Lec 10/ TB Ch 11

Question 1

• Lowest freq of harmonic spectrum: 2 names
• 1st vs 2nd harmonic freq
• Missing-fundamental effect
  
  • definition
  • LS graphs
  • RS graphs
    
    • connection to fundamental freq
• Rogues waves

Answer 1

Complex sounds

• Harmonics
  
  • Lowest frequency of harmonic spectrum:
    
    • Fundamental frequency
      
      • 2^nd harmonic is 2x as high as fundamental freq

Missing fundamental part 1

• Missing fundamental (x1) is not/barely noticed.
  
  • # 1 is delected/not played
  • • How do we perceive it?
  • Here a tone, we hum it back out
  • If the tone lacks the fundamental freq, we still perceive it
  • The auditory system completes the missing fundamental freq
  • It happens w/ a full or few harmonics
• More harmonics can be missing. How come? - demo
  
  • LS: pure tunes w/ increasing frequency
  • RS #1: when all freq are played together, we only perceive the fundamental frequency
  • RS#2: when all freq are played and the fundamental freq is removed, we still only perceive the fundamental freq
  • # 3&4: still hear faint fundamental freq
• RS#1: Looking at the sound pressure wave, they repeat the cycle of the fundamental freq
• This explains why the fundamental freq is perceived even though it is not there
• • X
• This is similar to rogue waves
  
  • In the ocean, waves can be superimposed
    *

Question 2

• graphs 1,2,3
  
  • how are they related
• Special feature of graph 4
• How can you hear 250 Hz?
• What explains this?
  
  • place code vs temporal code?

Answer 2

Missing fundamentals part 3

• Graphs #1,2,3: 500, 750, 1000 Hz (differ by increments of 250 Hz)
  
  • missing fundamental harmonic: 250 Hz
• How can you hear 250 Hz in Graph 4?
  
  • Graph 4: all 3 overlap, 2nd, 3rd & 4th harmonic overlap in peaks every 4 ms
    
    • these three waveforms come into alignment every 4 ms, which, conveniently, happens to be the period of the fundamental frequency for these three harmonics: 250 Hz.
    • Added up they yield a fluctuation in energy at 250 Hz
    • Indeed, every harmonic of 250 Hz will have an energy peak every 4 ms.
    • This explain why listeners perceive the pitch of this complex tone to be 250 Hz, even though the tone has no 250-Hz component.)
  • What explains this: place code vs temporal code?
    
    • For the basilar membrane, only regions sensitive to 500, 750, and 1000 are active
    • → The place code does not explain the missing fundamental effects
    • It must be the temporal code that explain the superimpositions, and population responses of neurons
      
      • They are responding particularly stronger at these particular rates (rate of fundamental freq)
      • Population code show increase activity across multiple neurons according to the volley principle
      • Reflecting fundamental freq 250 Hz

Question 3

• Timbre
• What does perception of timbre depend on?
• “Timbre contrast” or “timbre aftereffect experiment
  
  • vowel “ee”: # of formants?
  • Methods
    
    • Graph A
    • Graph b
    • Graph c
  • Conclusion
• Attack vs decay
  
  • definition
  • amplitude change

Answer 3

Complex sounds cont

• Timbre: Psychological sensation by which a listener can judge that two sounds that have the same loudness and pitch are dissimilar; conveyed by harmonics and other high frequencies
  
  • Possible to additional noise??
  • Perception of timbre depends on context in which sound is heard
  • Experiment by Summerfield et al. (1984) - “Timbre contrast” or “timbre aftereffect
• Study – freq spectrum of vowel ‘ee’
  
  • RS: “ee” has a peak at lower harmonic, a middle peak, and third peak (3 peaks in harmonic spectrum = fingerprints of vowels = formants)
  • # 1: Summerfield used synthesizer to create artificial sounds
    • Played the one on LS (a)
      
      • It looks like a broken comb
      • Some harmonics are deleted (3 patches)
      • Rs played this tone first where the harmonics are deleted (similar to color adaptation: stare at red spot)
    • Then played (b)
      
      • (similar to color adaptation: stare at white wall -> see afterimage of green dot)
      • Then you here an aftersound
      • Their perception were influenced by what they heard before
  • IOW: rs 1^st played the opposite of the vowel “ee”
  • Ppl adapt to
    
    • The first fundamental freq -> doesn’t appear as loud when
    • Same for the other red patches -> reduced ability to hear the red freq
    • But for the other intermediate freq -> you hear them well
    • As a result, what they hear sounds like the vowel “ee”
  • As such, the context (adapting sound @a) matters
• Attach and decay of sound
  
  • How sound starts/ends
  • Attack: Part of a sound during which amplitude increases (onset)
  • Decay: Part of a sound during which amplitude decrease
  • Ex. pluck string: attack is immediate
  • Ex. violin & bow: not as immediate

Question 4

• Auditory Scene Analysis
• Auditory scene
• Multiple sound sources in env (complex sounds) - How does auditory system sort out these sources?
  
  • 2 names
  • Example: frog + bird + splash
• strategies to segregate sound sources
  
  • 1: Spatial separation
    
    • example: frog & bird
  • 2: motion parallax
  • 3: spectral/temporal qualities/ auditory stream segregation
    
    • define
    • describe graphs
    • This aligns w/ which gestalt law?
    • Bach fugue example
  • 4: group by timbre
    
    • Bach organs & timbre
    • This aligns w/ which gestalt law?
  • 5: grouping by onset
    
    • definition
    • Rasch (1987): what helps to distinguish 2 notes from eo?
    • This aligns w/ which gestalt law?
      
      • Spectogram vs Spectrum dimensions
    • Did the bottle break
      
      • Top vs bottom
      • Timbre
• Multisensory integration
  
  • definition
  • Which 2 sensory modalities use this?
  • Ventriloquist effect

Answer 4

Hearing in the env

• Auditory scene: the entirety of sounds that is audible in a given moment and that conveys information about the events happening in that moment
• Multiple sound sources in env (complex sounds) - How does auditory system sort out these sources?
  
  • Source segregation, or auditory scene analysis (i.e. separate the sounds of frog + bird + splash)
  • Auditory Scene Analysis (how does the auditory system know when to put shit together/segregate them)
  • • # of strategies to segregate sound sources
  • # 1: Spatial separation between sounds (frog on left croaks; bird on right chirps)
    • # 2: motion parallax (when you move around, sound sources move pass your head -> spatial separation)
  • # 3: Separation on basis of sounds’ spectral or temporal qualities
  • Auditory stream segregation:
    
    • Perceptual organization of a complex acoustic signal into separate auditory events for which each stream is heard as a separate event
      
      • Y-axis = freq
      • X-axis = time
      • LS: Dalaldala
        
        • The freq/pitch of 2 sounds are close together
        • So our auditory system groups these 2 sounds together based on the Gestalt principle of similarity (Lec 4)
      • RS: The freq/pitch of 2 sounds are far -> don’t group them together
  • Gestalt law: similarity
    
    • Bach: One instrument playing, but it sounds like 2 is playing
• x
• # 4: Grouping by timbre
  • Organ can play same pitch w/ different timbres (2 type of pipes made of different material)
  • Can tell what belongs to what belongs to what
  • This resembles gestalt law of good continuation: based on the color
    
    • x
• # 5: Grouping by onset
• Harmonics of speech sound or music
  
  • If the sounds are played together -> we tend to group them together
  • If the sounds have diff onsets, we don’t group them together
• Grouping different harmonics into a single complex tone
• Rasch (1987): It is much easier to distinguish two notes from one another when onset of one precedes onset of other by very short time
• -> Resembles Gestalt law of common fate
  
  • Does the bottle break?
  • Spectogram: A pattern for sound analysis that provides a 3D display of intensity as a function of time and frequency
    
    • Spectrum: 2D (x-axis = freq; y-axis = power)
    • Spectrogrom: 3D, 3^rd dimension = time
      
      • Used color
      • 1^st dimension or vertical-axis = freq
      • 2^nd dimension or color = power or intensity
      • 3^rd dimension or horizontal-axis = time
    • • Based on the spectrogram, we can tell if the bottle is bouncing or breaking
    • Top graph = bounce
      
      • Fingers or spikes: certain frequencies have more power than other, and these frequencies re-occur (across) – have the same timbre
      • Since the timbre is constant, we can tell the bottle didn’t break
    • Bottom graph = broke
      
      • After drop, there are different pieces -> messy
      • The different pieces bounce off the floor at different times
      • Each vertical spectrum is different, the timbre are different

Auditory scene analysis

• Multisensory integration: vision (usually) helps audition to tell what belongs together
• Ventriloquist effect: An audio-visual illusion in which sound is misperceived as emanating from a source that can be seen to be moving appropriately when it actually emanates from a different invisible source.
  
  • When there’s 2 ppl, and only 1 person is moving their lips, we tend to perceive sounding coming from this person
  • Visual dominance for location.

Question 5

• restoration effect in visual modality - describe top 2 figures
• Which gestalt law if used here?
• Kluender and Jenison - restoration effect for sound
  
  • methods: 2 steps
    
    • 2 conditions
    • results
• Restoration effect for complex sounds
  
  • What sources of info is used?
  • Study: The mailman brought the letter.
    
    • 3 conditions (see spectrogram)
    • explanation

Answer 5

Continuity and restoration effects

• Restoration based on the gestalt law of good continuation: In spite of interruptions, one can still “hear” sound
  
  • LS: looks like a bunch of chicken feed
  • RS: when you occlude them partially, our visual system perceive the “chicken legs” as a continuous -> we see a cube
• • X
• Restoration based on the gestalt law of good continuation: In spite of interruptions, one can still “hear” sound
  
  • Experiments that use signal detection task (e.g., Kluender and Jenison) suggest that at some point, restored missing sounds are encoded in the brain as if they were actually present!
  • Methods:
    
    • # 1: Rs played a pure tone over time (red line)
    • # 2: At some time, they played some white noise (ex. radio w/ poor reception, all sorts of energy/freq) = grey box
  • • There are 2 conditions
      * Condition 1: you hear the pure tone, and the noise simultaneously
      * Condition 2: you hear the pure tone, then the noise, then the pure tone again
      
      • Results: used d’ experiment to detect if it is yes or not
      • Many people’s d’ = 0
      • They cannot tell if the sound is playing at the same time as the noise in the background
      • It seems the brain reconstructed a perception of the sound based on the gestalt law of good continuation
  • This also applies if Rs plays a pure tone that gradually increases in pitch
  • -> restoration effects for pure tones
• X
• Restoration of complex sound, (e.g., music, speech)
  
  • Here, we don’t just use gestalt law of good continuation/ auditory info, we use “Higher-order” sources of information
  • Ex. we may use semantic info
    
    • Ex. the phoneme “wh” is missing, replaced as white noise
      
      • Ex 1: “The *eel fell off the car.” (wheel)
        
        • we perceive this as “wheel”
      • Ex 2: “The *eel fell off … the table.” (meal)
        
        • we perceive this as “meal”

The restoration effect

• our perceptual system is taking into account there is smth occluding the sound you actually want to hear
• Phonemic restoration: Noise helps comprehension
  
  • Spectrogram for the sentence “The mailman brought the letter.”
    
    • For a = normal sentence
    • B: remove certain parts of the sentence (no noise) -> cannot understand sentence
      
      • There’s nothing, so the perceptual system doesn’t see a reason why we aren’t hearing this -> does not complete the sentence
    • C: replace certain parts of the sentences w/ white noise -> can understand it
      
      • Since there is white noise, perceptual system assumes smth is missing -> completes the sentence
  • • Similar to the chicken feet phenom earlier
      
      • LS: no bars, no reason to believe it is a cube
      • RS: there are bars occluding things -> likely to see a cube
        *

Question 6

• Pythagoras: music & math, planets & math
• Musical notes freq range
• Pitch
• Freq most audible for us
• Max freq musical freq can play
• Freq most played by instruments
• Why aren’t there many instruments playing 5000 Hz?
  
  • 2 reasons
• x
• Octave
• 2 dimensions of musical pitch
  
  • tone height
  • tone chroma
  • musical helix - describe
• Chords
• dissonant
• consonant
• distance b/w
  
  • 1st → 2nd harmonic
  • 2nd → 3rd harmonic
  • 3rd → 4th harmonic
  • 4th → 5th harmonic
• Ratios count
• x
• What relationship between notes is universal?
• Western versus Javanese on # of notes in an octave?
• Musicians’ estimates of intervals between notes vs infant?
• x
• Melody
  
  • definition
  • aka
  • Defined by relationship between ??? not ???
• Tempo
• Fugue
• Rhythm: is it specific to music?
• Botman 1894 study
  
  • mehod
  • result

Answer 6

Music

• Music is a way to express thoughts and emotions
• Pythagoras: Numbers and music intervals
  
  • Found that music has mathematical regularities; and thinks this is related to the distance b/w planets
• Some clinical psychologists practice music therapy
• X
• Musical notes
  
  • Sounds of music extend across a frequency range from about 25 to 4500 Hz (aka pitch)
  • Pitch: The psychological aspect of sounds related mainly to the fundamental frequency
  • Bottom of red curve = bottom of blue graph
    
    • Our best audibility based on these both graphs are 4500 Hz
• musical instruments can play up to 4500 Hz, not beyond
• Also note: we can perceive 5000 Hz better than 100 Hz (based on the red graph)
  
  • There are many instruments that can play 100 Hz (ex. guitar, harp, piano), but not 5000 Hz -> why?
  • Our temporal codes do not work at 5000 Hz, and temporal code is essential for enjoying music
    
    • Listeners: Great difficulty perceiving octave relationships between tones when one or both tones are greater than 5 kHz -> temporal frequency coding
    • Sounds below freq -> we can perceive patterns
• x
• Octave
• Octave: The interval between two sound frequencies having a ratio of 2:1
  
  • Example: Middle C (C4) has a fundamental frequency of 261.6 Hz; notes that are one octave from middle C are 130.8 Hz (C3) and 523.2 Hz (C5)
  • C3 (130.8 Hz) sounds more similar to C4 (261.6 Hz) than to E3 (164.8 Hz)
  • There is more to musical pitch than just frequency!

Musical pitch has 2 dimensions

• Tone height: A sound quality corresponding to the level of pitch. Tone height is monotonically related to frequency
• Tone chroma: A sound quality shared by tones that have the same octave interval
  
  • Each note on the musical scale (A–G) has a different chroma
• Musical helix: Can help visualize musical pitch
  
  • Notes go around the helix
  • The C2 and C3 are at the same angle (also represented green)

Music - patterns

• Chords: Created when two or more notes are played simultaneously
• Consonant: Have simple ratios of note frequencies (3:2, 4:3)
• Dissonant: Less elegant ratios of note frequencies (16:15, 45:32)
  
  • If you listen to more jazz, some dissonant sounds seem to be consonant
• 1^st harmonic -> 2^nd harmonic = octave
• 2^nd harmonic -> 3^rd harmonic = 5^th
• 3^rd -> 4^th harmonic = 4^th
• 4^th -> 5^th harmonic = 3rd
• Ratios count: chords across several octaves perceived as the “same”.
• X

Cultural differences

• Some relationships between notes, such as octaves, are universal
• Research on music perception: Western versus Javanese
  
  • Javanese culture: Fewer notes within an octave; greater variation in note’s acceptable frequencies
  • Pelog scale in Javanese music:
  • Musicians’ estimates of intervals between notes correspond to the music scale from their culture
  • Infants detect inappropriate notes within both scales (there’s smth innate in both of these scales)

Melody

• Melody: An arrangement of notes or chords in succession (chroma (not tone height) & rhythm) forming a gestalt
  
  • Examples: “Twinkle, Twinkle, Little Star” or “Baa Baa Black Sheep”
• Defined by relationship between successive notes, not specific sounds
• Melodies can change octaves or keys and still be the same melody
• Notes and chords vary in duration
• Tempo: The perceived speed of the presentation of sounds
• x
• Melodies/themes as (gestalts) building blocks of music
• Fugue: a compositional technique (in classical music) in two or more voices, built on a subject (theme) that is introduced at the beginning and then repeated at different pitches.
  
  • Long held chord -> pattern = theme in the fugue in different pitches

Rhythm: not just in music

• Lots of activities have rhythm: Walking, waving, finger tapping, etc.
• Bolton (1894) played sounds perfectly spaced in time. Listeners are predisposed to group sounds into rhythmic patterns
• Car and train rides

Question 7

• # of speech sounds
• vocal tract
• 3 parts of speech production & organs
• x
• respiration & phonation
  
  • process of initiating speech
  • children vs adult vocal folds
    
    • → voice pitch
• Articulation
  
  • What are sounds modulated by?
  • how do we change the shape of vocal tract?

Answer 7

Speech production

• Humans are capable of producing lots of different speech sounds:
  
  • 5000 languages spoken today, utilizing >850 different speech sounds ->
  • flexibility of the human vocal tract: can learn more than 850 diiff speech sounds
• Vocal tract: The airway above the larynx used for the production of speech
• X
• – Respiration (lungs): also diaphragm and trachea
• – Phonation (vocal cords): also larynx and vocal box
• – Articulation (vocal/oral tract): the circle area
  
  • Involves alveolar ridge, tongue, hard palate, soft palate/velum, teeth, lips, epiglottis, nasal tract, pharynx (air flow through)
• X
• Respiration and phonation
  
  • Initiating speech: Diaphragm pushes air out of lungs, through trachea, up to larynx
  • At larynx: Air must pass through two vocal folds
    
    • Children: Small vocal folds -> higher pitched
    • Adult men: Larger mass of vocal folds (or when you have a cold) -> lower pitch
    • Thus, larynx produces canonical sounds with vocal folds
• Articulation (sounds are modulated by vocal tract)
  
  • Area above larynx: Vocal tract
  • Humans can change the shape of their vocal tract by manipulating their jaws, lips, tongue body, tongue tip, and velum (soft palate)
    
    • Resonance characteristics created by changing size and shape of vocal tracts to affect sound frequency distribution

Question 8

• formants
• Describe top graphs: a → b → c
• What is the vocal tract doing?
• In the spectrogram, what do the 3 stripes correspond to?
  
  • Boot vs bat

Answer 8

Formants

• Peaks in speech spectrum: Formants
  
  • Labelled by number, from lowest to highest (F1, F2, F3, …)—concentrations in energy occur at different frequencies, depending on length of vocal tract
  • If LS top graph = the harmonic spectrum of what comes out of your vocal folds/larynx
    
    • Vocal tract will modify it to look like LS bottom graph
    • Ex. top graph: fundamental freq most powerful -> bottom: 5^th harmonic has the most power
    • These peaks = formants, also seen to timbre after effect
  • So the vocal tract is acting like a filter function, that amplify certain freq (F1,2,3)
  • This depends on shape of vocal tract
• Formants show up in spectograms as bands of acoustic energy that undulate up and down, depending on the speech sounds being produced.
• Spectrogram
  
  • X-axis = time
  • Y-axis = freq
  • Color: energy
    
    • 3 types of stripes = formants
• The speech sounds produced depend on the tongue position (front, back, high, low)
  
  • Ex. boot (back, high) vs bat (front, low)
• Spectrogram: A pattern for sound analysis that provides a three-dimensional display plotting time on the horizontal axis, frequency on the vertical axis, and intensity on a color or gray scale

Question 9

• what do VOWELS do to the vocal tract?
  
  • how do we produce diff vowels?
• 3 modes of obstruction when producing consonants
  
  • bah vs dah vs gah
  • s vs t
  • zoo vs sue
• Issue when we have to produce speech fast?
• Coarticulation
  
  • Definition
  • Explanation
• Bah vs boo spectrogram
  
  • why does the b look different?
• How are sounds perceived despite coarticulation?
• x
• Categorical perception
  
  • Sound: bah, dah, gah spectrum
    
    • For the sounds w/in categories, how do ppl perceive them?
  • Visual: blue → green spectrum
    
    • For the colors w/in categories, how do ppl perceive them?
• McGurk effect Exp
• x
• How do we use categorical perception despite a lack of invariances?
  
  • It depends on X??
  • Color cube
  • Baa vs Boo
  • Why does eebah and oodah look the same?
  • How do we perceive syllables?
• What other phenom is this also related?
• explain
• Baa vs boo

Answer 9

Classifying speech sounds

• Sound: Most often described in terms of articulation
• Vowels open the vocal tract, differently shaped tongue and lips
• Consonants obstruct the vocal tract: 3 modes that obstruction can happen
  
  • Place of articulation (e.g., at lips, at alveolar ridge, etc.)
    
    • Ex. “bah”: b is obstructing air flow w/ lips
    • Ex. dah: d obstructs air flow w/ alveolar ridge
    • Ex. gah: g obstructs air flow w/ back of the tongue
  • Manner of articulation (totally/partially/slightly obstructed airflow)
    
    • “s”: partially obstructed
    • “t”: totally obstructed
  • Voicing: Whether the vocal cords are vibrating or not
    
    • Ex. zoo from Z = vibrating starting w/ “z”
    • Ex. Sue = larynx is not really vibrating from “s”
    • • Speech production: Very fast
  • Since we need to move things (ex. thing in out vocal tract) really fast, things might not be at the ideal position to create the sound
• Coarticulation: The phenomenon in speech whereby attributes of successive speech units overlap in articulatory or acoustic patterns
  
  • Inertia prevents tongue, lips, jaw, etc. from moving too fast
    
    • Ex. you can only move your tongue so fast
    • Ex. Bah vs boo spectrogram
      
      • There’s overlap b/w vowel that comes after b
      • Here the b for both words look different
      • B is different depending on what comes next
  • Lack of invariance (there’s variance)
• -> How are sounds perceived despite coarticulation?
• • X
• Categorical perception
  
  • (Artificial) sound stimuli can vary continuously from “bah” (LS) to “dah” (middle) to “gah” (RS)
    
    • Aka continuum
    • • For visual perception: if it is blue -> green, we see blue, bluish grey, grey, greenish grey, green
  • But people do not perceive continuous variation, they perceive sharp categorical boundaries btw. stimuli; don’t perceive differences btw. sounds within category
    
    • Ex. we perceive bah, dah, gah
    • If you play things b/w bah and dah, they either say bah or gah and are absolutely convinced about it even though the stimuli are ambiguous
• McGurk effect Exp: See gah, hear bah. Perceive dah
  
  • audiovisual illusion that illustrates categorical perception
• x
• So then: how categorical perception despite a lack of invariances?
  
  • It depends on context
  • Similar to color perception
    
    • We see the red tile in the shadow as red as well as the red tile in the light as right
    • The top tile looks brown, the one in the shadow looks orange: they are physically the same but we perceive them differently
    • Perception of speech depends on the context (ex. what you hear b4)
• Back to coarticulation …
  
  • Ex. Baa vs boo = b are very different depending on the vowel coming after b/c of coarticulation
  • It’s more complicated
  • ‘bah’ after ‘ee’ (LS top graph) very similar sound as ‘dah’ after ‘oo’ (RS bottom graph)
    
    • Here “bah” and “dah” looks the same due to coarticulation
    • • We perceive syllables on the basis of the relative change in the spectrum: spectral contrast.
    • Spectral context (ex. we just heard “ee” or we just heard the “oo”)
    • Relative to the spectral context, we hear “b” and “d” differently
• This is also related to the timbre aftereffect
  
  • Spectral contrast results in adaptation to timbre
  • So we perceive the “b graph sound” as ee
  • We can modify things based on the spectral contrast/context
    *

Question 10

• do we use only one cue to recognize faces/objects/sounds?
• Does this word for languages we are learning?
• So what does perception depend on?
• Prenatals
  
  • what voice do their prefer?
  • What language do their prefer?
• “r”/”l” for Japanese vs English
• bb w/ English speaking parents vs Bb w/ Hindi speaking parents: describe graph
• Conclusion
• x
• 2nd challenge when learning a 2nd langauge
• x
• Saffran, Aslin, and Newport (1996):
  
  • Methods: learning new language
    
    • 3 steps
  • Statistical learning
    *

Answer 10

• Using multiple acoustic cues (for speech perception, we use multiple cues together to recognize speech sounds)
  
  • This is Similar to face recognition
    
    • Similar features can be used in different combinations and we can rely on multiple cues to recognize a face
• -> use multiple cues: this doesn’t work well on languages we are still learning
  
  • So, Perception depends on experience
• X
• Learning to listen
  
  • Babies learn to listen even before they are born!
  • Prenatal experience: Newborns prefer hearing their mother’s voice over other women’s voices
  • 4-day-old French babies prefer hearing French over Russian (prefer the language their mom’s speak)
• X
• Becoming a native listener
  
  • There are more then 850 speech sounds, but not all of these sounds are used in each language; we do not need the ability to distinguish between all of these speech sounds
  • Sound distinctions are specific to various languages
  • Example: “r” and “l” are not distinguished in Japanese
  • Infants begin filtering out irrelevant acoustics long before they start to say speech sounds
  • Exp: bb w/ English speaking parents vs Bb w/ Hindi speaking parents
    
    • When you play Hindi sounds to 6-8 mo bb w/ English speaking parents -> they are responsive
    • Their responses declines
    • 10-12 mo bb w/ English speaking parents: not responsive, they perceive them as irrelevant
    • 11-12 mo bb w. Hindi speaking parents -> still very responsive
  • Thus, as we grow, our minds spend more resources to learn our own language, so it only respond to sounds that are relevant in our language
• Another challenge when learning a 2^nd language
  
  • Learning words
    
    • How do we know where one word ends and another begins?
    • Ex. the spectrogram for the sentence “where are the silences b/w words”
      
      • Here “where are the” is spoken continuously w/ very little pauses
      • We can be choppy: where, are, the; but no one speaks like this
      • For new learners, if becomes difficult to tell where the words end/start
• X
• Bb learn new words really quickly: hear it once, knows it
• Research by Saffran, Aslin, and Newport (1996):
  
  • # 1: Created a novel language (w/ it’s own phonetic rules) and infants listened to sentences for two minutes
    • Phonetic rules: how likely is “s” coming after “p” (ex. spoon)
      
      • IOW: how likely are certain phonemes in this specific sequence (examine w/in words rather than b/w words?)
    • tupiro, golabu, dapiku, tilado
  • # 2: played novel words from that language bb have never heard b4 (those novel words have the same phonological rules)
  • Afterwards, infants could already distinguish between words and non-words in the novel language
• Statistical learning: Certain sounds (making words) are more likely to occur together and babies are sensitive to those probabilities
  
  • How can you tell if sounds belong to the same words vs 2 different words being pieced together
    
    • IOW: The phonological rules w/in words are regular; while the phonological rules b/w word (from one word to another) do not apply anymore
    • This is how we can tell and infants learn them fast
• Saffran suggests that bb learn to pick words out of the speech stream by accumulating experience w/ sounds that tend to occur together
  *

Question 11

• Why can’t we study speech in other species?
• Why other techniques do we use?
  
  • Brain damages: Why is this not an ideal technique?
  • Nroimaging: 2 main issues?
• Listening speech: what brain areas are more active?
• What brain areas are more activated for complex language tasks?
  
  • brain area?
  • L vs HR?
    *

Answer 11

Speech

Speech in the brain

• For visual system: we can conduct studies on animals as there are similarities b/w the visual system for humans and animals
  
  • We can’t do this for speech as other animals do not have similar capabilities
• Use different techniques
  
  • Identify brain areas for speech
  • We cannot damage patients’ brain to study this; cannot only study this when bad events (stroke, tumors) happen
    
    • Issue: stroke -> regular damage patterns to blood vessels
    • Language deficits symptoms look different for patients who have a stroke vs tumors removed
    • The patterns of blood vessels do not say anything specific about whether specific structures for language production are close together in the brain
    • Is it because of neighboring structures in the brain or is it about the neighboring structures that are perfused by the same blood vessels
    • This is the main difficulty of nropsych studies in patients
• Brain damage follows patterns of blood vessels, not brain function, so difficult to study
• TB: Damage from stroke may cover just part of a particular brain fx -> leaving some of the fx undamaged
• Or damage can cover a wide region that includes some or all of a particular brain fx and part of other fx
• Examine healthy ppl w/ EEG, MEG, PET and fMRI studies: Help to learn about speech processing in the brain
  
  • Correlational only
  • Hard to create well-controlled nonspeech stimuli because humans are so good at understanding even severely distorted speech
    
    • Challenges: difficult to find control conditions
    • Ex. there’s wind passing by in experiments, ppl interpret this pattern even though it’s just random noise in the lab
    • IOQ: perceptual system is so sensitive to speech that is attempts to perceive speech in control conditions
• Listening to speech: Left and right superior temporal lobes are activated more strongly in response to speech than to nonspeech sounds
• X
• Categorical perception tasks: Listeners attempt to discriminate sounds like “bah” and “dah” while having their brain scanned
• As sounds become more complex, they are processed by more anterior and ventral regions of superior temporal cortex – in both hemispheres
• As sounds become more speech-like, more activation in the left brain
• Research indicates that some “speech” areas become active when lip-reading (multisensory aspects of speech processing?)