Fast decisions about environmental information require categorization to distinguish between animate and non-animate things, plants and animals, vehicles and buildings, etc. (Fabre-Thorpe, 2011). Categorization serves not only distinction but also generalization when different objects are grouped on the basis of shared features (Keller and Soenfeld, 1950). The visual environment does not always favor perception: fog, poor lighting, absence of colors, low contrast, short flashes of an image allow only decisions made on the basis of coarse, global features or outlines of objects. In addition, sometimes only the periphery of the visual field is stimulated; still, we need to know whether this visual information has any relevance. For a detailed analysis on the other hand, fine details, colors and edges are important.

For fast and efficient categorization relevant information and actual goals should be considered. This process might root in the two major visual processing streams: the magnocellular pathway (MC) and the parvocellular pathway (PC). The majority of axons leaving the retina belong to either the MC or the PC. The MC runs (partly) to the frontal lobe, while the end of the PC stream is in the inferotemporal cortex (IT), a region essential for visual recognition. Instead of a detailed description (but see e.g., Mishkin and Ungerleider, 1982; Goodale and Milner, 1992) of the fundamental differences in the properties of the MC and the PC, here we focus only on those features of the MC which are relevant to our study. The MC pathway is very fast. Differences in conduction speed between the two pathways can be demonstrated as early as the lateral geniculate body (LGB): information arriving via the PC has some 20 ms delay as compared to the MC, and this difference is also present in V1 (Maunsell and Newsome, 1987; Nowak et al., 1995; Schmolesky et al., 1998). After V1 it takes only 6–9 ms to reach V3, the middle temporal area (MT), the middle superior temporal area (MST) or the frontal eye field (FEF) (Schmolesky et al., 1998).

On the basis of latency differences between the PC and the MC, Nowak and his colleagues suggested that visual signals processed in the MC might modulate activity in the PC through feed-forward, lateral or feed-back connections (Nowak and Bullier, 1997). Information carried rapidly by the MC toward the frontal areas may exert a top-down effect. In contrast with the hierarchical views of visual processing, this top-down effect is supposed to be able to modulate lower regions from higher cortical areas which have been activated earlier (Knierim and van Essen, 1992; Zipser et al., 1996). However, due to the fact that the MC is sensitive only to coarse features, the role of the MC in object recognition was not investigated for long. Recently published papers, however, suggest that when time is an issue, the MC carries sufficient data to extract relevant information, which—provided there is enough time—can be completed by colors and details carried by the PC. Several experiments (see below) were carried out in order to investigate rapid categorization by using pathway-specific stimulation.

Research on decisions concerning MC information can benefit from the fact that images projected on the peripheral retina almost exclusively stimulate the rod system. In a study by Thorpe and colleagues (Thorpe et al., 2001), participants had to decide about images and choose between animate/non-animate categories. Their results showed that eccentricity did not have an influence on the accuracy of the decisions and that low spatial frequency (LSF) information originating from the periphery of the retina was sufficient for categorization. It was also shown that rapid categorization is possible in the absence of colors (Delorme et al., 2010). The MC is sensitive to the achromatic differences in luminance; the pathway can be stimulated by stimuli having low (<8%) contrast and LSF (Tootell et al., 1988). Experiments on monkey and human participants using contrast differences (Mace et al., 2005, 2010) were performed and showed that images with sufficiently low contrast are invisible for the PC, so decisions concerning the stimuli must be based on information carried by the MC. If the PC were the only pathway involved in visual categorization, low contrast stimuli should cause a dramatic decrease in performance. However, at contrast values of 3% performance did not change significantly in either species, which suggests that it might be done on the basis of coarse information carried by the MC (Bar et al., 2001; Bar, 2003).

Different spatial frequencies carry different aspects of the visual stimuli. High spatial frequencies (HSFs) carry information about edges and patterns, while LSFs contain global information. The latter might be sufficient to make a first, global impression about the general shape of objects. Psychophysical studies show that LSF patterns (Sachs et al., 1971; De Valois et al., 1990) and complex sceneries (Schyns and Oliva, 1994; Mace et al., 2005, 2010) are perceived earlier than high SF. Electrophysiological results show that the first part of the activity of IT cells reflects global information (Sugase et al., 1999; Tamura and Tanaka, 2001) and only the later part of the responses, after some 51 ms, carries information about fine details (Sugase et al., 1999). This means that IT neurons respond first to low LSF and global features and only after that to fine details.

According to the studies mentioned above and based on their EEG findings, Thorpe and Fabre-Thorpe suggested an MC based, fast pathway which uses the same cortical areas as the ventral pathway. Thus, MC information arrives at the IT faster and reaches the prefrontal cortex and the motor cortex earlier than information carried by the PC if a fast decision is needed (Fabre-Thorpe et al., 2001; Thorpe and Fabre-Thorpe, 2003). Reaction times in monkeys performing rapid visual categorization are as short as 180 ms, which leaves time only for a feed-forward processing through the IT to the motor cortex via the prefrontal and premotor cortices (Fabre-Thorpe et al., 1998). It was also suggested that MC information supported PC processing through fast, local feed-back circuits along the ventral visual stream (Fabre-Thorpe, 2011).

Bar and his colleagues, on the other hand, hypothesized a top-down process which, using the rapid processing in the MC through the dorsal pathway could provide the IT with coarse but fast information through the orbitofrontal cortex (OFC). This top-down mechanism can limit the number of possible interpretations, decrease the amount of necessary computation and reduce the time needed. This global information is essential for making fast decisions for survival (Bar, 2003). In these experiments, the two pathways were stimulated selectively and categorization was required (Bar, 2003; Kveraga et al., 2007a,b). According to the findings, the critical structure in top-down processes is the OFC, whose early activation can be attributed to processing visual information in the MC (Bar, 2003; Kveraga et al., 2007b). In addition, a study investigating the functional coupling of cortical areas found phase coupling between V1 and the OFC, and the OFC and the IT (see Lin et al., 2004). Rokszin et al. (2016) investigated how the top-down effects are manifested in scalp ERPs when presenting low or high SF information. They found evidence of top-down, anterior effect for MC optimized images within the first 200 ms of visual processing (shorter N1 latencies and amplitude changes spreading to anterior scalp regions). The connection is provided by the fibers of the uncinate fascicle and the external capsule connecting the OFC with the IT (Cavada and Goldman-Rakic, 1989; Cavada et al., 2000; Fang et al., 2005).

It is important to note that although the MC is regarded as the main input for the dorsal or “Where?” pathway processing motion and serving spatial attention, nearly 50% of the MC fibers feed information into the ventral stream (Ferrera et al., 1992; Nealey and Maunsell, 1994). There is plenty of evidence supporting the role of the MC pathway in fast categorization; however, it is unclear whether this information after leaving V1 reaches the IT via the dorsal (a top-down process through the OFC) or the ventral pathway (local feed-forward or feed-back circuits preceding PC information) (Figure 1).

The goal of our study was to determine which of the above scenarios is more likely: does MC information responsible for fast visual decisions pass through the OFC or does it run together with the ventral pathway? One possible approach of the problem might be to interfere with the dorsal or ventral pathway to see whether the processing of those stimuli which are characteristic to the given pathway is affected or not. A logical choice is a non-invasive and reproducible electrical stimulation of the pathway(s).

Electrical stimulation manipulates the activity of cortical networks transitionally and reversibly in a non-invasive and painless way. The method consists of a weak transcranial current (tDCS) flowing through the brain using two large surface electrodes (Nitsche and Paulus, 2000; Manuel et al., 2014), which can influence cortical functions. In the past few years several studies investigated visual processing in humans using non-invasive electrical stimulation to directly modulate visual cortices in human subjects (Antal et al., 2001). The anodal stimulation over V1 increases the sensitivity of phosphenes (Antal et al., 2003a), contrast sensitivity, enhances the amplitude of N70 while the opposite effects were found using cathodal stimulation (Antal et al., 2003b,c, 2004a; Kraft et al., 2010). Futhermore, tDCS modulates human color discrimination in a pathway-specific manner (Costa et al., 2012). The anodal stimulation over MT improves learning of visually guided tracking movements (Antal et al., 2004c). After learning the anodal stimulation has no effect, but cathodal stimulation can increase the signal-to-noise ratio and improve the performance in the learned task (Antal et al., 2004b). The tDCS over the posterior parietal cortex modulates visuospatial processing (Sparing et al., 2009), bilateral stimulation over the anterior temporal lobe (right anodal, left cathodal) improves visual memory (Chi et al., 2010), cathodal stimulation of the temporo-parietal cortex reduces the magnitude of facial adaptation (Varga et al., 2007). Also, anodal stimulation improves implicit learning when the left prefrontal cortex is stimulated (Kincses et al., 2004) and enhances the recognition of facial expression when right OFC is stimulated (Willis et al., 2015). For a review see Antal et al. (2011) and Costa et al. (2015).

Effects of tDCS might be explained by the modulation of the resting membrane potentials of the stimulated area. Single cell recording studies have shown that cathodal stimulation can decrease firing activity, while the anodal stimulation have the opposite effect (Bindman et al., 1964; Purpura and McMurtry, 1965). In humans the tDCS has similar polarity dependent effects (Nitsche and Paulus, 2000, 2001). It seems that tDCS effects appear to be site specific but not site limited; the latter effects might be based on plasticity mechanisms.

Since tDCS seems to be a powerful technique for investigation visual processing, we applied cathodal or anodal tDCS and sham stimulation as a control in a decision making test, over the OFC (Nitsche et al., 2008; Dayan et al., 2013; Manuel et al., 2014; Willis et al., 2015). Our subjects were required to make a judgment on the real size of objects seen on the screen, i.e., whether they fit in a shoebox or not? There were two sessions; between the two sessions tDCS stimulation was applied.

There are two possible scenarios concerning the outcome. If stimulation of the OFC does not have an effect on decisions concerning both MC and PC optimized stimuli, or if the effects are similar using both stimuli that would support the idea that fast MC information is processed through the ventral pathway avoiding the OFC. Thus, only decision mechanisms were affected, but not the route of information flow. If, on the other hand, decisions about MC stimuli were affected selectively, it would support the hypothesis that MC information reaches the OFC, passes through it and is available for top-down modulation (Bar et al., 2006).

Before the stimulation, the three groups of volunteers performed the task under identical conditions (n = 48). Paired t-test was used for the statistical evaluation. The percentage of correct answers was 91.50 ± SD = 4.05 using MC stimuli, comparing with accuracy of PC stimuli (mean 90.06, ± SD = 4.69) the difference was not significant p = 0.12 (df = 47, t = 1.58, Figure 4A). Decisions about stimuli optimized for the MC yielded shorter response latencies than those for PC stimuli (mean MC latency = 0.90 s, ±SD = 0.20 s, mean PC = 0.98 s, ± SD = 0.23 s, p < 0.01, df = 47, t = −3.95, Figure 4B). These results suggest that the reaction time differences originate from the different processing times needed for MC and PC optimized stimuli, not from the differences in the recognizability of the MC and PC stimuli sets. This test verified that MC optimized stimuli are associated with shorter response latencies (Bar et al., 2006).

Here we report that we could selectively modulate the processing of magnocellular optimized stimuli by influencing the activity of the prefrontal cortex using tDCS. This result confirms the hypothesis that magnocellular information passes the orbitofrontal cortex, and therefore might be used for a top-down modulation of visual processing.

Several points have to be addressed when discussing the results.

The first question is whether our stimuli fit for the magno- and parvocellular pathways? It has been reported earlier that decisions concerning MC optimized stimuli are faster than those optimized for PC stimuli (Kveraga et al., 2007a,b). Our results confirmed that the stimuli used in this study are indeed suitable for driving the dorsal or ventral pathway specifically. The significant difference in response latency times before the stimulation favored MC optimized stimuli but did not favor PC optimized stimuli, indicating that pathway optimization was successful.

TDCS had a clear and significant effect on response accuracies. How can this be interpreted? The rationale behind our study was that transcranial stimulation may have a direct impact on baseline cortical excitability (Stagg and Nitsche, 2011) and the observation that predictions might accelerate the perception of our environment by pre-stretching or priming bottom-up processing. Most studies agree that the phenomenon is based on the information carried by the MC. The MC and the dorsal pathway, however, also feed information into the ventral, PC through different stages of the cortical visual system (Merigan et al., 1993; Chen et al., 2007) but it is not clear what the exact source of this information is. Is MC information processed simultaneously, together with PC information in the ventral pathway (Mace et al., 2005; Fabre-Thorpe, 2011) or does MC information arrive through top-down connections at the IT via the OFC (Bar et al., 2006; Kveraga et al., 2007a,b)? The question is further complicated by the observation that connections between areas V5, V4 and the IT, furthermore between the prefrontal cortex and the IT can facilitate object recognition (Tomita et al., 1999; Chen et al., 2007; Eger et al., 2007). Cathodal stimulation of the OFC exerts an inhibitory effect, since neurons under the stimulation electrode become less excitable and presumably decrease the level of neurotransmitter glutamate (Filmer et al., 2014). Anodal stimulation in our experiments supported OFC functions: accuracy improved considerably for LSF stimuli (HSF stimuli were not affected), while cathodal stimulation decreased accuracy. This is in line with the meta-analysis data reported by Jacobson et al. (2012), namely, in cognitive tasks anodal stimulation often improves performance. Also, several studies report a decrease in performance when applying cathodal stimulation (e.g., Stone and Tesche, 2009; Sparing et al., 2009; Kraft et al., 2010). While this might not be the case in general, i.e., that anodal stimulation improves, cathodal stimulation impairs cognitive function, in some cognitive fields like perception and attention studies the likelihood to get opposite effects after anodal and cathodal stimulation, respectively, is exceptionally high (Jacobson et al., 2012).

The OFC consists of two large regions: medial and lateral parts. The former plays a role in higher cognitive functions, associative, reward linked learning, processing emotions, integrating sensory modalities and, most importantly, making decisions (Kringelbach and Rolls, 2004; Wallis, 2012). The fact that stimulation affected only decisions about LSF images supports the idea that magnocellular information passes the OFC. According to Bar et al. (2006) this information might be used for top-down facilitation of decision making. The role of the OFC in decision making especially when previous knowledge or predictions are concerned was studied in fMRI experiments (Summerfield et al., 2006; Miall et al., 2014; Erez and Duncan, 2015).

The last question is how tDCS influences the motor cortex and thus behavioral response latencies? Response latency in psychophysical studies includes sensory processing, decision making and motor response. When interpreting our results, one must also consider that the arrangement of electrodes for modulating the OFC (Manuel et al., 2014) stimulates the motor cortex when cathodal stimulation is used, but inhibits it when anodal stimulation is applied. Results regarding the effects of tDCS on motor reactions are far from clear. The main effect of tDCS is biasing cortical excitability. The underlying mechanism is still debated but current work suggests that it shares similarities with the activity-dependent synaptic plasticity (Dayan et al., 2013). Most studies agree that there is a large variability among subjects when evaluating the effects of stimulation (e.g., Wiethoff et al., 2014; Pope et al., 2015; Davidson et al., 2016). The situation is further complicated by the fact that the same stimulating pair of electrodes will have obviously opposing effects on the motor cortex and on the OFC; factors influencing the motor component of the decision and responding process thus might mask the effects on the sensory part. In a meta-analytical review Jacobson et al. (2012) concluded, that it is quite common to see the AeCi effect (anodal stimulation, cathodal inhibition) on latency times in motor experiments where evoked potentials are studied; in this respect our study might be an exception, since no significant differences in response latencies could be shown. We have to note however, that only behavioral response latencies and no evoked potentials were analyzed in this study.

In summary, our behavioral results show that using these electrode positions we could modulate the cortical activity of the OFC, which has an effect on the top-down mechanism during the fast categorization of MC optimized stimuli (Bar et al., 2006). Our results do not exclude the possibility that magnocellular input fed into the ventral pathway may accelerate visual processing, but they give further evidence for the essential role of top-down processes originating from the OFC in visually based decisions. The goal of our study was to investigate the effects of bilateral stimulation of the orbitofrontal cortex, but for the correct interpretation of the reaction time changes another electrode arrangement is needed. Using electrodes on the two sides of the supraorbital region (Kincses et al., 2004; Fecteau et al., 2007; Ferrari et al., 2015) could enable the examination of dynamic changes of magnocellular processing and the differences between the function of the left and right OFC. However, the exact neuronal background and tracking the flow of information along the cortical pathways require electrophysiological methods (extracellular unit recording at several locations simultaneously) with a good temporal resolution.

AB: design of the work, critical revision, final approval, accountable for all aspects GCs: data acquisition, first draft, final approval, accountable for all aspects. MN: data acquisition, critical revision, approval, accountable for all aspects. PC: intrepretation of data, draft, approval, accountable for all aspects TK: intrepretation of data, draft, approval, accountable for all aspects GyS: statistical analysis, draft, approval, accountable for all aspects.