The 'early vs. late selection' attentional debate dates back to pioneering studies in the field of cognitive psychology. In 2004, a seminal paper was published in Psychological Review under the title Forty-five years after Broadbent (1958): Still no identification without attention (Lachter, Forster & Ruthruff). This paper argued that the ample evidence apparently demonstrating unattended word processing was in fact accounted for by rapid spatial-attention shifts to these words. Approximately at the same time, a new series of studies came out arguing that, in contrast to arbitrary or symbolic stimuli (such as letters or words), images of more 'natural' objects and scenes are in fact categorized and recognized ultra-rapidly, in the absence of focal attention (e.g., Li et al., PNAS, 2002). Other studies, however, failed to demonstrate such unattended high-level processing, bringing the 'early vs. late selection' dispute back to front stage. Here, I will present my own series of studies examining object and scene categorization, under various 'attended' vs. 'unattended' conditions. Despite some strong recent arguments in favor of automatic guidance of attention by scene semantics (e.g., Henderson & Hayes, Nat Hum Behave., 2017), my findings converge on Lachter et al's conclusions showing no categorization and identification of objects outside the main focus of spatial attention. I will share my insights on this disputable topic, and some further, open questions.