Morphological Roots and the Syntax Ammar Ammar ammar@mit

advertisement
Morphological Roots and
the Syntax
Ammar Ammar
ammar@mit.edu
In this paper, I try to investigate the existence of lexical roots or kernels as the atomic elements of the
lexicon. Assuming the existence of such roots, I also try to explore the nature of the interaction
between roots and patterns and the syntax, in phases and verb alternations.
Roots:
Traditionally, roots are considered to be the abstract elements that make up the lexicon and store the
necessary information for the formation of words. While root are essentially unpronounceable and
devoid of a definite meaning, their appearance in different words usually creates considerable semantic
and phonological relationships between these different derivatives of the root.
The significance of this approach to the formation of words stems from its agreement with the
mainstream models of generative grammar, which favors views in which the lexicon behaves like an
appendix to the grammar, storing only the abstract non-redundant information used by the other parts
of the computational system.
In languages like English, we the need for such abstraction is not obvious; this is due to the fact that
words (pronounceable lexical items) are the input for all syntactic and morphological processes in
overwhelming majority of cases. In contrast, the standard approach to morphology in Semitic
languages, like Arabic and Hebrew, is traditionally based on the existence of roots as the building
blocks of the lexicon. Words are then formed by applying various patterns and phonological process to
the root.
This root-based approach to semitic morphology is motivated by the nonconcatinative nature of
Semitic morphology, and the easily observable relations between words that share certain consonants,
namely the root or the pattern consonant. The following words which are some of derivatives of the
consonantal root (k-t-b) provide an example of the aforementioned phenomenon:
(1)
kataba
"wrote"
kitab
"book"
maktaba
"library"
maktoob
"letter"
kateb
"writer"
estaktaba
"asked to write"
kotiba
"was written"
As you can see, all the words above have something to do with writing, and they all contain the
consonants "k,t, and b". Moreover, verbs derived from different roots using the same patterns also
have semantic, phonological, and syntactic similarities, which are usually weaker, as we can see in the
following words which are derived from different patterns:
(2)
x-r-j + CaCaCa = xaraja
"exited" (intransitive)
x-r-j + oCCiCa = oxrija
" was caused to exit" (passive)
x-r-j + aCCaCa = axraja
" caused to exit" (transitive)
x-r-j + estaCCaCa = axraja
"asked X to exit"
d-x-l + CaCaCa = daxala
"entered" (intransitive)
d-x-l + oCCiCa = odxila
"was entered"(passive)
d-x-l + aCCaCa = adxala
"caused to enter" (transitive)
d-x-l + estaCCaCa = estad5ala
"asked to be entered"
In these verbs, the application of the same pattern to two different 1 roots results in verbs that have
similar semantic, phonological, and syntactic properties. The actual process of deriving verbs from
roots is actually more complicated, and involve changes on the phonology of the word. It is important
to note that the same procedure is used to derive other categories (e.g. nouns and adjective) from roots.
If we assume the Root Hypothesis, and consider the examples above to illustrate a type of derivation
processes, we can say that verbs derived this way form the majority of verbs in Arabic. Other verbs and
nouns are derived from other nouns that are either derived from the root or borrowed from other
languages; such verbs and noun are closely tied to their nominal stem in their semantic and
phonological features, as we can see in the following Arabic examples:
1 The two roots must be semantically related in order to yield results like one ones in the example.
(3)
telefon
n, "telephone"
talfana
v, "made a phone call"
talfaneh
n, "making a phone call"
Do Roots Exist?
As I mentioned before, it seems natural that researchers working on Semitic language would favor the
Root and Pattern approach to morphology, but that's not the case, and the Root hypothesis is not widely
assumed. Instead, the Lexicalist Hypothesis takes words to be the atomic elements of the lexicon, from
which everything else is built (Arad 2005). Bat-El (1994) and Ussishkin (1999), for example, has
argued that the root-based approach is inadequate to explain many of the phonological regularities in
denominal verbs in Hebrew, showing that we need the word based approach is needed to account for
such regularities.
While it is true that the word-based approach offers a simple explanation that can account for both
verb-derived and noun-derived verbs in all languages, I believe that storing words, as opposed to
storing abstract roots or kernels, involves less abstraction and more redundancy in the lexicon.
Moreover, the word-based approach misses the point by flattening the differences between verbderived and denominal verbs.
Arad (2005) argues for the existence of roots by observing the constraints on the interpretation of roots
in Hebrew denominal verbs. She notes that while root-derived verbs can be interpreted in multiple
ways, noun-derived verbs cannot, and must be tied in their meaning to the nouns from they are derived.
To illustrate this, consider the Hebrew words derived from the root "s-g-r" and the their interpretations:
(4)
Now consider the noun "misgeret" and the verb "misger" derived from it:
(5)
In addition to the root consonants, the verb "misger" contains the prefix "m-", which indicates its
nominal origin. According to Arad, to form the verb "misger", first the consonants of the root "s-g-r"
are combined with noun morphemes in "misgeret", using the pattern miCCeCet. The noun is then
embedded under a 'v' head to create the verb; this is usually accompanied by the truncation of nominal
suffixes and some other phonological changes. The trees below illustrate the process:
(6)
In the example above, we notice that the noun-derived verb "misger" is tied to the noun "misgeret" in
its meaning, and doesn't enjoy the same freedom of interpretation enjoyed by root-derived verbs seen in
the examples before. This might sound like a trivial observation, but Arad argues that this provides a
crucial distinction between the two classes of verbs; in her opinion, this distinction offers a good
empirical evidence for the existence of roots and world-level phases. This distinction goes beyond the
semantic behavior of the two verb classes to their phonology2. (Arad 2005)
Further evidence for the existence of roots can be found by studying the distributional patterns of verbs
and nouns in Semitic languages, through observation of patterns, regularities, and native speaker
judgments. For more arguments regarding the existence of roots, see work by Boudelaa and MarslenWilson on Arabic and Frost et al. , and Deutsch et al. on Hebrew cited by Ussishkin (2006).Roots
and Syntax:
1. The Syntax/Morphology Interface:
As the name implies, a theory of the interface between the morphology and the syntax needs to account
for the relationship between internal structure of words and the structure of phrases, and how the rules
for deriving complex words affect the derivation of syntactic structures.
One approach to this problem assumes a version of the Lexicalist Hypothesis mentioned in the previous
section. In this approach, words are considered to be, somehow, special and different from phrases. As
a result of this specialness, words are assumed to be a product of a special module of the grammar: the
lexicon. According to this approach, morphology and syntax share an opaque interface; that is to say,
even though words serve as the terminals in the syntactic derivation, their internal structure is
determined apart from the syntax.
An alternative approach to the same problem is the theory of Distributed Morphology, in which 'words'
are taken to be another product of the syntax. In this theory, 'words' lose their special status, and the
question of word structure is reduced to another question about the nature of the syntax; these questions
can then be answered by studying the primitive elements that form the inputs to the generative system,
and the set of rules and principles governing the work of this system, and its interaction with the
phonology. (Embick et Al. 2005)
While the lexicalist approach reliably provides a way to study words in separation from the complex
complications of the syntax, I believe that it sometimes fails to account for similarities between the
structure of phrases and words, especially in polysynthetic languages, where syntax and morphology
are hard to separate. In contrast, Distributed Morphology allows for a simpler, more transparent
2 for a thorough discussion of the topic, see Roots and Patterns by Maya Arad.
interaction interface between the lexicon and the syntax, and shifts a considerable part of the burden of
word formation from the former to the later, which gives more space for abstraction and eliminates
some of the redundancy. Distributed Morphology also makes place for more elegant explanations for a
number of syntactic phenomena which seem to have morphological origins, such as verb alternation
and phase spell out.
All that said, it's important to point out that when we are talking about abstract roots or lexical kernels,
the Distribute Morphology view of words becomes the natural choice, as roots become the natural
candidates for the inputs of the syntax assumed by the theory. From this point of view, the difference
between root-derived verbs and denominal verbs can be explained by invoking syntactic principles like
phase spell out. Furthermore, foreign words entering the language can be treated as new words (or
complex elements) and used to derive new words taking the phonological constraints of the language
into account. Similarly, the difference between languages like English and Semitic language can be
explained by saying that English simply has more lexical kernels than Hebrew.
2. Roots and Phases.
As mentioned in the previous section, the structure of words, which is traditionally the domain of
morphology, and that of phrases exhibit a number of interesting similarities. One such similarity comes
from the phenomenon known as Phase Spell-Out. Briefly speaking, Phase Theory tries to explain the
derivation complex sentences by dividing this derivation into separate derivations of local domains;
that is, every sentence goes through a number of phases before it reaches its final structure.
At the end of each phase, a semantic or phonological spell out is said to take place, resulting in some
constraints on the kind of operations that can be applied to the sentence derived so far. For example,
wh-movement might take place before a certain spell-out, but not after it. This also applies to certain
phonological operations.
Going back to Hebrew verbs, the semantic and phonological differences between root-derived and
noun-derived verbs can also be explained using phases that apply on the word level. In order to do
that, we need to distinguish roots (the abstract lexical kernels) from words (complex elements
constructed using roots and patterns). We also need to impose a constraint on the interpretations
assigned to a specific root, and ultimately on word formation from that root. This constraint can be in
the form of assigning semantic interpretations and phonological reincarnations to the root directly after
the first merge operation or the application of the first pattern. Once the semantic interpretation and the
phonological incarnations are assigned, they are carried out through the rest of the derivation process.
In this case assignment is considered to be a spell out point. (Arad 2003)
3. Roots and Verb Alternations:
As we noticed in the first section, another domain where the internal structure of words seems to affect
the structure of sentences is the domain of verb alternations. In Arabic, for example, the verb class is
determined by it's pattern, and a single root can create multiple alternations of the verb as needed.
Each of these alternation then determine the structure of the sentences it's used in. To illustrate this, lets
consider the roots "n-z-l" and "k-t-b", together with some of its possible alternations.
(7)
kataba
"wrote"
kotiba
"was written"
enkataba
"was written"
kaatab
"corresponded with"
kattab
"caused to write"
kuttib
"was caused to write"
estaktaba
"asked (someone) to write"
ektataba
"asked (for something) to
be written"
nazala
"descended"
anzala
"caused to descend"
onzila
"was caused to descend"
tanazzala
"descended heavily"
nazzala
"caused to descend
heavily"
nozzela
"was caused to descend
heavily"
estanzala
"asked to descend "
(8)
Conclusion:
In conclusion, I believe that while it's possible to explain some of the observations concerning the
interface between morphology and syntax without using the use of abstract roots or any other sort of
underspecified atomic lexical elements, the use of such elements offers and elegant way to account for
the interaction between the morphology and the syntax, and allows us to eliminate many of the
redundancy caused by viewing words as the smallest elements of the lexicon. References
- Arad, Maya (2003). Locality constraints on interpretation: the case of Hebrew denominal verbs.
Natural Language and Linguistic Theory 21: 737-778.
- Arad, Maya (2005). Roots and Patterns: Hebrew Morpho-Syntax. Dordrecht: Kluwer.
- Embick, David, and Rolf Noyer (2005). Distributed Morphology and the Syntax/Morphology
Interface. The Oxford Handbook of Linguistic Interfaces, Oxford University Press.
- Ussishkin, Adam (2006). Semitic Mophology: Root-based or word-based. Morphology, volume 16,
Number 1 / July 2006. Springer Netherlands.
Download