Processing of emotional words measured simultaneously with steady-state visually evoked potentials and near-infrared diffusing-wave spectroscopy

Emotionally significant stimuli automatically attract attention and have been shown to be processed preferentially in comparison to neutral stimuli. This effect is apparent for both unpleasant stimuli, such as angry faces, snakes or humiliation scenes, and in pleasant material like erotica or baby faces, as investigated in behavioral experiments, in studies using EEG, and in fMRI [1–3]. Functional MRI experiments indicate increased BOLD responses during presentation of intrinsically unpleasant stimuli as well as aversively conditioned stimuli in different brain areas, including - perhaps surprisingly - even primary visual areas [4–6]. Using near-infrared spectroscopy (NIRS), two recent studies [7, 8] reported increased metabolic activity in the occipital lobes for emotional, and in particular pleasant, compared to neutral pictures.

EEG event-related potential (ERPs) recordings found an enhanced early posterior negativity (EPN) between 200 ms and 300 ms after stimulus onset and an increased late positive potential (LPPs) around 500 ms for emotional compared to neutral pictures from the International Affective Picture Set (IAPS, [9]) [10–12]. Further, eliciting steady-state visual evoked potentials with oscillatory flickering IAPS pictures also leads to enlarged signal amplitudes for emotional pictures, pointing at higher activity in primary visual areas (e.g., [13, 14]).

Enhanced sensory processing of emotional stimuli is presumed to be mediated by feedback projections from the basal nucleus of the amygdala to the striate cortex and extrastriate, secondary visual and associative processing areas [15]. The amygdala is an almond-shaped structure in the anterior medial temporal lobe strongly involved in emotion processing [16]. Through this so-called "reentrant processing" the amygdala may therefore enhance neural responses to emotionally relevant stimuli [15].

Even purely symbolic stimuli such as words seem to be subject to the effects of enhanced visual processing and attention allocation, as shown in studies investigating event-related EEG potentials (ERPs) to emotionally significant words (see [17] for a review). Even in very early time windows emotionally significant words affect visual processing, as indicated by enlarged P2 s at parietal electrodes (around 180-250 ms after stimulus onset) for pleasant [18] or both pleasant and unpleasant words [19]. Similarly, an enhanced early posterior negativity (EPN) has been found at around 250 ms for emotional in comparison to neutral words [20–23].

Regarding late (> 300 ms) ERP effects, several studies found enhanced LPPs for both pleasant and unpleasant compared to neutral words (e.g., [23–25]). However, in some studies only pleasant words induced an enhanced LPP [19, 20, 26], although early effects were found for both pleasant and unpleasant words. Due to specific elaborative associative processing, later effects may have been restricted to pleasant stimuli [19, 20]. Evidence for deeper associative processing particularly of words with pleasant emotional content also comes from two studies probing the startle eye blink response [19, 27].

Event-related functional magnetic resonance imaging also revealed increased BOLD responses to emotionally relevant words in limbic areas and visual cortex [28, 29]. A recent study of single word reading found the largest effects during reading of pleasant words, as reflected in activation increase in both occipital visual areas and the amygdala. The increase in activation of these areas were correlated, in line with the model of re-entrant processing [30].

While it is becoming clear that emotionally significant words benefit from transient amplification of visual processing, it is an open question whether this enhanced activation is also seen in prolonged and sustained electrophysiological responses of the visual cortex, such as steady-state visual evoked potentials (SSVEPs), for which emotion effects have previously been found using IAPS picture stimuli [13, 14]. The present study addresses this issue combining electrophysiological and non-invasive optical measures of brain activity in response to the presentation of flickering word stimuli.

Diffusing-wave spectroscopy (DWS; also called diffuse correlation spectroscopy, DCS) is an emerging optical technique for the non-invasive detection of functional brain activity. The contrast mechanism of DWS is based on the dynamics of scatterers within tissue - such as erythrocytes - that is enhanced e.g. by the increase of regional blood flow associated with the vasodilatation induced by electrical activity [31–35]. The first entirely non-invasive applications of DWS to brain function detection [32, 34] focused on the human motor cortex stimulated by finger tapping tasks, which elicit large functional signals. However, such tasks interfere with systemic perfusion changes accompanying functional cortical activation.

Recent technical advances, which have significantly improved the signal-to-noise ratio of the DWS experiment [36], have allowed for detection of functional DWS signals from much smaller, activated areas located more deeply in the cortex, such as the human primary visual cortex where functional activation can be elicited without changes of systemic perfusion. For steady-state visual stimulation with 30 s full-field flickering, functional DWS signal changes of 3.0-3.8% were found [33] which can be traced back to the functional increase of rCBF of 60-80% measured with PET [37].

In addition to these slow DWS signals, Li et al. [35] reported that short (8.2s) visual stimulation blocks elicit a transient slowing of cortical dynamics localized on the left hemisphere of the visual cortex which is not explained by the current models of neurovascular coupling [38]. While NIRS and DWS signals yield complementary information on neurovascular coupling, the larger functional signal obtainable with DWS makes this technique attractive for studies in neuroscience.

So far, DWS has not been applied to experiments examining very subtle stimulus differences such as the emotional connotation of words, but it could provide additional and complementary information on hemodynamic changes in response to emotional stimuli.

Combining DWS and EEG requires an experimental paradigm that provides sustained stimulation over several seconds and at the same time a number of trials sufficient for a satisfying signal-to-noise ratio of the EEG response. Steady-state visual evoked potentials (SSVEPs) offer both these features. They are oscillatory, nearly sinusoidal scalp potentials in response to a repeated visual stimulation presented at frequencies of 4-6 Hz or higher [39, 40]. Their main advantage is a very good signal-to-noise ratio for moderate stimulation times, and a sharp peak at the stimulation frequency in the power spectrum of the signal [39]. Further, SSVEPs provide a unique possibility to study a well-defined and controllable electrophysiological brain response to prolonged stimulation. Hence, SSVEPs are ideally suited for investigating simultaneous EEG and metabolism-based techniques like optical measurements of brain activity. Larger SSVEP amplitudes were found during viewing of emotional in comparison to neutral pictures [13, 14], but so far no study has investigated SSVEP modulation during processing of emotional words. As detailed above, the neural dynamics underlying processing of emotional words are in many ways similar to the processing of emotional pictures and lead to increased attention and resource allocation. However, the emotional significance of words lies in their symbolic nature, which is not primarily conveyed by external stimulus characteristics, but by the activated associations. Therefore, during prolonged and sustained stimulation, emotion modulation for words may differ from those of pictures.

Here, we investigate SSVEPs and DWS signals evoked by pleasant, unpleasant, and neutral words. In line with previous literature [17, 20, 24, 28], we reasoned that emotional (pleasant and unpleasant) words would attract more attention than neutral words. Most SSVEP studies show higher SSVEPs for attended than for unattended stimuli [41, 42]. In parallel, extant SSVEP studies on affective picture processing [13, 14] have reported larger responses to emotional than to neutral pictures. However, two recent studies suggest that, under certain circumstances, SSVEP power can be reduced rather than increased by attention, e.g. when narrowing the spatial focus of endogenous attention [43] and depending on the oscillating brain network [44]. So far, no study examined SSVEP modulation by emotional words, precluding specific directional predictions. Based on previous research showing differential processing of emotional and neutral stimuli, we hypothesized that SSVEP power as well as the DWS signal would be different for emotional compared to neutral words.

By simultaneously measuring DWS and EEG, the present study aimed to investigate visual cortex activity in response to the presentation of flickering emotional and neutral words. Both methods revealed effects of stimulation, but an emotional modulation was only observed in the SSVEP, not in the DWS signal. DWS showed a statistically significant difference in mean decay time between stimulation and baseline periods for both long- and short-distance receivers R₁ and R₂. While the functional signal from the long-distance receiver replicates the previously reported transient slowing of cortical dynamics [35] and extends it to written word stimuli, we find a functional signal also in the data from the short-distance receiver which might arise from residual sensitivity of the short-distance sensor to cortical dynamics, or from increased blood flow in the scalp due to changes in peripheral perfusion induced by the stimulation. No differences in the DWS signals could be detected when comparing the three different emotional word conditions. Given the relatively subtle psychological difference between emotionally positive, negative and neutral words and the modest sample size of ten participants, DWS as implemented in this study may at present not be sensitive enough to detect these small effects. Theoretically, since the present DWS instrumentation allowed recording only from a part of the visual cortex, existing activation differences suggested by some previous fMRI studies [28–30] may show a regional distribution different from the one captured by the present design. The DWS sensors were placed 15 mm above Oz, thus one might argue that they were not measuring activity of the same cortical areas as reflected by the SSVEPs. However, given the limited conductivity of the skull leading to spatial low-pass filtering of the EEG signal, the SSVEPs recorded at the occipital electrodes reflects activation of a larger brain area, probably including the volumes probed by the optical experiment.

Two previous studies using a related optical technique, NIRS, found emotion effects, particularly for pleasant stimuli from the IAPS picture set [7, 8]. However, these effects were less clear-cut than corresponding EEG effects. This pattern could result from the lower sensitivity of metabolic contrasts to transient events. Moreover, emotional modulation of visual processing is generally considerably larger for colored pictures than for perceptually much simpler written words [22, 49, 50]. Thus, future DWS studies of affective processing using pictures from the IAPS set which are perceptually more engaging may well reveal significant effects. Conversely, since no previous NIRS study has investigated signal modulations by emotional words, it is unclear whether a corresponding experiment using conventional NIRS measurement would yield significant results. In the present data, no effect was obtained regarding the photon count rate which is the signal most closely reflecting NIRS. The transmitted light intensity measured by NIRS is proportional to the photon count rate measured in our DWS experiments. The present study found no statistically significant differences in photon count rate due to emotion categories. This could be due to the fact that the DWS photon count rate was measured from the speckle signal, which intrinsically has large fluctuations, and therefore leads to considerably lower signal-to-noise ratio than typical NIRS signals. An additional explanation for the absence of statistically significant functional NIRS signals in our data may be that the photon count rate measured at the laser wavelength of 802 nm is sensitive to changes of total hemoglobin concentration; these are smaller than the functional concentration changes of oxy-hemoglobin that are measured with standard NIRS instruments using two wavelengths located above and below the isosbestic point of hemoglobin. Given these important limitations, the question whether emotional content-dependent differences during word processing can be obtained using blood flow-based measures other than fMRI thus begs further research.

Regarding the SSVEPs, positive words elicited significantly smaller electrophysiological responses than neutral words at occipital electrode positions. The finding of larger differences between positive and neutral than negative and neutral words in emotion word processing is in line with some other previous reports [20, 23, 26, 30, 51], but the direction of our finding diverges from results of enhanced SSVEP power for emotional in comparison to neutral IAPS pictures [13, 14]. However, to our knowledge, the present study is the first to use a typical SSVEP paradigm to investigate the visual processing of emotional words, which may differ from the processing of emotional pictures.

The higher SSVEP amplitudes for neutral words cannot be due to confounding factors such as word length, frequency, or concreteness, as these factors were matched between conditions. Further, the ratings of valence and arousal of the word stimuli by the participants also confirmed the data previously obtained by Kissler et al. [21], ruling out the possibility that our sample of subjects was different with regard to valence or arousal judgments.

The present results may be accounted for by other factors: First, most previous studies investigating emotion effects on SSVEP amplitudes used a stimulation frequency between 10 and 13 Hz (e.g., [13, 52, 53]), which might be functionally different from lower frequencies. Investigating different driving frequencies, some authors found enhanced SSVEP amplitudes to attended stimuli for both 12 and 8.6 Hz [41], but the interplay between stimulation frequency and attentional modulation of SSVEPs is not entirely understood yet: In a first systematic investigation of the effects of spatial attention and driving frequency on SSVEPs, Ding and colleagues [44] found higher amplitudes for non-attended stimuli at a stimulation frequency of 9.2 Hz, when a competing attended stimulus was presented to the fovea, but smaller amplitudes when the competing attended stimulus was located in the periphery. Similarly, Chen et al. [43], using a driving frequency of 8.3 Hz and measuring MEG, found a reduction of steady-state visual evoked field amplitude to an attended stimulus when attention was directed to the center of the visual field.

However, it is important to note that these studies did not investigate emotional or exogenous attention but rather endogenous spatial attention. These two forms of attention have partly different underlying mechanisms and effects on visual processing. Whereas the deliberate (endogenous) allocation of attention to certain spatial positions requires top-down control of visual areas by fronto-parietal attention networks, enhanced attention to motivationally salient, emotional stimuli is considered to be automatic and bottom-up, involving the amygdala and its reentrant connections to modulate visual cortex activation [3, 54].

Therefore, our results might be due to the specific characteristics of word processing. SSVEPs are responses to stimulation of several seconds and they do not necessarily show the same effects as ERPs, which reflect the immediate first reaction to a stimulus. Regarding emotional picture stimuli, enhanced amplitudes have been observed in ERPs and in SSVEP [10–13, 55]. However, words differ from pictures in at least two very important aspects. First, they constitute perceptually relatively simple stimuli, in our case white letters on black ground. On a purely perceptual level, words are far easier to process than colorful photographs of complex scenes, for which visual analysis of all details and relevant aspects should take longer. Letter streams per se are very boring stimuli, as everyone will confirm who has been confronted with a text in an unfamiliar language. The import of a familiar word depends on its meaning, on the associated concepts and associated sensory representations it evokes. This mainly symbolic nature of words is the second and most important aspect distinguishing words from pictures, giving rise to the possibility of different reaction patterns to emotional words than to other, non-symbolic, emotional stimuli. While the initial response to reading an emotional word and seeing an emotional picture as reflected by ERP modulations seems very similar, subsequent sustained activity may diverge. Both emotional pictures and words activate subcortical emotion circuits [56–58]. However, as the emotional response is not caused by the letters on the screen themselves, but their semantic content, it is possible that subsequent stronger processing of emotional relevance takes places at a higher, associative level, which may result in a shift of attention to internal processes rather than external stimulation. This activity in associative cortices is probably not oscillating at the stimulation frequency, but may lead to a decrease of the SSVEP signal due to decreased primary visual processing. Indeed, electrocortical activity enhancements in response to emotional words have been most consistently found at processing stages that reflect activity outside the primary visual cortex [20, 21]. Moreover, in line with the possibility of reductions in purely perceptual processing, Herbert et al. [19, 27] showed that specifically pleasant adjectives are processed more deeply than neutral ones and that processing of pleasant adjectives leads to a stronger startle reaction in comparison to neutral words, reflecting a more internal focus of attention and higher costs of reorienting attention to external startle cues. From subjective experience, readers know the absorbing effects of exciting novels, leading almost to forgetting of external time and space, while creating an internal world of its own. Therefore, deeper internal processing of pleasant words may be associated with decreased activity in primary visual cortex as reflected by the observed reduced SSVEP amplitudes.

In summary, the present study demonstrates stimulation-driven modulation of both the DWS signal and the SSVEP. The SSVEP, but not the DWS signal, revealed differences due to the emotional content of the words used for stimulation. By showing a relatively smaller SSVEP response to stimulation with pleasant than with neutral words the direction of the emotion effect diverges from previous findings of studies using IAPS pictures. We suggest that this effect reflects specific processing characteristics induced by emotional words.

This is the first study to concurrently investigate the modulation of diffusing-wave spectroscopy, an optical measure of brain function, and steady state visually evoked potentials (SSVEPs), an electrophysiological measure of brain functioning, by stimulation with trains of flickering positive, negative, and neutral words. Both measures responded to the stimulation as reflected in increased activity as compared to baseline. However, only SSVEPs captured differences elicited by the stimulation's emotional content. Emotional modulation was reflected in a reduction in SSVEP power in response to pleasant as compared to neutral words. While a more pronounced modulation of cerebral activity during processing of pleasant words is in line with previous literature, the reduction of SSVEP power is a new finding, probably due to the symbolic nature of emotional language stimuli, which, similar to emotional pictures, elicits enhanced processing at temporally early stages, but then leads to increased internal processing at the cost of primary sensory activity during prolonged stimulation.