English Words

11.3.2
Two-stage models of lexical access in speech production
It has been proposed by a number of scholars that access to words stored in the mental lexicon takes place in
two distinct temporal stages. At stage one, the speaker looks up words in the SEMANTIC LEXICON where
meanings are stored. The second stage involves looking up the PHONOLOGICAL LEXICON for
pronunciation of the words accessed (Fromkin (1971, 1973); Butterworth (1983, 1989)).
As mentioned earlier, words are not stored alphabetically in the mental lexicon. Rather, it would seem,
they are stored in a manner more akin to that of a thesaurus. Groups of words with related meanings cluster
together. The relationship may be, for example, close semantic similarity (near synonymy) or, antonymy
(oppositeness) or some other kind of semantic relatedness unaccompanied by phonological similarity (cf.
[11.21], [11.22] and [11.23] above).
By contrast, in the second type of error a word with a similar pronunciation and belonging to the same
word-class but having an unrelated meaning is substituted for the right word. Many things can go wrong
with the phonological retrieval process without the semantic dimension of words being affected. Fromkin
(1971) cities many examples of phonological muddles caused by phonetic similarity, e.g.:
[11.24]
a.
Utterance:
bottle of page five
Target:
bottom of page five
b.
Utterance:
while the present—pressure indicates
Target:
while the pressure indicates.
c.
Utterance:
it spread like wild flower
Target:
it spread like wild fire
Fromkin recognises a number of different phonological errors:
[11.25]
a.
Phonological
anticipations
Utterance:
bake my bike
Target:
take my bike
b.
Phonological
preservations
Utterance:
pulled a pantrum
Target:
pulled a tantrum
In [11.25a] we see the replacement of the initial consonant of the first content word with that of the
second. But in [11.25b] the opposite happens. The initial consonant of the second word is replaced with that
of the first.
It is not always entire sounds that are substituted. Sometimes it is the DISTINCTIVE FEATURES of
which phonemes are made up that are exchanged or confused:
THE MENTAL LEXICON 171

---

[11.26]
a.
Utterance:
Cedars of Lemadon
Target:
Cedars of Lebanon
b.
Utterance:
flesh queer water
Taget:
fresh clear water
c.
Utterance:
blake fluid
Target:
brake fluid
(cf. Fromkin 1971)
What is involved in [11.26a] is the changing of oral [b] into nasal [m] and nasal [n] into oral [d]. The
phonetic feature [± nasal] is present in the word, but appears as part of the wrong segment in each case. In
[11.26b] and [11.26c] it is the feature [± lateral] that is swapped round and appears in the wrong word.
The phonological behaviour displayed in SPOONERISMS lends further support to the two-stage model of
lexical retrieval. In a spoonerism, a content word is moved from its intended position and turns up
elsewhere. The result is often an (un)intentional comic effect:
[11.27]
a. Utterance:
I told them to open their desks and leave them on the
books
.
Target:
I told them to open their books and leave them on the
desks
.
b. Utterance:
queer
old dean
Target:
dear
old queen
c. Utterance:
You hissed my mystery lectures.
Target:
you missed my history lectures
(The last two slips are attributed to the Revd William A. Spooner (1844–1930), Warden of New College,
Oxford, and the first to a school teacher friend of mine.)
In spoonerisms content words swap places without affecting the pattern of sentence stress as in [11.27a]
and [11.27b], Or, words are substituted with others that have similar sounds and stress patterns, as in [11.
27c]. This shows that assigning stress to a sentence is done separately from inserting content words.
By contrast, when function words are shunted round by mistake, the word that is moved normally takes with
it its stress:
[11.28]
a.
Utterance:
Can I turn off this.
Target:
Can I turn this off.
b.
Utterance:
Well I much would have preferred the owl.
Target:
Well I would have much preferred the owl.
(from Cutler and Isard (1980), cited in Butterworth (1980))
Clearly, content words differ from function words in the way in which they are processed in the mind.
Speech processing errors also give us a window on the storage and retrieval of grammatical elements.
Some errors show that grammatical function morphemes are stored in a very abstract manner. This can be
172 ENGLISH WORDS

---

clearly seen in errors involving NEG TRANSPORTATION. A disembodied notion of negation is stored
separately from the specific morphemes that represent it (e.g. in-, un-, non-, dis-, any-). Neg transportation
is a process whereby some negative prefix (e.g. in-, un-, non-, dis-, any-) is moved from its normal, intended
position and appears incorrectly in a different position where a different negative prefix may be substituted
for it. Fromkin (1973) gives these examples:
[11.29]
Utterance:
I disregard this as precise.
Target:
I regard this as imprecise.
Sometimes the negative morpheme is not transported to a new position in the sentence, but a different,
incorrect negative morpheme is used instead of the target one. Thus, in the sentence in [11.30], the clitic -
n’t
attached to was replaces the correct negative prefix, un-, that should appear before plugged:
[11.30]
Utterance:
I was unplugged in…
Target:
I wasn’t plugged in for a second.
Negative transportation lends further support to the assumption that content words are stored separately
from affixes. That is why you may succeed in retrieving the correct content word (plug) and produce an
incorrect sentence in the end if the wires get crossed so that the affix with the meaning you want (neg)
appears on a different site in the sentence (after was) from the one intended (as a prefix (un-) attached to
plugged
).
The claim that grammatical, functional morphemes are stored and retrieved separately from content
morphemes is further buttressed by additional instances of the uncoupling of affixes from stems and bases.
An interesting morphological error results in the root morphemes in different words exchanging places
leaving behind affixes that should have been attached to them. This phenomenon is called AFFIX
STRANDING. It can be seen in the following examples from Garrett (1980) (where the stranded suffixes
appear in upper case letters):
[11.31]
a.
Utterance:
You have to square it faceLY.
b.
Target:
You have to face it squarely.
[11.32]
a.
Utterance:
I’ve got a load of cookEN chickED.
b.
Target:
I’ve got a load of cooked chicken.
Garrett claims that affix stranding shows that affixes are not stored with stems as integrated wholes, but
that they are stored in separate places. The insertion of stems in syntactic ‘positional frames’ by syntactic
rules is distinct from morphological operations that attach affixes to stems. When all goes well these two
processes are fully synchronised. But a slight hitch may cause a mismatch between stems and affixes.
THE MENTAL LEXICON 173

---

But, as Butterworth (1989) points out, there are problems with Garrett’s account. In irregular morphology
there are examples of a word being placed in the wrong position with its correct irregular inflected form, as
in the following:
[11.33]
a.
Utterance:
I don’t know that I’d hear one if I knew it.
b.
Target:
I don’t know that I’d know one if I heard it.
The expected form according to Garrett’s theory is:
[11.34]
Utterance:
I don’t know that I’d hear one if I knowed it.
In this case the stranding predicted by Garrett does not materialise. Instead whole-word movement takes
place. Butterworth (1989) argues that the evidence for a separation of roots from affixes in the manner
proposed by Garrett is not so strong. In [11.34] the inflectional affix is integrated with the stem and moves
with it.
Arguably, Garrett’s position can be defended. If we assume a lexical morphology model with a
distinction between essentially irregular stratum 1 affixes and the regular stratum 2 affixes, we can hazard
the prediction that stranding only applies at stratum 2. Irregular inflection in verbs like knew, sang, rode and
nouns like feet, and mice, is integrated so early and so firmly into the stem that there is no chance of the
affixes being left stranded on the beach when the stem is moved elsewhere.