Prose

On the Monotonicity of Polar Adjectives

Negative polarity item licensing (1-4) and entailment patterns (6-8) indicate that certain adjectives generate monotone decreasing contexts, while others generate monotone increasing contexts (cf. Seuren 1978, Ladusaw 1979, Linebarger 1980, Sanchez-Valencia 1996?).

  1. It is difficult/*easy for him to admit that he has ever been wrong.
  2. It would be foolish/*clever of her to even bother to lift a finger to help.
  3. It is strange/*typical that any of those papers were accepted.
  4. It's lame/*cool that you even have to talk to any of these people at all.
  5. It's dangerous to drive in Rome. ==> <=/= It's dangerous to drive fast in Rome.
  6. It's safe to drive in Des Moines. =/=> <== It's safe to drive fast in Des Moines.
  7. It's weird to see people banging on drums in Sacramento. ==> <=/= It's weird to see people banging on homemade drums in Sacramento.
  8. It's common to see people banging on drums in Santa Cruz. =/=> <== It's common to see people banging on homemade drums in Santa Cruz.


Monotonicity properties represent one of several factors which have traditionally been used to classify gradable adjectives according to their "logical polarity" (in the sense of H. Klein 1996). Adjectives which license negative polarity items and downward entailments in clausal complements, such as difficult and strange, are classified as "negative", while adjectives which do not license negative polarity items but do permit upward inferences, such as easy and common, are classified as "positive" (see Seuren 1978 for early discussion of this issue). This paper addresses the question of how this distinction should be captured in the lexical semantics of gradable adjectives. I demonstrate that the monotonicity properties of polar adjectives can be derived if logical polarity is represented as an independently motivated sortal distinction between positive and negative adjectives in a model in which gradable adjectives denote relations between individuals and extents (intervals of a scale), rather than relations between individuals and degrees (points on a scale), as traditionally assumed (see e.g., Cresswell 1976, E. Klein 1991).