Accepted Manuscript Theta response in schizophrenia is indifferent to perceptual illusion Birgit Mathes, Christina Schmiedt-Fehr, Shwetha Kedilaya, Daniel Strüber, Andreas Brand, Canan Basar-Eroglu PII: DOI: Reference:
S1388-2457(15)00541-6 http://dx.doi.org/10.1016/j.clinph.2015.02.061 CLINPH 2007488
To appear in:
Clinical Neurophysiology
Accepted Date:
3 February 2015
Please cite this article as: Mathes, B., Schmiedt-Fehr, C., Kedilaya, S., Strüber, D., Brand, A., Basar-Eroglu, C., Theta response in schizophrenia is indifferent to perceptual illusion, Clinical Neurophysiology (2015), doi: http:// dx.doi.org/10.1016/j.clinph.2015.02.061
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Theta response in schizophrenia is indifferent to perceptual illusion
Birgit Mathes (1,2)*, Christina Schmiedt-Fehr (1,2), Shwetha Kedilaya (1,2), Daniel Strüber (3,4), Andreas Brand (1,2), Canan Basar-Eroglu (1,2)
(1) University of Bremen, Institute of Psychology and Cognition Research, Bremen, Germany (2) Centre for Cognitive Science, Bremen, Germany (3) Experimental Psychology Lab, Dpt. of Psychology, European Medical School, University of Oldenburg (4) Research Center Neurosensory Science, University of Oldenburg
* Corresponding author: Dr. Birgit Mathes University of Bremen, Institute of Psychology and Cognition Research, Grazer Str. 4, 28359 Bremen, Germany Tel.: +49-421-218 68707 Fax: +49-421-218 68719 E-mail: [email protected] URL: http://www.ipk.uni-bremen.de/en/start.html
Highlights -
Multistable perception allows examining the patient’s perceptual deficits while the stimulus remains unchanged.
-
This EEG study complements behavioural studies indicating that top-down driven organising principles of perception in schizophrenia is weak
-
The altered oscillatory theta response reflects impairments in the spatio-temporal integration of the neural information transfer in schizophrenia.
Abstract Objective: Patients with schizophrenia are impaired in maintaining coherent perceptual experiences. This is reflected in the oscillatory theta response and can be investigated by visual illusions. Ambiguous stimuli elicit illusory perceptual switches while the stimulus remains unchanged. Methods: Theta responses elicited by an ambiguous and unambiguous control stimulus were measured using the EEG during time periods of perceptual switching and perceptual stability (non-switching). Results: For the ambiguous task, theta activity increased during perceptual switching in healthy controls only. For the unambiguous task, the switching-related increase of theta activity was larger in controls than in patients. This reduced modulation of the theta response seems not to be related to a general decrease of theta activity in patients. Conclusions: These findings may be related to disturbances in the spatio-temporal integration of neural activity in patients. Reporting ambiguous and unambiguous perceptual switches seems to be more demanding for patients with schizophrenia than healthy controls. Significance: This is one of the first studies on the neurophysiologic correlates of illusory perception in schizophrenia. Focussing on the relation between different brain states (such as switching and non-switching) might integrate different findings about altered theta oscillations in schizophrenia.
Keywords: Brain oscillations; long-range coordination; top-down bottom-up; multistable perception; object perception.
2
1. Introduction Patients with schizophrenia are known to have problems in integrating sensory information into coherent perceptual experiences. Core symptoms of the illness, such as hallucinations, delusions and thought disorder, might be related to this impairment (Haesebaert et al., 2013; Hugdahl, 2009; Silverstein & Keane, 2011). 1.1 Top-down processes in schizophrenia investigated by visual illusions Object perception is driven by sensory information (bottom-up) as well expectations and prior experiences (top-down, Gregory, 1997; Güntekin & Basar, 2014; Long & Toppino, 2004). The importance and power of top-down processes in shaping coherent perception is revealed by visual illusions (Basar-Eroglu, Strüber, Stadler, Kruse, & Basar, 1993; Sterzer, Kleinschmidt, & Rees, 2009). Visual illusions arise by systematic discrepancies between a stimulus and the percept due to neural processes transforming sensory information into a perceptual experience (Gregory, 1997). Patients with schizophrenia seem to be less prone to visual illusions (for a review see Notredame, Pins, Deneve, & Jardri, 2014). From distance, healthy persons perceive a hollow mask as a proper face (Gregory, 1997) while patients with schizophrenia perceive the physically correct inverted face (Dima, Dietrich, Dillo, & Emrich, 2010; Dima et al., 2009). Their perception seems to be driven more strongly by the sensory information (i.e. depth cues) than by top-down driven organising principles (i.e., the implausibility of a person with an inward turned nose; Dima, et al., 2010; Dima, et al., 2009). Ambiguous stimuli elicit a special case of visual illusions. The sensory information of an ambiguous figure is inconclusive regarding the generation of coherent perception, because more than one meaningful solution is similarly probable to account for the current percept. During continuous observation, perception switches between all possible perceptual
3
alternatives although the physical properties remain unchanged (Long & Toppino, 2004, see Figure 1 for some examples). The individual speed of perceptual switching (the reversal rate) is variable (Strüber, BasarEroglu, Hoff, & Stadler, 2000; Strüber, Basar-Eroglu, Miener, & Stadler, 2001). It remains unclear whether the spontaneous reversal rate in schizophrenia differs from healthy persons. In the few studies available the findings range from acceleration to slowing of perceptual switching (Calvert et al., 1988; Eysenck, 1952; Keil, Elbert, Rockstroh, & Ray, 1998; Philip, 1953; Tschacher, Dubouloz, Meier, & Junghan, 2008). More consistent are reports about the lack of top-down control on the current percept of an ambiguous figure. The Schröder staircase has a semantic and, therefore, top-down processed bias, because perceiving the staircase from above is more realistic than perceiving it from below (see Strüber & Stadler, 1999 and Figure 1 for ambiguous figures with different semantic biases). Patients with schizophrenia perceive the more realistic perspective of the Schröder staircase for a shorter amount of time than healthy persons (Calvert, et al., 1988; Keil, et al., 1998). These studies indicate that semantic properties of an ambiguous stimulus have lower impact on the perception of patients than of healthy persons. The efficiency in maintaining a desired perceptual alternative or initiating a perceptual switch by voluntary control is also diminished (Eysenck, 1952; McBain, Norton, Kim, & Chen, 2011). Thus, behavioural studies indicate that the influence of top-down processes on the perception of ambiguous figures is diminished in schizophrenia. This is in accordance with the broader scope of the patient’s deficiencies in using contextual information during object perception (Silverstein & Keane, 2011) and implies that in schizophrenia top-down driven organising principles of perception are weak (John & Hemsley, 1992; Silverstein & Keane, 2011; Tsunoda et al., 2012).
4
To the best of our knowledge, this and our recent study (Basar-Eroglu, Mathes, Khalaidovski, Brand, & Schmiedt-Fehr, 2014) are the first that have investigated the neural processes underlying the perception of ambiguous figures in schizophrenia. This approach allows the investigation of top-down driven organising principles of visual perception in schizophrenia as they become apparent without changes of the external surrounding. ------------------- INSERT FIGURE 1 about here ----------------------------
1.2 Top-down processes investigated by theta oscillations Event-related theta oscillations (approx 4 – 7 Hz) have a frontal maximum (Basar-Eroglu & Demiralp, 2001) and play an important role for top-down regulated processes, such as focussed attention (Basar-Eroglu & Demiralp, 2001; Cahn, Delorme, & Polich, 2013; Polich, 2007), control mechanisms in working memory (Sauseng, Griesmayr, Freunberger, & Klimesch, 2010) or executive functions (Huster, Enriquez-Geppert, Lavallee, Falkenstein, & Herrmann, 2013; Schmiedt-Fehr & Basar-Eroglu, 2011; Schmiedt-Fehr, Dühl, & Basar-Eroglu, 2011; Yordanova, Falkenstein, Hohnsbein, & Kolev, 2004). Fronto-posterior phase-coupling of theta activity seems to be specifically important for top-down modulations of early visual areas during perceptual grouping (Volberg, Wutz, & Greenlee, 2013). We have recently shown that theta activity elicited by ambiguous figures is more pronounced at anterior in comparison to posterior sites than theta activity elicited by unambiguous figures. This may be indicative for an increased involvement of top-down processes during the perception of ambiguous figures (Mathes, Khalaidovski, Schmiedt-Fehr, & Basar-Eroglu, 2014). Theta oscillations are further known to increase during perceptual switches elicited by both, ambiguous and unambiguous figures (Ehm, Bach, & Kornmeier, 2011; Nakatani & van
5
Leeuwen, 2005). Thus, theta oscillations qualify as a neural marker to investigate possible impairments in top-down driven organising principles of visual perception in patients with schizophrenia. 1.3 Theta oscillations in schizophrenia Theta oscillations in schizophrenia are still not well understood. The highest consensus is that slow wave activity, including theta, in patients is enhanced during rest (Boutros et al., 2008). Task-related theta activity of patients may be reduced (A. T. Bates, Kiehl, Laurens, & Liddle, 2009; Doege, Jansen, Mallikarjun, Liddle, & Liddle, 2010; Doege et al., 2010; Haenschel et al., 2009; Kaser et al., 2013; Ramos-Loyo, Gonzalez-Garrido, Sanchez-Loyo, Medina, & Basar-Eroglu, 2009; Schmiedt, Brand, Hildebrandt, & Basar-Eroglu, 2005) or enhanced (Fehr et al., 2003; L.E. Hong, Moran, Du, O'Donnell, & Summerfelt, 2012; L. E. Hong, Summerfelt, Mitchell, O'Donnell, & Thaker, 2012; Missonnier et al., 2012). Reduced theta activity may be restricted to those recording sites that are involved in task processing in healthy persons, while other recording sites rather show increased activation (BasarEroglu, Schmiedt-Fehr, Marbach, Brand, & Mathes, 2008). The reduction of theta activity may not necessarily be a result of lower amplitudes, rather of a reduced consistency in timing the brain response with respect to a cognitive event (Basar-Eroglu, Schmiedt-Fehr, Mathes, Zimmermann, & Brand, 2009). Results like this indicate abnormal temporal integration and interregional connectivity of brain networks as a core disturbance in schizophrenia (for reviews see Basar, 2013; Basar & Güntekin, 2013; Ford, Krystal, & Mathalon, 2007; Friston & Frith, 1995; Uhlhaas, 2013). These disturbances might affect the entire perception-action cycle (Fuster, 2006). Thus, high theta oscillations during rest might be already indicative of a change in the spatial and temporal integration of neural activity in patients, possibly indicating that keeping the
6
perceptual experience stable and coherent is a great effort for patients already in the absence of specific cognitive demands. Changes in the amplitude (irrespective of an expected increase or decrease), changes in the topographical distribution and diminished consistency in response timing may all be indicators that theta activity cannot be effectively integrated in processing cognitive demands. The reported divergence of the results about increased or decreased theta oscillations in schizophrenia further implies the importance to learn more about the transitions between brain states. The general procedure of taking one brain state (e.g., the spontaneous EEG or the pre-stimulus activity) as a baseline to investigate another brain state (e.g., induced by a cognitive demand) has clear limitations if the baseline is already different between patients with schizophrenia and healthy persons. Comparing different perceptual task conditions with or without base-line correction might lead to different results already in healthy populations (Kaiser, Rahm, & Lutzenberger, 2008). Pioneering work in electrophysiological research has long stressed that spontaneous or pre-stimulus activity of the brain systematically impacts on task-related activity (Basar, Rahn, Demiralp, & Schürmann, 1998; Buzsaki, 2006; Klimesch, 1999). More recent studies specifically illustrate this link for alpha and theta oscillations (M. M. Doppelmayr, Klimesch, Pachinger, & Ripper, 1998; Hanslmayr, Volberg, Wimber, Dalal, & Greenlee, 2013; Makeig et al., 2002). Diminished modulation of theta activity with task demands has been reported to be more severe in patients with low performance during a selective visual attention task and with high theta activity during rest (Hanslmayr et al., 2013). The current findings about altered theta oscillations in schizophrenia, therefore, strongly point towards the necessity of considering altered transitions between brain states to better understand the neurophysiologic determinants of the illness. Ambiguous figures uniquely allow investigation of the transition between two
7
different brain states, namely perceptual switching and non-switching (that is perceptual stability), as they occur in an endogenous, self-generated succession. Taken together, this study combined the following three rationales: (1) We used ambiguous stimuli as they are ideally suited to reflect possible impairments in top-down driven organising principles of visual perception in schizophrenia. (2) We investigated theta oscillations, as they are known to reflect top-down processes. The topographical distribution of theta might differ between healthy controls and patients (Basar-Eroglu, et al., 2008). Frontal, central, parietal and occipital electrode positions were, therefore, included into the statistical comparison between groups. (3) Different findings about altered theta oscillations in schizophrenia indicate the importance to investigate the relation between different brain states, which has been accounted for by looking at time periods of perceptual switching and perceptual stability (non-switching). 1.4 Short outline of the study We analysed theta responses elicited by an ambiguous and an unambiguous control stimulus inducing apparent motion (see Figure 2A and B, for animations also Supplementary Figure S1). During the ambiguous task the perceived motion direction switched between horizontal and vertical motion or vice versa, even though the physical properties of the presentation did not change. During the unambiguous task a switch in the perceived motion direction was externally triggered by a slight reconfiguration of the stimulus. Visual perception is driven by sensory information (bottom-up) as well as by expectations and prior experiences (top-down processes). Given the reports mentioned above, we expected that in healthy persons perceptual switching would be accompanied by an increase of theta activity; reflective of top-down processes during both the ambiguous and
8
unambiguous task. In patients, we expected that this increase would be diminished, indicating weak influence of top-down processes on visual perception. This diminished increase would, however, not necessarily result from generally diminished theta activity during the entire experiment. ------------------- INSERT FIGURE 2 about here ----------------------------
2 Methods 2.1 Participants A total of 15 patients were measured. Fourteen patients were diagnosed with established schizophrenia and one patient with schizoaffective disorder. All patients were recruited as in-patients after clinical stabilization of an acute psychotic episode. The data of three patients had to be excluded from the EEG analysis, because the number of artefact-free epochs was too low for further processing. In two patients this was a result of low reversalrates during the ambiguous task. The mean age of the remaining twelve patients (4 females) was 30.0 years (SD: ± 9.6). Of those, eight patients were medicated on atypical antipsychotics, one took medication for anxiety disorder along with atypical antipsychotics and three patients were medicated on mixtures of atypical and typical antipsychotic medication. The mean duration of illness was 8 years (SD: ± 9.6 years). The educational level was assessed by rating the school leaving certificate between 1 (below highschool) to 4 (university-entrance diploma). Psychopathologic signs were measured by the Scale for the Assessment of Negative Symptoms (SANS) and by the Scale for the assessment of Positive Symptoms (SAPS, Andreasen, 1990; van Erp et al., 2014, see table 1 for details).
9
Twelve healthy volunteers (5 females) with a mean age of 31.0 (SD: 10.7) participated in the study. All participants gave written consent for their participation and had normal or corrected-to-normal vision. None of the participants reported any current substance use (for the last 14 days; excluding tobacco) or a diagnosed substance abuse disorder. Further exclusion criteria for patients and controls were a known history of learning difficulties, age > 50 years at time of assessment, evidence of head trauma with loss of consciousness, or disease of the central nervous system other than schizophrenia. All participants except one patient and one healthy volunteer were right-handed. Subsequent to the EEG measurements the participants underwent three standard neuropsychological tests to estimate group differences regarding their pre-morbid IQ (Mehrfachwahl-WortschatzIntelligenztest [MWT-B]; Lehrl, 2005), their non-verbal IQ (Zahlen-Verbindungs-Test [ZVT]; Vernon, 1993) and their ability of attention and concentration (d2 AufmerksamkeitsBelastungs-Test; M. E. Bates & Lemay, 2004). The ethics committee of the University of Bremen approved of the study. All participants signed an informed consent form prior to their participation in the study. ------------------- INSERT TABLE 1 about here ----------------------------
2.2 Stimuli and tasks Figure 2 depicts the stimuli presented to elicit apparent motion perception. An animated version of the ambiguous stimulus is illustrated in the Supplementary Figure S1. 2.2.1 Ambiguous task The Stroboscopic Alternative Motion (SAM) was presented during the ambiguous task. Fast alternation of two pictures (Amb1 and Amb2), each depicting two dots, generates the SAM 10
(Kruse, Stadler, & Wehner, 1986). For Amb1 the dots are displayed on the top left and bottom right and for Amb2 on the top right and bottom left (see Figure 2A). As indicated by its name, the SAM combines two aspects of visual illusions: (1) The display of dots in alternation with other spatially shifted dots induces illusory stroboscopic motion, that is, the illusion that dots move from one position to the other. (2) The diagonal organisation of the double-dot-displays renders the SAM ambiguous: The perceived motion direction might be either horizontal or vertical. During continuous viewing this attribution may be maintained or switched (Kruse, et al., 1986). Both, switching and non-switching time periods may, therefore, be compared (see below for the precise procedure to time-lock switching and non-switching time periods). 2.2.2 Unambiguous task For the unambiguous task a slightly modified stimulus allowed directional changes to be applied exogenously. The unambiguous vertical motion stimulus was generated by an alternating presentation of both dots either at the bottom or the top of the display (see Figure 2A, Vert1 and Vert2). The unambiguous horizontal motion stimulus was generated by an alternating presentation of both dots either at the left or the right side of the display (see Figure 2A, Horz1 and Horz2). The mean reversal rate (measured in reversals per minute) was set to 9.5, which matched the mean reversal rate during an ambiguous task in previous studies with healthy participants (Mathes, et al., 2014). For both tasks, participants were instructed to press a button immediately after the occurrence of a perceptual switch in perceived motion direction. At the subjects viewing distance of 150 cm, the horizontal and vertical distance between possible occurrences of the dots of the visual stimuli was 2.4 cm (visual angle: 0.92°) and 3.8 cm (1.45°), respectively. The dots were displayed in white on a black background. A 11
continuously presented central white dot served as fixation in both tasks. The stimuli were displayed for 165 ms and separated by an inter-stimulus interval of 85 ms during which only the central fixation dot was presented. Thus, as long as no external change of the stimulus pattern was applied (unambiguous task only) each stimulus configuration was repeated after 500 ms. All stimuli were presented on a 19’ CRT monitor controlled by a PC. Stimulus size, pace of the ongoing train of stimuli and the perceptuo-motor requirements of both tasks were identical. Recording time for each task was 6.7 minutes for the unambiguous task. Due to the partially low reversal rates in patients the recording time for the ambiguous task was extended to 10 minutes. A short learning session ensured understanding of the instructions. 2.3 EEG recording EEG was recorded with Ag-AgCl electrodes from standard locations (F3, F4, C3, C4, P3, P4, O1, O2) of the international 10-20 system, with linked earlobes serving as reference. The signal was amplified by means of a Nihon Kohden system (EEG-4421 G) with band limits between 0.1-70 Hz (24 dB/ octave) and an additional notch filter at 50 Hz. The data were digitized at a 500 Hz sampling rate and analyzed off-line. The EOG was registered from medial upper and lateral orbital rim of the right eye. 2.4 Definition of epochs and artifact rejection Figure 3 depicts schematically depicts the data segmentation procedure and the differential response between perceptual switching and perceptual stability (non-switching) in the timefrequency plane as defined below. 2.4.1 Definition of epochs containing a perceptual switch (“switching ”): Epochs were generated ranging from 1500 ms before to 1000 ms after the button press reporting a perceptual switch (“switching epochs”). If another button press within less than 12
1000 ms preceded the button press used to define the epoch, the epoch was discarded since the brain response elicited by the perceptual switch would possibly be distorted by postmotor activity of the preceding motor response. For the unambiguous task only epochs containing correctly indicated switches of perceived motion direction were used for the analysis. The length of the epoch helped to avoid filter artefacts and allowed inspecting the data for the determination of a shorter time window used for the statistical analysis (see below). 2.4.2 Definition of epochs containing no perceptual switch (“non-switching ”) For the definition of non-switching epochs an artificial trigger was generated every 225 ms for the entire recording session. The time range between 1500 ms before to 1000 ms after was defined for all artificial triggers to obtain possible epochs including non-switching time periods. In this way, the non-switching epochs were matched in length with the switching epochs to parallel the analysis, but the time-line or the definition of the 0 ms time point did not indicate a specific event as was the case for the switching epochs. All of those epochs were discarded if they contained any data points of epochs representing the switching periods or any data points of epochs that were rejected during the procedure of defining them. For the unambiguous task those epochs containing omissions were also rejected. In a consecutive manner, all remaining epochs were accepted as non-switching epochs, except they contained any data points of a previously accepted non-switching epoch (see Figure 3A). The entire procedure ensured that no data points were included in the analysis twice. Thus, the brain responses analysed as non-switching are clearly separated from brain responses related to perceptual switching. ------------------- INSERT FIGURE 3 about here ----------------------------
13
Switching and non-switching epochs contaminated by eye or other artefacts between 1100 ms before and 300 ms after the 0-ms demarcation were manually rejected offline. Single trials with amplitudes of more than 80 µV were generally discarded. By this procedure large spikes and drifts (resulting for example from muscle activity, movements affecting the electrode wires or changes in skin conductance) and large ocular artefacts (resulting from blinks and saccades) were excluded from further analysis. Visual inspection of each single trial ensured that artefacts of lower amplitude (e.g., minor blinks) were also discarded. The typical topographies and time courses to detect ocular artefacts are reviewed in more detail by Luck (2005) as well as Talsma and Woldorff (2005). Utilizing a random procedure, further epochs were excluded until the number of accepted epochs was adjusted to an average of 45 (SD: 5.6) epochs for each task (ambiguous and unambiguous), time period (switching and non-switching) and group (patient and control). 2.5 Time-frequency analysis and extraction of theta activity Time-frequency analysis was conducted in an analogous manner as in a previous study of our group (Mathes, Schmiedt, Schmiedt-Fehr, Pantelis, & Basar-Eroglu, 2012). All included epochs were transformed with a single Morlet wavelet of six cycles and a centre frequency ranging between 4 and 8 Hz in steps of 0.25 Hz using the toolbox of Torrence and Compo (1998). In order to calculate the transform
of the continuous recorded signal
convolved with the complex conjugate of the wavelet
centred at time
it was
and scaled with
:
.
14
The used mother wavelet
a Gaussian window where
was a sine wave modulated with
represents the dimensionless frequency determining the
width of the wavelet in time (for further information on wavelet analysis see review by Herrmann, Grigutsch, & Busch, 2005). For each single trial and frequency the amplitude A was determined for each participant and task (ambiguous and unambiguous) and calculated as the absolute value of the transformed data averaged over all included trials per condition (switching and non-switching):
This measure reflects the induced EEG response, which is the oscillatory activity following a sensory stimulation that is not necessarily time-locked to a stimulus (Basar-Eroglu, Strüber, Schürmann, Stadler, & Basar, 1996). The wavelets were normalised to have unit energy. The transformed data were multiplied by the square-root of the sampling interval to increase comparability with the signal magnitude if calculated by a Fourier transform (see e.g. Torrence and Compo (1998) for further information). To further visualize the specific modulations between switching and non-switching epochs for each group and task, a procedure described in detail by Delorme and Makeig (2004) was adapted: The mean amplitude was calculated over the time period between 750 before to 50 after the 0 ms demarcation for each artefact-free, non-switching epoch and averaged over all non-switching epochs separately for each task and participant. This was done only after ensuring that no systematic fluctuations in amplitude were found in the time course of the non-switching epochs (while preserving differences that may occur between the two groups and tasks). The resulting value was utilized as the baseline activity and the mean baseline log spectrum was subtracted from each spectral estimate of the switching time 15
periods, producing baseline-normalized time-frequency distribution for each task and participant. The colour at each image pixel then indicates amplification or attenuation (in dB) at a given frequency and latency for the switching time-period relative to the non-switching time period (see Figure 3B). 2.6 Definition of the analyzed time-frequency windows Time-frequency planes depicting the amplitude of the switching and non-switching epochs as well as their differentiation (see Figure 3B) were inspected separately for each task and participant and for all group grand averages. During the ambiguous task, an increase in amplitude during the switching time-period occurred broadly distributed in time and theta frequency. For the unambiguous task the increase of the theta amplitude was more restricted in the time domain. To reflect upon these observations, theta activity was analysed with a centre-frequency of F= 5.5 Hz and within a time-window ranging between 750 ms before until 50 ms after the 0-ms demarcation for the ambiguous and 600 ms before until 50 ms after the 0-ms demarcation for the unambiguous task, for both switching and non-switching trials. For that, the mean amplitude of the single-trial based measure A(t) of theta activity was calculated over all data points within the defined time windows. The approximate width in time of the centre frequency of the used wavelet was 500 ms, estimated as twice the folding time of the used Morlet wavelet, that is, the time after which the Gaussian window has dropped to exp(-2) ≈ 14% (Torrence & Compo, 1998). The frequency range as defined by two standard deviations of the wavelet in the frequency domain (i.e., of approximately 95.5% of the activity the wavelet reflects) was 3.7 to 7.3 Hz.
16
2.7 Statistical analysis 2.7.1 Statistical analysis of the behavioural results A button press occurring between 200 ms and 2500 ms after an externally applied switch of perceived motion direction was classified as a correct response for the unambiguous task. Due to the generally high performance, the error rate was investigated as the summed error score over all error types (commission errors, omission errors, too slow and too early responses). The median and the standard deviation of the reaction time (as a measure of response variability) was assessed from correct responses only. For the ambiguous task the reversal rate was defined as the number of button presses per minute. Not all participants finished neuropsychological testing (see table 1). With the remaining participants neuropsychological measures on pre-morbid functioning, non-verbal IQ and attention scores were calculated. For all group comparisons t-tests were applied. For each group the reversal rate for the ambiguous task was compared to the pre-defined setting of the 9.5 reversals per minute during the unambiguous task using nonparametric (Wilcoxon) statistics. 2.7.2 Statistical analysis of the theta response The analysis was performed separately for both perceptual tasks (ambiguous and unambiguous). For each task an analysis of variance was conducted including two groups (patients and controls) as between-subject factors as well as two conditions (switching and non-switching) and four locations (mean of F3/F4, C3/C4, P3/P4 and O1/O2, respectively) as within-subject factors. Huynh-Feldt corrected probabilities are reported to correct for any violations of the assumption of sphericity. Partial eta square values are reported as an estimate of effect size. Post-hoc comparisons using t-test statistics were applied following
17
significance in the overall ANOVA. All post-hoc tests were corrected using the Bonferroni procedure.
3 Results 3.1 Behavioural results The groups did not differ on the reversal rate during the ambiguous task or the median reaction time and error rate during the unambiguous task. The variability of the reaction time during the unambiguous task was larger in patients (p
S1388-2457(15)00541-6 http://dx.doi.org/10.1016/j.clinph.2015.02.061 CLINPH 2007488
To appear in:
Clinical Neurophysiology
Accepted Date:
3 February 2015
Please cite this article as: Mathes, B., Schmiedt-Fehr, C., Kedilaya, S., Strüber, D., Brand, A., Basar-Eroglu, C., Theta response in schizophrenia is indifferent to perceptual illusion, Clinical Neurophysiology (2015), doi: http:// dx.doi.org/10.1016/j.clinph.2015.02.061
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Theta response in schizophrenia is indifferent to perceptual illusion
Birgit Mathes (1,2)*, Christina Schmiedt-Fehr (1,2), Shwetha Kedilaya (1,2), Daniel Strüber (3,4), Andreas Brand (1,2), Canan Basar-Eroglu (1,2)
(1) University of Bremen, Institute of Psychology and Cognition Research, Bremen, Germany (2) Centre for Cognitive Science, Bremen, Germany (3) Experimental Psychology Lab, Dpt. of Psychology, European Medical School, University of Oldenburg (4) Research Center Neurosensory Science, University of Oldenburg
* Corresponding author: Dr. Birgit Mathes University of Bremen, Institute of Psychology and Cognition Research, Grazer Str. 4, 28359 Bremen, Germany Tel.: +49-421-218 68707 Fax: +49-421-218 68719 E-mail: [email protected] URL: http://www.ipk.uni-bremen.de/en/start.html
Highlights -
Multistable perception allows examining the patient’s perceptual deficits while the stimulus remains unchanged.
-
This EEG study complements behavioural studies indicating that top-down driven organising principles of perception in schizophrenia is weak
-
The altered oscillatory theta response reflects impairments in the spatio-temporal integration of the neural information transfer in schizophrenia.
Abstract Objective: Patients with schizophrenia are impaired in maintaining coherent perceptual experiences. This is reflected in the oscillatory theta response and can be investigated by visual illusions. Ambiguous stimuli elicit illusory perceptual switches while the stimulus remains unchanged. Methods: Theta responses elicited by an ambiguous and unambiguous control stimulus were measured using the EEG during time periods of perceptual switching and perceptual stability (non-switching). Results: For the ambiguous task, theta activity increased during perceptual switching in healthy controls only. For the unambiguous task, the switching-related increase of theta activity was larger in controls than in patients. This reduced modulation of the theta response seems not to be related to a general decrease of theta activity in patients. Conclusions: These findings may be related to disturbances in the spatio-temporal integration of neural activity in patients. Reporting ambiguous and unambiguous perceptual switches seems to be more demanding for patients with schizophrenia than healthy controls. Significance: This is one of the first studies on the neurophysiologic correlates of illusory perception in schizophrenia. Focussing on the relation between different brain states (such as switching and non-switching) might integrate different findings about altered theta oscillations in schizophrenia.
Keywords: Brain oscillations; long-range coordination; top-down bottom-up; multistable perception; object perception.
2
1. Introduction Patients with schizophrenia are known to have problems in integrating sensory information into coherent perceptual experiences. Core symptoms of the illness, such as hallucinations, delusions and thought disorder, might be related to this impairment (Haesebaert et al., 2013; Hugdahl, 2009; Silverstein & Keane, 2011). 1.1 Top-down processes in schizophrenia investigated by visual illusions Object perception is driven by sensory information (bottom-up) as well expectations and prior experiences (top-down, Gregory, 1997; Güntekin & Basar, 2014; Long & Toppino, 2004). The importance and power of top-down processes in shaping coherent perception is revealed by visual illusions (Basar-Eroglu, Strüber, Stadler, Kruse, & Basar, 1993; Sterzer, Kleinschmidt, & Rees, 2009). Visual illusions arise by systematic discrepancies between a stimulus and the percept due to neural processes transforming sensory information into a perceptual experience (Gregory, 1997). Patients with schizophrenia seem to be less prone to visual illusions (for a review see Notredame, Pins, Deneve, & Jardri, 2014). From distance, healthy persons perceive a hollow mask as a proper face (Gregory, 1997) while patients with schizophrenia perceive the physically correct inverted face (Dima, Dietrich, Dillo, & Emrich, 2010; Dima et al., 2009). Their perception seems to be driven more strongly by the sensory information (i.e. depth cues) than by top-down driven organising principles (i.e., the implausibility of a person with an inward turned nose; Dima, et al., 2010; Dima, et al., 2009). Ambiguous stimuli elicit a special case of visual illusions. The sensory information of an ambiguous figure is inconclusive regarding the generation of coherent perception, because more than one meaningful solution is similarly probable to account for the current percept. During continuous observation, perception switches between all possible perceptual
3
alternatives although the physical properties remain unchanged (Long & Toppino, 2004, see Figure 1 for some examples). The individual speed of perceptual switching (the reversal rate) is variable (Strüber, BasarEroglu, Hoff, & Stadler, 2000; Strüber, Basar-Eroglu, Miener, & Stadler, 2001). It remains unclear whether the spontaneous reversal rate in schizophrenia differs from healthy persons. In the few studies available the findings range from acceleration to slowing of perceptual switching (Calvert et al., 1988; Eysenck, 1952; Keil, Elbert, Rockstroh, & Ray, 1998; Philip, 1953; Tschacher, Dubouloz, Meier, & Junghan, 2008). More consistent are reports about the lack of top-down control on the current percept of an ambiguous figure. The Schröder staircase has a semantic and, therefore, top-down processed bias, because perceiving the staircase from above is more realistic than perceiving it from below (see Strüber & Stadler, 1999 and Figure 1 for ambiguous figures with different semantic biases). Patients with schizophrenia perceive the more realistic perspective of the Schröder staircase for a shorter amount of time than healthy persons (Calvert, et al., 1988; Keil, et al., 1998). These studies indicate that semantic properties of an ambiguous stimulus have lower impact on the perception of patients than of healthy persons. The efficiency in maintaining a desired perceptual alternative or initiating a perceptual switch by voluntary control is also diminished (Eysenck, 1952; McBain, Norton, Kim, & Chen, 2011). Thus, behavioural studies indicate that the influence of top-down processes on the perception of ambiguous figures is diminished in schizophrenia. This is in accordance with the broader scope of the patient’s deficiencies in using contextual information during object perception (Silverstein & Keane, 2011) and implies that in schizophrenia top-down driven organising principles of perception are weak (John & Hemsley, 1992; Silverstein & Keane, 2011; Tsunoda et al., 2012).
4
To the best of our knowledge, this and our recent study (Basar-Eroglu, Mathes, Khalaidovski, Brand, & Schmiedt-Fehr, 2014) are the first that have investigated the neural processes underlying the perception of ambiguous figures in schizophrenia. This approach allows the investigation of top-down driven organising principles of visual perception in schizophrenia as they become apparent without changes of the external surrounding. ------------------- INSERT FIGURE 1 about here ----------------------------
1.2 Top-down processes investigated by theta oscillations Event-related theta oscillations (approx 4 – 7 Hz) have a frontal maximum (Basar-Eroglu & Demiralp, 2001) and play an important role for top-down regulated processes, such as focussed attention (Basar-Eroglu & Demiralp, 2001; Cahn, Delorme, & Polich, 2013; Polich, 2007), control mechanisms in working memory (Sauseng, Griesmayr, Freunberger, & Klimesch, 2010) or executive functions (Huster, Enriquez-Geppert, Lavallee, Falkenstein, & Herrmann, 2013; Schmiedt-Fehr & Basar-Eroglu, 2011; Schmiedt-Fehr, Dühl, & Basar-Eroglu, 2011; Yordanova, Falkenstein, Hohnsbein, & Kolev, 2004). Fronto-posterior phase-coupling of theta activity seems to be specifically important for top-down modulations of early visual areas during perceptual grouping (Volberg, Wutz, & Greenlee, 2013). We have recently shown that theta activity elicited by ambiguous figures is more pronounced at anterior in comparison to posterior sites than theta activity elicited by unambiguous figures. This may be indicative for an increased involvement of top-down processes during the perception of ambiguous figures (Mathes, Khalaidovski, Schmiedt-Fehr, & Basar-Eroglu, 2014). Theta oscillations are further known to increase during perceptual switches elicited by both, ambiguous and unambiguous figures (Ehm, Bach, & Kornmeier, 2011; Nakatani & van
5
Leeuwen, 2005). Thus, theta oscillations qualify as a neural marker to investigate possible impairments in top-down driven organising principles of visual perception in patients with schizophrenia. 1.3 Theta oscillations in schizophrenia Theta oscillations in schizophrenia are still not well understood. The highest consensus is that slow wave activity, including theta, in patients is enhanced during rest (Boutros et al., 2008). Task-related theta activity of patients may be reduced (A. T. Bates, Kiehl, Laurens, & Liddle, 2009; Doege, Jansen, Mallikarjun, Liddle, & Liddle, 2010; Doege et al., 2010; Haenschel et al., 2009; Kaser et al., 2013; Ramos-Loyo, Gonzalez-Garrido, Sanchez-Loyo, Medina, & Basar-Eroglu, 2009; Schmiedt, Brand, Hildebrandt, & Basar-Eroglu, 2005) or enhanced (Fehr et al., 2003; L.E. Hong, Moran, Du, O'Donnell, & Summerfelt, 2012; L. E. Hong, Summerfelt, Mitchell, O'Donnell, & Thaker, 2012; Missonnier et al., 2012). Reduced theta activity may be restricted to those recording sites that are involved in task processing in healthy persons, while other recording sites rather show increased activation (BasarEroglu, Schmiedt-Fehr, Marbach, Brand, & Mathes, 2008). The reduction of theta activity may not necessarily be a result of lower amplitudes, rather of a reduced consistency in timing the brain response with respect to a cognitive event (Basar-Eroglu, Schmiedt-Fehr, Mathes, Zimmermann, & Brand, 2009). Results like this indicate abnormal temporal integration and interregional connectivity of brain networks as a core disturbance in schizophrenia (for reviews see Basar, 2013; Basar & Güntekin, 2013; Ford, Krystal, & Mathalon, 2007; Friston & Frith, 1995; Uhlhaas, 2013). These disturbances might affect the entire perception-action cycle (Fuster, 2006). Thus, high theta oscillations during rest might be already indicative of a change in the spatial and temporal integration of neural activity in patients, possibly indicating that keeping the
6
perceptual experience stable and coherent is a great effort for patients already in the absence of specific cognitive demands. Changes in the amplitude (irrespective of an expected increase or decrease), changes in the topographical distribution and diminished consistency in response timing may all be indicators that theta activity cannot be effectively integrated in processing cognitive demands. The reported divergence of the results about increased or decreased theta oscillations in schizophrenia further implies the importance to learn more about the transitions between brain states. The general procedure of taking one brain state (e.g., the spontaneous EEG or the pre-stimulus activity) as a baseline to investigate another brain state (e.g., induced by a cognitive demand) has clear limitations if the baseline is already different between patients with schizophrenia and healthy persons. Comparing different perceptual task conditions with or without base-line correction might lead to different results already in healthy populations (Kaiser, Rahm, & Lutzenberger, 2008). Pioneering work in electrophysiological research has long stressed that spontaneous or pre-stimulus activity of the brain systematically impacts on task-related activity (Basar, Rahn, Demiralp, & Schürmann, 1998; Buzsaki, 2006; Klimesch, 1999). More recent studies specifically illustrate this link for alpha and theta oscillations (M. M. Doppelmayr, Klimesch, Pachinger, & Ripper, 1998; Hanslmayr, Volberg, Wimber, Dalal, & Greenlee, 2013; Makeig et al., 2002). Diminished modulation of theta activity with task demands has been reported to be more severe in patients with low performance during a selective visual attention task and with high theta activity during rest (Hanslmayr et al., 2013). The current findings about altered theta oscillations in schizophrenia, therefore, strongly point towards the necessity of considering altered transitions between brain states to better understand the neurophysiologic determinants of the illness. Ambiguous figures uniquely allow investigation of the transition between two
7
different brain states, namely perceptual switching and non-switching (that is perceptual stability), as they occur in an endogenous, self-generated succession. Taken together, this study combined the following three rationales: (1) We used ambiguous stimuli as they are ideally suited to reflect possible impairments in top-down driven organising principles of visual perception in schizophrenia. (2) We investigated theta oscillations, as they are known to reflect top-down processes. The topographical distribution of theta might differ between healthy controls and patients (Basar-Eroglu, et al., 2008). Frontal, central, parietal and occipital electrode positions were, therefore, included into the statistical comparison between groups. (3) Different findings about altered theta oscillations in schizophrenia indicate the importance to investigate the relation between different brain states, which has been accounted for by looking at time periods of perceptual switching and perceptual stability (non-switching). 1.4 Short outline of the study We analysed theta responses elicited by an ambiguous and an unambiguous control stimulus inducing apparent motion (see Figure 2A and B, for animations also Supplementary Figure S1). During the ambiguous task the perceived motion direction switched between horizontal and vertical motion or vice versa, even though the physical properties of the presentation did not change. During the unambiguous task a switch in the perceived motion direction was externally triggered by a slight reconfiguration of the stimulus. Visual perception is driven by sensory information (bottom-up) as well as by expectations and prior experiences (top-down processes). Given the reports mentioned above, we expected that in healthy persons perceptual switching would be accompanied by an increase of theta activity; reflective of top-down processes during both the ambiguous and
8
unambiguous task. In patients, we expected that this increase would be diminished, indicating weak influence of top-down processes on visual perception. This diminished increase would, however, not necessarily result from generally diminished theta activity during the entire experiment. ------------------- INSERT FIGURE 2 about here ----------------------------
2 Methods 2.1 Participants A total of 15 patients were measured. Fourteen patients were diagnosed with established schizophrenia and one patient with schizoaffective disorder. All patients were recruited as in-patients after clinical stabilization of an acute psychotic episode. The data of three patients had to be excluded from the EEG analysis, because the number of artefact-free epochs was too low for further processing. In two patients this was a result of low reversalrates during the ambiguous task. The mean age of the remaining twelve patients (4 females) was 30.0 years (SD: ± 9.6). Of those, eight patients were medicated on atypical antipsychotics, one took medication for anxiety disorder along with atypical antipsychotics and three patients were medicated on mixtures of atypical and typical antipsychotic medication. The mean duration of illness was 8 years (SD: ± 9.6 years). The educational level was assessed by rating the school leaving certificate between 1 (below highschool) to 4 (university-entrance diploma). Psychopathologic signs were measured by the Scale for the Assessment of Negative Symptoms (SANS) and by the Scale for the assessment of Positive Symptoms (SAPS, Andreasen, 1990; van Erp et al., 2014, see table 1 for details).
9
Twelve healthy volunteers (5 females) with a mean age of 31.0 (SD: 10.7) participated in the study. All participants gave written consent for their participation and had normal or corrected-to-normal vision. None of the participants reported any current substance use (for the last 14 days; excluding tobacco) or a diagnosed substance abuse disorder. Further exclusion criteria for patients and controls were a known history of learning difficulties, age > 50 years at time of assessment, evidence of head trauma with loss of consciousness, or disease of the central nervous system other than schizophrenia. All participants except one patient and one healthy volunteer were right-handed. Subsequent to the EEG measurements the participants underwent three standard neuropsychological tests to estimate group differences regarding their pre-morbid IQ (Mehrfachwahl-WortschatzIntelligenztest [MWT-B]; Lehrl, 2005), their non-verbal IQ (Zahlen-Verbindungs-Test [ZVT]; Vernon, 1993) and their ability of attention and concentration (d2 AufmerksamkeitsBelastungs-Test; M. E. Bates & Lemay, 2004). The ethics committee of the University of Bremen approved of the study. All participants signed an informed consent form prior to their participation in the study. ------------------- INSERT TABLE 1 about here ----------------------------
2.2 Stimuli and tasks Figure 2 depicts the stimuli presented to elicit apparent motion perception. An animated version of the ambiguous stimulus is illustrated in the Supplementary Figure S1. 2.2.1 Ambiguous task The Stroboscopic Alternative Motion (SAM) was presented during the ambiguous task. Fast alternation of two pictures (Amb1 and Amb2), each depicting two dots, generates the SAM 10
(Kruse, Stadler, & Wehner, 1986). For Amb1 the dots are displayed on the top left and bottom right and for Amb2 on the top right and bottom left (see Figure 2A). As indicated by its name, the SAM combines two aspects of visual illusions: (1) The display of dots in alternation with other spatially shifted dots induces illusory stroboscopic motion, that is, the illusion that dots move from one position to the other. (2) The diagonal organisation of the double-dot-displays renders the SAM ambiguous: The perceived motion direction might be either horizontal or vertical. During continuous viewing this attribution may be maintained or switched (Kruse, et al., 1986). Both, switching and non-switching time periods may, therefore, be compared (see below for the precise procedure to time-lock switching and non-switching time periods). 2.2.2 Unambiguous task For the unambiguous task a slightly modified stimulus allowed directional changes to be applied exogenously. The unambiguous vertical motion stimulus was generated by an alternating presentation of both dots either at the bottom or the top of the display (see Figure 2A, Vert1 and Vert2). The unambiguous horizontal motion stimulus was generated by an alternating presentation of both dots either at the left or the right side of the display (see Figure 2A, Horz1 and Horz2). The mean reversal rate (measured in reversals per minute) was set to 9.5, which matched the mean reversal rate during an ambiguous task in previous studies with healthy participants (Mathes, et al., 2014). For both tasks, participants were instructed to press a button immediately after the occurrence of a perceptual switch in perceived motion direction. At the subjects viewing distance of 150 cm, the horizontal and vertical distance between possible occurrences of the dots of the visual stimuli was 2.4 cm (visual angle: 0.92°) and 3.8 cm (1.45°), respectively. The dots were displayed in white on a black background. A 11
continuously presented central white dot served as fixation in both tasks. The stimuli were displayed for 165 ms and separated by an inter-stimulus interval of 85 ms during which only the central fixation dot was presented. Thus, as long as no external change of the stimulus pattern was applied (unambiguous task only) each stimulus configuration was repeated after 500 ms. All stimuli were presented on a 19’ CRT monitor controlled by a PC. Stimulus size, pace of the ongoing train of stimuli and the perceptuo-motor requirements of both tasks were identical. Recording time for each task was 6.7 minutes for the unambiguous task. Due to the partially low reversal rates in patients the recording time for the ambiguous task was extended to 10 minutes. A short learning session ensured understanding of the instructions. 2.3 EEG recording EEG was recorded with Ag-AgCl electrodes from standard locations (F3, F4, C3, C4, P3, P4, O1, O2) of the international 10-20 system, with linked earlobes serving as reference. The signal was amplified by means of a Nihon Kohden system (EEG-4421 G) with band limits between 0.1-70 Hz (24 dB/ octave) and an additional notch filter at 50 Hz. The data were digitized at a 500 Hz sampling rate and analyzed off-line. The EOG was registered from medial upper and lateral orbital rim of the right eye. 2.4 Definition of epochs and artifact rejection Figure 3 depicts schematically depicts the data segmentation procedure and the differential response between perceptual switching and perceptual stability (non-switching) in the timefrequency plane as defined below. 2.4.1 Definition of epochs containing a perceptual switch (“switching ”): Epochs were generated ranging from 1500 ms before to 1000 ms after the button press reporting a perceptual switch (“switching epochs”). If another button press within less than 12
1000 ms preceded the button press used to define the epoch, the epoch was discarded since the brain response elicited by the perceptual switch would possibly be distorted by postmotor activity of the preceding motor response. For the unambiguous task only epochs containing correctly indicated switches of perceived motion direction were used for the analysis. The length of the epoch helped to avoid filter artefacts and allowed inspecting the data for the determination of a shorter time window used for the statistical analysis (see below). 2.4.2 Definition of epochs containing no perceptual switch (“non-switching ”) For the definition of non-switching epochs an artificial trigger was generated every 225 ms for the entire recording session. The time range between 1500 ms before to 1000 ms after was defined for all artificial triggers to obtain possible epochs including non-switching time periods. In this way, the non-switching epochs were matched in length with the switching epochs to parallel the analysis, but the time-line or the definition of the 0 ms time point did not indicate a specific event as was the case for the switching epochs. All of those epochs were discarded if they contained any data points of epochs representing the switching periods or any data points of epochs that were rejected during the procedure of defining them. For the unambiguous task those epochs containing omissions were also rejected. In a consecutive manner, all remaining epochs were accepted as non-switching epochs, except they contained any data points of a previously accepted non-switching epoch (see Figure 3A). The entire procedure ensured that no data points were included in the analysis twice. Thus, the brain responses analysed as non-switching are clearly separated from brain responses related to perceptual switching. ------------------- INSERT FIGURE 3 about here ----------------------------
13
Switching and non-switching epochs contaminated by eye or other artefacts between 1100 ms before and 300 ms after the 0-ms demarcation were manually rejected offline. Single trials with amplitudes of more than 80 µV were generally discarded. By this procedure large spikes and drifts (resulting for example from muscle activity, movements affecting the electrode wires or changes in skin conductance) and large ocular artefacts (resulting from blinks and saccades) were excluded from further analysis. Visual inspection of each single trial ensured that artefacts of lower amplitude (e.g., minor blinks) were also discarded. The typical topographies and time courses to detect ocular artefacts are reviewed in more detail by Luck (2005) as well as Talsma and Woldorff (2005). Utilizing a random procedure, further epochs were excluded until the number of accepted epochs was adjusted to an average of 45 (SD: 5.6) epochs for each task (ambiguous and unambiguous), time period (switching and non-switching) and group (patient and control). 2.5 Time-frequency analysis and extraction of theta activity Time-frequency analysis was conducted in an analogous manner as in a previous study of our group (Mathes, Schmiedt, Schmiedt-Fehr, Pantelis, & Basar-Eroglu, 2012). All included epochs were transformed with a single Morlet wavelet of six cycles and a centre frequency ranging between 4 and 8 Hz in steps of 0.25 Hz using the toolbox of Torrence and Compo (1998). In order to calculate the transform
of the continuous recorded signal
convolved with the complex conjugate of the wavelet
centred at time
it was
and scaled with
:
.
14
The used mother wavelet
a Gaussian window where
was a sine wave modulated with
represents the dimensionless frequency determining the
width of the wavelet in time (for further information on wavelet analysis see review by Herrmann, Grigutsch, & Busch, 2005). For each single trial and frequency the amplitude A was determined for each participant and task (ambiguous and unambiguous) and calculated as the absolute value of the transformed data averaged over all included trials per condition (switching and non-switching):
This measure reflects the induced EEG response, which is the oscillatory activity following a sensory stimulation that is not necessarily time-locked to a stimulus (Basar-Eroglu, Strüber, Schürmann, Stadler, & Basar, 1996). The wavelets were normalised to have unit energy. The transformed data were multiplied by the square-root of the sampling interval to increase comparability with the signal magnitude if calculated by a Fourier transform (see e.g. Torrence and Compo (1998) for further information). To further visualize the specific modulations between switching and non-switching epochs for each group and task, a procedure described in detail by Delorme and Makeig (2004) was adapted: The mean amplitude was calculated over the time period between 750 before to 50 after the 0 ms demarcation for each artefact-free, non-switching epoch and averaged over all non-switching epochs separately for each task and participant. This was done only after ensuring that no systematic fluctuations in amplitude were found in the time course of the non-switching epochs (while preserving differences that may occur between the two groups and tasks). The resulting value was utilized as the baseline activity and the mean baseline log spectrum was subtracted from each spectral estimate of the switching time 15
periods, producing baseline-normalized time-frequency distribution for each task and participant. The colour at each image pixel then indicates amplification or attenuation (in dB) at a given frequency and latency for the switching time-period relative to the non-switching time period (see Figure 3B). 2.6 Definition of the analyzed time-frequency windows Time-frequency planes depicting the amplitude of the switching and non-switching epochs as well as their differentiation (see Figure 3B) were inspected separately for each task and participant and for all group grand averages. During the ambiguous task, an increase in amplitude during the switching time-period occurred broadly distributed in time and theta frequency. For the unambiguous task the increase of the theta amplitude was more restricted in the time domain. To reflect upon these observations, theta activity was analysed with a centre-frequency of F= 5.5 Hz and within a time-window ranging between 750 ms before until 50 ms after the 0-ms demarcation for the ambiguous and 600 ms before until 50 ms after the 0-ms demarcation for the unambiguous task, for both switching and non-switching trials. For that, the mean amplitude of the single-trial based measure A(t) of theta activity was calculated over all data points within the defined time windows. The approximate width in time of the centre frequency of the used wavelet was 500 ms, estimated as twice the folding time of the used Morlet wavelet, that is, the time after which the Gaussian window has dropped to exp(-2) ≈ 14% (Torrence & Compo, 1998). The frequency range as defined by two standard deviations of the wavelet in the frequency domain (i.e., of approximately 95.5% of the activity the wavelet reflects) was 3.7 to 7.3 Hz.
16
2.7 Statistical analysis 2.7.1 Statistical analysis of the behavioural results A button press occurring between 200 ms and 2500 ms after an externally applied switch of perceived motion direction was classified as a correct response for the unambiguous task. Due to the generally high performance, the error rate was investigated as the summed error score over all error types (commission errors, omission errors, too slow and too early responses). The median and the standard deviation of the reaction time (as a measure of response variability) was assessed from correct responses only. For the ambiguous task the reversal rate was defined as the number of button presses per minute. Not all participants finished neuropsychological testing (see table 1). With the remaining participants neuropsychological measures on pre-morbid functioning, non-verbal IQ and attention scores were calculated. For all group comparisons t-tests were applied. For each group the reversal rate for the ambiguous task was compared to the pre-defined setting of the 9.5 reversals per minute during the unambiguous task using nonparametric (Wilcoxon) statistics. 2.7.2 Statistical analysis of the theta response The analysis was performed separately for both perceptual tasks (ambiguous and unambiguous). For each task an analysis of variance was conducted including two groups (patients and controls) as between-subject factors as well as two conditions (switching and non-switching) and four locations (mean of F3/F4, C3/C4, P3/P4 and O1/O2, respectively) as within-subject factors. Huynh-Feldt corrected probabilities are reported to correct for any violations of the assumption of sphericity. Partial eta square values are reported as an estimate of effect size. Post-hoc comparisons using t-test statistics were applied following
17
significance in the overall ANOVA. All post-hoc tests were corrected using the Bonferroni procedure.
3 Results 3.1 Behavioural results The groups did not differ on the reversal rate during the ambiguous task or the median reaction time and error rate during the unambiguous task. The variability of the reaction time during the unambiguous task was larger in patients (p
