Phase space is the idea that transfer, i.e the shipping of pieces from morpho-syntax to the phonology, may leave a footprint in the phonology in form of extra syllabic space inserted at the left edge of the phase. Thus when [A [B]] is spelt out, it may be received as [A CV+B] by the phonology, where CV is an empty onset-nucleus pair. Extra syllabic space may identify in a number of ways according to theoretical inclinations: an empty CV unit as in our example, or and x-slot, or a mora.
We first review the background of this approach to the interface: phonologically relevant computational domains, i.e. where items are computed in one go, are defined by syntactic phases, and by nothing else. Everybody agrees that such domains exist: for example, stress is penultimate in [párent] and [parént‑al], as well as in [[párent] hood] due to the fact that [pãrent] is a domain by itself when class 2 affixes like ‑hood are added. But the literature has found two distinct identities for these domains, either derivational for chunks below the word size (cycles / strata in Lexical / Stratal Phonology), or representational for bigger chunks (prosodic constituency, e.g. the Prosodic Word). Thus Kiparsky’s infamous statement from the 80s “post-lexical phonology is not cyclic”, i.e. there are no derivationally defined domains for chunks bigger than the word.
This statement was made in an environment where a whole sentece was computed in the syntax before it was sent to phonology. If more recent phase theory is correct (Chomsky 2000 and following), Kiparsky was wrong: phase theory precisely chunks the syntactic derivation of a sentence into smaller pieces, called phases, which are shipped to semantics and phonology piecemeal, i.e. when their computation is complete. Thus phonology does receive successive pieces of the size of words and larger units.
In a phase-based environment, computational domains in phonology are thus defined by the shipping of pieces for all chunk sizes. This means that there is no place for prosodic constituency, which is defines representational domains above the word size: the prosodic hierarchy is redundant and has to go.
Another aspect of phase space is its non-diacritic nature: whatever objects are inserted into phonological representations, they need to belong to the domain-specific vocabulary of phonology. Items like # or ω, which are or were commonly used to carry morpho-syntactic information into phonology, are therefore banned: they are interchangeable diacritics, and banana would do the exact same job (“l vocalizes to w before banana”). By contrast, x-slots, moras or CV units are truly phonological items: they are not interchangeable and produce predictable phonological effects.
Phase space will be illustrated by a number of data sets, including English function words (him is full in give him vs. reduced in give’m), German definite and indefinite articles (DAT dem is full in auf dem Beton “on the concrete” vs. reduced in auf’m Beton) and 3rd person ACC clitics in Italo-Romance dialects (/lu/ is full in enclisis cámma lu “call him”, but reduced to u in proclisis u cámma‑ni “they call him”, Zonza / Corsica).
One consequence of phase space is that phonology now may produce evidence for syntactic structure. Syntacticians may therefore want to revise their typical current practice “spell-out und forget”: in a phase-based environment where phases leave phonological footprints, they may want to talk to phonlogists about things other than the weather. The data set from Italo-Romance dialects will illustrate this issue: phonology tells syntax where exactly the clitics sit.
Readings
D’Alessandro, Roberta & Tobias Scheer 2015. Modular PIC. Linguistic Inquiry 46: 593-624. [pdf]
Scheer, Tobias 2021. Generative phonology: The interaction between phonology and morpho-syntax. The Oxford History of Phonology, edited by B. Elan Dresher & Harry van der Hulst. Oxford: OUP. [pdf]