MORPHOLOGY: STORAGE
Research has focused on how inflectional and derivational morphemes are stored and accessed, and on what constitutes a word primitive, the smallest unit of meaning stored in the lexicon. Three views are possible:
Each lexical item is stored in the lexicon as an individual entry. There are thus entries for HAPPY, UNHAPPY and HAPPINESS and for WALKED as well as WALK. However, there are close links between related items.
There are separate entries for productive bound morphemes: affixes like UN- and-NESS have entries alongside actual words such as HAPPY.
Inflected forms are stored under their root. In order to retrieve UNHAPPY, one has to access HAPPY.
Evidence from Slips of the Tongue and from aphasia strongly suggests that inflected forms are constructed during the planning of speech: i.e. that we access WALK and then attach-ED to it at a later stage. Any other solution would appear to be wasteful in terms of what is stored in the lexicon. However, certain items may possibly be stored as wholes because they are usually encountered in an inflected form (EYES, HAPPENED).
The situation for derived forms is less clear-cut. In forming UNHAPPY from HAPPY, the language effectively creates a new unit of meaning. Furthermore, the connection between HAPPY and UNHAPPY is not fixed by a rule like that between WALK and WALKED. There would have to be strong links between HAPPY and the prefix UN- to avoid misassembled forms such as DISHAPPY or INHAPPY.
The ‘separate storage’ account entails considerable complications for the reader or listener. In order to access the meaning of a word like UNHAPPY, a reader would have to ignore the prefix UN- and focus on the word’s root. Listeners would have to identify the prefix and store it in memory until the root had been heard. Since listening takes place in real-time, they would need to operate on the assumption that all sequences such as /In/or/rI/or/dIs/ were prefixes. This would mean that they ended up stripping off pseudo-prefixes such as re- in return or dis- in display, which would then have to be restored in order to identify the target word. There is some evidence that this is what happens: readers are said to be slower to recognise words with pseudo prefixes, perhaps because they lead to erroneous prefix-stripping.
But the process is not an efficient one– and the existence of large numbers of pseudo-prefixes in certain languages raises serious questions about the validity of the prefix-stripping account. It has been calculated that up to 80 per cent of words in English beginning with strings that appear to be prefixes are in fact pseudo-prefixed. One compromise solution is that lexical access of derived forms does not depend critically upon decomposition but that decomposition can occur.
It should be noted that the same complication does not arise with suffixes. Here, stripping is not necessary, since the root has already been identified by the listener before the suffix is reached. It is interesting to note that, across languages, suffixation is much more frequent than prefixation; this may be because suffixes cause fewer problems of processing.
Despite the above complications, an investigation by Marslen Wilson et al. (1994) appears to provide new evidence for separate storage of affixes. It made use of the priming task, which indicates how closely words (in this case, spoken words) are associated with each other. The results suggested two-way links between affixed forms and their roots. But whereas a prefixed form appears to give a boost to other forms bearing the same prefix, there is no such effect with suffixed forms. The explanation offered is that hearing a prefix or pseudo-prefix activates a whole group of associated words; but that no competition between the words takes place until the root is reached, so all the words remain highly activated. The researchers suggest that their results support a decomposition model in which potential affixes are clustered around a root (UN- and-NESS attached to the head word HAPPY).
See also: Access code, Lexical entry, Word primitive
Further reading: Aitchison (2003: Chap. 11); McQueen and Cutler (1998); Marslen-Wilson (1999); Stemberger (1998); Taft (1981)
