This research aimed to design a suitable pedagogic study on the effects of different error correction conditions (ECCs) on learner-initiated noticing. With this in mind, I sought to tackle comprehensive (eventually replaced by semi-comprehensive) error correction (EC) in an open, uncontrolled and learner-centred writing task. Despite numerous studies on written corrective feedback (WCF), little research has paid attention to what learners ‘notice’ while writing. Storch and Wigglesworth (2010) explain that noticing and the “processing of feedback [are] … less … researched … because it is difficult to access such learner-internal cognitive processes” (p. 305). Meanwhile Santos, López-Serrano and Manchón (2010 p. 132) indicate that the self-initiated character of writing problems turns writing into the perfect setting to study self-initiated noticing and focus-on-form processes. Framed within the ‘noticing’ debate, the ‘language learning potential of writing’ and the ‘writing-to-learn’ and ‘feedback-for-acquisition’ dimensions, this quasi-experimental study investigates how different ECCs influence the error types that learners ‘attend to’. The four-stage (composing/ error correction-noticing/ rewriting/ new writing) design involved 60 EFL undergraduates in three semi-comprehensive ECCs (Direct EC, Reformulation and Self-correction). Opinion essays, noticing sheets and a questionnaire constituted the data to analyse. The quantitative results support Truscott’s claim for the little value placed on ‘[semi-] comprehensive’ EC in grammar accuracy improvement. Qualitative analyses showed that the ECCs tested had different effects on learners’ noticing. Delayed self-correction elicited the most attention to form and is suggested as a more refined way to ‘notice the hole’; direct EC led to retention, reformulation resulted in the most engaging ECC.