I see what you mean: Theta power increases are involved in the retrieval of lexical semantic information
Introduction
It has been proposed repeatedly that semantic representations of words, and objects, are neurally implemented in distributed networks carrying sensory, motor and/or more abstract functional information (e.g., Damasio, 1990, Farah and McClelland, 1991, Warrington and McCarthy, 1987, Warrington and Shallice, 1984). A more specific hypothesis is that the exact topology of such networks is at least partly determined by the semantic properties of the word or object concept (Martin and Chao, 2001, Pulvermueller, 1999, Pulvermueller, 2001). This contrasts with the view put forward by others (e.g., Devlin et al., 2002, Tyler and Moss, 2001, Tyler et al., 2000) that semantic representations are stored in a distributed network in which there is no neural specificity as a function of semantic property.
According to the first view, manipulable objects such as tools are strongly linked to motor behavior, and therefore their representational networks should comprise a significant amount of neurons in motor cortex. Animals, which are most of the time (visually) perceived rather than manipulated, should be represented by networks that partly reside in visual cortex. Given anatomically confined lesions/damage to neural tissue, the topographical distinction between the representations of both categories is likely to lead to selective damage to the representations in one category. Support for this view comes from a large number of neuropsychological studies (see Forde and Humphreys, 1999, Saffran and Sholl, 1999 for reviews), that have shown the existence of category-specific semantic impairments in patients, for instance a relatively selective impairment for naming/recognizing animals as opposed to tools, or—less frequently—vice versa. These findings are at least compatible with the view that the semantic properties of words or objects partly determine the topography of their neural representation (although they can be interpreted in different ways, e.g., Tyler & Moss, 2001).
More direct support comes from a number of hemodynamic neuroimaging studies. In a PET experiment, Martin et al. (Martin, Haxby, Lalonde, Wiggs, & Ungerleider, 1995) found that generation of color words selectively activated a region in the ventral temporal lobe just anterior to the area involved in the perception of color, whereas generation of action words activated a region in the middle temporal gyrus just anterior to the area involved in the perception of motion. In another PET study (Martin, Wiggs, Ungerleider, & Haxby, 1996), these authors showed that naming pictures of tools activated premotor areas, while naming pictures of animals activated medial occipital areas. Within the language domain, a recent fMRI study (Hauk, Johnsrude, & Pulvermueller, 2004) showed that the visual presentation of verbs related to the execution of face, arm and leg movements (e.g., to lick, to pick and to kick) differentially activated areas in the motor cortex that were adjacent to or overlapping with the areas that were activated during actual movements of the corresponding body parts. Similarly, recent fMRI data (Vitali et al., 2005) show different effective connectivity patterns for viewing tools vs. animals, with tools engaging areas in frontal and premotor areas, and animals engaging visual association areas. It should be noted however that other studies failed to find such semantically specific activations during semantic processing (e.g., Noppeney & Price, 2002).
Together, the available data largely support the view that spatially distributed functional networks form the basis of semantic representations, and that the topographies of these networks correspond to the semantic properties of individual items. However, in order to answer the question of how such functional networks are formed, one has to gain insight into the temporal dynamics of their activation. Here, we present EEG data showing that oscillatory neuronal dynamics in the theta frequency range are involved in the activation of functional networks with semantically specific topography.
The temporal dynamics of neuronal activity can be investigated with neuroimaging methods that yield high temporal resolution such as EEG or MEG. While some studies have investigated the scalp topography of event-related potentials (ERPs) elicited by the presentation of different categories of words (Hauk and Pulvermueller, 2004, Pulvermueller et al., 2001), the oscillatory dynamics of EEG and MEG data presumably provide another window on the dynamic formation of functional networks. The reason for this is that an (indirect) link can be made between changes in neuronal synchrony on the one hand, and changes in the oscillatory activity that is present in scalp-recorded EEG and MEG on the other hand. The amplitude (or power) of oscillatory activity is indicative of synchronization changes because the synchronization of local groups of neurons, the activities of which are picked up by one and the same sensor, will result in larger EEG/MEG amplitudes for that sensor. It follows that local changes in synchrony of oscillatory firing patterns lead to changes in amplitude (or power) of rhythmic EEG/MEG activity at the single-trial level. The phase of EEG/MEG oscillations is important because synchronization of oscillatory firing patterns between spatially distant neuronal populations lead to increases in coherence, or in phase synchronization, between two (or more) concurrently measured EEG or MEG signals (Bastiaansen and Hagoort, 2003, Singer, 1999, Varela et al., 2001, see Pulvermueller, 1999 for arguments of how this mechanism applies to the representation of lexical semantics).
In sum, a topographical analysis of event-related changes of either power or coherence in oscillatory EEG or MEG activity during language processing tasks would be informative in establishing whether synchronous functional networks with distinct scalp topographies constitute the representation of different categories of words. A few studies have provided data that speak to this issue. In a number of studies aimed at characterizing EEG coherence changes elicited by different word categories, Weiss and colleagues (see Weiss & Mueller, 2003 for review) found different coherence patterns in the lower-beta frequency range (roughly 13–18 Hz) between concrete and abstract nouns, and between concrete nouns and verbs. However, the authors did not attempt to relate coherence topography to semantic properties in these studies. A study by Pulvermueller (Pulvermueller, Lutzenberger, & Preissl, 1999) did attempt to relate semantic properties to the scalp topography of EEG power changes. In a lexical decision task, nouns were contrasted with verbs, assuming that nouns are perception-related and should therefore have a relatively larger neuronal representations in occipital cortex, while verbs are action-related and should have representational network elements in motor cortex. The authors found power decreases in the gamma frequency range compared to baseline, that were smaller for verbs than for nouns at central electrodes, but smaller for nouns than for verbs over occipital electrodes. These data were interpreted to reflect different network topographies that followed the semantic properties of the stimuli.
Other studies have related semantic memory operations to power changes in the alpha frequency band (for review, see Klimesch, 1999). It should be noted however, that in most of these studies, the term semantic memory is used in the sense of declarative (as opposed to episodic) memory. In language comprehension theories however, the term semantic is used in a more narrow sense, referring to the meaning aspect of words (as opposed to syntax, phonology etc.). In the remainder of this paper, we will use the term semantic to refer to the more narrow, psycholinguistically defined process of attributing meaning to linguistic material. However, some studies do address semantic processing in this more narrow sense (e.g., Klimesch et al., 1997a, Klimesch et al., 1997b, Rohm et al., 2001). Generally, the results of these studies support the notion that semantic processing is indeed related to power decreases in the upper alpha band (roughly 10–12 Hz).
We further pursue the issue of whether words with different semantic properties are neurally represented in topographically distinct functional networks, by taking a different approach in constructing stimulus material. Rather than relying on a correlation between semantic features and specific word classes, we constructed two sets of nouns, one referring exclusively to colors and shapes, and the other referring exclusively to sounds. Thus, one set of stimuli only contained nouns with visual semantic properties, while the other set only contained nouns with auditory semantic properties. While their EEG was measured, subjects performed a visual lexical decision task, in which we contrasted the ‘visual’ nouns with the ‘auditory’ nouns (note that both were presented to the subjects in the visual modality). Given our hypothesis, the former set of nouns should activate a network with a relatively large component in the visual cortex, whereas the latter set should lead to the emergence of a network that extends into auditory cortex. Because event-related changes in power reflect changes in local synchrony, while event-related changes in phase coherence between different sensors reflect changes in long-range synchrony (Bastiaansen and Hagoort, 2003, Varela et al., 2001) we expected these local differences in network topography to be expressed as differential power increases over auditory and visual areas. More specifically, we expected EEG power to be larger over the auditory cortex following words with auditory semantic properties compared to words with visual semantic properties, while the opposite pattern was expected over the visual cortex. On the basis of the existing literature on oscillatory EEG/MEG dynamics during language processing, we hypothesized this differential power increase to be manifest in one (or more) of the following frequency bands: theta (4–7 Hz; based on Bastiaansen et al., 2002b, Bastiaansen et al., 2002c, Hagoort et al., 2004), alpha (8–12 Hz; based on the studies reviewed in Klimesch, 1999), lower-beta (13–18 Hz; based on the studies reviewed in Weiss & Mueller, 2003) or gamma (above 30 Hz; based on Hagoort et al., 2004, Pulvermueller et al., 1996, Pulvermueller et al., 1999).
Section snippets
Subjects
Sixteen students (4 male) from the Radboud University Nijmegen (aged 21–36 years), were paid for their participation after having given informed consent in writing, according to the Declaration of Helsinki. All were native Dutch speakers with normal or corrected-to-normal vision. They were right-handed, as assessed by a self-report. None of the participants had any neurological impairment, neurological trauma, and none used neuroleptics.
Stimulus material
The item set that is critical for the present study
ERP data
Fig. 1 shows the time-course of the ERPs elicited by the AUD and VIS words at representative channels.
A negative component is observed, with a (grand-average) peak at 400 ms after word onset. We identify this component as the classical N400. This is confirmed by the postcentral topographical distribution (Fig. 2a). A slight difference in scalp topography appears to be present between the two conditions, with the VIS words eliciting slightly more negativity at frontal channels than the AUD words
Discussion
We investigated event-related potentials (ERPs) and oscillatory brain dynamics of the EEG while subjects performed a visual lexical decision task. In agreement with our hypotheses, we found power increases in the theta band (4–7 Hz) and in the lower beta range (13–18 Hz). We also found the expected power decrease in the alpha (8–12 Hz) frequency range. However, in contrast to previous studies no evidence was found for power in- or decreases in the gamma frequency band.
The most striking finding of
References (59)
- et al.
Sub-second “temporal attention” modulates alpha rhythms. A high-resolution EEG study
Brain Research. Cognitive Brain Research
(2004) - et al.
Event-induced theta responses as a window on the dynamics of memory
Cortex
(2003) - et al.
Event-related alpha and theta responses in a visuo-spatial working memory task
Clinical Neurophysiology
(2002) - et al.
Event-related theta power increases in the human EEG during online sentence processing
Neuroscience Letters
(2002) - et al.
Syntactic processing modulates the theta rhythm of the human EEG
Neuroimage
(2002) - et al.
Functional connectivity of gamma EEG activity is modulated at low frequency during conscious recollection
International Journal of Psychophysiology
(2002) Category-related recognition defects as a clue to the neural substrates of knowledge
Trends in Neurosciences
(1990)- et al.
Is there an anatomical basis for category-specificity? Semantic memory studies in PET and fMRI
Neuropsychologia
(2002) - et al.
EEG theta and gamma responses to semantic violations in online sentence processing
Brain and Language
(2006) - et al.
Somatotopic representation of action words in human motor and premotor cortex
Neuron
(2004)
EEG alpha and theta oscillations reflect cognitive and memory performance: A review and analysis
Brain Research Reviews
Brain oscillations and human memory: EEG correlates in the upper alpha and theta band
Neuroscience Letters
Event-related desynchronization in the alpha band and the processing of semantic information
Brain Research. Cognitive Brain Research
Episodic retrieval is reflected by a process-specific increase in human electroencephalographic theta activity
Neuroscience Letters
Semantic memory and the brain: Structure and processes
Current Opinion in Neurobiology
Analysis of dynamic brain imaging data
Biophysics Journal
Retrieval of visual, auditory, and abstract semantics
Neuroimage
Event-related cortical desynchronization detected by power measurements of scalp EEG
Electroencephalography and Clinical Neurophysiology
Evaluation of event-related desynchronization (ERD) preceding and following voluntary self-paced movements
Electroencephalography and Clinical Neurophysiology
Patterns of cortical activation during planning of voluntary movement
Electroencephalography and Clinical Neurophysiology
The role of theta and alpha oscillations for language comprehension in the human electroencephalogram
Neuroscience Letters
Spatio-temporal frequency characteristics of intersensory components in audiovisually evoked potentials
Brain Research. Cognitive Brain Research
Neuronal synchrony: A versatile code for the definition of relations?
Neuron
Separable effects of priming and imageability on word processing: An ERP study
Brain Research. Cognitive Brain Research
Towards a distributed account of conceptual knowledge
Trends in Cognitive Science
Conceptual structure and the structure of concepts: A distributed account of category-specific deficits
Brain and Language
Generating animal and tool names: An fMRI study of effective connectivity
Brain and Language
The contribution of EEG coherence to the investigation of language
Brain and Language
The CELEX lexical database (CD-ROM)
Cited by (149)
Age and vocabulary knowledge differentially influence the N400 and theta responses during semantic retrieval
2023, Developmental Cognitive NeuroscienceTemporal and topographical changes in theta power between middle childhood and adolescence during sentence comprehension
2022, Developmental Cognitive Neuroscience