toward a prosodic-phrasing algorithm for tiberian...
TRANSCRIPT
0
Toward a Prosodic-Phrasing Algorithm for Tiberian Hebrew Vincent DeCaen
draft 9 (June 2011)1
Abstract
Algorithm
BINMAX >> PAUSE >> WRAP, LONGWD >> ALIGN >> BINMIN >> SYNTAX
1
0. Introduction: Generative Masoretics
0.0. The remarkable isomorphy of Biblical Hebrew (BH) syntactic structure (DeCaen 1995)
and the syntax of the Tiberian logogenic chant (Price 1990, 1996) cries out for a formal
explanation. The seminal study by Dresher (1994) explains this striking isomorphy by means of
an abstract, intermediate prosodic representation2 negotiated at the syntax-phonology interface.
0.1. Dresher thereby convincingly explains (a) how and why the chant deviates from BH
syntax; (b) the otherwise bizarre function of the accents in directly demarcating the domains in
which Tiberian Hebrew (TH) phonological rules operate; and (c) the compression and expansion
phenomena of the accent systems in terms of the tempo and grain of a natural, spoken language.
0.2. Dresher’s prosodic theory has tremendous explanatory power pace Churchyard (1999).
The scope of the theory can be extended and expanded into an ambitious programme in
generative metrics which I call Generative Masoretics (figure 1).
0.3. From the many phonological and musical transformations can be deduced a detailed
theory of TH metrical phonology (DeCaen 2008): a further interface intervening between the
syntax and the music. The resulting metrical grids can be leveraged into a metrical theory of BH
accentual-syllabic poetry (DeCaen 2009, 2011).
0.4. The implied complex boundary tones and down-stepping of highs and lows of the
prosodic representation naturally project intonational contours: the inherent music of the spoken
language. The parade of alternating highs and lows of these intonational contours is undoubtedly
the source of the logogenic chant (DeCaen 1999, see also Weil 1995).
2
x x x x
x x
x
Figure 1: Generative Metrics
2. Generative Masoretics and Computation
2.0. The prospect of robust, detailed theories of TH prosodic and metrical representation
opens out on a computational-linguistic vista. From arbitrary syntactic3 inputs, it should be
possible to algorithmically project a prosodic representation. From the prosodic structure, it
should be possible to algorithmically generate metrical grids. And from metrical grids it should
be possible to algorithmically project the musical grids of the two accent systems via musical
transformations.
2.1. The heart of such a computational project is the demarcation of phonological phrases
(φ), aligning the edges of φ with syntactic edges based on ranked preferences and constraints.
The optimality-theoretic (OT),4 node-correspondence-based Align/Wrap Theory of Selkirk
(2000, 2005) supplies the required constraint-based and edge-based mapping.
Prosody
Rhythm
Phonology Syntax
Poetry Music
3
3. Prosodic Phrasing of Simplified Inputs
3.0. It is not reasonable to try to draft a complex OT prosodic-phrasing algorithm for long
verses at one go. A reasonable strategy is to simplify inputs as much as possible to identify the
rudiments of such an algorithm.
3.1. One considerable simplification is to restrict inputs to intonational phrases and not whole
verses. It appears that aligning intonational phrases trumps aligning phonological phrases, as
would otherwise be expected. The nature and distribution of intonational phrases, as a first
approximation, is given in DeCaen (2011b).
3.2. Another simplification is to restrict inputs to the short, regular intonational phrases of the
simplest type of BH poetry. The short, end-stopped lines of Psalms 111-112 (DeCaen 2009,
2011a) and Job 3 (approximately 100 lines) offer a good place to start.
4. Continuous Dichotomy in OT Perspective
4.0. The essential feature that unites BH syntax and the syntax of the accent systems is
continuous dichotomy or top-down binary branching. Such continuous dichotomy can be
imposed by an absolute ranking of binary constraints (1) as in (2).
(1) BINARY MAXIMUM = BINMAX
Any prosodic phrase at any level of the prosodic hierarchy is
maximally binary-branching.
BINARY MINIMUM = BINMIN
Any prosodic phrase at any level of the prosodic hierarchy is
minimally binary-branching.
(2) BINMAX >> BINMIN
4
4.1. The absolute ranking of BINMAX ensures that no node in the prosodic representation is
ternary-branching. Notice that a phonological phrase (φ) consisting of one phonological word
(ω) incurs a penalty under BINMIN. Violations of this constraint should only arise, all things
being equal, where the input is odd-numbered.
4.2. N.B. Recursive phrasing is permitted to achieve conformity with BINMAX.
5. Two-Word Problem and “Long Words”
5.0. Continuous dichotomy imposed by (2) entails that two phonological words (ω) will
necessarily be phrased together as one phrase (φ). However, such is not always the case:
prosodic length can trump BINMIN. Consequently, a “long word” (3) must intervene as in (4).
(3) PROSODIC “LONG WORD” = LONGWD
A “long” phonological word (ω) must be aligned with its own phonological phrase (φ).
(4) BINMAX >> LONGWD >> BINMIN
5.1. N.B. The detailed definition of “metrical long word” or metrical dipod (DeCaen
2008), sufficient to explain the Tiberian musical transformations—or rather, lack thereof—is
NOT necessarily the same as “prosodic long word” that is revealed here in the prosodic-phrasing
algorithm for the poetic system of accentuation. More anon.
5.2. Job 3:21b is parsed first without the LONGWD constraint in (5), and then with the
intervening LONGWD constraint in (6). (As will become clear below, both words are “prosodic
long words”.)
(5) Job 3:21b wáyyaḥpǝrūhû mimmaṭmônîm
ω ω BINMAX BINMIN
(ω ω)
((ω) (ω)) *!*
5
(6) Job 3:21b wáyyaḥpǝrūhû mimmaṭmônîm
ω ω BINMAX LONGWD BINMIN
(ω ω) *!*
((ω) (ω)) **
5.3. The other instance of a two-word phrase in the limited database is Job 3:22a.
However, this token is rendered ambiguous by the obscuring musical transformation that is
triggered by metrical monopod ’ĕlê—gîl. (Nevertheless, as shown below, haśśǝmēḥîm is not a
“prosodic long word” according to the relevant distinction in §7.2.)
6. Three-Word Problem and Syntactic Edges
6.0. Continuous dichotomy does not resolve the three-word problem in (7), setting to one
side the odd long word. Because of the odd-numbered input, there is necessarily some violation
of BINMIN.
(7) Three-Word Problem
ω ω ω BINMAX BINMIN
(ω ω ω) *!
((ω ω) (ω)) *
((ω) (ω ω)) *
((ω) (ω) (ω)) *! ***
6.1. In such cases, the principal observation is that the prosodic phrasing is sensitive to
morphosyntactic “right edges”: ALIGNXP (8). Representative examples are provided in (9) and
(10).
(8) ALIGN R (XP, φ) = ALIGNXP
Align the right edge of an XP in syntactic (PF) representation with the right edge of
a phonological phrase (φ) in phonological (PR) representation.
6
(9) Ps 111:1b [N [N]NP]NP [N]NP
ω ω ω BINMAX BINMIN ALIGNXP
(ω ω ω) *! *
((ω ω) (ω)) *
((ω) (ω ω)) * *!
((ω) (ω) (ω)) *! ***
(10) Ps 111:2a [A]AP [N [N]NP]NP
ω ω ω BINMAX BINMIN ALIGNXP
(ω ω ω) *! *
((ω ω) (ω)) * *!
((ω) (ω ω)) *
((ω) (ω) (ω)) *! ***
6.2. However, in rare cases, it appears that the verb V is enforcing a syntactic edge
against expectations. Inspection shows that the simple application of ALIGNXP may be trumped
by higher-order syntactic considerations. The edge of a clause (CP) imposes a greater degree of
disjunction than the typical syntactic edge, corresponding in the isomorphic prosodic
representation to the right edge of an intonational phrase (I): ALIGNCP (11). Conjunction of
constituents must also be observed: CONJ (12). Representative cases are offered respectively in
(13)5 and (14).
6
(11) ALIGN R (CP, IP) = ALIGNCP
Align the right edge of any clause CP in syntactic (PF) representation with the right edge
of some intonational phrase IP in phonological (PR) representation.
(12) CONJUNCTION OF XPS = CONJ
The conjunction of XPs must be respected.
Conjunction is left-recursive (N.B. Ps 112:4b).
7
(13) Ps 112:10a [[A]AP V]CP [V]CP
ω ω ω ALIGNCP ALIGNXP
((ω ω) (ω)) *
((ω) (ω ω)) *
(14) Job 3:5a V [[N]NP & [N]NP]NP
ω ω ω CONJ ALIGNXP
((ω ω) (ω)) *
((ω) (ω ω)) *
6.3. Constraints such as (11) and (12) will multiply rapidly. The significant generalization
that is being missed is that constituents of the larger XP must be kept together. Following
Selkirk, a “wrap” constraint is introduced in (15).
(15) WRAPXP = WRAP
A phrase (XP) in syntactic (PF) representation (specifically, its terminal string)
must be contained within a prosodic phrase of the corresponding level (Selkirk 2005)
in phonological representation.
6.4. It should further be noted that there are many false friends in the application of
ALIGNXP: morphologically “free” versus “bound” nominal forms do not necessarily signal
edges. To take a simple example, there is no right edge between a noun and its following XP
complement (AP, e.g., Ps 111:10b; PP, e.g., Ps 112:2b): by rule, N' (read “N bar”) does not
impose a prosodic right edge (15).
(15) NP
N' XP
N
8
6.5. Thus, apparent exceptions to ALIGNXP are analyzed so that there is no right edge,
because there is no intervening XP. Two examples involving the complements of the participle
(Prt), and one the complements of adjective (A), are given in (16).
(16) Ps 111:8a [[[Prt]Prt' PP]Prt' PP]PrtP
Job 3:15b [[[Prt]Prt' NP]Prt' NP]PrtP
Job 3:19b [N]NP [[A]A' PP]AP
6.6. There are rare cases where ALIGNXP is unable to decide among candidates (18). In
such cases, the final arbiter (default) is the SYNTAX (17).
(17) SYNTAX
Prosodic phrasing is isomorphic with the syntactic structure.
(18) Job 3:18a [[Adv]AdvP [[N]NP V]TP]CP
ω ω ω ALIGNXP SYNTAX
((ω ω) (ω)) * *!
((ω) (ω ω)) *
7. Prosodic “Long” Word Again
7.0. Ex hypothesi, the only way a line can escape the constraints outlined above is if a
poetic “prosodic long word” is present. Only one three-word instance is found in the restricted
corpus: the PP lǝniplǝ’ōtāyw is “long” ex hypothesi in Ps 111:4a (19).
9
(19) Ps 111:4a NP V PP
ω ω ω BINMAX LONGWD BINMIN ALIGNXP SYNTAX
((ω ω) (ω)) * * *
((ω) (ω ω)) *! *
7.1. In this particular case, there is an exact minimum pair: NP V PP in Ps 111:5a. Even
more interesting, the PP in the latter case lîrē’āyw is also a TH “metrical long word” (metrical
dipod). In other words, from a TH metrical point of view, the two PPs are metrically identical
(20)!
(20) x x
(x x) (x x)
(x x) (x x) (x x) (x x)
lǝ nip lǝ ’ō tāyw lî rē ’āyw
7.2. However, the PPs are by no means identical from an accentual-syllabic point of
view. The accentual-syllabic parse employed in DeCaen (2011a) is projected downwards in (21).
The “prosodic long word” lǝniplǝ’ōtāyw has a complete group or foot; whereas the defective
word lîrē’āyw has an incomplete group or degenerate foot. Thus, the ambiguous case in Job
3:22a (22), noted above at §5.3, is disambiguated: it is not a “prosodic long word”, despite being
a “metrical long word”.
(21) x x
(x x) (x x)
(x x) (x x) (x x) (x x)
lǝ nip lǝ ’ō tāyw lî rē ’āyw
(* *) * (* *) * *) (* *) *
)* *) )* *)
* *
10
(22) x
(x x)
(x x) (x x)
haś śǝ mē ḥîm
*) * (* *) *
)* *)
*
7.3. The constraint LONGWD is accordingly modified as in (23).
(23) PROSODIC “LONG WORD” = LONGWD
A phonological word (ω) is “long” if, with reference to its accentual-syllabic parse,
its secondary foot is “complete”.
A “long” phonological word (ω) must be aligned with its own phonological phrase (φ).
8. Four-Word Problem
8.0. Continuous dichotomy will always select—absent a “prosodic long word”—the
candidate ((ω ω) (ω ω)) without some intervening constraint(s), as can be seen from the pruned
tableau in (24). One way to think of BINMIN is as a prosodic default among a block of lowest-
ranking defaults, including the proposed SYNTAX.
(24)
ω ω ω ω BINMAX BINMIN
((ω ω) (ω ω))
((ω ω ω) (ω)) *! *
(((ω ω) (ω)) (ω)) *!*
(((ω) (ω ω)) (ω)) *!*
((ω) (ω ω ω)) *! *
((ω) ((ω ω) (ω))) *!*
((ω) ((ω) (ω ω))) *!*
(ω ω ω ω) *!
((ω) (ω) (ω) (ω)) *! ****
8.1. Sitting immediately above the “defaults” must be the basic alignment constraints.
Following DeCaen (19xx: paper on autolexical NP), a block of alignment constraints ALIGN (25)
11
is proposed. ALIGNXP (26) was already introduced above in (8). ALIGNX (27) has the effect of
treating syntactic heads as prosodic clitics. The simple case of right recursion in (28) is
consequently mirrored in the right-recursive nesting of phonological phrases in (29).
(25) ALIGN = ALIGNXP, ALIGNX
(26) ALIGN R (XP, φ) = ALIGNXP
Align the right edge of an XP in syntactic (PF) representation with the right edge
of a phonological phrase (φ) in phonological (PR) representation.
(27) ALIGN L (X, φ) = ALIGNX
Align the head X of an XP in syntactic (PF) representation with the left edge
of a phonological phrase (φ) in phonological (PR) representation.
(28) NP
N NP
N NP
N NP
N
(29) [N [N [N [N]NP]NP]NP]NP
ω ω ω ω ALIGNX BINMIN
((ω ω) (ω ω)) *!
(((ω ω) (ω)) (ω)) *! **
(((ω) (ω ω)) (ω)) *! **
((ω) ((ω ω) (ω))) *! **
((ω) ((ω) (ω ω))) **
12
8.2. Examples involving ALIGNXP and ALIGNX are given in (30) and (31) respectively.
(30) Ps 112:4a V PP NP PP
ω ω ω ω ALIGNXP BINMIN
((ω ω) (ω ω)) *!
(((ω ω) (ω)) (ω)) **
(((ω) (ω ω)) (ω)) *! **
((ω) ((ω ω) (ω))) *! **
((ω) ((ω) (ω ω))) *! **
(31) Ps 112:8b P C V PP
ω ω ω ω ALIGNX BINMIN
((ω ω) (ω ω)) *!
(((ω ω) (ω)) (ω)) *! **
(((ω) (ω ω)) (ω)) *! **
((ω) ((ω ω) (ω))) *! **
((ω) ((ω) (ω ω))) **
8.3. The phrasing will ensure that arguments and clauses are not split over prosodic
phrases, but rather there is some basic integrity. Ensuring top-down integrity falls to the block of
“wrap” constraints in (15), where WRAP >> ALIGN. As noted, this “wrap” function subsumes a
myriad of preferences such as ALIGNCP and CONJ. An example with two XPs that must be
“wrapped” is given in (32).
(32) Ps 111:7a [N NP]NP [NP [& NP]]&P
ω ω ω ω WRAPXP ALIGN BINMIN
((ω ω) (ω ω)) *
(((ω ω) (ω)) (ω)) *! **
(((ω) (ω ω)) (ω)) *!* * **
((ω) ((ω ω) (ω))) *!* * **
((ω) ((ω) (ω ω))) *! * **
8.4. The most important top-down constraint is TH “pause” or prosodic disjunction. TH
pause is triggered by a number of syntactic structures subsumed under Comma Phrase (Selkirk
2005). As proposed in DeCaen (2011b), there are three ranked constraints (26)-(29) that can
13
taken as a block PAUSE (33), only dominated by BINMAX (1), but crucially adding here a default
that aligns a Comma Phrase with a phonological phrase (37). The example of Job 3:3b is given in
(38).
(33) PAUSE = ALIGNCOMMAP2 >> BINMIN(I) >> ALIGNCOMMAP1
>> ALIGNCOMMAP(φ)
(34) ALIGN (COMMAP1, I) = ALIGNCOMMAP1
Align the edges of a CommaP1 in syntactic (PF) representation with the edges
of a in phonological (PR) representation.
(35) ALIGN (COMMAP2, I) = ALIGNCOMMAP2
Align the edges of a CommaP2 in syntactic (PF) representation with the edges
of a corresponding intonational phrase in phonological (PR) representation.
(36) BINARY MINIMUM (I) = BINMIN(I)
An intonational phrase (I) must consist of at least two phonological phrases (φ).
(37) ALIGN R (COMMA PHRASE, φ) = ALIGNCOMMAP(φ)
Align the right edge of an XP in syntactic (PF) representation with the right edge of
a phonological phrase (φ) in phonological (PR) representation.
(38) Job 3:3b [NP V’amar]CommaP [ [V NP]CP]CommaP
ω ω ω ω PAUSE WRAP ALIGN BINMIN
((ω ω) (ω ω)) **
(((ω ω) (ω)) (ω)) *! *** **
(((ω) (ω ω)) (ω)) *! * * **
((ω) ((ω ω) (ω))) *! * * **
((ω) ((ω) (ω ω))) *! **
14
Bibliography
Archangeli, D., and D. Terence Langendoen. 1997. Optimality Theory: An Overview. Explaining
Linguistics, no. 1. Oxford: Blackwell.
Chiesa, Bruno. 1979. The Emergence of Hebrew Biblical Pointing: The Indirect Sources.
Judentum und Welt, no. 1. Frankfurt a.M.: Peter Lang.
Churchyard, Henry. 1999. “Topics in Tiberian Biblical Hebrew Metrical Phonology and
Prosodics”. Ph.D. diss., University of Texas at Austin.
Cohen, Miles B. 1969. The System of Accentuation in the Hebrew Bible. [Ph.D. diss., University
of Minnesota.] Minneapolis: Milco Press.
DeCaen, Vincent. 1995. “On the Placement and Interpretation of the Verb in Standard Biblical
Hebrew Prose.” Ph.D. diss., University of Toronto.
DeCaen, Vincent. 1999. “The Linguistic and Musical Reality behind the Tiberian Hebrew
Accents”. Symposium paper, University of Toronto, Jewish Studies Faculty Symposia, 2
February.
DeCaen, Vincent. 2004. “The Pausal Phrase in Tiberian Aramaic and the Reflexes of *i”.
Journal of Semitic Studies 49.2: 215-224.
DeCaen, Vincent. 2005. “On the Distribution of Major and Minor Pause in Tiberian Hebrew in
the Light of the Variants of the Second Person Independent Pronouns.” Journal of
Semitic Studies 50.2: 321-327.
DeCaen, Vincent. 2008. “On the Formal Definition of “Long Word” in Tiberian Hebrew:
Diagnostic, Database, Generative Analysis, and Implications”. IOMS/SBL paper
presented 24 Nov 2008, revised Feb 2009.
http://homes.chass.utoronto.ca/~decaen/papers/
DeCaen, Vincent. 2009. “Theme and Variation in Psalm 111: Phrase and Foot in Generative-
metrical Perspective”. Journal of Semitic Studies 54.1: 81-109.
DeCaen, Vincent. 2011a. “The Dark Matter of Hebrew Poetry: A Generative-Metrical Analysis
of the Biblical Tetrameter”. Ms.
DeCaen, Vincent. 2011b. “On the Intonational Phrase in Tiberian Hebrew: The Nature and
Distribution of Pausal Phonology”. Ms.
Dresher, B. Elan. 1994. “The Prosodic Basis of the Tiberian Hebrew System of Accents”.
Language 70.1: 1-52.
Goerwitz, Richard L. 1993. “Tiberian Hebrew Pausal Forms.” PhD diss., University of Chicago.
Kager, René. 1999. Optimality Theory. Cambridge Textbooks in Linguistics. Cambridge:
Cambridge University Press.
Kelley, Page H., Daniel S. Mynatt, Timothy G. Crawford. 1998. The Masorah of Biblica
Hebraica Stuttgartensia: Introduction and Annotated Glossary. Grand Rapids MI:
Eerdmans.
Khan, Geoffrey. 1987. “Vowel Length and Syllable Structure in the Tiberian Tradition of
Biblical Hebrew”. Journal of Semitic Studies 32.1: 23-82.
Khan, Geoffrey. 1995. “The Pronunciation of resh in the Tiberian Tradition of Biblical Hebrew”.
Hebrew Union College Annual 66: 67-80.
Khan, Geoffrey. 1996. “The Tiberian Pronunciation Tradition of Biblical Hebrew”. Zeitschrift
für Althebraistik 9.1: 1-23.
Mynatt, Daniel S. 1994. The Sub Loco Notes in the Torah of Biblia Hebraica Stuttgartensia.
BIBAL [Berkeley Institute of Biblical Archaeology & Literature] Dissertation Series, no.
2. North Richland Hills TX: BIBAL Press.
15
Price, James D. 1990. The Syntax of Masoretic Accents in the Hebrew Bible. Studies in the Bible
and Early Christianity, no. 27. Lewiston NY: Edwin Mellen.
Price, James D. 1996. Concordance of the Hebrew Accents in the Hebrew Bible. 5 vols. Studies
in the Bible and Early Christianity, no. 34A-E. Lewiston NY: Edwin Mellen.
Prince, Alan Sanford. 1975. “The Phonology and Morphology of Tiberian Hebrew”. Ph.D. diss.,
Massachusetts Institute of Technology.
Prince, Alan, and Paul Smolensky. 2004. Optimality Theory: Constraint Interaction in
Generative Grammar. Oxford: Blackwell.
Revell, E. J. 1980. “Pausal Forms in Biblical Hebrew: Their Function, Origin and Significance”.
Journal of Semitic Studies 25.2: 165-179.
Revell, E. J. 1981. “Pausal Forms and the Structure of Biblical Poetry”. Vetus Testamentum 31.2:
186-199.
Revell, E. J. 1987. Nesiga (Retraction of Word Stress) in Tiberian Hebrew. Textos y studios
“Cardenal Cisneros”, no. 39. Madrid: Instituto de Filología, Departemento de Filología
Bíblica y de Oriente Antiguo.
Revell, E. J. 1988. “Conjunctive Dagesh: A Preliminary Study”. Pp. 95-101 in VIII International
Congress of the International Organization for Masoretic Studies, Chicago 1988.
Masoretic Studies, no. 6. Edited by E. J. Revell. Scholars Press.
Revell, E. J. 1989. “Deḥiq: Exceptions to the Standard Pattern”. Pp. 85-99 in Proceedings of the
Ninth Congress of the International Organization for Masoretic Studies, 1989. Masoretic
Studies, no. 7. Edited by Aron Dotan. Scholars Press.
Selkirk, Elisabeth. 2005. “Comments on Intonational Phrasing in English”. In Prosodies:
Selected Papers from the Phonetics and Phonology in Iberia Conference, 2003.
Phonetics and Phonology. Edited by S. Frota, M. Vigario, M. and J. Freitas. Mouton de
Gruyter. Ms.
Selkirk, Elisabeth. 2000. “The Interaction of Constraints on Prosodic Phrasing”. Pp. 231-261 in
Prosody: Theory and Experiment: Studies Presented to Gösta Bruce. Text, Speech and
Language Technology, no. 14. Edited by Merle Horne. Dordrecht: Kluwer.
Trask, R. L. 1997. The Penguin Guide to Punctuation. London: Penguin.
Weil, Daniel Meir. 1995. The Masoretic Chant of the Bible. Jerusalem: Magnus.
Widawski, Lea. 1990. “The Paseq in the Hebrew Bible: Occurrences in Medieval Manuscripts,
Characteristics and Relation to the Accentuation System” [Hebrew]. Ph.D. dissertation,
Bar-Ilan University.
16
End Notes 1 Acknowledgements…
2 To be clear, the accentuation is not itself a prosodic representation. This might be the mistaken impression from
reading, e.g., Churchyard (1999: 222-252). The way the accentuation varies relative to the syntax simply suggests that it is sensitive to an abstract intermediate syntax that Dresher (1994) identifies as prosodic. We know independently that the accentuation has its own rhyme and reason, and has its own constraints and transformations. There are thus several intermediate “levels” separating the syntax from the music. 3 [make note of Kirk Lowery’s work on the syntax parser, update, prospects.] 4 On Optimality Theory, start with Prince & Smolensky (2004); then see Archangeli & Langendoen (1997) and Kager (1999). 5 See also Ps 112:9a, 10b; Job 3:11b, 21a.
6 See also 111:3a, 4b, 8b, 9c.