And phrases likepre-lecture preparationandpre-preparation lectureoccasion no surprise; nor do I have any notion whether
I've ever heard the mbefore. This property of productive morphology is of particular interest in a heavily affixing
language such as Turkish, in which word forms like (2) are not unusual (thefirst morpheme is the stem, the rest are
suffixes):
(2) gel- emi- yebel- ir- im
come-unable-possible-aorist-isg‘I may be unable to come.’(Van Valin and LaPolla 1997, 44)
çalış-tır-ıl- ma- malıy-mış
work-cause-passive-neg-oblig- infer‘Supposedly he ought not to be made to work’(Spencer 1991,189)
It is of no consequence whether or not a speaker of Turkish has heard these particular forms before: they can be
created and interpreted on the spot.
Now consider what it might mean for such forms to be created“in the lexicon.”If“in the lexicon”means“stored in
long-ter m me mory,”the implication is that speakers store all regular forms. On this hypothesis, whenone hears a new
noun ste mlikewug, oneimmediatelyadds notjustwugbut alsowugsto one's long-ter mstore.Si milarly, Dutch speakers
who learn a new noun also add to their long-term memory the diminutive form. For a more striking case, there is a
register of American English that permits one to insert an expletive into the interior of a word for emphasis, creating
such forms asmanu-fuckin-facture, ele-fuckin-mentary, and so on (McCarthy 1982). (British speakers, I' mtold, insertbloody
instead.) The insertion is possible only if the word has an appropriate stress pattern: roughly, its main stress must be
preceded by a secondary stress and preferablyan unstressed syllable—hence*ele-fuckin-phantis notpossible.But within
thisconstraint,themorphologicalprocess is perfectlyproductive. Thus thehypothesis thatthese forms are created“in
the lexicon”implies that whenever one learns a new word of appropriate stress pattern, one also tucks away in
memory the -fuckin- for mas well!
These cases are at least technicallypossible.The case of massivelyaffixing languages like Turkish is more problematic.
Hankamer (1989) observes that in such languages every verb has tens of thousands of possible forms, all constructed
through productive morphology. If all these forms are created“in the lexicon,”then the demands on long-term
memory are multiplied by four orfive orders of magnitude over a meagerly inflected language like English. When it
comes tothecapacityof long-term memoryoneis hard-pressed tosayhowmuchis implausible,but cases like Turkish
do stretch the imagination.
In fact, the productivity of such large inflectional systems is probably more extensive than that of, say, the Determiner
syste mof English. (3) illustrates,