David Willis

Maintaining the historical directionality of syntactic change in numeral phrases
David Willis (University of Cambridge)
Parametric approaches to syntactic change (Lightfoot 1991, 1999) suggest that languages
undergo radical restructuring of certain parts of their grammar at particular times. A single
underlying change is reflected in multiple surface changes manifested at roughly the same
time. Conversely, it has also long been recognized that a series of apparently related changes
may take many centuries to run to completion (Sapir 1949 [1921]: 150). These two
observations present a paradox: on the one hand, we wish to unify a series of changes by
deriving them from a single underlying shift; yet, disparate timings of individual changes
seem to necessitate a separate representation for each in the grammar. The long-term nature of
change presents a further problem: how can successive generations move their grammars one
step towards the new parametric setting in an apparently directed way?
This paper examines these issues with respect to changes in the syntax of noun
phrases containing numerals in Welsh, in the comparative context of similar developments in
Russian. I argue that a series of changes in the syntax of Welsh numeral phrases since the
medieval period reflects a single fundamental shift: Middle Welsh (MW) numeral phrases are
sensitive to properties (features) of the numeral (dual, numerative vs. singular, plural), but not
to properties of the noun (gender); Present-day Welsh (PDW) numeral phrases, on the other
hand, are sensitive only to the gender of the headnoun. This can be interpreted as a shift in the
semantico-syntactic status of numerals: in MW, number was represented as an interpretable
syntactic feature on numerals, [iNum: DUAL/NUMERATIVE]; today, the number of numerals is
purely semantic and not represented syntactically (i.e. the iNum feature has been lost). In the
short run, speakers ‘patch up’ the grammar, replicating much of the surface output of the
number-based agreement system, but without recourse to a number feature. In the long run,
loss of this feature leads to the spread of gender-based rather than number-based agreement in
the noun phrase. In this way, changes are unified as reflecting a single change, namely the
loss of the interpretable number feature, yet the long-term staged nature of the developments
is successfully modelled.
Evidence for the shift comes from changing patterns of agreement and initial
consonant mutation (data idealized from searching a substantial body of electronically
available texts throughout the period in question). In MW, the form of a noun in a numeral
phrase varies according to context: it may be formally singular or plural, or in a special
‘numerative’ form used only with numerals (e.g. ‘son’: singular mab, numerative meib, plural
meibion, see (1)–(3)). Modifying adjectives manifest complex patterns of agreement and
initial consonant mutation (‘soft’ mutation indicated using bold underline in (2)). On the
surface, these patterns mix items of different specifications (singular, numerative, plural) in
unexpected combinations. However, they become regular if we suggest that the entire phrase
is dual after ‘two’, numerative after higher numerals, and singular or plural as appropriate if
numerals are absent. On this analysis, agreement copies the value of an interpretable number
feature on the numeral to all heads within the noun phrase. The apparent complexity is due to
morphological syncretism e.g. cryfion is both the numerative and plural of cryf ‘strong’. Soft
mutation then appears throughout (2) because, with the numeral ‘two’, the feature [iNum:
DUAL] spreads to every head within the phrase, and this feature triggers soft mutation on the
following word.
In PDW, exemplified in (4)–(6), nouns and adjectives in numeral phrases are
uniformly singular. Agreement and mutation patterns distinguish between masculine and
feminine nouns, but are insensitive to the numeral itself. Adjectives in numeral phrase must
now be singular; soft mutation occurs on the numeral ‘two’ and the word immediately
following, but is otherwise determined by gender: soft mutation has disappeared with
masculine nouns in contexts that previously required it, while it has been innovated with
feminine nouns where it was previously excluded (e.g. MW tair mam cryfion ‘three strong
mothers’ > PDW tair mam gryf, with innovation of soft mutation on gryf).
Loss of the interpretable [iNum] feature is best placed early, during the MW period.
However, while some of the surface changes appear then, others are later, and some are not
complete even in PDW. Parts of the older system are maintained on the surface using ad-hoc
lexical entries for some time. The chronology of these changes suggests that a naïve
interpretation of parametric change cannot be maintained, and that some notion of
actualization must be incorporated into it. The relative ordering of changes may nevertheless
be significant: less salient contexts and those which are best specified in structural terms (e.g.
adjectives after feminine nouns with numerals higher than three) come into line with the new
parameter setting at an earlier date than contexts which are more salient or which can be
specified lexically.
Finally, we will consider support offered for this interpretation by the parallel of
changes in the syntax of Russian numerals (Babby 1987, Rappaport 2002). Here, loss of dual
has led to a far-reaching realignment of patterns of case and agreement, with existing surface
forms being partially maintained by reinterpreting them as instances of a new category of
quantitative genitive. Again, the new system is implemented over the space of several
hundred years, and the central shift is best understood as being in place earlier and triggering
micro-changes, rather than deriving from them.
mab cryf
meibion cryfion
son.SG strong.SG
son.PL strong.PL
‘a strong son’ (MW)
‘strong sons’ (MW)
ddau fab
son.SG strong.PL
the.DU two.DU son.DU strong.DU
‘the two strong sons’ (MW)
son.NUM strong.PL
the.NUM three.NUM son.NUM strong.NUM
‘the three strong sons’ (MW)
meibion cryf(ion)
son.PL strong.(PL)
‘a strong son’ (PDW)
‘strong sons’ (PDW)
y ddau fab
the two
son.SG strong.SG
‘the two strong sons’ (PDW)
y tri
the three son.SG strong.SG
‘the three strong sons’ (PDW)
Babby, Leonard H. 1987. Case, prequantifiers, and discontinuous agreement in Russian.
Natural Language and Linguistic Theory 5, 91–138.
Lightfoot, David W. 1991. How to set parameters. Cambridge, Mass.: MIT Press.
Lightfoot, David W. 1999. The development of language. Oxford: Blackwell.
Rappaport, Gilbert C. 2002. Numeral phrases in Russian: A minimalist approach. Journal of
Slavic Linguistics 10, 329–42.
Sapir, Edward. 1949 [1921]. Language. London: Harvest.