The previous section dealt with influences of affective stimuli on experimentally observable behavior. Automatic and uncontrolled processes were focused within simple tasks in priming paradigms. The current section extends the approach and addresses influences from affective stimuli on explicit associative learning tasks. Indicators for an altered subjective meaning of affective words in PTSD patients are presumed both in behavioral (e.g., learning performance) and neurophysiological data. The current section thus demonstrates insights from previous evidence on both fields. The indicators for a pathological appraisal will be outlined with respect to the associative (fear) structure perspective outlined above.
Contributions from associative learning to the associative structure perspective
Associative fear structures (Foa & Kozak, 1986; P. J. Lang, 1979) have been described as a particular example of propositional (P. J. Lang, 1979) or representational networks of highly interrelated elements. Each of the elements is supposed to A) represent an individual concept (e.g., Collins & Loftus, 1975) and B) be at some level of activation. The activation itself may spread along connecting links as described above for the spreading activation mechanism. Individual or different samples of concepts are accessible to consciousness if its activation reaches a certain threshold.
The current section proposes that associative learning provides insights in individual associative structures for the following reasons: Associative learning should be enhanced or facilitated for subjectively related concepts. In other words: Two concepts that have been repeatedly activated simultaneously have a higher probability to be more or less directly linked within the representational network. If thus one of the concepts is activated and its activation spreads along upcoming links, activation probability of the other (target) concept is enhanced. The concept’s activation is necessary for an explicit processing task. The easier the activation of particular concept’s representation, the higher is the encoding (learning) success. Concept pairs with higher subjective relations should thus be easier to learn. By contrast, a better learning indicates subjective relations between the stimulus materials.
Associative learning is thus seen as a possible approach to detect subjective relationships. The following paragraphs outline previous evidence for the presumed coherences on different response levels that are investigated in the current dissertation:
both behavioral and neurophysiological measures. The next paragraph emphasizes indicators for pathologically shifted relations reflected in associative learning performance. Afterwards, neurophysiological indicators for associative learning in general are outlined as well as studies among clinical (in general and PTSD) samples.
Associative learning in PTSD patients
The associative learning method to investigate pathologically altered associative structures in PTSD is relatively rare: Only two respective studies were published by Gurvits and colleagues (1993) as well as Golier and colleagues (2002).
Gurvits and her colleagues (1993) studied Vietnam combat veterans with and without PTSD. The design comprised a neurological examination, neuropsychological testing and an EEG following sleep deprivation. Verbal associative learning was part of the neuropsychological test battery. Results revealed several neurological “soft signs” in PTSD and significant correlations between the neuropsychological test scores and the neurological soft signs. For instance, motor coordination, motor sequencing, sensory integration, primitive reflexes, and apraxia or agnosia correlated with the neurological measures. But no significant group differences were found for verbal associate learning.
Golier and her group (2002) investigated Holocaust survivors with and without PTSD. Additional healthy Jewish adults who had not been exposed to the Holocaust participated in a memory performance task. All participants were asked to learn 12 semantically related and unrelated word pairs. Learning performance was measured in both a recall and a word stem completion task (the latter consisted of both words previously shown and new words). Holocaust victims with PTSD demonstrated fewer recall performance for semantically unrelated words (compared to survivors without PTSD) and semantically related words (compared to healthy controls). An incremental performance decrease with increasing age resulted in the PTSD group. No group effects were found for the word stem completion task in contrast. In conclusion, explicit but not implicit memory performance seems to be impaired in the clinical group. It remains unanswered whether this is rather a risk factor or a consequence of PTSD. Explicit memory performance requires a conscious recollection of the memory elements that again involves hippocampal functioning. Hippocampal structures are discussed to be
impaired in trauma victims suffering from PTSD (Bremner, 2002; Buckley et al., 2000) and in advanced age (Golomb et al., 1993). Hence, a final decision between PTSD and age causing the low performance is not possible but explains the specific memory impairment for explicit learning. Unfortunately, no valence- or trauma-related effects were investigated in this study.
What about further indicators for a PTSD-specific memory deficit?
Buckley and colleagues (2000) further reviewed the literature considering information processing in PTSD patients but this paragraph is restricted to findings pertaining to the current dissertation. A poorer general verbal (and nonverbal, Jelinek et al., 2006) memory performance in PTSD patients has been found in some studies (e.g., Bremner et al., 1993; Buckley et al., 2000, for a review; Jelinek et al., 2006; Yehuda et al., 1995) but not in others (e.g., Gurvits et al., 1993). Unfortunately, these studies were not run within associative learning paradigms. More specifically, Buckley et al. (2000) conclude that there is substantial evidence that PTSD patients show memory deficits even for trauma-neutral information. Anyhow, it is not yet defined at which stage of information processing these deficits occur. Yehuda et al. (1995) claim that it is rather a circumscribed cognitive deficit at the retrieval stage of processing, and does not affect the encoding process. To investigate this question, some researchers have adopted the classical Stroop paradigm, and replaced stimuli with disorder-related words (e.g. “heart attack” for panic disorder patients) and other word categories. Stimuli are presented above or below (with subsequent stimulus masks) conscious perception threshold.
Reaction time differences between the different word categories are seen in relation to disorder-specific attention. Harvey et al. (1996) for instance found delayed vocal response latencies in a color naming task for trauma-related words (masked and unmasked) among car accident victims. McNally et al. (1996) in contrast did not so for ex-Vietnam combatants. McNally et al. (1996) presented trauma, positive, and neutral words to Ex-Vietnam combatants in a first run and then mixed the words with new ones in a masked condition. The authors expected PTSD patients to identify masked trauma words better than controls, but they did not.
However, Buckley et al. (2000) conclude that a final distinction whether experimental effects within the modified Stroop paradigm are due to disorder-specific processing or related to generally negatively valenced material is not yet possible. To clarify this question, experimental designs should include apart from healthy Controls
also groups with other anxiety disorders. But unfortunately, so far too few such studies have been published.
With respect to the question if PTSD patients are biased to automatically process threat stimuli, evidence is somewhat inconsistent, too (Bryant & Harvey, 1997; Harvey et al., 1996; McNally & Amir, 1996). Buckley et al. (2000) summarize that there is limited evidence that PTSD patients process threatening (disorder-specific) stimuli at a prerecognition level of information processing. Inconsistent results are mainly explained by methodological reasons. In addition, Buckley et al. (2000) take into account that differences might also be due to the duration of PTSD: Predominance of clinical symptoms such as intrusion symptoms and emotional numbing differ in the course of PTSD and might have different impact on the observable behavior.
Furthermore, ecological validity of such experiments is a problem since patients do not suffer from intrusions from single words.
McNally et al. (1994) further investigated combat veterans’ memory processes within an autobiographical memory paradigm: Their participants watched either a combat related or a neutral video tape. Afterwards, the subjects retrieved a specific autobiographic event that can be associated to a presented word (positively and negatively valenced and neutral). Results revealed a tendency for an over generalized memory retrieval to positive cues within the PTSD group. That is PTSD patients had difficulties in identifying specific personal memories in response to positively valenced cues. This “over generalized” memory retrieval is enhanced when subjects were primed with a combat-related video tape compared to the neutral tape. The results were interpreted as indicating an emotional numbing in PTSD and will of particular interest in the discussion of section 9.
Insights from associative learning experiments among other clinical samples
Hermans, Pieters, and Eelen (1998) conducted a study among eating disorder patients and controls testing disorder-related abnormalities both for implicit and explicit memory. The explicit memory test was operationalized in cued recall tasks for 4 word categories: pleasant, unpleasant, neutral, and anorexia-related. Anorexia patients showed better recall performance for anorexia-related words compared to both their performance in the other word categories and the Controls’ performance. Implicit memory was tested within a word stem completion task. No disorder-related effects appeared in this case. Hermans et al. (1998) interpret this as an indicator for enhanced
self-related memory schemata and a selective content-specific processing bias which might in some ways be comparable to trauma-specific memory schemata.
Lundh, Czyzykow, and Öst (1997) ran a similar experiment among 30 panic patients with agoraphobia and 30 controls. The authors presented positive, neutral, social threat, and physical threat words. Disorder-specific better learning for physical threat words was again found in a cued recall task, but not in a word stem completion test. The disorder-specific bias even correlated with anxiety and avoidance measures.
This is to be seen in line with the assumed fear structure bias in PTSD patients.
Disorder-specific alterations in memory performance have been additionally investigated in depressive patients (Watkins, Vache, Verney, Muller, & Mathews, 1996). Watkins, Vache, Verney, Muller, and Mathews (1996) instructed patients with clinical depression to study words with positive, neutral, and negative valences and to produce free associations to various cues afterwards. The authors aimed at measuring the amount of words previously studied (primed). Indeed, the depressive patients named more negative, i.e. mood-congruent words from the study list, whereas controls did so for positive words. The authors discuss this finding as proof for the mood-congruent memory bias, which is one of the sustaining mechanisms in depression. This refers to the tendency of depressed or sad people to remember particularly information consistent with the own mood. A vicious cycle results in the tendency to remember more mood-congruent negative memories, which in turn reinforce the negative mood and so on.
Comparable results were previously obtained by Bradley, Mogg, and Williams (1994) for sub-clinically depressive participants.
In summary, evidence among other clinical samples suggest that memory tests may detect disorder-specific biases.
Moderating factors in verbal learning paradigms
Several factors have been shown to improve verbal learning performance in general (Buchanan, Etzel, Adolphs, & Tranel, 2006): Performance of healthy subjects in free recall or recognition tests is best for both highly arousing taboo and semantically related word pairs. Performance decreased for word pairs that are only semantically related but only moderately arousing and performance is even less for unrelated and less arousing word pairs. Hence, additive effects of semantic relatedness and emotional arousal on memory were reported. In a similar vein, Talmi and Moscowich (2004) already investigated whether semantic relatedness explains the enhancement of memory
for emotional words with three experiments. The authors aimed at differentiating between the performance enhancing moderators “emotional connotation of the words”
and “semantic relatedness” and presented three types of stimuli: emotional words, categorized neutral words (semantically related from two categories: kitchen and music), and random neutral words. Results revealed better performance for categorized neutral compared to emotional words. In addition, performance of both was better than for random neutral words. Semantic relatedness thus has stronger effects on word pair learning performance than emotionality.
Neurophysiological indicators of associative learning
The current dissertation further investigates neurophysiological indicators of associative learning. The following paragraphs summarize cortical indicators for stimulus encoding, rehearsal, context embedding, etc. that goes beyond the event-related fields/potentials described in section 4.2. That section generally dealt with phasic alterations in the activity level in short time windows that follow a specific event.
These activity level alterations are referred to as components that are distinguishable with respect to the experimental paradigm on the basis of its latency, topography, and relative polarity (EEG).
This section additionally emphasizes slow event-related brain potentials, so-called slow waves. The term refers to tonic voltage changes that last for at least 500 ms, eventually persisting for several seconds (Rosler et al., 1993). EEG measurements during sentence reading for instance reveal a phasic negativity in the left anterior cortex, a particular slow wave example. The so-called LAN-phenomenon (left anterior negativity) is reported to be related to the encoding of words and syntax processing that can not assigned to roles and propositions immediately (Rösler et al., 1997). Storage of verbal information into long-term memory is additionally related with negative slow potentials over the frontal cortex (Rosler et al., 1993). The negative potential over left frontal cortex is thus accepted as a correlate of long-term memory search. The amplitude of this slow wave increases with memory load and evaluation of the stimulus significance (Birbaumer, Elbert, Canavan, & Rockstroh, 1990) and is further related to the task difficulty. The topographic distribution in contrast depends on the task quality (Rösler et al., 1997). The topography of slow ERPs is assumed to reflect the relative activation/inactivation of distinct cortical cell assemblies whereas the amplitude of the
negative maximum is related the amount how much a certain cell assembly is activated at a particular time (Rosler et al., 1997).
Physiologically, slow waves are associated with the synaptic activity at the apical dendrites of the cortical pyramid cells (Birbaumer et al., 1990) and indicate a state of a relative activity enhancement/decrease of the respective cortical area. Negative slow waves are seen as manifestation of enhanced excitatory postsynaptic potentials in the upper cortical layers, thus as an indicator for apical dendrites depolarization. Positive slow waves in contrast indicate less excitatory postsynaptic potentials (hyperpolarization of apical dendrites). Thus, negative slow waves indicate increased excitability of the cortical tissue and a state of relative activation, whereas positive slow waves indicate the contrast: decreased excitability and relative inactivation or inhibition of cortical tissue (Rosler et al., 1993; Rosler et al., 1997).
Section 9 examines subjective associative structures with associative word pair learning under EEG-measurements. A consideration of how and where objects or concepts are represented in the brain is thus of particular interest. Recent models postulate in contrast to earlier assumptions no specific regions that are specialized for the representation of a specific object. It is rather postulated that the same cortical pattern is activated for the retrieval and for the original stimulus confrontation. Hence, situation-specific retrieval pattern are evidenced instead of particular (modality-specific) locations for episodic contents. Neocortical structures are consequently assumed to perform both information processing and information storage (Rösler et al., 1997). For instance, slow waves are assumed to reflect memory retrieval. Their respective maximum was found in the same areas that are involved in online information processing of the same stimulus type: verbal stimuli evoked slow wave activity over parietal and left frontal areas during encoding and retrieval (Rösler et al., 1997).
The consideration of cortical responses of the simultaneous presentation of word pairs (or whole sentences) is further important for the current study. The EEG characteristics change during presentation of word pairs and depend on whether the two words match in terms of content. A phasic negativity between 300 and 500 ms has been found for words presented at the end of a sentence when it did not match the content of the sentence. This “N400-effect” amplitude is the more expressed, the more the stimulus deviates from the context (Rösler & Hahne, 1992). The effect illustrates that additional activated network elements are necessary to integrate a certain word within its context.
Contributions from previous slow wave studies among healthy subjects
The first event-related potential within a learning paradigm was shown in animal (monkey) studies (Stamm & Rosen, 1972). The authors report a slow negative potential shift in the principalis cortex related to the learning success. In humans, event-related recordings indicated that left frontocortical activation is associated with memorizing verbal material in a paired associate learning paradigm (M. Lang et al., 1987). Lang et al. (1987) instructed their subjects to built permanent associations between letters and Morse codes by trial and error. Left frontal activity was particularly found in the beginning of learning, whereas an increasing slow frontal negativity was found in later stages of learning. The results suggest that cognitive learning processes are related to slow negative potential shifts as these slow negative frontal shifts were specific for a learning task compared to a control task. Furthermore, a facilitation of learning was found when stimulus material is presented during spontaneous occurrence of frontal negative potential shifts (Bauer & Nirnberger, 1980). Hence, Lang et al. (1987) hypothesize that the frontal lobe is more activated the more the subjects utilize the experience of previous trials.
The same work group (W. Lang et al., 1988) continued analyzing slow potential shifts by instructing subjects to memorize pairs of words, either meaningful or meaningless. The word pairs were either pre-experimentally (in semantics) related or to be learned during the experiment. A larger slow potential shift was found for to be learned than for the pre-experimentally existing word pairs. The authors assume the potential shift to indicate the importance of the left frontal lobe for elaborative encoding of verbal material.
The group continued their work (Uhl, Lang, Lang, Kornhuber, & Deecke, 1990) and addressed the specific contributions from semantic and verbal access in general.
The subjects learned associations between complex stimuli such as unfamiliar black and white faces and melodies. Uhl et al. (1990) report a bilateral slow negativity at frontal electrodes in the non-semantic and non-verbal paired associate learning task. Hence, the nonverbal stimuli resulted in symmetric frontal activation in contrast to a left hemispheric dominance.
Rosler, Heil, and Glowalla (1993) investigated slow wave changes during long-term memory retrieval in a Fan effect paradigm: Subjects are instructed to learn several (e.g. three) lists of words. A single word appears either in only one, in two, or in all three lists. The target run comprises word pairs. The participants have to decide whether
there is a relationship between the two words. Decision latencies increased with the number of associations in the formerly learned lists. Neurophysiological parameter revealed a negative potential over left frontal areas during retrieval. The amplitude of a (bilaterally distributed) frontal negative slow wave increased with the diversity of associative structure that had to be searched for (number of lists that includes the target word).
But is the left lateralization due to semantic relationships within the verbal material (Rosler et al., 1993)? Three studies investigated long-term memory retrieval emphasizing slow frontal negativity (W. Lang et al., 1988; Rosler et al., 1993; Uhl et al., 1990): The two studies stimulating with semantic material (W. Lang et al., 1988;
Rosler et al., 1993) found a lateralization within the left hemisphere, whereas Uhl et al.
(1990) report a bilateral frontal negativity for non-semantic material.
A more recent hint for late activity in language processing was given by Spironelli and Angrilli (2006). The authors aimed at demonstrating that the cortical activation pattern is less word specific than rather task dependent. The same set of words was presented either in a phonological, semantic, or orthographic task within a reaction time experiment and EEG measurement. Indeed, the authors found negative potential shifts in two late time windows: 1000 2000 ms after stimulus onset (initial CNV) and 2000 -3000 ms (terminal CNV). The negative potential shifts were bilaterally expressed only in semantic tasks. Results parallel the literature on ERPs and semantic processes
A more recent hint for late activity in language processing was given by Spironelli and Angrilli (2006). The authors aimed at demonstrating that the cortical activation pattern is less word specific than rather task dependent. The same set of words was presented either in a phonological, semantic, or orthographic task within a reaction time experiment and EEG measurement. Indeed, the authors found negative potential shifts in two late time windows: 1000 2000 ms after stimulus onset (initial CNV) and 2000 -3000 ms (terminal CNV). The negative potential shifts were bilaterally expressed only in semantic tasks. Results parallel the literature on ERPs and semantic processes