You might at this point be wondering if it really makes a difference whether we conceive of morphemes as things with discrete meanings and their own lexical entries, or as parts of morphological rules that realize unanalyzable words with complex meanings. Our conception of what morphological rules look like really does matter, because it makes predictions about what sorts of morphology we should expect to find in the languages of the world. On the one hand, Item and Arrangement theories predict that morphology ideally should be agglutinative, with words segmentable into several pieces, each of which has a distinct meaning. In some languages this is the case – recall our sketch of Turkish. On the other hand, Word and Paradigm theories, although they do not strictly preclude agglutinative morphology, lead us to expect that morphology typically ought not to be agglutinative; rather, it should contain lots of multiple exponence, as is the case in Latin, but not in Turkish. The problem, of course, is that neither of these predictions is quite right. Some languages are more agglutinative than others. Some languages have lots of multiple exponence, and others have little or none.

So it does make a difference which kind of theory we choose. But the choice is made difficult by the complexity and variety of morphology we actually do find in the languages of the world. It is made even more difficult by the fact that each of these models is not really just a single theory, but a kind of umbrella that encompasses a number of different theories. For example, in their simplest form IA models say that the correspondence between meanings and ‘pieces’ ought to be one-to-one, but they need not say this. It is possible to propose an IA model in which the relationship between pieces and meanings is ideally, but not strictly, one-to-one. Similarly, there are a number of different versions of WP or realizational models that are plausible. We will not be able to go into the various theoretical possibilities here, but you will no doubt encounter them in a more advanced course in morphology. It seems safe to say, though, that each theory must be tested against a wide variety of languages with all different sorts of word formation, and a wide variety of inflectional and derivational word formation processes including not only affixation, but also compounding, conversion, reduplication, and templatic morphology. We will end this section by simply saying that the jury is still out on whether IA, IP, or realizational models of morphology constitute better models of how morphology is organized in the human mind.