78 A LEXICON WITH LAYERS

does not necessarily mean that Ms and Mr Average Native-Speaker are aware of it too. What in principle
should be compositional for some speakers may not be for all speakers.
The classic example of this is provided by the cranberry words (cf. Bloomfield 1933). These words
highlight the problem of treating morph-like units that do not belong to any recognisable morpheme in the
language as it is used today. A cranberry is a kind of berry. So, we can identify berry as a morpheme in that
word. But that leaves us the problem of what to do with the apparently meaningless unit cran- that is left
behind. The only loophole available is to treat cranberry as a compound word made up of the free root
morpheme berry and the bound root morpheme cran- which has the peculiarity of only occurring as part of
the word cranberry. Admittedly, the word cran does exist, meaning ‘a measure of herrings (about 750
fish)’, but that word is unrelated to the cran- in cranberry.
We are in a quandry. You might be willing to accept that cranis a non-recurrent morpheme with a very
restricted distribution— occurring only in cranberry. But you may still be wondering where that leaves the
definition of the morpheme as a minimal meaningful unit. It seems that definition has to be modified. It has
to be weakened so that we make allowances for some morphemes which are not meaningful units on their
own. All words have to mean something when they occur in isolation. However, morphemes are not
absolutely required to have a clear, identifiable meaning on their own. They may be chameleon-like, with
meanings that change somewhat in different words (cf. [4.2] on p. 55).
Cran- and similar words make a revision of the definition of the morpheme necessary. To overcome the
problems some linguists have defined the morpheme in purely distributional terms: a linguistic entity
‘uniquely identifiable in terms of phonemic elements, and occurring in stated environments of other
morphemic segments (or in stated utterances)’ (cf. Harris 1951:171). This is a definition that highlights
distribution. ‘Cran- ’ can be isolated as a morpheme on distributional grounds, like the problematic Latinate
root morphemes in [4.2]. For instance, in the case of cran- we can use the substitution test to show that cran-
is a morpheme. In some other berry words such as blueberry and blackberry we have the morphemes blue
and black occurring in the same slot as cran-. (See also Aronoff 1976:15.)
In all this the difficulty is the fact that between the productive morpheme used actively in the construction
of words (e.g. agentive noun suffix -er) and dormant ones like -ery that are fossilised, there exists a myriad
of possibilities. The difference between productive and unproductive morphemes is a gradient, not a
dichotomy. There are numerous morphemes that are productive to different degrees. For instance -aire as in
doctrinaire is found in only a few existing words and is unlikely to be used in the formation of new ones.
Relatively few speakers recognise it as a suffix and its meaning is obscure. Nevertheless it is possible that it
is moribund rather than dead. It could still be taken off the shelf and dusted down and used to form new
words when the need arises. For instance, in the early 1990s those in British politics opposed to an ever
closer union of the states that make up the European Community used the term communautaire (borrowed
from French) as an insult for those firmly committed to the goal of European political unity. Although it is
not likely that the average speaker of English is aware of -aire being a suffix in communautaire, it is very
likely that those who use this term in political discourse, being quite erudite, are able to identify -aire as a
suffix here.
One of the intriguing phenomena in morphology is that acrossthe-board productivity in word-formation is
very rare. Often an otherwise general word-formation process is blocked in some circumstances. This is
particularly true in derivational morphology which tends to be less predictable than inflectional morphology.
Word-formation is often subject to various constraints which may be phonological, morphological or syntactic
(cf. Aronoff 1976, Katamba 1993).
Here we will limit the discussion to some phonological constraints. We will see how the affixation of an
otherwise productive affix may be inhibited for phonological reasons. We will illustrate this with the -al
 ENGLISH WORDS 79

suffix which derives action nouns from verbs. According to Siegel (1974), for a verb base to be eligible for
the suffixation of -al, it must meet these phonological requirements:

(i) The verb base must end in a stressed syllable.

(ii) The verb base must end phonologically in either a vowel, or a single consonant, or at most two
consonants.
(iii) If the base does not end in a vowel, its final consonant must be made in the front of the mouth (i.e. it
must be a labial or alveolar consonant).

These conditions are exemplified in [6.12]. (Consider the pronunciation, not the spelling):

[6.12]
a. -al suffix after vowel-final bases trial denial betrayal renewal
b. -al after verb bases ending in a labial consonant revival arrival approval removal

c. -al after verb bases ending in a dental or alveolar consonant

betrothal appraisal acquittal recital referral
d. -al after verb bases ending in a (alveolar) consonant preceded by another consonant
rehearsal dispersal reversal rental

However, if a verb ends in a palatal or velar consonant as in [6.13], suffixation of this -al is blocked:

[6.13]
*judgeal *attackal *approachal *rebukal *encroachal

The stress on the final syllable is always a vital factor. A base that otherwise meets the conditions for -al
suffixation will be ineligible if stress is not on the last syllable. Hence the ill-formedness of *`audital,
*`combatal and *`limital.
As Siegel’s analysis predicts, a base like *a`ttemptal, which ends in a three-consonant cluster, cannot take
the suffix either even though it has got stress in the right place, on the final syllable.
Even then there are inexplicable exceptions. Some seemingly well-formed potential words are disallowed
although they appear to meet all the conditions. For instance, for no apparent reason, *con`testal, *e`scortal
and *a`llowal are not allowed through the net.

6.7
PEEPING BEYOND THE LEXICON
After words have been formed in the lexicon, they are used in the syntax to form phrases and sentences.
Even at this stage they are still subject to phonological rules. The phonological rules that affect words at this
stage are called PHRASAL RULES (or POST-LEXICAL RULES: cf. section 8.3). Post-lexical
phonological rules differ from the phonological rules which operate in the lexicon in that they are not
sensitive to the idiosyncratic lexical properties of words. They apply mechanically whenever the right
phonetic circumstances are present, regardless of any special morphological or syntactic properties that
words may have. We have seen that, in the lexicon, certain phonological processes apply only if certain
morphemes are present. For example, STRESS SHIFT applies only when certain stratum 1 suffixes are
80 A LEXICON WITH LAYERS

attached. Unlike lexical rules, post-lexical rules do not take into account any of the peculiarities of the
words they affect. They are general and exceptionless. They are automatically triggered by the presence of
the appropriate phonological input.
An example of a post-lexical rule is the optional assimilation of an alveolar consonant at the end of one word
to the place of articulation of the consonant at the beginning of the next word in fast casual speech. Thus,
the word mad by itself is pronounced [mæd]. But in mad girl and mad man it may be uttered with a final [g]
and [b] respectively so that we get [mæg g :l] and [mæb mæn] when the /d/ assimilates to the place of
articulation of the following velar /g/ and labial /b/. The phrasal, place of articulation assimilation rule
applies in an environment that spans two phonological words. It is not a lexical rule confined to morphemes
and single words. The assimilation can affect any word with the appropriate phonetic make-up if it appears
in this phrasal context. However, since our concern in this book is morphology, the study of internal
structure of words, we will not explore phrasal rules any further. We will stick to processes that affect
morphemes and single words.

6.8
SUMMARY
In this chapter we have explored the kinds of lexical information that need to be included in the lexicon. We
have established that the lexicon needs to contain semantic, phonological, morphological, syntactic
properties. Word-formation interacts strongly with processes that apply in the other modules of the
grammar since they all need access to various kinds of information about morphemes and words found in
the lexicon.
Most of the chapter has been given over to characterising the relationship between morphological and
phonological rules found in the lexicon using the multi-layered theory of lexical morphology. (For further
exploration of this model see van der Hulst and Smith 1982a, 1982b, 1982c, Katamba 1989 and Goldsmith
1990.)
We have seen that affix morphemes play a pivotal role since the division of the lexicon into strata is
based on the phonological properties of affixes. Affixes are put on two strata depending on whether they are
phonologically neutral or non-neutral. This enables us to handle simply several important morphological
phenomena such as the sequencing of morphemes in words, the relative productivity of morphemes, and
blocking.
In the course of the discussion we re-examined the role of the morpheme as the minimal unit used to
signal meaning. We saw that some morphemes have very elusive meanings and we considered a
distributional approach as a possible way of characterising the morpheme.
Finally, we concluded by briefly contrasting the idiosyncratic nature of lexical phonological rules with
the predictable nature of phrasal rules that affect words after they have been processed through the syntax.

EXERCISES

1.

a. What kinds of information must the lexicon contain?

b. Explain why the lexicon must be much more than a long list of words that are found in a language.
What kinds of generalisations should be captured in the lexicon?