Suppose one has a strong divine command metaethics (SDCM) that conceptually analyzes "x is obligated to A" as "God commands x to A". Then one faces the problem of distinguishing commands from other speech acts. It seems very plausible that a part of the story about what commands are is going to involve an intention to generate an obligation or an intention to engage in a speech act of a sort defined by a certain kind of generation of obligation. In any case, it seems very plausible that the concept of obligation is going to figure in the story of what a command is. And hence it is circular to conceptually analyze obligation in terms of commands.
The above summarizes the argument, but we should make two friendly amendments to SDCM. The first is that SDCM can be more modestly taken as only conceptually analyzing moral obligation.
The second is that commands always need to be understood in terms of a role or social relationship. Unless all obligations are moral obligations (which I am inclined to think, but which almost nobody else thinks), it should be possible for God to issue commands in a role that does not generate moral obligation. For instance, imagine Jesus and other kids playing Simon Says. Jesus says: "Simon says, run a mile." A kid who doesn't do it and is willing to accept, fair and square, a loss in the game is not automatically sinning through disobedience to God. For Jesus did not command as creator and master of the universe, but only as a Simon in Simon Says. And the same could happen without an Incarnation. There is nothing to bar God engaging in some game with humans. (Maybe one can argue that even in Simon Says, one has a prima facie moral obligation to obey the Simon. That would be controversial, but would force a modification to some of my arguments.) Of course, normally when God issues something that sounds like a command, we reasonably assume that it is a divine command, just as normally when one's superior officer issues something that sounds like a command, we reasonably assume that it is a military command. But in both cases, these presumptions can be defeated by context or explicit qualification.
So it is not a necessary truth that x is morally obligated to A if and only if God commands x to A. Let C be that role which creatures have in regard to their creator that paradigmatic universal divine commands like "Thou shalt not kill" are issued in respect of. For instance, C might be the role of owing gratitude to God for everything (cf. Evans, but Evans does not accept SDCM as he isn't analyzing moral obligation) or of being created by a loving God (cf. Adams, but Evans tells me that Adams also isn't analyzing moral obligation), or C might be the role of being in the image of God, or the like. And then say that a C-command to x is a command issued by God to x in virtue of x's filling C.
So our SDCM now says that "x is morally obligated to A" is to be analyzed as "God C-commands x to A."
But now, what is a C-command? It is very plausible that a defining part of being a C-command is an intention to generate a C-obligation, or at least being the sort of speech act that is intended to generate a C-obligation. In other words, the notion of a C-command depends on that of a C-obligation.
So now we have a puzzle. What is the relationship between a C
-obligation and a moral obligation? On the SDCM, every moral obligation necessarily is a C
-obligation, since presumably each of God's (valid) C
-commands necessarily succeeds in generating the C
-obligation that it is intended to generate.
What about the converse? Could there be a C-obligation that isn't a moral obligation? That would be very odd for at least two reasons. First, the features of C that make it plausible that C-commands necessarily generate moral obligations make it just as plausible that all C-obligations necessarily are moral obligations. Second, intuitively if C-obligations aren't all necessarily moral obligations, God should be able to issue a C-command that generates something that merely C-obligates and doesn't morally obligate.
Thus, given SDCM, C-obligations and moral obligations are necessarily coextensive.
But now why analyze moral obligations in terms of C-commands instead of in terms of C-obligations? After all, there are very few, if any, other roles where all role obligations are generated only by commands. I can't think of any actual ones, though one might be able to gerrymander one. But if necessarily moral obligations are all generated by C-commands, and necessarily moral obligation and C-obligation are coextensive, then necessarily all C-obligations are generated by C-commands, which just does not seem very plausible. (We might even give an example of a plausible C-obligation that don't seem to be generated by a divine command: say, the C-obligation to find out what C-commands have been issued--it would be odd if God had to command that to us before it became C-obligatory.) And, second, even if C happens to be a rare role such that necessarily all C-obligations are generated by C-commands, it is more plausible to analyze moral obligations in terms of C-obligations than in terms of C-commands. For, plausibly, it is only because C-commands generate C-obligations that C-commands generate moral obligations. (After all if per impossibile there were a C-obligation that didn't result from a C-command, it would surely still be a moral obligation. But I don't lay much stress on this counterpossible.)
So, if we've gone down this road, we should analyze moral obligations in terms of C-obligations. But since the concept of a C-command conceptually depends on that of a C-obligation, we will hit the circularity problem if we further try to analyze C-obligations in terms of C-commands.
If this line of thought is correct, then those who want to analyze moral obligation in terms of divine commands should instead switch to a view on which moral obligation is to be analyzed in terms of our creaturely role obligation to God. One can then supplement that--though I think it would be implausible--with a claim that necessarily all creaturely role obligations are generated by divine commands. But that won't be an analysis of moral obligation.
The same form of argument works against strong legal positivism, the doctrine that "x is legally obligated to A" is to be analyzed as "legislators legally-command x to A". For the notion of a legal-command depends on that of legal obligation.
Objection: Perhaps one can analyze C-commands and legal-commands in terms of commitments to punish. Thus, God C-commands A (or the state legally-commands A) if and only if God (or the state) commits himself to inflict harsh treatment on those who do not A.
Response: This is mistaken. When we drive to San Antonio on the tollway around Austin, we automatically get charged--via cameras trained on our license plate--about $9 in toll. This is a very modest piece of "harsh treatment" (still, it is a real piece of it--a court might assign a minimal fine of a $1 for something, and that's still a punishment). Imagine it was more expensive--say, $100. Then we might say it is a harsh treatment. But it would not follow that this harsh treatment would constitute a command to avoid that road! That would defeat the purpose of the tollway.
Acknowledgment: I am grateful to Steve Evans for making clear the limitedness of the target of my argument.