1. Gamma-band activity (GBA)
Gamma-band activity refers to those oscillations in electrophysiological recordings that lie in the higher frequency range of the temporal spectrum, typically above 20 Hz. A useful approach for a classification of the brain’s oscillatory activity is the frequently cited nomenclature introduced by Galambos (1992). Galambos distinguished between (1) Spontaneous gamma rhythms, which are not related to any stimulus, (2) Evoked gamma band responses, which are elicited and precisely time-locked to the onset of an external stimulus, (3) Emitted gamma band oscillations, which are time-locked to a stimulus that has been omitted, and (4) induced gamma band rhythms that are initiated by but not time- and phase-locked to a stimulus. In the present thesis, we will focus on the latter phenomenon as well as on the steady-state visual evoked potential (see section I.C.2), which can be regarded as a special case of evoked oscillatory activity (Tallon-Baudry and Bertrand, 1999), in that the spectral characteristics of SSVEP are mainly determined by the frequency of the external driving stimulus.
Although most of the pioneering work in the field of induced oscillations has been
conducted in animals, the GBA has also been recorded in humans as early as in the 1950’s by Lesse (1957), using deep electrodes in the human amygdala. However, the functional meaning of this phenomenon is still under debate. One major group of approaches suggests that synchronous responses of grouped cells including their timing is related to the representation of visual objects and scenes (Milner, 1974; von der Malsburg and Schneider, 1986; von der Malsburg, 1995; Engel et al., 1997). The most popular hypothesis in this context has been the ‘temporal binding hypothesis’ (or temporal correlation hypothesis), originally proposed by von der Malsburg (von der Malsburg and Schneider, 1986). According to this hypothesis, GBA is related to ‘feature binding’ or ‘perceptual integration’. This position has been advocated by authors such as Singer (Singer and Gray, 1995), and Tallon-Baudry & Bertrand (Tallon-Baudry and Bertrand, 1999), among others. In detail, the synchronized activity of neurons in distributed assemblies is assumed to encode the integrity of different aspects or features belonging to one perceived object or visual scene. This hypothetical mechanism is illustrated in Figure I.4.
Figure I.4: Illustration of the temporal correlation hypothesis of feature binding (adapted from (Singer et al., 1997);). The integrity of the triangle is encoded by the synchronized firing of the neurons that are
tuned to the same stimulus orientation.
Other proposals emphasize the importance of Hebbian cell assemblies that are formed on the basis of learned associations (Pulvermüller, 1999; Pulvermüller et al., 1999). This perspective assumes that oscillatory activity arises from well-ordered activity in networks that have been shaped by learning. This type of activity is the equivalent of the ‘reverberation’ in cell assemblies as proposed by Donald Hebb (1949). It is assumed that this mechanism is the basis for all forms of learning, including the learning of percepts during ontogenesis. If a learned association is being activated, the oscillatory activity in the respective assembly will be the correlate of activation of the learned percept. Synchronization is viewed as but one special case of organized spatio-temporal patterns.
Further suggestions as to the role of oscillatory processes include encoding the stimulus
synchronous synchronous
non-itself (Freeman, 1995). In this view, the stimulus is encoded by a complex temporal and spatial pattern that might be visualized as the ‘attractor’ of the system. One of the major features of nonlinear systems and their respective ‘strange attractors’ is that they can account for the irregularities usually found in biological systems (Elbert et al., 1994). A final role that has been posited for oscillatory processes is that of a timing mechanism (Fukai, 1995). Since various neuronal processes involve the frequency of spikes from an individual cell’s contribution to the cell assembly, a mechanism must be available for summing this process over a finite time period.
Oscillations might provide the clock signal, for example by creating a phase signature, needed to update the contribution of the individual neurons.
Empirically, synchronous gamma-band oscillations of spatially distributed cells have been reported to occur in the visual cortex of the anesthetized (Eckhorn et al., 1988; Gray and Singer, 1989; Engel et al., 1991; Freiwald et al., 1995) or alert cat (Gray and Viana Di Prisco, 1997) as well as in the awake monkey (Eckhorn and Obermueller, 1993; Frien et al., 1994; Kreiter and Singer, 1996). Furthermore, the role of gamma oscillations in plasticity, learning and memory has been emphasized by more recent animal studies (e.g. Lisman and Idiart, 1995; Murthy and Fetz, 1996a, b; Roelfsema et al., 1997). The GBA can also be recorded in humans using traditional electrophysiological recording techniques. Early work (Sheer, 1970; Spydell et al., 1979; Spydell and Sheer, 1982) has shown faster EEG oscillatory activity to be associated with human learning, cognitive performance and memory processes, which has been interpreted as an index of ‘focused arousal’ in the task-relevant neuronal assemblies. More recent reports in this area have demonstrated that modulation of GBA is associated with the presentation of real triangles and illusory triangles (i.e., Kanizsa illusion), but not visually similar nontriangles (Tallon et al., 1995). Viewing a coherently moving light bar has also been reported to increase the GBA as compared to viewing two bars moving in opposite directions (Müller et al., 1996; 1997a).
Likewise, an illusory waterfall (i.e., wavy lines in the vertical), but not wavy lines without direction, enhances gamma band activity in specific time and frequency windows, displaying topographical distributions that follow the retinotopic organization of visual cortex (Lutzenberger et al., 1995). With regard to more complex tasks, enhanced GBA was observed when subjects were required to activate an object’s internal representation of a hidden object in a visual search task Baudry et al., 1997) as well as during the delay of a short-term memory task (Tallon-Baudry et al., 1998), thus suggesting that a temporal correlation of neuronal activity also takes place in top-down processing of visual objects. A series of experiments performed by Llinas and collaborators demonstrated some evidence for the involvement of 40-Hz oscillations in states of consciousness (e.g. Llinas and Ribary, 1993). In this work, this phenomenon is interpreted as the
oscillatory manifestation of cortico-thalamic loops, that is said to constitute a neural correlate of consciousness (Llinas et al., 1998). Recently, it has been demonstrated that GBA is also modulated by spatial selective attention, lending further support to its functional relevance (Gruber et al., 1999). Makeig and Jung (1996) reported that during drowsiness, gamma band power modulations in the human EEG are related to periods of alert performance in an auditory target detection task. GBA might thus also be regarded a correlate of conscious awareness.
GBA has been controversial both in terms of methodology and theoretical implications (e.g. Bair et al., 1994; Jürgens et al., 1995). For example, since high frequency oscillations have been reported to occur in the insect brain in which integrative perceptual mechanisms are not required (Kirschfeld, 1992), the question arises as to whether human GBA is actually a signature of cognitive processes such as perception or active memory or constitutes an epiphenomenon of visual processing without functional relevance. Likewise, the functional role of GBA for feature integration processes has been doubted on empirical and theoretical grounds. More precisely, it has not yet been shown empirically that the synchronized activity of distant cells is a specific and necessary condition of feature integration (Hardcastle, 1997). Rather, alternative explanations include network coupling during perception or non-specific horizontal neuronal connections. For instance, Lamme and Spekreijse (1998) failed to observe systematic relations between the firing synchrony in primary visual cortex and properties of the visual scene. They conclude that synchrony is not necessary for feature integration and synchrony may be due to horizontal cortical connections. In an important methodological critique, Menon et al. (1996) reported that inter-electrode synchronization as measured by electrocorticogram in human showed high spatial frequencies and was restricted to small circumscribed areas. The authors conclude that measurement of brief instances of synchronized activity is not possible using electrodes on the surface of the scalp. However, this conclusion seems not to be warranted given the evidence for high-frequency neuronal responses in the human brain reviewed above. Furthermore, it has been argued that the data presented by Menon and co-workers are partly consistent with a functional role of synchronized activity in perceptual processes (Lutzenberger et al., 1997).
2. Steady-state visual evoked potentials (SSVEP)
SSVEPs are continuous brain responses elicited by a repetitive visual stimulus that is periodically intensity-modulated at a fixed rate of 6-8 Hz or higher. Employing electrophysiological recording techniques, they can be recorded at the scalp as a nearly sinusoidal oscillatory waveform having the same fundamental frequency as the driving stimulus, often
including higher harmonics (Regan, 1989). SSVEPs are able to reflect cortical responses in the time range of several seconds up to minutes and hours. Thus, they can have similar time course of measurement as have functional blood flow methods. Given these properties, SSVEPs appear to be especially suitable for monitoring processes related to fluctuations in transient states of the central nervous system such as attention. An additional advantage can be seen in the fact that the relevant parameters (i.e. amplitude and phase of the SSVEP) can be easily extracted by means of FFT analysis techniques and submitted to further analyses (Picton et al., 1987; Müller et al., 1998a, b). Because of the rapid presentation of stimuli, a high number of recording epochs can be obtained. Accordingly, better signal-to-noise ratio than in traditional ERP studies can be attained if the evoked response for each repetition is in the amplitude range of the ERP. This is usually the case for stimulation frequencies below 20 Hz (Müller et al., 1998a). Thus, SSVEPs may be better suitable for dipole modeling techniques than are some of the ‘cognitive’ ERP studies, which do not allow for obtaining sufficient numbers of trials. In terms of functional correlates, steady-state potentials in the auditory or visual modality have been shown to be sensitive for tonic subjective states such as anesthesia (Plourde and Picton, 1990), sleep (Picton et al., 1987), or vigilance (Silberstein et al., 1990). In addition, the amplitudes of SSVEPs are modulated as a function of phasic changes in the individual such as cognitive performance (Silberstein et al., 1995) or visual spatial selective attention (Morgan et al, 1996; Müller et al, 1997b, 1998b). Recent work has used the steady-state visual evoked field, recorded by means of MEG, for investigating transitions between percepts in an eye rivalry paradigm (Srinivasan et al., 1999). Besides amplitude changes, changes in SSVEP phase have been used to describe effects of experimental manipulations on timing of the SSVEP (Silberstein et al., 1995). It is not clear, however, whether this parameter can be interpreted unambiguously, because there is no absolute external reference for phase lags.
Rather, phase delays are expressed with respect to sine or cosine functions.
These techniques have to date not been used to examine affective processes. The present thesis aims at evaluating and testing the usefulness of SSVEPs in this context. It is expected that SSVEP amplitude and phase may be useful tools for the investigation into human emotional perception, both in terms of response timing and function localization.