Auditory Event Related Potentids to Phonologid and Semantic Violations in
Sentence Endings in Preschool Age Children
by
Shannon Eiizabeth MacLean
Submitted in partial fiilfiUment of the requirements for the degree of
Master of Science in Audiology
at Daihousie University
Halifax, Nova Scotia
Apd, 1998.
O Copyright by Shannon Elizabeth MacLean, 1998
Acquisitions and
Bibliogmphic Services
Acquisitions et
senrices bibliographiques
The author bas granted a nonexclusive licence allowing the
National Libmy of Caaada to
reproduce, loan, distribute or sell
copies of this thesis in microform,
paper or electronic formats.
The author retains ownership of the
copyright in this thesis. Neither the
thesis nor substantid extracts fkom it
may be printed or otherwise
reproduced without the author's
permission.
L'auteur a accordé une licence non
exclusive permettant à la
Bibliothèque nationale du Canada de
reproduire, prêter, distriiuer ou
vendre des copies de cette thèse sous
la forme de micr&che/nlm, de
reproduction sur papier ou sur format
électronique.
L'auteur conserve la propriété du
droit d'auteur qui protège cette thèse.
Ni la Wse ni des extraits substantiels
de celle-ci ne doivent être imprimés
ou autrement reproduits sans son
autorisation.
For Keith.
Table of Contents
ListofIllustrationsandTables ...........................................
vi
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
...
Acknowledgements .................................................. w u
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
AuditoryStimuii . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Produres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Resuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Endnotes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
AppendOr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
List o f IlIwtrations and Tables
Table 1. Summary of the four sentence conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Figure 1. Scheme ofRecording Sites...................................... 11
Table 2. Recording Parameters .......................................... 13
Table 3 . Response Accuracy and Reaction Times for Sentence Condition .......... 16
Figue 2. Boxplot of Readon Tvne by Sentence Condition. . . . . . . . . . . . . . . . . . . . . 17
Figure 3 . Boxplot o f Percentage Correct by Sentence Condition. . . . . . . . . . . . . . . . . 17
Figure 4.(a) Grand average wavdorms c o m p a ~ PM-SM
g
to PMM-SM .......... 19
Figure 4. (b) Grand average wavefonns wrnparing PM-SM to PM-SMU . . . . . . . . . 20
Figure 4. (c) Grand average waveforms comparing PM-SMto PMM-SMM. . . . . . . . 21
Table 4. Mean peak latencies and amplitudes for the PMN. . . . . . . . . . . . . . . . . . . . . . 22
Table 5 . Mean peak latencies and amplitudes for the N400. . . . . . . . . . . . . . . . . . . . . . 22
Figure 5 . (a) Amplitude of PMN by Sentence Condition. . . . . . . . . . . . . . . . . . . . . . . . 23
Figure 5 . (b) Latency o f PMN by Sentence Condition. . . . . . . . . . . . . . . . . . . . . . . . . . 23
Figure 6. (a) Amphtude of N400 by Sentence Condition. . . . . . . . . . . . . . . . . . . . . . . . 24
Figure 6. (b) Latency of N400 by Sentence Condition. . . . . . . . . . . . . . . . . . . . . . . . . . 24
Abmact
Event-related ptentials (ERP), reaaion t h e (RT) and response accuacy measures were
obtained duruig a semantic ciasdication of tenninal words of spoken sentences in nine
preschool age girls with hearing sensitivity and receptive one word vocabulary within
normal k t s . The sentences were constnicted to yield four conditions in which the
terminal word: 1) met both phonologicai and semantic expectancies (e-g., Giraffi have
long necks); 2) met initiai phonological expectancies only and violated sernantic
expectancies (e-g., GrafEes have longs nets); 3) violated phonological expectancy but met
a lower order semantic expectancy (e.g., Giraffes have long legs); and 4) violated both
phonologicai or sernantic expctmcies (e.g., Girâffes have long pots). The ERP latency of
the phonologicai mismatch negatvity (PMN), a response associateci with phonological
processing, was significantly longer in response to sentences where the initial phoneme
was expected but the semantic expectancy was violated. However, this longer processing
did not impacted upon the RT of the behavioral response for this condition. S i g n i f i d y
low aca>racy and long RTs were found in behavioral responses to sentences in which
sernantic expectancies were met but were les expected (low Cloze probability). The
findings are discussed with reference to phonologicai and semantic processing and
prereading abiiity, as weli as with reference to the sensitivity of ERP and behavioral
response to violations with semantic and phonological processùig.
The author 6rst thanks the childien and theu caregivers for taking the time to participate
in this study. Without their kind assistance this study wodd not have been possible. Very
special thanks are extended to Dr. Joseph Byme for his untë*ng enthusiasq optimism,
and support for this project. Thanks are also extended to hs.John Comoly and
Christine Sloan for thei.assistance in stimuli prepamtion and data interpretation Special
thanks also to Dr. Edward Yang and Melanie Campbeii for their assistance in manuscript
p r e p d o n and thanks to Gordon Whitehead and Michel Comeau for th& assistance with
equipment cali'bration. Also thanks to Heather Mason and Christine Santilii for helping to
locate necessary equipment. The author gratefully acknowledges the assistance of TrÏcia
Beattie in preparing the data for analysis. This work was supporteci by the IWK- Grace
Student Research Scholarship and Dalhousie Graduate Scholarship.
Introduction
At present the literature shows an increasing interest in explorhg the possible links
between the processes of speech perception and phonologid abiiity (Lovrich, Cheng, &
Velting, 1996; Tdai, 1980; Watson & Miller, 1993; Werker & Pegg, 1992). Knowledge
of the phonological structure of language, that is, the knowledge that spoken words
wnsist of smaller sound units, is considerd fundamental in learning to rad. For example,
Stuart and Coltheart (1988) found that phonologicai awareness in preschool children was
related to their subsequent progress during their first year of learning to read and
cuncluded that young children could make use of grapheme-phoneme correspondences
nom the outset if they had adequate phonological awareness. Studies on n o m d y
developing children have shown that the ability to analyze spoken words into smaller units
and performance on standardized reading tests were highiy correlateci (Wagner &
Torgesen, 1987; Watson & Miller, 1993) and that phonological awareness assessed during
kindergarten was the most signiticant predictor of word identification and s p e h g skilis in
school-age children (MacDonald & Comwaii, 1995). Evidence for the importance of
phonological awareness in iiteracy acquisition also cornes fi-omstudies of children with
readiig and spelling difliculties. In rnany cases these chiidren often perform poorly on
tests of phonological awareness (Bird, Bishop, & Freeman, 1995).
One possible explanation for why some children with readiing difnculties perform
poorly on tests of phonological awareness is that these children, while able to discriminate
between sound segments in isolation, t&il to identify or to be aware of these segments
within the intemal structure of spoken words p i r d & Bishop, 1992; Bir& Bishop &
Freeman, 1995). In the context of fluent speech, children may perceive a collection of
=und segments as a single sound unit. For example, a syllable composed oftwo or more
phonemes may be heard sirnply as a single sound unit. Behavioral studies of children's
phonological skills have cunfirmed that phonological awareness, like other sMs, follows a
developmental progression (Goswami, 1991; for review see Goswami & Bryant, 1990).
Initiaiiy, young children find it an easier task to segment fîuent speech into syllables as
these units of sound roughly correspond to articulatory acts (Kuhl& Meltzoq 1982).
Once children are able to isolate syllables, they then begin to divide syilables into onsets
and rimes' but have difncuity apprehending the interna1 sound segments of complex onsets
and rimes. At this stage, chiidren can analyze a spoken word like /no/ into /rd, a one-
phoneme onset, and /O/,a one-phoneme rime. They can divide /plan/ into the onset /pi/
and the rime /an/. However, they have more difEculty analyzing the two-phoneme onset
of this word into /p/ and N, and the two-phoneme rime into /a/ and /ni. Mer young
children have an awareness of the onset-rime units, they can segment words into individual
sounds nich as phonemes (Treiman, 1991).
Isolating smaller segments of sounds, such as phonemes, is more chalienging as
they are more deeply embedded within the utterance @ttrouer & Studdert-Kennedy,
1986). Because of coarticulation, any point within a spoken word refiects both carry-over
effects fiom previous phonemes and anticipatory
of phonemes foilowing, that is,
the articulation of one segment is not always accomplished before the articulation of the
beginning of the next segment. For example, it is typical in English to begin opening the
velo-pharyngeal port for a nasal consonant during the articulation of a p r d m g vowel.
This r d t s in a nasalized vowel. This anticipation of an dculatory geshire is d e d
antitipatory coarticulation Coartidation c m also occur in the other direction For
example, the lip rounding gesture necessary for a rounded vowel may still be present in a
consonant several phonemes later (for o v e ~ e w
of martidation, see MacKay, 1987).
Because of coartidatory effects, a given segment contains some acoustic information
about the ~ ~ ~ o u n d isegments
ing
and may therefore provide a clue to their perception.
Coartidation pennits the rapid production of phonetic strings but obscures the
underlying segmental nature of speech. As a remit, information fkom several phonemes
are folded into a single unit of sound. This leaves no overt cue to the underlying
segmentai nature of speech, hence the spoken word may appear seamless. It has been
suggested that perception et the level of the phoneme rnay develop later in childhood,
perhaps as a consequence of learning to read (Gomami & Bryant, 1990; Gosw-
1995;
Treiman, 199 1).
In studying the specific phonological and semantic processes that might be
Uivolved in speech perception, the auditory event related potentials
technique
offers information that behavioral methods do not- W~thERPs it is possible to follow a
continuous course of brain d v i t y , and thus obtah idormation regarding the end product
of processing and of the sequence, timing and stages of specific processes. ERPs dso
render valuable information wncerning auditory processing, even in certain cases when no
deliberate attention is directeci to the sthdus event. This is a distinct advantage over
behavioral tests of phonologid awareness which r w e conscious attention to the
internai structure of words: for example, asking a child to demonstrate her phonologid
awareness by counting the sounds in words deleting sounds (e.g., say belt without the
iü), rnanipulating sounds (e.g., revershg phonemes) or categorizing sounds (i.e.,
categorizing words by initiai, medial and final sounds). Such tasks require the listener to
deviate ffom the more fiuniliai task of attending to the meaning of words to a less famüar
task ofreflecting on the intemal sound structure of words. This type of task manipulation
requires a certain metalinguistic abiiity which may or may not be present at an early age.
Because of the relatively high level of abstraction needed to demonstrate phonologïcal
awareness, many young children may not be able to d m i e what they achiaiiy know
regarding the phonology oftheir native language. Altematively, a particulariy bright child
may learn to perform a behavioral task with practice and corrective feedback, but lack the
phonological skills to perform dierent or unfamiliai tasb (Morais 1991).
The use of auditory ERPs for the study of the processing of linguistically
significant feahues of speech by the brain has been extensive in the past decade (Maiste,
Wiens, Hunt, Scherg, & Picton, 1995; Kraus et al., 1993; see Kutas & van Petten, 1988
and MoKese, 1983 for review). However, a large proportion of this research has been
limitaito the study of isolated speech stimuli (e-g., voice onset times, consonant-vowel
syllables, single words). Because the acoustic characteristics of speech spoken as isolated
independent u h are often very difEerent fiom those of mnning speech, studies of isolated
acoustic and phonetic features rnay provide limited idonnation toward the understanding
of the necessary events that lead to phonological and semantic processes necessary to
comprehend everyday fluent speech (Jerger, Jerger, & Abrams, 1983).
Recent studies Ugng spoken sentences varying in contexhd constraint have
demonstrated that auditory ERPs may provide additional information toward the
understanding of the phonological and semantic processes involved in the perception of
nuullng speech. In a shidy designeci to examine the N400' response, an ERP component
linked to semantic processing (for review see, Kutas & van Petten, 1988), Comoiiy,
Stewart, and P W p s (1990) found a negative response preceding the N400 by
approximately 200 ms in adults to the terminal words oflow contextually constrained
spoken sentences. In a follow-up study, Comolly, Phillips, Stewart, and Brake (1992)
found that the N200 was d é c t e d by the introduction of a masking stimulus while the
latency of the N400 was delayed significantlyand concluded that the N200 and N400
reflected different processing. Because spoken terminal words took approximately 300
rns to present, Connoliy and PWps (1994) proposed that the N200 response may index
pre-attentive or automatic auditory processing of the initial sound segments of words
termhating spoken sentences.
Comoily and Phillips (1994) demonstcated the hnctional independence of the
N200 and N4ûû response in adults using spoken sentences terminating in words that
containecl phonological, semantic or both phonological-semantic violations. A semantic
violation was defined as a terminai word whidi did not fit semanticdy with the preceding
sentence c o n t a . Both high or medium probability terminai words were defineci as
semantically appropriate. A phonological violation was define.as a terminal word that
had an initial phoneme which Wered £tomthe initial phoneme for a highly probable
semantic ending. For example, ifthe expected terminal word for a particular sentence was
6
neck, and the sentence terminateci with the word baving the uiitial phoneme as something
other than ln, it would be viewed as a phonologicai violation As in previous shidies, they
found that termina words that contained semantic violations elicited the N400 response
whereas terminal words that confhned semantic expectancies (Le., were semantically
appropriate) did not evoke a N4ûû nsponse. Interestingly, only in conditions thaî
contained phonological violations did an N200 response appear. Because the N200
response appeared sensitive to violation of phonological expectancies, the N200 was
renamed as the phonological mismatch negativity PMN) to avoid confusion with other
ERP components occurring w i t h the same latency range. Because the PMN appears
sensitive to the acoustic-phonetic processing of the initial segment of the terminal word
and given that it can be elicited fiom the listener at an automatic level of processing, the
PMN rnay have potentid application in the investigation of the phonological processing
abilities of young chiidren.
A study which could be interpreted as refiecting PMN response in children cornes
fiom Holcomb, Coffey and Neville (1992). They studied the dwelopmental course of the
N400 in the auditory and visuai modality using sentences that tenninated with either a
semanticaliy appropriate (highly expected) or semantically inappropriate (unexpected)
word. In their experiment with participants ages 5 to 26 years, it was found the onset of
the N400 in the auditory modality was earlier, larger and spanned a wider latency range
than the onset of the N400 observeci in the visual modality. These authors concluded that
the processing requirement in the two modaIities d E d . It could be argued that the
eariier onset of the N400 in the auditory modality d e c t s the presence of a separate
wmponent, the PMN.
Given that prereaders have an awareness of individual phonemes at the beginnings
of words, where these initial phonemes constitute single phoneme omets, they are able to
determine when words share the same initial soumis (e.g., get, gan) and when they do not
(e.g., pan, m). It wouid be U i f o d v e to detamine how this SMoperates in the
context of fluent speech ushg an adaptation of the Comoliy and Phillips (1994) p d g m
with sentences containhg phonologid and semantic violations. Prereaders may be able
to quickly identify an inappropriate terminal word by its onset sound and therefore elicit a
PMN response to sentences cuntaining phonological violations. The combined shidy of
the PMN and the N400 enables one to obtain information concerning not only the bottomup process of employing information contained in the awustic signai to decode meanllig
Cui this case s
d e r phonological segments, nich as word onset or the initial phoneme),
but also of the top-down innuence o f meaning on the perception of sound. Two
questions were addresseci in the m e n t study: a) can the PMN and N400 wmponents be
obtained in preschool-aged girls in response tu spoken sentences containing phonological
and semantic violations? and b) ifelicited, are the PMN and N400 responses observed in
the chiid simüar to the responses observed in the adult as shown by Comofly and Phillips
(1 994)?
Method
Subiects
-
N i e preschool-age girls with a mean age of 5-5 years (range = 5.0 5.9 years)
participated in this shidy. AU participants were right handed as v d e d by parental
report. Children were recnUted nom local daycares in the W a x area (pop. 350,000).
Ali children met the following criteria: 1) English as a first language, 2) hearing sensitivity
within normal limits, that is, hearing thresholds equal to or better than 15 dB HL for pure
tones fiom 500 to 4000 IL in octave steps in both ears, 3) no history of chronic ear
infection, 4) no history of neurological problems, and 5) receptive single-word vocabulary
within average range as measured by the Peabody Picture Vocabulary Test-Revised
(PPVT-R; DUM & Dunn, 198 1). Data coilected fiom an additional four children were not
included in this analysis because of their fdure to satie d inclusion criteria (n=3) or
because of equipment dficulties (n= 1).
Pnor to this study, 20 grade prirnary students (16 females and four males) between
the ages of five and s u years of age participated in establishing the Clore probability of
the auditory stimuli. Thwe children were recxuited h m three grade primary classes
within the Annapolis Vdey District School Board.
dito-
. .
The practice and test stimuli were age-appropriate sentences taken or adapted
fiom Holcomb, Coffey, and Neville' (1992). The test stimuli were generated nom 23
' ~ h i author
s
is very gratefbl for the permission to use these stimuli.
8
9
sentences which had a Cloze probabüity of equal to or be#a than .BO as judged h m the
sarnple of 20 grade primary students. By varying the terminai word of each sentence, four
conditions were aeated, brioging the total number of sentences to 92 (23 per condition).
The sentences were assignecl to one of four conditions in which the terminal word of
sentences varied accordhg to phonologid andlor semantic appropriataiess. Sentences
ending with words that met the phonological or semantic expectancies were descnied as a
"match", whiie those that violated expectancies were descrïïed as a "mismatch". For the
phonological match-semantic match condition (PM-SM) the final word was higtily
predictable (Cloze p r o b a b w z 30)given the p r e d i i g sentence context (e-g., Girmes
haw long rzeckE). In the phonological match-semantic mismatch condition (PM-SMM)
the onset phoneme of the terminal word was identicai to the hi& Cloze probabiiity word
in the PM-SM condition, but it was semanticitlly inappropriate (e.g., G i w e s hnve long
pets). In the phonologicai mismatch-semantic match condition ('MM-SM), sentences
terminateci with words of low Cloze probability: the terminal word was semantically
appropriate, but the onset phoneme was relatively unsrpeaed (e.g., GkMes have long
&)-
Finaily, in the phonologid mismatch-semantic mismatch condition (PMM-SMM),
the terminal word of each sentence was semantidy inappropriate and the onset ofthe
word was unexpected (e-g., Giraes have long-).
is listed in Table 1.
A summary of the four conditions
of the Four-
Table 1.
Condition
..
C~N&UL
Manipulation
Sentence
- -
-
-- -
PM-SM
Giraffes have Iong necks.
none
PMM-SM
G M e s have long legs.
phonolo@caIviolation ody
PM-SMM
Guaffes have long nets.
semantic violation only
PMM-SMM
Giranes have long pots.
phonologid and semantic violation
Ali temiinal words began with a single phoneme onset. Sentences varied in length f?om
four to 12 words (mean = 7.8). Each sentence was recurded using an adult fernale at a
n a d speakhg rate, digitized a -a samphg rate of 20,000 Hz. The presentation of each
sentence ranged in duration fiom 3 to 5
S.
The sentences were presented d i o t i d y at a
cornfortable listening level(80 dB SPL). The complete list of stimuli is listeci in the
PrDcedureS
While the cldd watched an age-appropriate vide0 the researcher measured the
chüd's head to identifii electrode placement. The area was marked with a cosmetic pend,
and cleaned with a pumice solution to reduce the Unpedance between recording electrodes
and skin.
Each c h i 'was tested individually in a sound attenuated booth. Ten surfhce
electrodes (nlled with wnductive gel) were placed on the scalp: three on rnidline sites
(Fq Cq and Pz), two on temporal sites ( T 5 and T6), two on reference sites at eariobes,
two on occuIarmotor sites (above and beiow the nght eye), and one on the forearrn (as the
cornmon electrode). A scheme of electrode placement is shown in Fig. 1.
Figure 1. Scheme ofRecording Situ
12
The task was presented to the child as a cornputer game. The child was instructed
to relax and sit quietly while listening to sentences through insert earphones. A 1 0 0 Hi
tone was presented for 300 ms at 65 dB SPL pnor to each sentence to cue the child when
to begin listening. The child indicated whether or not the sentence was semantidy
correct or incorrect ("OK"or "siliy", respectively) by pressing a button with a
conesponding pichire of a happy or silly face. The chiid was instructed to fixate on the
response pad while listening to the sentences. Response hand was counterbalanced across
subjects. The task was self-pacad, each triai beginning when the subject pressed a button.
Each child engaged in a practice task made up of 23 sentences so that at least five
incidences of each condition were presented.
The task was presented in four blocks in the same order for each chiid. Each block
consisted of 23 sentences pseudo randomly distributeci by type with no sentence type
occurrhg more than h c e in sequence. Sentence stems were not repeated within a block.
Each block lasted approxhately five minutes. Short breaks of approximately three
minutes were &en between blocks during the testing procedure so that the chiid couid
rest. Auditory ERP recordings were acquired in continuous mode. The recording
parameters are shown in Table 2. Following the auditory ERP testing, a hearing screening
test was performed. Following the hearing screening, the child was administered the
Auditory Anaiysis Test (AM; Rosner & Simon, 197l), and the Peabody Picture
Vocabulary Test-Revised (PPVT-R; DUM & DUM, 1981) to obtain estirnates of early
phonologid awareness skiU and single-word receptive vocabdary, respectively.
Table 2. RecorTone Duration
300 ms
--
Sentence Duration
SOOO ms
Tone Intensity
70 dB SPL
Sentence Intensity
variable 55 - 65 dB SPL
s d f ~ d
Rate
Number
92
Ear Presentation
Binaural
Mode Presentation
Air Conduction
Transducer
Insert Earphone
Amplification
20,000 x
70 pV
Sensitivity
Analysis T h e
Data Points
1100 ms
512
Fz - Al + ~2'
Channel 1
- -
- -
Channel 3
Pz-Al+A2
Channel 4
T3 - A l +A2
Channel 5
T4-Al+A2
Common
nonceptialic (forearm)
Analvses
Behavioral Data
The overd reaction times and percentage of correct responses for each of the four
sentence conditions were calculated. For an individuai, a total of 23 correct responses
were possible for each condition. Repeated measures anaiysis of variance (ANOVA)
were used to analyze the response accuacy and reaction times (RTs) of the response to
the four sentence conditions.
ERP Data
off-line amiysis tirne commenced 100 ms before the onset of the terminal word and
continueci for 1000 ms post terminal word onset. Latency of the ERP components was
measured as the tirne fiom stimulus onset of terminal word to the most negative point
reached within the time window of specific components. The latency of the N400 was
measured as the most negative point between 500-850 rns (i 10Y0).The PMN was
identified as the most negative point occuning between 300 rns and the peak of the N400
(300450 ms
10%). Amplitude of the N400 and PMN was measured as the voltage
digerence between the mean activity for the 100 ms prestimulus period (Le., badine) and
the point of greatest negative deflection within the defined latency range.
Repeated meaSuTes analysis of variance (ANOVA) using Greenhouse-Geisser
conservative degrees of fieedom (Greenhouse & hisser, 1959) were used to analyze the
amplitude and latency measures of the components. The goal of statistical data analysis
was to evaluate whether or not the amplitude or latency of the PMN component in
1s
conditions containing a phonologid violation @honologicalmismatch) was ciBiirent fiom
those conditions &out
a phonological violation, Likewise, d y s e s ofthe N400 were
to determine whether or not the amplitude or latency of
the N400 component in those
conditions containhg a semantic mismatch was different fiom those conditions without a
semantic mismatch Post-hoc cornparisons were made with the Tukey Honestly
Sigrifkant Difference (HSD) test.
Resuits
The percentage of correct responses and mean reaction times (RTs) in the four sentence
conditions, as weii as the r d t s ofthe statistical anaiyses an reporied in Table 3. There
were signincantly more mors committed in the PMM-SMcondition a3,24]=55.30
p<.00 1). Mean RTs were significantiy longer for the PMM-SMcondition (EI3,24]=3 -40
~<.05). See Figures 2 and 3 for boxplotss of RTs and percentage correct by sentence
condition.
-
Reaction Tirne
Percentage Correct
Condition
M
SR
- . ... -. -
hl
-
sr2
- - - . - ---
PM-SM
85.5 1
13.22
5.64 s
O. 75
PMM-SM
54-59'
13.60
6.40 s**
1.16
PM-SMM
92-90
3.45
5.58 s
0.25
5.41
5.83 s
0.79
PMM-SMM
94.58
Note&* E.01 level. ** g<.05 level.
The eariy sensory components (e-g.,NI,P2) were not evident because natural rates of
speech precluded a clear and abrupt onset. The absence of these early components has
been noted in eariier studies using running speech stimuli (Connoiiy et al., 1990, 1992).
The grand mean ERPs for the group are presented in Figure 3 (a to c) for each sentence
Figure 2. Boxplot ofReaction Tiie by Sentence Condition
Figure 3. Boxplot for Percentage Correct by Sentence Condition
18
condition Despite appearances of amplitude differences baween the condition
wavefiorms, there were no statisticaily sipificant main &ects of amplihide or interaction
e f f i of amplitude by site for either the PMN or the N4ûû component. That is, the
mean amplitudes of the PMN for the four conditions were comparable and the mean
amphdes ofthe N W did not M e r across the conditions. No sigmficant scaip
d i m i o n H k c t s were seen for either the amplihide or latency. The mean latemcy and
amplitude of the PMN and N400 respective@are show in Tables 4 and 5 . The boxplots
for amplitude and latency of the PMN are shown in Figures 5 (a and b), respectively. nie
boxplots for amplitude and latency of the N400 are s h o w Figure 6 (a and b), respectively.
Howewer, a main effect for latency was found for the PMN (condiion main effect,
E(2.50,24)=6.48,
-004). The peak PMN latency of PM-SMM was delayed by over 43
ms relative to the PMM-SMM condition and delayed by over 27 ms relative to the PM-
SM condition.
Figure 4. (a) Grand Average Waveforms Comparing PM-SM to PMM-SM across Site.
Figure 4. (b) Grand Average Wavefom Comparing PM-SM to PM-SMM across Site.
Table 4.
PMN Latency in ms
PMN Amplitude in PV
Condition
PM-SM
366-29
6-82
4.27
0.68
PMM-SM
3 76-46
7.96
4.66
1.O8
PM-SMM
394.18*
22.5
-4.9 1
0.32
351-17
5.19
-6.26
0.80
PMM-SMM
Note. *p<.OS level
Table 5.
N400.
N4ûû Latency in ms
N400 Amplitude in pV
Condition
M
SIT
M
32
PM-SM
692.72
32.25
-5.48
0.89
PMM-SM
664.26
14.65
-5.23
0.40
PM-SMM
679.19
44-22
-7.00
0.59
PMM-SMM
633.43
48.43
-6.24
0.67
Figure 5. (a) Amplitude ofthe PMN by Sentence Condition
Figure 5. (b) Latency ofthe PMN by Sentaice Condition
Figure 6. (a) Amplitude of the N40 by Sentence Condition
Figure 6.(b) Latency of the N4ûû by Sentence Condition
Discussion
The results of this study indicated that the mean amplitudes of the PMN and the
N400 for were comparable across the four conditions. One interpretation for the absence
of significant PMN or N4ûû amplitude differences among the four conditions is that the
amount of processing reflected by the PMN and N400 may have been comparable across
conditions. That is, the negativity found across aIl four sentence conditions may be the
result of the iimited experience of the children with singlesentence processing and
semantic classification. More speci6cally, N400 amplitude has been shown to inaease as
task demands required more involveci processing of the identity of the stimulus. Lovrich,
Cheng and Velting (1996) have shown that inreased N400 amplitude corresponded to
inmeases in task demands. In addition, the amplitude of the N400 has been s h o w to be
smder in response to primed, repeated, or matching stimuli, because such (extended?)
exposure would reduce demands on memory access and integration (e.g., Holcomb et al.,
1992). Children's relatively limited expenence with altemate word endings for the same
sentence stem could have produced less primeci responses or more in-depth processing
for aU four conditions, resulting in comparable negativity.
This experiment had an unexpected findimg in that terminal words which best fit
with the preceeding sentence context (PM-SM), elicited an N400 response comparable to
the anamoulous fit in children. Holcomb et al. (1992) found that younger subjects (age 5
to 14 years) exhibited an N400-like negativity for the bat-completion sentence condition
(e-g. PM-SM) as well as for the anomalous-completion sentence condition ( P M M - S m .
However, negativity was found to be greater in the anomalous-completion condition. In
26
the present shidy, the negativity found in the PMM-SMis consistent with a previous study
in the visuai modality by Kutas and Hiliyard (1984) which demonsûated N400s to terminai
words that were semanticaily correct but of lower Cloze probabiiity. Kutas and HiIlyard
(1984) reporteci a gradation in N400 amplitude to words varyhg in Cloze probability that
ended medium constraint sentences. The PMM-SMcondition in the present study rnay
have contained a wnfound that diminished the abiiity to Merentiate between the PMN
and N400 on the basis of phonological and semantic processing. That is, a terminal word
in which the initial phoneme was unsrpected, but semantidy correct, required the use of
a lower Cloze probability word. This in tuni, would be more k e l y to elicit an N400
(Kutas & Hillyard, 1984). Although Connoily and Phillips (1994) did not elicit an N400
response to PMM-SM in adults, this response may be more robust in preschool-age
children who may be less experienced in interpreting terminal words of lower Cloze.
Within the semantic mismatch conditions (PM-SMM and PMM-ShAhQ behavioral
response accuracy and RTs were not significantly different fiom each other. One
interpretation for this findimg is that the children were not mislead by the phonological
match at word onset into thinking the terminai word was a semantic match as well. It
appears that the shared onset of the PM-SM and PM-SMM conditions was not
signincantly powemil enough cue to lead chüdren to mistake one word for another (e.g.,
"nets" for "necks"). aven the onset /ni,one does not have sutncient Somation upon
which to spenilate what word is being uttered, because many words in the vocabuiary of a
young child that &are the omet ln/. Onset sirnply be@
to narrow down possible word
candidates at the onset l d . Howeva, once amustic aies f?om the rime unit are heard,
one can begh to m
w the possible candidates d o m considerably, perhaps to a single
word given the additional iofluence of semantic mes. Therefore, preschool-age children
may find the mismatch at the rime unit (e-g., n&-na
more d e n t because of the
influence of both acoustic aies and semantic aies. Treiman and Zukowski (1992) in fact
f o n d that rimes were easier to recognize than onsets in an auditory matching task during
which children had to make same/Merent judgements on the basis of either onsets (plank-
piea) or rimes (spl&w&). ûoswarni (199 1) found thai English-speaking school-age
children w d d more &y
read words that share the same rime unit (&-ln&) than words
that share the same word onset and vowel (e-g. w-b)
where the shared spelling unit
required segmenthg the rime. It may be that the rime is the more salient unit with the
syllable (MacKay, 1987). It seems reasonable then, that when one is expecting the word
"neckn,and hemNnet ",a violation has ocairred in the salient parameter of the rime unit.
It is easy to judge when something is wrong when the more salient parameter of rime has
been violateci than to be mislead by the less salient onset match. The PM-SMMcondition
may be relatively easy because the terminal words are rnismatched at the rime level and
therefore can be easily recognized as inappropriate.
Another more speculative explanation is that young children may be able to
recognize words based on more holistic recognition strategieq given that the fine-grained
discrimination processes necessary for decidiig among words are not as necessary as they
are for adult perceivers who tend to know many words that share simüar awustic
properties. However, this explanation cannot account for the pattern of ERP findings,
which suggest ciiildren are indeed making fine discriminations.
28
The latency of the PMN was Sgnificantiy longer in the PM-SMMcondition
relative to the other conditions. One explanMion for this hding is that the onset match in
the presence of semantic mismatch may have produceci tirne delay where f i d e r
processing was needed to resolve the information of bottom-up and top-down which âuled
to wmplement one another. However, it was the N400 latency which was significanfly
longer in the PM-SMMin the Connoily and Phillips (1994) study with adults. Given that
the latencies are comparable, both the PMN in this study and the N400 in Comolly et al.
(1994) occurred around a mean latency of approxixnately 400 ms window, the latencies
are comparable and could wnceivably reflect a W a r stage of stimulus andysis.
However, ERP responses &om children are typically have larger amplitudes and longer
latencies wmpared to those responses elicited fiom adults.
It is noteworrhy that the increased PMN latency for the PM-SMMcondition did
not impact on the readon tirne (RT) of the behavioral response. However, the
wmparatively low accuracy and long RTs of the behaviord response for the PMM-SM
condition were significantly diierent suggests that the sentence were very dernanding for
children to process. Given the possible higher impact a mismatch at the rime level, the
phonological mismatch in both the rime and onset unif as in the PMM-SM, may have
primed the preschool-age child to dismiss a semantic match as inappropnate. An altemate
exphnation for the di&iculty with the PMM-SMcondition is that chiidren this age with
limited language experience may have accepted oniy one semantic metch, the highest
Cloze terminal word, to fit the preceding sentence contexi (e.g., "GUaffes have long
necks"). Preschwl-age children might have often heard giraffes desçnied as havhg long
29
necks and so wuid judge the tmthfihess of this statement with relative ease. However,
judging the truthfùlness of "GirMeshave long legs", may pose the greater challenge for
chüdren who have heard @es
les often descnied as havir~glong legs andor they may
not have made the observation themselves. Additionally, the phonological mismatch at
word omet and rime may have primeci childnn to beliew that the terminal word was
semanticaliy wrong. Phonological mismatches at both the onset and rime unit in the
presence of a perceived semantic match may have caused the chüdren to rethink their
initial assumptions, thus producing relatively long RTs for the PMM-SM condition. The
ERP and overt behavioral fidings suggest that the PMN may be sensitive to bottom-up
processing of amustic mes and that RT may be sensitive to top-down processing
important for verifjing the semantic congniity of the sentence. The m e n t fïndings
support the conclusion that RT and ERP do not necessarily reflect identical processing
operations (Lovrich, Cheng & Velting, 1996).
How do the findings of this study contniute to the understanding of the
relationship between speech perception and phonological abîlity and ultimately reading
ability? One theory of speech pereption explains that in order to efficiently process a
steady and rapid Stream of running speech, both bottom-up and top-down information
must work in tandem (for review of the paraiiel processing theory of speech perception
see Ryds 1996). Using both sources of information may ease the burden of on-line
processing of the acoustic signal, reducing dependence on the acoustic information
present in the speech signal. Prior knowledge of words and expressions can ease the
burden of on-line bottom up p r o c d g . A m t e expex%ttionsmay ease the perception
30
of speech. Topdown information allows an over-arching fhmework to which may be
added acoustic information,
It is reasonable that children who rely mostly on bottom-up processîng rnay requise
a longer time to analyze the information they hear as the processing is largely sequential
and dependent on complete acoustic mes. This in&cient strategy could result in
processing breakdowns during the rapid presentation of numerous acoustic mes, and may
account for why some cNdren perforrn listening task pwriy with degraded acoustic
signais or in hostile acoustic environrnents. Failure to make use of topdown processing,
robs the child of an over-arching fhmework in which to fit pamal awustic information.
Processing acoustic cues sequentdy in time with little or no prior knowledge of what is
to l o g i d y foiiow is both in&cient and hard to maintain. Auditory memory may be
taxed as attempts are made to store cues not aiready analyzed. Depending on the speed of
bottom-up processing and the capacity of auditory memory, entire syllables or words may
f d to be processed. Similady, children who are hearing Unpaireci and have trouble
detecting which phonemes are present in the acoustic signai, rnay have less information to
integrate with topdown information. The consequence o f f h g to integrate bottom-up
and top-down idonnation is that one becornes less skiiled at anticipating words fiom
partial acoustic signais and has less exposure to language use. Such children would have
restricted vocabuiaries and would be l e s expenenced in making the comection between
sound and symbols.
Further research is required to more accurately define the characteristics of the
PMN,its sensitivity to phonological processses, its relationship to semantic processing and
ultirnately phonological awareness. Ideally this study shodd be replicated and extendeci
with a larger sample of children. Replicating this study with a larger sample would
confinn the present hdings, ailowing for a more powedid separafion ofthe effects found
to date.
In the present study a small sample of sentences stems were used. This was
required to match the four sentence conditions in terrns of age appropnateness, Cloze
probability and phonetic structure. The overali number of trials was determinecl by what
was judged to be reasonable test session duration, given the attention span and alertness of
preschool-age children. However, the issue of ERP variability is signincant enough that a
longer test session may be requind. Shply increasing the number of sentence stems
thereby increasing the number of sweeps which contribute to the averages certainly would
serve to reduce the noisiness of the data and iikely would reduce the variability associated
with the both the PMN and N400 responses. Future research in phonological and
semantic processing using ERP teçhnology rnight compare sentence stems which are
repeated across condition to unique sentence stems aaoss condition d e t e d e the e f f i s
on ERP waveforms.
Endnotes
1. The ternis onset and rime refkr to units of sound within a syllable. Onset refers to the
initial consonant sound(s) preced'mg the vowel, and the rime corresponds to the vowel
sound and any remaniing consonant sound(s) in the rest of the syliable.
The auditory ERP is a recording of the changes in electric potentiai in the brain in
response to auditory stimuiation The signal averaghg technique aiiows for tbis datively
small amplitude activity to be extraded fkom the random changes in elecfrical actniity.
nie auditory ERP is usually represented as a cornplex wavefm rdeding changes in the
amplitude and fiequency of electricai potentials over the. The fluctuations in amplitude
and latency of certain peaks and troughs dong the response wavefonn are thought to
reflect changes in brah activity. The amplitude and latency of neurai activity foiiowing a
target stimulus may provide a meanire of neural information processing reflecting stimulus
perceptions and task decisions by an individuai (Jbfoü", 1983).
2.
3. The nomenclature ofERe components simply refers to the typical latency and polarity
of the wavefonn; the N400 is so narned because it is characterked by a negative voltage
response with a latency of about 400 rns post stimulus onset.
4.
The result is a b a l a n d reference that qualires the contributions of each ear to each of
the sites (scalgelectrode arrays). Thus, dirences arnong wavefonns for each site (e.g
Fz and Cz) can be atln'buted either to hemispheric or C N S factors, rather than to
conmiutions from the ear or &£Ferencesin the orientation of the array with respect to the
stimulus side.
5. Boxplots are sumrnary plots based on the median, quartiles, and extreme values.
Boxplots are formed fkom "boxes", which contain the 50% of values e g between the
25th and 75th percentiles, and the "whiskers", lines that extend from the box to the highest
and lowest values, excluding outliers. A line a a a s s the box indicates the median. Boxplots
dso wnvey information about spread and skewness.
Practice Stimuli List*
1. The squirrel clirnbed the tree and hid some
a. nuts.
2. The boy caught a fish with his fishing
a. rod.
3. J i saw the lightning and heard the
a. thunder.
4.
On Saturday morning I like to watch
d. bath.
5. The girls danced to the
d. bean.
6. John threw rocks into the
d. summer.
7. Every ~ g h Kathy
t
sets the table before
a. supper.
8. Susan put away her crayons and colouring
b. bed.
9. For lunch I Iike a bowl of tomato
d. shampoo.
10. Sally drew a beautitiil
a. picture.
11. J i saw the lightning and heard the
b. Thursday.
12. The squirrel ciimbed the tree and hid some
b. noise.
13.
On Saturday momhg I Iike to watch
c. t.v.
'Note: a. PM-SM, b. P M - S m c. PMM-SM, d. PMM-SMM.
14.
The boy caught a fish with h% fishhg
c. net.
15. Susan put away her crayons and colouring
a book.
16. Every night Kathy seds the table before
c. dinner.
17. The comic book CO-
nfty
a cents.
18.
Sdy drew a beautifid
b, noise.
19.
O n Saturday morning 1 me to watch
b. camds.
20. Jiil saw the iightning and heard the
c. min.
2 1. The squirrel climbed the tree and hid some
c. food.
22. For Iunch 1 like a bowI of tomato
a, soup.
23. The boy caught a fish with his fishing
b. rock,
Test Stimuli List
1. The girls walked to the store to buy an ice cream
a, cone
b. coat
c. sandwich
d. mitten
2. The bùd laid her eggs in the
a. nest
b. neck
c. barn
d. pet
3. Betty jumped into the swimming
a. pool
b. purse
c. hole
d- bus
4. Mike's dad planted a tree in the back
a yard
b. year
c. field
d. night
5 . Marty sang his favourite
a- Song
b. sock
c. tune
d. house
6.Mary had a Little
a. lamb
b. land
c. sister
d. mountain
7. It's hot in the summer and cold in the
a. winter
b. window
c.
fd
d. hug
8.1 Wear a sweater when its
a cold
b. mat
c. chiuy
d. pend
9. After supper Saily heiped wash the
a dishes
b. dimer
c. table
d. summer
10. Bobby played in the rain and got soaking
a wet
b. web
c. feet
d. kite
11. G
i
r
a
f
f
e
s have long
a. necks
b. nets
c. legs
d. pots
12. Johnny took the dog for a
a walk
b. watch
C. nln
d- fish
13. Tomm)'s aunt gave him a hug and
a. kiss
b. kid
c. giît
d- farm
14. Dogs have four
a- legs
b. lips
c. feet
d. hills
15.1 dont go to school on Saturdays or
a. Sundays
b. suppers
c. holidays
d. picîures
16.J
m
i drank a giass of orange
a juice
b. jail
=- POP
d*tool
17. Spiders catch bugs in their
a. web
b. wet
c. mouth
d. Iake
18. The boys ate aü the chocolate chip
a cookies
b. colows
c, muffins
d. water
19. Ellen was punished because she told a
a. lie
b. light
c. fib
d, bat
20.Jii planted some flowers Ui the
a. garden
b. garbage
c. yard
d. feet
21. She cut the paper with the
a scissors
b. sister
C. cutter
d. pancake
22. Betsy is learning to ride a
a, bike
b. bite
c. horse
d. pie
23.1srnile when I'rn happy and cry when I'm
a sad
b- sat
c. hurt
d- rock
References
Bird, J. & Bishop, D.V. M.(1992). Perception and awareness of phonemes in
J o u d of D
phonologically impaireci chüdren
i
s
o
r
.d
. m 77,
289-3 11.
Bird, J., Bishop, D. V. M., & Freernan, N.H. (1995). Phonologicd awareness and
iiteracy development in children with expressive phonological irnpairments.
h a d Hearing Reswch 38.446-462.
Comoliy, J. F. & Phillips, N.A (1994). Event-related potentiai components
reflect phonological and semantic processing of the tenninal word of spoken sentences.
256-266,
ConnoUy, J. F., Phillips, N.A, Stewart, S. K, & Brake, W. G. (1992). Eventrelated potential sensitivity to amustic and semantic properties of t e r d words in
sentences.
n and- .1
1-18.
C o ~ o l i yJ., F., Stewart, S. H.,& Phillips, N.A. (1990). The effms of processing
requirements on neurophysiologicd responses to spoken sentences.
n
a
39.302-3 18.
DUM,L.M. & DUM,L.M. (1 98 1). Peabody Picture Vocabulary Test-Revised.
Minnesota: Amencan Guidance Service.
Goswami, U.(1995). Phonologicd development and read'ig by d o g y : what is
analogy, and what is t not?
Journal 18ReadinP.a139445.
Goswami, U. (1931). Leaming about speliing sequences: The role of onsets and
rimes in analogies in reading. çhild Develo-
62.110-1123.
40
and to
Goswami, U.,& Bryant, P. E. (1990). Phon&gical
Hillsdale, NJ: Lawrence Erlbaum.
Greenhouse, S.W. & Geisser, S. (1 959). On methods in the analysis of pronle data.
c h o m s , 95-1 12,
Holwmb, P.J., Coffq, S . A . & Nevilie, H.J. (1992) Visuai and auditory sentence
processing: A developmentai analysis ushg event-related brain potentials. &velPpmenfd
Neuropychology&
8.03-24 1.
Jerger, S., Jerger, J., & Abram, S. (1 983). Speech audiometry in the young chad.
4(1), 56-66.
Kraus, N.,McGee, T., Micco, A, Sharma, A, Carre& T.,& Nicol, T.(1993).
Mismatch negativity in school-age chiidren to speech stimuli that are just perceptually
..
difrent. m c t r o e n w Ch& Neu-
123- 130.
Kuhl, P. K. & Meltzoe AN. (1982). The bimodal perception of speech in infancy.
1& 1138-1 144,
Kutas, M., & Hillyard, S. A (1980). Reading sense1ess sentences: Brain potentials
reflect semantic incon@ty.
Science 207,203-205.
Kutas, M.,& Wyard, S. (1 984). Brain potentials during reading reflect word
expectancy and semantic association. m e . 307. 16 1- 163.
Kutas, M. & van Petten, C. (1988). Event-related brain potentials studies of
language. In P X . Ackles, J.R Jennings, and MG-H- Coles (Eds.).-ces
in
o l o w Greenwich, CT:JAI Press.
LoMich,D.,Cheng, J. C., & V e l a D. M. (1996) Late cognitive brain
41
potentialq phonologid and semantic clessification ofspoken words, and reading ab* .
children. Jounial of C h c d an-
in
161-177.
Ne~rpgsychnlpev~
1
MacDonald, G. W. & Cornwall, A. (1995). The relationship between phonological
awareness and reading and s p e h g achievement eleven years later. J o u d of-
.
*..
Disabilrties_523-527.
MacKay, 1. R A. (1987).
The science of w
-CS:
h production(2nd 4.).
Toronto: AUyn and Bacon.
Maiste, A C.,Wiens, A. S., Hunt,M. J., Scherg, M., & Picton, T. (1995). Eventrelated potentids and the categorid perception of speech sounds.
H
u
68-90.
Marslen-Wdson, W. (1987). Fundonal parallelisrn in spoken word recogntion.
71-102.
MON', D.L. (1983). Event-related potentials and language processes. In kW.K.
Gaillard & W. Ritter (Eds.), T
.u .
t
o
&
~
~ c ouw o nse n k
Amsterdam: North-Holland. pp. 345-368.
Morais J. (199 1). Phonological awareness: A bridge between language and
literacy. ln D.I. Sawyer and B.J. Fox (Eds.),-cal
on of CU-ectiva
aw-ess
r
e
@p. 161-189). New York: Springer Verlag.
Nittrouer, S. & Studdert-Kennedy, M. (1986). The role of coartidatory effects
in the perception of fricatives by chilcûen and ad&.
ort on S~eechRe&
-es
S
u
SR-93/94,1-2 1.
Rosner, J. & Simon, D. P. (197 1). The Auditory Analysis Test: An initial report.
42
384-392.
Pa-)
San n e g o : .
Singuiar
Ryalls, J. (1 %%
to
A
Publishing Group.
Stuart, M.& Coltheart, M. (1988) Does reading develop in a sequence of stages?
* .
0muti0~
30. 139-181.
TaUal, P. (1980). Auditory temporal perception, phonics, and reading disabilities in
children.
andLanmi182-198.
ane
Treiman, R (1991). Phonological awareness and its roles in leaming to read and
awarenes in rea-
spell. In D.J. Sawyer and B.J. Fox (Eds.),
ent va
The
@p. 161-189). New York: Springer Verlag.
Treiman, R, Br Zukowski, A. (1992). Levels of phonological awareness. In S.
Brady & D.Shankweiier (Eds.), P
~
c !roc-d
. a
.
Hillsdde, NJ:
in
Erlbaum.
Wagner, RK. & Torgesen, J.K.(1987). The nature of phonological processing
Bulletin. 10L 192-
and its causal role in the acquisition of readiig skiIls. -cal
2 12.
Watson, B .U.& Miller, T.K.(1 993). Auditory perception, phonologicd
processing, and readiig abilityldisability. JourggJof aeech
H
-
36.
850-863.
Werker, J. F. & Pegg, J. E. (1992). Infant speech perception and phonological
acquisition. In C.A Ferguson, L. Menn & C. Stoel-Gammon, (Eds), P h o w
d e v e l o m . Timoniwn, Maryland: York Press.
IMAGE EVALUATION
TEST TARGET (QA-3)
APPLIED 1 IMAGE. lnc
S
-.-=
---
1653 East Main Sûeeî
Rochester, NY 14609 USA
Phone: 716/482-a300
Fax: 716/288-5989