We investigated whether covert ensembles of high-(emotion), and low-level (brightness) visual information, extracted from peripheral faces (presentation/encoding:200 ms), unintentionally influences perception of a central target face stimulus in individuals typically developing (TD) and with autism spectrum condition (ASC). Graded alterations in the summary intensities of the emotion and brightness of the peripheral stimuli modulated the perceptions of the target face in both TD and ASC. In contrast, when we measured goal-directed (overt) ensemble face-emotion and brightness perception, we found that in half of ASC the overt ensemble emotion perception was impaired than TD. Additionally, we repeated both experiments with a backward visual mask to restrict not just encoding but also background processing in the visual system to 200 ms. This revealed that the effect of peripheral ensembles on centre perception was present only with brightness at least in TD but of overt ensembles was evident with both emotion and brightness in TD and ASC alike. These results suggest that while ensemble statistics of low-level information derived automatically and rapidly (200 ms) from contextualized faces are used for target face perception, the same takes longer with high-level information. However, overt facial ensembles are rapidly processed in both TD and ASC.Humans can efficiently extract summary statistics (e.g. mean, variance, range) from groups of visual entities, also known as 'ensemble representation' 1 . This summarizing has not only been reported across simple objects and low-level visual features, such as size 2,3 , orientation 4,5 , brightness 6 , and colour 7 , but also higher-level visual features, such as facial-identity 8,9 , gender 10 , attractiveness 11 , and emotion 10,12 . Forming ensemble representation across a plethora of visual information allows the brain to statistically summarize related and relevant information into meaningful percepts that can efficiently guide behaviour and is generally accepted as a 'global representation' 1 . However, if and to what extent the ensemble (global) representation computed from parts of an image influences the perception of a given target (local visual target of interest) within that image has been less explored to date.Some human studies have demonstrated that the above phenomenon operates with various low-to mid-level features and objects as visual stimuli. Memory of features of an object (e.g. colour) was reported to be influenced by spatial context 13 , while remembered size of individual circles was biased by the mean of other circles 14 . Likewise, spatial frequencies of gabor-patches were biased by the previously presented patches 15 . Influence on visual short-term memory of shape, orientation, and colour of objects was found by manipulating the respective features of the non-target object sets in a change detection task 16 . More recently, this phenomenon was also reported with high-level visual stimuli, such as faces which were perceived as more attractive in ...