If the nasal vowel was represented as monosegmental, one would expect it to be followed most often by the palatal approximant.
The items with an approximant have a longer friction duration.
What happens, in fact, is that the major par t of the approximant itself is devoiced.
Although amplitude is still relatively high, there is a greater reduction than in the case of weak approximant segments.
However, it occurs only when the righthand environment is a high front vowel, an approximant, or a fricative.
Moreover, each local approximant has to be evaluated efficiently.
Replacing by a rational approximant the problem reduces to a periodic case and a similar picture is obtained.
The second difference between their patterns concerns the outcome of coalescence for stop+approximant clusters.
