Heave-heaven

Overview

There is a debate in the reading literature about whether and the extent to which orthographic wordforms are decomposed based on their morphological composition. An influential paper, Rastle, Davis, & New (2004) from Psychonomic Bulletin and Review, claimed that the process of morphological decomposition happens early and automatically in orthographic processing based solely on the appearance of compositional structure in orthography (visual domain). In their lexical decision task they took sets of prime-target pairs that fell into one of three categories:

  • Morphologically related pairs, like: ACID (prime) -> acidic (target)
    • These were called “transparent” pairs, alluding to the fact that the morphological relationship is obvious and true.
  • Obvious, morphologically unrelated pairs, like: HEAVE ->heaven
    • They called these “form” pairs, emphasizing that there is some aspect of orthographic form shared across the items, but nothing else.
  • Pairs that had the appearance of a morphological relationship, but no actual semantic relationship, like: ARCH -> archer
    • They called this condition the “opaque” condition to emphasize that there is some appearance of a relationship but such a relationship is not immediately clear given the fact that the pairs are not actually semantically related (more on that later though)

They found a statistically reliable difference between the primed responses in the transparent and opaque conditions as compared to the form condition, which they took to be evidence of early automatic morphological decomposition of the surface morphology of words found in those conditions relative to the other. The problem with this interpretation is (a) they didn’t consider the phonological aspects of the words included as stimuli, especially the differences in phonological overlap of prime-target pairs, (2) they didn’t account fully for the semantic relationship of prime-target pairs across conditions, and (3) effect estimates are based on limited statistical methods that don’t adequately account for these random trends by item. Our work here is aimed at getting to the bottom of this phenomenon by more fully accounting for these phonological and semantic properties of primes and targets.

Status updates

For status updates to workflow on curriculum modeling see the wiki page relevant to the specific project you are interested in. For example, if you are interested in unordered candide, see the wiki page for unordered candide.

Data

  • Overview analyses of effects, including preliminary account of phonological overlap

Research team

  • Mark Seidenberg (PI; seidenberg@wisc.edu)
  • Matt Cooper Borkenhagen (cooperborken@wisc.edu)
  • Madison Barker (madisonbarker852@gmail.com)
  • Past contributors:
    • Cher Yang (RA, grad student at U Washington)
    • Sarah Engel
    • Sarah Heinze

Comments are closed