Author: Alec Marantz (Page 4 of 11)

I’m one of the co-editors of a forthcoming Handbook of Distributed Morphology (I’m the junior editor, in the sense that my co-editors are accomplishing way more than me towards getting the volume together. Senior Editors: Artemis Alexiadou, Ruth Kramer, and Isabel Oltra-Massuet.) Collectively, with the help we hope of our chapter writers and the broader community, the editors will put together an introductory chapter that explains some of the basics of DM in the context of the origins of the theory in the 1990’s. I’ll be Blogging some paragraphs for this chapter, in the hopes of generating questions and corrections. Please help!

Distributed Morphology, in particular as presented in Halle & Marantz (1993), emerges from a specific time and place within the trajectory of Generative Grammar: MIT Linguistics in the early 1990’s. The 1980’s at MIT had seen much work centered around the connection between the internal structure of words and the syntax and semantics of the sentences in which the words occurred. This included in particular the issue of whether the order of morphemes in words “mirrored” (to use Muysken’s observation (1981), later popularized by Baker (1985)) the hierarchical structure of sentences syntactically and semantically. And also bracketing paradoxes (Pesetsky (1985), Sproat (1988)). This work, however, was not at the heart of the development of DM. It did, however, rely on the piece-y-ness and abstractness of morphemes, as well as the parallelism if not identity between word structure and sentences structure.

The center of the DM paper was Halle’s concerns about allomorphy. Abstract morphemes distributed in the syntax were at the center of Chomsky’s (1957) Syntactic Structures’ approach to the English auxiliary system. Abstract -en, for example, came with have in a syntactic tree and was spelled out as (mostly) -ed and sometimes -en by context-sensitive phrase structure rules after affix-hopping. In the years after SS, lexicalist approaches to morphemes dominated phonology in particular, where -en and -ed would be separate morphemes in complementary distribution as perfect participle endings. Spell out of abstract morphemes in the SS manner allows for a straightforward approach to syncretism, where more specified allomorphs are inserted into an abstract morpheme position when their environmental requirements are met; elsewhere a default allomorph would appear. For, e.g., the English present tense paradigm, this seems the best analysis of marked -s (for third person singular subjects) and default zero, as opposed to a lexicalist position that needs to postulate 5 distinct zero present tense morphemes to fill out the paradigm.

Halle was working on a proposal that morphemes with suppletive allomorphs should be given the underlying phonological form of Q, a placeholder for phonological features. Then morphophonological rules would spell out the various allomorphs of Q in particular contexts. The Q attached to present tense features, for example, would spell out as -s in the context of third singular features and zero elsewhere. Morphemes without suppletive allomorphs, like progressive -ing – the concrete morphemes – would carry their phonological forms with them in the syntax. In discussions with Morris, although as a card-carrying lexicalist in Lieber’s (1980) sense, believing that morphemes contained phonological forms, I thought that the Q vs. concrete distinction was unnecessary within any theory that allowed Q’s – that allowed the phonological form of a morpheme to be spelled out in the phonology after syntax. If you have Q’s, every morpheme could be a Q, with concrete morphemes simply having only default realizations, not contextually sensitive ones. Once I got to work with Morris on complex inflectional systems with rampant syncretisms, I drank the Q-laid, abandoned the lexicalist assumption about phonological forms, and worked with him on a fully realizational morphology (all Qs all the time). Here we were particularly influenced by Robert Beard’s (e.g., 1986) work.

But the immediate impetus for the work in the original Distributed Morphology paper was Anderson’s A-Morphous Morphology research (1992). Anderson endorsed a realizational approach to inflection, but not to derivation – a split that we would put pressure on in later work but that was not under discussion in the 1993 paper. However, for Anderson the phonological forms of inflectional morphemes were not pieces – not units in the phonology. Rather, they were the by-product of morphophonological rules, identifiable as units by the linguist but not by the grammar. Anderson made two moves. First, he assimilated affixation to so-called “process” morphemes such as reduplication, metathesis, ablaut/umlaut and truncation. Prefixation and suffixation were just the addition or subtraction of phonemes, as in epenthesis. Second, he claimed that the order of overt inflectional material with respect to other such material and with respect to the stem was determined by the (stipulated) order of morphophonological rule blocks, which also served to explain some of the complimentary distribution of inflectional material. This order didn’t “mirror” syntax in any sense, and was in an important sense arbitrary. As part of this system of morphophonological spell-out, Anderson proposed a number of principles that yielded complementary distribution among certain inflectional spell-outs, to explain why we don’t have oxens and mices, for example.

The 1993 paper was clearly a response to Anderson. We re-analyzed illustrative data sets important to his framework from Georgian and Potawatomi and explicitly evaluated his approach to complementary distribution. Meanwhile, of course, research in what has been labelled “prosodic morphology” was demonstrating (e.g., Marantz (1982)) why so-called process morphemes might best be seen as the addition of a piece of phonological material that may have the effect of reducing phonological structure (as in hypocharistics and truncation) or rearranging phonology as in metathesis. We emphasized in the paper, however, that even if the phonological form of some morphemes should be seen as a process rather than as a piece, the processes could be seen to occur in a position in a hierarchical structure that was the spell out of the syntax. That is, process phonology could be considered a morphophonological “piece” in a particular hierarchicacal position within a syntactic tree structure of morphemes.

To be continued…..

Anderson, S. R (1992). A-morphous morphology (Vol. 62). Cambridge University Press.
Baker, M. (1985). The mirror principle and morphosyntactic explanation. Linguistic inquiry, 16(3), 373-415.
Beard, R. (1986). On the separation of Derivation from Morphology: Toward a Lexeme. Indiana University Linguistics Club.
Chomsky, N. (1957). Syntactic Structures. The Hague: Mouton.
Halle, M., & Marantz, A. (1993). Distributed morphology and the pieces of inflection. Hale, K. & SJ Keyser (eds.), The View from Building 20.
Marantz, A. (1982). Re reduplication. Linguistic inquiry, 13(3), 435-482.
Muysken, P. C. (1981). Quechua causatives and logical form: a case study in markedness.
Pesetsky, D. (1985). Morphology and logical form. Linguistic inquiry, 16(2), 193-246.
Sproat, R. (1988). Bracketing paradoxes, cliticization and other topics: The mapping between syntactic and phonological structure. Morphology and modularity, 339-360.

The work from the MorphLab on morphologically complex words in English has generally investigated words with one (overt) suffix. An exception is Yohei Oseki’s dissertation research (published in Oseki et al. 2019 and Oseki & Marantz 2020), which investigated combinations of prefixes and suffixes with an eye to long-distance relationships between the prefix and suffix. For example, un– attaches as a negative prefix to adjectives and as a “reversative” prefix to a subset of reversible verbs. The string unbeat is ill-formed as is, since beat is not one of the verbs semantically compatible with reversative un-, but unbeatable is fine, since the –able suffix creates an adjective from the verb, to which adjective the negative un– may attach. English prefixes are limited, and these combinations of prefixes and suffixes constitute a very small part of the general English vocabulary.

What about combinations of suffixes, as in nation-al-iz-ation? Here a consideration of the possible combinations yields some insights into the classification of morphemes cross-linguistically as “derivational” vs. “inflectional” and raises questions about whether we should expect morphological processing to reflect real differences in the function and nature of different morphemes.

A good starting point for a dive into multiple suffixation in English is Nigel Fabb’s (1988) article on affix order. While the literature has since produced many critiques questioning some of Fabb’s data and conclusions, his general overview of the data and the issues has proved foundational for further studies. Sticking only to English “word” suffixes (those that attach to stems that can be used as independent words, as opposed to bound stems like toler– in tolerate, tolerance and tolerable), Fabb finds strikingly few possible combinations of suffixes, even accounting for the category preferences of the suffixes (e.g. –able attaches to verbs). Most suffixes attach only to a finite set of stems and to a specific list of other suffixes. Some suffixes generally attach only to stems.

Particularly striking is Fabb’s conclusion (p. 535):

“A few suffixes are subject to no selectional restrictions other than those involving part-of-speech. These are:

(3)        [1] -able
            [17] deverbal -er
            [37] -ness”

(The numbers in square brackets reflect the affixes’ order in Fabb’s list of 43 suffixes.) Note that –able and –er attach to verbs while –ness attaches to adjectives. Fabb finds no selectional-free affixes that attach to nouns. Conspicuous in their absence from Fabb’s list are the participle endings –ed and –ing that create adjectival/stative passives and event nominalizations respectively; it’s not clear what motivated him to leave these out. The availability of an adjectival passive for a given verb depends on certain semantic and syntactic properties, not morphological or phonological ones. Similarly, –ing event nominalizations are generally OK on any verb semantically compatible with an event nominal. If we add –ing and –ed to the suffixes in (3), we get a plausible list of category-changing morphemes that could be part of the extended projection of verbs and adjectives. That is, the meanings in this list – for verbs, agentive nominals, event nominals, abilitative adjectives, result/state adjectives; for adjectives, property nominals – may be universal uses of verbs and adjectives and part of the paradigm for any (semantically compatible) verb or adjective in a language.

Note that nounhood is a dead end for this set of derivational suffixes. Verbs can yield nouns and adjectives, and adjectives can yield nouns, but none of the suffixes mentioned take nouns as their input. The maximum depth of derivation given these suffixes is two: verb-able/ed–ness. Examples would include: brok-en-ness, say–able-ness, close–d-ness, etc. However, nouns do enter into productive derivations, e.g. with –ish and –like: dissertation–like, farmer–ish, etc. For some reason, Fabb leaves –like off his suffix list and claims that –ish doesn’t attach outside other suffixes. However, –like and –ish seems amenable to an analysis similar to that of compounds, particularly since they may attach to phrases: I-don’t-care-like attitude, je-ne-sait-quoi-ish expression.

To go beyond three suffixes of the non-compound-ish type – and to consider the possibility of recursion in derivational morphology – we need to look at a set of Latinate affixes that Fabb identifies, including –al, –ion, –ity, –ism, –ist, and –ize, which interact with Latinate –ify, –ate, –ic, –ive, –an, –ous, and the previously mentioned –able. In particular, –al somehow gets us around the usual dead end of nouns, creating adjectives from nouns, and –ize allows the creation of verbs from adjectives. Since we can create nouns from verbs, we now can get adjectives from these nouns and verbs from the resulting adjectives, giving us the recursion noted by Lieber and others: nation-al-iz-ation-al-iz-ation-al-…  Where do –al and –ize fit into a universal set of morphemes perhaps associated with the extended projection of nouns and adjectives respectively?

Adjective forming –al creates an adjective from a noun with the meaning, “of or pertaining to the noun.” In English, for nouns like lake that resist the –al suffix, the noun itself can serve as a pre-nominal modifier: lake house, lake boat. For the same modification in predicative position, the noun would appear with a preposition: a house on a lake, a boat for the lake. It’s an interesting question whether languages generally have a way of creating a “of or pertaining to” modifier from a noun via affixation. Perhaps this question has been addressed in the literature, but if so, I haven’t seen the work.

For –ize, the question is whether languages generally have a way to create a causative accomplishment verb from an adjective. In English, the default construction for this meaning is a periphrastic “make NP Adj” construction (I made them happy, *I happyized them), but English lacks a productive derivational method to create causatives even from e.g. intransitive change of state verbs. So one could imagine a language with affixal causative constructions also having causatives formed from adjectival stems (or not).

This limited recursivity of derivation for English (and Indo-European languages in general) is not really parallel to what we see in agglutinative languages like Turkish. An example from Ergin, Morgan and O’Donnell (2020) illustrates the contrast.

Kur: set up
Kur-um: institution
Kur-um-sal: institutional
Kur-um-sal-laş: become institutional
Kur-um-sal-laş-tır: institutionalize
Kur-um-sal-laş-tır-ıver: institutionalize quickly
Kur-um-sal-laş-tır-ıver-e-me: cannot institutionalize quickly
Kur-um-sal-laş-tır-ıver-e-me-yecek: will not be able to institutionalize quickly
Kur-um-sal-laş-tır-ıver-e-me-yecek-ler: they will not be able to institutionalize quickly
Kur-um-sal-laş-tır-ıver-e-me-yecek-ler-imiz: those that we will not be able to institutionalize quickly
Kur-um-sal-laş-tır-ıver-e-me-yecek-ler-imiz-den: one of those that we will not be able to institutionalize quickly
Kur-um-sal-laş-tır-ıver-e-me-yecek-ler-imiz-den-miş: it has been heard/reported that (it) is one of those that we will not be able to institutionalize quickly
Kur-um-sal-laş-tır-ıver-e-me-yecek-ler-imiz-den-miş-siniz: it has been heard/reported that you are one of those that we will not be able to institutionalize quickly
Kur-um-sal-laş-tır-ıver-e-me-yecek-ler-imiz-den-miş-siniz-ce-sine: it has been heard/reported that you are as if one of those that we will not be able to institutionalize quickly

The first stages of the illustrated Turkish affixation mirror English institut-ion-al-ize, with the added feature of distinguishing the formation of the accomplishment verb from the adjective as an unaccusative (-laş) and the causativization of the unaccusative verb (-tır). After institutionalize, however, Turkish veers into non-English directions, piling on heads that correspond to independent words in English sentences, including adverbs (-yecek  ‘quickly’) and evidential-like heads (-miş  ‘it has been heard/reported’). It’s not clear whether these suffixes are derivational or inflectional on standard definitions, or whether it’s correct or insightful to call them “paradigmatic,” except if we use “paradigmatic” to describe words formed along the extended projection of a verb.

Various (sets of) affixes seem to have their own linguistic stories, resisting a general classification e.g. as “derivational” or “inflectional” suffixes. It is therefore unclear whether or not our processing models should themselves treat suffixes in a uniform manner. Time will tell.

References

Ergin, R., O’Donnell, T., & Morgan, E. (2020). Storage and Computation of Multimorphemic Words in Turkish. Cognitive Science Society 2020.

Fabb, N. (1988). English suffixation is constrained only by selectional restrictions. Natural language & Linguistic theory, 6(4), 527-539.

Oseki, Y., Yang, C., & Marantz, A. (2019). Modeling Hierarchical Syntactic Structures in Morphological Processing. Proceedings of the Workshop on Cognitive Modeling and Computational Linguistics, 43-52.

Oseki, Y., & Marantz, A. (2020). Modeling Morphological Processing in Human Magnetoencephalography. Proceedings of the Society for Computation in Linguistics, 3(1), 209-219.

In the post on phoneme surprisal, we asked whether the processing of morphemes within words, as measured by phoneme surprisal (the neural response to a cohort-based phoneme surprisal variable), might be less sensitive to contextual variables than we might have assumed. For example, the recognition of a suffix might involve surprisal computed not over a cohort of morphemes that are compatible with the stem (with candidate frequency within the cohort itself modulated by the transition probability from the stem to the candidate suffix), but over a cohort of morphemes compatible simply with the first phoneme of the suffix independent of morphological context.

Similar considerations might apply for visual word recognition and the M170 response from the Visual Word Form Area (VWFA). Instead of trying to explain the M170 using a single surprisal value computed from the likelihood of the word given the grammar, we should explore the possibility that the M170 is modulated by a number of variables that aren’t fully captured by a single surprisal measure.

The landmark paper exploring morphological variables modulating the M170 is Solomyak and Marantz (2010), which itself built on the work of Zweig and Pylkkänen (2009), which demonstrated that morphologically complex words yielded larger M170 responses than matched morphologically simple words. Solomyak and Marantz explored M170 reactions to a diverse set of derived English words, with a variety of suffixes. We included both free stem words (farmer, where farm can occur by itself) and bound stem words (tolerable, where toler– occurs also in e.g. tolerate, but not without a suffix). We also included “unique stem” words, like amity, whose stem arguably occurs only in the presented word. This last group did not yield absolutely clear results, but a return to this type of word in Gwilliams and Marantz (2018) provided evidence that they, too, are decomposed at the M170 response.

The oft-cited result from Solomyak and Marantz is that, for free stem and bound stem words, the M170 is modulated by the transition probability from stem to suffix but not by the surface frequency of the word, once the variance associated with transition probability is removed. What is not often remembered is that other variables did in fact modulate the M170 independent of transition probability, specifically stem and affix frequencies. That is, in addition to the contextual variable transition probability, a-contextual unigram stem and affix frequencies also showed significant effects at the M170 response. These results are variously replicated in subsequent studies.

Consider the possibility that in the general M170 brain region and time interval, multiple connected processes are performed. The look-up of the representations of morphological forms is pursued along with the evaluation of the syntactic structure connecting multiple morphemes, if more than one morpheme is recovered from the input. Here, as suggested for auditory processing, the recognition of individual morphemes based on the visual input might be governed by contextual variables as well as a-contextual variables. If this suggestion is correct, we should be able to distinguish a number of sub-responses to different stimulus variables in perhaps different sub-regions of the VWFA and neighboring cortex and at different time points within the general M170 interval.

Another possibility exists, where a single measure of visual word surprisal, computed by us from the variables discussed and perhaps other variables, can account for all the relevant variation in the M170 response, without decomposition of this response. However, if our cognitive understanding of the various variables involved (e.g., stem frequency, affix frequency, transition probability) implicate different processes in a cognitive model, then even if we’re successful in accounting for the M170 response in terms of this single, composite variable, we can’t be said to have explained the M170 – we won’t understand why this variable works. Rather, we would need to re-think our cognitive theory of complex word recognition to make sense of why that single variable would be key.

References

Gwilliams, L., & Marantz, A. (2018). Morphological representations are extrapolated from morpho-syntactic rules. Neuropsychologia, 114, 77-87.

Solomyak, O., & Marantz, A. (2010). Evidence for early morphological decomposition in visual word recognition. Journal of Cognitive Neuroscience, 22(9), 2042-2057.

Zweig, E., & Pylkkänen, L. (2009). A visual M170 effect of morphological complexity. Language and Cognitive Processes, 24(3), 412-439.