In any event,lexical entries can be matched withthe phonology blackboard only through their phonological structure.
It doesn't matter what contextual information is present on the conceptual blackboard, because this material is not
linked to the phonology blackboard. The processor, trying to understand the incoming sentence, can get around to
integrating withthe context only when it comes to conceptual integration. But we're notthere yet—we're stilltryingto
find a phonological match.^101
Suppose a match is found. Then the phonological structure of the lexical item comes to be bound to the material
alreadyontheblackboard. This bindingisthestepoftheprocessorsaying“thisisthewordI' mhearing.”Theliterature
on lexical access traditionally calls this part of the process“selection”of a long-ter m me mory lexical ite mas a
candidate in working memory.
In practice, the lexicon will often contain many different items or combinations of items that match the phonological
structure in working memory (e.g.a parentandapparent). The processor, which for the moment must take a purely
phonologicalpointofview, cannotdistinguishamongthese—thisawaitsintegrationontheotherblackboards. Henceit
stands to reason that we should obtain the well-known results dating back to Swinney (1979) and Tanenhaus et al.
(1979) showing that lexical access is semantically promiscuous—it activates every lexical ite mthat has the right
phonology, regardless of meaning. In the presentapproach, we can interpret thisas multiple items binding to thesame
phonologicalstructure in workingmemory, in competition with each other. (Cutler and Clifton1999) offer a surveyof
models of word recognition that incorporate concurrent competing analyses.)
When a lexical ite mestablishes a phonological connection to working me mory, this also activates its syntactic and
conceptual structures and connects the mto the appropriate depart ments of working me mory as well—in competition
with other candidates, of course. In practice, the activation of syntax and semantics is not instantaneous, given that
activation has to spread through the lexical structure. The syntactic and conceptual material, unlike the phonological
material, is not yet integrated: this is the function of the syntactic and conceptual integrative processors. And, as
mentioned earlier, conceptual
208 ARCHITECTURAL FOUNDATIONS
(^101) Actually, inthegeneralcase a matchwiththephoneticinput isnotwhathas tobesought.A pointrarelyaddressedinstudiesof lexicalaccess inperceptionis thenecessityto
negotiate between phonetic input and the morphophonological form in which the item is stored. I am thinking, for instance, of cases in which rapid speech degrades the
phonetics, of cases where regular sound change creates predictable phonetic alternants, and—for an extreme situation—cases where a morphological affix is realized as
reduplicationor wherethelexicalphonologicalfor mis brokenup byan infix, as informs suchasMononga-fuckin-hela(seesection6.2.1). Here the“calltothelexicon”must
be mediated by the phonology integrative processor, which has to construct active hypotheses about what morphophonological forms to attempt to call.