π Lecture 34 β Linguistic Phonetics-I II
π Overview: This lecture explores the phonetic dimensions of individual speech production, focusing on how articulatory movements are controlled and how speech is stored in memory. It addresses the balance between speaker and listener needs, explaining the phonetic forces that shape sound patterns and historical changes in pronunciation.
ποΈ Topics Covered
The lecture examines controlling articulatory movements through muscular complexity, memory for speech and the lack of phonetic invariance, exemplar theory's explanation of speech memory, the balance between phonetic forces like ease of articulation and perceptual separation, and how languages maintain this balance through assimilation and contrast maintenance.
π Lecture Summary
Topic-171: Controlling Articulatory Movements
This section explores the speech motor control underlying individual phonetic production. For example, producing a voiceless bilabial [p] involves an array of muscular complexity including dozens of muscles in the chest, abdomen, larynx, tongue, throat, and face β all contracted with varying degrees of tension in specific sequence and duration. For lip closure, two main muscles (depressor labii inferior and incisivus inferior) are activated to create 'too much' and 'enough tension'. Simultaneously, jaw muscles trade with lip muscles for closing and opening. This structure specifies an overall task "close the lips" at the top node, with subtasks like "raise the lower lip" and "lower the upper lip" coordinated to accomplish the overall goal. Additionally, for voiceless bilabial [p], the glottis needs to be wide apart for free air passage. Thus, exploiting the air passage, keeping it voiceless at larynx, and creating closure by lips and jaws with many subtasks are achieved through controlling articulatory movements, enabling understanding of individual variation in speech production.
π‘ Why this matters: This shows that even a simple sound like [p] involves complex motor coordination, explaining why individual speakers produce sounds differently.
Topic-172: Memory for Speech
Speech is diverse and complex, particularly regarding the phonetics of the individual. Different speakers of the same language will have somewhat different productions depending on vocal tract physiology, their own habits of speech motor coordination, and their memory of speech. Sociolinguistic features also influence production β we are exposed to speech styles ranging from careful pronunciations in public speaking to casual styles typical between friends. This leads to lack of phonetic invariants (or variability of invariability). This 'lack of phonetic invariance' poses an important problem for phonetic theory as we try to reconcile that shared phonetic knowledge can be described using IPA symbols and phonological features with the fact that individual phonetic forms span a very great range. This problem has great practical significance for language engineers trying to get computers to produce and recognize speech. The solution is facilitated by 'the phonetic implementation approach' that focuses on how experiences are encoded in memory. According to this view, words are stored in speech memory in their most basic phonetic form and used when needed.
π Definition β Lack of phonetic invariance: The phenomenon where the same linguistic unit (e.g., a phoneme) is produced with highly variable acoustic and articulatory characteristics across different contexts, speakers, and speaking styles.
Topic-173: Explaining Memory for Speech
The role of memory for speech under exemplar theory suggests that many instances of each word are stored in memory, and their phonetic variability is memorized rather than computed. The main postulates include:
- Language universal features: Broad phonetic classes (e.g., aspirated vs. unaspirated) derive from physiological constraints on speaking or hearing, but their detailed phonetic definitions are arbitrary β a matter of community norms.
- Speaking styles: No one style is basic (from which others are derived), because all are stored in memory. Bilingual speakers store two systems.
- Generalization and productivity: Exemplar theory says generalization is possible within productivity. However, productivity β the hallmark of linguistic knowledge in the phonetic implementation approach β is the least developed aspect of exemplar theory.
- Sound change: Sound change is phonetically gradual and operates across the whole lexicon. It is a gradual shift as new instances keep on adding.
π Definition β Exemplar theory: A theory of memory for speech which posits that all experienced instances (exemplars) of words are stored in memory, and phonetic categories emerge from the similarity structure of these stored tokens.
Topic-174: The Balance Between Phonetic Forces
To explain the sound patterns of a language, the views of both speaker and listener are considered. Both prefer to use the least possible articulatory effort (except when producing very clear speech), resulting in numerous assimilations, with some segments left out and others reduced to minimum. Thus a speaker uses language with ease of articulation (e.g., co-articulation and secondary articulation). This tendency toward maximum ease of articulation leads to change in pronunciation of words. For example, in co-articulations, a change in the place of the nasal and following stop occurred in words like improper and impossible before these words came into English through Norman French. In these words, the [n] from the prefix in- (as in intolerable and indecent) changed to [m], reflected even in spelling. These are historical cases of assimilation, where one or more segments are affected by adjacent segments for economy of articulation. Articulation with least possible effort enables us to speak by keeping the balance between phonetic forces, ultimately leading to changes in pronunciation.
π‘ Why this matters: Historical sound changes like in- becoming im- before labials show how ease of articulation shapes language over time, even affecting spelling.
Topic-175: Explaining the Balance Between Phonetic Forces: Explanation
When producing sounds with maximum ease of articulation, only similar sounds are affected. The focus of speakers is always on maintaining a sufficient perceptual distance between sounds that occur in a contrasting set (e.g., vowels in stressed-monosyllabic words beat, bit, bet, and bat). This principle of perceptual separation does not usually result in one sound affecting an adjacent sound (as in maximum ease of articulation). Instead, perceptual separation affects the set of sounds that potentially can occur at a given position in a word β such as the vowel position in stressed monosyllables β so that perceptual separation is maximized. The principle of 'maximum perceptual separation' also accounts for some differences between languages. These examples illustrate how languages maintain a balance between the requirements of the speaker (pressure for easier articulations) and those of the listener (sufficient perceptual contrast between sounds that affect meaning).
π Definition β Maximum perceptual separation: The principle that sounds in a contrasting set (e.g., vowels in minimal pairs) tend to be maintained at sufficient perceptual distance to preserve distinctiveness and avoid confusion in meaning.
β Key Takeaways
The lecture emphasizes that speech production involves complex motor control where multiple muscles coordinate hierarchically to achieve articulatory goals like lip closure for [p]. Memory for speech, explained through exemplar theory, stores many instances of words with their phonetic variability rather than computing them, which accounts for lack of phonetic invariance across speakers and styles. Languages balance two opposing phonetic forces: the speaker's drive for ease of articulation (producing assimilations like [n] becoming [m] before labials) and the listener's need for perceptual separation between contrasting sounds (e.g., maintaining distinct vowels in beat, bit, bet, bat). This balance explains both synchronic variation and historical sound change, with phonetic implementation approach and exemplar theory providing frameworks for understanding how individual phonetic forms relate to shared linguistic knowledge.
π§ Quick Revision Questions
- What muscular actions are involved in producing the lip closure for [p], and how do subtasks coordinate?
- What is the "lack of phonetic invariance" and why is it a problem for phonetic theory and language engineering?
- According to exemplar theory, how are words stored in memory and how does this account for phonetic variability?
- How does the principle of ease of articulation explain the historical change from [n] to [m] in words like improper?
- What is the principle of maximum perceptual separation and how does it balance speaker and listener needs in contrasting sets like beat, bit, bet, bat?
π Lecture 35 β PHONOTACTICS AND SYLLABIC TEMPLATES
π Overview: This lecture explores phonotactics and syllabic templates, focusing on how languages organize sounds into syllables and the permissible sequences of phonemes. It explains syllabification, the universal CV syllabic pattern, and how languages classify syllable structures as simple, moderate, or complex based on consonant clusters. Understanding these concepts is crucial for analyzing phonological well-formedness and cross-linguistic variation.
ποΈ Topics Covered
The lecture covers three main topics: Syllabic Templates and Syllabification, where the syllable as a unit of pronunciation is defined and the universal CV pattern is introduced; Explaining Syllabification, which discusses internal syllable structure, consonant clusters, and the classification of languages into simple, moderate, and complex syllabic patterns; and Phonotactics, which examines the study of permissible sound sequences in languages, including sequential constraints, phonotactic rules, and examples from English phonotactics.
π Lecture Summary
Topic-176: Syllabic Templates and Syllabification
A syllable is the unit of pronunciation typically larger than a single sound and smaller than a word. Words may be divided into parts, as in ne-ver-the-less, and dictionaries indicate syllabic divisions for hyphenation. Syllabification refers to the division of a word into syllables, while resyllabification is a reanalysis that alters syllable boundary locations. A word with one syllable is monosyllabic; with more than one, it is polysyllabic. The CV pattern (one consonant at the onset followed by a vowel as its peak) is found in all languages of the world β it is the universal syllable pattern (Max Onset Principle). Some languages only allow CV templates, e.g., Honolulu (CVCVCVCV) and Waikiki (CVCVCV). This pattern is also common in nicknames across languages: kami, nana, baba, papa, mani, rani. Children first acquire the CV pattern of their mother tongue during L1 acquisition.
Topic-177: Explaining Syllabification
Languages differ in their internal syllable structure, and syllabic patterns vary across languages and families. Consonant sequences are called clusters (e.g., CC β two consonants or CCC β three consonants). Most phonotactic analyses are based on syllable structures and syllabic templates. On the basis of consonant clusters at edges (onset and coda), three types of syllabic patterns are identified:
π Definition β Simple syllabic pattern: CV (e.g., to, be) π Definition β Moderate syllabic pattern: CVC(G)(N) β where G stands for Glide and N stands for Nasal (specific consonants only) π Definition β Complex syllabic pattern: CCVCC to CCCVCC β involving bipartite (CC) and tripartite (CCC) clusters
Topic-178: Phonotactics
Phonotactics is the study of phonemes and their order found in syllables β the study of sound sequences in a language. Languages do not allow all phonemes to appear in any order. For example, an English speaker knows that /streΕΞΈs/ makes the English word strengths and that /bleidΚ/ would be acceptable as blage (though it does not exist), but /lvm/ could not be part of an English word.
Phonotactic analyses reveal interesting findings: Why do words like bump, lump, hump, rump, mump(s), clump associate with large blunt shapes? Why do words ending with a plosive and a syllabic /l/ (e.g., muddle, fumble, straddle, cuddle, fiddle, buckle, struggle, wriggle) relate to clumsy, awkward, or difficult action? Why can't English syllables begin with /pw/, /bw/, /tl/, /dl/ when /pl/, /bl/, /tw/, /dw/ are acceptable? All such discussion is called phonotactics of the language.
179. Explaining Phonotactics
Phonotactics refers to the order (sequential arrangements or tactic behavior) of segments (sounds or phonological units) that occur in a language. It shows what counts as a phonologically well-formed structure of a word. The allowed and restricted sound patterns of language are found through phonotactics. For example, in English, consonant sequences like /fs/ and /spm/ do not occur initially in words, and there are many other restrictions on possible consonant+vowel combinations. By analyzing data, sequential constraints of a language can be stated as phonotactic rules.
According to Generative phonotactics, no phonological principles can refer to morphological structure; phonological patterns sensitive to morphology (e.g., affixation) are represented only in the morphological component of grammar, not in phonology.
Examples of English phonotactic patterns:
| Pattern | Example |
|---|---|
| One phoneme (V) | I, oh, owe |
| Two phoneme (CV) | to, be, see |
| Three phoneme (CVC) | cat, dog, run |
| Four phoneme (CCVC) | stick, click, brick |
| Five phoneme (CCVCC) | brisk, treats, speaks |
| Six phoneme (CCCVCC) | streets, strand, strips |
| Seven phoneme (CCCVCCC) | strengths |
Other possible patterns: CCV (try), CCCVC (stroke), CCCV (straw), VCC (eggs), CVCC (risk), CVCCC (risks).
β Key Takeaways
The most critical points from this lecture are: (1) The CV pattern is the universal syllable template found in all languages, and children acquire it first; (2) Syllable structures are classified as simple (CV), moderate (CVC(G)(N)), or complex (CCVCC to CCCVCC) based on consonant clusters at onset and coda; (3) Phonotactics studies permissible sound sequences in a language, revealing both allowed and restricted patterns; (4) Languages impose sequential constraints β certain consonant clusters are allowed initially (e.g., /pl/, /tw/) while others are not (e.g., /pw/, /tl/) in English; (5) Generative phonotactics holds that phonological principles cannot refer to morphological structure β patterns sensitive to morphology belong to the morphological component, not phonology.
π§ Quick Revision Questions
- What is the universal syllable pattern found in all languages, and why is it significant?
- How are languages classified based on their syllabic patterns, and what do the terms "simple," "moderate," and "complex" refer to?
- Define phonotactics and provide two examples of permissible and two examples of impermissible English syllable-initial consonant clusters.
- According to Generative phonotactics, where should morphological sensitive phonological patterns be represented?
- Give one example each of a five-phoneme, six-phoneme, and seven-phoneme English word pattern, and explain what each pattern indicates about consonant clusters.
π Lecture 36 β Using PRAAT-I
π Overview: This lecture introduces the PRAAT software, a free and powerful tool for phonetic analysis, synthesis, and manipulation of speech. It is designed to equip students with fundamental skills for recording, displaying, segmenting, labeling, and exporting speech data, which are essential for acoustic phonetics research.
ποΈ Topics Covered
The lecture begins with an introduction to PRAAT, its manual, and basic exploration of the software. It then guides students through the practical steps of recording and displaying speech, segmenting and labeling sound files using textgrids, and exporting visual displays of waveforms and spectrograms to a Word document for reporting purposes.
π Lecture Summary
Topic-180: Introduction to PRAAT
PRAAT is a freeware program created by Paul Boersma and David Weenink at the Institute of Phonetic Sciences, University of Amsterdam. It can be freely downloaded from http://www.praat.org or http://www.fon.hum.uva.nl/praat/. The software allows you to analyze, synthesize, and manipulate speech, and create high-quality pictures for articles and theses. Active discussions and blogs are available on the 'yahoo Praat group', and an introductory tutorial is on the homepage. π‘ Why this matters: PRAAT is the standard tool in modern phonetics and phonology research, replacing many expensive alternatives.
π Definition β PRAAT: A computer program used for analyzing, synthesizing, and manipulating speech, and for creating high-quality visual displays.
Topic-181: Introduction to PRAAT Manual
The manual used in this course was developed by Sonya Bird and Qian Wang at the University of Victoria, Canada. It contains carefully designed labs with worksheets. This course will conduct important experiments from the first five labs, enabling students to continue independently. PRAAT requires only modest computer experience, and queries can be addressed to the developers or the 'yahoo Praat group'. It is essential for learning acoustic phonetics and future research work.
Topic-182: Exploring PRAAT
Start by downloading PRAAT from www.praat.org based on your PC or Mac specifications. Initially, explore the two main windows: the Object window and the Picture window. To begin, double-click on the icon to open PRAAT and get to know its layout, including buttons on the Object window, main menu, analysis and synthesis tools, and the Picture window options.
Topic-183: Recording and Displaying
This is the first practical experiment. To record a new mono sound:
- Go to NEW > Record mono-sound (with a sampling rate of 44100 Hz).
- Ensure the volume bar fluctuates; if not, you're not recording or speaking loudly enough.
- Watch out for clipping: if the recording level goes into the red on the volume bar, you'll get a clipped signal, which is bad for speech analysis.
- Give the recording a name in the "Save to list" box and click "Save to list."
To display the recording:
- In the Objects window, highlight the sound file and click Edit (on the "Analysis and synthesis tools" panel).
- To extract a portion for study, select the desired part, then go to File > Extract Selection. The extracted selection appears as 'sound untitled' in the Objects window. Close the Edit window.
π Definition β Clipping: A distortion that occurs when the recording level is too high, causing the signal to be cut off at the peaks; it is detrimental to speech analysis.
Topic-184: Segmenting and Labeling
Segmenting and labeling is crucial for acoustic analysis, allowing you to place segmental symbols and add textgrids to spectrograms. Follow these steps:
-
Create a textgrid: Highlight the sound file in the Objects window, then go to Annotate > To TextGrid. Create two tiers (e.g., 'word segment') and enter them in the 'All tier names' cell.
-
Open sound file and textgrid together: Hold down Ctrl and click on both files to highlight them. Click Edit to see the waveform (top), spectrogram (middle), and textgrid (bottom).
-
Segment the file: Place the cursor at the beginning of a target name on the spectrogram/waveform to show a boundary line. Click in the little circle at the top of the word tier to create a boundary. To remove a boundary, highlight it and go to Boundary > Remove or hit Alt+backspace.
-
Label the intervals: Select/highlight the target interval by clicking between two boundaries (the interval turns yellow). To input or change text, edit in the Textbox above the spectrogram. Give each interval a name. For phonetic symbols, go to the TextGrid editing window and select Help > Phonetic symbols.
π Definition β Textgrid: An annotation file in PRAAT that contains tiers with boundaries and intervals for labeling segments of a sound file.
Topic-185: Exporting Visual Display to Word File
Exporting the visual display to Word is part of the experiment's write-up. To export your labeled waveform and spectrogram:
- Maximize the Edit window by clicking the square at the top right.
- Hit the PrtScr (Print Screen) button on your keyboard.
- Open a Word document, provide an introduction as required, and hit Ctrl+V to paste the PRAAT image.
- Give the figure a number and title below the image.
β Key Takeaways
The critical points from this lecture are: PRAAT is a free and essential tool for acoustic phonetics used to analyze, synthesize, and manipulate speech. The first practical step is recording a mono sound at a 44100 Hz sampling rate while avoiding clipping. Segmenting and labeling require creating a textgrid with tiers, opening it alongside the sound file, and placing boundaries between intervals on the waveform/spectrogram. Finally, exporting the visual display to Word is done by maximizing the Edit window, using the Print Screen key, and pasting the image. Mastering these basics enables independent future work in phonetic research.
π§ Quick Revision Questions
- What is PRAAT, and who created it?
- What is the recommended sampling rate for recording mono sound in PRAAT, and why is clipping harmful?
- Describe the steps to create a textgrid and open it together with a sound file for editing.
- How do you create and remove a boundary on a textgrid tier?
- What is the simple method to export a labeled waveform and spectrogram from PRAAT into a Word document?
π Lecture 37 β Using Praat-II
π Overview: This lecture focuses on the practical application of Praat for acoustic analysis of speech sounds, particularly sonorous sounds like vowels. It explains the Source-Filter Model of speech production and provides step-by-step instructions for measuring fundamental frequency, harmonics, and formants using Praat software. Understanding these measurements is crucial for analyzing the acoustic components of speech.
ποΈ Topics Covered
This lecture covers the Source-Filter Model of speech, followed by three measurement techniques: measuring fundamental frequency (F0/pitch) using three methods, measuring harmonics using narrow-band spectrograms, measuring formants using wide-band spectrograms and three techniques, and finally exploring the relationship between harmonics and formants. The lecture emphasizes the distinction between harmonics (source/laryngeal) and formants (filter/vocal tract).
π Lecture Summary
Topic-186: The Source-Filter Model of Speech
The Source-Filter theory is crucial for understanding the basic components of speech sounds and the nature of acoustic signals (the physics of speech sounds). It is particularly important for acoustic analysis of vowels and vowel-like sonorous sounds. The major goal is to understand and explore basic acoustic components of sonorous speech sounds such as fundamental frequency, harmonics, and formants. These components create the acoustic signals associated with speech, and understanding them is crucial to understanding what we actually hear.
Topic-187: Measuring the Fundamental Frequency
Fundamental frequency (F0 or Pitch) is an important component of source-filter theory. It can be taken from the middle of the sound (e.g., vowel) using three techniques: displaying the Pitch track (automatically), measuring from Pitch track manually, and by looking at the waveform by highlighting one cycle.
Method 1: Displaying the pitch track and allowing Praat to measure pitch automatically:
- Display the pitch track: Pitch > Show pitch.
- Place your cursor in the middle β a stable portion of the vowel.
- Go to Pitch > Get pitch β a box will appear with the pitch value in it.
Method 2: Displaying the pitch track and measuring pitch manually:
- Display the pitch track - Pitch > Show pitch.
- Click on the blue pitch track in the middle of the vowel.
- A red horizontal bar should appear with the pitch value (in dark blue) on the right side of the window.
Method 3: By looking at the waveform (top of the display):
- Zoom into a small piece of the waveform in the middle of the vowel and measure the period by highlighting one complete cycle and noting the time associated with it.
Topic-188: Measuring the Harmonics
Harmonics are the multiple integers of the fundamental frequency which are basically the result of vocal fold vibration (complex wave). We need the narrow band spectrogram for measuring the H (which we can set by fixing the spectrum setting at 0.025). By starting measuring the frequency of the first three harmonics, we go to the H10 (H1, H2, H3 β H10). When vocal folds vibrate, the result is a complex wave, consisting of the fundamental frequency plus other higher frequencies called harmonics.
To see harmonics, we need a narrow-band spectrogram, which is more precise along the frequency domain than the default wide-band spectrogram.
Procedure:
- Display a narrow-band spectrogram: Spectrum > Spectrogram settings.
- Change the window length to 0.025s (the default window length is 0.005s for wide-band spectrogram).
- Notice the grey horizontal bands corresponding to harmonics.
- For each vowel, measure the frequencies of the first 3 harmonics (H1-H3) and the 10th harmonic (H10).
- Click on the center of each harmonic in the center of each vowel.
- A red horizontal bar should appear with the frequency value on the left side of the window in red.
π Definition β Harmonics: Multiple integers of fundamental frequency that result from vocal fold vibration (complex wave).
π Formula: Harmonics = n Γ F0 (where n = integer, F0 = fundamental frequency) β Harmonics are whole-number multiples of the fundamental frequency.
Topic-189: Measuring the Formants
Formants are the overtone resonances. Acoustically, to plot vowels on a chart, F1-F2 are very important. Narrow band spectrograms are required for measuring harmonics whereas we need the wide bands for measuring formants (important characteristics of sonorant speech sounds β vowels). On spectrogram, formants are thick bands (darkness corresponds to loudness; the darkest harmonics are the most amplified). These amplified harmonics form the formants characteristic of sonorant speech sounds.
Method 1: Displaying the formants (red dots on the spectrogram) automatically:
- Display the formant track: Formant > Show formants.
- Place cursor in the middle, stable portion of the vowel.
- Go to Formant > Formant listing: a box appears with time point and first four formants.
Method 2: Displaying the formants and measuring the frequency manually:
- Display the pitch track: Formant > Show formants.
- Place cursor in the center of each formant, in the middle of the vowel.
- A red horizontal bar appears with the frequency value on the left side of the window in red.
Method 3: Measuring the frequency without displaying Praat formants (if Praat's formant tracking goes wonky):
- Remove Praat's formant tracking: Formant > Show formants (unclick).
- Place cursor in the center of each formant, in the middle of the vowel.
- A red horizontal bar appears with the frequency value on the left (in red).
π Definition β Formants: Overtone resonances; amplified harmonics that are characteristic of sonorant speech sounds.
π‘ Why this matters: F1-F2 are essential for plotting vowels on a vowel chart, which is fundamental for acoustic phonetic analysis.
Topic-190: Relationship Between Harmonics and Formants
Having taken measurements for both formants and harmonics, we compare them to explore the relationship between the two. Captured in the Source-Filter Model of speech, from comparison of the two values, it is clear that harmonic numbers are different for one type of sounds but the formants are the same. The relationship is: harmonics are related to the laryngeal activity (source) and formants are the output of the vocal tract (filter).
The figure illustrates: Source (vocal folds produce complex wave with harmonics) β Filter (vocal tract shape amplifies certain harmonics to create formants) β output speech sound.
β Key Takeaways
Fundamental frequency (F0) can be measured using three methods: automatic pitch display, manual pitch tracking, or waveform cycle measurement. Harmonics are integer multiples of F0 resulting from vocal fold vibration and are measured using narrow-band spectrograms (window length 0.025s). Formants are amplified harmonics shaped by the vocal tract and are measured using wide-band spectrograms; the first two formants (F1, F2) are essential for vowel identification. The Source-Filter Model explains that harmonics come from the laryngeal source while formants are produced by the vocal tract filter. A key observation is that harmonic numbers vary across sounds but formant frequencies remain consistent for the same vowel.
π§ Quick Revision Questions
- What are the three methods for measuring fundamental frequency in Praat?
- What spectrogram setting is required for measuring harmonics, and why?
- What is the relationship between harmonics and formants according to the Source-Filter Model?
- Why are F1 and F2 measurements particularly important in acoustic phonetics?
- How can you measure formants if Praat's automatic formant tracking is unreliable?
π Lecture 38 β Using Praat-III
π Overview: This lecture guides students through the practical acoustic analysis of vowels using Praat software. It covers measuring intrinsic pitch (F0) and the first two formants (F1 and F2) for eight American English vowels, culminating in plotting these values on a chart in Excel. This hands-on session is critical for connecting acoustic measurements to articulatory properties of vowels.
ποΈ Topics Covered
The lecture comprises four main topics: first, exploring vowel properties by recording eight specific American English vowels. Second, measuring intrinsic pitch (F0) for each vowel using Praat's tools. Third, analyzing the spectral make-up by measuring the first two formant values (F1 and F2). Fourth, plotting the formant values on a chart in Excel, including reversing axes to create the traditional vowel space.
π Lecture Summary
Topic-191: Vowel Properties
To explore the acoustics of vowels, you need to record the eight vowels from American English: heed, hid, head, had, hod, hawed, hood, and whoβd. After recording, you will measure three things: intrinsic pitch, spectral make up (formants), and plot them in an Excel sheet. First, look at your vowels in the Edit window to ensure you can clearly see the vowel formants; if not, review previous labs.
Topic-192: Intrinsic Pitch
Measure the pitch (F0) for each of the eight recorded vowels using any of the three methods discussed in Topic-187. If values seem strange, use another method for confirmationβmeasuring the waveform is the safest way. To confirm pitch via a spectral slice:
- Select a 70-80ms portion of the vowel.
- Go to Spectrum > View spectral slice.
- Click on the first (big) peak = H1 = F0 (ignore small spikes, which may be noise). Note the frequency of this peak at the top of the vertical bar.
Use the confirmed pitch values to plot the pitch of each vowel on your Excel sheet. Label the y-axis with a scale that spreads out measurements. Create a cluster chart, export it to Word, and give it a figure number and title.
Topic-193: Spectral Make-up (Formant Values)
To determine the spectral make-up of the vowels, measure the first two formants (F1 and F2) for each of the eight vowels. Use either automatic formant tracking or manual measurement (trust your judgment over Praat's). Use the default wide-band spectrogram for measuring F1 and F2. Also, answer these questions by revisiting chapters on acoustics:
- a. Why do formants (F1 and F2) differ across vowels?
- b. What does F1 correspond to in terms of articulation?
- c. What does F2 correspond to in terms of articulation?
π‘ Why this matters: These questions directly link acoustic measurements (formants) to articulatory features (tongue height and frontness), forming the basis of vowel classification.
Topic-194: Plotting Vowels on Chart
Using the F1 and F2 measurements, plot the values on an Excel chart. Organize the data in three columns: vowels in the first column, difference between F2 and F1 in the second column, and F1 in the third column. Use the Scatter chart type. To make the chart correspond to the traditional vowel space, reverse the values for both formants on both axes (Y and X), so that zero for F1 and F2 is at the right corner. After plotting, note that F1 is inversely related to vowel height, and the difference between F2 and F1 is related to vowel frontness. Export the chart to Word and give it a number and title.
β Key Takeaways
The critical skills from this lecture are recording and analyzing the eight American English vowels in Praat, measuring both intrinsic pitch (F0) and the first two formants (F1 and F2). You must be able to confirm pitch measurements using spectral slices and use the wide-band spectrogram for manual formant tracking. The final, crucial step is plotting F1 against the F2-F1 difference in Excel, reversing both axes to produce the standard vowel space chart. This process directly demonstrates that F1 corresponds to vowel height (open/close) and F2-F1 difference to vowel frontness (front/back).
π§ Quick Revision Questions
- What are the eight American English vowels (as words) you must record for this lecture?
- Describe the "safest" method for measuring pitch (F0) if Praat's automatic values seem strange.
- How can you use a spectral slice to confirm your pitch (F0) measurement?
- In the Excel chart for formants, which two values do you plot on the axes, and which axis is reversed to create the traditional vowel space?
- What articulatory property does F1 correspond to, and what property does the difference between F2 and F1 correspond to?
π Lecture 39 β Using PRAAT - IV
π Overview: This lecture explores the acoustic analysis of sonorants (nasals, glides, and liquids) using Praat software. It focuses on measuring formants and identifying distinctive acoustic patterns that differentiate these sounds from vowels, which is essential for understanding acoustic phonetics.
ποΈ Topics Covered
The lecture covers sonorants and their formants, nasal formants and their distinctive waveforms including anti-formants and formant transitions, and glide formants and their comparison with vowels. It guides students through hands-on recording and analysis of sequences like /ama/, /ana/, /aΕa/, /wi/, and /ju/ to explore systematic acoustic patterns.
π Lecture Summary
Topic-195: Sonorants and Their Formants
Sonorants are vowel-like sounds, specifically nasals and glides. They are called sonorants because they have formants (the acoustic correlates of resonance). However, they differ from vowels because they generally have lower amplitude, which makes them behave like consonants. To analyze these sounds, record the sequences: /ama/ - /ana/ - /aΕa/ - /wi/ - /ju/.
π Definition β Sonorants: Vowel-like consonants (nasals and glides) that have formants but lower amplitude than vowels.
π‘ Why this matters: Distinguishing sonorants from vowels is crucial for identifying consonant types in spectrograms.
Having recorded these sequences, explore the features by measuring F1, F2, and F3 for each sound. Compare these measurements with those of vowels. Look for systematic patterns in the formant values before proceeding to the next topic.
Topic-196: Nasal Formants
Formants for nasal sounds are essential for acoustic analysis. Measure the first three formants (F1, F2, and F3) of nasals from the recorded files, using the previously learned method of measuring formants. Remember that nasals have very distinctive waveforms that differ from vowels. They have distinctive forms of anti-formants, which are bands of frequencies that are damped, and formant transitions.
π Definition β Anti-formants: Bands of frequencies that are damped or reduced in amplitude, characteristic of nasal sounds.
When measurements are complete, answer these questions:
- Are there any systematic patterns across nasals?
- Is there one formant with a similar frequency for all places of articulation?
- Is there one formant that has much higher amplitude than the others across nasals?
- Do you see any overall differences between the nasals on the one hand and [a] on the other?
π Example: Comparing /m/, /n/, and /Ε/ in the sequence /ama/-/ana/-/aΕa/ β the F1 value might be consistently around 250-300 Hz for all nasals (answer to Q2), while F2 varies with place of articulation (e.g., lower F2 for /m/ due to labial closure).
Topic-197: Glide and Their Formants
Glides, such as /w/ and /j/, are also sonorants that are vowel-like because they have formants. From the recorded files, take the first three formants (F1, F2, and F3) from the middle of the sounds (the midpoint of [w] and [j]). Analyze how similar the formant structure of glides is to that of vowels and nasals. Draw lines to indicate F1, F2, and F3 on the spectrogram and compare them with vowels.
π Definition β Glides: Sonorant consonants (e.g., /w/ and /j/) that have formant structures similar to vowels but function as consonants in syllable structure.
After analysis, answer these questions:
- Which vowel does each glide resemble? (e.g., /w/ resembles [u] and /j/ resembles [i])
- Are there any differences between the contours, i.e., between the transitions out of the two glides?
- What do you think may cause these differences (in terms of articulation)?
π Example: For /wi/ β the F1 and F2 of [w] will be low (like [u], F1300Hz, F2700Hz), while the [i] part will show a rapid transition to high F2 (~2300Hz). For /ju/ β [j] will show low F1 (~300Hz) but high F2 (~2300Hz, like [i]), transitioning to lower F2 for [u].
β Key Takeaways
This lecture emphasizes that sonorants (nasals and glides) are vowel-like consonants with formants but lower amplitude than vowels, making them acoustically distinct. Nasals are identified by their anti-formants and distinctive waveforms, while glides show formant structures that resemble specific vowels (/w/ like [u], /j/ like [i]). Systematic patterns across nasals include a consistent formant frequency for all places of articulation, and glides exhibit formant transitions that differ based on their articulatory origins. Understanding these acoustic features is critical for distinguishing sonorants from vowels in spectrographic analysis.
π§ Quick Revision Questions
- What makes sonorants different from vowels in terms of amplitude?
- What are anti-formants, and why are they important for identifying nasal sounds?
- Which vowel does the glide /w/ resemble, and which does /j/ resemble, based on formant structure?
- When measuring formants of nasals, what is the first step in finding systematic patterns?
- What causes the differences in formant contours between the two glides /w/ and /j/?
π Lecture 40 β Using PRAAT-V
π Overview: This lecture provides a practical, hands-on guide to analyzing acoustic features of obstruents (stops, fricatives, and affricates) using PRAAT software. It focuses specifically on how to measure and interpret voicing correlates in stops, including the voice bar, Voice Onset Time (VOT) , and preceding vowel duration, enabling students to distinguish voiced, voiceless, and aspirated stops acoustically.
ποΈ Topics Covered
This lecture covers the acoustic analysis of stop consonants using PRAAT. It begins with an examination of stop voicing on spectrograph, focusing on the voice bar and the duration of preceding vowels. It then moves to measuring Voice Onset Time (VOT) , with detailed steps for calculating and comparing negative, zero, and positive VOT across voiced, voiceless, and aspirated stops like /apa/, /aba/, /ata/, /ada/, /apha/, and /atha/. The lecture concludes with guided questions for self-assessment.
π Lecture Summary
Topic-198: Stop Voicing on Spectrograph
There are three important acoustic correlates of voicing in stops: the voice bar, VOT (Voice Onset Time), and the duration of the preceding vowel. To analyze these, record tokens like /apa/, /aba/, /ata/, /ada/, /apha/, and /atha/. For each stop in the file, take three measurements: first, see the voicing or the voice bar by exploring features of the stop. Second, explore features related to the place of articulation (e.g., any bilabial feature for /p/ or /b/ in comparison with non-bilabial sounds). Third, check the duration of the preceding vowels. Note down the presence of voicing. The voice bar appears as a low-frequency energy band during the closure period of a voiced stop. If the voice bar disappears, it may be due to a lack of glottal vibration or a very short closure duration.
π Definition β Voice Bar: A low-frequency energy band visible on a spectrogram during the closure of a voiced stop, indicating ongoing vocal fold vibration. π No formula; this is a visual measurement technique. π Example: When analyzing the word /aba/ in PRAAT, zoom into the spectrogram during the /b/ closure. Look for a dark horizontal line near the bottom of the spectrogram (around 0β200 Hz). If it is present throughout the closure, the stop is fully voiced; if it disappears mid-closure, the voicing may have ceased.
β Guided Questions:
- Do you see the voice bar? If the voice bar is present at all, does it last through the duration of the closure? Why do you think it might go away? (Possible reason: subglottal pressure equalizes, stopping vocal fold vibration.)
- Can you point out the features related to the place of articulation? Comment on them (e.g., bilabial stops show a diffuse burst spectrum; alveolar stops show a more compact burst).
- How about the duration of the preceding vowel? Is it affected by the voicing? (Vowels are typically longer before voiced stops than before voiceless stops.)
π‘ Why this matters: Distinguishing voiced from voiceless stops in a spectrogram is foundational for acoustic phonetic analysis and helps in understanding phonological contrasts.
Topic-199: Measuring Voice Onset Time (VOT)
Another aspect of acoustic correlate of stops is VOT, which is a characteristic of voiced, voiceless, and aspirated stop sounds. There are very easy steps to calculate VOT. Record /apa/, /aba/, /ata/, /ada/, /apha/, and /atha/. Zoom in through your stop sounds so that you can analyze the patterns and find the difference among the three types of VOT (negative, zero, and positive). Measure the VOT of each stop and compare voiced/voiceless counterparts (p/b, t/d, k/g). Similarly, zoom in so that you can clearly see the stop closure followed by the beginning of the vowel. You can measure the time between the end of the stop closure (the beginning of the release burst) and the onset of voicing in the following vowel (the onset of regular pitch pulses in the waveform). This measurement is Voice Onset Time or VOT.
π Definition β Voice Onset Time (VOT): The time interval between the release of a stop closure (the burst) and the onset of vocal fold vibration (voicing) for the following vowel. π Formula: VOT = Time(onset of voicing) β Time(release burst). Measured in milliseconds (ms). π Example: For a voiceless unaspirated stop like /p/ in /apa/, the VOT is near zero (voicing starts almost immediately after the burst). For a voiced stop like /b/ in /aba/, VOT is negative (voicing begins during the closure, before the burst). For an aspirated stop like /ph/ in /apha/, VOT is positive and long (there is a delay of 50β100 ms between the burst and voicing onset).
β Guided Questions:
- How does VOT differ in voiced and voiceless stops and what articulatory explanation can you come up with for this? Answer: Voiced stops have negative or short positive VOT; voiceless stops have longer positive VOT. Articulation: In voiced stops, vocal folds are already adducted and vibrating; in voiceless stops, they are abducted, needing time to adduct and begin vibrating.
- How does the duration of the preceding vowel differ depending on the voicing of the following consonant? Answer: Vowels are longer before voiced stops than before voiceless stops (a phenomenon called pre-fortis clipping).
π‘ Why this matters: VOT is a critical acoustic cue that helps listeners distinguish between phonemic categories like /b/ vs. /p/ in English, and its measurement is central to acoustic phonetics research.
β Key Takeaways
The most critical concepts from this lecture for the exam are the three acoustic correlates of stop voicing: the voice bar (a low-frequency energy band during closure for voiced stops), Voice Onset Time (VOT) (the time between stop release and voicing onset, which can be negative, zero, or positive), and the duration of the preceding vowel (longer before voiced stops). Students must be able to measure VOT in PRAAT by zooming in on the waveform to identify the release burst and the onset of pitch pulses, and then interpret these values to classify stops as voiced, voiceless, or aspirated. The lecture emphasizes practical data collection and analysis using recordings of minimal pairs like /apa/ vs. /aba/.
π§ Quick Revision Questions
- What are the three acoustic correlates of voicing in stops mentioned in this lecture?
- How do you identify a voice bar on a spectrogram, and what does its presence indicate?
- Define Voice Onset Time (VOT) and state the formula for measuring it from a waveform.
- Compare the typical VOT patterns for a voiced stop /b/, a voiceless stop /p/, and an aspirated stop /ph/.
- How does the duration of a preceding vowel change depending on whether the following stop consonant is voiced or voiceless?