Oscillatory MEG gamma band activity dissociates perceptual and conceptual aspects of visual object processing: A combined repetition/conceptual priming study

Abstract

We used a combined repetition/conceptual priming task to investigate attenuations of induced gamma-band activity (iGBA) due to prior experience. We hypothesized that distinguishable iGBA suppression effects can be related to the processing of (a) perceptual aspects, and (b) conceptual aspects of cortical object representations. Participants were asked to perform a semantic classification task with pictures of real world objects and their semantically corresponding words, using a design that isolated distinct levels of the neural suppression effect. By means of volumetric source analysis we located stimulus domain-specific iGBA repetition suppression effects (60–90 Hz) in temporal, parietal, and occipital areas of the human cortex. In contrast, domain-unspecific iGBA repetition suppression, corresponding to conceptual priming, was restricted to left temporal brain regions. We propose that the selective involvement of left temporal areas points to the activation of conceptual representations, whereas more posterior temporal, parietal, and occipital areas probably reflect perceptual aspects of higher-order visual object processing.

Introduction

The repeated processing of stimuli commonly leads to faster and more accurate responses in a variety of tasks (repetition priming), even if the repetition itself is task-irrelevant. With respect to the underlying brain mechanisms, neuroimaging studies using positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) have associated repetition priming predominantly with decreased brain activity following stimulus processing. Consequently, this phenomenon has been termed repetition suppression (for reviews see Henson, 2003, Grill-Spector et al., 2006, Schacter et al., 2007). According to Desimone (1996) repetition suppression might represent a by-product of the “sharpening” of stimulus representations in the cortex. In this view, objects can be cortically processed by sparser neuronal assemblies following repeated occurrences of a given stimulus which in turn leads to decreased activation as seen in studies using PET and fMRI (Wiggs and Martin, 1998). Further empirical evidence supporting the sharpening hypothesis originates from electrophysiological studies using electroencephalography (EEG) and magnetoencephalography (MEG). Even though repetition-dependent changes are already evident in the amplitude reduction of the averaged scalp-recorded potentials, i.e. ERPs (e.g., Race et al., 2010, Voss et al., 2010), the analysis of oscillatory brain activity in the gamma-band frequency range (> 30 Hz) seems to be particularly suited to unravel the nature of cortical object processing in greater detail.

Resting on a broad scope of electrophysiological evidence, including data from animal and human patient studies using intracranially implanted electrodes, transient induced gamma band activity (iGBA) is assumed to reflect the synchronized oscillatory activity of distributed neuronal populations engaged in the representation of objects (Tallon-Baudry and Bertrand, 1999, Varela et al., 2001, Engel et al., 2001, Jensen et al., 2007). Several studies have shown that scalp-recorded power and phase synchronization of iGBA decreases for repeatedly presented pictures of objects or words (Gruber and Müller, 2002, Gruber and Müller, 2005, Fiebach et al., 2005, Conrad et al., 2007), but the precise origins of these attenuation effects are still unknown. In particular, it is unclear to what extent iGBA mirrors perceptual and/or conceptual aspects of visual object processing. Most studies employed repetition priming paradigms such that physically identical stimuli were presented as prime and target stimuli. It is therefore not possible to conclude whether iGBA attenuation on repeated occurrences of the same stimulus results only from facilitated processing of perceptual stimulus features, or if conceptual processing also plays a role. Some evidence that gamma band activity is not merely reflecting low-level perceptual features comes from studies on crossmodal sensory processing (Schneider et al., 2008, Schneider et al., 2011). For instance, studying the integration of visual and auditory information Schneider et al. (2008) presented pictures of objects followed by congruent or incongruent environmental targets sounds (e.g., the picture of a sheep and the sound produced by sheep or by a bell). Induced GBA was reported to be larger if the bimodal information was congruent, suggesting that gamma band oscillations are related to the formation of multisensory conceptual representations. Further indications for an involvement of gamma band oscillations at higher processing levels come from a word repetition priming experiment using words and word homophones (Matsumoto and Iidaka, 2008). An iGBA suppression effect was found if a prime word (e.g., “pair”) was followed by a homophone target word (e.g., “pear”), indicating a role of gamma band oscillations at the phonological level above the orthographical representation.

Recently, a compelling study showed that scalp-recorded EEG signals can be contaminated with ocular artifacts (miniature eye movements) that in particular might impact the interpretation of iGBA (Yuval-Greenberg et al., 2008). In detail, the authors presented evidence that iGBA, commonly found in scalp-recorded EEG approximately 200–300 ms following stimulus onset, is time-locked to – and therefore incorporates – the so called saccadic spike potential. Up to now, different algorithms have been proposed to detect and remove saccadic spike potentials from the EEG (Keren et al., 2010, Hassler et al., 2011), but the validity of these methods has yet to be established. In our study we recorded MEG which represents a reference-free recording technique and therefore should not be susceptible to the broad and distant spatial transmission of frontal eye movement artifacts due to referencing (see Discussion). In addition, we analyzed iGBA in source space, and none of the signals' sources in the gamma frequency band was located at frontal brain sites or around the orbital cavities of the eyes. Furthermore, we investigated two time windows, i.e. 200–400 ms and 400–700 ms after stimulus onset. The 400–700 ms time interval is set notably later than the reported critical microsaccadic iGBA effects, and it is located within the ongoing iGBA that can often be observed until the offset of a stimulus (Gruber and Müller, 2002, Fries et al., 2008, Haenschel et al., 2009, Hassler et al., 2011).

In the current study, we employed a combined repetition and conceptual priming paradigm to dissociate iGBA sharpening effects directly between perceptual and conceptual levels. The stimuli consisted of pictures of real world objects and their semantically matched words, so that within a given prime-target stimulus-pair a word could be used as a semantic prime for a subsequent picture (target) and vice versa. In the present experiment, participants were instructed to classify each stimulus, i.e. pictures or their semantically matched words to be either “natural” or “man-made”. Stimuli were repeated either in the same domain as during the initial presentation (picture–picture or word–word) or in a different domain (picture–word or word–picture). Our main objective was to demonstrate repetition-dependent iGBA attenuation during visual object processing in response to prime-target stimulus pairs which were identical at the conceptual level but different at the perceptual level. This supposition follows directly from the sharpening approach but – to the best of our knowledge – has not been addressed yet. More precisely, we expected to identify iGBA reductions after repeating the same physical stimulus (e.g., picture follows picture), as well as after repeating the semantic information by using a cross-domain switch in stimulus presentation (picture follows semantically corresponding word). Inferring from priming studies with fMRI acquisitions, we assumed that within-domain suppression effects should reside in cortical areas associated mostly with perceptual processing (Wagner et al., 2000, Henson and Rugg, 2003). In contrast, cross-domain suppression effects should be restricted primarily to areas involved in conceptual processing. Schacter et al. (2007) proposed that, for visual stimuli, perceptual cortices show a gradient of stimulus specificity with greater specificity occurring in posterior regions compared to more anterior sites. Importantly, regions of the lateral temporal cortex were reported to respond to conceptual components of repetition priming across different stimulus domains. Therefore, if prior experience-dependent iGBA reductions reflect perceptual as well as conceptual aspects of priming, we expect that these distinct levels of visual object processing are associated with neural suppression in dissociable regions of mainly temporal, parietal, and occipital cortex. Specifically, we hypothesized that conceptual priming, i.e. cross-domain iGBA repetition suppression, would be associated with temporal cortex areas. To this end, we combined Morlet wavelet analysis of oscillatory brain activity (Bertrand and Pantev, 1994) with a volumetric source localization approach (Variable Resolution Electromagnetic Tomography, VARETA; Bosch-Bayard et al., 2001) to characterize the underlying brain sources of gamma band activity.