irregular forms. However, phonological integration has no idea that thewrong verbs are there; itcan't see theintended
meaning. So it goes its merry way producingknew. Another such example, this time at the phonological level, isan
unkey's muncle[intended:a monkey's uncle]. Here, since themhas been removed frommonkey, the word now starts with a
vowel. This triggers the automatic phonetic process of replacingawithan; this process doesn't care whether what it
precedes is a real word or not.
Garrett (1975), through an extensive categorization of speech errors, maps out a sequence of stages in speech
production that is close to what emerges from the present model. At the beginning of the process is lexical selection,
whichresults in thesemantic substitutionsand blend errors mentioned in the previous section. The other main classes
involve different stages of integration.
- Sound exchanges and shifts such asour queer old deananda but-gusting meal[intended:gut-busting].
- Word exchanges and shifts:examplesareany guys you time...[intended:any time you guys...] andthe crazy idea who
had a good idea.... [intended:the crazy guy...]. Here words are moved around complete with their affixes. This
case contrasts with - Morpheme exchanges and shifts: examples areI'd hear it if I knew it; naming a wear tag [wearing a name tag]; I hate
raining on a hitchy day [hitching on a rainy day]. In these examples, stems move around, leaving affixes in place.^107
Garrett shows that each type of error has a characteristic maximal span over which shifts can take place, and
characteristic constraints on what can substitute for what. Subsequent research (e.g. Shattuck-Hufnagel 1979; Levelt
1989; Dell 1986; Bock 1995) has refined the constraints and tendencies found in these different classes of errors. For
presentpurposes, themain pointis thatLashley was basically rightabout errors resultingfrom improper integration of
material lying around in working memory.
Moreover, the stages of processing identified by Garrett correspond nicely to stages of integration in the present
model. Sound exchanges and shifts are clearly problems in phonological integration. The word exchanges/shifts and
morpheme exchanges/shifts are both problems in syntactic integration. Section 5.6 suggested dividing the syntactic
component into phrasal and morphosyntactic tiers. Given this division, the word errors fall naturally under phrasal
integration and the morpheme errors under morphosyntactic integration. This is hardly news. However, the close
correspondence between the competence and
216 ARCHITECTURAL FOUNDATIONS
(^107) Most of these examples are from Shattuck-Hufnagel's (1979) corpus of actually observed speech errors.