Arnold Zwicky zwicky at CSLI.STANFORD.EDU
Fri Nov 2 18:59:35 UTC 2001

there are several, at least partially distinct, phenomena at issue

one is the stressing of latinate-prefix (con-, per-, etc.) verbs
(which show some tendency to be root-stressed) and related nouns
(which show some tendency to be prefix-stressed) and verbs
zero-derived from those nouns (which tend to preserve the stress
pattern of the nouns).  there's a considerable literature on the
subject.  it's known that there is variation at every point in these
general patterns and that some of this variation is associated with
region, age, etc., though often in an item-specific way.

another phenomenon is variation between afterstress and forestress in
a collection of specific words that do not involve latinate prefixes
and are not compounds: cement, police, tennessee, etc.  i believe
there is also a fair literature on this front-shifting (as it would be
seen from the viewpoint of standard varieties outside the american
south).  these variants are strongly associated with region, but there
is also considerable variation on which items are affected; it's
certainly not an all-or-nothing matter.

then there's an alternation between afterstress and forestress, for
certain words, within the speech of individuals, depending on the
prosodic context.  many people have, for instance
  thirTEEN reJECtions
  THIRteen NO votes.
the alternation avoids successive syllables with primary stress, and
promotes alternating stress within phrases.  there is again a
considerable literature on the subject; search for "Rhythm Rule"
in works on theoretical phonology.

fourth, there is the placement of accent within phrases for the
purposes of emphasis or contrast:
  I don't want to DISplace them; I want to REplace them.
this strategic accent placement overrides the stress patterns
associated with lexical items.

finally, there are differences between speakers on the stressing of
some noun-noun compounds: ICE cream vs. ice CREAM, CHERRY pie vs.
cherry PIE etc.  (here i give stress patterns in isolation; these
compounds may be subject to the rhythm rule - someone who has "ice
CREAM" in isolation or before unstressed syllables might well have
"ICE cream CONE" and "Ice cream SANDwich" - and are certainly
subject to strategic accent placement - someone who usually has
"cherry PIE" will have "I want CHERRY pie, not APPLE").

there is a certain amount of literature on forestress vs. afterstress
in compounds; much of this is summarized in my paper "Forestress
and Afterstress", unfortunately available only in the Ohio State
University Working Papers in Linguistics 32.46-62 (1986).  [it was
intended for a festschrift for dwight bolinger that never came to be.]

"ICE cream" vs. "ICE cream" is mentioned already in bloomfield's
Language (1933:228), where bloomfield takes the position that
forestressed "ICE cream" (which he spells "ice-cream") is a compound,
while afterstressed "ice CREAM" (which he spells "ice cream") is a
phrase, "although there is no denotative difference of meaning" (nor,
i should point out, any difference in internal syntax).  bloomfield
tried to stick resolutely to his definitions, wherever they led him.

in my paper, i argue that there are noun-noun compounds (in some
cases, whole classes of them) with afterstress.  for instance,
there's a contrast for many speakers between the afterstressed
pattern of proper names like
and the forestressed pattern of otherwise parallel proper names
with "street" as the second element:

some of these *patterns* differ from dialect to dialect, and in
addition, a fair number of *individual compounds* (like "ice cream"
and "cherry pie", the latter noted in lees's (1960) discussion of the
matter) have different stressings for different speakers.  i'm not
sure to what extent these item-specific stressings are associated with
region, rurality, age, class, etc., though i suspect that the
associations are weak.  (for what it's worth, like many of the other
northeast u.s. speakers on this list, i'm an "ICE cream" speaker.  but
i'm also a "cherry PIE" speaker.)  weak associations with social
factors (including region) wouldn't be surprising, since pretty much
everyone has some afterstressed compounds, in addition to the default
forestressed ones, and everyone has alternations in stress patterns
due to the rhythm rule and strategic accent placement, so that
language learners are likely to have both models available to them.

not all variation is closely tied to social factors, after all.  when
there are two variants available, and both are well represented in the
speech of a community, and the basis for choice is not entirely
transparent (especially if there are competing patterns), learners are
likely to pick a variant on the basis of which one they experience (or
notice) first, especially for very frequent items.  in this way,
item-specific speaker-specific variants can persist for quite some

not that i'm *excluding* the possibility of a strong geographical
association for, say, the stressing of "ice cream".  i just wouldn't
be surprised (or dismayed) if it turned out there wasn't one.

arnold (zwicky at

More information about the Ads-l mailing list