Factors Regulating English Secondary Stress
Young-Ah Do
(Seoul national university)
Do, Young-Ah, 2006, Factors regulating English Secondary Stress.SNU. Phonology2. This paper analyses the factors that are relevant in determining English secondary stress.From generative grammar to Optimality Theory, literature on secondary stress has concerned mainly about stem stress preservation and avoidance stress clash. Recently, there suggested other view which concern the frequency of the words. By examining the secondary stress patterns in English words from Longman Pronunciation Dictionary (Wells 1990), however, it can be demonstrated that mere phonological constraint or actual usage cannot not explain secondary stress pattern alone. The aim of this paper is to suggest the importance of lexically- determined parsing system and the meaning of prefix in regulating English secondary stress placement.
Keywords: secondary stress, frequency, vowel reduction, lexical-specific structure, rhythmic constraints
- Introduction
English stress patterns are so complicated that there seem to be no complete account that is able to give a perfect explanation to them. Among such complexities, the existence of secondary stress is crucial. Since English words can have more than one level of stress, i.e. sàtisfáction,, each word necessarily has one primary stressed syllable and may have one or more secondary stressed syllables.
Secondary stress is generally treated together with primary stress, because the two can be assigned by similar rules or constraints. (Wensky2004, p.11) Most works which have been so far regard English secondary stress as the result of interaction of two forces: stem-stress preservation and avoidance of stress clash. In the course of making such apparatuses, they mainly deal with the word with suffixes.[1]By contrast, prefixes are not discussed frequently.
This paper examines some other factors that have correlation with secondary stressing by relying on the mono-morphemic items and words with prefix. (e.g. dis- a-, mis-). The aim of this discussion is to show that in spite of a significant role of the weight and structure of syllable in determining the place of secondary stress, items can have variants regardless of them, and lexically pre-determined parsing system and the meaning of prefix cannot be underestimated in regulating the pattern.
As the borderline between prefix and compound-initial are not clear-cut and both types comprise bound morphemes (Fudge1984 139), the list of prefix in Fudge will be accepted as the starting point of overall classification.
- Previous studies
2.1 Fudge (1984)
The work done by Fudge is about the influence of word-initial bound morphemes(prefix and compound-initial) on the secondary stressing of English words. As his analysis is based upon the examination of a vast amount of English words, he gives very thorough and precise description of these data rather than providing a formal model for the stress-system of English. He arranges word-initial bound morphemes into subgroups according to their influence on the stressing of words.
(1) Factors that are examined by Fudge (138-192)
1) whether the morpheme in question is capable of carrying primary word stress
e.g. isóchronous vs. ìntracéllular
2) whether it is attached to free stems
e.g. unwanted vs. apostolic
3) whether the morpheme has a constant meaning
e.g. unearth vs. confine
4) whether the final consonant of the morpheme is lost if it is attached to a stem starting with the same consonant
e.g. unnatural() vs. connect ()
5) whether the final vowel of the morpheme is long
e.g. homotaxis vs. homo:gonous
On the basis upon the result of these factors, he classifies prefix into two groups; stress-neutral and stress repellent. (Fudge 165) Both cases roughly show complementary distribution in there behaviors.
2.2. Burzio (1994)
Burzio(1994) incorporates Fudge’s quantitative finding and develops a constraint based theory which can provide a certain string of stress pattern in words with suffixation. Two of his constraints that are related to the discussion of this paper and one of the lexically determined foot structures are below.
(2) 1) a. #(ΦL → #L( (Burzio 101)
b. #H →(ΦH)
2) Metrical alignment (Burzio 149)
*(σ…H…)
Where “…”: includes no foot boundaries
3) lexical representation which includes right foot boundaries.(Burzio 227)
e.g. …a)ble ,icΦ)
With these kinds of constraints and pre-determined foot structure, his theory predicts that though there exists preferable stress pattern, normally more than one acceptable parsing is possible. The list of allowed patterns for a string generally contains one that is most preferred. For prefix, however, he did not give any account.
2.3. Pater (2000)
Pater(2000) indicates that lexically-specific constraints should be proposed to explain lexical-specific English secondary stress. Upon this assumption, he divides English words in two groups(S1, S2) and gives each group different kinds and rankings of constraints. He finally argues that English secondary stress shows non-uniformity.
This paper will show the inadequacy of Fudge’s grouping by showing that lexically-determined structure of prefix play an important role in placing the secondary stress. Burzio’s constraints based approach is more effective, in that the theory permits the violation of constraints and it can give satisfactory account for the variants of parsing system. However, the result from Longman pronunciation dictionary (Wells 1990) will suggest some modification to his constraints. Pre-determined foot structure of him will be accepted in case of prefix too. Pater’s lexically- specific constraints will be evaluated not adequate for making generalization of the pattern, since he relies too much on lexical specific factors. The examination in this paper will be able to give some further steps before moving into lexical-specific constraints for regulation.
3. Statistical results
Statistical data will be based upon quantitative analyses of Longman Pronunciation Dictionary (henceforth: Longman). For this purpose, this chapter presents the analysis of 271 words that are primary stressed on their fourth syllable. The idea of selecting this group is on the basis of Nora Wensky’s previous study[2] (2004). Wensky analyzed 737 words but did not give further modification on these raw data. To incorporate quantitative data into the relation of secondary stress pattern and prefix, I perform classification and analysis of the original data from Wensky. And 14 cases excluded from Wensky are found from Longman and included in my analysis.
<Table 1> The size of data in Longman Pronunciation Dictionary
primary stressed on 4th syllable / analyzed data among 751 itemstokens / 751 / 271(36.1%)
3.1. Syllable weight and secondary stress
There is general consensus that it is unmarked or natural for heavy syllables to be parsed and have a stress on it. As given in (2)-1) above, Burzio suggested that unparsed heavy syllable in word initial position tends to be parsed. He takes this idea as one of his constraints that regulate the stress pattern. The data from Longman, however, show that this constraint is violated in copious cases. Although candidates with some violation can be the output in constraint-based theory, pattern 1 which includes about half of the data that has unparsed heavy syllable gives significant doubt whether this constraint can operate properly.
<Table 2> Pattern 1 #σ( σ̀σ )( σ́
first three syllables / Tokens1.LLL / 37
2. HLL / 31
3.HHL / 26
4.HnLL / 19
5.LHL / 18
6.HnHL / 13
7.HnHnL / 7
8.HnHHn / 1
9.HLHn / 1
10. HHnHn / 1
Total / 149
The distribution of initial unparsed syllable is: 55% of light syllables and 45% of heavy ones. Among heavy syllable Hn -CVC ending with s or sonorous consonant-(definition from Burzio 110) holds24.8% and true Heavy(heavy except Hn) occupies 75.2%. Even If Hn is classified as a light syllable, as Burzio did, the proportion of truly heavy syllable is still high. Indeed, 15 items that have long vowel in initial syllable but have no secondary stress were found from Longman.
(3) unstressed long vowel in initial syllable(15 items)
co:habite co:habitation de:fibrillation de:population de:regulation de:salination de:segregation de:toxication de:decasylabl pre:destination pre:farication pre:meditation re:decoration re:forestation tri:nitrotoluene
In spite of evidences above, the fact that initial stressed syllable are all parsed and get the secondary stress gives power to the tendency of parsing heavy syllable. The tokens by the type of first three syllables are summarized in <Table 3> below.
<Table 3> Pattern 2 #(σ̀̀̀ )(σ̀σ )(σ́
first three syllables / Tokens1.HnHn Hn / 4
2. HLL / 4
3.HHnL / 4
4.HnHnL / 3
5.HnHL / 3
6.HnLL / 2
7.HnHHn / 1
8.HLHn / 1
9.HHL / 1
Total / 23
All of the data in <Table 3> contain initial heavy syllables. It is noticeable that word initial heavy syllable are all parsed and get the secondary stress, even though the stressing creates Head clash.
The probability of violation of constraint makes it reluctant to fully agree with the argument above. Even if data from <Table 2> do not sufficiently support my argument, there still exists the evidence that can be suggested for this string of idea. Some words in Longman have one or more ‘variants’ in regard with the secondary stress. I classify them into three cases on the basis upon the pattern they show.
<Table 4> Pattern 2 and 1
#σ( σ̀σ )( σ́ / #(σ̀̀̀ )(σ̀σ )(σ́
first three syllables / Tokens1.HnLL / 20
2. HLL / 15
3.HnHnL / 13
4.HnHL / 9
5.LLL / 4
6.HHL / 4
7.LHL / 3
8.HLH / 3
9.LLH / 2
10.LHnL / 2
11.HnLHn / 2
12.HnLHn / 2
13.HHnHn / 2
14.HnHHn / 1
15.HHnH / 1
total / 65
Type of word initial syllable L:11(17.1%) H:54(82.9%)
<Table 5> Pattern 3 and 1
#( σ̀σσ )( σ́ / #σ( σ̀σ )( σ́
first three syllables / Tokens1.HnLL / 10
2. HLL / 7
3.LLL / 6
4.HnHL / 4
5.LHnL / 1
6.LHL / 1
7.HnHnHn / 1
8.HLH / 1
9.HHL / 1
10.LHnL / 2
11.HnLHn / 2
12.HnLHn / 2
13.HHnHn / 2
14.HnHHn / 1
15.HHNH / 1
Total / 25
Type of word-initial syllable L:10 (40%) H:15(60%)
<Table 6> Pattern 3, 2 and 1
#( σ̀ )(σ́σ)( σ́ / #σ( σ̀σ )( σ́ #( σ̀σσ )( σ́
first three syllables / Tokens1.HnLL / 3
2.HnLHn / 1
3.HLL / 1
4.HHL / 1
Total / 5
Type of word initial syllable H:5 (100%)
Variants in parsing and second stressing system present a couple of remarkable insights of English second stressing. First of all, it is worth noticing that the second stress pattern has it’s variants in 95cases among 271. Approximately one third of the words mined in this paper take it’s variants and it seems to be hard to discover idiosyncratic character among them. The result makes us doubt the possibility of regulation the pattern of secondary stress pattern on the basis upon syllable structure. It will be continuallydiscussed in examination below.
It can be roughly seen in every three classes that word initial heavy syllables are more likely to have it’s variant forms than light ones. Heavy syllable marked more than double the amount in sum of three tables as it’s initial morpheme. This might partly indicate the uncertainty or ambiguity of parsing heavy syllable. Surprisingly light syllable does not appear at all in the pattern that has two variant forms, as summarized in <Table 5> above. This can be another evidence to the assertion that most of prefix with heavy syllable structure is not likely to have pre-determined parsing or foot system at the level of it’s lexicon.
3.2. Other factors about syllable and secondary stressing
There is much literature concerning the syllable-related factors in deciding placement of secondary stress. As alluded in 3.1. above, however, syllable structure itself cannot give full explanation to secondary stress pattern. Together with the weight, the tendency of parsing syllable (Parse σ) and avoiding the succession of head (*Clash-Head) cannot play critical roles in determining the secondary stress in themselves. To assert this claim, the patterns of variants from <Table 4> to <Table 6> are again quoted in (4)
(4) 1) <Table 4> Pattern 2 and 1 #σ( σ̀σ )( σ́ / #(σ̀̀̀ )(σ̀σ )(σ́
2) <Table 5> Pattern 3 and 1 #( σ̀σσ )( σ́ / #σ( σ̀σ )( σ́
3) <Table 6> Pattern 3, 2 and 1 #( σ̀ )(σ́σ)( σ́ / #σ( σ̀σ)( σ́ #( σ̀σσ )( σ́
Variants in (4) –1) indicate that whether to parse syllable and give secondary stress is dependent not primarily on structure or weight of the syllable. Rather, the parsing system can be varied regardless of it’s syllable structure. This argument seems supported also by (4)-2) and 3) in that they vary not only in secondary stress but in the way of parsing. Though there can be shown some features which are common in (4), such as Foot Type Troche, other constraints- especially Parse Syllable and *Clash Head- cannot operate by themselves but have to be consolidated with other factors that will be discussed in remaining part of this paper.
Let us now turn to Stress Preservation, which has been considered for the secondary stressing in the volume of studies that have been so far. The analysis of words that have no variants proved that Stress Preservation is the main factor in deciding the place of secondary stress in derived forms with prefix. Among 172 items in <Table 2> and <Table 3>, which do not have variant form, Stress Preservation is overridden only in 5 cases. Wensky examined larger size of data than mine (Wensky 133) and the result also supports the idea that Stress Preservation is reponsible for the place of secondary stresses in the overwhelming majority of derived words. Wensky’s analysis shows that Stress Preservation was sometimes overridden by the preference of the ternary foot ( σLσ) over the parsing σ(Lσ), but this only occurred in 17 words compared to the almost 300 words , in which stress preservation was not violated.
4. Lexically determined prefix
By those observations so far, the need of regulating prefixes which are lexical-specific is alluded above. In 4.1., one case which was traditionally considered having obligatory secondary stress regardless of it’s environment will be tested. Then the result will justify that all the prefixes in this group are not lexically determined ones. On the other hand, in 4.2. two prefixes will be suggested as pre-determined prefix. As this analysis considers the items in limited aspects and the amount of data is not large enough to ensure this fact, the discussion will just stay in experimental level. Much more researches have to done to test the propriety of my suggestion.
4.1. Not pre-determined prefix
Let us first look at a group of prefixes which is not pre-determined, contrary to Fudge’s assumption. Fudge(165) thought that some prefixes carry obligatory secondary stress, so the prefix is stressed even if secondary stress is not required by other principle. And he gives some example of lexical entries.
<Table 7> Autostressed prefixes. (Fudge 165)
prefix / Prefix / prefix / prefix / prefixa-/an-(neg) / mis-(wrongly) / ex-(formerly) / re-(again) / mal-(badly)
example / Example / example / example / example
amoral / mis-spell / ex-husband / reappear / malfunction
Longman gives some counterexamples to his analysis. Below are items which hasve so called autostressed prefixes. Some words always keep secondary stress on an autostressed prefix but there are 11 cases which lack stress on it.
(5) always stressed on an autostressed prefix (10 items)
misapprehend misapprehension miscalculation nonlcohilic noninterferencenonintervention redistribution reeducation selfabnegation selfpreservation
(6) lack stress on an autostressed prefix(11items)
achondroplasia achondroplastic misrepresent redecoration reduplication reforestation regeneration regurgitation rejuvenation rejuvenescene resuscitation
Examples in (6) suggest that all of so called autostressed prefix do not have lexically determined underlying structure.
4.2 Pre-determined prefix
4.2.1 Prefix ‘un-‘ vs. ‘in-‘
Wensky argues that ‘un-‘ is pre-determined prefix by following cluster split property of Burzio. Data from Longmen confirms his claim on prefix ‘un-‘. It’s lexically determined character will be even more justified when prefix ‘un-‘ is compared to prefix ‘in-‘, which has the similarmeaning with ‘un-‘ but are not lexically marked before taking parsing or footing.
The behaviors that prefix ‘un-‘ shows are below.
(7) ‘n’ in ‘un-‘ never form a onset of following syllable, even if the immediate following syllable bears a primary stress.
e.g. ùn.ái.de.d, ùn.éa.se – VC.V
(8) no degemination occurs
e.g. unnatural = un-natural
One of the properties of ‘un-‘ can become more clear if it is compared with prefix ‘in-‘ Prefix ‘in-‘ become identical with the following consonant. Fudge explain this phenomenon as assimilation.(Fudge 180) As the result of assimilation, originally two consonants , one in the coda of prefix and the other in the onset of stem, become a single consonant.(e.g. irrelevant) That possibility of occurring degemination varies from case to case and it is different from the property of ‘un-‘ in (7) and(8) This can be an evidence that ‘in-‘should be excluded in lexically-determined prefix and ‘un-‘ should be marked as pre-determined prefix.
4.2.2. Primary stressed prefixes
A minor group of prefixes appears in nouns that have a verbal counterpart and the two are stressed differently. e.g. ábstract (noun)~ àbstráct(verb)(Wensky, 92) Since the word bears the same structure in both noun and verb form, no constraints or principles can work to explain the different stress pattern between the two. In this point of view, I claim that this group should have fully marked stress system in their lexicon originally.
5. Meaning of prefix
The last factor to consider is not within the scope of phonology. In this discussion, the meaning of the prefix itself and whether it maintains its constant meaning after affixation is relevant. Two groups of words are classified in <Table 8> and <Table 9>, by whether they get the secondary stress or not. When the prefix has ambiguous meaning in the word, the full word in mined at Oxford English Dictionary.
<Table 8> #(σ̀̀̀ )(σ̀σ )(σ́
type of prefix / frequencya- (neg) / 1
de-(get rid of) / 1
dis-(neg) / 2
mis-(wrongly) / 3
non-(neg.) / 3
pre-(formerly) / 3
re-(again) / 2
self- ( / 2
trans-( / 1
un-(neg.) / 4
total / 22
<Table 9> #σ( σ̀σ )( σ́
type of prefix / frequency / type of prefix / frequencya- / 6 / hy- / 1
ac- / 5 / il- / 1
ad- / 1 / im- / 1
af- / 1 / in- / 9
ag- / 1 / i- / 1
al- / 1 / ma- / 2
an- / 1 / mo- / 1
ap- / 2 / o- / 2
ar- / 1 / of- / 1
as- / 3 / pe- / 1
au- / 2 / per- / 3
ca- / 1 / po- / 1
co- / 1 / pre- / 2
col- / 1 / pro- / 4
com- / 3 / re- / 7
cor- / 2 / ref- / 1
de- / 9 / res- / 2
di- / 2 / so- / 2
dis- / 5 / sub- / 1
do- / 1 / sug- / 1
ec- / 2 / sus- / 1
e- / 20 / Syl- / 1
en- / 1 / ve- / L
ex- / 7 / vi- / 1
fe- / 1 / vo- / 1
ges- / 1 / total / 149
ha- / 2
hal- / 1
When the prefix does not bear a secondary stress, almost any kinds of prefixes appear, as demonstrated in <Table 9>. By contrary, the prefix which has no variants and always gets the secondary stress consist of so called stress-neutral prefix (Fudge 165) Fudge define stress-neutral prefix as a group of prefixes which is attached o free stems and have constant meaning.(Fudge 169) The attempt to compare <Table 8>, which has secondary stress at prefix with <Table 10> reveals that the prefixes that has tendency to bear secondary stress has clear meaning and that is also constant in the word it is included.