Modulation of induced gamma band activity in the human EEG by attention and visual information processing
Introduction
Oscillatory cortical activities in the human brain above 20 Hz, i.e. in the gamma band, do not form a homogeneous class of responses. Experimental evidence demonstrates diversity with respect to both perceptual and neurophysiological mechanisms. In 1992, Galambos introduced the following classification of EEG responses with power in the gamma band (Galambos, 1992):
- 1.
Spontaneous gamma rhythms. Such rhythms appear without obvious relation to an external event.
- 2.
Evoked gamma band responses. These are event-related responses elicited by and precisely time-locked to an external stimulus. In expansion to the classification by Galambos, one should further divide this category into (a) transient evoked gamma band responses and (b) driven responses like the steady-state auditory or visual evoked response.
- 3.
Emitted gamma band oscillations. Gamma band activity time-locked to a stimulus that has been omitted. Conceptually, but not technically, these may be subsumed under category 2.
- 4.
Induced gamma band rhythms are initiated by an event but are not time- and phase-locked to the eliciting stimulus. This category in itself encompasses a variety of different processes.
In accordance with the topic of this volume, we focus on induced gamma band responses in the human brain. A series of hypotheses seeks to attribute meaning to oscillating cell assemblies. One of the earlier, empirically based hypothesis was put forward by Walter Freeman. According to Freeman (1975, see also Freeman, 1996, Freeman and van Dijk, 1987, Freeman and di Prisco, 1986) the stimulus itself is coded in oscillatory patterns or, more precisely, as a dynamic state of a non-linear system that appears to the observer as oscillations. On the basis of intracortical recordings from the olfactory bulb of the rabbit, Freeman and di Prisco (1986) concluded that after the presentation of a learned odor, the system switches from a spatially and temporally unpatterned chaotic state to a global odor-specific state which is characterized by a single near-limit attractor. In other words, the attractor governing the dynamic pattern is related to stimulus encoding, everything that is not coherent to this signal is dismissed as noise.
The most popular hypothesis that in fact could be a consequence of Freeman's model, the temporal binding hypothesis, predicts that synchronized oscillatory neural activity is the mechanism by which various brain regions form one percept and are, therefore, the key mechanism for feature binding (Abeles, 1982, Eckhorn et al., 1990, Gray et al., 1989, Gray et al., 1990, Gray and Singer, 1987, Malsburg and Schneider, 1986, Milner, 1974, Singer and Gray, 1995). Thus, oscillations in the gamma band range were considered to be functionally different from oscillations in the alpha range. It was hypothesized that alpha band oscillations reflect idling in neural mass systems (Hari and Salmelin, 1997, Pfurtscheller, 1992, Pfurtscheller and Aranibar, 1979, Pfurtscheller et al., 1993, Pfurtscheller and Klimesch, 1992). This idling state would allow the system to start more rapidly than by a ‘cold start’ (Hari and Salmelin, 1997).
In the last decade, most of the studies that investigated induced gamma band responses were conducted in animal research (see Singer and Gray, 1995 for an overview and Engel et al. this volume). One of the early animal experiments has demonstrated that long-range synchronization in area 17 of the anaesthetized cat reflects global stimulus properties (Gray and Singer, 1989). In this experiment, multiunit activity was recorded from two sites which preferred vertical orientations and were separated by 7 mm. The corresponding receptive fields were non-overlapping and co-linearly arranged. This arrangement allowed three different stimulus conditions: (a) a long continuous light bar moving across both fields; (b) two independent light bars moving in the same direction; and (c) the same two bars moving in opposite directions. The results showed that the induced gamma band responses of the two receptive fields were synchronized when the long bar was presented. In the case of the two independent bars moving in the same direction, the synchronization across the two receptive fields became weaker and totally disappeared if the motion of the stimuli was incoherent. These results gave raise to the notion that induced gamma band activity is related to the features of the stimuli and thus were considered to support the temporal binding hypothesis.
In one series of experiments on induced gamma band activity in the human EEG (Müller et al., 1996), we aimed at mimicking the work on animals, e.g. the Gray et al. (1989) study, as closely as possible. We used the same stimulus configuration consisting of a long coherently moving bar and two small bars moving in opposite directions. We hypothesized that if synchronized neural activity is related to cortical object representation, we should see enhanced power in the gamma band when subjects processed the long bar. In the condition where two incoherently moving bars were presented, we expected a marked reduction in gamma band power as compared to the coherently moving long bar since — on the basis of a simplified model — two cell assemblies with no phase coherence should code the two bars. In the macroscopic EEG recording, a reduction in gamma power would be the consequence. In accordance with our hypothesis, we observed significantly increased gamma band power on posterior electrode sites when subjects attended the long bar as compared to the condition during which subjects attended to the two bars moving in opposite directions. This finding was replicated in a subsequent follow-up study (Müller et al., 1997a, Müller et al., 1997b).
There are now a series of reports that confirm a modification of human EEG gamma band power as a function of stimulus properties. Lutzenberger et al. (1995) found increased gamma band power when lines in a visual quarter field moved coherently, giving the impression of a waterfall, as compared to a condition during which the lines were moving randomly. In a series of experiments Tallon-Baudry and colleagues reported a link between stimulus features and the representation of a stimulus and induced gamma band activity (Tallon et al., 1995, Tallon-Baudry and Bertrand, 1999, Tallon-Baudry et al., 1996, Tallon-Baudry et al., 1997a, Tallon-Baudry et al., 1997b). In addition, these authors have pointed out a possible link between gamma band activity and memory processes (Tallon-Baudry et al., 1998). The experiments by the Lyon group are fully described elsewhere in this volume, we refer the reader to that chapter. In using illusionary triangles and squares, Herrmann et al. (1999) have replicated the findings of Tallon-Baudry and colleagues but only with respect to evoked gamma band responses. The work is also fully described elsewhere in this volume.
Recently, Rodriguez et al. (1999) have shown an increase in gamma band power when subjects were confronted with so-called Mooney faces (black/white shapes of faces) as compared to when these figures were presented upside-down, which did not allow subjects to identify a face. In addition to the studies mentioned so far, induced gamma band responses in humans have been reported in the auditory cortex (Jokeit and Makeig, 1994, Tallon-Baudry and Bertrand, 1999), in the sensorimotor cortex (Kristeva-Feige et al., 1993), during the processing of words in contrast to pseudo words (Eulitz et al., 1996, Lutzenberger et al., 1994, Pulvermüller et al., 1996, Pulvermüller et al., 1995) and in a classical conditioning paradigm (Miltner et al., 1999).
In the following, we present a series of four experimental studies. Three of them (studies 2, 3 and 4) have been published previously (Gruber et al., 1999, Keil et al., 1999, Müller et al., 1999).
The first two experiments were conducted to demonstrate and validate a link between the modulation of induced gamma band activity and spatial selective visual attention. In previous research, it has been suggested that attended information within the visual field, or — to express it with a widely used metaphor — information which falls within the attentional ‘spotlight’ elicits larger sensory-evoked responses as compared to the information outside the ‘spotlight’ (Hillyard and Anllo-Vento, 1998, LaBerge, 1995, Mangun, 1995, Posner and Petersen, 1990, Posner and Dehaene, 1994). Such an amplification mechanism has been demonstrated in EEG recordings by showing an amplitude augmentation of particular components of the visual evoked potential (VEP) when a stimulus was attended as compared to when the stimulus was ignored (Anllo-Vento and Hillyard, 1996, Hillyard and Anllo-Vento, 1998, Hillyard et al., 1998, Luck and Ford, 1998, Mangun, 1995). It has been validated using positron emission tomography (PET); blood flow was increased in those regions of the visual cortex that were related to the sensory processing of an attended stimulus (Corbetta et al., 1993, Corbetta et al., 1995, Heinze et al., 1994, Mangun et al., 1997). Recently, these findings were confirmed by means of functional magnetic resonance imaging (fMRI; Beauchamp et al., 1997, Haug et al., 1998, Martinez et al., 1999). In sum, experimental results support the idea that visual information in attended sensory pathways triggers an amplification, i.e. stronger neuronal responses as compared to unattended pathways, resulting in a larger signal. Consequently, if induced gamma band activity is related to visual information processing, EEG spectral gamma power should be enhanced when subjects attend to a certain stimulus as compared to when subjects ignore that stimulus. The experimental outcome presented in Section 2 confirms this hypothesis.
The next study demonstrates a link between the perception of a Gestalt and induced gamma band activity. A traditional approach used to study visual perception exploits the properties of ambiguous (bistable) figures (Attneave, 1971, Kanizsa and Luccio, 1995). We used an ambiguous figure that, when rotated, biased the subjects’ perception to that of either a sad or happy face. On the basis of the law of good figure (gute Gestalt), we assumed that induced gamma band responses will be associated with the perception of the sad and happy face rather than the continuous visual input.
In the last experimental study presented here, we investigated whether processing emotional pictures modulates gamma band activity differentially depending on the valence of the stimulus. Theoretically, affect and emotion in humans have been related to the integrated activity of brain circuits including structures such as the amygdala, hippocampus or anterior cingulate in addition to the neocortex particularly sensory representational zones (Davidson and Hugdahl, 1995, Derryberry and Tucker, 1992, LeDoux, 1995b, Liotti and Tucker, 1995, Tucker and Dawson, 1984). In particular, the prefrontal cortex has been suggested to be an important part of such a widespread affective network (Damasio, 1995, Davidson, 1992, Davidson et al., 1990). Based on the findings reported so far, it seems reasonable to predict that neural activity subserving affective processing might use high-frequency oscillatory coding in the relevant circuits as a means to integrate the activity of distributed structures.
Section snippets
Attentional modulation of induced gamma band activity
The following two studies were conducted to investigate whether EEG induced gamma band activity is modulated by visual selective spatial attention. As mentioned above, a consistent finding is an amplitude augmentation of certain components of the VEP when a stimulus is attended. This is seen as an attentional cortical facilitation in visual areas corresponding to the processing of the stimulus (Hillyard and Anllo-Vento, 1998, Hillyard et al., 1998). This facilitation mechanism should also apply
General discussion
In the present paper we presented four studies which aimed at further demonstrating that induced gamma band activity in the human brain is linked to visual perceptual mechanisms and information processing. First, we showed that induced gamma band activity is modulated by spatial selective visual attention. The increase in spectral power when a subject attended to a certain stimulus as compared to when the same stimulus was ignored was replicated in the follow-up study. Thus, the results are in
Acknowledgements
We would like to thank Eva Bonna, Ursula Lommen, Heidi Messmer, Klaus Lang and Jürgen Wolf for help in data acquisition and Lisa Green for editorial support. We also thank Prof. Peter Lang, University of Florida, for help in selecting the picture set of study 4 and Prof. Thomas Elbert, University of Konstanz, for constructive discussions. Research was supported by grants from the Deutsche Forschungsgemeinschaft and the Human Frontier Science Program.
References (108)
- et al.
Non-linear dynamic complexity of the human EEG during evoked emotions
Int. J. Psychophysiol.
(1998) - et al.
Artifactually high coherences result from using spherical spline computation of scalp current density
Electroencephalogr. Clin. Neurophysiol.
(1991) Emotion in the perspective of an integrated nervous system
Brain Res. Rev.
(1998)Anterior cerebral asymmetry and the nature of emotion
Brain Cogn.
(1992)- et al.
Emulation of somatosensory evoked potential (SEP) components with the 3-shell head model and the problem of ‘ghost potential fields’ when using an average reference in brain mapping
Electroencephalogr. Clin. Neurophysiol.
(1990) 40-Hz asymmetry during recall of emotional events
Int. J. Psychophsiol.
(1989)- et al.
Oscillatory neuromagnetic activity induced by language and non-language stimuli
Cogn. Brain Res.
(1996) Feedback models of gamma rhythms
TINS
(1996)- et al.
Spatial patterns of visual cortical fast EEG during conditioned reflex in a rhesus monkey
Brain Res.
(1987) - et al.
Selective visual-spatial attention alters induced gamma band responses in the human EEG
Clin. Neurophysiol.
(1999)