Skip to main content
. Author manuscript; available in PMC: 2026 Sep 24.
Published before final editing as: Nat Hum Behav. 2026 Aug 13:10.1038/s41562-026-02546-w. doi: 10.1038/s41562-026-02546-w

Figure 1. Pseudo-word repetition task to study planning and execution for speech production.

Figure 1.

a. Speech planning and execution at the sublexical level (post word formation) involves the composition of a syllable frame and the hierarchical incorporation of phonemes within this frame for the execution of the ordered speech sequence. Graphical representation of speech neural mechanisms from the planning of intended speech units to the execution of speech sound sequences. Speech production involves planning prior to execution, and execution which involves motor articulation followed by auditory monitoring of vocal feedback. b. Location of all 8106 recorded intracranial electrodes from 52 patients in both left and right hemispheres. c. Schematic of pseudo-word repetition task. Color bars indicate the duration of the auditory stimulus (blue), response time from the Go cue (orange), and response duration (green). d. Pseudo-word tokens used for the speech repetition task differed on their syllabic frame: consonant-vowel-consonant (CVC, green) and vowel-consonant-vowel (VCV, blue). The phonemes were chosen from a fixed for distribution for both consonants (5) and vowels (5) at each position within the pseudo-word. e. Histogram distributions of response duration for both CVC and VCV trials. The durations of VCV trials were significantly longer than the CVC trials (linear mixed effects [LME] modeling, t(9,451) = 38.3 β = 0.098, p < 0.001, 95% CI = 0.09–0.1, two-sided). The dark blue line indicates the beta estimate of the actual LME model, and the histograms (light blue) indicate the estimated distributions from permuted models (1000 permutations). f. Histogram distributions of response times for both CVC and VCV trials. The response time of CVC trials were significantly faster than the VCV trials (linear mixed effects [LME] modeling, t(9,451) = 8.56, β = 0.045, p < 0.001, 95% CI = 0.044–0.069, two-sided). The dark blue line indicates the estimate of the actual LME model, and the histograms (light blue) indicate the estimate distributions from permuted models (1000 permutations). Longer response times for di- vs. mono- syllabic utterances suggest that the syllable frame is used as a sub-lexical unit for speech planning.