DOCUMENT RESUME

ED 394 341 FL 023 753

AUTHOR Adams, Beverly Colwell; Wade, Melissa M.TITLE The Impact of the Comma on Sentence Processing.PUB DATE Apr 96NOTE 14p.; Paper presented at the Annual Meeting of the

American Educational Research Association (New York,NY, April 8-12, 1996).

PUB TYPE Reports Evaluative/Feasibility (142)Speeches/Conference Papers (150)

EDRS PRICE MF01/PC01 Plus Postage.DESCRIPTORS Adolescents; Adults; *Age Differences; *Ambiguity;

Children; College Students; Grade 4; Grade 6;Grammar; Higher Educatio,,; Intermediate Grades;*Language Processing; *Punctuation; ReadingComprehension; Sentence Structure; *Syntax

IDENTIFIERS Ambiguity Detection; *Commas; *Parsing

ABSTRACTA study investigated whether children and adolescents

use commas and the principle of Late Closure to guide sentenceparsing decisions as adults do in processing syntactically ambiguoussentences. The study consisted of three experiments, conductedsimilarly but with different subject groups: 24 university students;24 fourth-graders; and 19 sixth-graders. In each, subjects readsentences containing different arrangements of pre-posed and mainclauses and comma use, and answered simple comprehension questions,all presented on a computer using a self-paced, one-wordmoving-window procedure. Results indicate that all groups had shorterreading times on the main clause verb in late closure sentences thanin early closure sentences, supporting the dominati.in of late closureparsing strategy. While results for adults supported the predictedpatterns of comprehension speed involving closure and comma, theywere not statistically significant. The fourth-graders appeared notto use information about the comma to override the late closureparsing strategy. The sixth-graders were clearly sensitive to thelate closure strategy, but did not use comma information aseffectively as did adults. Reading times were similar to those offourth-graders. Four figures and one table support the data. (MSE)

************************************************************************ Reproductions supplied by EDRS are the best that can be made *

* from the original document. *

***********************************************************************

The Impact of the Comma on Sentence Processing

Beverly Colwell Adams

Melissa M. Wade

University of Virginia

American Educational Research Association

New York, New York

April 8 - 12, 1996

PERMISSION IL) HE- r'RODUCE ANP

rjiF MINATE THIS MATEHI

HAS i3EEN C,HANTED

10 E LUUCA 3NAI HE

INF OHM ATION cuNTf P

S DEPARTmENT OF EDUCAT,eN

EDUCATIONAL REEOURCCS INFORMATION

skims rcirmf,t rrrir; roprndir, er,CENTER 'ERIC'

r et eived frt."; pre Del,orfrrIrjrndfrod it

0 Minor changes havp hr r*.arte tinImprovo rebnoductirril qurthfy

Prirnts of VIOAr on npinuon claterl

proyhrm II 01.1,y

2 BEST COPY AVAILABLE

Introduction

When adults read sentences, they make grammatical assignments about thewords. This process, called PARSING, follows rudimentary principles requiredin the syntactic and semantic evaluation of text. One of the principles, LateClosure, guides the parser to attach incoming lexical items into the current phraseunder construction. Different kinds of syntactically ambiguous sentences (e.g.,garden-path sentences, such as (1) While the crowd cheered the popular magiciansawed the woman in half.) serve as an effective means for investigating the parsingprocess.

When adult readers process (1), they typically misanalyze the verb,CHEERED, by selecting the transitive form and attaching the noun phrase, THEPOPULAR MAGICIAN to it. Comparing reading times for the verb in the mainclause in ambiguous (1) and unambiguous sentences (e.g., (2) While the crowdcheered him(,) the popular magician sawed the woman in half.) indicatessignificantly longer latencies in the ambiguous sentences at the verb, SAWED.

Longer RTs at the verb in the main clause in ambiguous sentences such as(1) argue for Late Closure guidance. Late Closure guides the parser down theinappropriate path. The attachment error is realized at the verb in the main clause(SAWED). In sentences such as (2), Late Closure guides the parser appropriatelyto attach the NP, HIM, as the object of the preposed verb, CHEERED. Thegarden-path error in (1) compared to the "errorless" reading in (2) at the verb inthe main clause, reflects support for the Late Closure strategy.

Late Closure misanalysis of sentences such as (1) can be avoided by theinsertion of a comma, separating the preposed clause and the main clause (e.g.,While the crowd cheered, the popular magician sawed the woman in half.) Mostadults identify the comma as a grammatical convention that guides the parser toavoid the Late Closure misanalysis.

Although the Late Closure principle has been investigated extensively in the-adult population, research investigating children's parsing strategies is rare. (Butsee Adams & Royer, 1993; 1995). From an educational and developmentalperspective, the following questions become interesting:

(1) Do young children and adults use different parsing strategies when processingsyntactically ambiguous rentences?

3

2

(2) Do children and adults comprehend the comma as a guide to disambiguatesyntactically ambiguous sentences?

The goals of the study were to examine whether children and adolesL.ents useCommas and the principle of Late Closure to guide parsing decisions as adults do.

MethodParticipants

Twenty-one undergraduate volunteers at the University of Massachusettsreceived course credit for their participation.

Materials, Design, and Procedure

An example of the 20 sets of sentence conditions appears in Table 1. Thetargets included a preposed clause and a main clause. The four forms of eachsentence were developed by the (2 X 2) factorial combination of Early versus LateClosure strategies and the presence versus absence of a Comma. The commaconditions were included as control conditions.

Each participant was tested individually. Each participant read a toy' of 109sentences (20 target, 89 fillers). The materials were presented on a computerscreen in a self-paced, one-word moving window procedure. Approximately 1/3of the items was followed by a simple comprehension question.

Results and Discussion

The pattern of results demonstrated that adult readers read the verb in themain clause more quickly in Early Closure sentences than in Late Closuresentences, F1(1,20) = 8.15, p < .01, MSE = 42272.64. Furthermore, adultreaders read the verb in the main clause more quickly in sentences with commasthan sentences without commas, F1(1,20) = 7.18, p < .05, MSE = 22152.40. Ascan be seen in Figure 1, adult subjects read the verb in the main clause aspredicted. Although predicted, the interaction of Closure and Comma did notachieve significance. Thus, it seems that the Late Closure principle guides theadult parser to attach the postverbal NP to the preposed verb, close the initialclause, and then process the rest of the sentence without difficultly. For adults,the presence of a comma speededEapdhigutha on the verb in the main clause.

3

Participants

In order to determine the pattern of results for younger readers, 24 fourthgraders (9-10 years old), participated in Experiment 2. The fourth graders wereelementary school students who attended a school in Northampton, MA.

Materials, Design, and Procedure

The same materials were used in Experiment 2 as in Experiment 1. Thefourth graders were tested in a quiet room in the school library. Fourth graders,just as adults, read the materials on a computer screen using a self-paced, oneword moving-window procedure.

Results and Discussion

There was a main effect of Closure for fourth graders; i.e., the verb in themain clause was read more slowly in Early Closure sentences than Late Closuresentences, F1(1,23) = 7.41, p < .05, MSE = 57909.52. Although the pattern wasin the predicted direction (840.5 ms versus 880 ms), the Comma effect did notachieve significance. When a comma was included in a Late Closure sentence oran Early Closure sentence, the time to read the verb in the main clause was notfaster than when a comma was absent. The interaction of Closure X Comma didnot achieve significance. It seems that fourth graders did not use informationabout the comma to override the Late Closure strategy. Figure 2 illustrates the RTpattern graphically.

Experiment 3

Participants

Nineteen sixth graders, 11-12 years old, participated in Experiment 3. Thesixth graders attended the same elementary school as the fourth graders.

Materials, Design, and ProcedureThe same materials were used in Experiment 3 as in Experiments 1 and 2.

The sixth graders were tested in the same quiet room in the school library as thefourth graders. Sixth graders, just as adults and fourth graders, read the materialson a computer screen using a self-paced, one word moving-window procedure.

4

Results and Discussion

There was a main effect of Closure for sixth graders; i.e., the verb in themain clause was read more slowly in Early Closure sentences than Late Closuresentences, F1(1,18) = 11.14, p < .01, MSE = 17040.97. Neither the Commaeffect nor the interaction of Closure and Comma achieved significance. Figure 3charts the pattern graphically. Sixth graders were clearly sensitive to the LateClosure strategy but did not use comma information as effectively as adults.Surprisingly, the additional 2 years of reading experience relative to fourth gradersdid not produce differential reading times when the comma was present.

Experiment 4

Participants

Twenty-eight adolescent readers, 14-16 years old, participate :n Experiment4. The adolescent readers were students from three high schools in centralVirginia. The adolescents were tested at the University of Virginia whilepartici?ating in a summer program.

Materials, Design, and Procedure

The same materials were used in Experiment 4 as in Experiments 1, 2, and3. Adolescents read the materials from a computer screen and progressed througheach sentence using a self-paced, one-word moving window procedure.

Results and Discussion

The adolescent readers' data showed a main effect for Closure; i.e., the verbin the main clause was read more slowly in Early Closure sentences than LateClosure sentences, F1(1,27) = 7.38, p < .05, MSE = 94602.77. There was nomain effect of Comma and the interaction of Closure and Comma did not achievesignificance. Most surprisingly, adolescent readers did not use the commainformation to isolate the preposed clause from the main clause to avoid

misanalysis. Although the pattern of reading times represented the predictedpattern (666.3 ms versus 738.8 ms), adolescent readers did not use commainformation to avoid the garden path as efficiently as adults.

6

General Discussion

5

The present study used syntactically ambiguous and unambiguous sentencesto investigate the guidance of Late Closure and the impact of the Comma onsentence processing. All readers (adults, children, and adolescents) displayed thedomination of Late Closure by shorter reading en the verb in the main clauseof Late Closure sentences (While the crowd cheere him(,) the popular magiciansawed the woman in half.) compared to Early Closure sentences (While the crowdcheered(,) the popular magician sawed the woman in half).

In contrast, only adult readers demonstrated a Comma effect. The presenceof a comma did not shorten reading time in the disambiguating region (the verbin the main clause) for children or adolescents. The time to read Late Closuresentenc-s (comma/ + or comma/-) was not predicted to differ. The data for allreaders supported this prediction. However, the presence of a comma waspredicted to eliminate garden-path effects. The adult data demonstrated thepredicted pattern although no statistical analysis of the Closure X Comma effectachieved significance.

Educational Implications

One of the functions of the comma, to disambiguate sentences by separatingclauses, was acknowledged and used by only college-educated readers. Thissimple bit of grammatical and syntactic knowledge may impact the teaching ofreading in such a way that the comprehension of less skilled readers will beenhanced significantly.

7

6

References

Adams, B.C. & Royer, J.M. (1993). Parsing strategies in sentence parsing:A developmental study. Presentation at the Society for Research in ChildDevelopment (SRCD), New Orleans, LA.

Adams, B.C. & Royer, M. (1995). Developmental differences in sentenceprocessing strategies. Presentation at the Society for Research in ChildDevelopment (SRCD), Indianapolis,

Table 1

An exam le sentence in theJour conditions

Condition Sentence

Early Closure/No Comma While the crowd cheered the popular magician sawed the woman in half.

Early Closure/Comma While the crowd cheered, the popular magician sawed the woman in half.

Late Closure/No Comma While the crowd cheered bim the popular magician sawed the woman in half.

Late Closure/Comma While the crowd cheered him, the popular magician sawed the woman in half.

45 0

Figure 1. Mean Reading Time of the Verbin the Main Clause for Adults

444 444444 44,444444444444444444444444444,4444444444444444444444444444444,44,444444444444444444444444444444444444444444444444,4444444444444444444444444444444444444444+,4444444444444444444.44444444

444 444444 44,404 44444, 444444 444444444444444444,4444,444444444444444444444,4444444444444444444440444444444444444444,444444444444444444444444444444444444444444444444444444444444444444444444444444444444444444440444444444444444444044444444

;x14 6C4

(3)

0.)

;.4

co

4054

C.) ;84

E

c4c.D4MA

Ft 0)Ci)

(1)

Co)

.t5.1

cD

cl) 4

0.1) P.64PRI

` "101%4444444kkak.....................................................

C.)

4I-0-:r

: :::::::::::::::::::::::::::::::::::::::: PLV

4,44044444,++,++441.4)+4)+++10**44+++A04.4444++44.44.49.4).0.#4)40+4)44+410**44.11.444444404)..++44+.4#44+4,4414+411)40.44+++++14+4+41044,444444. (a)

**+4+4)41141)44444)+++++40+4,444+++4,44.44+4-+94,4***944404)+4+49+#444044.4.4)+944094+404.404*

C\2r4

toto

()awl) attu &uipta ueatv

C)

r.

a)ti.44

ci) pc,

464 Cul

E :r4

ttk0 4:24

0 L.)

raia)

r-4

14 41 ti II

F---

tr)

+4.4,404+404.41++444+at4+.......440.444414+++

4'441.4144,40++++++++++++.4++++.0.440,.......0140..1+404.++++++++++++.++++++4-.44.4404,40++++4.+.1+.+.4.410+.1004044+40.4+ 114.4).....+4,40..++.41.10..+4.+.44.++1 rg ra+.441+444++.44.40+.4.41++++++.+.43+441+ +++.44.......44.4++++++.4+.404.4+.+++.1...04+4....+011.40.4.+4++.0.+++4 ^++....0.4+40+1

InIfl

(aastu) attu iuipnj utlatv

rva Au. fmaia.A.11..

WWWWWWWWSWWWWWVIIMSNWNSWilM%V*VVVVVVVV*VVVVVVVVVVVVVVVVVVV**VVVVVVWVVWWWWVVVVVVVVVVVVVV*VVVVVVVVVVVVVNwww*VVVVVV*VWVVVVVVVVVVVVVVVVVVVVVVVV40WWWWWW**VVVVVVVVVVVVVVVWVVVVVVVVV*

attallaelagglaltallagall

4444** ++.40.40+++.0404444444.+444404.tr4.40+I++40*******44.0+++440444++++++::::It* +,* 40.44,,,,...,444+4444,444++...+4to

4444414+++40.4044.,44+4.44++.4140.+444.....4.+++.444.44..<044+******44...444444..4.+4441044,4444+404...444.4.4444.41404.04.414444,44.440,4+.....40.0.4.

+++444.+4.40-04044.44.4++++...40404,.....*...+40............+++.4404-014++0+++++44+404.4.4.444414..44440.#4444**41444.400.....44444+.44.**.+4.40444.++4444 44444.41444..4.41404444.44.440444++++++40414++++++.....44...44.44400044+++44041.0411+...

(aasuu) auta uuNAT