In Kannada, visual features are arranged in blocks called "akshara," making this a visually more complex writing system than typical alphabetic orthographies. Akshara knowledge was assessed concurrently and 8 months later in 113 children in the first years of reading instruction (aged 4-7 years). Mixed effects logistic regression models showed that both symbol-level (visual complexity, phonological complexity, frequency) and child-level (phoneme awareness, vocabulary, rapid automatized naming) measures predicted variations in akshara knowledge. Thus predictors of akshara knowledge appear to be related to aspects of both visual and phonological processing. It is argued that the visual complexity of the Kannada script pose somewhat different challenges to the child learning to read than those reported for visually less complex orthographies.