Age-of-acquisition effects in native speakers and second-language learners

Contrasting effects of age of acquisitionin lexical decision and letter detection

EGBERT M. H. ASSINK, SONJA VAN WELL, ANDPAUL P. N. A. KNUIJTUtrecht University

The American Journal of PsychologyFall 2003, Vol. 116, No. 3, pp. 367–388

Content in the AJP database is intended for personal, noncommercial use only.You may not reproduce, publish, distribute, transmit, participate in the transferor sale of, modify, create derivative works from, display, or in any way exploitthe AJP content in whole or in part without the written permission of thecopyright holder.

To request permission to reprint material from The American Journal ofPsychology, please find us online at:

http://www.press.uillinois.edu/about/permission.html

or email us at:

[email protected]

© 2003 by the Board of Trustees of the University of Illinois

AMERICAN JOURNAL OF PSYCHOLOGYFall 2003, Vol. 116, No. 3, pp. 367–387© 2003 by the Board of Trustees of the University of Illinois

Contrasting effects of age of acquisitionin lexical decision and letter detectionEGBERT M. H. ASSINK, SONJA VAN WELL, andPAUL P. N. A. KNUIJTUtrecht University

Two experiments studied attention in beginning and skilled readers of Dutchto letter information in function words and content words. Early and late ac-quired nouns and function words were presented to third-grade students andskilled adolescent readers. Target words were presented in short story contexts,as in the study of Greenberg, Koriat, and Vellutino (1998). Target nouns werematched on word frequency. Predictions of the structural account hypothesisof letter detection (Koriat, Greenberg, & Goldshmid, 1991) were confirmed.No age-of-acquisition effect was found. In contrast, a separately conducted lex-ical decision experiment using the same content word stimulus sets showedshorter decision latencies for early acquired words. The combined results sug-gest that during silent reading, when attention is focused on meaning, phono-logical processes may play a less prominent role than in lexical decision tasksthat demand explicit control of phonological codes. The letter detection re-sults confirmed predictions of the structural account hypothesis for both be-ginning and skilled readers. Taken together, these studies demonstrate thatphonological processes in silent reading may play a less prominent role andthat the structural account of letter processing is valid for languages other thanHebrew and English but probably is not the unique mechanism involved inletter detection.

A well-documented phenomenon in reading research is the missing-letter effect (Corcoran, 1966; Healy, 1976; Healy & Drewnowski, 1983;Read, 1983). This effect is obtained in letter detection tasks. In a letterdetection task the participant is instructed to read a printed text andto encircle a specified target letter during silent reading. A consistent-ly found result is that letters in short, highly frequent words such as the,with, and for are missed more often than letters in less frequent wordssuch as ball and sun.

Several attempts have been undertaken to explain missing letter de-tection phenomena (Healy, 1976; Drewnowski & Healy, 1977; Healy &Drewnowski, 1983). Healy and colleagues have developed an accountknown as the unitization theory. In the unitization theory it is assumedthat word frequency is the major underlying factor for explaining miss-

03.367-388.AJP116_3 10/29/03, 3:30 PM367

368 assink et al.

ing letter effects. The theory proposes a hierarchy of processing mech-anisms, such as the letter feature level, the letter level, the word level,and the sentence level. High-frequency words are recognized as a whole,whereas unfamiliar or rare words are broken down into their compo-nent letters. If recognition is successful at a given higher level, readersstop processing lower-level units and proceed with the next word orword group.

Several studies support the unitization theory (e.g., Proctor & Healy,1985; Healy, Oliver, & McNamara, 1987). Studies manipulating percep-tual factors such as introducing letter case alternation (Drewnowski &Healy, 1977) and presenting text in unfamiliar letter type (Schneider& Healy, 1993) showed better letter detection performance for high-frequency words.

Koriat and coworkers have challenged unitization theory (Koriat &Greenberg, 1991; Koriat, Greenberg, & Goldshmid, 1991; Koriat &Greenberg, 1994). The core of their criticism of the unitization theoryis that it explains letter detection phenomena at the isolated whole wordlevel, whereas letter detection occurs during reading of continuous text.Koriat and his coworkers claimed that syntactic processes play a majorrole during the processing of orthographic information (Koriat et al.,1991). Their alternative theory, known as the structural account, ex-plains selective attention to letters in text primarily by linguistic prop-erties of words in sentences.

According to the structural account, during reading for meaning bothsyntactic structure and meaning are coded. In the early stages of pro-cessing the focus is on structure, then attention shifts to meaning. Inthe first stage function words such as prepositions, articles, and conjunc-tion particles play a major role. They trigger syntactic structure, andwhen this is accomplished the reader’s efforts shift to comprehension.In the early stages attention is focused on function words; if structurehas been completed, attention shifts to content words. In this stagefunction word information is less accessible.

To reconcile the competing accounts discussed here, Moravcsik andHealy (1995) focused their research on the effect of meaning on letterdetection. In a series of six experiments they tested the hypothesis thatthe amount of time spent processing a given word is a major determi-nant of letter detection errors on that word. This hypothesis, which hasbecome known as the processing time hypothesis, is used in the presentstudy as the framework for investigating the role of age of acquisition(AOA) during letter detection. The processing time account of letterdetection may be considered an extension of the unitization account andelegantly explains syntactic factors and interaction effects of word scram-bling and word function. Thus, the structural account and the updated

03.367-388.AJP116_3 10/29/03, 3:30 PM368

contrasting effects of age of acquisition 369

version of the unitization account may be regarded as complementaryrather than competitive. For this reason the letter detection experimentsreported in the present study focus on these complementary aspects.

More recent research on letter detection has focused on further ver-ification of the structural account in relation to developmental issues.In a study using elementary school children, Greenberg, Koriat, andVellutino (1998) tried to verify whether well-established missing-lettereffects found in skilled readers can already be found in younger read-ers. A related question was how missing-letter effects develop with ageand increasing reading skill. Developmental trends in letter detectionhave been reported by Drewnowski (1978, 1981). In a study with firstthrough fifth graders and college students it was found that older par-ticipants missed more letters in function words than younger partici-pants. Additionally, among first-grade readers, the good readers weremore likely to exhibit the missing-letter effect than were poor readers.Moreover, scrambling the words within a passage reduced the error ratejust for the good readers. Greenberg et al. (1998) attributed these find-ings to the developing skill of efficient reading for meaning.

The results of Greenberg et al. (1998) support the structural accountof letter detection. In their discussion, the authors pointed out that theirresults should not be considered as a final word in the discussion of thisissue. Additional questions have to be addressed. Two of these are pre-sented in this article: the role of word frequency, age of acquisition(AOA) of words, and their interrelationship. As to word frequency,Greenberg et al. (1998) correctly remarked that their second experi-ment also provides evidence in favor of the unitization theory, for thesimple reason that high-frequency content words produced also moreomission errors than low-frequency content words. Unlike in Healy(1976), this effect was found in connected text. Moravcsik and Healy(1995, Experiment 6) also showed a large effect of word frequency inconnected text with syntactic function controlled in their comparisonof the test words you, thou, and they. More work is needed to furtherdisentangle developmental changes accounted for by joint effects ofincreased unitization of high frequency words and increased unitizationof function words in order to establish structural frames in reading.

The AOA concept is directly associated with development and orga-nization of the mental lexicon. In combination with word frequency,AOA is a theoretically powerful concept for a deeper and more analyt-ic account of the development of word decoding mechanisms. AOA hasbeen studied extensively, and recently a number of interesting studieshave been published (Gerhand & Barry, 1998, 1999; Morrison & Ellis,1995; Ellis & Morrison, 1998; Turner, Valentine, & Ellis, 1998; Morrison,Ellis, & Quinlan, 1992).

03.367-388.AJP116_3 10/29/03, 3:30 PM369

370 assink et al.

Including acquisition age in letter detection studies is interestingbecause contrasting AOA effects in lexical decision and letter detectionmight well provide new information about how phonological process-es are involved during silent reading. Effects of acquisition age havebeen reported mainly in studies using lexical decision (both visual andauditory), picture and word naming, word reading, and memory tasks.In this type of task the participant performs a specific mental operationon the target word (e.g., judging lexical status, reading aloud, remem-bering). In contrast, during letter detection the participant processescontinuous text for meaning while checking the presence of a specifiedtarget letter. Thus, it is possible that during letter detection attentionalfocus is less on phonological aspects of the target word than on seman-tic aspects of the text.

How do phonological processes work during silent reading and let-ter detection? Healy (1994) described two ways in which phonologicalprocesses might be involved in letter detection tasks: phonological pro-cesses needed for lexical access and phonological processes needed forstorage in short-term memory. Both processes might be involved inprocessing of letters in early- and late-acquired words. The question ofinterest here is how. We tested the hypothesis that early-acquired wordsare accessed with more facility, in accordance with the phonologicalcompleteness hypothesis (Brown & Watson, 1987). As a consequenceof this facilitated access, little time is necessary to process the meaningsof early-acquired words. Because of the minimal effort these words take,more letters will be missed, as predicted by the processing time hypoth-esis of Moravcsik and Healy (1995). In sum, early-acquired words areeasy to access, so less time is needed to process their meanings, whichresults in higher letter detection error rates. Thus, completeness ofphonological codes in early-acquired words mediates or facilitates se-mantic access during reading for meaning.

Do the structural and the unitization accounts predict different out-comes for letter detection experiments manipulating AOA? Yes. Bothaccounts differ importantly in the locus of the effects under discussion.In the structural account these effects occur after lexical access is com-plete, whereas in the unitization account these effects occur duringlexical processing. The structural account does not predict age of ac-quisition effects because no structural (syntactic) factor is involved, andit locates the processes as postlexical. In the unitization account, AOAeffects are possible, under the condition that we allow for a mechanismdescribed in the previous paragraph. Therefore, the unitization accountallows for these effects if phonological processes are involved in a sim-ilar way as in lexical processing.

In letter detection, are the effects of AOA similar to the effects report-ed in the earlier-discussed studies using lexical decision, naming, and

03.367-388.AJP116_3 10/29/03, 3:30 PM370


memory tasks? There are two possibilities. If in letter detection phono-logical processes are involved in a way similar to the way they are usedduring lexical decision, AOA effects will be found. If that is the case, itmay be expected that letters in early-acquired words will be missed moreoften than letters in late-acquired words. However, if in the silent read-ing the use of phonology is minimal, a widely held theoretical position(cf. Ellis, 1993, p. 36), no effect of AOA is predicted. A major goal ofthe current study was to find evidence for one of the competing pre-dictions.

In line with Greenberg et al. (1998), the focus is on how in functionand content words AOA and word frequency affect letter detection inparticipants of varying levels of reading development. We attempted toreplicate effects reported earlier by Greenberg et al. (1998) with Dutchparticipants and Dutch word materials. To summarize, six questionswere investigated in the three experiments reported here. First, doesthe structural role of function words in text contribute to the missing-letter effect in Dutch? Second, can it be shown that the missing-lettereffect in Dutch increases with reading skill? Third, does AOA contrib-ute to the missing-letter effect if word frequency is controlled for?Fourth, does the effect of AOA increase with growing reading skill?Fifth, does word frequency affect letter detection if AOA is controlledfor? Sixth, and finally, does the effect of word frequency increase withreading development?

EXPERIMENT 1

Experiment 1 was conducted to verify whether in a lexical decisiontask setting reliable AOA and word frequency effects could be obtainedusing the content word stimulus sets of the letter detection words re-ported in Experiments 2 and 3. In accordance with previous research(Gerhand & Barry, 1999), effects for both factors were predicted.

METHOD

Participants

A group of 30 university students, all skilled readers of Dutch, participatedin this lexical decision experiment. All participants had normal or corrected-to-normal vision.

Materials

The sets of early- and late-acquired content words (Appendix A) and high-and low-frequency content words (Appendix B) were the same as used in theletter detection experiment (Experiment 2). Experiment 1 was conducted toverify the presence of well-established baseline effects reported for AOA and

03.367-388.AJP116_3 10/29/03, 3:30 PM371

372 assink et al.

word frequency in the literature (e.g., Brysbaert, Lange, & Van Wijnendaele,2000; Turner et al., 1998).

Procedure

A computerized lexical decision task using Micro Experimental Laboratorysoftware (Schneider, 1988) was presented on an IBM PC using all experimen-tal content words (see Appendixes A and B) and 40 additional filler words as“yes” items. This set was completed with an equal number of pronounceablepseudoword items, based on real words and changing one phoneme. Wordsappeared in random order centrally on the screen in lowercase bold using 12–point Helvetica font. Each stimulus item was preceded by a fixation point, whichwas replaced by the stimulus item after 200 ms. There was a 500-ms delay be-tween the participant’s response and the onset of the next trial. Nonwords werematched to the words for letter length. The participants sat at a comfortablereading distance from the screen (approximately 65 cm). Reaction times weremeasured from the onset of each word.

RESULTS

Reaction time on correct decisions was used as dependent variable.As expected, an analysis of variance of the reaction time data with ac-quisition age (early or late) and frequency (high or low) as within-par-ticipants factors revealed significant effects of both acquisition age F(1,29) = 16.45, p < .001 (523 ms for early-acquired words and 576 ms forlate-acquired words) and frequency F(1, 29) = 17.23, p < .001 (540 msfor high-frequency words and 581 ms for low-frequency words). Becauseage of acquisition and word frequency were not factorially manipulat-ed, no interaction effect could be established.

EXPERIMENT 2

METHOD

Participants

Two groups participated in the experiment: 51 third-grade pupils (8–9 years)and 49 adolescent final grade grammar school students (17 years). All partici-pants volunteered for the experiment. All participants were native speakers ofDutch; they had normal reading performance skills and had good to normalvisual discrimination. A premium for best performance was announced at theintroduction of the experimental task.

Materials

Following the design used in the Greenberg et al. (1998) experiment, thestimulus set consisted of function words and content words, presented in storycontexts. All words were selected using the CELEX lexical database of English

03.367-388.AJP116_3 10/29/03, 3:30 PM372


and Dutch (Burnage, 1990; Baayen, Piepenbrock, & Van Rijn, 1993). All tar-get words had a final letter t. Two Dutch function words were used: the con-junction “WANT” and the preposition “NAAST.” Both words are in the earlyAOA range (Kohnstamm, Schaerlaekens, De Vries, Akkerhuis, & Froonincksx,1981).1 Both function words were paired six times with an early-acquired con-tent word with final t and six times with a late-acquired content word with finalt. This resulted in two sets of 12 function–content word pairs. Each contentword occurred only once in the text. The number of letters was kept constantacross word pairs. Mean frequency for early-acquired words did not differsignificantly from that of late-acquired words. Appendix A presents a completelist of the stimulus set used.

For each function word (“NAAST” and “WANT”) the 12 function–contentword pairs were incorporated in the text of two short stories, six pairs in eachstory. In each story there were three pairs with an early-acquired word and threepairs with a late-acquired word. This resulted in four short stories: “NAAST-1,”“NAAST-2,” “WANT-1,” and “WANT-2.” Each story contained 30 sentences. Inaddition to the sentences containing the six target function–content word pairs,each story contained 12 filler sentences with a single target letter t but not intarget words and 6 sentences containing no target letter. Besides this, otherinstances of the target letter t occurred in the sentences containing the targetfunction–content word pairs.2

For each of the four stories a misplacement version was made. In a misplace-ment version the function word and its corresponding content word wereswapped, placing the function word at the place of a content word and viceversa. These versions will be referred to in this article as misplaced story ver-sions.

The sets of early- and late-acquired content words (Appendix A) and high-and low-frequency content words (Appendix A) were also used in a separatelyconducted visual lexical decision study (Experiment 1). This lexical decisionexperiment was conducted to verify the presence of well-established baselineeffects reported for AOA and word frequency in the literature (Brysbaert et al.,2000; Turner et al., 1998).

The word recognition memory test booklet (test items in Appendix B) wasused to check whether the participant had read the passages with sufficientaccuracy. This testing procedure was adopted because the focus was on verify-ing whether the target words had been identified in the text. The test bookletcontained two pages: one for the normal story and one for the misplaced textstory. The test for each story consisted of a list of 40 words, alphabetically ar-ranged in four columns. Ten words were taken exactly from the story, 10 wordswere semantically related items (e.g., pear for apple), 10 words were orthographicneighbors (e.g., mouse for house), and 10 words were completely new. Partici-pants had to identify the words they had seen in the story text.

Procedure

The story materials were used for constructing test booklets. A story book-let consisted of three pages. Page 1 contained the letter detection instruction,page 2 contained a (normal or scrambled) story version, and page 3 contained

03.367-388.AJP116_3 10/29/03, 3:30 PM373

374 assink et al.

the complementary story version. The misplaced story version was always aversion with a different function word. This resulted in four different storybooklets. These four versions were administered in about equal numbers in allparticipating groups. The task was administered in groups. The instructionswere read through in groups. Participants were instructed to read the two sto-ries in the booklet at a normal speed, focusing on comprehending the gist ofthe story. Participants were told that they would be tested for reading accura-cy after they had finished reading. They were also told that they might encoun-ter sentences with misplaced words but that they could ignore this. In addition,they were told that their task was to encircle every letter t encountered duringreading and that they were not permitted to look back in any sentence. Subse-quently they were presented with a short specimen text. The specimen alsocontained misplaced words. Finally, all instructions were summarized and afinancial reward of 15 Dutch guilders for the best participant in terms of textcomprehension was announced. When a participant had finished the task, heor she received the word recognition memory test booklet. After participantshad finished their task, they left the testing room.

RESULTS

Reading accuracy test data

Hits and false alarm rates on the word recognition test were comput-ed for both groups. Third graders detected 45% of the words (hits) andproduced 4% false alarms. For the skilled readers these values were 47%hits and 3% false alarms. Third graders and skilled readers both select-ed 17% of the semantically related words. For the orthographically re-lated foils these values were 10% and 2%, respectively.

Letter detection data

Because AOA (early or late) and word type (function or content) werenot crossed factors, the data were analyzed in two separate designs. Thefirst design focused on the effect of misplacement on the function–con-tent word pairs, a replication the Greenberg et al. (1998) study withDutch words. The second design focused on the effect of misplacementon AOA in content words. The first design is a word type (function vs.content) by story type (normal vs. scrambled) by group (third grade vs.skilled) design. The second one is a story type (normal vs. misplaced)by acquisition age (early vs. late) by group (third grade vs. skilled) de-sign. For both designs, analyses of variance were used.

The data for design 1 are presented in Table 1.A word type (function vs. content) by story type (normal vs. scram-

bled) by group (third grade vs. skilled) analysis of variance on omissionerror scores produced a significant main effect of word type, F(1, 98) =20.58, p < .001. Detection of target letters in function words was lowerthan in content words (18% omissions and 12% for content words).

03.367-388.AJP116_3 10/29/03, 3:30 PM374


There was a main effect of misplacement, F(1, 98) = 18.86, p < .001;letter detection in normal context was worse (19% omissions for nor-mal and 11% in misplaced context). There was a main effect of group,with detection performance improving with age, F(1, 98) = 24.53, p <.001 (21% omissions for beginners and 9% for skilled readers). Therewas a first-order interaction, word type (function vs. content) by mis-placement, F(1, 98) = 8.72, p < .004, showing the predicted missing let-ter effect. Single-degree-of-freedom comparisons (Keppel, 1982) wereconducted for in-depth analysis of the interaction. The effect of wordclass (difference between function words and content words) was estab-lished in the normal context condition, F(1, 98) = 20.97, p < .001, where-as this difference was absent in the misplaced context condition, F(1,98) = 0.53, ns. In addition, the effect of misplacement was present infunction words, F(1, 98) = 17.22, p < .001, whereas the same effect wasabsent in content words, F(1, 98) = 2.97, ns.

To study the effect of misplacement on AOA in content words, a sto-ry type (normal vs. misplaced) by acquisition age (early vs. late) by group(third grade vs. skilled) analysis of variance was computed. The data forthis analysis are presented in Table 2.

Table 1. Omission errors (% missed [SD]) in normal and misplaced contextfor function words and content words in beginning and skilled readers,Experiment 2

Normal version Misplaced version Difference(a) (b) (a − b)

Function Content Function Content Function Contentwords words words words words words

Beginners 34 (28) 18 (17) 17 (21) 16 (15) 17 2Skilled 16 (25) 8 (10) 7 (16) 5 (12) 9 3

Overall missing letter effect (% missed) 13.0 2.5

Table 2. Omission errors (% missed [SD]) in normal and misplaced contextfor early and late acquired words in beginning and skilled readers,Experiment 2

Normal version Misplaced version Difference(a) (b) (a − b)

Early Late Early Late Early Late

Beginners 20 (21) 17 (23) 18 (22) 14 (24) 2 3Skilled 10 (16) 17 (14) 5 (14) 5 (17) 5 12

Overall missing letter effect (% missed) 3.5 7.5

03.367-388.AJP116_3 10/29/03, 3:30 PM375

376 assink et al.

This analysis showed a main effect of group, F(1, 98) = 21.71, p < .001;letter detection on early- and late-acquired content words was worse forthe beginners or improved with reading experience (17% omissions intarget words for beginners and 7% for skilled readers). No main effectsof misplacement or acquisition age were found in this analysis, nor wasthere any interactions between these factors.

DISCUSSION

The post hoc word recognition memory test controlling for real read-ing during the letter detection task showed for both third graders andskilled readers acceptable hit and false alarm rates. At least, this may beconcluded if one realizes that the hit and false alarm performances areincidental learning rates. This a priori memory check enabled us toanalyze the detection performance results found.

Experiment 2 addressed the first three of the total set of six questionsformulated at the beginning of this article. The first question was wheth-er an effect could be found of the structural role of function words incontinuous text—in other words, whether evidence could be found forthe missing letter effect in Dutch. Our data clearly support this hypoth-esis, as is illustrated in Table 1. Letter information in function words wasdifferentially processed in normal and misplaced text. Moreover, thepredicted difference between function and content words in normalcontext disappeared in the misplaced context condition. Our dataconfirm the findings reported in Greenberg et al. (1998).

As to the second question, whether the effect increases with readingskill, no interactions were found with group membership, so our datado not permit us to conclude that the two groups of participants in ourexperiment responded differently to the experimental manipulations.Thus, unlike those of Greenberg et al. (1998), our data suggest a gen-eral missing letter effect across participants showing no developmentaleffect in letter detection. The trend in our data is that the youngerparticipants showed even a larger effect of misplacement. This is a re-markable outcome in comparison with the data reported by Greenberget al. (1998). Also remarkable is that the younger participants showedhigher omission rates for both function and content words, contrary tothe results reported by Drewnowski (1981). The reading accuracy testdata do not show any between-group differences, so the effect cannotbe explained by attentional differences between both groups.

With regard to the third question, the effect of AOA in letter detec-tion, no effect of this factor was found. This is an interesting resultbecause it allows us to specify in more detail its potential relationshipwith word frequency. The effect of word frequency on letter detection

03.367-388.AJP116_3 10/29/03, 3:30 PM376


is well documented (Proctor & Healy, 1985; Healy, Oliver, & McNama-ra, 1987).

Turner et al. (1998), who presented auditory and visual lexical deci-sion tasks to skilled readers, reported an interesting finding in this re-spect. In the auditory tasks only effects of AOA were found, whereas inthe visual experiments effects of both word frequency and AOA werefound. A possible explanation for this intriguing finding is that wordfrequency, as compared with AOA, is strongly associated with the printedcode (cf. Stanovich & West, 1989). On the other hand, AOA appearsto be more directly associated with phonology. Brown and Watson(1987) proposed the phonological completeness hypothesis of the ef-fect of AOA, which claims that early-learned words have more completephonological representations than later-acquired words. This idea in-duced us to design a follow-up experiment focusing on the complemen-tary effects of syntactic function and word frequency on letter detection,controlling for AOA.

EXPERIMENT 3

The design of this follow-up study was similar to that used in Exper-iment 2. Experiment 3 focused on two distinct aspects of letter detec-tion performance: the role of syntactic function and surface word fre-quency. To study the first aspect, high-frequency function words werecompared with high-frequency content words. To study the second as-pect, late-acquired high-frequency content words were compared withlow-frequency content words. The main difference was that in Experi-ment 3 the misplacement manipulation was omitted. The reason for thiswas that the focus of interest was on frequency effects in content wordsthat were experimentally controlled for AOA. Experiment 2 showed thatreliable missing letter effects can be found in skilled and less skilledreaders of Dutch, thus supporting the structural account of letter pro-cessing during reading of continuous text. Experiment 3 was an attemptto answer complementary questions of Moravcsik and Healy (1995)referred to earlier in this article. More specifically, the goal of Experi-ment 3 was to assess the size and scope of additional letter detectionmechanisms associated with word frequency, which cannot be account-ed for by the structural account. A letter detection experiment was con-ducted with high- and low-frequency words in connected text.

According to predictions of the structural account, letter detectionin high-frequency function words is predicted to be lower than detec-tion in high-frequency content words. As regards the effect of manipu-lating frequency in content words, the structural account does not pre-

03.367-388.AJP116_3 10/29/03, 3:30 PM377

378 assink et al.

dict an effect here, whereas the unitization account does. Thus, anydetection differences found between high- and low-frequency contentwords may be taken as evidence for letter detection mechanisms work-ing beyond the mechanisms proposed by the structural account of let-ter detection.

METHOD

Participants

Two groups of participants took part in the experiment: 53 third-grade pu-pils and 41 adolescent, final grade grammar school students. All participantswere native speakers of Dutch; they had normal reading performance skills andhad good to normal visual discrimination. None had participated in Experi-ments 1 and 2. The conditions for participation were the same as described forExperiment 2.

Materials

The stimulus set consisted of three groups of words: high-frequency func-tion words, high-frequency content words, and low-frequency content words.Stimuli were presented in story contexts. Function words were the same as usedin Experiment 2: the conjunction “WANT” and the preposition “NAAST.” Alltarget words had a final letter t. Both function words were paired six times witha high-frequency content word with a final t and six times with a low-frequen-cy content word with a final t. This resulted in two sets of 12 function–contentword pairs. Each content word occurred only once in the text. The number ofletters was kept constant across word pairs. All content words used were lateacquired. Appendix A presents a complete list of the stimulus set used. For eachfunction word (“NAAST” and “WANT”) the 12 function–content word pairswere incorporated in the text of two short stories, six pairs in each story. Eachstory contained six sentences with the target function word, three sentenceswith a target high-frequency content word, and three sentences with a targetlow-frequency content word. In addition to the sentences containing targetwords, each story contained six filler sentences with a single target letter t andsix filler sentences containing no target letter.

Procedure

The procedure followed was the same as described for Experiment 2. Theonly difference was that the story booklets did not contain a page with a mis-placed story version. Booklet versions were constructed such that a bookletalways contained the two different function words. This resulted in two possi-ble combinations with two corresponding story booklet versions. A bookletconsisted of three pages. Page 1 contained the letter detection instruction, page2 contained the text of the first story, and page 3 contained the second textversion. Booklet versions were randomly administered across groups. Instruc-tion and administration were identical to those of Experiment 2, including theword recognition memory test method used (for test items, see Appendix B).

03.367-388.AJP116_3 10/29/03, 3:30 PM378


RESULTS

Reading accuracy test data

Hits and false alarm rates on the word recognition test were comput-ed for both groups. Third graders detected 40% of the words (hits) andproduced 3% false alarms. For the skilled readers these values were 43%hits and 2% false alarms. Third graders selected 16% and skilled read-ers selected 17% of the semantically related words. For the orthograph-ically related foils these values were 10% and 4%, respectively.

Letter detection data

The results of the letter detection in function and content words andthe high- and low-frequency words are presented in Table 3.

To assess the effects of word function, a word type (function vs. con-tent) by group (third grade vs. skilled) analysis of variance with partic-ipant means as observation unit was performed. This analysis showed amain effect of group, F(1, 92) = 5.90, p < .016, with third graders miss-ing more letters (23% omission errors vs. 17% for skilled participants),and a word type effect, F(1, 92) = 36.55, p < .001. Omission error rateof target letters was 26% for function words and 15% for content words.No interactions were found. A complementary analysis of variance tak-ing item means as observation units and taking surface frequency ascovariate showed that the word function effect was accounted for byfrequency. Appendix A shows that function words have a substantiallyhigher frequency than the high-frequency content words that met theselection criteria for inclusion in the stimulus list in the experiment.

The letter detection data of high- and low-frequency content wordsare presented in Table 3. The data for high- and low-frequency wordswere analyzed using a frequency (high vs. low) by group (third gradevs. skilled) analysis of variance. This analysis revealed a significant maineffect of group, F(1, 92) = 12.53, p < .001, with third graders missingmore target letters (15% omissions vs. 9% for skilled readers). The ef-fect of frequency in content words was significant, F(1, 92) = 22.06, p <

Table 3. Mean omission error scores (% missed [SD]) in function andcontent words and in high- and low-frequency words, Experiment 3

Function words Content words

Beginners 28 (28) 18 (18)Skilled 23 (21) 11 (11)

Low-frequency words High-frequency words

Beginners 11 (16) 18 (17)Skilled 6 (12) 11 (11)

03.367-388.AJP116_3 10/29/03, 3:30 PM379

380 assink et al.

.001; the omission error rate was 15% for high-frequency content wordsand 9% for low-frequency content words. There was no interaction ef-fect with group.

DISCUSSION

The results of Experiment 3 illustrate that word frequency is an in-dependent factor contributing to the missing-letter effect. This demon-strates that unitization effects and effects of syntactic structure in coop-eration account for letter detection phenomena. This is in agreementwith more recent thinking in this domain of reading research, as ex-pressed in the processing time hypothesis (Moravcsik & Healy, 1995,1998). Thus, the outcomes of Experiment 3 confirm the convergingviews in recent research.

GENERAL DISCUSSION

Six questions were addressed in Experiments 2 and 3. The first ques-tion was whether evidence could be found for the missing-letter effectin Dutch. The results of Experiments 2 and 3 may be summarized asfollows. A comparison of function words and content words showed astrong detection difference in the direction, as predicted by the struc-tural account hypothesis (Koriat et al., 1991; Greenberg et al., 1998).Additional analyses of the source of this effect made clear that syntac-tic effects were responsible for it because misplacement one-sidedlyenhanced detection in function words. Word frequency alone cannotexplain this effect of misplacement unless the notion of unitization isexpanded beyond the single word level, that is, if the notion of phraselevel unitization is introduced (Healy, Conboy, & Drewnowski, 1987).

The second question dealt with developmental aspects in the missing-letter effect. A major finding in the present experiments was that theless skilled reader group showed the same detection patterns as theexperienced readers. Although statistically not significant, the trend inthe data was that the differences between function and content wordswere even slightly larger in the third-grader group. This is inconsistentwith the Greenberg et al. (1998) study, where a developmental trendwas found, with younger participants showing a less strong experimen-tal effect. Thus, a major conclusion from the present experiments is thateven readers who have been reading for only a few years are sensitiveto syntactic cues affecting letter processing in words.

A novel feature introduced in the present experiments was the intro-duction of AOA as a possible factor involved in letter detection. This

03.367-388.AJP116_3 10/29/03, 3:30 PM380


was the third question addressed in the present study. There are strongarguments for assuming such a relationship. Recent research has dem-onstrated that phonological factors are differentially involved in process-ing early-acquired words (Gerhand & Barry, 1998). This suggests thatearly-acquired words might be attended to differently during readingcontinuous text. Experiment 2 did not provide evidence for this, andthere was also no evidence of a developmental trend, the fourth ques-tion of interest. However, the lexical decision experiment conductedseparately with the same content word stimuli showed strong AOA andfrequency effects. These combined outcomes of letter detection andlexical decision are theoretically interesting because they provide evi-dence that in silent reading, as exemplified in the letter detection task,phonological factors play a less prominent role, as is observed duringlexical decision. We choose the words “less prominent” to avoid unwar-ranted conclusions from our findings. Other studies of letter detectionhave pointed to the role of phonological processes in that task. Forexample, Read (1983) and Schneider, Healy, and Gesi (1991) showedthat people make more errors in detecting the letter f than the letter oin the word of and attributed that result to phonological processing.Thus, our contrasting effects of AOA in lexical decision and letter de-tection should not be interpreted to imply that phonological processesare not involved during reading for meaning. Our conclusion is thatthey are differently involved and that this difference can be understoodif we look at the kind of reading task used.

In the letter detection experiment reported in this study, no cleardevelopmental trends were found. The less skilled participants showedmissing letter effects comparable to those of the skilled group. The bestexplanation we can give at present is that on closer inspection of read-ing level, our elementary school participants showed a high level ofreading performance. Inspection of reading achievement percentiledata on the annually administered national achievement test showedperformance scores in the 67th percentile. This might well suggest thepresence of fully developed syntactic frames involved in fluent reading.This might also explain why in the misplaced condition the higheromission rate in high-frequency function words disappeared for allparticipants. The powerful manipulation of swapping function andcontent words may well have focused attention on both misplaced wordsin an equal manner, with the effect of reducing omission rates to thesame level. It is difficult to determine whether swapping affected detec-tion of both word types to the same degree.

AOA effects have been reported in a variety of tasks, such as naming(Morrison & Ellis, 2000), lexical decision (Turner et al., 1998), and rec-ognition and recall (Dewhurst, Hitch, & Barry, 1998). Further contras-

03.367-388.AJP116_3 10/29/03, 3:30 PM381

382 assink et al.

tive research using multiple experimental tasks such as lexical decision,letter detection, and memory are needed to further explore the effectsreported in this study. One possibility is to study frequency and AOA inexperimental studies using letter detection tasks in combination withincidental and intentional memory tasks. If Gerhand and Barry (1998)are right in their suggestion that phonology and orthography are differ-ently involved in processing early- and late-acquired words, these exper-iments could reveal interesting new perspectives on this issue.

With regard to the effects of frequency in late-acquired content words,the fifth and sixth questions investigated in this article, the manipula-tion of word frequency in content words in Experiment 3 turned outto be successful. This demonstrates that there is no single-mechanismaccount for letter detection in reading of continuous text for compre-hension. The additional analysis using item means as observation unitsalso showed that detection of function words in normal text is affectedby both syntactic slot information and word frequency. It would be toosimple to propose a single-mechanism account of letter processingduring reading. The idea that during normal reading multiple mecha-nisms contribute to the quality of letter perception in text has beenexplicitly forwarded by a variety of authors in more recent publications(Moravcsik & Healy, 1995; Greenberg et al, 1998).

If a single-mechanism account of letter detection proves to be unten-able, it is appropriate to ask what major factors are involved in letterdetection during normal reading. At present we have evidence thatthese factors include visual aspects such as typeface (Drewnowski &Healy, 1977; Schneider & Healy, 1993), syntactic structure (Koriat et al.,1991), word frequency (Cunningham et al., 1988), and the semanticrole of words in texts (Moravcsik & Healy, 1995). A challenge for futureresearch is to show how these mechanisms in concert operate duringnormal reading.

Appendix A. Stimuli for letter detection task, Experiments 2 and 3

Frequency is surface frequency in the CELEX 1–million-word corpus, with thelog frequency, in millions, in parentheses (Baayen et al., 1993).

Age of acquisition is the proportion of teachers who claim that this particularword occurring in a spoken sentence context will be understood by a 6–year-old child (Kohnstamm et al., 1981). An “x” indicates that this word does notoccur in the list.

03.367-388.AJP116_3 10/29/03, 3:30 PM382


Function words used in Experiments 2 and 3

Stimulus Frequency Acquisition age

WANT 134,656 (2.91) 85NAAST 15,207 (2.55) 96

Early- and late-acquired nouns used in Experiment 2:

Early-acquired nouns Late-acquired nouns

Stimulus Frequency Acquisition Stimulus Frequency Acquisitionage age

RUIT 523 (1.07) 60 BUIT 504 (1.07) xCENT 756 (1.25) 97 FORT 574 (1.14) 40NEST 809 (1.27) 98 PEST 597 (1.14) xPOST 736 (1.71) 97 KUST 1,961 (1.66) 42KAST 1,518 (1.55) 92 LAST 2,429 (1.75) 52VOET 4,069 (1.98) 100 FEIT 8,592 (2.30) xKLANT 873 (1.32) 71 FRONT 816 (1.27) xPOORT 1,158 (1.43) 92 KWART 1,042 (1.39) xBORST 2,973 (1.84) 88 KOMST 1,395 (1.51) xLIJST 1,567 (1.56) 69 WINST 1,594 (1.57) xKRANT 3,041 (1.85) 98 KUNST 3,830 (1.95) 47FEEST 1,686 (1.60) 100 GEEST 7,408 (2.24) x

High- and low–surface-frequency nouns used in Experiment 3:

High-frequency content words Low-frequency content words

Stimulus Frequency Acquisition Stimulus Frequency Acquisitionage age

BUIT 504 (1.07) x MOOT 9 (0) xFORT 574 (1.14) 40 KROT 63 (0) xPEST 597 (1.14) x FRET 11 (0) xKUST 1,961 (1.66) 42 KNOT 24 (0) xLAST 2,429 (1.75) 52 KRAT 58 (0) xFEIT 8,592 (2.30) x BOUT 60 (0) xFRONT 816 (1.27) x FLIRT 41 (0) xKWART 1,042 (1.39) x SPURT 13 (0) xKOMST 1,395 (1.51) x KLUIT 57 (0) xWINST 1,594 (1.57) x PLINT 26 (0) xKUNST 3,830 (1.95) 47 SPLIT 26 (0) xGEEST 7,408 (2.24) x NOEST 1 (0) x

03.367-388.AJP116_3 10/29/03, 3:30 PM383

384 assink et al.

Appendix B. Reading accuracy control test, Experiment 2

Words from the stimulus set printed in CAPITAL LETTERS:

Words from Semantically Orthographicallythe text related related Unrelated

Story 1 BUIT roof ruit eeuwFORT vesting port vormPEST hekel post zaakKUST zee kast kerkLAST hinder list weekFEIT gegeven geit aard

MOOT stuk noot volkKROT bouwval grot reeksFRET marmot pret stijlKNOT bol snot proefKRAT bak wrat stoepBOUT moer fout storm

Story 2 FRONT voorkant klont briefKWART deel zwart filmsKOMST bezoek korst kringWINST verlies minst brandKUNST beeld gunst strafGEEST spook gerst graafFLIRT sjans shirt slangSPURT sprint sport hondKLUIT klomp fluit maanPLINT leest print stofSPLIT spleet spuit arts

NOEST kwast hoest vuur

Notes

The authors would like to thank the pupils and the teaching staff of theKohnstamm School and the Bonifatius College in Utrecht and the De WilgeSchool and the Alberdingk Thijm College in Hilversum for their enthusiasticparticipation in the experiments. Correspondence about this article should besent to Egbert M. H. Assink, Psychology Department, Utrecht University, Heidel-berglaan 2, 3584 CS, Utrecht, The Netherlands (e-mail: [email protected]).Received for publication October 25, 2001; revision received July 17, 2002.

1. This list contains the vocabulary found in elementary school reading books.A panel of elementary school teachers has judged these words. Early-acquiredwords are words judged by 90% or more of the panel to belong to the standardvocabulary (active or passive) of a 5–year-old child. Late-acquired words are

03.367-388.AJP116_3 10/29/03, 3:30 PM384


words judged to have been acquired at the age of 5 by 50% or less of the judg-es of the panel (Kohnstamm et al., 1981) or words that do not occur in theelementary school vocabulary list.

2. The text of these narratives is available from the author.

References

Baayen, R. H., Piepenbrock, R., & Van Rijn, H. (1993). The CELEX lexical data-base (CD-ROM). Philadelphia: Linguistic Data Consortium, University ofPennsylvania.

Brown, G. D. A., & Watson, F. L. (1987). First in, first out: Word learning ageand spoken word frequency as predictors of word familiarity and wordnaming latency. Memory & Cognition, 15, 208–216.

Brysbaert, M., Lange, M., & Van Wijnendaele, I. (2000). The effects of age ofacquisition and frequency of occurrence in visual word recognition: Fur-ther evidence from the Dutch language. European Journal of Cognitive Psy-chology, 12, 65–85.

Burnage, G. (1990). CELEX. A guide for users. Nijmegen, The Netherlands:CELEX.

Corcoran, D. W. J. (1966). An acoustic factor in letter cancellation. Nature, 210,658.

Cunningham, T. F., Healy, A. F., Kanengiser, N., Chizzick, L., & Willitts, R. L.(1988). Investigating the boundaries of reading units across ages and read-ing levels. Journal of Experimental Child Psychology, 45, 175–208.

Dewhurst, S. A., Hitch, G. J., & Barry, C. (1998). Separate effects of word fre-quency and age of acquisition in recognition and recall. Journal of Experi-mental Psychology: Learning, Memory, and Cognition, 24, 284–298.

Drewnowski, A. (1978). Detection errors on the word the: Evidence for theacquisition of reading levels. Memory & Cognition, 6, 403–409.

Drewnowski, A. (1981). Missing -ing in reading: Developmental changes inreading units. Journal of Experimental Child Psychology, 31, 154–168.

Drewnowski, A., & Healy, A. F. (1977). Detection errors on the and and: Evidencefor reading units larger than the word. Memory & Cognition, 5, 636–647.

Ellis, A. W. (1993). Reading, writing and dyslexia. A cognitive analysis. Hove, UK:Psychology Press.

Ellis, A. W., & Morrison, C. M. (1998). Real age-of-acquisition effects in lexicalretrieval. Journal of Experimental Psychology: Learning, Memory, and Cognition,24, 515–523.

Gerhand, S., & Barry, C. (1998). Word frequency effects in oral reading are notmerely age-of-acquisition effects in disguise. Journal of Experimental Psychol-ogy: Learning, Memory, and Cognition, 24, 267–283.

Gerhand, S., & Barry, C. (1999). Age of acquisition, word frequency, and therole of phonology in the lexical decision task. Memory & Cognition, 27, 592–602.

Greenberg, S. N., Koriat, A., & Vellutino, F. R. (1998). Age changes in the miss-ing-letter effect reflect the reader’s growing ability to extract the structurefrom text. Journal of Experimental Child Psychology, 69, 175–198.

03.367-388.AJP116_3 10/29/03, 3:30 PM385

386 assink et al.

Healy, A. F. (1976). Detection errors on the word THE: Evidence for readingunits larger than letters. Journal of Experimental Psychology: Human Perceptionand Performance, 2, 235–242.

Healy, A. F. (1994). Letter detection: A window to unitization and other cogni-tive processes in reading text. Psychonomic Bulletin & Review, 1, 333–344.

Healy, A. F., Conboy, G. L., & Drewnowski, A. (1987). Characterizing the pro-cessing units of reading: Effects of intra- and interword spaces in a letterdetection task. In B. K. Britton & S. M. Glynn (Eds.), Executive control pro-cesses in reading (pp. 279–297). Hillsdale, NJ: Erlbaum.

Healy, A. F., & Drewnowski, A. (1983). Investigating boundaries of reading units:Letter detection in misspelled words. Journal of Experimental Psychology:Human Perception and Performance, 9, 413–426.

Healy, A. F., Oliver, W. L., & McNamara, T. P. (1987). Detecting letters in con-tinuous text: Effects of display size. Journal of Experimental Psychology: Hu-man Perception and Performance, 13, 279–290.

Keppel, G. (1982). Design and analysis: A researcher’s handbook. Englewood Cliffs,NJ: Prentice Hall.

Kohnstamm, G. A., Schaerlaekens, A. M., De Vries, A. K., Akkerhuis, G. W., &Froonincksx, M. F. (1981). Nieuwe streeflijst woordenschat voor 6–jarigen [Newtarget vocabulary list for 6–year-old children]. Lisse, The Netherlands:Swets & Zeitlinger.

Koriat, A., & Greenberg, S. N. (1991). Syntactic control of letter detection:Evidence from English and Hebrew nonwords. Journal of Experimental Psy-chology: Learning, Memory, and Cognition, 17, 1033–1048.

Koriat, A., & Greenberg, S. N. (1994). The extraction of phrase structure dur-ing reading: Evidence from letter detection errors. Psychonomic Bulletin &Review, 1, 345–356.

Koriat, A., Greenberg, S. N., & Goldshmid, Y. (1991). The missing-letter effectin Hebrew: Word frequency or word function? Journal of Experimental Psy-chology: Learning, Memory, and Cognition, 17, 66–80.

Moravcsik, J. E., & Healy, A. F. (1995). Effect of meaning on letter detection.Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 82–95.

Moravcsik, J. E., & Healy, A. F. (1998). Effect of syntactic role and syntacticprominence on letter detection. Psychonomic Bulletin & Review, 5(1), 96–100.

Morrison, C. M., & Ellis, A. W. (1995). Roles of word frequency and age ofacquisition in word naming and lexical decision. Journal of ExperimentalPsychology: Learning, Memory, and Cognition, 21, 116–133.

Morrison, C. M., & Ellis, A. W. (2000). Real age of acquisition effects in wordnaming and lexical decision. British Journal of Psychology, 91, 167–180.

Morrison, C. M., Ellis, A. W., & Quinlan, P. T. (1992). Age of acquisition, notword frequency, affects object naming, not object recognition. Memory &Cognition, 20, 705–714.

Proctor, H. D., & Healy, A. F. (1985). A secondary task analysis of a word famil-iarity effect. Journal of Experimental Psychology: Human Perception and Perfor-mance, 3, 286–303.

03.367-388.AJP116_3 10/29/03, 3:30 PM386


Read, J. D. (1983). Detection of F’s in a single statement: The role of phono-logical recoding. Memory & Cognition, 11, 390–399.

Schneider, V. I., & Healy, A. F. (1993). Detecting phonemes and letters in text:Interactions between different types and levels of processes. Memory &Cognition, 21, 739–751.

Schneider, V. I., Healy, A. F., & Gesi, A. T. (1991). The role of phonetic pro-cesses in letter detection: A reevaluation. Journal of Memory and Language,30, 294–318.

Schneider, W. (1988). Micro Experimental Laboratory user’s guide: Computer tech-niques for real time experimental experimentation. Pittsburgh: Psychology Soft-ware Tools, Inc.

Stanovich, K. E., & West, R. F. (1989). Exposure to print and orthographicprocessing. Reading Research Quarterly, 24, 402–433.

Turner, J. E., Valentine, T., & Ellis, A. W. (1998). Contrasting effects of age ofacquisition and word frequency on auditory and visual lexical decision.Memory & Cognition, 26, 1282–1291.

03.367-388.AJP116_3 10/29/03, 3:30 PM387

x

03.367-388.AJP116_3 10/29/03, 3:30 PM388

Age-of-acquisition effects in native speakers and second-language learners

Documents

Transcript of Age-of-acquisition effects in native speakers and second-language learners