 Vowels have a fully developed formant pattern and can be classified on the basis of their first two formants F1 and F2. These resonance frequencies can approximately be associated with the size of two cavities separated by a constriction in the vocal tract. F1 with the size of the pharynx and F2 with the size of the aural cavity. Let's compare two more vowels to illustrate that. Cardinal vowel 1, E, with a front constriction, a large pharyngeal cavity and a small aural cavity and cardinal vowel number 5, or with a back constriction and thus a much smaller pharyngeal cavity and a relatively large aural cavity. And now look at their spectrograms. Cardinal vowel number 1 is characterized by a low F1 value at around 400 Hz and a very high F2 value at around 2,500 Hz whereas cardinal vowel number 5 has a slightly higher F1 value at around 600 Hz and a much lower F2 value of around a thousand Hz. Here are the spectrograms of all eight cardinal vowels. In plotting the frequencies of F1, the vertical axis, against the frequency of F2 on the horizontal axis, we find a pattern that is remarkably similar to the cardinal vowel chart. Yet the match is not exact because the articulatory cardinal vowel chart is based on the point of the greatest tongue constriction whereas the acoustic chart takes its data from all vocal tract resonances. Let us now identify the vowels within more complex spectrograms. In our first example, we can clearly identify at least two vowels. The first with formant 1 and formant 2 very close together, thus it must be a low vowel such as R. The second vowel involves a transition. The first part is clearly a low vowel again with F1 and F2 close together and in the second part F1 and F2 are drifting wide apart. So the result is a diphthong, in this case I. And here is a solution, I said, sunshine. Note that the nasal consonants also involve some sort of form patterns, yet they are less well developed. Here is another example. Again it contains two vowels. The first one looks like a front vowel with an F2 value of around 1600 hertz, hence air, ur or ur are likely candidates. The second vowel must be a diphthong with a high vowel offset such as E and an onset similar to our first vowel, so A is a likely candidate. And here is a result. I said earthquake, where our two vowels are surrounded by fricatives and plosives. So the identification of vowels within complex spectrograms involves the identification of a clearly defined formant structure and the extraction of the frequency values of F1 and F2.