LINGUIST List 13.134

Mon Jan 21 2002

Sum: Head Prominence and Floating Tones

Editor for this issue: Marie Klopfenstein <>


  1. Mike Cahill, head prominence and floating tones

Message 1: head prominence and floating tones

Date: Fri, 18 Jan 2002 13:58:11 -0500
From: Mike Cahill <>
Subject: head prominence and floating tones

Summary: head prominence and floating tones

The original questions were:
1) "Floating tones" are abundant in African languages, and
have been documented in a number of Mexican languages as
well. Does anyone know of their existence in other
languages of the world?
2) It seems intuitively obvious that heads of phrases
should be more phonetically prominent than non-heads.
But does anyone know of cases or actual studies where
this has been shown? Or counterexamples?

This summary, a bit long, I'm afraid, is organized as
- Why I sent the queries
- Responses about floating tones (4)
- Responses about head prominence (12)

I conflate a summary of these because they are related.
What prompted my query was the behavior of tonal
associative morphemes in African languages. In the 31
languages for which I have data, the associative
(possessive, genitive) construction (e.g. 'horse's leg")
consists of two adjacent nouns, with a specific tone,
usually High, added to one. Underlyingly, it is Noun1 +
High tone + Noun2. (The High tone is a remnant of an
associative morpheme which had a segmental as well as a
tonal component.) The head noun in some languages is Noun1;
in other languages it is Noun2. Now in these 31 languages,
this High tone morpheme sometimes shows up on Noun1,
sometimes on Noun2. Where Noun1 is head, the tone may
appear on either Noun1 (13 languages) or Noun2 (11
languages). Where Noun2 is head, the tone appears only on
Noun2 (8 languages). (One language has two patterns, so the
counts add to 32.) What you never find is that the tone
appears on Noun1 if Noun1 is the dependent. So two factors
emerge. The tone either goes to the noun that's the head,
or it goes rightward. It does not go left to a dependent

When I presented this at the 1998 LSA, I talked of two
constraints: TONE-Right and HEAD-Prominence. My present
queries are in pursuit of more information about these
patterns. I have received relatively few responses this
time about floating tones, but I previously had good
support for TONE-Right. However, the responses the past few
days about the phonetic prominence of heads showed that
heads are generally NOT prominent, and thus I need another
explanation for why the associative tone is attracted to
head nouns. I speculate a bit on this below. Now on to the


I admit I wasn't specific enough on this query. My
reference to "floating tones" was aimed at morphemes
consisting solely of tone, or else a tone connected
with one word that is nonetheless pronounced on an adjacent
word. Some responses also had this in mind, others didn't.
I thank all who responded, and since these are interesting
in their own right, I summarize them below (most I've
condensed to the essentials).

Cory Sheedy:
Tsuut'ina (also known as Sarcee, Sarsi). There is evidence
for a "floating" High tone that triggers regressive upstep
and indicates 2nd person singular. This upstep tone is
morphemic but it's a little more complicated than that. It
only surfaces once the segmental 2SG.SUBJ is deleted by
another process. What happens is it raises the preceding
tone by one level (low-to-mid or mid-to-high).

Mark Donohue:
They're found in at least the Skou family in New Guinea,
where they have L for past tense on verbs, overwriting the
lexical tone.
 The possessive suffixes are formed from the free
pronouns with a falling tone imposed on them, except for
3sg.m, which has a high tone.
 And the dative is made with a change of vowel to [e],
and a low tone. This reflects an historical pattern.
 I think that Kevin Ford has written, in Language and
Linguistics in Melanesia, about floating tones in the
Goroka languages in PNG. I think, though am not sure, that
the same sort of thing happens in Lani. Oh, and the Lakes
Plains languages in Irian have tone marking tense/aspect.

John Coontz:
Omaha-Ponca has a system of pitch accent. The accent is
normally structured as a HL pattern across the word, with
the H being assigned to the first or second syllable and
applying to any preceding syllable as well. Subsequent
syllables are L. I've noticed enough oddities to think
this is at best a first approximation of a description.
Anyway, if a word is a monosyllable or disyllable,
pronounced in isolation, then the L manifests on the end of
the last H syllable, producing a falling pitch. I think
the H is fixed to the accented syllable, and the L floats
to whatever follows.

Paul Boersma:
Limburgian has a change of tone if one goes from a
disyllable to a monosyllable, at least if the intervening
consonant is voiced, e.g. dru:HveL 'grapes' ~ dru:HLf
'grape'. At least this is a possible analysis. Underlying
forms could be dru:HveLn ~ dru:HveL, respectively, with
dropping of final e and n afterwards, and moving of the L
of the dropping e to the new final syllable.


Though there were a few positive responses, most gave
examples where the head of a phrase is NOT prominent. (This
had been obvious to me about PPs in English from the
start.) In terms of the tonal associative morphemes, then,
head-prominence is almost certainly not the factor that
attracts the High tone to the head. More promising may be
head-markedness, and I will probably explore that next. (I
welcome correspondence on this.) I do thank all who wrote
in; I list their responses below, with some condensing.
It's a step toward truth to know when you're mistaken!

David Odden:
I know of counterexamples, in the form of rules which
"downgrade" the prosodic properties of the phrasal head,
and nothing else. The classical case is Kimatuumbi:
"Shorten long vowels in the head of a phrase". There are a
number of phrasal rules in languages which shorten vowels,
delete high tones, and perhaps other things. Shortening and
de-toning "work against" phonetic prominence; then, if
there were such a thing as a desire for the head of a
phrase to be prominent, you would expect to find a language
where phrasal heads *resist* some kind of sandhi. But that
does not happen: rather, phrasal heads are specifically
sought out as targets of degradation, and as far as I know
are *never* specifically immune to degradation.

Richard Coates:
It's not true, is it, that heads are always phonetically
prominent (if you mean 'more prominent than other material
in the same phrase'), even in English? Only in marked
circumstances is P most prominent in PP, V in VP or D in

Mark Liberman:
A counterexample: The verb is usually considered to be the
head of the VP, but it is also usually less prosodically
prominent than an object NP. In the case of an SO language
like English, you could take this to be because the object
is later and therefore pragmatically "newer". But I believe
that the same thing is roughly true in Japanese, where
verbs normally follow objects.

The usually cited factors in prominence are things like
1. new >> old
2. later >> earlier
3. referential >> anaphoric/cataphoric
4. noun >> verb
5. contentful >> functional
There is some literature on what kind of "newness" is
relevant (some and not others apparently). When things are
equally new (e.g. a string of unrepeated unknown digits)
later things tend to be more prominent. Pronoun-like things
are usually low in prominence regardless of order. Verbs
tend to be low in prominence relative to nouns, other
things equal. Functional or grammatical formatives tend
to be weak.

Bill Poser:
Yes, I think that Mark is right about Japanese. A typical
SOV sentence in Japanese will have just two minor phrases:
[S] [OV], with the result that the second phrase,
containing the [OV], will have at most a single pitch
accent. And since the rule for resolution of pitch accent
conflicts at the phrase level is "leftmost wins", if the
object is accented its accent survives and that on the
verb, if any, is deleted.
Another set of counterexamples would come from PPs.
Presumably the head of a PP is a P, but adpositions are to
my knowledge always little dinky things that are not
prominent. The same would be true of DPs, if it is right
that what we used to think were NPs are really headed by
the determiners.

Doug Whalen:
Cecile Fougeron and Pat Keating have been looking at
prosodic strengthening, in which sounds at the onsets of
prosodic domains are longer and more "forceful" (have
larger movements, etc.). This may be what you want. One
reference is below, and it will cite others.
FOUGERON, CECILE. 2001. Articulatory properties of
initial segments in several prosodic constituents in
French. Journal of Phonetics, 29.109-35.

Philip Carr:
Yes, one *would* expect heads of phrases to be
perceptually/phonetically more prominent. But if one
subscribes to the 'Determiner Phrase' analysis of NPs in
English, then at least some determiners are going to be
less prominent, qua 'heads', than the noun. PPs in English
raise an interesting question: if the P is head, as has
been claimed, then they are not normally more salient than
the nouns which follow them.

Peter T. Daniels:
Blackbird. I would expect the opposite, because mightn't a
head be more likely to be a topic, and a non-head a
comment? (as in the compound noun offered, which may or may
not qualify as a phrase!)

Michael Johnstone:
I'm not sure, but I think the language I was thinking of
that stresses prepositions is Czech.

Taylor Roberts:
One possible counterexample can be seen in Pashto and I
expect in other languages with second-position clitics. A
clause-initial PP containing a 'full' NP (e.g., 'from the
man') is 'heavy' enough to host a second-position clitic.
However, if the clause-initial PP instead contains a
'lighter' NP, such as a pronoun (e.g., 'from him'), the PP
will not be heavy enough to host the clitic, and the clitic
appears further to the right (wherever it finds a heavy-
enough host to its left). As the preposition (or
postposition) is the head of the PP, its merely being the
head of that phrase is obviously not sufficient for it to
be prosodically prominent. Your generalization, though, is
surely more accurate for 'lexical' categories such as NP
and VP. For info on Pashto, see Taylor Roberts, 'The
optimal second position in Pashto,' 367-401 in Geert Booij
and Jeroen van de Weijer (eds.), _Phonology in Progress -
Progress in Phonology: HIL Phonology Papers III_ (The
Hague: Holland Academic Graphics).

Daniel Loehr:
In terms of possible counterexamples, Cruttenden (1997), on
p. 86 and thereabouts, has a few examples of the
intonational nucleus on function words. For example,

I put my bag in your study because there was nowhere
else TO put it.
Please make sure you bring all your belongings WITH

Not being a syntactician, I can't speculate on whether TO
and WITH in the above examples are heads (of the
infinitival clause, and of the prepositional clause,
respectively). Cruttenden, Alan (1997) Intonation, 2nd ed.
Cambridge University Press.

Paul Boersma:
Whether the language is object-verb or verb-object, it
seems to be the object that is stressed in at least 90% of
the cases.

Alex Monaghan:
an interesting notion, but i don't see why it's intuitive
that heads of (syntactic? semantic?) phrases should be
prominent. firstly, if the same information were encoded at
every linguistic level that would be massively redundant -
even more redundant than language already is. secondly,
since syn/sem heads are not generally marked at other
levels of linguistic structure (morphology,
syllabification, etc.), why should they be marked in
prosody? thirdly, there is a great deal of evidence - all
the way back to SPE - that syntax and prosody are not
generally congruent, so syntactic heads will not generally
be prosodic heads.

i can think of two excellent reasons why syntactic/semantic
heads will not coincide with prosodic heads:
1 - one major function of prosody is to encode the phrase-
level structure of speech. to this end, many languages have
regularly right-headed or left-headed prosody. french is a
classic example. this makes the phrases easy to identify,
but makes it difficult for syn/sem heads to coincide with
prosodic heads unless the former are similarly constrained.
2 - another major function of prosody is to distinguish
between supposedly "new" and "given" information. unless
syn/sem heads are always presented as "new", they will not
be prosodic heads: thus, in an english sentence such as
"John bought a red jeep, and Mary bought a yellow jeep.",
it would be extremely odd to put particular prominence on
either occurrence of "jeep" because it is the "given" part
of the meaning.

for a longer discussion of the factors determining prosodic
prominence, see my article "What Determines Accentuation?"
in Journal of Pragmatics 19 (1993) pp.559-584 (an online
version is available at

Mike Cahill
Summer Institute of Linguistics
(now SIL International)
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue