Autism spectrum conditions (ASC) are characterized by atypical neurodevelopmental patterns, often leading to impairments in social interactions, communication, and inflexible behavior. Additionally, perceptual anomalies are an important aspect of ASC (Dakin and Frith, 2005; Simmons et al., 2009; Mottron, 2011; Robertson and Baron-Cohen, 2017), as outlined by the weak central coherence framework (Happé and Frith, 2006) and the enhanced perceptual functioning hypothesis (Mottron et al., 2006). The core feature of these theories is the abnormal interaction between bottom-up and top-down processes (Pellicano and Burr, 2012).

During low-level (early-stage) visual perception, the sensory system automatically extracts elementary information from external objects (e.g., lightness, color, stereopsis, and motion). Low-level vision is bottom-up and data-driven because it originates with the stimulation of rods and cones in the retina, and this information is immediately forwarded to the visual cortex via the lateral geniculate nucleus. In contrast, top-down processes refer to the effect of previous knowledge and mental effort on perception and recognition. In other words, the brain makes predictions and inferences based on past experiences and memories (Pellicano and Burr, 2012). A typical example is mental imagery (visualizing or “seeing in the mind’s eye”), which is similar to perceptual experience, but occurs in the absence of external objects: the individual retrieves images from the memory and intentionally maintains this information in the focus of consciousness (Kosslyn et al., 2006).

A paradigm shift in the research of sensory cortical areas led to the recognition that the primary visual cortex (V1) is not a passive “blackboard” for the bottom-up perception of basic object features. It turned out that top-down signals from higher-level cortical areas (e.g., prefrontal cortex and anterior cingulate cortex) modulate neural activity even in the V1 during working memory, retrieval, and mental imagery (Pearson et al., 2015; Roelfsema and de Lange, 2016). Caron et al. (2006) showed enhanced performance in the detection of simple visual stimuli and superior discrimination of first-order gratings in ASC, which indicates a heightened functioning of V1. However, despite extensive research efforts, we behold scarce knowledge about the interaction between bottom-up and top-down processes in the earliest stage of visual information processing in ASC.

The lateral masking task provides a unique opportunity for the behavioral assessment of bottom-up perception and top-down imagery at the level of the V1 (Ishai and Sagi, 1995; Kéri, 2015; Maróthi and Kéri, 2018; Figure 1). In the perceptual part of the task, participants detect a central low-contrast target (Gabor patch) vertically flanked by two high-contrast masks. Gabor patches are small, bean-shaped objects consisting of alternating, brighter and darker regions, providing an optimal stimulus for the V1 (Ishai and Sagi, 1997; Figure 1). The luminance contrast of the target Gabor patch is defined by the difference in its brightness and that of the background. The flanking masks facilitate the detection of the central target if they fall within an intermediate distance from the target: participants are able to detect a target with low contrast. However, the masks have an inhibitory effect if the target-mask distance is small, that is, a high contrast is necessary for target detection. It has been postulated that the effect of masks can be attributed to lateral interactions between neuronal groups and their feedback modulation in the V1. These lateral (horizontal) interactions are thought to be mediated by short-range connections between neurons responding to similar visual stimuli (Polat and Sagi, 1993; Polat et al., 1998; Angelucci et al., 2002; Crook et al., 2002). Interestingly, we can observe similar masking effects when participants only imagine the previously presented masks, which indicates an interplay between bottom-up and top-down processes (Ishai and Sagi, 1995). At the neural level, cortical cells responding to simple visual stimuli and their lateral interactions might be activated by signals from higher-level extrastriate and prefrontal areas in the absence of external stimuli (Freeman et al., 2003). Moreover, the imagery-induced facilitation of target detection persists for several minutes after viewing the masks, which may point to the existence of a monocular, orientation-specific, low-level iconic memory system that stores the sensory trace of the masks and enables the reactivation of these low-level representations during top-down mental imagery (Ishai and Sagi, 1995). It has also been shown that the top-down imagery effect is weaker than the facilitation induced by the physical presence of the masks and is especially pronounced at intermediate target-mask distances (Ishai and Sagi, 1997). It must be underlined that higher-level top-down control during imagery is different from early-stage feedback mechanisms related to short-term memory and to the modulation of target-mask interactions (Gilbert et al., 2000; Angelucci et al., 2002; Freeman et al., 2003; Summerfield and Egner, 2009). Anatomical and physiological measurements from the macaque mapped the role of intra-areal V1 lateral connections and inter-areal feedback connections to V1 in spatial summation. Monosynaptic lateral connections within V1 mediated interactions in the spatial summation field of neurons, but feedback circuits from extrastriate cortex to V1 were needed for the full spatial range of center-surround interactions necessary for the contextual modulation and global-to-local integration of visual signals (Angelucci et al., 2002).

Kéïta et al. (2011) applied the lateral masking task to assess the functional integrity of lateral interactions in the visual cortex of individuals with ASC. The authors measured contrast thresholds for a centrally presented target Gabor patch flanked by collinear masks at different distances. As expected, both ASC and control groups showed heightened target sensitivity when the target-mask distance was intermediate. Strikingly, this facilitation was significantly greater in the ASC group relative to the control group (Kéïta et al., 2011). The authors concluded that atypical visual functions in ASC can be explained by altered lateral interactions in the visual cortex responsible for the earliest stage of feature extraction (e.g., luminance, hue, spatial frequency, and orientation of objects). However, other groups failed to replicate these findings (Jachim et al., 2015; Dickinson et al., 2018), possibly because of substantial differences in samples and psychophysical methods. There are no studies aimed to evaluate imagery in the same experimental setting.

To resolve these controversial results, it is critical to take into consideration that individuals with ASC exhibit a high degree of variation in the development of visuospatial functions (Muth et al., 2014), and it may be related to mental imagery (Soulières et al., 2011). Some of them display outstanding performances (phenotypic peaks) on tasks assessing visuoconstructional abilities (e.g., Block Design and Raven’s Progressive Matrices) relative to other cognitive functions (Caron et al., 2006). The Block Design test is a part of non-verbal IQ assessment, during which participants are asked to rearrange blocks with different color patterns on their sides to match a predefined template (Wechsler, 1997). Regarding mental imagery, individuals with ASC who exhibit a Block Design peak performed better on a mental rotation task than non-autistic controls and ASC individuals with no Block Design peak. This indicates their heightened imagery ability to form, access, and manipulate mental images (Soulières et al., 2011), which is in line with the classic idea that some people with ASC think in pictures (Grandin, 1995; Kana et al., 2006; Heaton et al., 2008; Soulières et al., 2009; Sahyoun et al., 2010). However, it is not known whether enhanced visuoconstructional and imagery abilities are related to the earliest level of visual information processing, and how top-down factors interact with bottom-up perception.

Therefore, we tested the following hypotheses: (1) individuals with ASC exhibiting a Block Design peak (ASC Peak) show higher facilitation in the lateral masking task relative to neurotypical controls and ASC with no such phenotypic peak (ASC Non-Peak); and (2) based on the results of Soulières et al. (2011), we hypothesized that individuals with ASC Peak show enhanced mental imagery abilities.

The results from the control group are in keeping with the findings of prior experiments (Polat and Sagi, 1993). However, we only found partial support for our hypotheses in ASC. In accordance with the results of Kéïta et al. (2011), we observed higher facilitation of contrast detection in ASC relative to the controls when the target-mask distance was intermediate (3λ). However, this effect was confined to the ASC Peak group. In addition, Block Design performances specifically correlated with lateral facilitation in the ASC Peak group. This suggests that altered developmental trajectories, leading to outstanding higher-level visuoconstructional abilities, may also result in the enhancement (less inhibition) of lateral connections in the V1. From another point of view, atypical higher-level perception may stem from altered low-level processes in ASC (Perreault et al., 2015). Vandenbroucke et al. (2008) used an electrophysiological and neural network approach to map feedforward, horizontal, and recurrent feedback processing in ASC. They found abnormal object boundary detection as early as 120 ms after stimulus presentation, which may be a marker of dysfunctional lateral connections in early visual areas. Interestingly, ASC individuals were characterized by an enhanced subsequent occipital activity (225 ms), whereas recurrent feedback processing from extrastriate areas (260 ms) was spared (Vandenbroucke et al., 2008).

We also found weak inhibition at small target-mask distances (1λ) in the ASC Peak group. This is consistent with a previous functional magnetic resonance imaging study demonstrating a correlation between weaker surround suppression in the V1 and autistic traits in the general population (Flevaris and Murray, 2014), although brain imaging studies using the lateral masking paradigm have not been performed. Therefore, the interpretation of our behavioral results at the neuronal level is indirect. However, Dickinson et al. (2018) showed that greater severity of autistic symptoms was associated with increased short-range lateral inhibition. The authors concluded that lateral connections were not generally dysfunctional in ASC, and subtle alterations and individual variations might explain the heterogeneity of visual perceptual phenotype (Bertone et al., 2005; Dickinson et al., 2018). These divergent results may stem from different methods and from the heterogeneity of ASC.

Classic models suggest that the facilitatory effect of collinear masks is due to the spreading of excitatory signals from the mask-responsive to the target-responsive cells via lateral connections in the V1 (Livingstone and Hubel, 1984; Gilbert and Wiesel, 1989; Polat and Sagi, 1993). Feedback from higher cortical areas may strengthen the attentional modulation of collinear masks and targets (Gilbert et al., 2000; Freeman et al., 2003), although lateral facilitation seems to be too fast for such feedback effects (Yu et al., 2002). Specifically, lateral facilitation by collinear masks was demonstrated at a stimulus duration as brief as 8 ms, a timing too short for cortical feedback (Yu et al., 2002). According to the models of temporal dynamics, there are two possible components of facilitation. The first is a fast and short-acting component of facilitation that involves a single spatially-elongated perceptual channel defining the classic receptive field without long-range connections (Georgeson and Georgeson, 1987). The second is a delayed and long-acting component that is based on long-range connections between collinear flanker masks and target (Polat and Sagi, 2006). Huang and Hess (2008) showed that these models are not sufficient and that the dynamics of perceptual facilitation are both fast and sustained. The authors proposed a two-component model of facilitation consisting of a rapid signal across large retinal distances based on feedback from higher centers and a sustained, lower-level response involving the temporal integration of locally responsive mechanisms (Huang and Hess, 2008). Therefore, the main point of the Huang and Hess (2008) model, which is based on a series of psychophysical studies, is that facilitation is neither delayed and long-lasting, not fast and short-lived. Instead, facilitation is fast and sustained. Contrary to the classic view, the fast signal is based on feedback, whereas the sustained response is mediated by low-level mechanisms. In sharp contrast to lateral facilitation, masks can inhibit target detection if the target-mask distance is small (<2λ). This lateral inhibition is thought to be related to short-range inhibitory connections (Blakemore et al., 1970; Polat and Sagi, 1993; Shushruth et al., 2013). At the behavioral level, our results indicate that both excitatory and inhibitory lateral connections are altered in ASC Peak, characterized by enhanced excitation at intermediate target-mask distances (3λ) and reduced inhibition at small target-mask distances (1λ). This raises the possibility that autism-related developmental alterations in visual cortical circuits are not confined to a single mechanism, and that these can be detected especially in individuals with ASC who display exceptional visuoconstructional abilities.

We also identified a robust dissociation between perception and imagery: individuals with ASC, irrespective their visuoconstructional abilities, displayed a negligible imagery effect. It is possible that they failed to imagine the masks or the imagined masks failed to modulate the target. However, given that there were normal (ASC Non-Peak) or enhanced (ASC Peak) lateral interactions in the perceptual task, it is not likely that successfully imagined masks failed to act on the target. Overall, the dissociation between perception and imagery indicates enhanced early bottom-up perception and dampened top-down control in the ASC Peak group, whereas in the ASC Non-Peak group, weak top-down control was accompanied by unaltered perception.

Our results provide a plausible explanation for an important controversy in the literature. While Kéïta et al. (2011) reported increased lateral facilitation in individuals with ASC relative to controls, Jachim et al. (2015) found the opposite phenomenon, that is, weak facilitation in ASC. Notably, the sample of Jachim et al. (2015) included individuals with Asperger’s syndrome, which was not the case in the Kéïta et al.’s (2011) study. Differences in the experimental design could also contribute to the discrepancy of results. Kéïta et al. (2011) used long exposure time and added gray scale noise to the stimulus display, raising the possibility that noise differently interfered with perceptual processing in ASC and healthy controls. In our study, the stimulus exposure time was short, and no noise was applied, excluding the possibility that these parameters accounted for the difference in previous studies. Moreover, we controlled individual variations in cognitive developmental patterns (ASC Peak vs. ASC Non-Peak). However, it cannot be excluded that individuals with Asperger’s syndrome exhibit a distinct pattern of lateral masking performance. Furthermore, it is important to take into account that our participants were male, and the results may not be generalized to female individuals.

The finding that heightened Block Design performance was associated with poor imagery seems to be counterintuitive, because in a previous study Block Design performance was related to a more efficient imagery and manipulation of mental representations (Soulières et al., 2011). The authors conducted mental imagery experiments in 23 individuals with ASC (11 with Block Design peak) and 14 matched neurotypical controls. In the first experiment, participants were asked to imagine a letter inside a circle, whereas the second experiment included mental rotation tasks with two- and three-dimensional shapes, hands and letters. Individuals with ASC, especially those with Block Design peak, were more accurate in the imagery tasks relative to the controls. These results can be explained in the framework of a global advantage in perceptual processing (enhanced perceptual functioning model) or by an eminence in veridical mapping, which refers to the ability to detect isomorphisms among objects (Soulières et al., 2011). However, mental imagery abilities depended on the task type: the performance was more pronounced when the task included the mental rotation of three-dimensional objects rather than more simple two-dimensional shapes, hands, or letters (Soulières et al., 2011). Therefore, it is possible that imagery is even less effective in the case of low-level luminance contrast gratings, as shown in the present study. At the neuronal level, there is evidence that people with ASC show higher activation in a widespread parietal and occipital network during the Block Design task relative to non-autistic controls, but this enhanced neuronal network does not include the V1, which is critical in the lateral masking task (Polat and Sagi, 1993; Kana et al., 2013).

Decreased performances on the mental imagery task may be related to the impaired top-down control of information processing in ASC. At the neuronal level, it may be due to deficits in visual cortical areas, control areas (prefrontal cortex), or the connectivity between the sensory and control regions (Frith, 2003; Loth et al., 2010; Cook et al., 2012; Van Der Cruys et al., 2018). Our study could not explore these possibilities at the neural level, and it is not clear whether individuals with ASC failed to imagine the masks or the successfully imagined mask failed to exert an effect on the target. However, previous studies indicated that people with ASC exhibit normal (Kana et al., 2006) or even enhanced visual imagery (Soulières et al., 2011) using more complex stimuli. It has also been shown that non-visual cognitive control might play a compensatory role in imagery in ASC. When participants were asked to memorize a map of a fictitious island, verbal IQ and working memory were positively associated with image scanning performance in ASC, but not in healthy controls (Maras et al., 2014). In the present study, verbal compensation was not possible because of the simplicity of stimuli.

In the framework of the Bayesian decision theory, Pellicano and Burr (2012) suggested that atypical perceptual experiences in ASC can be explained by a deficient interaction between incoming sensory information and a prior knowledge of the world. Specifically, attenuated top-down influences (Bayesian “hypo-priors”) may induce a hyper-realistic perception, which is weakly influenced by past experiences. In our paradigm, prior experiences did not influence the detection of the target stimulus when participants were requested to retrieve and to imagine the previously presented masks. This suggests that the diminished effect of prior experiences can be revealed at the early stage of visual processing in ASC.

The general loss of imagery in ASC may stem from nonspecific factors (e.g., a misunderstanding of the instructions or a lack of motivation), and this is a main limitation of the study. However, individuals with ASC excellently performed on challenging neuropsychological and psychophysical tests in this study, even displaying better performances than healthy control subjects in some domains. Generally, as discussed above, individuals with ASC do not show impaired visual mental imagery (Kana et al., 2006; Soulières et al., 2011; Maras et al., 2014). The imagery condition was matched to the perception condition in task structure and general difficulty, and we made sure during the practice block that individuals with ASC understood the task correctly. Moreover, in our previous study, individuals with schizotypal personality disorder, who often find testing situations very challenging due to their decreased motivation, distractibility, and suspiciousness, exhibited enhanced (more effective) mental imagery relative to control participants, which is against the possibility that the imagery task is more difficult to understand or motivationally more demanding than the perceptual task (Maróthi and Kéri, 2018). Finally, a rating scale for fatigue and motivation did not indicate significant differences between individuals with ASC and healthy control participants.

In conclusion, our results provide psychophysical evidence for a unique pattern of lateral interactions in the V1 of individuals with ASC Peak: a decreased inhibition at short distance (1λ of target-mask distance) and an enhanced facilitation at intermediate distances (3λ of target-mask distance). The decreased horizontal inhibition in early-stage vision may be relevant to several assumptions of the enhanced perceptual theory and the weak central coherence hypothesis of autism (Happé and Frith, 2006; Mottron et al., 2006). Because of the decreased inhibition of adjacent stimuli (masks) on a central target, ASC Peak individuals may exhibit a detail-focused processing rather than a holistic perception of larger images (Happé and Frith, 2006). Specifically, decreased lateral inhibition exerted on a small spatial region (i.e., the target Gabor patch in the lateral masking task) may lead to the pop-out and enhanced perceptual awareness of this spot. From another point of view, low inhibition and high facilitation in the lateral connections of the V1 may lead to a better detection of elementary visual features and simple contours (Mottron et al., 2006; Loffler, 2008). The biological bases of these findings remain to be explored: inhibition-excitation ratio in the V1 is a possible target for future investigations (Robertson et al., 2016).

The datasets generated for this study are available on request to the corresponding author.

The studies involving human participants were reviewed and approved by National Medical Research Council. The patients/participants provided their written informed consent to participate in this study.

SK, KC, and RM designed the study. RM coordinated data collection and measurements. RM and SK analyzed the data. RM, KC, and SK wrote the first draft of the article, which was reviewed, edited, and approved by both authors.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.