24.962 Page 1 5/23/05 Cyclicity and its OT translations Outline: •Characterize and illustrate the cyclic syndrome.•Laws predicted by cyclic rule application•An example of what cyclic application does/does not deriveProperties of cyclic rule application (1) The cyclic syndrome: an unexpected phonological similarity between base and derivative. •Two lexically related forms A and B.•B = [aff//A] (B is a derivative of A) •A derived property P is common to A and B (e.g. stress on same syllable in A and B)• Property P is predicted by the phonology in A but not in B. Example: stress on 2nd syllable from left in perípheral, perìpherálity; sublíminal, sublìminálity; solídify, solìdificátion. Stress on 2nd syllable is predictable (conforms to relevant constraint hierarchy on stress distribution) in perípheral, sublíminal, solídify, but not in perìpherálity; sublìminálity, solìdificátion (cf. Tàtamagóuchi, hàmamèlidánthemum). The anticipated solution: let O-O Corresp constraints outrank phonotactic constraints. (2) The derivational characterization of the cyclic syndrome: • Cyclic domain (1st attempt): any morphosyntactic constituent •Addition of any morpheme to a given constituent defines a new cyclic domain.• In a morphosynt. constituent [X//[Y//[Z]]], 1st cycle = the innermost constituent, [Z]. • Cyclic rule application:A sequence of ordered rules R1-Rn applies first on the 1st cycle. The sequence R1-Rn reapplies on successively larger cycles until the largest constituent has been processed. •As always with rules: the input to Rulei is the output of the previous rule Ri-1 Therefore the input to Cycle n is the output of Cycle n-1 Rules (vastly simplified) : Stress antepenult; Stress 1st, unless clash Stress demotion (1…1 -> 2…1) Cycle 1, input: lexical entry peripheral Stress antepenult: perípheral Stress 1st, unless clash: N/A Stress demotion: n/a Cycle 2, input perípheral-ity Stress antepenult perípherál-ity Stress 1st, unless clash: N/A Stress demotion perìpherál-ity Compare Tàtamagóuchi, Mèditerránean: longer 1st cycle domain allows Stress 1st to apply successfully. A non-cyclic application of stress to peripherality would yield same stress as in those forms. (3) Laws predicted by cyclic rule application (i)Occasional base-derivative similarity: Bases and derivatives are predicted to be occasionally more similar to each other than expected from the simple non-cyclic application of phonological processes/constraints. E.g. perípheral and perípheràlity vs. Tàtamagóuchi24.962 Page 2 5/23/05 (ii) But No Guarantees: The simple fact that rules apply cyclically does not guarantee base-derivative similarity: cyclically applied rules can decrease base-to-derivative similarity. E.g. rémedy and rémedying are similar because stress didn’t reapply. Had stress re-applied on the –ing cycle, we’d get {rémedy, *remédying} like {órigin, oríginal} (iii) Base priority: The phonology of bases influences the phonology of derivatives, not vice versa [How COULD it could be otherwise? Imagine that for every pair of base and derivative, rules apply from outermost cycle inwards, that is on the derivative. On every cycle the outermost affix is stripped; rules reapply on the inner domain, if their SD is met: Cycle 1: input peripheral-ity Stress antepenult: peripherálity Stress 1st unless clash: péripherálity Stress demotion: pèripherál-ity Cycle 2: input pèripherál-ity Strip affix: pèripherál Stress antepenult: blocked in clash No further rules apply Output: *pèripherál (like Kàlamazóo) Resulting paradigm *{ pèripherálity, pèripherál} is uniform wrt stress, just like the real paradigm {peripheral, perípheràlity }. But it’s the wrong paradigm. General point here: it’s possible to imagine what paradigms consisting of similar bases and derivatives would look like if Base priority did not hold. It’s an empirical fact that the world wasn’t made that way.] (iv) Proximate base effect: The proximate base of [X//[Y//[Z]]] is [Y//[Z]], not [Z]. The phonological properties of the derivative are determined by the phonological properties of its proximate base. E.g. stress in orìginálity is determined by stress in oríginal , not in órigin. (v) No transderivational similarity: Suppose you have two co-derivatives of Z: [X//[Y//[Z]]] and [W//[V//[Z]]]: they will share phonological properties only because they share [Z] or, possibly, because X-Y and W-V are accidentally similar. E.g. stresses of demónstrative and demónstrable: based on démonstràte Or bureáucratism and bureáucracy: based on búreaucràt. Accidental identity? (v) No foresight: Information that is available on Cycle n+1 is not seen on Cycle n. E.g. English liquids becomes syllabic if not adjacent to a V: Syllabic is liquid in cycling, puzzling: V unseen on cycle 1. (4) Not all processes are cyclic: Flapping ('V_V) á[|]om a[tÓ]ómic, *a[|]ómic [in general allophonic differences tend to be less noticeable (most naïve speakers aren’t even aware of the flap-stop difference) and it is the more noticeable properties that tend to have reported cyclic effects. But see Steriade 2000 in Labphon 5 for examples of allophonic cyclicity in English and French.]24.962 Page 3 5/23/05 (5) Which types of processes are cyclic? Unclear answer in LP a. Mascaró 1976: all and only obligatory neutralization rules are cyclic. [Russian devoicing is neutralizing and non-cyclic: goro[t], goro[d]-a Sundanese nasalization is non-neutralizing and cyclic: ma) rai) n]rios vs. ≠a) ) a)b. Mascaró 1976 and Kiparsky 1982: all and only rules blocked in NDE are cyclic. [ NDE = non-derived environments, i.e. contexts present as such in some morpheme's UR.] [Finnish CGradation (Kiparsky 1993) blocked in NDE but not cyclic: hattu 'hat', hatu-n 'hat-GEN'; but appelsiini 'orange' hattu-n-si -> hattusi 'hat-GEN-yours' (ns -> s)] (6) Are the laws predicted by cyclic application empirically true? (i) Occasional base-derivative similarity: yes, but see below. (ii) No guarantees: see below. (iii) Base priority: yes, this is empirically verified. (iv) Proximate base effect: variable, generally holds, but ask about Indonesian stress. (v) No transderivational similarity: perhaps not, but debated. (vi) No foresight: not investigated A famous example to illustrate these points (7) Effects of cyclic