When looking for a word, you may have resorted to producing its sound instead, for example saying “quack” instead of “duck”. Such lexicalisations of real-word sounds are called onomatopoeias, and are mainly used by children before they learn the arbitrary words, typically calling a dog “woof”. Research has shown that onomatopoeic words are easier to learn at any age, supporting the grounded cognition theory in which language and sensory-motor systems interact. However, little is known about the role of onomatopoeias in the adult language and their interplay with perception. To help fill this gap, we compared visual detection of animal and object photographs under conditions of reduced attention in an attentional blink task. The photographs (e.g., a phone) were preceded either by a word (e.g., “phone"), a sound (e.g., of a telephone), an onomatopoeia (e.g., “dring”), or neutral versions (prosody-only). We found no advantage of iconic language on perception. Instead, inanimate onomatopoeias were associated with reduced detection rates compared to every other type of cue (animate onomatopoeias, words and sounds), suggesting that inanimate onomatopoeias interfere with perception, supporting the grounded cognition theory. Considering the higher number of false alarms after onomatopoeic cues, it seems that they elicit higher mental imagery without enhancing detection, but such lower accuracy could also be explained by a general lower familiarity by adults with this type of words.