I like The Sounds of the World’s Languages a lot. If you’re used to think in phonological terms, it greatly expands your simple model with real-world phonetic quirks. (For example, it turns out vowel “height” and “backness” are not actually very correlated with the position of the highest point of the tongue, though the two properties are real enough in the acoustic level).
One fun way to browse a book like this is to look for “your” languages in the index and peek ahead. I learned quite a few neat things about Chinese and Icelandic this way, but since the focus of this blog is Japanese, have some Japanese-related tidbits to whet your appetite:
As we noted earlier, bilabial stops differ from stops at the majority of other places of articulation in that they involve an active movement of the articulator on the upper surface of the vocal tract. The upper lip moves down to meet the upward-moving lower lip to form the closure and moves back up at the release. [Figure ommited.] Partly because both articulators are in motion, and therefore the increase in the cross-sectional area of the oral escape channel occurs more rapidly, the release phase of a bilabial stop is typically shorter than that of stops at other places of articulation. On many occasions […] aerodynamic forces may also cause the lips to separate more rapidly than they close (Fujimura 1961). The fact that both articulators are soft tissue may explain why bilabial stops are more likely to be produced with incomplete closure than stops in which the active articulator contacts a less flexible surface. This results in patterns such as that in Japanese, where in native and Sino-Japanese vocabulary /p/ is pronounced word-initially as [ɸ] or [h] (depending on the following vowel), and as [p] in a cluster with another consonant, whereas [t], [k] do not alternate with fricatives (McCawley 1968, Shibatani 1990:166–7).
Geminate stops in many languages are limited to word-medial position where they usually close the preceding syllable, shortening its vowel to some degree, as well as serving as the onset of the following syllable (Maddieson 1985). Japanese exemplifies the rare case where a preceding syllable is essentially unaffected by a long following stop (Homma 1981, Smith 1992). The acoustic record does not reveal whether a long stop is produced with two separate articulatory gestures, the first corresponding to the syllable-closing part and the second corresponding to the syllable-opening part, as was proposed long ago by Sievers (1876). A number of studies have looked a this issue using either electromyography or methods of tracking articulatory movements over time. An electromyographic study by Lehiste, Morton and Tatham (1973) showed that two peaks of activity of the orbicularis oris muscle can occur both for word-medial geminate [pp] in Estonian and for [p#p] across a word boundary in English for one speaker of each of these languages. On the other hand Barry’s (1985) dynamic palatographic data on English [k#k] sequences did not show any evidence of two articulatory peaks and Smith’s (1992, 1995) x-ray microbeam study of word-medial geminates in Italian and Japanese did not show double peaks in the articulatory movement of the lips for geminate bilabial stops in either of these languages, nor was there evidence of any double peaks in the tongue blade movements for alveolar geminate stops in this study. […] It thus seems evident that geminates can be produced with a repeated articulatory movement under some circumstances, but that this is unlikely to be the most common articulatory pattern. Moreover, the presence or absence of a second articulatory peak cannot be taken as diagnostic of whether a long closure represents a geminate stop or a sequence of two identical stops.
[Following a discussion on the relationship between rhotics and laterals, and the lateral flap as belonging to both classes.] Lateral flaps are auditorily reminiscent of both [ɾ] and [l]. Some of the reports of alternations between [ɾ] and [l] in a variety of languages may be attributable to different perceptions of what is in fact a consistent articulation […]. In general, back vowels seem to predispose toward the production (or perception) of lateral variants, and front vowels toward rhotic variants. […]
In Japanese, Shimizu and Dantsuji (1987:16) note that
Some Japanese use both a lateral approximant [l] and a flap [ɾ] as completely free variants. Some Japanese use a lateral approximant [l] in the word initial position and use a flap [ɾ] in the intervocalic position. Some use a lateral approximant [l] in each position. Other use a retroflex voiced stop [ɖ] in addition to these sounds.
Thus there are patterns of alternation between rhotics and laterals that associate these two classes together. As noted earlier, there are also distributional similarities between rhotics and laterals. The most typical members of these classes are relatively sonorous, but both classes include sounds that are far from being so.
The great majority of the world’s languages have a predictable relationship between the phonetic Backness and Rounding dimensions. Front vowels are usually unrounded and back vowels are usually rounded. However, as shown above for Bavarian German, front vowels with a rounded lip position also occur. In addition, back vowels without lip rounding can be found, sometimes simply because a language has relaxed the linkage between Backness and Rounding (as for the high back vowel of Japanese) […]
[After a lenghty discussion of the “weird” high-front vowel [ʉ] in Swedish, which is proposed to have “vertical lip compression”.] All these observations, together with our own investigations of these languages, lead us to conclude that there are two lip position parameters for vowels, vertical lip compression and protusion. In most languages these parameters are implemented jointly (and, also, linked to the front-back dimension), and it is sufficient to distinguish rounded (either compressed or protuded) vowels from unrounded vowels. In a small number of languages the two parameters are independently controlled. Some languages may choose to use lip compression rather than the form of rounding that has lip protusion. Edwin Pulleyblank (personal communication) notes that Japanese /u/ can be regarded as having compressed lips rather than being simply unrounded. This vowel shows its labiality by the fact that it alternates with /w/ in verbal inflections. Pulleyblank also notes that the Japanese allophone of /h/ that occurs before /u/ is bilabial [ɸ], with what we here call compressed, rather than protuded, lips. We have not investigated the acoustic characteristics of lip compression. They are presumably similar to those of lip rouding and protusion insofar as any decrease in lip aperture tends to lower all formant frequencies, but compression and protusion differ with respect to their distinct effects on the length of the vocal tract.
I’ve seen speakers alternating [l] and [ɾ] freely but never one that consistently uses [l] word-initially as in Korean. I wonder if it could be a dialect thing. Also, some people say they notice [l] being used for a “cute” effect, for example in pop-idol songs.