Objective: Changes in second formant frequency fluctuations (FFF2) were examined in adults who stutter (AWS) and adults who do not stutter (ANS) when producing nonwords under varying emotional conditions. Methods: Ten AWS and 10 ANS viewed images selected from the International Affective Picture System representing dimensions of arousal (e.g., excited versus bored) and hedonic valence (e.g., happy versus sad). Immediately following picture presentation, participants produced a consonant-vowel + final /t/ (CVt) nonword consisting of the initial sounds /p/, /b/, /s/, or /z/, followed by a vowel (/i/, /u/, /ε/) and a final /t/. CVt tokens were assessed for word duration and FFF2. Results: Significantly slower word durations were shown in the AWS compared to the ANS across conditions. Although these differences appeared to increase under arousing conditions, no interaction was found. Results for FFF2 revealed a significant group-condition interaction. Post hoc analysis indicated that this was due to the AWS showing significantly greater FFF2 when speaking under conditions eliciting increases in arousal and unpleasantness. ANS showed little change in FFF2 across conditions. Conclusions: The results suggest that AWS’ articulatory stability is more susceptible to breakdown under negative emotional influences.