On Syntactic Features

In discussions of the morphophonological mapping from syntax to phonology, we aim to establish generalizations based on the features of morphemes in the syntax.  But the featural content of morphemes in the syntax is often undertheorized by syntacticians, who are often motivated by a desire to keep the syntax somehow transparent to semantic interpretation.  What sorts of features should morphemes carry in the syntax?

Syntactically relevant features?  What does the syntax actually accomplish with the features?

Features related to agreement and case marking:  “Unvalued” or “valued” person, number, and gender features.  Unvalued or valued case features?

Features related to syntactic selection?  Here we can separate proposals into two categories.  First, we have EPP or Merge features, i.e., features that ask for Merger, often requiring Merger of a phrase with a categorical (+DP) or “criterial” (+wh) feature.  Second, we have “selectional” features, whose operation is up for theoretical grabs.  If one head “selects” for another head, when in the course of syntactic operations is this selectional feature “checked” (e.g., at Merge or perhaps at the interface with PF).  Some syntacticians either propose or act as if they are assuming that c-selection or syntactic feature selection could produce the order of f-sequences, e.g., in the extended projection of the verb.

The case and agreement features, plus the Merge/EPP features, would constitute a set of clearly syntactic-relevant features, features that, to some extent, seem to drive the syntax.  Selectional features are not clearly syntactic, i.e., do not necessarily implicate Merge and/or the valuing of syntactic features.

The mapping to phonological structure is often thought to depend on OTHER syntactic features, features that may seem to be important also to the semantics but that don’t obviously determine syntactic structure or behavior.

Features like Tense features:  +/-Past, Perf, Prog.  Or Participle features.  Or adverbial type features (repetitive, pluractional, negative).

These “other” features figure in arguments for the “Morphome,” as in Aronoff’s observations about the Latin -t stems.  The -t stem is used in verb forms and in derived forms in cases where there is no semantic trace of “pastness” or “perfectness.”  That’s enough for Aronoff to argue that there is no feature that words containing the -t stem share.  However, the -t stem is, syntactically, an adjectival participle.  The morpheme or morphemes forming it from the verb stem in Latin must include the unvalued gender features associated with adjectives.  In addition, the morpheme that closes off the -t stem as a participle must close off the extended projection of the verb:  no further verbal inflectional heads may be added.  In terms of the larger morphological system of Latin, there’s no problem having the perfect participle stem used for the various word formation processes Aronoff gives for the Latin -t stem.  The syntactic features necessary for this stem don’t require “past” or “perfect” features, just as the English perfect participle can be built with a nominal head without needing semantic features, which do no work in the syntax 

But these observations about the -t stem don’t explain the syntax of the -t stem, i.e., why Latin can put these morphemes together in the way it does, and how a morpheme closes off an extended projection.

Additionally, we need to ask which heads carry what information relevant for the formation of complex heads in a head projection and for the ordering of heads with phrases.  Suppose we have Merged an extended projection of v.  Some section of heads in this projection will form a complex head for the morphophonology and be ordered with respect to phrases while others might form independent phonological words.  Is there some feature shared by say Tense heads in English with Participle Heads and Plural heads that tell the morphophonology that these heads complete a morphophonological word such that further inflection is not possible?