I’m looking into how to select a phrase with a specific head (a PP headed by “by” for example) within the Grammar Matrix. The Matrix has a singular feature FORM which seems to serve as a combination of the MINORS and VFORM features within the ERG. I was wondering about the reason behind the split in the ERG?

Hi @Dan I’m wondering if you have any wisdom for us here!


The MINORS features are used in the ERG to enable selection of complements or specifiers where the constraints pick out a lexical entry or a small set of entries, including the most typical use of a verb selecting for a particle or a PP headed by a specific preposition. Other uses in the ERG include (1) specifiers selected by adjectives, where positive (ordinary) adjectives require a “very”-type specifier, while comparative adjectives in contrast select a “much”-type specifier (“very tall” vs. “much taller”); (2) specifiers selected by some classes of nouns, such as adjectivally-derived plural nouns like “rich” or “old” that require a definite article (“the rich will thrive” but not “*some rich will thrive”, and similarly for “the oldest were honored” but not “*some oldest”); (3) more fine-grained uses such as day-denoting nouns selecting for part-of-day nouns to admit “Tuesday morning” or “tomorrow afternoon” treated as head+complement structures.
In contrast, the Matrix (and indeed standard HPSG) use of FORM for selecting among the elements of the inflectional paradigm for a particular verb lexeme seems different in kind from the kind of selection illustrated above. Here, selection by an auxiliary verb for a verbal complement with a particular verb form (“is singing” vs. “has sung”) draws a morphological distinction, while the uses above contrast (often small) sets of distinct lexemes with other such sets.
It might have been technically possible to use the same feature for both purposes in the ERG, as is done in the Matrix, but the type hierarchy that would be needed to constrain the possible values of that feature would not be easy to defend or maintain. Further, the MINORS.MIN feature values are often closely correlated with semantic groupings of lexemes, though perhaps grammaticized (or fixed), as with “rely on/upon”, or the dependencies with degree specifiers of color adjectives as in “pale/light blue”, “dark/bright red”). Treating the verbal FORM feature separately has allowed a more adventurous exploration of uses of the MINORS.MIN feature for a wide variety of selectional dependencies in the ERG, enabling usefully fine-grained co-occurrence constraints.


Thank you, Dan! I think it’s probably fair to say that the Matrix so far only uses FORM for roughly what you use VFORM for, and we haven’t yet done much in the MINORS.MIN space. This is why we’re interested, as we move into that space!

Thank you!