Probing sensorimotor memory through the human speech-audiomotor system

Abstract

Our knowledge of human sensorimotor learning and memory is predominantly based on the visuospatial workspace and limb movements. Humans also have a remarkable ability to produce and perceive speech sounds. We asked whether the human speech-auditory system could serve as a model to characterize the retention of sensorimotor memory in a workspace that is functionally independent of the visuospatial one. Using adaptation to altered auditory feedback, we investigated the durability of a newly acquired speech-acoustical memory (8- and 24-h delay), its sensitivity to the manner of acquisition (abrupt vs. gradual perturbation), and factors affecting memory retrieval. We observed extensive retention of learning (∼70%) but found no evidence for offline gains. The speech-acoustical memory was insensitive to the manner of its acquisition. To assess factors affecting memory retrieval, tests were first done in the absence of auditory feedback (with masking noise). Under these conditions, it appeared there was no memory for prior learning as if after an overnight delay, speakers had returned to their habitual speech production modes. However, when speech was reintroduced, resulting in speech error feedback, speakers returned immediately to their fully adapted state. This rapid switch shows that the two modes of speech production (adapted and habitual) can coexist in parallel in sensorimotor memory. The findings demonstrate extensive persistence of speech-acoustical memory and reveal context-specific memory retrieval processes in speech-motor learning. We conclude that the human speech-auditory system can be used to characterize sensorimotor memory in a workspace that is distinct from the visuospatial workspace.NEW & NOTEWORTHY There is extensive retention of speech-motor learning. Two parallel modes exist in speech motor memory, one with access to everyday habitual speech and the other with access to newly learned speech-acoustical maps. The availability of speech error feedback triggers a switch between these two modes. Properties of sensorimotor memory in the human speech-auditory system are behaviorally similar to, but functionally independent of, their visuospatial counterparts.

Keywords: auditory perception; contextual memory retrieval; implicit adaptation; speech memory retention; speech motor learning.

Figure 1:

Experimental design (A) Experimental setup with a participant seated in front of a computer monitor, which displayed the stimulus to be read aloud. The microphone and associated audio setup recorded the participant’s speech and played it back through headphones. (B) Schematic of formant perturbation of a speech signal (time domain on top, spectrogram on the bottom) (C) Study design. Vertical gray stripes indicate noise feedback trials, the numbers in parentheses indicate number of trials for that condition.

Figure 2:

Extensive retention and context-based retrieval following speech motor learning. (A) Percentage change in the first formant (F1) from baseline during learning and retention. The horizontal dashed line indicates zero percentage change in F1 from baseline. Each solid circle gives the average over three consecutive trials (1 bin), the shaded region indicates standard error. Vertical gray bars indicate bins with speech-modulated noise feedback. During the baseline block, these trials were interspersed with non-perturbed trials, but are combined and shown at the end of baseline block for visualization purposes. Insets show initial trialwise data for visit-1 and visit-2. (B) Retention as a percentage of learning (second bin after reintroducing speech feedback on visit-2 normalized by average of last ten bins with speech feedback on visit-1). (C) Comparison of learning (last bins) and retention (first bins) with speech feedback (darker shades) as well as those with speech-modulated noise feedback (lighter shades). Asterisks indicate p-values less than 0.05.

Figure 3:

Changes in F2 do not contribute to F1 retention. (A) Percentage change in the second formant (F2) from baseline during learning and retention. The horizontal dashed line indicates zero percentage change in F2 from baseline. Each solid circle shows the average value over three consecutive trials (1 bin) and the shaded region gives the standard error. Vertical gray bars indicate bins with speech-modulated noise feedback. During the baseline block, these trials were interspersed with non-perturbed trials, but are combined and shown at the end of baseline block for visualization purposes. (B) Scatter plots showing correlations between retention of F1 and learning of F1 (bottom left) or F2 (bottom right). Each solid circle indicates data from one participant.

Figure 4:

New learning stabilizes within a novel region of the vowel space. (A) Percentage change in first formant frequency during learning. Each trial bin consists of an average over three consecutive trials. Baseline data (16 bins) until the vertical dashed line, followed by a learning block (75 bins) in which the first formant frequency is shifted up. All utterances here are for the stimulus ‘dep’. The horizontal dashed line indicates 0% change relative to baseline. (B) Map of the vowel space for /Ɛ/ (dep), /æ/ (dap), /I/ (dip) during baseline and for /Ɛ/ (dep) at the end of learning. Each solid circle represents averaged data for 50 trials of that vowel for one participant; each ellipse marks the 95-percentile boundary for a given vowel. Arrows represent changes relative to /Ɛ/ with the convention that the arrow points to the vowel which is represented by the letter. The subscript indicates the experimental condition (e.g., A_BASELINE or A_LEARN). (C) Polar plot showing vectors from baseline of /Ɛ/ (dep) to each vowel and from /Ɛ/ following learning to /æ/ and /I/. Numbers around the circle indicate direction in degrees and the magnitude is indicated by smaller to larger circles. (D) Comparison of direction and magnitude of vectors from /Ɛ/ (dep baseline) to /Ɛ/ (dep at the end of learning). (E) Comparison of vectorial direction and magnitude from /Ɛ/ (dep baseline) to /æ/ (dap baseline) versus /Ɛ/ (dep following learning) to /æ/ (dap baseline). (F) Comparison of direction and magnitude of vectors from /Ɛ/ (dep baseline) to /I/ (dip baseline) versus /Ɛ/ (dep following learning) to /I/ (dip baseline). For plots (D), (E), and (F), each solid circle of similar color indicates the value for each participant and asterisks indicate p-values of less than 0.05.