The ability to actively maintain tactile information is a prerequisite for complex behaviors such as object recognition. Tactile WM is crucial for blind people who predominantly rely on their tactile sense in everyday life, e.g. for reading Braille writing. Moreover, the tactile modality is being increasingly applied as an additional channel to provide information in environments with visual and acoustic information overload, e.g. in a cockpit. However, as opposed to WM for visuo-spatial or phonological information, WM for tactile stimuli has been largely neglected by psychological and neuroscientific research. The present dissertation, therefore, aimed to shed more light on the mechanisms and the neural basis of tactile WM. This dissertation was guided by two methodological premises. First, the investigation of the neural basis of tactile WM should start with a relatively simple task manipulating only specific stimulus features and involving few cognitive processes. Second, to get a complete understanding of tactile WM, converging evidence provided by different methods and integration of findings over species is necessary. Therefore, in all experiments of this dissertation a vibrotactile delayed discrimination task was employed which has been used in previous studies using the methods of single-unit recordings in non-human primates and behavioral experiments in humans. Our experiments aimed to add evidence according the neural network supporting the same task in humans.
First, this dissertation addressed the question of which neural network is related to vibrotactile WM. In Study I, fMRI was used to identify the brain regions supporting task performance in the vibrotactile delayed discrimination task. A distributed network including S1, S2, ventral and medial PMC, lateral PFC and PPC exhibited increased activity during the different task periods (encoding, maintenance, decision making). This network is similar to - but comprises more brain regions - than the network previously studied in non-human primates. However, although distinct patterns of brain activity could be related to the different task periods, it is not clear which specific function the respective regions have for vibrotactile WM. Studies manipulating specific task processes might elucidate this issue further. For instance, two recent studies demonstrated the importance of the lateral PFC for cognitive control in vibrotactile WM. One study confirmed that the difficulty of the decision in vibrotactile WM depends on Weber's law and demonstrated that the activity of the lateral PFC increases with increasing decision difficulty (Pleger et al., 06). A second study showed that active retrieval of information about vibrotactile stimuli is accompanied by an increased connectivity between lateral PFC with S2 and PPC (Kostopoulos et al., 07).
The EEG study (Study II) showed that, in addition to brain regions activated during the maintenance of vibrotactile stimuli, the parieto-occipital and primary somatosensory cortex seem to be functionally inhibited during the delay period. It has been suggested that the function of this inhibition is to prevent task irrelevant and possibly interfering information from entering the brain regions actively maintaining the memory trace (Klimesch et al., 07). In addition, the EEG study showed that alpha power was increased during the delay period at prefrontal and posterior parietal sites. Prefrontal alpha power was further modulated by delay length. Neural synchronization in the alpha frequency range between PFC and PPC has been suggested to reflect the functioning of a global neural workspace controlling cognitive processing including WM (Palva und Palva, 07). The interaction between brain regions has not been investigated in this dissertation. However, in future studies the manipulation of task demands could be related to the amount of synchronization or coherence between brain regions as in indicator of fronto-parietal top-down modulation. In addition, the analysis of oscillatory activity between brain regions is a promising correlate of actively maintained memory representations (von Stein und Sarnthein, 00).
While it is accepted that S1 is crucial for the encoding of vibrotactile stimuli into WM, it is still unclear what role this brain region plays for the active maintenance of tactile information. The EEG experiment (Study II) and the experiment using concurrent subliminal stimulation (Study III) addressed this issue and suggested that the activation level of S1 is still enhanced during the early delay period. In the EEG study, the amplitude of the rolandic rhythms over the contralateral somatomotor cortex was still reduced compared to the pre-trial baseline and only reached its maximum during the middle portion of the delay period. Concurrent subliminal stimulation, previously shown to inhibit S1, impaired discrimination performance when applied during the early delay period for subjects receiving relatively high levels of subliminal stimulation indicating that S1 activity during this period is functionally relevant. In addition, the performance comparison between high and low performing subjects in the EEG study suggested that the inconsistent finding between human subjects and monkeys might be related to the velocity and efficiency with which subjects can encode the vibrotactile stimulus. This interpretation further implies that S1 is not actively involved in the maintenance of the vibrotactile memory trace but instead that some subjects still consolidate the representation of the vibrotactile stimulus during the early delay period.
In both studies the clearest finding regarding the role of S1 for vibrotactile WM was found in the pre-trial periods. The lower amplitude of the rolandic rhythms in WM compared to the control condition probably reflects anticipatory attention associated with a tonic up-regulation of S1 in the WM trials. The results of the study applying concurrent subliminal stimulation additionally suggest that reducing the background noise in S1 by increasing local inhibition improves performance. This performance improvement might indicate that lower levels of noise in S1 before the first stimulus is applied prepares S1 to process the upcoming stimulus and subsequently facilitates encoding. Linking the two studies, it has been shown that increasing inhibition abolishes alpha-like oscillations and bursting activity in the rat somatosensory "barrel" cortex and that this physiological state is beneficial for faithful encoding as opposed to stimulus detection (Nicolelis, 05; Swadlow, 03). Our data might also suggest that subliminal stimulation reduces synchronous alpha activity in S1. In the pre-trial period this would reflect enabling a physiological state related to increased anticipation and decreased noisy background activity which is beneficial for performance. Further studies could investigate the relationship between oscillatory activity over the somatomotor cortex and the effects of subliminal stimulation.
Together, the results of this dissertation regarding the role of S1 in vibrotactile WM suggest that S1 is not actively involved in the maintenance of the WM trace but that the physiological state of S1 is dynamically regulated to optimize task performance. The proposed dependence of S1 activity during the early delay period on the subjects' encoding efficiency could be tested by training subjects in the vibrotactile delayed discrimination task and analyzing changes in performance and S1 activity over time.
Already Cowan (93; 88) emphasized the importance of activated LTM representations for WM. More and more behavioral evidence suggests that in magnitude discrimination tasks, implicit LTM representations about the average value of the stimulus set presented during the experiment influence task performance giving rise to the TOE (Sinclair und Burton, 96; Hellstrom, 85; Masin und Fanton, 89). Results of the behavioral experiments of this dissertation indicate that the TOE is a robust phenomenon in vibrotactile discrimination tasks independent of specific task parameters (Study IV). The fMRI data (Study V), additionally suggest, that during the presentation and delay period current sensory evidence provided by the vibrotactile standard stimulus is integrated with the representation about the average stimulus. The somatosensory cortices and the PPC, which have been shown to be involved in encoding and maintaining the vibrotactile stimulus, are associated with this integration process. These results provide neural evidence for sensory accounts that explain the TOE by assimilation or integration processes (Helson, 64; Hellstrom, 85). In addition to this pre-decisional source, it could be that in trials where the sensory evidence about the first stimulus is weak or completely lost, subjects could base their discrimination on the average representation only (Masin und Fanton, 89). The study was not designed to decide between pre-decisional and decisional accounts of the TOE but it shows that the TOE at least partly results from pre-decisional processes. Further studies are necessary to disentangle the relative impact of these implicit LTM before and during decision making. Interestingly, the subjects were not aware that they used average information. Therefore, it seems that the TOE in vibrotactile discrimination is another example of how average or prototype information implicitly influences behavior (Hellstrom, 85).
I would like to conclude by integrating the findings of this dissertation and previous studies on vibrotactile WM (see Figure 3).
The neural representation format of the frequency of vibrotactile flutter stimuli in WM is a firing rate code, i.e., the neuronal firing rate parametrically varies depending on the frequency of the presented vibration. This firing rate code is a unique feature of vibrotactile WM and substantially different from representation formats for object or spatial information; whereas object WM has been reported to rely on a categorical code with neurons selectively firing for a specific object or color (Fuster und Jervey, 81), spatial WM has been demonstrated to rely on a code that is tuned to a specific location with neurons firing maximally for the preferred location and decreasing firing rates with increasing location from the preferred location (Funahashi et al., 89). S1 generates the rate firing code and is, therefore, crucially involved in the encoding of vibrotactile information into WM. In humans, this encoding process is supported by verbal processes. The findings of this dissertation indicate that the vibrotactile WM trace is not actively maintained in S1. This is in line with findings in the visual and auditory modality where modality-specific storage of stimulus features is supported by the modality-specific sensory associated cortex but not primary sensory cortex (for an extensive review see (Pasternak und Greenlee, 05)). The tactile posterior association areas actively maintaining the vibrotactile memory trace are S2 for maintenance in the early delay period and the PPC for sustained maintenance. This suggests that the neural structure of WM is similar for the different modalities with stimulus encoding in primary sensory brain regions and maintenance in a distributed network including the modality-specific posterior association cortex. However, during the early delay period, activity in S1 can still be enhanced reflecting ongoing consolidation processes. The occurrence of S1 activity during the early period is related to encoding efficiency which depends on the subjects' ability and training. From S1, the rate code is transmitted to S2 and the inferior PPC. Both regions convey the information to lateral PFC and PMC. All these regions exhibit increased activity during vibrotactile stimulation which is parametrically modulated by vibration frequency. The existence of two populations of neurons with positively and negatively modulated firing rates in these regions increases the fidelity of the memory representation. The maintenance of a vibrotactile memory trace and vibrotactile decision making are supported by a distributed network of overlapping brain regions including S2, the PPC, the lateral PFC, and the ventral and medial PMC. The decision process evolves gradually in these brain regions and is realized by neurons encoding the difference between both vibrations.
Together, many brain regions conjointly enable vibrotactile WM. In addition, there is a gradual specialization of brain regions rather than an absolute specialization. Whereas the inferior PPC with its close connections to the somatosensory cortices might be relatively more important for the sustained maintenance of the vibrotactile memory trace, the PMC as an interface between sensory and motor regions might be especially suited to initiate the comparison process evolving into a decision. On the other hand, the lateral PFC is involved in active maintenance and decision making but its major role is the control of ongoing information processing in the other regions. The control function of lateral PFC during vibrotactile decision making is reflected by increased connectivity of lateral PFC with S2 and with the inferior PPC during active memory retrieval and by increased activity of the lateral PFC with increased task difficulty determined by Weber's law.
LTM representations reflecting the average vibration frequency heavily influence vibrotactile WM. First, in somatosensory brain regions sensory evidence about the current vibration is integrated with this average representation during stimulus encoding and maintenance. Additionally, the average information is used as a standard for the decision process when sensory evidence about the first vibration is weak or lost.
The physiological state of S1 in the pre-trial period determines the vibrotactile WM performance. On the one hand, top-down modulation in form of spatial selective attention towards the target finger primes S1 and facilitates subsequent processing. On the other hand, less noisy thalamo-cortical bottom-up processing also facilitates task performance.
|Figure 3. An integrated model of vibrotactile WM. a) summarizes the major claims of the different studies of this dissertation. b) shows the main regions of the neural network related to the different task periods. Please note that neural network supporting vibrotactile WM is much broader than depicted here. Abbreviations: S1, primary somatosensory cortex; S2, secondary somatosensory cortex; M1, motor cortex; mPMC, medial premotor cortex; vPMC, ventral premotor cortex; PFC, prefrontal cortex; IPL, inferior parietal lobe; SPL, superior parietal lobe.|
Of course, WM for vibrotactile information lacks obvious ecological validity. However, I believe that the vibrotactile delayed discrimination task provides a useful model for studying the neural basis of the multitude of processes linking perception and action as well as known psychophysical phenomena such as the TOE or Weber's law that commonly contribute to WM. At a later stage, this knowledge can be used to study more complex cognitive activities such as Braille reading or tactile object recognition which heavily relies on the interaction between the different somatosensory modalities and motor functions.
|© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.|
|DiML DTD Version 4.0||Zertifizierter Dokumentenserver|
der Humboldt-Universität zu Berlin
|HTML-Version erstellt am: