Moving up and down parameter hierarchies: markedness, third factors and diachrony

Add to your list(s) Download to your calendar using vCal

• Prof Ian Roberts (University of Cambridge)
• Thursday 11 May 2023, 16:30-18:00
• English Faculty, GR05.

If you have a question about this talk, please contact CLF.

Parameter hierarchies of the kind proposed and developed at length in Roberts (2019) give us an automatic definition of markedness: the lower an option is in a hierarchy, the more marked it is. It follows that any diachronic change which involves a system moving from a higher to a lower option on a parameter hierarchy is in fact a change from a relatively less to a relatively more marked option. Conversely, any change which involves movement upwards in a parameter hierarchy involves movement from a more to a less marked option.

I will discuss and illustrate several examples of changes involving feature loss and concomitant movement upwards in a parameter hierarchy (the loss of V-to-T movement in Early Modern English, the loss of V2 in the history of English, the loss of multiple wh-movement between Latin and Romance and the loss of wh-movement altogether in various languages including Chinese; see the parameter hierarchy for wh-movement in Roberts 2019:529).

On the other hand, there are also examples of changes in which a system moves downwards in a hierarchy. For example, the Final Over Final Condition (FOFC) requires word-order change from head-final to head-initial to affect higher heads in a given Extended Projection before, or at least not later than, it affects lower ones. I will present evidence that TP changed from head-final to head-initial in both the history of English and the history of Latin/Romance. The evidence for successive shifts from head-final to head-initial orders in successively lower categories in the clause is evidence of systems moving down the word-order parameter hierarchy over time, and hence change from relatively less to relatively more marked. A further case of this type is the change from consistent to partial to non-null-subject-language, seen in the history of several Germanic languages as well as French.

On the one hand, it is desirable to have a mechanism allowing change from less to more marked systems: our theory of change must accommodate the fact that the overall range of diversity in the world’s languages does not, as far as can be discerned, seem to have changed over time. Certainly there is no evidence for any kind of general tendency towards a steady-state maximally unmarked system. On the other hand, if the notion of markedness is to have real content, there should be at least a weak dispreference for change from less to marked in contrast to change in the opposite direction. I will suggest that exactly this near-equilibrium can be achieved by the right approach to the third factors in language design (see Chomsky 2005).

Consider the three third factors Feature Economy (FE), Input Generalisation (IG) and the Subset Principle:

(1) (i) Feature Economy (FE) (see Roberts & Roussou (2003: 201)): Postulate as few FFs as possible, given the PLD . (ii) Input Generalisation (IG) (see Roberts (2007a: 275)): Maximise available FFs. (iii) Subset Principle (SP) (Berwick 1982): Postulate the smallest grammar compatible with the Primary Linguistic Data.

FE does not always disfavour movement down a parameter hierarchy: it frequently alters the distribution of features in a system without changing the inventory of features. The lower options in a hierarchy are, however, successively more marked in relation to IG, but not necessarily in relation to FE. Both FE and IG favour higher positions in parameter hierarchies and therefore favour upward parametric changes, but only IG actively disfavours lower positions; FE, to the extent that new features are not introduced in lower positions but only the distribution of features is affected, is neutral. Furthermore, the Subset Principle can also be seen in this light: if we think of the Subset Principle as a preference to avoid superset traps by preferring more restricted feature distributions over less restricted ones, then it will always favour going down the hierarchy as an acquisition strategy. To this extent, the Subject Principle favours lower, and therefore more marked, parameter setting. But the effects of the Subject Principle are slight and indirect as far as syntactic change is concerned.

Taking the three third factors together then, we have the following general picture: Feature Economy Favoured Neutral Input Generalisation Favoured Disfavoured Subset Principle Disfavoured Favoured

Upward change  Downward change

TABLE 3 .1 Third factors in relation to direction of change in a parameter hierarchy

We see that, taken together, the third factors give rise to a slight preference for upward over downward change in parameter hierarchies, i.e. from more to less marked. Given that markedness always involves preferences, this seems like a good result.

This talk is part of the Cambridge Linguistics Forum series.

This talk is included in these lists:

• Cambridge Forum of Science and Humanities
• Cambridge Language Sciences
• Cambridge Linguistics Forum
• Cambridge talks
• Chris Davis' list
• English Faculty, GR05
• Featured lists
• Guy Emerson's list
• Language Sciences for Graduate Students

Note that ex-directory lists are not shown.