PART I: THE BACKGROUND
2. LITERATURE REVIEW
2.7 Halle (1998)
2.7 Halle (1998)
The stress-system of Halle (1998)(H98) is based on ordered rules, and stress levels are represented by a bracketed grid, as in HV. Here, however, foot boundaries are marked by unmatched parentheses, following Idsardi (1992). Stress-bearers are syllables, which may be marked extrametrical at the right edge of a domain by Edge-marking (cf. (76) below).
Consonants at the right edge may also be invisible to the stress rules, but this is not stated by H98 at all, but his derivations seem to suggest this (on this issue see a more detailed discussion in Wenszky, 1999).
Derivation is cyclic, but this approach does not recognise stress-preservation—which is one of the greatest drawbacks of this system—, and thus all structure is erased before a new cycle is started. The cycle is only needed because other rules of phonology also work in cycles and they might need information provided by stress rules on previous passes. It is assumed that cyclic rules are applied every time a cyclic affix is added to the stem. Non-cyclic rules, however, only pass through the word once, when all the affixes have been added. Each constituent of a word is marked for cyclicity ([±cyclic]), and only the relevant rules work on them (H98: 554, cf.
HV: 79–81). If a non-cyclic affix attaches to the stem, stresses remain untouched (e.g. in expréssion®expréssionless).
At the beginning of a new cycle, after the erasure of all stresses and structure, derivation starts again, as if the word—even if it is a derived item—were monomorphemic. Each syllable head (as a stress-bearer) is assigned a line 0 asterisk, and then some morphemes (e.g.-ure, -y (H98: 557)) are marked unstressable. An unstressable syllable (maximally one syllable per word) always appears at the right edge of the syntactic domain and is represented by a dot on the grid.
The next step in the derivation is to mark edges: the Edge-marking rules (76) select the boundary of a foot on line 0 by inserting unmatched parentheses into the grid. The two rules are disjunctively ordered, i.e. if (76a) can apply, (76b) is blocked, if (76a) cannot apply, (76b) will come into play. There are several lexical exceptions to the rules in (76), such as the majority of verbs and unsuffixed adjectives, which are exempt from both kinds of edge-marking.
(76)Edge-marking rules(H98: 549) (76a) RLR Edge-marking
Æ ®] / * ___ * ## line 0 Condition J: Final * projects short vowel.
(76b) LLR Edge-marking Æ ®[ / * ___ * ## line 0
RLR Edge-marking (76a), which inserts a Right parenthesis to the Left of the Rightmost syllable (hence the name), partly does the work of rules traditionally referred to as rules of extrametricality. When an asterisk is followed by a right parenthesis or preceded by a left parenthesis without intervening parentheses, it will belong to a foot. The two kinds of boundary marks (i.e. brackets [ ] inserted by Edge-marking and parentheses ( ) inserted by the MSR) have
2.7 Halle (1998) 58 2. Literature review
the same effect, they are just differentiated by H98 to make reading of grids easier. If an asterisk—and thus the syllable head and the syllable—remains unparsed it will not receive any stress in the cyclic stratum. If such an unparsed asterisk represents the final syllable of the word, the MSR (77) will not see it, since its computation starts at the boundary inserted before this asterisk. This corresponds to the case of syllable extrametricality, which is needed for most nouns in English.
LLR Edge-marking (76b), as it inserts a left parenthesis, creates a foot at the end of the word, but its head will only be marked by the Main Stress Rule (77c). This applies right after Edge-marking and marks the syllable which will carry the primary stress of the word. The MSR (77) is different from the English Stress Rule of LP (p. 278) and the MSR of HV (p. 228) in that it is not iterative, therefore it does not create superfluous secondary stresses. This way, conflation is successfully avoided, which is a great advantage, since no effort is wasted to create metrical structure that is erased later. However, as mentioned above, the cyclic application of rules does exactly this—it creates stresses that are erased at the beginning of the next cycle.
The MSR (77) below actually arises by collapsing two very similar rules into one. In one case the MSR starts from an already existing boundary (marked by P in the rules), and in the other from the end of the word (##). The meaning of < > is that if there is an edge-marking boundary in the grid, the MSR (77) will count from the boundary, not from the end. The word boundary functions as a starting point only when there are no edge-marks in the sequence, i.e.
in words that are exempt from both types of edge-marking. The MSR works from right to left.
(77b) only applies if (77a) cannot. (77a) skips a syllable ending in V(C) and another one before a boundary or the end of the word, and places a left parenthesis before these two skipped syllables. (77b) skips only one syllable and places the boundary there. This is actually the Weak Retraction pattern of LP, which LP only apply for secondary stresses. The third part of the rule, (77c), puts an asterisk above the leftmost asterisk of the foot created by LLR Edge-marking (76b) or MSR itself ((77a–b). Naturally, all words go through the MSR, because every English word must have a primary stressed syllable.
(77)Main Stress Rule(MSR) (H98: 549)
a.Æ ®( / ___ * * <P *> ## line 0 Condition K: Second * projects vowel in light rime.
b.Æ ®( / ___ * <P *> ## line 0 c. Line 0 heads are leftmost.
The P stands for a boundary of either kind: ] or [.
The rules discussed up to this point select the most prominent syllable of the whole word and post-tonic secondary stresses. At this point these are both represented by a line 1 asterisk. The work of these rules is illustrated in (78) below.
2. Literature review 59 2.7 Halle (1998)
(78)The work of Edge-marking and the MSR(based on H98: 548–550) (78a) RLR Edge-marking
MSR *
* * * * RLR Edge-m. * * *] * a, c * (* *] *
A me ri ca ® A me ri ca ® A me ri ca
MSR *
* * * RLR Edge-m. * *] * b, c * (*] *
a gen da ® a gen da ® a gen da
(78b) LLR Edge-marking
MSR * *
* * * LLR Edge-m. * * [* a, c (* * [*
ma la chite ® ma la chite ® ma la chite
MSR * *
* * * LLR Edge-m. * (* [* a, c * (* [*
sta lag mite ® sta lag mite ® sta lag mite
(78c) Exception to Edge-marking
MSR a, c *
* * * No Edge-m. * (* *
de ve lop ® de ve lop
MSR b, c *
* * No Edge-m. * (*
u surp ® u surp
At this point all stressed syllables that are marked have an equally high column of asterisks. Later, in the output of stress rules, however, secondary stresses will be represented by a lower column of asterisks than that of the primary stress, i.e. if line 0 marks stress-bearers, at least two other lines are needed to represent a word with three levels of stress. Secondary stresses occur either before or after the primary stress, and these cases are dealt with by two different rules in H98. Let us examine post-tonic secondary stresses first.
Post-tonic secondary stresses emerge as the result of LLR Edge-marking (76b) and the Rhythm Rule (79). RR is actually an edge-marking rule on line 1, which inserts a left parenthesis before the first asterisk on line 1. Since only those words have two asterisks on line 1 that have undergone LLR, only these will show the effect of (79). Here the primary stress will be on the first foot-head rather than on the second one, i.e. this is the way to derive post-tonic secondary stresses. In other cases, the grid gets one extra level, but the primary stress will automatically be
2.7 Halle (1998) 60 2. Literature review
on the syllable marked by the MSR.9The unbounded foot that emerges is left-headed, similarly to line 0 feet.
(79)Rhythm Rule(H98: 550)
a.Æ ®( / ## ___ * line 1 LLL b. Line 1 heads are leftmost.
As far as pre-tonic secondary stresses are concerned, H98 does not give a detailed account, though a great number of English words have these. The rule that is responsible for them is based on Halle—Kenstowicz (1991) and is stated in H98 as follows (80).
(80)Iterative Foot Construction (IFC)(H98: 565)
Construct binary feet by inserting right parentheses iteratively from left to right.
As a result of this rule, the asterisks that remain before the main stress on line 0 are mechanically arranged into binary feet, in a way that every odd-numbered syllable will get secondary stress. Though not stated formally in the article, these feet should also be left-headed. Since the Rhythm Rule (79) is a cyclic rule, while IFC (80) is a non-cyclic one that always follows the cyclic rules, the line 1 grid-marks constructed by (80) will never carry the primary stress (IFC counterfeeds RR). Furthermore, even numbered syllables can never carry secondary stress, which is contrary to the facts, as our examples will show.
Let us derive our example words now. As already noted, the real stress pattern of a word emerges in the last cycle, because all structure is erased at the beginning of every cycle.
This is why only the last cycle is shown in the derivations below. The stress rules of H98 can only derive one pattern foracademician(81), because IFC can only promote odd-numbered syllables to secondary stress. If-ianis one syllable, the word must be an exception to edge-marking, because otherwise stress would fall on-de-.If, however,-ianis disyllabic (as in all previous derivations), the word undergoes RLR Edge-marking. The result is the same. Since there is only one syllable on line 1, the Rhythm Rule cannot retract stress, it only builds one more line.
9Whether the application of the Rhythm Rule in words with one asterisk on line 1 is necessary is a theoretical question.
If the minimal sufficient grid is aimed at, the Rhythm Rule should not work in these cases. If primary stresses of two different words should have equally high grids (so that they would be comparable more easily), the work of the Rhythm Rule is indispensable in every case.
2. Literature review 61 2.7 Halle (1998)
(81)àcademícian ~ acàdemícian (81a)àcademícian
* *
MSR a, c * RR (* IFC * (*
RLR * * * (* *] * * * * (* *] * * *) * (* *] *
® a ca de mi ci an ® a ca de mi ci an ® a ca de mi ci an
(81b)acàdemícian—underivable
If the last syllable ofdissimilarity(82) is marked unstressable (82a.i), because it is a word-final-y,the word must be exempt from Edge-marking. If the final syllable is visible to the stress-rules, RLR Edge-marking works (82a.ii). Subsequently, the Rhythm Rule adds another line and IFC places secondary stress on the first syllable. The other variants cannot be derived because IFC cannot place stress on an even-numbered syllable and adjacent stresses cannot be derived either.
(82)dìssimilárity ~ dissìmilárity ~ dìssìmilárity (82a)dìssimilárity
(82a.i) Unstressables
no Edge-marking *
MSR a, c * * * (* * .
® dis si mi lar i ty
(82a.ii)
RLR *
MSR a, c * * * (* *] *
® dis si mi lar i ty
* *
RR (* IFC * (*
* * * (* * . * *) * (* * .
® dis si mi lar i ty ® dis si mi lar i ty
(82b)dissìmilárity—underivable (82c)dìssìmilárity—underivable
Though H98’s system has serious problems deriving the patterns of the first two example words, all four variants ofemanatorycan be derived in his system, because he has special machinery to do that (H98: 561–563). There is a special rule (83), which makes-at-in
2.7 Halle (1998) 62 2. Literature review
the ending-atoryunstressable. This is highly exceptional, since normally only the last syllable of a domain can be unstressable.
(83)-atoryshortening (optional)(H98: 562)
In-at-orythe suffix -at- is shortened. In addition, -at- becomes unstressable if the preceding syllable ends with a light rime and in certain lexically marked cases.
Another special device is the optional rule (84) that deletes a boundary before the endings-ary/-ory.
(84)-ory/-aryreduction(H98: 558)10 (®Ø / (* ___ * ## line 0
| +o/ary
In the first variant ofemanatory(85a) the whole ending-oryis rendered unstressable. It is exceptional to treat two syllables as extrametrical. From here the derivation proceeds as normal: LLR Edge-marking puts a mark before-ate-, and the MSR and the Rhythm Rule derives the pattern needed. The word in (85b) is regular: only the last syllable is unstressable, and derivation proceeds as normal. In (85c)-atoryshortening makes the syllable-at-unstressable in the middle of the word, and due to RLR Edge-marking there will be no post-tonic secondary stress.-atoryshortening also works in (85d), but here LLR Edge-marking induces secondary stress on the ending-ory, while main stress will be on the first syllable, similarly to (85c).
(85)émanàtory ~ èmanátory ~ émanatory ~ émanatòry (85a)émanàtory
Unstressabless
MSR a, c * *
LLR * * [* . . (* * [* . .
® e ma nat o ry ® e ma nat o ry
*
RR (* *
(* * [* . .
® e ma nat o ry
10The seemingly similar ending-erydoes not belong here, because it does not induce the same stress patterns. F84 does not mention this ending. Wells (p. 251) says that this “stress-neutral suffix is used only after a strong-vowelled syllable (machínery); the variant-ryis used otherwise (déntistry).”
2. Literature review 63 2.7 Halle (1998)
(85b)èmanátory
Unstressables
MSR b, c *
RLR * * *] * . * * (*] * .
® e ma nat o ry ® e ma nat o ry
* *
RR (* IFC * (*
* * (*] * . * *) (*] * .
® e ma nat o ry ® e ma nat o ry
(85c)émanatory
Unstressables
-atoryshortening MSR a, c *
RLR * * .] * . (* * .] * .
® e ma nat o ry ® e ma nat o ry
*
RR (*
(* * .] * .
® e ma nat o ry
(85d)émanatòry
Unstressables *
-atoryshortening MSR a, c * * RR (* *
LLR * * . [* . (* * . [* . (* * . [* .
® e ma nat o ry ® e ma nat o ry ® e ma nat o ry
H98’s rules cannot derive pre-tonic secondary stresses in several cases, and variation is only possible in words ending in-atory(and-ative,which is treated in a similar manner)(cf.
Section 9),though in reality this is not the only class of words that display variation in stress patterns. The treatment of the-atoryclass needs special machinery and is not in line with the rest of the rules. Furthermore, as this system does not recognise the preservation of stresses, a lot of superfluous derivation is done and information produced in earlier cycles is lost.