• No se han encontrado resultados

Empiece con una oración: Padre nuestro que estas en los cielos. Santificado sea tu nombre

Incluir en el informe que realizan al final de cada clase

Lección 6 - El uso de las llaves / Los papeles del hombre y la mujer

2. Empiece con una oración: Padre nuestro que estas en los cielos. Santificado sea tu nombre

Over the last decade, researchers have used multivariate pattern-classification analyses (MVPA) of fMRI BOLD to decode various behavioral and cognitive states (e.g. perceiving, attending to, and imagining features, objects, and scenes) from neural activity (for reviews, see Haynes & Rees, 2006, Haynes, 2015, Naselaris et al., 2011, Norman et al., 2006, Pereira et al., 2009 and Tong & Pratte, 2012). In contrast to univariate approaches that analyse the time course of each voxel independently, MVPA is able to reveal patterns of activity across the entire brain (or smaller ROIs), taking into consideration the activation and deactivation of each voxel. MVPA is comprised of a multitude of methods for analyzing neuroimaging data. The common element that unites these different methods is that they take into consideration the relationship between multiple variables (e.g., voxels in fMRI or channels in MEG/EEG), instead of treating them as independent of one another and measuring relative activation strengths. Previous neuroimaging studies using univariate analyses, have suggested that there are different brain areas that are specialized for different perceptual processes (e.g. Gauthier

et al., 1999) and for specific categories of stimuli (e.g. Kanwisher et al., 1997;

Hesselmann et al., 2008). For example, a well-documented effect is the higher activation of the FFA in both detection and identification of faces vs. non-face objects (e.g.

activation in other regions of the ventral occipitotemporal cortex when perceiving cars; Grill-Spector et al., 2004; Kleinschmidt & Cohen, 2006). However, in a study conducted by Haxby et al. (2001) using a form of multivariate pattern analysis, the results suggest that the representation of faces and different categories of objects are widely distributed and overlapping. In particular, they found that one could decode whether the participant was seeing a face or another category of stimulus even from the pattern of brain activity outside of well-known ‘face-specific’ regions. Patterns of activity that discriminated among all categories (faces and objects) were found even within cortical regions that have been previously reported as being category specific. Their results suggest that regions such as the PPA (Parahippocampal Place Area) or the FFA are not specific to only spatial arrangements or human faces but, rather, are part of a more extended activation pattern for all categories. The results from Haxby et al. (2001)’s study

contrast with the typical univariate analysis, which would not pick up differences based on the activation pattern across voxels but rather the comparisons in the activation levels of individual voxels. The type of analysis conducted by Haxby et al. (2001) is known as decoding. It is the most popular application of MVPA, and the approach that we use in our studies.

The term “decoding” refers to the prediction of an outcome (e.g., stimulus category being viewed; reversal trial or not) based on the pattern of brain activity. Consider a simple experimental design in which the participant viewed images of

upright faces and inverted faces while their brain activity was recorded (Figure 2.4A). The goal of the decoding analysis is to test whether we can predict, with above chance (50%) accuracy, whether the participant was viewing an upright face or an inverted one based on the pattern of activity across the brain (in response to the different

stimuli/experimental conditions recorded using neuroimaging techniques such as EEG, MEG, fMRI, etc). If the experimental stimuli can be successfully “decoded” from the participant’s patterns of brain activation, a conclusion can be made that some

information relevant to the experimental manipulation exists in the neuroimaging data (Grootswaggers et al., 2017). For instance, if one can accurately decode whether trials are reversals or not, this suggests that the widespread pattern of brain activity holds information about this process. See Figure 2.4 for an illustration of the decoding approach adapted from Grootswaggers et al. (2017)’s figure (Figure 1, p.679).

Decoding in EEG is typically conducted based on the pattern of activity across the scalp. Importantly, to take advantage of the temporal resolution of EEG, the scalp activity pattern can be decoded at each time point across the trial (Figure 2.4E) to obtain a time-course of prediction accuracy values. Timeframes with above chance decoding accuracy hold information about the condition or stimulus category whereas those time frames with chance accuracy do not. This type of analysis is compatible for answering my research questions on identifying the temporal profile of reversal-related brain activity in the pre-stimulus and post-stimulus periods (Chapters 4&5). Seeing as it incorporates information across all electrodes, I expect that it will reveal reversal-related mechanisms beyond those that have been detected using averaged data with univariate methods. This idea is in line with previous findings by Ronconi et al. (2017). In the

study mentioned previously (in Chapter 1), conducted by Ronconi et al. (2017), results showed that the perceptual outcome from bistable stimulus perception could be reliably decoded from the phase of prestimulus oscillations in right parieto-occiptal channels. This type of activity cannot be detected with univariate approaches because the pre- stimulus period is usually flattened due to averaging. Decoding analyses can result in earlier detection of differences in activation patterns (Cauchoix et al., 2012; Cauchoix et al., 2014).

Recently, researchers have started applying pattern-classification analyses to electroencephalography (EEG) data (e.g. List et al., 2017; Ronconi et al., 2017). This application to EEG has furthered the findings from the standard ERP analyses. For example, List et al. (2017), using pattern classification, have found that stimulus

Figure 2.4. An illustration of the decoding approach. (A) Brain responses to participants viewing inverted and upright faces. (B) Patterns of activity evoked by the two stimuli represented in multiple dimension. (C) A classifier is trained on a subset of the data with the aim of discriminating accurately the patterns of activity associated with the different

stimuli. (D) Testing the predictions of the trained classifier on the data not used in training to obtain decoding accuracy. This process is repeated (E) across time for EEG data and (F) over regions for fMRI data. This figure was adapted from Grootswager et al. (2017).

250ms with a spatio-temporal pattern that mirrored and extended previous ERP findings linked to these to these two types of stimuli. For instance, in one of their results, their topographic maps on faces versus gabors discrimination revealed that this discrimination emerges primarily from the left posterior electrode sites (for most of their participants) with a time frame that is consistent with the N170 component. Moreover, in a second experiment, List et al. (2017) revealed that this type of analysis (MVPA) can decode the scope of attention from EEG data. Their pattern classification analyses identified linear topographies of EEG signals that successfully distinguish, on a trial-by-trial basis, locally-focused vs. globally-focused attention.

Although ERP analyses offer a great deal of information on the underlying mechanisms of perceptual reversals, as indexed by findings in the literature, and would help us identify what factors influence the changes we might observe in the RP and the RN in our experiments (Experiments 1 & 2), there could potentially be effects that are not detected due to the univariate nature of these analyses. This poses a problem for some of the aims of my thesis, seeing as I am setting to explore whether there is any pre- stimulus and post-stimulus activity that has not been identified previously. Experiments 1, 2 (Chapter 4), 3 and 4 (Chapter 5) use MVPA in order to identify the temporal windows in the pre-stimulus (Chapter 5) and post-stimulus (Chapter 4) periods during which the activity and frequency modulations (explained in more detail in Chapter 4 and Chapter 5) are linked to perceptual reversals and to percept choice in ambiguous figure perception. This analysis will help further our (Chapter 3) and previous ERP findings that might not have been able to detect this activity due to the univariate nature of these

analyses (Chapter 4). Moreover, they will help us understand the nature of the pre- stimulus activity that has been linked to ambiguous figure perception (Chapters 5).

Chapter 3: Response Dependence of Reversal Related ERP Components in

Documento similar