Affective processing has been investigated on the neurophysiological level as well. Since this is important for the studies outlined in section 7 and 9, previous findings to this topic are introduced in more detail.
Differences in the amplitudes of particular voltage peaks of event-related potentials have been initially investigated in a “viewing affective pictures” design. This valid laboratory paradigm offers a systematic variation of affective stimulation (Cuthbert et al., 2000; Keil et al., 2001). Cuthbert, Schupp, Bradley, Birbaumer, and Lang (2000) for instance conducted a typical „viewing affective pictures“ design:
Emotionally arousing and neutral pictures of the International Affective Picture System (IAPS, 1999) were presented to healthy participants for 6 seconds each while event-related potentials were observed.
Introduction to emotional brain research with affective pictures
The following typical shape of event-related potentials was observed in the
“viewing affective pictures” design (Cuthbert et al., 2000; Junghofer, Keil et al., 2003;
Keil et al., 2001): A prominent, early negative peak emerged at approximately 100 ms shifting into positive waves between 200 and 300 ms as well as between 300 and 400 ms. In frontocentral areas, a prominent negative peak was further observed between 400 and 700 ms. After these peaks, a positive slow potential was measured, lasting up to several seconds. With respect to the different emotional stimulation, midline differences between the various emotional picture categories started between 200 and 300 ms:
Significantly greater positivity was observed during presentation of pleasant pictures.
The same effect is true for the P3 region. Even the frontocentral negativity between 400 and 700 ms is more prominent in pleasant pictures. In contrast, the following positive shift is more expressed for both pleasant and unpleasant pictures compared to neutral ones.
To summarize emotionality effects, slow positive voltage changes are significantly larger for affective than neutral stimuli. The positive voltage shift begins generally around P3 with a maximum at 1 second. The P3 amplitude has been described as a marker for attention, stimulus probability and task relevance (Donchin & Coles, 1988). Keil et al. (2002; 2001) further found stronger positive activation for affective compared to neutral pictures in posterior regions for the P1, independent of the content.
In contrast, N1, early P3, late P3, and slow waves had significantly higher amplitudes for pleasant (N1) as well as pleasant and unpleasant (P3, slow wave window) pictures, particularly in posterior regions. The effects were expressed more strongly in the right hemisphere. Cuthbert et al. (2000) conclude that a) differential brain responses to emotional stimulation were found even earlier than 200 ms and b) the positive voltage shift represents a selective processing of emotional stimuli. The advantage of emotional stimuli is assumed to be regulated by the above mentioned elementary motivational systems, the appetitive and defensive system. Observing ERP components while participants are confronted with emotional material therefore seems to be appropriate to investigate evoked emotional states. The correlation of the positive shift in affective materials with the subjects’ arousal ratings encourages this conclusion.
In addition to these typical “affective picture viewing” studies, early posterior negativity as well as and augmented positive voltage shift were also found in response to pictures of emotional (threatening) faces (Schupp, Ohman et al., 2004). The results
suggest a neurophysiological index for a facilitated processing of threatening stimuli.
The conclusion is consistent with Lang’s biphasic emotion theory (P. J. Lang, 1979; P.
J. Lang et al., 1993) which includes a natural selective attention to evolutionary significant stimuli in the environment.
To investigate the speed the human brain requires to detect the emotional impact of a stimulus, Junghöfer, Bradley, Elbert, and Lang (2001) applied a particular paradigm: The authors studied differential brain responses to varying affective stimuli in rapid serial visual presentation. Junghöfer et al. (2001) showed a stream of 700 IAPS pictures (1999) with 3 and 5 Hz varying in affective arousal to naive students. Rapid selective discrimination between emotional arousing and less arousing material has been shown: The ERP waveforms in response to high arousal pictures became more negative than the low arousal ERPs (beginning around 150 ms). The effect was localized in the occipito-parietal area. In summary, the visual brain was found to rapidly distinguish between emotionally arousing and neutral stimuli (Junghofer et al., 2001).
Comparable results were even found for presentation times of 120 ms (Schupp, Junghofer et al., 2004). Both results thus support the hypothesis of a very short-term conceptional memory store (Potter, 1993). Potter (1993) claims that there is a very rapid meaningful representation of a stimulus which might not even be consolidated but which allows the detection of a particular content or emotional valence, for instance.
Junghofer et al. (2001) point out that event-related potentials may add an answer to the question if a very early conceptual analysis is able to discriminate different emotional contents. This emotional discrimination seems to be mediated by primary and secondary visual processing areas of the brain.
The below described studies focus on emotional processing of affective words.
Hence, the following paragraphs emphasize previous evidence upon processing of (affective) words.
Emotional brain research with affective words
The first attempts to empirically demonstrate the affective connotation of words went in line with the above mentioned shift from the categorical to dimensional emotion theories (see section 4.1).
Early discriminative responses for emotionally high versus low arousing verbal stimuli have been found that parallel findings with affective pictures. Discrimination responses were even found when presented subliminally or at perceptual threshold and
also in clinical populations (Kissler et al., 2006). Nevertheless, the results on the different response variables are somewhat more complicated (Kissler et al., 2006).
A possible explanation that accounts for more complex results with affective words should be considered prior to a description of concrete experimental findings.
Emotions are seen as old and universal activation systems, controlled by subcortical brain structures. Their evolutionary aim is assumed as managing critical situations in which quick action decisions are required. Reading and writing, in contrast, are related to more recent developments, so that processing of emotional words may include very different brain areas. These different backgrounds and assumed neural circuitries may account for the complexity in involved processes.
Nevertheless, “linguistic expressions are stored within semantic networks that encompass links to all aspects of their linguistic and pragmatic usage and emotional connotations” (Kissler et al., 2006, p. 5). That is, a general dynamic network is assumed to exist with all aspects of a single representation included, organized in various sub-networks. The above described associative fear structure represents such general dynamic network. According to Lang and colleagues (1993), an associative (fear) structure can also include physiological and motor response information which have been co-activated once or several times with the relevant emotional representation.
Consequently, the investigation of emotional word processing seems reasonable for an empirical demonstration of an associative fear structure.
However, research on how emotional contents influence different stages of word processing is relatively rare. The following paragraphs briefly describe empirical findings concerning this topic.
Affective words among healthy subjects: A traditional perspective versus experimental evidence
Traditionally, perceptual but no meaning-related features are expected in ERPs until 150 - 200 ms after word stimulus onset (Posner, Abdullaev, McCandliss, &
Sereno, 1999). Semantic integration is assumed for the N400 (Kutas & Federmeier, 2000).
This traditional view was “attacked” by some early studies. The first studies in the field were run in the 1960ths. After Lifshitz’ (1966) early work, Begleiter and Platz (1969) reported differences between emotional (“taboo words”), neutral words, and blank flashes. In contrast to the traditional statement, these differential responses were
observed within the first 200 ms in a group of 20 healthy male subjects. The participants performed a passive viewing task and a naming condition (presentation times minimally above perceptual threshold). Amplitudes on one measured electrode namely the right occipital (O2 according to 10-20 system) were significantly larger for the taboo words than for neutral words or blank flashes. Effects were more pronounced in the naming condition.
Another important contribution to this field was given by the Chapman work group (Chapman, McCrary, Chapman, & Bragdon, 1978). The authors presented words, which scored either relatively high or low on one of the Osgood dimensions. The healthy subjects were instructed to pronounce the briefly (510 ms) presented words.
Grand average ERPs measured over the centro-parietal area differed systematically between stimuli loading high and low on one of the different dimensions.
Vanderploeg, Brown, and Marsh (1987) extended the previous findings to a comparison between ERPs to emotional words and emotional faces. The authors investigated healthy male subjects in a two-phase design: First, emotional words and faces were presented and rated by the participants with respect to emotional content while ERPs were recorded (6 electrodes). The presentation of faces was paired with auditory presentations of the words in a second conditioning phase. Afterwards, the complete stimulus set was rated again under parallel ERP recording. Word presentation results revealed only slight emotion-specific differences in late components. Amplitudes for emotional words were slightly (but not significantly) larger. In combination with strong emotionality effects for face presentations, the authors concluded that words are less powerful stimuli than faces or pictures. In addition, a larger intra-class variance within the word category compared to the face category might have reduced the words’
effects.
A few years later, Naumann et al. (1992) addressed affective word processing and implemented a structural decision versus an affective processing task. The design resulted in a long lasting positive shift for an active, conscious evaluation of the adjectives’ affective meaning. In addition, an enhanced P3 component was found for emotional compared to neutral stimuli. Both effects were evidenced in a pilot study and a replication experiment.
Twenty years later than Chapman and colleagues, Skrandies (1998) extended their approach: Skrandies (1998) presented extreme nouns on the Osgood dimensions as evaluated with the semantic differential. The stimuli were presented in a rapid serial
visual presentation (RSVP) design: The different stimuli were shown subsequently in a stream for one second each. ERPs were recorded parallelly (30 channels). Subjects were instructed to memorize the words for a later test. The results further contrasted the traditional statement of semantic processing starting around the N400. Differences in response strength and latency occurred between 80 and 265 ms after stimulus onset.
Additional and more recent evidence endorses the “attack” on the traditional view.
Schapkin, Gusev, and Kuhl (2000) report differential cortical responses to emotional and neutral words that occurred earlier than 400 ms. Their healthy participants categorized the unilaterally presented stimuli according to their emotional connotation.
The earliest differences between emotional and neutral words were found around 230 ms after stimulus onset (amplitudes for pleasant > unpleasant and neutral at central sites). The effect was interpreted as one of general evaluation of emotional significance.
P3 was further enhanced for pleasant compared to neutral words. Slow positive wave results depended upon emotion at anterior and posterior sites.
Bernat, Bunce, and Shevrin (2001) ran another study showing early emotionality effects. The authors demonstrated that even stimuli presented below conscious perception elicit differential brain responses in a simple viewing condition. More positive amplitudes were found in response to unpleasant compared to pleasant adjectives (particularly P3 and late positivity). Stimulus presentation time varied:
Responses to both sub- and supra-threshold presented stimuli were more pronounced in the left hemisphere (P1 and N1). A bilateral activity in later components (P3 and LP) was only found for supraliminal presentation. This study provides another evidence that differences related to the affective meaning of stimuli emerge even within 100 ms.
These effects even occur when stimuli are processed unconsciously.
Ortigue, Michel, Murray, Mohr, Caronnel, and Landis (2004) conducted another study using a dense array ERP system. Their healthy male subjects incidentally processed (lexical decision with bilaterally presented word/non-word or non-word/non-word pairs) the shown non-word/non-words. This design aimed at clarifying the apparent paradox that linguistic stimuli elicit differential brain responses even in “prelexical” stages (before stimuli are recognized as words). Lexical decision performance was better for emotional words, particularly when presented in the right visual field. But the emotionality effect was more expressed for stimuli presented in the left field. ERPs between 100 – 140 ms after stimulus onset were related to the different experimental conditions and identified (linear source estimation, LAURA) in bilateral lateral-occipital sources with a
maximum in the right hemisphere. The results suggest a specialized brain network, which is activated very early by the emotional connotation of words.
Kissler and colleagues (2006) further confirmed early differences in ERPs evoked by emotionally arousing (pleasant and unpleasant) and neutral words. The authors presented adjectives and nouns previously rated for valence and arousal. Stimulus condition sets were matched for word length and frequency. Presentation time in a RSVP design varied between 333, 666, and 1000 ms. The subjects were instructed to simply read the words. Results revealed a left hemisphere dominance in processing linguistic stimuli: An occipito-temporal (left) negativity with a maximum at 260 ms after stimulus onset differed between emotionally arousing and neutral words. The different presentation times elicit only slight differences. The finding of larger N4 amplitudes for neutral compared to emotional words is explained by unequal stimulus probabilities and respective expectancy effects (one third = neutral; two thirds = emotional).
Herbert and colleagues (Herbert, Kissler, Junghofer, Peyk, & Rockstroh, 2006) presented similar stimulus material (only adjectives) for 4 s to healthy subjects. The P2 component differed once more for emotional compared to neutral words, but not for pleasant and unpleasant words. The effect was more pronounced the higher the perceived (previously rated) stimulus arousal. While effects for P3a were driven by the arousal but not the valence, this was not true for the late positivity: In contrast to Bernat and colleagues (2001), amplitudes were larger for pleasant than for unpleasant and neutral words and correlated with the auditory startle response.
Tradition versus evidence: What is concluded for the access to affective words?
In summary, the evidence contrasts the traditional view of no semantic context processing and thus no meaning related effects before the P3. Future research should thus include early time windows in the analysis of emotionality effects evoked by affective words. Nevertheless, the analysis of N400 effects is still important.
Conscious processing is supposed to start around P3/N400 and is influenced by semantic expectancy, task relevance and depth of mental engagement (Kissler et al., 2006). Detection and processing of emotional contents are considered to be part of semantic processing. The N400 is therefore regarded as an appropriate “classical” ERP component for emotion effects and electrophysiological indicator for semantic processing (Kissler et al., 2006). Kissler and colleagues (2006) add on the one hand,
that other effects of semantic processing have not been found because of an N400 analyzing bias. On the other hand, they consider N400 as indicator of semantic integration within a larger context. That is, N400 modulations will only occur if single word stimuli are actively connected to a broader emotional content or if the subjects are already in a certain mood when the stimulus is presented (Kissler et al., 2006).
Moderating factors in affective word processing
The debate about the time window for affective word processing will be briefly added by a consideration of moderating factors in affective word processing.
First, the relation of the valence and arousal dimension should be taken into account: At low stimulus arousal levels, a stronger response to pleasant stimuli has been found more often (positivity offset). In contrast, responses to unpleasant stimuli are facilitated at high stimulus arousal levels (negativity bias, Cacioppo, 2000). Words are generally less arousing than pictures. The interpretation of ERPs to visually presented word stimuli thus have to particularly consider the positivity offset.
Another point of interest addresses the experimental conditions that favor early emotionality effects. Emotionality effects in response to affective words are evidenced even in early (preconscious) stages. Some points remain unanswered though when and how cortical responses are exactly influenced by the emotional stimulus content, the role of neutral stimuli, and the role of the subjects’ task, for instance (Kissler et al., 2006). Early (P1, N1) indicators for affective word processing are found more often for very brief stimulus presentation times near or below perception threshold (Begleiter &
Platz, 1969; Bernat et al., 2001; Chapman et al., 1978; Flor et al., 1997; Knost, Flor, Braun, & Birbaumer, 1997; Ortigue et al., 2004; Pauli et al., 2005) as well as for repeated presentation of stimulus sets (Begleiter & Platz, 1969; Chapman et al., 1978;
Ortigue et al., 2004; Skrandies, 1998).
Mediating structures in affective word processing
This paragraph will briefly summarize the literature considering mediating structures in affective word processing. Animal experiments suggest different short cut routes in emotional processing with a central position of the amygdala. This is particularly true for fear-relevant or at least highly arousing stimuli (LeDoux, 1995).
The role of the amygdala within fear regulation was systematically investigated in classical conditioning paradigms and pharmacological studies. These studies aimed at
investigating antero- and retrograde transport in the respective axons. LeDoux and co-workers managed to identify afferent and efferent connections within a neural network that are involved in both emotional appraisal as well as regulation of the (fear) response.
The pivotal finding showed that the general neural pathway assumed for an ingoing acoustic startle stimulus consists of miscellaneous connections. It is generally assumed that the sensory information is transferred via the auditory pathway to the thalamic Corpus geniculatum mediale (CGM). The auditory pathway continues its way to the primary auditory cortex in the temporal lobe via the acoustic radiation. But lesion studies prove that additional connections exist: A conditioned fear reaction in response to an acoustic stimulus was still observed when the general pathway via the acoustic radiation was lesioned (animal experiment). In contrast, for earlier lesions e.g. in the CGM or even the Colliculi inferiores, the conditioned fear reaction disappeared (LeDoux, 1995). The evidence thus suggests further connections between thalamus and auditory cortex. Additional lesion studies suggest that this connection is mainly afforded by the amygdala. Its role is assumed in integrating sensory information from the thalamus and the affective appraisal (lateral nucleus). In addition, the amygdala is hypothesized to regulate the efferent (fear) response in the central nucleus (LeDoux, 1992). According to Turner and Herkenham (1991), respective findings have been reported for other modalities, too. The assumed short cut routes might account for the facilitation of emotionality effects in very brief stimulus presentation designs. Another problem emerges with respect to affective word paradigms: The amygdala model still implies some kind of reading ability on subcortical (thalamus) level.
Indeed, indicators of prelexical word processing were found at least starting around 150 ms after stimulus onset: These indicators consist e.g. in a left occipito-temporal negativity for written words (Dehaene, Le Clec, Poline, Le Bihan, & Cohen, 2002). Dehaene and colleagues (2002) investigated the hypothesis of a visual word form area (VWFA) in a fMRI study. The VWFA is assumed to be located in the left fusiform gyrus and to enable a prelexical representation of visually presented words. These assumptions are based on consistently resulting brain activity in that area when skilled readers respond to visually presented words. The hypotheses are further encouraged by deficits in word reading if the left posterior fusiform gyrus is lesioned. Dehaene and colleagues (2002) instructed healthy adults to perform a same-different judgment task with words from five different semantic categories (tools, body parts, action verbs, animals, numbers) and pseudo words. Indeed, greater activation for written than for
spoken stimuli was found in the entire occipito-ventral stream (from occipital extrastriate cortex to the anterior fusiform gyrus), left anterior cingulate, left and right posterior intraparietal regions, and the left anterior insula. Pseudo words evoked a strong response in the VWFA, too. Pseudo words thus imply an involvement in prelexical processing of that area, but do not differentiate for different semantic categories or semantic meaning at all.
Such occipito-temporal negativity empirically occurred e.g. in RSVP designs sensitive to semantic aspects of visual word processing. The negativity maximum
Such occipito-temporal negativity empirically occurred e.g. in RSVP designs sensitive to semantic aspects of visual word processing. The negativity maximum