when there are multiple potential contingent responses from different speakers, should we prefer the speaker who most recently was logged as contingently prompting/responding to the target child? if so, does that constraint outweight stretch/strict considerations?
when there are multiple potential contingent responses from different speakers, should we prefer the speaker who most recently was logged as contingently prompting/responding to the target child? if so, does that constraint outweight stretch/strict considerations?