The high variance in the subjective evaluations enables stimulus selection as intended. The curvilinear relationship of the valence and arousal dimensions parallels previous evidence (P. J. Lang et al., 1997). Nouns rated as pleasant or unpleasant are evaluated as high arousing stimuli whereas neutral nouns elicit less arousal. Thus, both expectations are confirmed and stimulus selection for the studies presented below bases on these self-report measures.
In comparison to the original ratings of German native speaker students (Herbert et al., 2006), some differences in the affective categorization emerged. This might be due to the age difference on the one hand. On the other hand, cultural influences might have biased the subjective evaluations as the current sample consisted of refugee
adolescents. Stimulus categorization on the basis of this prestudy seems thus reasonable for the subsequent studies among refugee adolescents.
The clear nonlinear relationship between valence and arousal ratings replicates the reliable findings of Peter Lang’s group (e.g., P. J. Lang, 1995; P. J. Lang et al., 1997, 1998; P. J. Lang et al., 1993). The evaluation of affective stimuli as more arousing than neutral stimuli thus seems to be independent of age or ethnic background.
The amount of situational descriptions, cognitions, and emotions within the free associations further encourages the idea of personal experiences represented in associative networks as described above. The idea to measure semantic relatedness and subjective associations with free associations is in line with Nelson and colleagues (2005).
7 Altered affective word processing in PTSD adolescents:
an MEG study
7.1 Introduction
Evidence demonstrating neurophysiological indicators of fear structures in humans is relatively rare. The present study assumes that the mental representation of a traumatic experience moderates manifestation and chronification of PTSD symptoms.
Thus, the current study addresses an experimental paradigm that allows a conclusion to such representations. Since visual presentation of affective stimuli is well documented to evoke a number of cognitive, physiological, and behavioral responses, such paradigm is an appropriate method to provide insights into pathologically altered processing of affective stimuli (Elbert, 1998). Mental representations of life-threatening events are assumed to be organized in associative networks such as the fear structure described by Lang (1979). The pathological fear structure is hypothesized as extremely large and strongly interconnected (Ehlers & Clark, 2000; Foa & Kozak, 1986) comprising elements of the traumatic experience in all modalities. That refers to sensory, cognitive, physiological, and behavioral information. In addition to these stimulus components inherent in the original traumatic situation, the associative structure contains the individuals’ own response components, e.g. fear responses.
Based on the Hebbian thesis of strengthened associations between cell assemblies due to synchronous activation (1949), interconnections within the associative structure are strengthened with every recall. The structure additionally includes new elements from the recall environments (Borgelt, 2006; Elbert et al., 2006). Thus, the probability of future activation through triggering by one of its elements and subsequent activation of the whole structure is enhanced by every recall (Elbert et al., 2006). Since response components are part of the structure, too, fear responses might be reactivated as well.
The adequate processing of the trauma is thus hindered with every recall – a process that is supported by the PTSD avoidance symptoms (Foa et al., 1995). This vicious circle is assumed to play a crucial role in the chronification of PTSD symptoms and the clinically observed facilitation of intrusion symptoms.
The present study aims at demonstrating fear structure activation in an experimental context. The experiment is based on the previous ERP evidence regarding the temporal course of emotional stimuli processing. Provided that elements in all
modalities as well as representations of stimuli and original (e.g. fear) responses are included in the fear structure, emotional arousal itself should be part of the structure.
Thus, an emotional arousal should trigger the reactivation of the structure even if not directly associated with the individual trauma (Blomhoff et al., 1998; Elbert et al., 2006).
Cortical responses to affective pictures
Section 4.2 points out that responses to affective stimuli are indicated even on a neurophysiological level and reflect regulatory mechanisms for sensory input into the cortex (Paige et al., 1990).
Numerous studies addressed cortical responses to affective pictures within a rapid serial visual pictures (RSVP) design comparable to the current study. Schupp et al.
(2004) as well as Junghöfer et al. (2001; 2003) summarize two event-related components that are evoked and moderated by emotional stimuli. An early posterior negativity (EPN) was observed that starts 100 – 150 ms after stimulus onset and peaks around 250 – 260 ms. The effect was the more prominent the higher the emotional arousal of the stimulus was. A second component was identified as enhanced late positive potentials for highly arousing stimuli that occur from 350 – 400 ms onwards.
Cuthbert et al. (2000) specify that late positive voltage changes are significantly larger for affective than neutral stimuli, starting 200 – 300 ms after stimulus onset with a maximum at 1 second. Cuthbert et al. (2000) conclude that the late positive shift represents a selective processing of emotional stimuli. The report of Keil et al. (2001) parallels these findings: These authors describe significantly higher N1, early P3, late P3, and slow wave amplitudes for pleasant (N1) as well as pleasant and unpleasant (P3, slow wave window) pictures, particularly in posterior regions.
The advantage of emotional stimuli is assumed to be regulated by the elementary motivational systems as described by Lang (1979) in the biphasic emotion theory. In return, larger positive slow wave activity in response to a certain stimulus suggests a more intense emotional connotation for the participant. Cortical responses thus provide the opportunity to investigate experimentally manipulated emotional states within and across subjects: Differential brain responses to emotionally arousing material are reflected in the time course and source analysis of the responses (Keil et al., 2002). The conclusion is in line with the concept of motivated attention: Motivationally relevant stimuli automatically activate attentional resources.
Junghöfer and colleagues (2001) and Schupp et al. (2004) additionally addressed the speed in human emotion processing: How long does the stimulus have to be presented in order to provoke the emotion-specific cortical responses? Designs with 3 Hz presentation rate per affective picture (Junghofer et al., 2001) and 120 ms for threatening faces (Schupp, Ohman et al., 2004) revealed an enormous speed in human emotional processing. 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 enables the detection of a particular content. This is of particular interest to the current study as a fast, involuntary, and possibly unconscious emotional alarm response is hypothesized for PTSD patients.
Cortical responses to affective words
The current study presents affective verbal material instead of pictures. The presentation of stimuli from the International Affective Picture System (1999) to refugee adolescents was considered to be difficult for several reasons: Picture ratings refer to a healthy adult American sample who is not comparable to the current sample.
In addition, a high drop-out rate was feared when presenting emotionally arousing IAPS pictures to PTSD adolescents, e.g. mutilation pictures. Since adolescents were recruited for the current sample, both methodological and ethical considerations prompted the selection of affective words. There is evidence that visually presented verbal material provokes comparable emotion effects among healthy (e.g., Bernat et al., 2001; Herbert et al., 2006; Kissler et al., 2006; Ortigue et al., 2004; Schapkin et al., 2000; Skrandies, 1998; Vanderploeg et al., 1987), and other clinical samples (Flor et al., 1997; Kissler et al., 2006; Knost et al., 1997; Pauli et al., 2005; Williamson et al., 1991).
Nevertheless, presenting affective words instead of pictures implies several disadvantages. For instance, a long-lasting debate about the time course of semantic access to affective words has to be considered with respect to the current experiment: A traditional perspective claims that perceptual but no meaning-related features are expected in ERPs until 150 - 200 ms after word stimulus onset (Posner et al., 1999).
Semantic access is assumed to occur around the N400 (Kutas & Federmeier, 2000). But the evidence contrasts this traditional view and reports emotion-related effects from 80 ms onwards (e.g., Bernat et al., 2001; Kissler et al., 2006, for a review; Ortigue et al., 2004; Posner et al., 1999; Skrandies, 1998).
For instance, Skrandies (1998) reports differential cortical responses to extreme nouns on the Osgood dimensions. Differences occurred from 80 ms onwards. Bernat and colleagues (2001) demonstrate different brain responses between unpleasant and pleasant adjectives for P1/N1, P3, and late positivity. Ortigue et al. (2004) show emotion-specific ERP differences for visually presented words in a lexical decision task. Kissler and colleagues (2006) report an occipito-temporal (left) negativity with a maximum at 260 ms after stimulus onset that differed for emotionally arousing and neutral words (nouns and adjectives). Herbert and colleagues (2006) further report differential P2 components for emotional (both pleasant and unpleasant) compared to neutral adjectives.
The literature discusses neurophysiological short-cut routes that mediate very early emotion-specific responses (e.g., LeDoux, 1995). Nevertheless, these routes between thalamus, amygdala, and cortex imply “reading skills” on subcortical levels.
Neurophysiological indicators for this “prelexical” word processing were found to start at least 150 ms after stimulus onset. A left occipito-temporal negativity was reliably demonstrated for written words (Dehaene et al., 2002). Dehaene and colleagues (2002) claim the existence of a visual word form area (VWFA) located in the left fusiform gyrus that enables a prelexical representation of visually presented words. Deficits in word reading for VWFA lesions further support this hypothesis as well as an involvement of the VWFA in word reading for both real words and pseudo-words (Dehaene et al., 2002).
In conclusion, the current experiment will focus on early effects as well as P3 and late positivity.
Contributions from previous evidence among PTSD samples
So far, to the author’s awareness, only a single study presented written words to PTSD patients (Blomhoff et al., 1998). Meaningful emotional and meaningless (distractor) words were presented within an oddball paradigm to survivors of a ship fire disaster with high and low PTSD symptoms. A biphasic waveform resulted in response to the verbal distractor stimuli: An early negative deflection was observed in the first phase whereas the second phase was characterized by a positive deflection. High and low symptom groups differed: The high symptom group showed stronger responses in the first phase. This was true for the low symptom group in the second phase. The authors referenced to the activation of a specific emotional or semantic network.
In addition, several neurophysiological studies presenting affective pictures already provided insights in emotional stimulus processing among PTSD patients. On the one hand, there are indicators for a hypersensitivity for unpleasant environmental stimuli referenced by very early cortical responses that start around 100 ms after stimulus onset in the visual cortex (Junghofer, Schauer et al., 2003) and orbito-frontal regions (Borgelt, 2006). On the other hand, results are still somewhat inconsistent, as e.g. Saleptsi (2005) did not find such early effects. Participants in all studies were recruited among adult refugees who survived organized violence and torture. Results were compared to healthy but ethnically matched controls (Borgelt, 2006; Junghofer, Schauer et al., 2003; Saleptsi, 2005) as well as a clinical schizophrenic control group (Borgelt, 2006; Saleptsi, 2005), and healthy German controls (Borgelt, 2006).
Participants viewed emotionally arousing and neutral pictures of the international affective picture system (IAPS, 1999) in various presentation times (300 ms – 6 s each).
Besides the very early effects, further indices of an altered emotion processing were reported, e.g., in the (early) P3 window: Junghöfer et al. (2003) report an EPN between 100 and 300 ms after stimulus onset over extended visual areas in response to affectively arousing pictures. The EPN in frontal lobe areas was prolonged in PTSD patients and particularly for unpleasant pictures. Saleptsi (2005) shows enhanced prefrontal and parietal P3 amplitudes for unpleasant stimuli. Borgelt (2006) describes a failed habituation in the cortical PTSD “alarm response” to unpleasant pictures between 200 and 300 ms. Selective processing of unpleasant pictures in late time windows (up to 600 ms) in temporal regions was further reported for PTSD patients. In addition, both Saleptsi (2005) and Borgelt (2006) report significant correlations between visually evoked cortical responses and clinical avoidance measures (Saleptsi, 2005) or arousal and overall symptom score (Borgelt, 2006).
The evidence encourages the idea that the fear structure is easily activated by highly arousing unpleasant stimuli (Borgelt, 2006; Junghofer, Schauer et al., 2003).
Furthermore, references for facilitated inclusion of additional elements in the associative structure were investigated, too: A relative selective recognition performance for unpleasant compared to neutral and pleasant stimuli within PTSD patients points in that direction (Borgelt, 2006).
Karl, Malta, and Maercker (2006) recently published a meta-analytic review that considers studies addressing event-related components among PTSD participants.
Unfortunately, comparability to the current experiment was low as the evidence neither
refers to refugees nor to adolescents. In addition, the included studies did not emphasize affective stimulus processing. Nevertheless, some important details with respect to pathologically altered event-related responses will be considered in the current discussion.
General expectations
Indicators for “normal” word processing are expected for healthy controls. More specifically, the current analysis will focus on indicators for an early word processing as described by several authors (Bernat et al., 2001; Kissler et al., 2006; Ortigue et al., 2004; Posner et al., 1999; Skrandies, 1998): A preference for emotional (particularly pleasant) compared to neutral words should be observable in time windows for N13, early P3, late P3 and late positivity. Current source localization should indicate visual word processing in the ventral stream. That is, early responses should occur in the primary visual cortex, whereas prelexical stages of word processing should emerge in the VWFA as described by Dehaene and colleagues (2002). Enhanced frontal activity in the early P3 time window should indicate the preference for emotional (pleasant) words in higher amplitudes. (Late) P3 amplitude as an index of attention, stimulus probability, and task relevance (Donchin & Coles, 1988) should reflect the allocation of attentional resources. A contextual update of stimuli with the working memory and the embedding of the stimulus in its semantic context that starts around 400 ms (Kissler et al., 2006) should indicate a continued stimulus processing of control participants – with a particular preference for emotional (pleasant) stimuli.
PTSD-specific expectations
The current study aims at investigating cortical responses to affective words among PTSD adolescents that might activate the fear structure. It is expected that even a negative emotional arousal is enough to touch a fear network element, which might lead to altered cortical activations during affective processing.
Brain activity related to affective processing should thus be different in the PTSD group compared to ethnically matched healthy controls. More specifically, an early posterior negativity starting around 100 ms after stimulus onset is expected for affective
3 Note that the historical EEG nomenclature is used to label the analyzed time windows. The terms have been chosen in order to facilitate the comparability of the time windows with previous evidence although the current data refer to MEG data.
(unpleasant) words in the PTSD group. Respective findings have been reported for occipital (Junghofer, Schauer et al., 2003) and prefrontal regions (Borgelt, 2006).
Furthermore, differences in frontal activity during processing of unpleasant stimuli are expected for the P3 time window and late positivity: The inclusion of emotional arousal in a pathological fear structure suggests both differences for the “healthy” preference of emotional (pleasant) words reflected in larger P3 amplitudes and the late positive shift indicating the embedding in individual associative networks.
7.2 Methods
Sample recruitment and description
Altogether, N = 26 adolescent refugees participated; n = 10 of them fulfilled the DSM-IV criteria of PTSD at the time of evaluation, n = 16 subjects without such a diagnosis were assigned to the control group. Sample recruitment occurred within an epidemiologic study addressing the prevalence of psychiatric disorders among adolescent refugees (Ruf, in preparation). Three participants contacted the Psychotrauma Research and Outpatient Clinic for Refugees of the University of Konstanz by themselves and were invited to participate in the study. None of the subjects was a native German speaker. To guarantee sufficient reading ability, study participation was only offered to adolescents speaking and reading German fluently. All participants were right-handed. Besides offering psychological treatment to those suffering from PTSD, all subjects were paid 10 Euros for participation.
Movement artifacts in the neurophysiological data led to the exclusion of one PTSD and four control subjects. Data of three additional control subjects were excluded from the analyses, since these subjects previously suffered from PTSD but were successfully treated at the time of evaluation (Ruf, in preparation). Finally, data of n = 9 adolescents in each group were analyzed.
A survey of demographic variables is shown in Table 7.1 (next page). Groups did not differ in age (t(16) = .00; p = 1.00), general intelligence (t(13) = -1.06; p = .31), and concentration performance (t(13) = -.93; p = .37 for overall performance and t(13) = -1.12;
p = .28 for error rate).
Table 7.1 Demographic specifications of participants
Seven out of nine control participants reported having experienced at least one traumatic event in their lives (1 to 3 different event categories4). The adolescents named e.g. seeing dead people (excluding funerals), witnessing somebody else been physically attacked or severe accidents (witnessed and experienced by themselves). PTSD participants reported more and a broader range of traumatic experiences (1 to 10 different traumatic event categories): Witnessing a physical attack against family members (mostly the parents during detentions by soldiers) was reported most often, followed by war experiences such as bombing, shelling or burning houses. As at least one traumatic experience is requested for PTSD diagnosis, none of these adolescents reported no traumatic experience at all.
Diagnostic and experimental procedure
Diagnostic interviews and the experimental procedure were performed individually in the laboratories of the University of Konstanz. The diagnostic interviews
4 Note that the number of different reported event categories is listed, not the total number of traumatic experiences.
comprised a set of demographic variables and both the UCLA Posttraumatic Stress Disorder (PTSD) Reaction Index for children (Rodriguez et al., 1999) and the Mini-international neuropsychiatric interview (M.I.N.I. child version, Sheehan et al., 1998).
In addition, general intelligence was assessed with the Colored/Standard Progressive Matrices (Raven, 1962, 1976) as well as general concentration performance by Test d2 (Brickenkamp, 2002).
During the MEG measurement, participants were lying in a comfortable position and were instructed to read the words presented on the ceiling. A video camera was installed to observe the patient’s well-being inside the magnetically shielded measurement cabin. In addition, an assistant joined the participants during the whole experiment.
Stimulus selection and presentation
The stimulus set consisted of 120 German nouns. Forty pleasant, unpleasant, and neutral words each were selected on the basis of the prestudy among refugee adolescents (see section 6) and German students (Herbert et al., 2006). The stimulus set was repeated three times to enhance trials per condition. Thus, 360 stimulus presentations resulted in total. The sets of pleasant, unpleasant, and neutral words were matched regarding word length and frequency (F(2,117) = .05, p = .95 for word length;
F(2,117) = .61, p = .55 for word frequency). The latter was determined by the CELEX
data base (Baayan, Piepenbrock, & van Rijn, 1993).
Stimuli were presented in an RSVP design with a 1 Hz frequency and without an inter-stimulus interval. The list of stimuli was presented randomly. The stimulus presentation was realized with Presentation and took 6 minutes and 20 seconds. The full set of stimuli is depicted in Table 12.3 in the appendix.
MEG recordings and preprocessing
MEG data were recorded using a whole-head magnetometer (148 channels, MAGNES; 2500 WH, BTi, San Diego, CA) with a sampling rate of 508.63 Hz and an online 0.1 Hz high-pass filter (bandwidth 100 Hz). The individual head shape of each participant and five index points (left and right preauricular points and three locations on the forehead) were digitized in order to calculate the relative head position in the scanner during the experiment. Furthermore, vertical and horizontal eye-movements (EOG) were derived from above and below the eye and from the outer canthi to allow
off-line artifact correction during MEG measurement. For the same purpose, an electrocardiogram (ECG) was assessed with one electrode on each forearm during stimulus presentation. Both physiological measurements were realized by a SynAmps amplifier (NEUROSCAN) using AgCl electrodes.
Stimulus presentation was provided via video beamer (JVCTM, DLA-G11E) and a mirror system projecting stimuli onto the ceiling of the measurement cabin.
Artifact correction was realized off-line. First, data were corrected for noise and cardiac activity. Bad channels were identified and interpolated. Afterwards, the continuous data stream was corrected for eye-blinks using the multiple source eye correction MSEC algorithm as implemented in BESA (Berg & Scherg, 1994), extracting typical eye-blinks from the continuous data stream and then identifying similar artifacts during the experimental run.
Subsequently, the different stimulus epochs of 1000 ms post-stimulus length were
Subsequently, the different stimulus epochs of 1000 ms post-stimulus length were