10 20 30 40 50 60 70 80 90 100 PercentH it R ats C annotD iscrim inate Spectrally Sim ilar G lides rad lad n.s. 10 20 30 40 50 60 70 80 90 100 PercentH it Sim ilar Vow els are D ifficultto D iscrim inate dad dead dud deed dood 10 20 30 40 50 60 70 80 90 100 PercentH it G eneralization Across Talker and G ender dad -Fem ale1 Fem ale2 Fem a le3 Male 1 Male2 Male3 tad -Fem ale 1 Fem ale2 Fem ale3 Male1 Male2 Male3 10 20 30 40 50 60 70 80 90 100 PercentH it Early Training on N N N H inders Later D iscrim ination ofH H H HLN HHH LLL NNN 10 20 30 40 50 60 70 80 90 100 PercentH it Tone Sequence Sim ple Frequency D iscrim ination LLL (5 kH z) H H H (12 kH z) 10 20 30 40 50 60 70 80 90 100 PercentH it H LN is Indistinguishable from the R eversed Sequence HLN NLH 10 20 30 40 50 60 70 80 90 100 PercentH it C S -Frequency (octaves above C S+) Frequency D iscrim ination Threshold CS+ ( 5.6kH z) 0.13 0.20 0.26 0.32 0.37 0.43 0.53 10 20 30 40 50 60 70 80 90 100 PercentH it H LN is D iscrim inable from Triplets ofEach Segm ent HLN HHH LLL NNN 10 20 30 40 50 60 70 80 90 100 PercentH it Increm entalTraining Enhances R eversed Sequence D iscrim ination HLN NLH 10 20 30 40 50 60 70 80 90 100 PercentH it Enrichm entD oes N otIm prove R eversed Sequence D iscrim ination HLN NLH 0 1 2 3 Dprim e Sum m ary:R anked D -Prim e Perform ance on allSequence Tasks NL H E nrich -N LH NN N First S ing l et H H H Firs t Inc rem -NLH Frequency Behavioral data was collected from 46 rats over 4661 total daily training sessions. Methods All animals performed Go/ No-Go discrimination tasks in the same operant training booths. The stimuli used as CS+ and CS- varied among different training tasks, but the general timeline of training and procedures were the same for all tasks. Stages of Training Shaping Goal: Animals learn to press lever to receive food rewards Time Course: ~ 6 sessions (3 days). Continued until 3 sessions in which animals retrieved 50 pellets independently. Detection Training Goal: Animals learn to press to receive rewards only after they hear a sound. Animals must learn to avoid hitting during silent periods. Time course: ~30 sessions (3 weeks). Continued until 10 sessions in which animals respond significantly more to CS+ trials than to catch trials (ie, d’ >= 1.5 for 10 sessions) Discrimination Training Complex Sound Discrimination Abilities in Rats and the Effects of Multiple Training Manipulations A.C. Puckett, C.T. Novitski, N.D. Engineer, A.L. McMenamy, M.S. Perry, C.A. Perez, P. Kan, Y.H. Chen, V. Jakkamsetti, C.L. Heydrick, M.P. Kilgard, The University of Texas at Dallas, Richardson, TX Introduction The current experiments have examined the ability of rats to perform discriminations among complex sounds, including tone-noise sequences and speech stimuli. Understanding how normal animals are able to discriminate complex sounds is necessary in order to fully understand the auditory cortex and how injury, learning or plasticity can change perception and cortical functioning. SPEECH DISCRIMINATION SEQUENCE DISCRIMINATION Conclusions • Frequency discriminations are easy, but other sequence discriminations are difficult. • Onsets are the most salient elements of sequences. • Sequences beginning with the same element are difficult to discriminate. • Reversed sequence discrimination was impossible, despite different initial elements. •Sequences may be normally processed as a ‘unit’ rather than discrete elements. • Discrimination strategies may be changed by training. • If rats learn a poor strategy early in training, they will not learn to discriminate effectively. • Intermediate exemplars allowed animals to adopt effective strategies. HHH LLL NNN Silence TRAIN IN G DAYS HLN (C S+) HHH LLL NNN Silence TRAIN IN G DAYS HLN (C S+) NNN LLL HHH Silence TR AINING DAYS HLN (C S+) NNN LLL HHH Silence TR AINING DAYS HLN (C S+) Conclusions • Large spectral differences are easy for rats to discriminate, even if differences are not in the onset of the speech sound. • Onsets/consonants: /dad/ vs. /bad/ & /gad/ • Middle of stimulus/vowels: /dad/ vs. /deed/ & /dood/ • Subtle spectral differences are difficult for rats to discriminate (and are difficult for humans as well). • Onsets/consonants: /rad/ vs. /lad/ • Middle of stimulus/vowels: /dad/ vs. /dead/ & /dud/ • Rats can generalize among several variants of a speech stimulus. • Rats could generalize across several compressed exemplars, indicating that VOT wasn’t the only cue. • Rats could generalize across several different speakers. Rats were able to generalize across many different fundamental frequencies, temporal patterns of enunciation, and other idiosyncrasies. • Speech sounds seem to be more easily discriminated than tone- noise sequences. Speech Stimulus Creation •Recording: Monosyllabic words spoken by female native English speaker in a sound- proof chamber. –The words ‘dad’ and ‘tad’ were recorded from 5 other native English speakers (3 male, 2 female) to assess speaker generalization. •Frequency shifting: The frequency of the fundamental and all other formants were shifted into the rat’s hearing range by doubling their frequency. –Compressed versions of ‘dad’ and ‘tad’ were generated to assess temporal generalization. •Noise reduction: Background noise was subtracted from each signal. •Filtering: Each signal was filtered to correct for the frequency-response curve of the booth speaker. •Intensity adjustment: The RMS-values of the signals were adjusted so that the loudest 100 ms of each vowel was 60 dB SPL. Future Directions – www.utd.edu/~kilgard • Assessment of changes in perceptual abilities after NB-stimulation pairing Amanda Puckett • Measurement of changes in auditory cortex after long-term sequence training Dr. Navzer Engineer • Measurement of responses of auditory cortex after long-term speech training Crystal Novitski • Assessment of cortical processing of speech sounds after environmental enrichment Vikram Jakkamsetti • Assessment of loss of perceptual abilities after cortical injury Dr. Owen Floody • Information theory analysis of cortical responses to speech sounds Helen Chen • Sounds are delivered from a speaker placed outside the cage. Speaker is positioned so that sounds are usually delivered to the animal’s left ear. • L – Rat responds to sound stimuli by pressing on the lever. Only presses within 3 seconds of CS+ stimuli are rewarded. • P – Pellet dispenser (MED Associates) located outside the sound- proof chamber delivers a 45 mg food reward (Bio- Serv) to the rat after correct responses. • H – House light is extinguished after false alarms or late responses P H CS + sounds CS - sounds L P H CS + sounds CS - sounds L Apparat us 0 1 2 3 Dprim e Sum m ary:R anked D -Prim e Perform ance on allSpeech Tasks r ad vs.lad tad - 10% tad -m ale dead tad -50% du d tad t a d -fem ale do od gad de ed bad dad vs. HLN vs. 0 10 20 30 40 50 60 70 80 10 20 30 40 50 60 70 PercentH it Days ofTraining Increm entalTim e C ourse HLN silence NNN NLN NNH NLH 0 10 20 30 40 50 60 10 20 30 40 50 60 70 80 90 100 PercentH it D ays ofTraining Exam ple Tim e C ourse ofTraining S ilence LLL HHH In Loving Memory of Matt Perry (March 1981 - September 2005) In honor of the neuroscientist who delved into the mysteries and whims of life with wholehearted delight. Thank you to one who touched the minds he sought to understand. As a true inquirer into philosophy, pharmacology and literature, Mat enriched rats as well as people. It is an honor to have shared a common path with you. Acknowledgements n.s. n.s. n.s. n.s. 10 20 30 40 50 60 70 80 90 100 PercentH it Stop C onsonants are Easily D iscrim inable dad bad gad 10 20 30 40 50 60 70 80 90 100 PercentH it H LN is D iscrim inable From its C om ponentSegm ents HLN H L N 10 20 30 40 50 60 70 80 90 100 PercentH it % oforiginal stim ulus length G ood G eneralization to the M ajority ofC om pressed Variants dad -100 90 80 70 60 50 40 30 20 10 tad - 1 00 90 80 70 60 50 40 30 20 10 10 20 30 40 50 60 70 80 90 100 PercentH it Voice O nsetTim e D ifferences are D iscrim inable dad tad A)Frequency discrim ination C )Tripletdistractor- H igh first E)Tripletdistractor- N oise first F)R everse O rder H L N L L L H H H N L H H L N H L N discrim ination C )Tripletdistractor- H igh first D )Tripletdistractor- N oise first E)R everse O rder Frequency (kH z) H L N L L L H H H H H H H H H H H H H H H L L L L L L L L L L L L N N N N N N N N N N N N N L H H L N H L N F)Increm ental Training 225 m s H L N Tim e (w eeks) N L N N N H N L H B)Sequence elem ent discrim ination N L H H L N Target (C S+) Task D istracter (C S-) N N N N N N A)Frequency discrim ination C )Tripletdistractor- H igh first E)Tripletdistractor- N oise first F)R everse O rder H L N L L L H H H N L H H L N H L N discrim ination C )Tripletdistractor- H igh first D )Tripletdistractor- N oise first E)R everse O rder Frequency (kH z) H L N L L L H H H H H H H H H H H H H H H H H H H H H H H H H H H L L L L L L L L L L L L L L L L L L L L L L L L N N N N N N N N N N N N N N N N N N N N N N N N N L H H L N H L N F)Increm ental Training 225 m s H L N Tim e (w eeks) N L N N L N N N H N N H N L H N L H B)Sequence elem ent discrim ination N N L L H H H L N Target (C S+) Target (C S+) Task D istracter (C S-) D istracter (C S-) N N N N N N N N N N N N D ad R ad D ad D ead Dud D eed Dood Lad Tad D ad 10% D ad B ad G ad D ad 50% D ad Tad Tad 50% Tad 10% D ad M ale 1 D ad Fem ale 2 D ad Fem ale 1 Tad Fem ale 1 Tad Fem ale 2 Tad M ale 1 A )Place ofA rticulation B )V oice O nsetTim e C)VO T C om pression D)VO T M ultiple Talker E)Vow els F)G lides TARGET (C S+) DISTRACTER (C S-) D ad D ad R ad R ad D ad D ad D ead D ead Dud Dud D eed D eed Dood Dood Lad Lad Tad Tad D ad 10% D ad 10% D ad D ad B ad B ad G ad G ad D ad 50% D ad 50% D ad D ad Tad Tad Tad 50% Tad 50% Tad 10% Tad 10% D ad M ale 1 D ad M ale 1 D ad Fem ale 2 D ad Fem ale 2 D ad Fem ale 1 D ad Fem ale 1 Tad Fem ale 1 Tad Fem ale 1 Tad Fem ale 2 Tad Fem ale 2 Tad M ale 1 Tad M ale 1 A )Place ofA rticulation B )V oice O nsetTim e C)VO T C om pression D)VO T M ultiple Talker E)Vow els F)G lides TARGET (C S+) DISTRACTER (C S-) Rad Rad Lad Lad Dad Dad Tad Tad HLN HLN NLH NLH
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
10
20
30
40
50
60
70
80
90
100
Per
cen
t H
it
Rats Cannot Discriminate Spectrally Similar Glides
rad lad
n.s.
10
20
30
40
50
60
70
80
90
100
Per
cen
t H
it
Similar Vowels are Difficult to Discriminate
dad dead dud deed dood
10
20
30
40
50
60
70
80
90
100
Pe
rce
nt
Hit
Generalization Across Talker and Gender
dad -
Female
1
Femal
e2
Femal
e3
Mal
e1
Mal
e2
Mal
e3
tad -
Femal
e1
Femal
e2
Femal
e3
Mal
e1
Mal
e2
Mal
e3
10
20
30
40
50
60
70
80
90
100
Per
cen
t H
it
Early Training on NNN Hinders Later Discrimination of HHH
HLN HHH LLL NNN
10
20
30
40
50
60
70
80
90
100
Per
cen
t H
it
Tone Sequence
Simple Frequency Discrimination
LLL (5 kHz) HHH (12 kHz)
10
20
30
40
50
60
70
80
90
100
Per
cen
t H
it
HLN is Indistinguishable from the Reversed Sequence
HLN NLH
10
20
30
40
50
60
70
80
90
100
Pe
rce
nt
Hit
CS- Frequency (octaves above CS+)
Frequency Discrimination Threshold
CS+(5.
6kHz)
0.13
0.20
0.26
0.32
0.37
0.43
0.53
10
20
30
40
50
60
70
80
90
100
Per
cen
t H
it
HLN is Discriminable from Triplets of Each Segment
HLN HHH LLL NNN
10
20
30
40
50
60
70
80
90
100
Per
cen
t H
it
Incremental Training Enhances Reversed Sequence Discrimination
HLN NLH
10
20
30
40
50
60
70
80
90
100
Per
cen
t H
it
Enrichment Does Not Improve Reversed Sequence Discrimination
HLN NLH
0
1
2
3
Dp
rim
e
Summary: Ranked D-Prime Performance on all Sequence Tasks
NLH
Enric
h - NLH
NNN Firs
t
Single
t
HHH Firs
t
Incr
em -
NLH
Freque
ncy
Behavioral data was collected from 46 rats over 4661 total daily training sessions.
MethodsAll animals performed Go/ No-Go discrimination tasks in
the same operant training booths. The stimuli used as CS+ and CS- varied among different training tasks, but the general timeline of training and procedures were the same for all tasks.
Stages of TrainingShaping
Goal: Animals learn to press lever to receive food rewardsTime Course: ~ 6 sessions (3 days). Continued until 3 sessions in which animals retrieved 50 pellets independently.
Detection TrainingGoal: Animals learn to press to receive rewards only after they hear a sound. Animals must learn to avoid hitting during silent periods.Time course: ~30 sessions (3 weeks). Continued until 10 sessions in which animals respond significantly more to CS+ trials than to catch trials (ie, d’ >= 1.5 for 10 sessions)
Discrimination TrainingGoal: Animals learn to press to receive rewards only after they hear a CS+. Animals must avoid hitting after CS- stimuli or during silent periods.Time course: Variable (20 – 300 sessions).
Sequence Discrimination: Animals trained for up to 300 sessions or until behavioral criteria were met.
Speech Discrimination: Animals progressed through a variety of CS- conditions after 2 weeks of training on each task, regardless of behavioral performance.
Complex Sound Discrimination Abilities in Rats and the Effects of Multiple Training ManipulationsA.C. Puckett, C.T. Novitski, N.D. Engineer, A.L. McMenamy, M.S. Perry, C.A. Perez, P. Kan, Y.H. Chen, V. Jakkamsetti, C.L. Heydrick, M.P. Kilgard, The University of Texas at Dallas, Richardson, TX
Introduction
The current experiments have examined the ability of rats to perform discriminations among complex sounds, including tone-noise sequences and speech stimuli. Understanding how normal animals are able to discriminate complex sounds is necessary in order to fully understand the auditory cortex and how injury, learning or plasticity can change perception and cortical functioning.
SPEECH DISCRIMINATIONSEQUENCE DISCRIMINATION
Conclusions• Frequency discriminations are easy, but other sequence discriminations are difficult. • Onsets are the most salient elements of sequences.
• Sequences beginning with the same element are difficult to discriminate.
• Reversed sequence discrimination was impossible, despite different initial elements.
•Sequences may be normally processed as a ‘unit’ rather than discrete elements.
• Discrimination strategies may be changed by training. • If rats learn a poor strategy early in training, they will not learn to discriminate effectively.• Intermediate exemplars allowed animals to adopt effective strategies.
HHH
LLL
NNN
Silence
TRAINING DAYS
HLN (CS+)
High first
HHH
LLL
NNN
Silence
TRAINING DAYS
HLN (CS+)
High first
NNN
LLL
HHH
Silence
TRAINING DAYS
HLN (CS+)
Noise first
NNN
LLL
HHH
Silence
TRAINING DAYS
HLN (CS+)
Noise first
Conclusions• Large spectral differences are easy for rats to discriminate, even if differences are not in the onset of the speech sound.
• Onsets/consonants: /dad/ vs. /bad/ & /gad/ • Middle of stimulus/vowels: /dad/ vs. /deed/ & /dood/
• Subtle spectral differences are difficult for rats to discriminate (and are difficult for humans as well).
• Onsets/consonants: /rad/ vs. /lad/• Middle of stimulus/vowels: /dad/ vs. /dead/ & /dud/
• Rats can generalize among several variants of a speech stimulus. • Rats could generalize across several compressed exemplars, indicating that VOT wasn’t the only cue. • Rats could generalize across several different speakers. Rats were able to generalize across many different fundamental frequencies, temporal patterns of enunciation, and other idiosyncrasies.
• Speech sounds seem to be more easily discriminated than tone-noise sequences.
Speech Stimulus Creation
•Recording: Monosyllabic words spoken by female native English speaker in a sound-proof chamber.
–The words ‘dad’ and ‘tad’ were recorded from 5 other native English speakers (3 male, 2 female) to assess speaker generalization.
•Frequency shifting: The frequency of the fundamental and all other formants were shifted into the rat’s hearing range by doubling their frequency.
–Compressed versions of ‘dad’ and ‘tad’ were generated to assess temporal generalization.
•Noise reduction: Background noise was subtracted from each signal.•Filtering: Each signal was filtered to correct for the frequency-response curve of the booth speaker. •Intensity adjustment: The RMS-values of the signals were adjusted so that the loudest 100 ms of each vowel was 60 dB SPL.
Future Directions – www.utd.edu/~kilgard• Assessment of changes in perceptual abilities after NB-stimulation pairing Amanda Puckett• Measurement of changes in auditory cortex after long-term sequence training Dr. Navzer Engineer• Measurement of responses of auditory cortex after long-term speech training Crystal Novitski• Assessment of cortical processing of speech sounds after environmental enrichment Vikram Jakkamsetti• Assessment of loss of perceptual abilities after cortical injury Dr. Owen Floody• Information theory analysis of cortical responses to speech sounds Helen Chen
• Sounds are delivered from a speaker placed outside the cage. Speaker is positioned so that sounds are usually delivered to the animal’s left ear.• L – Rat responds to sound stimuli by pressing on the lever. Only presses within 3 seconds of CS+ stimuli are rewarded. • P – Pellet dispenser (MED Associates) located outside the sound-proof chamber delivers a 45 mg food reward (Bio-Serv) to the rat after correct responses.• H – House light is extinguished after false alarms or late responses
P
H
CS + soundsCS - sounds
L
P
H
CS + soundsCS - sounds
L
Apparatus
0
1
2
3
Dp
rim
e
Summary: Ranked D-Prime Performance on all Speech Tasks
rad v
s. la
d
tad -
10%
tad -
mal
edea
d
tad -
50%
dudta
d
tad -
fem
ale
dood gaddee
dbad
dad vs. HLN vs.
0 10 20 30 40 50 60 70 80
10
20
30
40
50
60
70
80
90
100
Pe
rce
nt
Hit
Days of Training
Incremental Time CourseHLNsilence
NNN
NLN
NNHNLH
0 10 20 30 40 50 60
10
20
30
40
50
60
70
80
90
100
Pe
rce
nt
Hit
Days of Training
Example Time Course of Training
Silence
LLL
HHH
In Loving Memory of Matt Perry (March 1981 - September 2005) In honor of the neuroscientist who delved into the mysteries and whims of life with wholehearted delight. Thank you to one who touched the minds he sought to understand. As a true inquirer into philosophy, pharmacology and literature, Mat enriched rats as well as people. It is an honor to have shared a common path with you. AcknowledgementsWe wish to thank all the members of the Kilgard Lab behavioral team. Research supported by NIH R21#1R15DC00662401
n.s.
n.s. n.s.
n.s.
10
20
30
40
50
60
70
80
90
100
Per
cen
t H
it
Stop Consonants are Easily Discriminable
dad bad gad
10
20
30
40
50
60
70
80
90
100
Per
cen
t H
it
HLN is Discriminable From its Component Segments
HLN H L N
10
20
30
40
50
60
70
80
90
100
Pe
rce
nt
Hit
% of original stimulus length
Good Generalization to the Majority of Compressed Variants
dad -
100 90 80 70 60 50 40 30 20 10
tad -
100 90 80 70 60 50 40 30 20 10
10
20
30
40
50
60
70
80
90
100
Per
cen
t H
it
Voice Onset Time Differences are Discriminable
dad tad
A) Frequency discrimination
C) Triplet distractor-High first
E) Triplet distractor-Noise first
F) Reverse OrderH L N
L L L H H H
N L H
H L N
H L N
discrimination
C) Triplet distractor-High first
D) Triplet distractor-Noise first
E) Reverse Order
Fre
qu
enc
y (k
Hz)
H L N
L L L H H H
H H HH H H
H H HH H H
L L LL L L
L L LL L L
N N NN N N
N N NN N N
N L H
H L N
H L N
F) Incremental Training
225 ms
H L N
Time (weeks)
N L N N N H N L H
B) Sequence elementdiscrimination NLHH L N
Target(CS+)
Task Distracter(CS-)
N N NN N N
A) Frequency discrimination
C) Triplet distractor-High first
E) Triplet distractor-Noise first
F) Reverse OrderH L N
L L L H H H
N L H
H L N
H L N
discrimination
C) Triplet distractor-High first
D) Triplet distractor-Noise first
E) Reverse Order
Fre
qu
enc
y (k
Hz)
H L N
L L L H H H
H H HH H HH H HH H H
H H HH H HH H HH H H
L L LL L LL L LL L L
L L LL L LL L LL L L
N N NN N NN N NN N N
N N NN N NN N NN N N
N L H
H L N
H L N
F) Incremental Training
225 ms
H L N
Time (weeks)
N L NN L N N N HN N H N L HN L H
B) Sequence elementdiscrimination NNLLHHH L N
Target(CS+)Target(CS+)
Task Distracter(CS-)
Distracter(CS-)
N N NN N NN N NN N N
Dad
Rad
Dad
Dead Dud Deed Dood
Lad
Tad
Dad 10%
Dad Bad Gad
Dad 50% Dad Tad Tad 50% Tad 10%
Dad Male 1 Dad Female 2 Dad Female 1 Tad Female 1 Tad Female 2 Tad Male 1
A) Place of Articulation
B) Voice Onset Time
C) VOT Compression
D) VOT Multiple Talker
E) Vowels
F) Glides
TARGET (CS+)
DISTRACTER (CS-)
DadDad
RadRad
DadDad
DeadDead DudDud DeedDeed DoodDood
LadLad
TadTad
Dad 10%Dad 10%
DadDad BadBad GadGad
Dad 50%Dad 50% DadDad TadTad Tad 50%Tad 50% Tad 10%Tad 10%
Dad Male 1Dad Male 1 Dad Female 2Dad Female 2 Dad Female 1Dad Female 1 Tad Female 1Tad Female 1 Tad Female 2Tad Female 2 Tad Male 1Tad Male 1
A) Place of Articulation
B) Voice Onset Time
C) VOT Compression
D) VOT Multiple Talker
E) Vowels
F) Glides
TARGET (CS+)
DISTRACTER (CS-)
Rad LadRad Lad
Dad TadDad Tad
HLN NLHHLN NLH
Related Documents
