ1. (b) Determination of the effective duration (τ e) of the alpha rhythm by estimating the slope of the envelope of the autocorrelation function and determining the delay at which it reaches 10% of its maximal, zero-lag value. Effective duration measures duration of temporal coherence, i.e., the duration for which repetitive structure persists in a signal

corresponded mostly with the center frequency in the range 8–13 Hz and thus was not analyzed.

Referring to the results in Section 5.3, the integration interval 2T was selected as 2.5 s in the running ACF analysis to obtain the values of τe, (0), and φ1. Table 15.2 shows results of the one-way ANOVA for τe, (0), and φ1 values of the alpha wave for the 10 subjects. Significant effects were found when the period was varied and the mean luminance was fixed and when both period and luminance were varied. However, significant effects were not found when the period was fixed and the luminance was varied except for (0) at O1.

Only when the period was varied were the values of τe, (0), and φ1 for the most preferred stimuli larger than those for the less preferred stimuli for all subjects, as shown in Figs. 15.6, 15.7, and 15.8. The tendency was especially clear in the

Fig. 15.6 Effective durations τe of EEG alpha rhythms at different electrode positions in response to a change in flicker period using preferred and less preferred rates. Error bars represent 95% confidence interval. ◦ , higher preference; •, lower preference

Fig. 15.7 Magnitudes (0) of EEG alpha rhythms at different electrode positions in response to a change in flicker period using preferred and less preferred rates. Error bars represent 95% confidence interval. ◦ , higher preference; •, lower preference

Fig. 15.8 Amplitudes φ1 of EEG alpha rhythms at different electrode positions in response to a change in flicker period using preferred and less preferred rates. Error bars represent 95% confidence interval. ◦ , higher preference; •, lower preference