Recent research suggests that performance on Statistical Learning (SL) tasks may be lower in children with dyslexia in deep orthographies such as English. However, it is debated whether the observed difficulties may vary depending on the modality and stimulus of the task, opening a broad discussion about whether SL is a domain-general or domain-specific construct. Besides, little is known about SL in children with dyslexia who learn transparent orthographies, where the transparency of grapheme-phoneme correspondences might reduce the reliance on implicit learning processes. The present study investigates the impact of SL in Spanish, a transparent orthography, among 50 children aged 9 to 12 years, with and without dyslexia. For this purpose, we used four SL tasks to evaluate two modalities (auditory/visual) and two stimulus type (verbal/nonverbal) and evaluated both accuracy and response times on each condition. The findings reveal that children with dyslexia in Spanish exhibit lower performance on SL tasks (accuracy) compared to the control group, regardless of the modality and stimulus type used. However, children with dyslexia struggle the most with tasks that involve visual material. This indicates that children with dyslexia in transparent orthographies have particular difficulties in extracting distributional probabilistic information in the absence of explicit learning instructions. Notably, difficulties were more pronounced in visual tasks involving verbal stimuli. The present results help to better understand the underlying mechanisms involved in reading acquisition in children with dyslexia.