314 Ahmed Moutaouakil
tional relations, pragmatic functions are clearly text-based notions (cf. Dik
1997a: 309–338), although they may be expressed within the clause.
As shown above, term structure can be said to run parallel to clause
structure. However, the layering of term structure typically does not go be-
yond the lower layer in the interpersonal level, i.e. the modal layer. It is
hard to speak of a ‘term illocutionary layer’, at least as far as explicit term
structure is concerned. The values of the modal operator are more restricted
in the term than they are in the clause: I have shown elsewhere (Mou-
taouakil 1993) that only volitional and emotional subjective modality
distinctions can occur in a modalized term (excluding evidential distinc-
tions).
In languages with a concatenative morphology, word formation is
achieved, as is well known, through an affixation (prefixation, suffixation,
infixation or circumfixation) rule which applies to a stem. Approached in
terms of the GPH, word structure can be conceived of as a partial actualiza-
tion of structure (17): on a stem standing as a nucleus, prefixes and suffixes
build hierarchically organized layers. As one might expect, the actualiza-
tion of structure (17) in the word is more restricted than it is in the other
discourse categories. First, affixes can express all the features of the repre-
sentational level through a quality layer (Republican, Chinese, panelling,
anti-social etc.), a quantity layer (hypercritical, supernatural, overdressed,
subhuman etc.), and a locality layer (subway, transatlantic, pre-marital,
post-classical, unfair, etc.). Word layering can attain the interpersonal
level, as in the case of words with pejorative affixes such as: (misleading,
malodorous etc.). However, it cannot go beyond the modal layer: I am
aware of no affixes which can be said to express illocutionary features.
Second, not all the operator values may obtain at the word level. For in-
stance, the modal values are restricted, it seems to me, to pejorative
features. Third and more importantly, words do not have, as one might ex-
pect, the same hosting capacity as terms and clauses. A complex word can
hardly contain more than two prefixes. This means that it would be diffi-
cult to find words involving more than two layers at once.
A final word on the text-clause parallelism: the structural similarity be-
tween the clause and the text is commonly described in terms of the
projection of the structure of the former onto the latter. The facts examined
above clearly show that the projection process – if we can speak of such a
process at all – should be conceived of as taking place in the reverse direc-
tion, i.e. from the text to the clause. With the approach proposed here we
no longer need the notion of projection: the similarity between the different
discourse categories is not due to a structural projection from one category