According to our high-level visual dysfunction hypothesis, reading problems of some dyslexic readers are a salient manifestation of a more general deficit of visual cognition stemming from disrupted functioning of high-level visual regions. These regions play a fundamental role in the recognition of words and other objects. Supporting the hypothesis, some previous studies indicate that people with dyslexia have problems with tasks thought to rely on high-level regions of the ventral visual stream, but results are mixed. Here we report the first preliminary results of a large-scale preregistered study (https://osf.io/4dr3f) where we put the hypothesized deficient functioning of high-level visual mechanisms in dyslexia to a direct test using a combination of electrophysiological and behavioral methods. The study involves the recording of evoked visual potentials (both event-related and steady state) as well as behavioral measures on the processing of faces, objects, and words. Here, we focus on behavioral results for faces and houses. Participants performed a matching task with faces or houses of different spatial frequency content (low vs. high spatial frequencies). Accuracy was in general equivalent for high and low spatial frequency faces, but considerably poorer for low compared to high spatial frequency houses, consistent with the diagnostic role of low spatial frequencies for faces compared to other objects. Results furthermore indicate that dyslexic readers tend to be worse than typical readers at recognizing both high and low spatial frequency faces (large Cohen's d), while the group difference was smaller for high and low spatial frequency houses (small to medium Cohen's d), consistent with specific ties between visual face and word processing. Behavioral performance of the dyslexic group appears to be bimodal; some dyslexic readers perform well while others cannot do the task at all, suggesting that only some people with developmental dyslexia might have a disorder of high-level vision.