Understanding how listeners process speech under noisy conditions is critical for refining models of speech perception and lexical access. This experimental study investigated whether increasing background noise affects listeners’ reliance on consonants versus vowels during lexical reconstruction. Native Dutch participants completed a nonword reconstruction task under three conditions: no noise, low noise, and high noise. They were asked to convert nonwords into real Dutch words by altering a single phoneme, either a consonant or a vowel. While no clear phoneme preference emerged in the no-noise condition, vowel substitutions declined significantly under both noise conditions, suggesting a shift toward greater reliance on consonant cues. This shift did not significantly differ between the low- and high-noise conditions, possibly indicating a threshold effect in response to auditory degradation. These findings point to the need for refinements in current models of lexical access and support the idea that listeners increasingly rely on predictive mechanisms when auditory input is compromised.