Glossary of silent letters. Part one.
The answer to this question is so long that it needs to be divided into two parts. The following is part one:
How you deal with the question of silent letters depends pretty much on how you analyze sound-to-spelling correspondences. In American English Spelling (AES) I chose to minimize the number of letters that I called silent, simply because I wanted to avoid having a unit of silence floating around in the system. Thus, instead I analyzed to complex spellings that had evolved over the centuries due to simplification and sound change. For instance, rather than call the final <b>’s in words like bomb silent, I analyzed to the complex sound-to-spelling correspondence [m] = <mb>, recognizing the simplification of the earlier [mb] pronunciation. I pretty much ignored the question of the <d>’s in words like grandfather and handkerchief, which I dismissed as “only apparent silent <d>’s” (pp. 337-38) – though I’m not sure now exactly what “apparent” means there.
Since then my thinking has gotten somewhat more sophisticated as various aspects of orthographic reality impressed themselves on me. There is an important distinction, for instance, between diacritic and non-diacritic silent letters. The most common diacritic silent letters are silent final <e>’s used to mark preceding long vowels (mate, baste), soft <c>’s and <g>’s (ounce, change), voiced <th>’s (breathe, loathe), and used to insulate otherwise word-final <s>, <z>, <u>, and <v> (tense, bronze, plaque, salve) (AES, pp. 145-48), and to provide the third letter in words that would otherwise have only two letters (as in eye, awe) which would contradict the Short Word Rule (AES, pp. 87-89).
A second example of a diacritic silent letter is the <u> that sometimes is used between <c> or <g> and <e, i, y> to avoid the appearance of soft <c> and <g>: biscuit, guest, guilty, plague.
There are also non-diacritical silent final <e>’s as in French adoptions like cigarette, where the <e> no longer acts as the feminine marker that it does in French. A number of similar non-diacritical silent final <e>’s are discussed in AES, pp. 148-54. Non-diacritical medial silent <e>’s occur in past tense forms like licked and boomed, where the <e> is silent. There are also the non-diacritical silent <e>’s associated with syllabic <l> and <n>, as in battle and forgotten.
Related to these silent <e>’s are other letters that fall silent due to simplification of sound clusters: the silent <h> in exhaust, the silent <d> in grandfather, handkerchief, the silent first <o> in laboratory, the silent <th> in isthmus, the silent <n> in monsieur, the silent <i> in business. It seems to me that in these cases, we really have to recognize silent letters that are serving no diacritical purpose.
End of part one.