Salmoneus wrote:It's not a spelling pronunciation.
It's also not quite as simple as "thinking there's a /g/ because there's a [g]". Most people do not, for instance, believe there to be /b/ in "debt", or /s/ in "island".
The problem arises in this particular case because people know that there's not just an /n/ in these words - "sing" and "sin" aren't homophones, for instance. But because they have no words (or concepts) to describe the /N/ sound - and because, to be fair, the /N/ sound does sound a bit /g/-ish - they can only describe /N/ as being /n/ plus /g/, as it's spelled (and sort of sounds). So if you tell them there's no /g/, they get confused, because they know there's not just /n/, and they don't have the conscious concept of /N/ as an alternative for what might be there.
After all, if you "drop the G" from "walking", you get "walkin'", with an /n/... it's got an apostrophe and everything! And surely if 'dropping the G' were just turning something from /N/ to /n/ - a change with no loss of elements or gain in simplicity - then it wouldn't be such a common indicator of lazy speech!?
[similarly, some people think there's an /h/ in /S/ and /T/. But this is less common, probably because we don't have any productive, shibbolethical "dropping the H" process].
And even beyond that, because of the history, there are still ways in which [ŋ] behaves synchronically in the phonology of English as if it were a cluster /ng/: it doesn't occur after long vowels/diphthongs, except for in onomatopoeic or expressive words, it alternates with [ŋg] in the comparative and superlative forms of
long, strong, young, and of course, it doesn't occur word-initially or after other consonants. Facts like these make the status of /ŋ/ as a phoneme of English somewhat disputed even among phonetically knowledgeable analysts (apparently, a number of generative phonologists are in favor of analyzing English [ŋ] as synchronically derived from /ng/ (or /n/, before /k/): see this page from
The Phonology of Swedish, Tomas Riad, for some references).
When learning how to do phonetic and phonemic transcriptions, you don't just have to learn about phones, but also about what the traditional/conventional abstractions are for representing specific languages' sound systems. No phonetic transcription is perfectly narrow, and of course a phonemic transcription will only represent one particular analysis of a language—which it is doubtful that everyone will agree on in all respects.