Writing systems vary in the way they express the sounds and meanings of spoken language. Alphabetic writing systems contain information about phonological structure within the orthography due to systematic relations between graphemes and phonemes. Logographic writing systems encode less-fine-grained information about phonological structure via more arbitrary mappings between characters and syllables. We tested whether differences in orthographic structure impact on reading acquisition and the spoken language representations that underpin reading (e.g., Rastle et al., 2011). 24 adults were trained on two artificial languages with alphabetic and logographic writing systems. Each language contained 24 words denoted by phonological, orthographic, and semantic components. Learning was measured using computerised training tasks over 10 days before performance was assessed using behavioural tests. Following training, neural activity was recorded using fMRI whilst participants made meaning judgements about trained spoken and written stimuli. We assessed the development of mappings between sounds, spellings, and meanings of words and the division of labour between dorsal and ventral reading pathways. Performance was compared using linear mixed-effect models and we contrasted neural activity using paired-samples t-tests. Representational similarity analysis assessed whether alphabetic and logographic systems influence neural sensitivity to phonemic, orthographic, and semantic structure during reading and listening. Overall, alphabetic words exhibited stronger orthography–phonology mappings while orthography–semantic mappings were stronger for logographic words. The dorsal pathway showed greater activity for alphabetic written words; the ventral pathway was more active for logographic written words. Representations only encoded the phonemic structure of alphabetic written words and the orthographic structure of alphabetic spoken words. These findings advance our knowledge of how writing systems shape reading acquisition and spoken language representations. They suggest different strategies are used to learn alphabetic and logographic languages and that orthographic transparency can impact on the division of labour and nature of neural representations.