For individuals with grapheme-colour synaesthesia, letters, numbers and words elicit vivid and highly consistent colour experiences. A critical question in determining the mechanisms underlying the phenomenon is whether synaesthetic colours arise early in visual processing, prior to the allocation of focused attention, or at some later stage following explicit recognition of the inducing form. If the synaesthetic colour elicited by an achromatic target emerges early in visual processing, then the target should be relatively easy to find in an array of achromatic distractor items, provided the target and distractors elicit different synaesthetic colours. Here we present data from 14 grapheme-colour synaesthetes and 14 matched non-synaesthetic controls, each of whom performed a visual search task in which a target digit was distinguished from surrounding distractors either by its unique synaesthetic colour or by its unique display colour. Participants searched displays of 8, 16 or 24 items for a specific target. In the chromatic condition, target and distractor digits were presented in different colours (e.g., a yellow '2' amongst blue '5's). In the achromatic condition, all digits in the display were black, but targets elicited a different synaesthetic colour from that induced by the distractors. Both synaesthetes and controls showed the expected efficient (pop-out) search slopes when the target was defined by a unique display colour. In contrast, search slopes for both groups were equally inefficient when the target and distractors were achromatic, despite eliciting distinct colours for the synaesthetes under normal viewing conditions. These results indicate that, at least for the majority of individuals, synaesthetic colours do not arise early enough in visual processing to guide or attract focal attention. Our findings are consistent with the hypothesis that graphemic inducers must be selectively attended to elicit their synaesthetic colours.