Auditory cortical monitoring prevents speech errors before they happen Caroline A. Niziolek UCSF Depts. of Radiology and Otolaryngology – Head and Neck Surgery Biomagnetic Imaging Laboratory Speech Neuroscience Laboratory
The Speech Chain Denes & Pinson, 1993
The Speech Chain Denes & Pinson, 1993
The Speech Chain Denes & Pinson, 1993
The Speech Chain Denes & Pinson, 1993
The Speech Chain Denes & Pinson, 1993 deafness
The Speech Chain Denes & Pinson, 1993 deafness decrease in intelligib ility
The Speech Chain Denes & Pinson, 1993 deafness decrease in intelligib ility frontal motor lesion
The Speech Chain Denes & Pinson, 1993 deafness decrease in intelligib ility frontal motor lesion perceptual deficits
Understanding speech production via perception
MEG fMRI ECoG
Understanding speech production via perception Houde, Niziolek, et al., 2014, ISSP MEG fMRI ECoG
Understanding speech production via perception
How does auditory feedback affect speech output? Understanding speech production via perception
How does auditory feedback affect speech output? How do we detect deviations from what we intend to say? Understanding speech production via perception
How does auditory feedback affect speech output? How do we detect deviations from what we intend to say? What is the corrective behavior when a deviation is detected? Understanding speech production via perception
How does auditory feedback affect speech output? How do we detect deviations from what we intend to say? What is the corrective behavior when a deviation is detected? Understanding speech production via perception
Auditory feedback causes compensatory changes in our speech output
Focal acoustic changes evoke focal compensation frequenc y amplitud e
Focal acoustic changes evoke focal compensation frequenc y amplitud e f0 (pitch)
Focal acoustic changes evoke focal compensation frequenc y amplitud e formants (vowel)
Focal acoustic changes evoke focal compensation frequenc y amplitud e amplitude (loudness )
Chang, Niziolek, et al., 2013, PNAS f0 shift during vocalization
f0 shift of emphatic stress f0 contrast distance Patel et al., 2011, JSLHR
f0 shift of emphatic stress f0 contrast distance Patel et al., 2011, JSLHR ampl contrast distance
Real-time formant alteration Niziolek & Guenther, 2013, J. Neurosci. Niziolek & Guenther, 2014, Frontiers for Kids
Real-time formant alteration Niziolek & Guenther, 2013, J. Neurosci. Niziolek & Guenther, 2014, Frontiers for Kids x y frequency amplitude
Real-time formant alteration Niziolek & Guenther, 2013, J. Neurosci. Niziolek & Guenther, 2014, Frontiers for Kids x y frequency amplitude
Real-time formant alteration Niziolek & Guenther, 2013, J. Neurosci. Niziolek & Guenther, 2014, Frontiers for Kids x y frequency amplitude
Real-time formant alteration
Greater compensation to vowels shifted across a boundary Niziolek & Guenther, 2013, J. Neurosci.
Enhanced auditory error to shifts across a boundary Niziolek & Guenther, 2013, J. Neurosci.
Chang, Niziolek, et al., 2013, PNAS Niziolek & Guenther, 2013, J. Neurosci. f0formants Patel et al., 2011, JSLHR f0 contract distance
How does auditory feedback affect speech output? How do we detect deviations from what we intend to say? What is the corrective behavior when a deviation is detected? Understanding speech production via perception
How does auditory feedback affect speech output? How do we detect deviations from what we intend to say? What is the corrective behavior when a deviation is detected? Understanding speech production via perception
Neural responses to feedback Chang, Niziolek, et al., 2013, PNAS
Neural responses to feedback Chang, Niziolek, et al., 2013, PNAS NORMAL
Neural responses to feedback Chang, Niziolek, et al., 2013, PNAS NORMAL ALTERED
Neural responses to feedback Chang, Niziolek, et al., 2013, PNAS NORMAL ALTERED < feedback match: suppression
Neural responses to feedback Chang, Niziolek, et al., 2013, PNAS NORMAL ALTERED < feedback match: suppression feedback mismatch: no suppression ≥
Auditory cortical neurons are suppressed during speech, but only when the feedback matches what is expected.
What causes this selective suppression? Can it be used for error detection?
Models of speech motor control
CONTROLLER : premotor, motor cortex Models of speech motor control
CONTROLLER : premotor, motor cortex OUTPUT : vocal tract, articulato rs Models of speech motor control motor command
CONTROLLER : premotor, motor cortex OUTPUT : vocal tract, articulato rs COMPARISON : auditory cortex internal prediction Models of speech motor control motor command
CONTROLLER : premotor, motor cortex OUTPUT : vocal tract, articulato rs COMPARISON : auditory cortex auditory signal internal prediction Models of speech motor control motor command
CONTROLLER : premotor, motor cortex OUTPUT : vocal tract, articulato rs COMPARISON : auditory cortex auditory signal internal prediction Models of speech motor control motor command
CONTROLLER : premotor, motor cortex OUTPUT : vocal tract, articulato rs COMPARISON : auditory cortex auditory signal internal prediction corrective error signal Models of speech motor control motor command
CONTROLLER : premotor, motor cortex OUTPUT : vocal tract, articulato rs COMPARISON : auditory cortex auditory signal internal prediction corrective error signal Models of speech motor control motor command
Artificially-altered feedback
Chang, Niziolek, et al., 2013, PNAS release from suppression
Artificially-altered feedback Chang, Niziolek, et al., 2013, PNAS change in speech output release from suppression
Are these error- correction processes at work in natural speech?
Multi-vowel task eat spea k Ed add list en eat Ed add Niziolek et al., 2013, J. Neurosci
Auditory M100
Owen et al., 2012, NeuroImageNiziolek et al., 2013, J. Neurosci Source localization: Champagne
Vowel space: center vs. periphery Niziolek et al., 2013, J. Neurosci
Vowel space: center vs. periphery Niziolek et al., 2013, J. Neurosci “eat ” “Ed” “add”
Are peripheral productions processed as errors?
centerperiphery F1 F2
centerperiphery F1 F2 predict variability off center = expected
centerperiphery F1 F2 predict variability off center = expected
centerperiphery F1 F2 predict variability off center = expected
centerperiphery F1 F2 predict variability off center = expected don’t predict variability off center = error
center periphery F1 F2 predict variability off center = expected don’t predict variability off center = error
center periphery F1 F2 predict variability off center = expected don’t predict variability off center = error
center periphery F1 F2 predict variability off center = expected don’t predict variability off center = error
Individual subject decrease in suppression Niziolek et al., 2013, J. Neurosci
Decrease in suppression is consistent in left AC p = Niziolek et al., 2013, J. Neurosci
Decrease in suppression is consistent in left AC Niziolek et al., 2013, J. Neurosci
center periphery F1 F2 predict variability off center = expected don’t predict variability off center = error
center periphery F1 F2 predict variability off center = expected don’t predict variability off center = error WINNER
acoustic distance decrease in suppressio n ∝ Niziolek et al., 2013, J. Neurosci
acoustic distance decrease in suppressio n ∝ Niziolek et al., 2013, J. Neurosci
acoustic distance decrease in suppressio n ∝ Niziolek et al., 2013, J. Neurosci
acoustic distance decrease in suppressio n ∝ Niziolek et al., 2013, J. Neurosci
acoustic distance decrease in suppressio n ∝ Niziolek et al., 2013, J. Neurosci
acoustic distance decrease in suppressio n ∝ Niziolek et al., 2013, J. Neurosci
acoustic distance decrease in suppressio n ∝ Niziolek et al., 2013, J. Neurosci
Suppression falls off at the vowel periphery Niziolek et al., 2013, J. Neurosci Fall-off increases with acoustic distance
Acoustic error is coded by auditory suppression
The efferent prediction may reflect an acoustic target or goal (not merely a “copy” of motor commands)
How does auditory feedback affect speech output? How do we detect deviations from what we intend to say? What is the corrective behavior when a deviation is detected? Understanding speech production via perception
How does auditory feedback affect speech output? How do we detect deviations from what we intend to say? What is the corrective behavior when a deviation is detected? Understanding speech production via perception
Does error-like processing have behavioral consequences?
Behavioral consequences
Niziolek et al., 2013, J. Neurosci Behavioral consequences
If centering is partly driven by auditory feedback, it should: correlate with cortical responses to feedback decrease when you can’t hear yourself (i.e., in masking noise)
Centering correlates with suppression Niziolek et al., 2013, J. Neurosci
Masking noise reduces centering Niziolek et al., submitted quiet
Masking noise reduces centering quiet masking noise Niziolek et al., submitted
Masking noise reduces centering quiet masking noise Implications for hearing loss Niziolek et al., submitted
We detect deviations from an expected target Decreased suppression ongoing error detection process (we make “errors” all the time!)
We correct our deviations before they become errors Decreased suppression ongoing error detection process (we make “errors” all the time!) Centering ongoing error correction process (partly mediated by feedback)
Ongoing & future directions: Assess error detection and correction capacities in patients with speech disorders
Speech error detection and correction in aphasia
detection impaired
Speech error detection and correction in aphasia centerperiphery F1 F2 detection impaired
Speech error detection and correction in aphasia centerperiphery F1 F2 detection impaired detection preserved
Speech error detection and correction in aphasia center periphery F1 F2 detection impaired detection preserved
Speech error detection and correction in aphasia center periphery F1 F2 detection impaired detection preserved sensory feedback training
Speech error detection and correction in aphasia center periphery F1 F2 detection impaired detection preserved sensory feedback training motor skill training
Ongoing & future directions: How task-specific is the perceived error?
Niziolek et al., in prep Redefining center and periphery
Sort by formant Niziolek et al., in prep Redefining center and periphery F1 F2
Sort by formant Niziolek et al., in prep Redefining center and periphery F1 F2 ✓
F0 Sort by formant Sort by pitch Redefining center and periphery F1 F2 ✓
F0 Sort by formant Sort by pitch Redefining center and periphery F1 F2 ✓ ✗
Multi-pitch task Niziolek et al., in prep eat spea k eat list en eat z xc z xc
Owen et al., 2012, NeuroImageNiziolek et al., in prep Source localization: right hemi
F0 Sort by pitch Pitch task auditory suppression
F0 Sort by pitch Pitch task auditory suppression ✓ Niziolek et al., in prep p = 0.048
F0 Sort by formant Sort by pitch Pitch task auditory suppression F1 F2 ✓ Niziolek et al., in prep p = 0.048
F0 Sort by formant Sort by pitch Pitch task auditory suppression F1 F2 ✗ ✓ Niziolek et al., in prep p = 0.27 (n.s.) p = 0.048
Task-specific suppression F0 forman t pitch F1 F2 vowel task pitch task
Task-specific suppression F0 forman t pitch F1 F2 vowel task pitch task ✓ ✓
Task-specific suppression F0 forman t pitch F1 F2 vowel task pitch task ✓ ✓ (n.s.) ✗
Pitch centering Niziolek et al., in prep
Pitch centering Niziolek et al., in prep
Understanding speech production via perception
Altered auditory feedback causes compensatory changes to speech output Understanding speech production via perception
Altered auditory feedback causes compensatory changes to speech output Degree of auditory cortical suppression allows us to detect deviations from what we intend to say Understanding speech production via perception
Altered auditory feedback causes compensatory changes to speech output Degree of auditory cortical suppression allows us to detect deviations from what we intend to say This suppression may underlie a corrective behavior that serves to bring speech back on track Understanding speech production via perception
Thank you! NIH F32DC NIH R01DC NSF BCS Frank Guenther John Houde Sri Nagarajan Eddie Chang Danielle Mizuiri Susanne Honma
Decrease in suppression in both speak and listen Niziolek et al., 2013, J. Neurosci
No systematic changes in suppression in right AC Niziolek et al., 2013, J. Neurosci
No systematic changes in suppression in right AC Niziolek et al., 2013, J. Neurosci
Formant variability
Anti-centering masking noise Niziolek et al., in prep
Direct cortical recordings Chang, Niziolek, et al., 2013, PNAS
Altered feedback Normal feedback Chang, Niziolek, et al., 2013, PNAS
“SIS falloff” across acoustic space Niziolek et al., 2013, J. Neurosci