As you say, it’s related to instantiation, specifically the probability of instantiation... less likely choices are more marked. That gets you into knots with tone 1 imperatives here, which you say are ‘unmarked’ but are no more frequent than other tones in your data. A problem may be ambiguity of the term ‘neutral’. It can mean unmarked, but in Halliday’s key system it means neutral force, in contrast to mild, strong etc.
PS Key treats grammar/tone couplings as delicate mood choices, but another perspective is delicate speech function and appraisal choices... ‘what a command is trying to get done ‘
In contrast, marked/unmarked are used technically as feature labels in textual systems, theme and information focus. Could we perhaps avoid using these terms quasi-technically in pedagogic contexts, and reserve them for textual features? The ‘good reason’ is textual prominence.
Blogger Comments:
[1] To be clear, there is no problem here because markedness is not defined by frequency. Halliday (2005 [1991]: 68):
An unmarked term is a default condition: that which is selected unless there is good reason for selecting some other term. It is not defined by frequency, but it is likely to correspond to the more probable term in a system whose probabilities are skew.
Halliday & Matthiessen (2014: 169) identify the different speech functions realised by imperative mood realised by different tones as follows:
command: tone 1 (unmarked in positive)
invitation: tone 3 (unmarked in negative)
request (marked polarity): tone 13, with tonic on do/don’t
plea: tone 4
The frequency of a selected tone will thus vary with the frequency of a selected speech function realised by imperative mood in a given text type, and this varies with the tenor relations of the interlocutors.
[2] This misunderstands the distinction between a feature and a classification of feature. In Halliday's system of KEY, Halliday & Greaves (2008: 208) 'neutral' is one of the features, not a classification of a feature as neutral or unmarked:
[3] This misrepresents the SFL model. To be clear, SPEECH FUNCTION (semantics) is realised by MOOD (lexicogrammar), and different features of KEY (lexicogrammar), realised by different features of TONE (phonology), differentiate the SPEECH FUNCTION of a MOOD into more delicate types. An example is the case of commands being differentiated into invitations, requests and pleas according to the tone selection, as in [1] above.
[4] To be clear, the unmarked/marked distinction is not restricted to features of textual systems, as demonstrated by the unmarked/marked distinction for the system of TONE, which realises interpersonal systems. Another interpersonal example is the unmarked choice of Subject; Halliday & Matthiessen (2014: 194):
For any clause, there is one choice of Subject that is ‘unmarked’ – that is assumed, in the absence of evidence to the contrary. In a giving clause (offer or statement), the unmarked Subject is ‘I’; while in a demanding clause (question or command), the unmarked Subject is ‘you’.
An experiential example is unmarked/marked distinction for present tense, which varies according to PROCESS TYPE.
[5] This misunderstands the 'good reason' principle. To be clear, 'textual prominence' is what choosing an element as Theme or New affords, and this is the case whether the choice is unmarked or marked. The 'good reason' principle, on the other hand, is what lies behind the choice of a marked option in preference to the unmarked option.
