processing of morphological and semantic cues in russian ... · 134 kempeandmacwhinney...

Processing of Morphological and Semantic Cues inRussian and German

Vera KempeState University of New York at Oswego

Brian MacWhinneyCarnegie Mellon University

This study examines the on-line processing of morphological cues to sentenceinterpretation in Russian and German with the goal of evaluating the relativeimpacts of cue availability and cue reliability Both Russian and German usethe cues of word order animacy case-marking and subjectndashverb agreementto identify the agent of active transitive sentences However the availabilityof the case-marking cue is higher in Russian than in German Using apicture-choice paradigm we contrasted case-marking and animacy inRussian and German The reaction times showed larger effects of case-marking in Russian than in German and effects of animacy in German butnot in Russian These results suggest that the higher the availability of a cuethe larger the processing benets associated with the presence of this cue andthe smaller the impact of other converging information A recurrentcascaded backpropagation network was designed to simulate these effectsThe network succeeded in capturing the essential language differences in thereaction times thereby illustrating how the statistical properties of cues in alanguage can affect the time-course of activation of alternative interpreta-tions during sentence processing

INTRODUCTION

Crosslinguistic studies of sentence processing have documented pervasivedifferences between languages (Cuetos amp Mitchell 1988 Frazier ampdrsquoArcais 1989 Vigliocco Butterworth amp Semenza 1994) The Competi-

Requests for reprints should be addressed to Vera Kempe Department of PsychologySUNY Oswego Oswego NY 13126 USA E-mail kempeoswegoedu

This research was supported by a fellowship from the German Academic Exchange Board(DAAD) to the rst author We would like to thank Don Mitchell Roman Taraban and veanonymous reviewers for helpful comments on an earlier draft of this paper

c 1999 Psychology Press Ltd

LANGUAGE AND COGNITIVE PROCESSES 1999 14 (2) 129ndash171

130 KEMPE AND MACWHINNEY

tion Model (MacWhinney amp Bates 1989) has been proposed as a way ofrelating the observed processing differences to variations in languagestructure In particular the Competition Model focuses on the role ofsurface cues such as word order noun animacy subjectndashverb agreementand noun case-marking The strength of a cue is viewed as depending onthree factors (i) its availability dened as the proportion of times a cue ispresent and can be used for accessing the underlying function (ii) itsreliability dened as the proportion of times a cue signals the correctinterpretation given that it was present and (iii) its cost which depends onthe perceptual salience of the cue and the load it places on workingmemory

Studies within the Competition Model framework have examinedprocessing in over 15 languages (Bates McNew MacWhinney Devescoviamp Smith 1982 Kail 1989 MacWhinney Bates amp Kliegl 1984MacWhinney Pleh amp Bates 1985 McDonald 1986 Sokolov 1989) Ina typical Competition Model experiment subjects are presented withsimple transitive sentences and are asked to decide which noun refers tothe agent of the sentence A consistent nding from these studies is that ifthe cues for agentivity are placed in competition with each other thechoice of one of the nouns as the agent can be modelled by multiplicativecue integration (McDonald amp MacWhinney 1989) In these off-linestudies it has been found that the primary determinant of cue strength iscue reliability In other words when subjects are given enough time topermit full deliberate consideration of all competing cues and interpreta-tions they integrate cues in a way that maximises the probable correctnessof their nal interpretation

Newer work in the Competition Model framework (Hernandez Batesamp Avila 1994 Kilborn 1989 Li Bates amp MacWhinney 1993 McDonaldamp MacWhinney 1995 Mimica Sullivan amp Smith 1994) has attempted toextend the earlier off-line models to the study of on-line sentenceprocessing The paradigm used in many of these studies is a speeded cross-modal picture-choice task where participants listen to a transitive sentencewhile making a choice between two candidate agents that are presentedvisually Both candidates are named in the sentence but only one can beselected as the agent Participants are encouraged to respond even beforethe sentence is nished if they are condent of their interpretation Theuse of this speeded picture-choice task provides insights into how thecomplex interaction between various cues unfolds over time There aretwo ways in which the results from this on-line technique further supportthe emphasis on cue reliability that emerged from the earlier off-linestudies First it has been found that the strongest and most reliable cueslead to the fastest reaction times Second it has been found that any typeof competition or disagreement between cues results in inhibition and

LANGUAGE DIFFERENCES IN ON-LINE PROCESSING 131

slower reaction times These two effects are well-predicted by a simplemultiplicative cue integration model in which cue strength is largelypredicted by reliability

However there are two other effects found in these on-line studies thatpoint towards the need for a more complicated model First certain cueshave a more pronounced effect on the reaction times than on the choice ofone of the nouns as an agent Second reaction times are not as clearlyaffected by cue convergence as are patterns of agent choice It appears thatstrong cues tend to saturate the on-line processing system so thatproviding additional evidence when a strong cue is already present haslittle additional effect on reaction times There may also be a trade-offbetween the benets gained from obtaining more evidence from aconvergent cue and the costs associated with processing this cue (Kail1989 Mimica et al 1994)

We noted earlier that the Competition Model relates cue strength to twoseparate cue validity factors reliability and availability Reliability is ameasure of cue consistency and dependability whereas availability is ameasure of frequency There is some reason to believe that the frequencyof a syntactic structure has a stronger effect upon on-line role assignmentprocesses than upon off-line interpretation In related work frequencyeffects have been demonstrated for relative clause attachment (Cuetos ampMitchell 1998 Mitchell amp Cuetos 1991 Mitchell Cuetos amp Zagar 1990)as well as for the processing of syntactically ambiguous lexical items underthe inuence of verb frame structures (MacDonald Pearlmutter ampSeidenberg 1994 Trueswell Tanenhaus amp Garnsey 1994 Trueswell1996) This related literature would lead us to suspect that frequencywould also play a major role in determining agent role assignment In factthe model developed by MacDonald et al (1994) corresponds quite closelyto one offered by MacWhinney (1987) In both accounts alternativeattachment structures compete in terms of alternative lexical forms orhomonyms The candidacy of each competing attachment is supported bycues that vary in strength MacDonald et al (1994) link cue strength moreto frequency whereas the Competition Model has often linked it more toreliability It would be interesting if this contrast reects the greateremphasis given to on-line data by MacDonald et al (1994) or Mitchell etal (1990) If so we would expect to nd a stronger effect of frequency onCompetition Model experiments when they include a stronger on-linefocus

It is possible to disentangle the effects of frequency and reliability byutilising the fact that many languages have fully grammaticised cues thatare completely reliable but not fully available The specic inectionalparadigm of the language determines the extent to which a fully reliablecue is also available in particular sentences A comparison of two


languages in which reliable morphological cues differ only in availabilityallows us to measure the impact of availability during on-line sentenceprocessing In this study we will compare on-line processing in Russianand German two languages which have an identical repertoire of cues foragent role assignment but which differ with respect to the availability ofmorphological cues If processing differences between these two languagescan be found this will provide evidence for differential effects of cueavailability during on-line sentence processing

The distribution of cues in Russian and German

The present study focuses on the cues for agentivity in simple transitivesentences that take the form non-verb-noun (NVN) A typical sentence ofthis type is lsquoThe boy kicked the ballrsquo This sentence type was chosenbecause it is the one used in many crosslinguistic studies that have beenconducted in the past

The cues provided in Russian and German for determining who didwhat to whom in a simple active transitive sentence are word orderanimacy case-marking and subjectndashverb agreement We will rst describeeach of these four cues and then provide quantitative information abouttheir availability and reliability in Russian and German

Word Order The ordering of the nouns and the verb in a sentence canprovide information about which noun is most likely to be the agent It hasbeen shown for example that positioning of a noun before the verb is avery strong cue for agentivity in English (MacWhinney et al 1984) InGerman there is a strict rule that requires that the tense-bearing verbshould appear in second position If tense is marked by an auxiliary verb itis this verb that appears in second position and other verbs will appear innal position As long as the tense-second rule is obeyed German permitsvariation in the placement of the nouns and other arguments around theverb In particular both SVO (subject-verb-object) and OVS (object-verb-subject) orders are permitted in German main clauses although SVOorder is much more common than OVS order The selection of either SVOor OVS is based on pragmatic factors Because of the verb-second ruleGerman does not typically permit SOV or OSV in main clauses Howeverthe order AVSO can arise when an adjunct (adverb participle orprepositional phrase) is placed in rst position In such cases AVSO isstrongly preferred over AVOS Russian on the other hand permits all sixbasic word orders (SVO OVS VSO VOS SOV OSV) and the selectionof any one of these orders is based on pragmatic factors For the SVO andOVS sentences that will be the focus of the current study both languagesmake a choice between the two orders on the basis of pragmatic factors


Animacy There is a general tendency in all languages for agents to beanimate In a sentence with two nouns that differ in relative animacy theanimacy cue can be used to favour the selection of the most animate nounas agent In most sentences the availability of the animacy cue tends toconverge with and blur into the lsquoprobable eventrsquo cue since the mostanimate noun is usually most likely to be the actor However CompetitionModel studies that have been devoted to disentangling the relative effectsof the two cues indicate that animacy is the dominant cue (Bates et al1982 Bates MacWhinney Caselli Devescovi Natale amp Venza 1984) Inthe current study we will be examining a fairly rigid form of the animacycue since we will not be including a large range of noun-verbcombinations As Corrigan (1986 1988) has shown animacy informationinteracts strongly with the details of the relations between specic nounsand specic verbs However the details of these interactive processes lieoutside the scope of the current study

The availability of animacy contrasts depends on the pragmatics of thediscourse context and is unlikely to differ signicantly across languagesHowever it has been shown that languages differ markedly in terms of theactual reliance that they place on the animacy cue when it is presentLanguages such as Italian Spanish and German (eg Bates et al 1982MacWhinney et al 1984 Hernandez et al 1994) place a greater relianceon animacy than languages such as English and Hungarian In the case ofEnglish it appears that the strong word order cue completely obviates theneed to rely on additional information from animacy Similarly inHungarian the strong case-marking cue makes animacy relatively lessimportant In Italian Spanish and German on the other hand theoccasional absence of some other highly determinant cue increases the rolethat the animacy cue can play during sentence processing

In order to fully evaluate the evidence provided by an animacy contrastthe listener has to process the whole clause However in the speededpicture-choice paradigm the animacy of each individual noun cancontribute incremental evidence to the ongoing decision process Becauseof this possible incremental use of animacy our corpus analyses code thereliability of the animacy cue in terms of its effect on the choice ofindividual nouns as agent

Case-marking Unlike a pragmatic cue such as animacy the case-marking cue is a fully conventionalised cue that can independently signalthe case role status of each noun phrase Although this cue is fullygrammaticised it may often be ambiguous or absent In German case-marking varies according to the number and gender of the noun Theactual markings are mainly expressed on the articles or adjectives thatprecede the noun For some cases and some nouns case is also marked on


sufxes In Russian case-marking varies according to the number genderand animacy of the noun The forms used to mark case appear exclusivelyas sufxes that follow the noun Although both languages possessnominative and accusative marking there are signicant differences inthe patterns of neutralisation within the two declensional paradigms Thesedifferences are crucial to the current study because they account for thedifferential availability of nominative and accusative marking in Russianand German In German nominative and accusative are neutralised infeminine (die-die) neuter (das-das) and plural (die-die) nouns Thus inorder to reliably indicate either nominative (der) or accusative (den)marking at least one masculine noun has to be present in the sentence InRussian nominative-accusative neutralisation occurs mainly when thereare masculine inanimate or neuter nouns which end in a consonant Inthese nouns genitive dative instrumental and locative are marked bydifferent vowel sufxes whereas nominative and accusative both end in anull-morpheme Additionally feminine nouns ending in an end-palatalisedconsonant (eg matrsquo)1 also exhibit nominative-accusative neutralisationSince end-palatalised feminine nouns are quite infrequent and mostRussian neuter nouns are inanimate nominative-accusative neutralisationmainly co-occurs when both nouns are inanimate However text counts(Zubin 1977 1979) show that sentences with two inanimate nouns and atransitive verb are rather rare in languages These different neutralisationpatterns and the ways in which they interact with animacy congurationsare responsible for the lower availability of the case-marking cue intransitive sentences in German as compared to Russian

Subject-verb agreement Subject-verb agreement is the other morpho-logical cue that can provide reliable evidence for agentivity In bothRussian and German the verb agrees in number and person with thesubject of the sentence In Russian singular nouns and past tense verbsalso agree in gender The fact that Russian requires agreement on moremorphosyntactic dimensions reects a generally greater reliance on theagreement cue in Russian as compared with German

Corpus-based estimations In order to quantify our assumptionsregarding the differential availability of morphological cues in Russianand German we extracted samples of active transitive sentences from avariety of sources and coded the presence or absence of each of these cues

1 End-palatalisation of consonants in Russian is graphemically expressed by the soft-signwhich is transcribed by an apostrophe In the transcription of Russian we follow thetransliteration rules used by Comrie and Corbett (1992)


For each language a corpus of 250 sentences was selected fromcontemporary novels newspaper editorials and childrenrsquos books Thereferences are listed in Appendix 1 For each source several pages wereselected at random from which all sentences containing a single transitiveverb were coded In Russian negative sentences were excluded becausethese sentences permit both accusative and genitive marking of the directobject We also excluded sentences with null subjects Since subjectomission is more frequent in Russian it was necessary to consult asomewhat larger text corpus in order to nd a matching number ofqualifying sentences

We are aware of the fact that written language is not necessarily anaccurate reection of the distributional characteristics of the entirelanguage Unfortunately large samples of adult oral speech are not yetavailable2 In an oral corpus we would expect more word order variabilityHowever since the variation between SVO and OVS is determined bypragmatic or discourse factors in both languages this cue should affectboth languages to a similar degree We would also expect more subjectellipsis particularly in Russian However the omission of the subject doesnot affect the reliability or availability of the subjectndashverb agreement cueThere is little reason to believe that animacy effects are different in the twocultures Finally the reliability and availability of the case-marking cue isbased on distributional effects that should be similar in written and oralcorpora Although written corpora give us a good initial idea of the relativeavailability and reliability of these cues it is clear that for future researchit would be desirable to base availability and reliability estimations oncorpora of oral speech as well

Since we are interested in the immediate effects of cues during sentencecomprehension animacy information and case-marking were coded aslocal cues for each individual noun The ten specic cues that we trackedwere

1 animacy of the rst noun (N1-Animacy)2 inanimacy of the rst noun (N1-Inanimacy)3 animacy of the second noun (N2-Animacy)4 inanimacy of the second noun (N2-Inanimacy)5 nominative marking of the rst noun (N1-Nominative)

2 A good example of the use of oral corpora in psycholinguistic research can be found in theCHILDES database of child language and adult second language samples (MacWhinney1995) where spoken interactions from over 25 languages including Russian and German areavailable Unfortunately crosslinguistic studies of adult language production have notdeveloped a similar focus on the collection transcription and analysis of spoken corpora Thisis clearly a priority for future development in this eld


6 nominative marking of the second noun (N2-Nominative)7 accusative marking of the rst noun (N1-Accusative)8 accusative marking of the second noun (N2-Accusative)9 verb agreement with the rst noun (N1-Agreement)

10 verb agreement with the second noun (N2-Agreement)

Additionally the word order (NVN NNV VNN) of the sentence wascoded Availability of a cue was calculated as the proportion of codeablesentences in which the cue was present

Reliability of the cue was calculated as the proportion of times the cuewas associated with the rst noun as agent given that the cue was present(ie p Agent | Cue) Thus a reliability of 10 indicates that whenever thecue is present it always points to the rst noun as the agent A reliability of00 indicates that a cue always points to the second noun as the agent Areliability of 05 indicates that the cue has no value at all in terms ofhelping us to decide whether the rst noun is the agent Both values closeto 00 as well as values close to 10 are indicators of cues with extremelyhigh reliability The availability and reliability estimations for the variouscues in both languages are given in Table 1

As can be seen from Table 1 there are only minor differences betweenRussian and German in the reliability of the cues When they are presentand contrastive the morphological cues of case-marking and subjectndashverbagreement are absolutely reliable The reliability of animacy vs inanimacy

TABLE 1Cue Availability and Cue Reliability in German and Russian

Availability ReliabilityRussian German Russian German

NVN 0722 0392 0897 0885NNV 0204 0465 0720 0798VNN 0074 0143 0500 0943N1-Animacy 0779 0776 0931 0968N1-Inanimacy 0221 0224 0481 0454N2-Animacy 0377 0424 0641 0673N2-Inanimacy 0623 0576 0947 0986N1-Nominative 0775 0445 1000 1000N2-Nominative 0143 0078 0000 0000N1-Accusative 0078 0029 0000 0000N2-Accusative 0373 0212 1000 1000N1-Agreement 0586 0294 1000 1000N2-Agreement 0114 0049 0000 0000

The reliability indicates how often a cue points to the rst noun as the agent NVN noun-verb-noun etc


of the rst and second noun is very similar in both languages and reectsthe fact that animacy information is a semantic cue rather than alanguage-specic cue The reliability of word order is also quite similar inRussian and German Despite the differences in permitted word ordersboth languages have a strong tendency for the rst noun to be the agentCalculated over all word order types the proportion of sentences wherethe rst noun was the agent is 0832 in Russian and 0853 in German

Russian and German differ primarily in terms of cue availability Whilethere were only small differences with respect to the availability of theanimacy cue the corpus analyses indicate that the availability of case-marking and subjectndashverb agreement is much higher in Russian than inGerman Moreover this pattern holds for both SO and OS congurationsThe overall proportion of sentences without any morphological cue was0025 in Russian and 0223 in German The corpus-based availability andreliability estimations therefore suport the claim that Russian andGerman differ markedly with respect to the availability of morphologicalcues

STUDY 1 SPEEDED PICTURE CHOICEEXPERIMENT

Our rst study used the speeded picture choice task to examine possiblelanguage differences in on-line processing of the cues for agentivity Sincethe reliability of these cues does not differ between Russian and Germanwe do not expect signicant differences in the distribution of agent choicesHowever we expect to nd differences in reaction times due to differentialeffects of cue availability

A complete crossing of all available cues described in the corpusanalyses would result in 288 grammatically correct sentence types foreach language Since testing all these conditions is not feasible within onestudy we conned the design to a systematic crossing of animacysubjectndashobject conguration and case-marking in NVN sentences If thelower availability of case-marking in German has an effect on on-lineprocessing we would expect the reaction time benet for sentences withcase-marked nouns as compared to sentences with unmarked nouns to belarger in Russian than in German Moreover the Competition Modelleads us to expect that any such weakness in the use of the case-markingcue in German should be matched by a compensatory reliance onanimacy information If this is the case then reaction time benets fromconverging animacy information should be larger in German than inRussian


MethodParticipants Participants were 14 native speakers of German and 14

native speakers of Russian The speakers of German were studying orworking in the United States Seven of the Russian participants wereimmigrants and the remaining seven were enrolled in a student exchangeprogramme or were relatives of exchange students All butone participant inthe Russian group had received partial or complete college or universityeducation in their native language None of the participants had lived in anEnglish-speaking environment for longer than 2 years The participantsrsquoexposure to other languages was assessed through a background ques-tionnaire which indicated that none of the participants had been exposed toany of their second languages before the age of 11 The mean age of theparticipants was nearly identical in both groups (Russian 257 yearsGerman 266 years) All participants were paid $500 for their participation

Materials and Design Language was varied as a between-subjectsfactor The stimulus materials consisted of simple active transitive NVNsentences which were grammatically correct in both Russian and GermanWithin each language the sentences were varied according to thefollowing ve factors

1 animacy of the rst noun (N1-Animacy animate vs inanimate)2 animacy of the second noun (N2-Animacy animate vs inanimate)3 conguration (SVO vs OVS)4 case-marking of the rst noun (N1-Marking marked vs unmarked)5 case-marking of the second noun (N2-Marking marked vs un-

marked)

Appendix 2 presents the composition of the 32 cells of this design Theconguration factor species the type of marking on the two nouns forSVO sentences case-marking on the rst noun was nominative and on thesecond noun accusative In OVS sentences the pattern was the oppositecase-marking on the rst noun was accusative and on the second nounnominative Language was varied as a between-subjects factor When case-marking is neutralised (unmarked-unmarked) it is impossible to distin-guish between the SVO and OVS congurations and listeners willnormally impose an SVO interpretation This neutralisation occurs forfour cell pairs in Appendix 2 1 and 5 9 and 13 17 and 21 and 25 and 29When interpreting the results it is important to remember that sentenceswith unmarked case in the OVS condition will be interpreted as SVO

The sentences were composed from combinations of four animatenouns four inanimate nouns and two verbs In order to minimise semantic


differences between languages exact translation equivalents for all tenwords were used The English translations of the eight nouns werelsquomotherrsquo lsquodaughterrsquo lsquofatherrsquo lsquosonrsquo lsquoowerrsquo lsquoplatersquo lsquospoonrsquo and lsquocakersquoThe verbs were lsquolooking forrsquo and lsquondingrsquo In German the nouns lsquomotherrsquolsquodaughterrsquo lsquoowerrsquo and lsquocakersquo are feminine and exhibit nominative-accusative neutralisation In Russian there is also nominative-accusativeneutralisation for these four nouns because lsquomotherrsquo and lsquodaughterrsquo havenal palatalisation and because lsquoowerrsquo and lsquocakersquo are masculineinanimate Since all nouns were singular and all verbs in present tenseverb agreement was not available as a cue

For each condition four sentences were constructed by rst combiningthe relevant nouns in all possible ways and then counterbalancing the twoverbs This resulted in 128 grammatically correct sentences Although allof the sentences were grammatical half of them were semanticallyimplausible Implausible sentences correspond to English sentences suchas lsquoThe cake nds the fatherrsquo In these sentences animacy is in competitionwith case-marking andor conguration Examples for all 32 sentence typesare also presented in Appendix 2

The eight nouns and two verbs were recorded by a female native speakerand digitised with a 16-bit sampling rate at 22 kHz using SoundEdit16Combining single word recordings into sentences ensured that theintonation pattern was identical for all sentences and that no prosodiccues were available to the listener The presentation of the digitisedauditory stimuli and of the eight pictures depicting the nouns in thesentence were controlled by the PsyScope experimental control pro-gramme (Cohen MacWhinney Flatt amp Provost 1993) The pictures of thefour inanimate nouns were taken from the Snodgrass and Vanderwart(1980) materials The pictures of the four animate nouns were taken fromother sources

Procedure The participants were familiarised with the pictures and thenouns by presenting each picture one at a time in the middle of the screenalong with the correct label for each picture Next participants saw pairs ofpictures As soon as the pictures were displayed the label of one of the twopictures was presented through headphones Participants placed their leftand right index ngers on the two outer buttons of the CMU button boxand were given these instructions lsquolsquoAs fast as possible press the buttoncloser to the object being namedrsquorsquo Thus if the left picture was named theleft button had to be pressed and vice versa The purpose of thispreparation phase was to acquaint the participants with the speeded forcedchoice task and to familiarise them with the specic picture-nouncombinations Participants saw all 54 possible combinations of the eightpictures in individually randomised order


In the main experiment participants were told that they were going tohear a series of simple transitive sentences accompanied by the pictures ofboth nouns Again their task was lsquolsquoAs fast as possible press the buttoncloser to the picture that is depicting who or what was carrying out theactionrsquorsquo If the left picture depicted the agent the left button had to bepressed and vice versa Participants were also told that some of thesentences might be implausible and that they should make a choiceregardless of whether the sentence made sense to them or not It wasstressed that they had to make a choice as soon as they felt certain enougheven if they had not yet nished hearing the sentence

For each participant ten practice sentences were selected randomlyfrom the larger pool of sentences After the ten practice trials all 128sentences were presented Order of presentation was randomisedindividually by the PsyScope program A trial consisted of the presentationof a xation point in the middle of the screen for 1000 msec followed bysimultaneous presentation of the sentence and the two pictures After theparticipants had executed their response the next trial followed with anISI of 500 msec Participantsrsquo choice and reaction times were recordedThe experiment was carried out on a Macintosh Centris 660 AV

Results

Choice Responses Table 2 presents the choice data for all conditionsas proportions of rst noun choices Although the rst noun is always thecorrect choice in case-marked SVO sentences and the second noun isalways the correct choice in case-marked OVS sentences we do notpresent the choice data as errors This is because in the morphologicallyambiguous sentences there was no reliable cue for determining a correctinterpretation

The arcsin-transformed proportions of rst noun choices for eachlanguage group were submitted to an omnibus ANOVA The rst columnof Appendix 3 lists all effects that reached or almost reached signicance atthe P lt 005 level both in the analyses by subjects (F1) and by items (F2)This analysis revealed no signicant language differences The large maineffect of Conguration indicates that both in Russian and in Germanagent choice was determined by case-marking When case-marking wasavailable participants chose the rst noun in SVO sentences and thesecond noun in OVS sentences This main effect is specied by the 2-wayinteractions of Conguration with N1-Marking Conguration with N2-Marking and N1-Marking with N2-Marking as well as the 3-wayinteraction of Conguration N1-Marking and N2-Marking which isdepicted by the solid lines in Fig 1

FIG 1 Proportions of rst noun choices in the experimental data (solid lines) andasymptotic activation in the model (dotted lines) as a function of Conguration N1-Markingand N2-Marking for Russian (upper panel) and German (lower panel)

141

TAB

LE2

Prop

ortio

nof

Firs

tN

oun

Cho

ices

per

Con

ditio

nin

the

Hum

anD

ata

and

inth

eM

odel

Rus

sian

Ger

man

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngD

ata

Net

wor

kD

ata

Net

wor

k

anin

anim

SVO

unm

ark

unm

ark

086

095

093

097

mar

ked

100

100

098

099

mar

ked

unm

ark

098

099

100

098

mar

ked

098

100

097

099

OV

Sun

mar

kun

mar

k0

920

950

930

97m

arke

d0

270

020

320

08m

arke

dun

mar

k0

020

000

040

00m

arke

d0

020

000

070

00

inan

imSV

Oun

mar

kun

mar

k0

940

990

970

98m

arke

d0

971

000

980

99m

arke

dun

mar

k1

000

991

001

00m

arke

d0

951

000

961

00O

VS

unm

ark

unm

ark

092

099

100

098

mar

ked

028

002

045

080

mar

ked

unm

ark

006

000

010

003

mar

ked

003

000

009

000

inan

iman

imSV

Oun

mar

kun

mar

k0

580

780

550

50m

arke

d0

881

000

850

99m

arke

dun

mar

k0

860

890

860

48m

arke

d0

951

000

940

99O

VS

unm

ark

unm

ark

066

078

056

050

mar

ked

025

001

018

000

mar

ked

unm

ark

001

000

006

000

mar

ked

002

000

002

000

142

TAB

LE2

(Con

td)

Prop

ortio

nof

Firs

tN

oun

Cho

ices

per

Con

ditio

nin

the

Hum

anD

ata

and

inth

eM

odel

Rus

sian

Ger

man

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngD

ata

Net

wor

kD

ata

Net

wor

k

inan

imSV

Oun

mar

kun

mar

k0

890

770

830

92m

arke

d1

001

000

960

99m

arke

dun

mar

k0

970

740

980

97m

arke

d0

971

001

001

00O

VS

unm

ark

unm

ark

088

077

081

092

mar

ked

027

000

039

010

mar

ked

unm

ark

009

000

007

020

mar

ked

003

000

012

000

143


The fourth and fth columns in Appendix 3 display the percent ofexperimental variance accounted for by each effect or interaction in boththe analysis by subjects and by items Together the various effectsinvolving case-marking and conguration accounted for approximately93 of the experimental variance The participants chose the rst noun82 of the time when sentences were not marked for case as compared toalmost 100 of the time in case-marked SVO sentences The fact that inthe absence of case-marking the rst noun is chosen above chance suggeststhat in both languages there is a default bias towards interpreting the rstnoun as the agent if no additional cues are available This bias is howeverrapidly reversed towards a correct second noun choice if an accusativemarker occurs on the rst noun In sentences with unmarked rst nouns ndnominative-marked second nouns the initial lsquorst nounrsquo interpretation hastime to become fairly well consolidated by the time the listener encountersthe conicting nominative case-marker later in the sentence This leads toa substantial amount of erroneous rst noun choices The averageproportion of erroneous rst noun choices in this condition was 027 inRussian and 034 in German In both languages the proportion of errorsin nominative-marked OVS sentences is signicantly higher than inaccusative-marked OVS sentences (all ts gt 94 all Ps lt 0001)

The main effect of N1-Animacy indicates that animate rst nouns aremore readily interpreted as agents This tendency is however easilyoverridden by case-marking as the interaction of N1-Animacy and N1-Marking indicates (see Fig 2 solid lines) The effects of N2-Animacy wasrelatively weak indicating that the second noun had much less inuence onthe decisions than the rst one

Reaction Times All outliers above 4000 msec were truncated therebyexcluding 35 of the data points Since all sentences were grammaticallycorrect we included only correct responses into the statistical analysis Forcase-marked sentences correct responses were those indicated by thenominative andor accusative case-markers For unmarked sentences rstnoun responses were dened as the correct ones reecting the fact thatSVO is the canonical unmarked order in both languages This is justiedby the nding that both language groups exhibited a strong rst noun biasin these sentences Overall another 12 of the data points were excludedby these criteria Due to the exclusion of these data points 13 conditionmeans per subject were missing and had to be substituted following aprocedure recommended by Winer (1971 p 487)

The analysis of the reaction times is complicated by the fact that theauditory presentation of the stimulus material induced an additionalsource of variability due to the different duration of the various parts ofthe sentence For example Russian animate nouns were always 154 msec

FIG 2 Proportions of rst noun choices in the experimental data (solid lines) andasymptotic activation in the model (dotted lines) as a function of N1-Marking and N1-Animacy for Russian (upper panel) and German (lower panel)

145


shorter than Russian inanimate nouns as opposed to a difference of only77 msec in German Similarly some of the Russian accusative markedmasculine animate nouns were longer than their unmarked nominativeform because the addition of the sufx changes the syllabic structure ofthese nouns [eg syn (NOM)mdashsyna (ACC)] Moreover the languagesdiffered with respect to the position of the case markers within thesentences In German case-marking affects the article which precedes thenoun whereas in Russian it affects the sufx at the end of the noun Thisalso has consequences for how fast a decision about the agent of thesentence can be reached Figure 3 illustrates the differences in thetemporal structure of a typical Russian and German experimental sentenceby displaying the position of the case-markers in relation to the averageduration of the two nouns and the verb These duration differences need tobe accounted for if reliable conclusions about the on-line effects of thevarious cues are to be drawn particularly in regard to the languagedifferences

In order to account for the duration differences we adjusted thereaction times to various reference points in the sentences and performedthe statistical analyses on these adjusted reaction times First we adjustedthe reaction times to the end of the rst noun by subtracting the durationof the rst noun from the reaction time for each sentence This adjustmentcontrols for potential effects due to differences in the duration of the rstnouns Second we adjusted the reaction times to the end of the verbthereby controlling for duration differences of the rst noun and the verbcombined Finally we adjusted the reaction times to the end of thesentence which additionally accounts for duration effects due to the second

FIG 3 Average duration of the noun phrases and the verb as well as positioning of case-markers in the Russian and German sentences


noun Omnibus ANOVAs by subjects and by items were performed on allthree adjusted reaction times Unless specied in the text all effectsinterpreted below are statistically reliable for all three reaction timeadjustments This restriction is necessary since effects that are not similaracross adjusted reaction times may be confounded with differences in theduration of the various parts of the sentence3 The corresponding F-valuesand percentages of experimental variance accounted for by the variouseffects and interactions are listed in Appendix 4

We will rst describe the effects that were similar in both languagegroups ie effects that did not interact with the factor of Language Thestrongest effect in both language groups was the effect of case-marking onthe rst noun which accounted for about 40 of the experimentalvariance If the rst noun was case-marked reaction times were on average318 msec faster than if it was unmarked This effect was specied by aninteraction of N1-Marking and Conguration which is due to the fact thatthe average reaction time benet of 499 msec from accusative marking onthe rst noun in OVS sentences is larger than the average benet of151 msec from nominative marking on the rst noun in SVO sentences

There was also a signicant albeit weaker effect of case-marking on thesecond noun This effect was further specied by a series of 2-way and 3-way interactions with N1-Marking and Conguration These interactionsare due to longer reaction times in sentences with unmarked rst andnominative marked second nouns of the type Die Tochter sucht der Vaterin German and DocIuml isIuml cIuml et otec in Russian as compared to unmarkedsentences like Die Tochter sucht die Mutter or DocIuml isIuml cIuml et matrsquo For thesesentences an additional 327 msec were required to restructure the initialinterpretation of the rst noun as the agent as soon as the unexpectednominative marker on the second noun is encountered The initialincorrect interpretation was further strengthened if the rst noun wasanimate as indicated by the signicant 3-way interaction of CongurationN1-Animacy and N2-Marking However case-marking of the secondnoun had no additional effect if the rst noun was unambiguously case-marked

Next we will describe those effects that were different in both languagegroups as indicated by interactions with the factor of Language First there

3 The use of ANCOVAs to partial out the effects of word duration is not a viable alternativebecause the assumption of constant slopes is violated Thus the duration of the varioussentence parts has different effects in the two languages For example the duration of thenouns has much more of an effect in Russian because the case marker appears at the end ofthe noun so that the listener will be much more inclined to process the whole noun asopposed to German where decisions can be made right after the article and before the wholenoun has been processed


was a signicant interaction between Language and N1-Marking in theanalyses by items However this effect fell short of signicance in theanalyses by subject (P = 006 for reaction times adjusted to the end of therst noun and P = 008 for reaction times adjusted to the end of the verband the second noun) This interaction indicates that the reaction timebenet of 400 msec from rst noun marking in Russian was larger than thebenet of 228 msec in German Note that this interaction was found inreaction times adjusted to the end of the nouns which did not account forrelative duration differences between articles and nouns in the variousconditions in German Still the relative benet from case-marking on therst noun was larger in Russian despite the fact that the average durationof the German case-marked nouns was slightly shorter in terms of numberof syllables This suggests that the language differences in the processingbenet from case-marking cannot be attributed to word duration effectsFurthermore the interaction between Language and N1-Marking wasspecied by a 3-way interaction involving the factor Conguration whichindicates that the language differences in the benet from rst noun-marking were most pronounced for accusative marked rst nouns Theupper panels of Fig 4 depict this interaction in the reaction times adjustedto the end of the rst noun

The other important reaction time difference between Russian andGerman was related to the effects of animacy of the rst noun This issuggested by the interaction between Language and N1-Animacy whichreached signicance in four out of the six ANOVAs and fell short ofsignicance in the analyses by items for reaction times adjusted to theend of the verb (P = 008) and to the end of the second noun (P =

009) In the German group but not in the Russian group decisions arefaster when the rst noun is animate Apparently the relatively weakervalidity of nominative-marking in German leaves more room foranimacy information to impact on-line performance We also found a3-way interaction between Language Conguration and N1-Animacy inthe analyses by items However this effect fell short of signicance inthe analyses by subjects (P = 01 for reaction times adjusted to the endof the rst noun P = 007 for reaction times adjusted to the end of theverb and the second noun) This suggests that the N1-Animacy benetobserved in German seems to be conned to SVO sentences in whichthe presence of an animate rst noun serves to further support thelsquoagent rstrsquo bias Apparently in OVS sentences accusative-marking onthe rst noun tends to override any effects of N1-Animacy in bothlanguages thereby supporting the earlier nding that accusative markersare stronger cues than nominative markers The upper panels of Fig 5depict this interaction in the reaction times adjusted to the end of therst noun


Discussion

The off-line results of the speeded agent choice task demonstrate that inmany regards speakers of Russian and German behave similarly in theirinterpretation of the sentences In the vast majority of trials the nalinterpretation is determined by the most reliable cue which is case-

FIG 4 Reaction times adjusted to the end of the rst noun in experiment 1 (upper panels)and in the reaction time estimates derived from the model (lower panels) as a function ofnominative-marking (SVO) and accusative-marking (OVS) in Russian and German


marking This comes as no surprise In both languages we found that theanimacy cue had a greater effect on speeding up reaction times when itoccurred on the rst noun than when it occurred on the second nounThis effect matches up well with the idea that the rst noun plays aunique role in organising and grounding the conceptual structure of the

FIG 5 Reaction times adjusted to the end of the rst noun in experiment 1 (upper panels)and reaction time estimates derived from the model (lower panels) as a function of N1-Animacy and Conguration in Russian and German


rest of the sentence (MacWhinney 1977 Gernsbacher Hargreaves ampBeeman 1989)

However these ndings regarding similarities between the twolanguages were supplemented by some interesting language differencesin on-line processing which had been predicted by our extended versionof the Competition Model Specically the corpus analysis predicted thatcase-marking should be a weaker cue in German than in Russian Thismeans that in relative terms animacy should be a stronger cue inGerman than in Russian Two ndings conrmed these predictions Firstthe magnitude of the processing benets from case-marking on the rstnoun was larger in Russian supporting the idea that case-marking ingeneral is a stronger cue in Russian than in German Both the processingbenet of rst noun case-marking and the language difference in themagnitude of this effect were most pronounced for accusative-markingThis aspect of the result is somewhat puzzling The smaller effect ofnominative-marking as compared to accusative-marking did not fall outobviously from the corpus-based availability estimations Collapsed overall nouns the difference in availability of the nominative and theaccusative was of similar size for Russian and German Looking only atreliability and availability we cannot account for the stronger effect ofaccusative case-marking Second the reaction time effects attributable toanimacy of the rst noun were different in Russian and German InRussian the strong use of nominative-marking swamps any effects ofanimacy of the rst noun Basically speakers of Russian look for rstnoun case If there is unambiguous case-marking it is used rapidlyignoring any additional information In German however rst nounanimacy leads to similar reaction time benets for both unmarked andnominative-marked rst nouns thus indicating that speakers of Germantake both case and animacy cues into account when processing the rstnoun

The fact that all sentences in the experiment were grammatical and didnot contain unusual cue combinations supports the ecological validity ofthe task and the generalisability of this interpretation It can be arguedhowever that the speeded nature of the task might have induced unusualperformance patterns In this regard it is noteworthy that the analyses ofthe choice data revealed no language differences while the analyses of thereaction times did This suggests that the obtained processing differencesbetween the two languages cannot be attributed to task-induced strategiesor to a trade-off between speed and accuracy but reect basic differencesin automatised aspects of sentence processing

Finally the results revealed an interesting pattern of cue interactionThe interaction between case-marking of the rst and second nounrevealed that a second case-marker leads to no additional benet if the


rst noun is already reliably case-marked This nding replicates resultsobtained by Kail (1989) and Mimica et al (1994) with respect to theconvergence of cues which have shown that in the presence of a strongcue a weaker supporting cue might show no additional effect In thesituation of a speeded task this can lead to early decisions that are maderight after the encounter of the strong cue and before all other cues in asentence have been fully processed These non-additive patterns of cueinteraction cannot be predicted from cue validity estimations for each ofthe cues separately

Taraban and Roark (1996) have shown that the performance of abackpropagation learning model showed a better t to the patterns of cueinteraction than estimations based on the reliability of individual cuesThey have argued that connectionist models provide a more accurate wayto account for effects of cue competition and cue convergence The nextsection describes an attempt to extend this approach towards cueinteraction in on-line processing We will try to simulate the obtainedempirical results using a recurrent cascaded backpropagation network Theaim of this simulation is to explore to what degree the language differencesand patterns of cue interaction in the reaction times can be accounted forby principles of parallel interactive activation In evaluating the networkperformance we will concentrate on the following ve main results fromthe experiment

1 Availability effects The discrepancy between choice data andreaction times showed that differences in cue availability tend toaffect the time course but not the nal result of processing

2 Case-marking effects The case-marking cue has an immediate effectwhich is stronger in Russian than in German This languagedifference has been attributed to the higher availability of this cuein Russian

3 Animacy effects The language differences in the availability of case-marking result in reliance on the animacy cue in German but not inRussian

4 On-line cue interaction The stronger an early cue the more it tendsto override later cues in on-line processing without displaying anypatterns of information integration On the other hand the weakera cue the more does processing rely on other supporting cues

5 Accusative superiority There was an overall stronger benet fromaccusative-marking as compared to nominative-marking in thereaction times This effect was not predicted by the corpus-basedvalidity estimations


STUDY 2 CONNECTIONIST SIMULATION

The Competition Model notions of availability and reliability have beenuseful as guides to off-line choice behaviour They also work to predictcertain basic aspects of language differences in on-line processingHowever they do not offer a satisfactory account for the specic patternsof cue interaction The connectionist model developed in this sectionattempts to go beyond these two rather simplistic constructs and toformulate a fuller account of on-line processing effects The modelpresented here is designed to capture basic language differences in the cuedistributions It is not presented as an attempt to model the full structure ofeither Russian or German Instead it uses a simple architecture and alimited training corpus to model patterns of cue interaction in the reactiontime data of Study 1

The model emphasises the sequential temporal nature of the sentenceinterpretation task This temporal sequentiality is captured by using asimple recurrent network (Elman 1990) which takes in input one word at atime The network keeps track of previous activation states while newinformation comes in As the input is presented the activation of variousalternative interpretations builds up gradually Cascaded networks(McClelland 1979) in which the activation of units is calculated as therunning average of their net input over time have been shown to exhibit agood t to reaction time data (Cohen Dunbar amp McClelland 1990)Combining the advantages of cascading with those of the recurrentnetwork we created a recurrent cascaded network to simulate the on-lineeffects of availability reliability and cue interaction Learning occurs viathe backpropagation rule (Rumelhart Hinton amp Williams 1986) whichmodies connection strength as a function of feedback The output of themodel is compared to a target signal and the learning algorithm adjusts theconnection weights in a way that reduces overall error Back-propagationnetworks have been shown to be sensitive to the frequency of inputpatterns since frequent patterns make large contributions to the overallerror reduction

The network consisted of four layers an input layer with ve units acontext layer with ve units a hidden layer with ve units and a singleoutput unit The ve input units coded the following properties of an inputstring (i) phrasal word class (noun phrase vs verb phrase) (ii) animacy(iii) nominative-marking (iv) accusative-marking and (v) a single codingfor any of the morphosyntactic dimensions (gender or number) on whichnouns and verbs can agree If a string was coded as a verb the animacy andcase-marking units were always set to 0 The single output unit codedwhether the rst noun was the agent of the sentence The values of thetarget output were set to 1 for a lsquorst nounrsquo-response and to 0 for a lsquosecond


nounrsquo-response Thus the activation strength in the output unit corre-sponds to the strength of the lsquorst-nounrsquo-interpretation which can bedirectly compared to the proportion of rst noun choices in the empiricaldata

At each trial the input corresponding to a sentence is processed in thefollowing way after the rst string is presented activation is fed forwardand compared to the target Using the backpropagation algorithm allweights are changed in a way that minimises the overall error term Theactivation state of the hidden units is then copied back to the context unitsand presented as input together with the second string After processingthe second string the newly obtained activation state is copied back andpresented as context to the input of the third string When processing ofthree input strings is completed the values of the context units are clearedThe architecture of the network is presented in Fig 6

The recurrent architecture of the network allowed us to match thedistribution of cues over the course of the sentence The local cues ofanimacy and case-marking were coded as features within an input stringwhereas the topological cue of verb agreement was coded as a match ofactivation in the input units over time For example the input coding forthe three strings in an SVO sentence containing two animate case-markednouns and verb agreement was 111010000111010 Agreement betweenthe rst noun and the verb is coded by virtue of the match of the value lsquolsquo1rsquorsquo

FIG 6 Architecture of the recurrent cascaded backpropagation model


in the nal unit of the rst and the second strings In order to prevent thenetwork from developing verb-agreement biases associated with aparticular value of the unit half of the patterns had agreement coded asa match in the value of 1 (as in the example above) and the other half as amatch in the value of 0

We constructed two sets of training patterns in order to model theessential differences in the cue distributions between Russian and GermanEach pattern in the training sets corresponded to a specic sentence typeThe sentence types varied on the dimensions of word order animacy of thetwo nouns case-marking of the two nouns and verb agreement Thetraining sets resembled the frequencies of all sentence types obtained fromthe corpus analyses thereby reproducing the essential language differencesin the availability of the cues The German corpus included 71 differentsentence types the Russian corpus included only 51 different sentencetypes For the ambiguous sentence types the agent was determined on thebasis of the contextual information given in the corpus

We used a very low learning rate of 001 and a momentum of 07Furthermore the proportion of activation of the hidden units that wascopied back to the context units (m) also had to be set to the fairly lowvalue of 04 in order to allow for drastic activation changes induced bycompeting evidence presented at different points in the input patternsEach network was trained for 3000 epochs on the Russian vs the Germantraining patterns until the overall error term had reached an asymptoticvalue in both sets During learning the connection weights were updatedafter each sweep through the whole set of training patterns For bothlanguages we ran ve training simulations each with a different set ofrandom weights at the outset of learning After training the learningperformance was tested for each of the ve runs in each language The testpatterns coded the 32 sentence types used in Experiment 1 Ten of the 32test patterns had been present in the German training set and 18 of the 32test patterns had been present in the Russian training set In the test phasethe network was presented with the 32 test patterns and was allowed tocycle until the output unit had reached an asymptotic value The number ofcycles for each input string was set to 200 It has been shown that theasymptotic values reached in cascaded feedforward networks are identicalto the activation reached in a single step in standard one-pass computa-tions (McClelland 1979) However the cascaded nature of the networkallows us to examine the gradual build-up of activation

In order to compare the network performance with the empirical datawe saved the asymptotic activation values reached after presentation ofeach of the three input strings in a pattern The asymptotic activationarrived at after the third string which corresponds to the last segment ofthe sentence was compared with the overall proportion of rst noun


choices for the corresponding sentence type obtained in Experiment 1Furthermore for each pattern we determined the number of cyclesnecessary to reach a threshold which was taken as an estimation of thereaction time The threshold values were set to 098 for a rst nounresponse and to 002 for a second noun response This threshold preventedus from coding an incorrect response in patterns where animacy contrastconicted with case-marking on the second noun Note that this thresholdcan be reached after the rst string (within 0ndash200 cycles) after the secondstring (within 200ndash400 cycles) or after the third string (within 400ndash600cycles) thereby accounting for the fact that in a speeded choice taskresponses can be given at any time during sentence comprehension If theasymptotic activation did not reach the threshold at all we coded themaximum number of 600 cycles as reaction time estimation The nextsection evaluates the t of the model to the human data with respect to theproportions of the rst noun choices and the reaction times

Fit to Choice Data The mean asymptotic activation values calculatedover the ve runs in each language can also be found in Table 2 Theoverall t of the model to the experimental data was determined through aregression of the mean activations onto the proportions of rst nounchoices This analysis revealed an R2 of 095 (P lt 0001) for Russian and092 (P lt 0001) for German thus exhibiting an excellent t of the model tothe human data It demonstrates that even for this limited training set arecurrent backpropagation network is able to learn the correct interpreta-tions of a wide variety of Russian and German sentence patterns even ifthey were not present in the learning input

In order to better evaluate the extent to which the network reproducedthe qualitative prole of the choice data we submitted the arcsin-transformed asymptotic activations for the ve runs per language to anomnibus ANOVA Since the variance over the ve runs was quite smallmany of the factorial effects reached signicance Therefore we will notpresent the F-values but rather compare the effect sizes of the model andthe human data in terms of percentage of variance accounted for by eacheffect and interaction The model and human effect sizes calculated as thepercentage of experimental variance accounted for by each effect arepresented in Appendix 3

As in the human data no signicant language differences were foundThe hierarchy of the effect sizes is the same in the network and in thehuman data the largest effect size was found for the factor ofConguration (about 61) followed by the interaction of N1-Markingand Conguration (about 14) and of N2-Marking and Conguration(about 5) This indicates that in the network as in the human data nounchoice is overwhelmingly determined by case-marking Furthermore the


analysis revealed interactions between rst noun marking and rst nounanimacy conrming that as in the human data case-marking overridesanimacy information in the networkrsquos nal result of sentence interpreta-tion The dotted lines in Fig 1 show the model performance with respect tothe effects of Conguration N1-Marking and N2-Marking and in relationto the human data As can be seen the direction of the effects as well asthe overall effect sizes of model and data are almost identical

One source of minor discrepancies between the learning outcome of thenetwork and the human choice data are patterns with unmarked rst nounsand nominative-marked second nouns (see Table 2) Recall that thesepatterns correspond to the sentences requiring a restructuring of the initiallsquoagent rstrsquo-interpretation Here the Russian network exhibited moreaccurate performance than the speakers of Russian This discrepancy ismost likely a result of the speeded nature of the task which caused theparticipants to terminate processing before they had settled into a stableresponse In contrast the network always reaches asymptotic activationWhile the performance of the German network was also superior for mostof these patterns the speakers of German outperformed the network whensecond noun nominative-marking conicted with the animacy contrast inAI sentences For this pattern the erroneous rst noun activation of thenetwork (080) was clearly higher than in the human data (045) (seeTable 2) which indicates that the German network weighted the animacycue somewhat higher than the Russian network This suggests that theeffect of noun animacy was a bit exaggerated in the German networkHowever the fact that this cue is weighted higher in German is inaccordance with the stronger reliance on animacy found in the Germanreaction times Taken together the learning outcome of the networkrevealed a remarkable t to the choice data of the experiment

Fit to Reaction Times In order to evaluate the general t of thereaction time estimations derived from the model we determined howmuch variance in the human reaction times is accounted for by the numberof cycles necessary to reach the threshold Because the reaction times inthe experimental data are inuenced by the actual duration of the words inmilliseconds we need to partial out these stimulus effects in order toproperly evaluate the match of the model to the data To do this weperformed a stepwise regression on the mean reaction times per sentencetype with duration of the rst and the second noun entered at the rst stepand number of cycles entered at the second step Duration of the verb wasnot entered since the counterbalancing scheme used for constructing theexperimental materials had resulted in identical mean verb duration persentence type The percentage of variance accounted for by the number ofcycles above the duration of the rst and the second noun was 24 for


Russian [F(128) = 116 P lt 001] and 38 for German [F(128) = 393 Plt 0001] Thus the network performance accounted for a substantialpercentage of variance in the human data The general t of the model wassomewhat better for German

An omnibus ANOVA was performed on the number of cycles in theve runs in order to evaluate the qualitative prole of the networkperformance with respect to the time course of processing The effectsizes are presented in the last column of Appendix 4 We will comparethe performance of the network to the human data with respect to thelanguage differences in on-line processing of case-marking and nounanimacy First as in the human data the largest proportion of variancewas accounted for by the effect of N1-Marking The average difference inthe number of cycles between unmarked and marked rst nouns was 463in the Russian and 314 in the German network However the overallmagnitude of this effect is exaggerated in the network where it accountedfor 814 of variance as compared with only an average of 42 ofvariance in the human data This effect was further specied by theinteraction of Language and N1-Marking which indicates that themagnitude of the N1-Marking benet was larger in the Russian networkthan in the German network Thus the network captures the essentiallanguage difference in the effect of case-marking on the rst nounHowever a comparison between the upper and lower panels of Fig 4demonstrates that the network did not differentiate between effects ofrst noun-nominative-marking and rst noun accusative-marking asevidenced in the absence of any interaction of N1-Marking withConguration in the network performance This suggests that the largerbenet from accusative-marking cannot be explained by the distribu-tional characteristics of cues in Russian and German transitive sentencesIn the Discussion section we will suggest other explanations for thisnding

Second the model correctly reproduces the essential language differ-ences in the effect of rst noun animacy by demonstrating a signicantinteraction between Language and N1-Animacy Moreover the lowerpanel of Fig 5 demonstrates that in German the benet from animate rstnouns is more pronounced in SVO sentences This result matches thehuman data quite well

Finally any effects and interactions involving N2-Marking did notaccount for a signicant proportion of variance in the networkperformance This supports the experimental nding that strong cuesprovided early in the sentence result in fast immediate responses whichare not affected by any information provided at a later point unless theyresult in a complete revision of the initial interpretation However thenetwork underestimated the magnitude of the reaction time increase that


is required for this restructuring of the initial interpretation in sentenceswith unmarked rst and nominative-marked second nouns

Taken together the network succeeded in capturing the essentiallanguage differences by exhibiting larger effects of rst noun case-markingin Russian and larger effects of rst noun animacy in German although itwas unable to reproduce the differences between nominative andaccusative-marking This supports the assumption that the differentialeffects of nominative and accusative are unrelated to the distributionalcharacteristics of case-markers in Russian and German

Discussion

The recurrent cascaded backpropagation network matched the beha-vioural data obtained in the experiment closely in a variety of aspectswhich are summarised below At the same time there was one aspect inwhich the model failed to match up well with human reaction time data

Availability effects Both in the model and in the data no signicantlanguage differences were found with respect to the nal outcome ofprocessing The crosslinguistic differences related to the availability of cuesmanifest themselves in the time course of processing Figure 7 illustrateshow this effect comes about While the asymptotic activation valuesarrived at in both languages are very similar the time needed to reach theactivation threshold may be different These differences are larger at thebeginning of the sentence suggesting that availability effects can best bedetected in speeded tasks

Case-marking effects The model correctly simulates the fact that theprocessing benet from case-marking is larger in Russian than in GermanWe will briey describe the nature of this effect with respect to thepatterns illustrated in Fig 7 In Russian threshold activation for sentenceswithout case-markers is reached at the average after the third string (Fig 7rst panel) The Russian network takes more time than the Germannetwork because at the beginning of each pattern the Russian networkstarts out with an activation value close to 05 as opposed to 09 inGerman If the rst noun is case-marked the activation threshold can bereached after the rst string and the relative gain in time is quitesubstantial (Fig 7 second and fourth panels) In the German network theinitial activation strongly favours the lsquoagent rstrsquo interpretation (around09) Thus in patterns with two unmarked nouns the threshold can bereached at the second string (Fig 7 upper panel) despite the absence of areliable marker Compared to this faster response the relative gain fromcase-marking on the rst noun is smaller in German than in Russian This

FIG 7 Time course of activation build-up for selected patterns in the Russian (thicker line)and German (thinner line) networks

160


analysis indicates that the larger benet from case-marking in Russian is arelative one caused by different initial biases at the outset of processing Itshould be noted however that the magnitude of this effect was greatlyexaggerated in the model This is most likely a consequence of the fact thatthe model was exposed to a very limited set of training patterns whichcoded only nominative and accusative-marking An inclusion of a largervariety of sentence types containing all possible case-markers shoulddiminish this exaggeration and lead to more realistic effect sizes Howeverthe main emphasis of this simulation was to replicate the languagedifferences and in this respect it performed well

Animacy effects The network correctly captures the fact that animacyis a stronger cue in German than in Russian This occurs not becauseanimacy is more reliable or more readily available in German but becausecase-marking is less available in German than in Russian The simulationshows that the strength of any individual cue depends on the relativestrength of other cues which in turn is determined by the distributionalcharacteristics of the primary linguistic data

On-line cue interaction As in the human data cues on the second noundid not lead to any additional benet if they merely supported cuesencountered earlier in the sentence Thus early cues appear to triggerimmediate responses the network equivalent of which is a fast attainmentof the criterion activation value Only if later cues are in conict withearlier ones additional processing time for restructuring the initialinterpretation is required This aspect of the data was not matched bythe network performance However a closer look at the networkperformance shows that this is a consequence of the way reaction timeswere estimated even though the asymptotic activation value at 600 cycleswas below criterion in patterns with unmarked rst nouns and nominative-marked second nouns the cut-off at 600 cycles leads to a systematicunderestimation of any time required for additional processing

Accusative superiority The principal failure of the model was itsinability to simulate the stronger effect of accusative-marking overnominative-marking of the rst noun in the reaction times We believethat the accusative superiority effect cannot be attributed to distributionalcharacteristics of the nominative and the accusative in Russian andGerman but rather is a result of the structure of the inectionalparadigms in the two languages Each of the Russian and Germaninections occurring in the nominative and the accusative case also servesas a marker for a variety of other cases dependent on the number andgender of the noun For example the German article der is not only


associated with nominative in masculine nouns but also with genitive anddative in feminine nouns and with genitive in plural nouns Similarly theGerman article den is not only a marker for accusative in masculine nounsbut also a marker for dative in plural nouns The fact that den is associatedwith only two possible alternatives (accusative dative) whereas der isassociated with three alternatives (nominative genitive dative) suggeststhat the former is a less ambiguous case-marker than the latter Thisproblem of inectional syncretism is even more severe in Russian where avery limited number of sufxes is distributed over a paradigm with 72 cellsThe accusative superiority observed in the experiment might be aconsequence of the fact that the inections associated with the accusativeare less syncretic ie refer to a smaller number of cases across thelanguage than the inections associated with the nominative even if theyare perfectly reliable with respect to the individual noun In order tocapture this in a connectionist simulation we will need to provide a richerrepresentation of the complete inectional paradigm This is clearly a taskfor future modelling work

GENERAL DISCUSSION

Two studies were designed to extend the Competition Model frameworkto account for crosslinguistic differences in on-line sentence processing Inparticular we were interested in assessing the possible role of cueavailability as a determinant of on-line processing The results of theexperiment conrmed the existence of language differences in Russian andGerman on-line processing despite the fact that both languages providethe same repertoire of reliable cues We argue that the observedcrosslinguistic differences in on-line processing must be attributedgenerally speaking to differences in cue availability Moreover theconcepts of reliability and availability taken alone are not sufcient topredict more complex patterns of cue interactions To account moresatisfactorily for these effects we constructed a recurrent cascadedbackpropagation model that exhibited a good t to the human behaviourobserved in the experiment and was able to capture the essential languagedifferences as well as the patterns of cue interactions

The simulation provided a compelling illustration of the ways in whichthe effects of various cues unfold through real time In the networkcomprehension is modelled as incremental accumulation of evidence forcompeting alternative interpretations (eg lsquothe rst noun is the agentrsquo vslsquothe second noun is the agentrsquo) Both the simulation and the data indicatethat stronger cues permit to essentially ignore the presence of weaker cueswhereas weaker cues allow other information to inuence the timing andoutcome of the sentence interpretation The on-line cue interactions reect


relative differences in cue strength that can be traced to differences in cueavailability stemming from neutralisations in the inectional paradigms

The development of these processing mechanisms can be described as atype of lsquolinguistic tuningrsquo (Cuetos amp Mitchell 1988) in that the statisticalproperties of the language may tune the processing system to use theavailable cognitive resources in the most efcient way The notions of cueintegration and linguistic tuning are also compatible with both theinteractive constraint-based view of sentence processing (MacDonald etal 1994) and the referential-support approach (Altmann 1988 Altmannet al 1992 Altmann amp Steedman 1988) According to these modelssemantic referential and contextual information as well as the statisticalproperties of the language all contribute to the selection of interpretationalternatives What differentiates our approach from these complementaryrelated perspectives is our crosslinguistic methodological focus While themajority of research in the domain of sentence processing has concentratedon the factors contributing to lexical and syntactic ambiguity resolution wehave focused on crosslinguistic differences in the processing of both non-ambiguous and ambiguous sentences in order to show how the statisticalproperties of the language modify the constraints operating in on-lineprocessing We have shown that ambiguity in the inectional paradigmsdiminishes the availability of morphological cues which in turn decreasesthe on-line processing benets associated with this cue Use of thiscrosslinguistic methodology broadens the empirical basis for thesetheoretical claims beyond the scope of English to the realm of propertiesthat may be true of language as a whole

The fact that statistical properties of the language inuence on-lineprocessing is not immediately compatible with an approach that viewslanguage comprehension as the application of non-competitive rules orparsing principles (Frazier 1987 1989 Frazier amp Fodor 1978) Howeverthe lack of research in the area of morphosyntactic processing within thisapproach makes it difcult to evaluate its specic predictions with respectto the effects of morphological cues The fact that the orthogonal crossingof animacy and case-marking in the experiment revealed a strongeranimacy benet in German seems to argue against a view of morphosyntaxas an encapsulated module insulated against semantic information (BockLoebell amp Morey 1992 Fodor 1983) Instead the larger animacy effect inGerman suggests that language processing operates in a highly interactivefashion which permits the relatively weak evidence from the morphosyn-tactic domain to be supplemented by evidence from the semantic domain

However the experimental results revealed differences in the on-lineeffects of nominative-marking and accusative-marking that could not becaptured by the simulation The failure to account for the accusativesuperiority effect is most likely a consequence of the limited number of


sentence types provided as input to the model and the fact that our codingof linguistic information represented nominative-marking and accusative-marking as dichotomous features thereby neglecting the specic inec-tional changes associated with case Future modelling attempts will have toinclude a sampling of the entire inectional paradigm across a largervariety of sentence types in order to get a more precise picture of howhuman language processing adapts to the statistical properties of thelanguage

Taken together this study has yielded ve major ndings each of cleartheoretical importance First we have shown that cue availability has amajor on-line impact on sentence processing Second we have providedadditional evidence for the nonmultiplicative effects of redundant cues foron-line processing Third we have shown that the absence of a singlestrong on-line cue in a language like German tends to provide room for theexpansion of a secondary cue such as animacy Fourth we have shown thatthese on-line cue processing effects can be captured through a connec-tionist model that utilises a cascaded recurrent backpropagation archi-tecture Fifth we have demonstrated crosslinguistic differences in theimmediacy of processing of morphological cues such as case markers Wehave shown that even two languages that have a parallel set of reliablemorphological cues can differ in regard to on-line processing Given thiswe can expect even stronger crosslinguistic differences in on-lineprocessing between languages that have markedly different cue reper-toires such as Navajo Quechua Arabic or Georgian We hope that theseinitial results further encourage other researchers to consider the use ofsystematic crosslinguistic comparisons as ways of understanding funda-mental issues in language processing

Manuscript received September 1996Revised manuscript received January 1998

REFERENCESAltmann GT (1988) Ambiguity parsing strategies and computational models Language

and Cognitive Processes 3 73ndash97Altmann GT amp Steedman M (1988) Interaction with context during human sentence

processing Cognition 30 191ndash238Altmann GT Garnham A amp Dennis Y (1992) Avoiding the garden path eye

movements in context Journal of Memory and Language 31 685ndash712Bates E MacWhinney B Caselli C Devescovi A Natale F amp Venza V (1984) A

crosslinguistic study of the development of sentence interpretation strategies ChildDevelopment 55 341ndash354

Bates E McNew S MacWhinney B Devescovi A amp Smith S (1982) Functionalconstraints on sentence processing A cross-linguistic study Cognition 11 245ndash299


Bock K Loebell H amp Morey R (1992) From conceptual roles to structural relationsBridging the syntactic cleft Psychological Review 99 150ndash171

Cohen J Dunbar K amp McClelland JL (1990) On the control of automatic processes aparallel distributed processing account of the Stroop-effect Psychological Review 97 332ndash361

Cohen J MacWhinney B Flatt M amp Provost J (1993) PsyScope An interactivegraphical system for designing and controlling experiments in the Psychology laboratoryusing Macintosh computers Behavioral Research Methods Instrumentation and Compu-tation 25 257ndash271

Comrie B amp Corbett G (Eds) (1992) The Slavonic Languages London RoutledgeCorrigan R (1986) The internal structure of English transitive sentences Memory and

Cognition 14 420ndash431Corrigan R (1988) Who dun it The inuence of actor-patient animacy and type of

verb in the making of causal attributions Journal of Memory and Language 27447ndash465

Cuetos F amp Mitchell DC (1988) Cross-linguistic differences in parsing Restrictions onthe use of the Late Closure strategy in Spanish Cognition 30 73ndash105

Elman J (1990 Finding structure in time Cognitive Science 14 179ndash212Fodor JA (1983) The Modularity of the Mind Cambridge MA MIT PressFrazier L (1987) Sentence processing A tutorial review In M Coltheart (Ed) Attention

and performance XII (pp 601ndash681) Hove UK Lawrence Erlbaum Associates LtdFrazier L (1989) Against lexical generation of syntax In W Marslen-Wilson (Ed) Lexical

Representation and Process Cambridge MA MIT PressFrazier L amp DrsquoArcais GBF (1989) Filler driven parsing A study of gap lling in Dutch

Journal of Memory and Language 28 331ndash344Frazier L amp Fodor JD (1978) The sausage-machine A new two-stage parsing model

Cognition 6 291ndash325Gernsbacher MA Hargreaves DJ amp Beeman M (1989) Building and accessing clausal

representations The advantage of rst mention versus the advantage of clause recencyJournal of Memory and Language 28 735ndash755

Hernandez A Bates E amp Avila L (1994) On-line sentence interpretation in Spanish-English bilinguals What does it mean to be lsquolsquoin betweenrsquorsquo Applied Psycholinguistics 15417ndash446

Kail M (1989) Cue validity cue cost and processing types in French sentencecomprehension In B MacWhinney amp E Bates (Eds) The Crosslinguistic Study ofLanguage Processing (77ndash117) New York Cambridge University Press

Kilborn K (1989) Sentence processing in a second language The timing of transferLanguage and Speech 32 1ndash23

Li P Bates E amp MacWhinney B (1993) Processing a language without inections Areaction time study of sentence interpretation in Chinese Journal of Memory andLanguage 32 169ndash192

MacDonald MC Pearlmutter NJ amp Seidenberg MS (1994) The lexical nature ofsyntactic ambiguity resolution Psychological Review 101 676ndash703

MacWhinney B (1977) Starting points Language 53 152ndash168MacWhinney B (1987) Toward a psycholinguistically plausible parser In S Thomason

(Ed) Proceedings of the Eastern States Conference on Linguistics (pp 1ndash8) Ohio StateUniversity Columbus Ohio

MacWhinney B (1995) The CHILDES project Tools for analyzing talk (2nd ed)Hillsdale NJ Erlbaum

MacWhinney B amp Bates E (Eds) (1989) The Crosslinguistic Study of SentenceProcessing New York Cambridge University Press


MacWhinney B Bates E amp Kliegl R (1984) Cue validity and sentence interpretation inEnglish German and Italian Journal ofVerbal Learning and Verbal Behavior 23 127ndash150

MacWhinney B Pleh C amp Bates E (1985) The development of sentence interpretationin Hungarian Cognitive Psychology 17 178ndash209

McClelland JL (1979) On the time-relations of mental processes An examination ofsystems of processes in cascade Psychological Review 86 287ndash330

McDonald JL (1986) The development of sentence comprehension strategies in Englishand Dutch Journal of Experimental Child Psychology 41 317ndash335

McDonald JL amp MacWhinney B (1989) Maximum likelihood models for sentenceprocessing research In B MacWhinney amp E Bates (Eds) The Crosslinguistic Study ofSentence Processing (pp 397ndash421) New York Cambridge University Press

McDonald JL amp MacWhinney BJ (1985) The time course of anaphor resolution Effectsof implicit verb causality and gender Journal of Memory and Language 34 543ndash566

Mimica I Sullivan M amp Smith S (1994) An on-line study of sentence interpretation innative Croatian speakers Applied Psycholinguistics 15 237ndash261

Mitchell DC amp Cuetos F (1991) The origins of parsing strategies In C Smith (Ed)Current Issues in Natural Language Processing (pp 1ndash12) Center of Cognitive ScienceUniversity of Austin Texas

Mitchell DC Cuetos F amp Zagar D (1990) Reading in different languages Is there auniversal mechanism for parsing sentences In DA Balota GB Flores drsquoArcais KRayner (Eds) Comprehension Processes in Reading (pp 285ndash302) Hillsdale NJLawrence Erlbaum Associates

Rumelhart D Hinton G amp Williams R (1986) Learning internal representations byback propagation In D Rumelhart amp J McClelland (Eds) Parallel DistributedProcessing Explorations in the Microstructure of Cognition (pp 318ndash362) CambridgeMA MIT Press

Snodgrass JG amp Vanderwart M (1980) A standardized set of 260 pictures Norms forname agreement image agreement and visual complexity Journal of ExperimentalPsychology Human Learning and Memory 6 174ndash215

Sokolov J (1989) The development of role assignment in Hebrew In B MacWhinney ampE Bates (Eds) The Crosslinguistic Study of Sentence Processing (pp 118ndash157) New YorkCambridge University Press

Taraban R amp Roark B (1996) Competition in learning language-based categoriesApplied Psycholinguistics 17 125ndash148

Trueswell JC (1996) The role of lexical frequency in syntactic ambiguity resolutionJournal of Memory and Language 35 566ndash585

Trueswell JC Tanenhaus MK amp Garnsey SM (1994) Semantic inuences on parsingUse of thematic role information in syntactic ambiguity resolution Journal of Memory andLanguage 33 285ndash318

Vigliocco G Butterworth B amp Semenza C (1994) Constructing subject-verb agreementin speech The role of semantic and morphological factors Journal of Memory andLanguage 34 186ndash215

Winer BJ (1971) Statistical Principles in Experimental Design New York McGraw HillZubin D (1979) Discourse function of morphology The focus system in German In T

Givon (Ed) Syntax and Semantics Discourse and Syntax (Vol 12 pp 469ndash504) NewYork Academic Press

Zubin DA (1977) The semantic basis of case alternation in German In RW Fasold ampRW Shuy (Eds) Studies in Language Variation Semantics Syntax PhonologyPragmatics Social Situations Ethnographic Approaches (pp 88ndash89) Washington DCGeorgetown University Press


APPENDIX 1Source references for the corpus analysis in Russian and German

GermanDanella U (1989) Die Unbesiegte Munchen Wilhelm Heyne Pp 169ndash173 459ndash463Die Kinder- und Hausmarchen der Bruder Grimm (1967) Der Kinderbuchverlag Berlin

Pp 136ndash139Rickenhaus Fr (1993) Wieviel Programme ertragt ein Mensch Die Zeit 48 (Nov 23) 13ndash

15Schmidt-Hauer Chr (1993) Die Diktatur des Demokraten Die Zeit 47 (Nov 19)Simmel JM (1988) Die im Dunkeln sieht man nicht Munchen Knauer Pp 75ndash80 388ndash

392

RussianAntonov S (1976) Povesti i passkazy Moskva Molodaja Gvardija 86ndash90 210ndash214Bitov A (1993) Ozhidanije obesrsquorsquojan (Fragment romana) Moskovskije Novosti 45 Nov 7Gorbunov A amp Kokesnikov A (1993) Shturm lsquolsquoBelogo Domarsquorsquo lsquolsquoAlrsquofarsquorsquo ne khotela krovi

Moskovskije Novosti 45 Nov 7Navrozov L (1993) U nas etogo citatrsquo ne budut Literaturnaja gazeta 47 Nov 24Russkije Skazki (1991) Moskva Sovetskaja Rossija 34ndash44Strokanrsquo S (1993) Yadernaja Ukrainamdashbolrsquoshoj blef ili bolrsquoshaja ugroza Moskovskije

Novosti 45 Nov 7Vainer AA amp Vainer GA (1985) Era miloserdija Kishinrsquoov Lumina 115ndash119 279ndash283

377ndash381Vassilrsquoev G (1993) Somnenie estrsquo kompliment iskusstvu Moskovskije Novosti 45 Nov 11

APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169


APPENDIX 3F-values and effect sizes in the rst noun choices (human data and network)

Proportion N1 choice experimental variance

Effect F1(126) F2(1193) By subjects By items Network

N1-Animacy 185 464 11 16 16N2-Animacy 96 425 14Conguration 8649 1784 612 606 601N1-Marking 2965 2258 81 77 29N2-Marking 529 212 13 10 36Conguration N1-Marking 6198 4402 144 149 135Conguration N2-Marking 1385 1426 48 48 48N1-Animacy N2-Animacy 74 188N1-Animacy N1-Marking 238 224 18N1-Animacy N2-Marking 168 101 18N1-Marking N2-Marking 553 272 13 28Conguration

N1-Marking N2-Marking1044 1099 42 37 36

Effect sizes (in percentage of experimental variance) are given only if the effect or interactionaccounted for more than 1 of variance P lt 0001 P lt 001 P lt 005

171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1


tion Model (MacWhinney amp Bates 1989) has been proposed as a way ofrelating the observed processing differences to variations in languagestructure In particular the Competition Model focuses on the role ofsurface cues such as word order noun animacy subjectndashverb agreementand noun case-marking The strength of a cue is viewed as depending onthree factors (i) its availability dened as the proportion of times a cue ispresent and can be used for accessing the underlying function (ii) itsreliability dened as the proportion of times a cue signals the correctinterpretation given that it was present and (iii) its cost which depends onthe perceptual salience of the cue and the load it places on workingmemory

Studies within the Competition Model framework have examinedprocessing in over 15 languages (Bates McNew MacWhinney Devescoviamp Smith 1982 Kail 1989 MacWhinney Bates amp Kliegl 1984MacWhinney Pleh amp Bates 1985 McDonald 1986 Sokolov 1989) Ina typical Competition Model experiment subjects are presented withsimple transitive sentences and are asked to decide which noun refers tothe agent of the sentence A consistent nding from these studies is that ifthe cues for agentivity are placed in competition with each other thechoice of one of the nouns as the agent can be modelled by multiplicativecue integration (McDonald amp MacWhinney 1989) In these off-linestudies it has been found that the primary determinant of cue strength iscue reliability In other words when subjects are given enough time topermit full deliberate consideration of all competing cues and interpreta-tions they integrate cues in a way that maximises the probable correctnessof their nal interpretation

Newer work in the Competition Model framework (Hernandez Batesamp Avila 1994 Kilborn 1989 Li Bates amp MacWhinney 1993 McDonaldamp MacWhinney 1995 Mimica Sullivan amp Smith 1994) has attempted toextend the earlier off-line models to the study of on-line sentenceprocessing The paradigm used in many of these studies is a speeded cross-modal picture-choice task where participants listen to a transitive sentencewhile making a choice between two candidate agents that are presentedvisually Both candidates are named in the sentence but only one can beselected as the agent Participants are encouraged to respond even beforethe sentence is nished if they are condent of their interpretation Theuse of this speeded picture-choice task provides insights into how thecomplex interaction between various cues unfolds over time There aretwo ways in which the results from this on-line technique further supportthe emphasis on cue reliability that emerged from the earlier off-linestudies First it has been found that the strongest and most reliable cueslead to the fastest reaction times Second it has been found that any typeof competition or disagreement between cues results in inhibition and

















































marked)











Results




141

TAB

LE2

Prop

ortio

nof

Firs

tN

oun

Cho

ices

per

Con

ditio

nin

the

Hum

anD

ata

and

inth

eM

odel

Rus

sian

Ger

man

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngD

ata

Net

wor

kD

ata

Net

wor

k

anin

anim

SVO

unm

ark

unm

ark

086

095

093

097

mar

ked

100

100

098

099

mar

ked

unm

ark

098

099

100

098

mar

ked

098

100

097

099

OV

Sun

mar

kun

mar

k0

920

950

930

97m

arke

d0

270

020

320

08m

arke

dun

mar

k0

020

000

040

00m

arke

d0

020

000

070

00

inan

imSV

Oun

mar

kun

mar

k0

940

990

970

98m

arke

d0

971

000

980

99m

arke

dun

mar

k1

000

991

001

00m

arke

d0

951

000

961

00O

VS

unm

ark

unm

ark

092

099

100

098

mar

ked

028

002

045

080

mar

ked

unm

ark

006

000

010

003

mar

ked

003

000

009

000

inan

iman

imSV

Oun

mar

kun

mar

k0

580

780

550

50m

arke

d0

881

000

850

99m

arke

dun

mar

k0

860

890

860

48m

arke

d0

951

000

940

99O

VS

unm

ark

unm

ark

066

078

056

050

mar

ked

025

001

018

000

mar

ked

unm

ark

001

000

006

000

mar

ked

002

000

002

000

142

TAB

LE2

(Con

td)

Prop

ortio

nof

Firs

tN

oun

Cho

ices

per

Con

ditio

nin

the

Hum

anD

ata

and

inth

eM

odel

Rus

sian

Ger

man

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngD

ata

Net

wor

kD

ata

Net

wor

k

inan

imSV

Oun

mar

kun

mar

k0

890

770

830

92m

arke

d1

001

000

960

99m

arke

dun

mar

k0

970

740

980

97m

arke

d0

971

001

001

00O

VS

unm

ark

unm

ark

088

077

081

092

mar

ked

027

000

039

010

mar

ked

unm

ark

009

000

007

020

mar

ked

003

000

012

000

143







145
















Discussion



















































Discussion





160








GENERAL DISCUSSION


































































392





APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1

















































marked)











Results




141

TAB

LE2

Prop

ortio

nof

Firs

tN

oun

Cho

ices

per

Con

ditio

nin

the

Hum

anD

ata

and

inth

eM

odel

Rus

sian

Ger

man

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngD

ata

Net

wor

kD

ata

Net

wor

k

anin

anim

SVO

unm

ark

unm

ark

086

095

093

097

mar

ked

100

100

098

099

mar

ked

unm

ark

098

099

100

098

mar

ked

098

100

097

099

OV

Sun

mar

kun

mar

k0

920

950

930

97m

arke

d0

270

020

320

08m

arke

dun

mar

k0

020

000

040

00m

arke

d0

020

000

070

00

inan

imSV

Oun

mar

kun

mar

k0

940

990

970

98m

arke

d0

971

000

980

99m

arke

dun

mar

k1

000

991

001

00m

arke

d0

951

000

961

00O

VS

unm

ark

unm

ark

092

099

100

098

mar

ked

028

002

045

080

mar

ked

unm

ark

006

000

010

003

mar

ked

003

000

009

000

inan

iman

imSV

Oun

mar

kun

mar

k0

580

780

550

50m

arke

d0

881

000

850

99m

arke

dun

mar

k0

860

890

860

48m

arke

d0

951

000

940

99O

VS

unm

ark

unm

ark

066

078

056

050

mar

ked

025

001

018

000

mar

ked

unm

ark

001

000

006

000

mar

ked

002

000

002

000

142

TAB

LE2

(Con

td)

Prop

ortio

nof

Firs

tN

oun

Cho

ices

per

Con

ditio

nin

the

Hum

anD

ata

and

inth

eM

odel

Rus

sian

Ger

man

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngD

ata

Net

wor

kD

ata

Net

wor

k

inan

imSV

Oun

mar

kun

mar

k0

890

770

830

92m

arke

d1

001

000

960

99m

arke

dun

mar

k0

970

740

980

97m

arke

d0

971

001

001

00O

VS

unm

ark

unm

ark

088

077

081

092

mar

ked

027

000

039

010

mar

ked

unm

ark

009

000

007

020

mar

ked

003

000

012

000

143







145
















Discussion



















































Discussion





160








GENERAL DISCUSSION


































































392





APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1









Results




141

TAB

LE2

Prop

ortio

nof

Firs

tN

oun

Cho

ices

per

Con

ditio

nin

the

Hum

anD

ata

and

inth

eM

odel

Rus

sian

Ger

man

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngD

ata

Net

wor

kD

ata

Net

wor

k

anin

anim

SVO

unm

ark

unm

ark

086

095

093

097

mar

ked

100

100

098

099

mar

ked

unm

ark

098

099

100

098

mar

ked

098

100

097

099

OV

Sun

mar

kun

mar

k0

920

950

930

97m

arke

d0

270

020

320

08m

arke

dun

mar

k0

020

000

040

00m

arke

d0

020

000

070

00

inan

imSV

Oun

mar

kun

mar

k0

940

990

970

98m

arke

d0

971

000

980

99m

arke

dun

mar

k1

000

991

001

00m

arke

d0

951

000

961

00O

VS

unm

ark

unm

ark

092

099

100

098

mar

ked

028

002

045

080

mar

ked

unm

ark

006

000

010

003

mar

ked

003

000

009

000

inan

iman

imSV

Oun

mar

kun

mar

k0

580

780

550

50m

arke

d0

881

000

850

99m

arke

dun

mar

k0

860

890

860

48m

arke

d0

951

000

940

99O

VS

unm

ark

unm

ark

066

078

056

050

mar

ked

025

001

018

000

mar

ked

unm

ark

001

000

006

000

mar

ked

002

000

002

000

142

TAB

LE2

(Con

td)

Prop

ortio

nof

Firs

tN

oun

Cho

ices

per

Con

ditio

nin

the

Hum

anD

ata

and

inth

eM

odel

Rus

sian

Ger

man

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngD

ata

Net

wor

kD

ata

Net

wor

k

inan

imSV

Oun

mar

kun

mar

k0

890

770

830

92m

arke

d1

001

000

960

99m

arke

dun

mar

k0

970

740

980

97m

arke

d0

971

001

001

00O

VS

unm

ark

unm

ark

088

077

081

092

mar

ked

027

000

039

010

mar

ked

unm

ark

009

000

007

020

mar

ked

003

000

012

000

143







145
















Discussion



















































Discussion





160








GENERAL DISCUSSION


































































392





APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1


141

TAB

LE2

Prop

ortio

nof

Firs

tN

oun

Cho

ices

per

Con

ditio

nin

the

Hum

anD

ata

and

inth

eM

odel

Rus

sian

Ger

man

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngD

ata

Net

wor

kD

ata

Net

wor

k

anin

anim

SVO

unm

ark

unm

ark

086

095

093

097

mar

ked

100

100

098

099

mar

ked

unm

ark

098

099

100

098

mar

ked

098

100

097

099

OV

Sun

mar

kun

mar

k0

920

950

930

97m

arke

d0

270

020

320

08m

arke

dun

mar

k0

020

000

040

00m

arke

d0

020

000

070

00

inan

imSV

Oun

mar

kun

mar

k0

940

990

970

98m

arke

d0

971

000

980

99m

arke

dun

mar

k1

000

991

001

00m

arke

d0

951

000

961

00O

VS

unm

ark

unm

ark

092

099

100

098

mar

ked

028

002

045

080

mar

ked

unm

ark

006

000

010

003

mar

ked

003

000

009

000

inan

iman

imSV

Oun

mar

kun

mar

k0

580

780

550

50m

arke

d0

881

000

850

99m

arke

dun

mar

k0

860

890

860

48m

arke

d0

951

000

940

99O

VS

unm

ark

unm

ark

066

078

056

050

mar

ked

025

001

018

000

mar

ked

unm

ark

001

000

006

000

mar

ked

002

000

002

000

142

TAB

LE2

(Con

td)

Prop

ortio

nof

Firs

tN

oun

Cho

ices

per

Con

ditio

nin

the

Hum

anD

ata

and

inth

eM

odel

Rus

sian

Ger

man

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngD

ata

Net

wor

kD

ata

Net

wor

k

inan

imSV

Oun

mar

kun

mar

k0

890

770

830

92m

arke

d1

001

000

960

99m

arke

dun

mar

k0

970

740

980

97m

arke

d0

971

001

001

00O

VS

unm

ark

unm

ark

088

077

081

092

mar

ked

027

000

039

010

mar

ked

unm

ark

009

000

007

020

mar

ked

003

000

012

000

143







145
















Discussion



















































Discussion





160








GENERAL DISCUSSION


































































392





APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1

TAB

LE2

Prop

ortio

nof

Firs

tN

oun

Cho

ices

per

Con

ditio

nin

the

Hum

anD

ata

and

inth

eM

odel

Rus

sian

Ger

man

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngD

ata

Net

wor

kD

ata

Net

wor

k

anin

anim

SVO

unm

ark

unm

ark

086

095

093

097

mar

ked

100

100

098

099

mar

ked

unm

ark

098

099

100

098

mar

ked

098

100

097

099

OV

Sun

mar

kun

mar

k0

920

950

930

97m

arke

d0

270

020

320

08m

arke

dun

mar

k0

020

000

040

00m

arke

d0

020

000

070

00

inan

imSV

Oun

mar

kun

mar

k0

940

990

970

98m

arke

d0

971

000

980

99m

arke

dun

mar

k1

000

991

001

00m

arke

d0

951

000

961

00O

VS

unm

ark

unm

ark

092

099

100

098

mar

ked

028

002

045

080

mar

ked

unm

ark

006

000

010

003

mar

ked

003

000

009

000

inan

iman

imSV

Oun

mar

kun

mar

k0

580

780

550

50m

arke

d0

881

000

850

99m

arke

dun

mar

k0

860

890

860

48m

arke

d0

951

000

940

99O

VS

unm

ark

unm

ark

066

078

056

050

mar

ked

025

001

018

000

mar

ked

unm

ark

001

000

006

000

mar

ked

002

000

002

000

142

TAB

LE2

(Con

td)

Prop

ortio

nof

Firs

tN

oun

Cho

ices

per

Con

ditio

nin

the

Hum

anD

ata

and

inth

eM

odel

Rus

sian

Ger

man

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngD

ata

Net

wor

kD

ata

Net

wor

k

inan

imSV

Oun

mar

kun

mar

k0

890

770

830

92m

arke

d1

001

000

960

99m

arke

dun

mar

k0

970

740

980

97m

arke

d0

971

001

001

00O

VS

unm

ark

unm

ark

088

077

081

092

mar

ked

027

000

039

010

mar

ked

unm

ark

009

000

007

020

mar

ked

003

000

012

000

143







145
















Discussion



















































Discussion





160








GENERAL DISCUSSION


































































392





APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1

TAB

LE2

(Con

td)

Prop

ortio

nof

Firs

tN

oun

Cho

ices

per

Con

ditio

nin

the

Hum

anD

ata

and

inth

eM

odel

Rus

sian

Ger

man

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngD

ata

Net

wor

kD

ata

Net

wor

k

inan

imSV

Oun

mar

kun

mar

k0

890

770

830

92m

arke

d1

001

000

960

99m

arke

dun

mar

k0

970

740

980

97m

arke

d0

971

001

001

00O

VS

unm

ark

unm

ark

088

077

081

092

mar

ked

027

000

039

010

mar

ked

unm

ark

009

000

007

020

mar

ked

003

000

012

000

143







145
















Discussion



















































Discussion





160








GENERAL DISCUSSION


































































392





APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1







145
















Discussion



















































Discussion





160








GENERAL DISCUSSION


































































392





APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1
















Discussion



















































Discussion





160








GENERAL DISCUSSION


































































392





APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1

















































Discussion





160








GENERAL DISCUSSION


































































392





APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1


160








GENERAL DISCUSSION


































































392





APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1








GENERAL DISCUSSION


































































392





APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1
































































392





APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1

APP

END

IX2

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N

1-A

nim

acy

N2-

Ani

mac

yC

ong

urat

ion

N1-

Mar

king

N2-

Mar

king

Ger

man

Rus

sian

1an

iman

imSV

Oun

mar

kun

mar

kD

ieM

utte

rsu

cht

die

Toc

hter

M

atotildeisIuml

cIumlet

docIumlotilde

2m

arke

dD

ieM

utte

rsu

cht

den

Sohn

M

atotildeisIuml

cIumlet

syna

3

mar

ked

unm

ark

Der

Vat

ersu

cht

die

Toc

hter

O

tec

isIumlcIuml e

tdo

cIumlotilde

4m

arke

dD

erV

ater

such

tde

nSo

hn

Ote

cisIuml

cIuml et

syna

5

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Mut

ter

Doc

IumlotildeisIuml

cIumlet

mat

otilde 6

mar

ked

Die

Toc

hter

such

tde

rV

ater

D

ocIumlotilde

isIumlcIumle

tot

ec

7m

arke

dun

mar

kD

enSo

hnsu

cht

die

Mut

ter

Syna

isIumlcIumle

tm

atotilde

8m

arke

dD

enSo

hnsu

cht

der

Vat

er

Syna

isIumlcIumle

tot

ec

9in

anim

SVO

unm

ark

unm

ark

Die

Mut

ter

such

tdi

eT

orte

M

atotildeisIuml

cIumlet

tort

10

mar

ked

Die

Mut

ter

such

tde

nT

elle

rM

atotildeisIuml

cIumlet

tare

lku

11m

arke

dun

mar

kD

erV

ater

such

tdi

eT

orte

O

tec

isIumlcIumle

tto

rt

12m

arke

dD

erV

ater

such

tde

nT

elle

rO

tec

isIumlcIuml e

tta

relk

u13

OV

Sun

mar

kun

mar

kD

ieT

ocht

ersu

cht

die

Blu

me

Doc

IumlotildeisIuml

cIumlet

cvet

ok

14m

arke

dD

ieT

ocht

ersu

cht

der

Lof

fel

Doc

IumlotildeisIuml

cIuml et

lozIumlk

a15

mar

ked

unm

ark

Den

Sohn

such

tdi

eB

lum

eSy

naisIuml

cIumlet

cvet

ok

16m

arke

dD

enSo

hnsu

cht

der

Lof

fel

Syna

isIumlcIuml e

tlo

zIuml ka

17in

anim

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

ocht

er

Cve

tok

isIumlcIuml e

tdo

cIumlotilde 18

mar

ked

Die

Blu

me

such

tde

nSo

hn

Cve

tok

isIumlcIumle

tsy

na

19m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

ocht

er

Loz

Iumlka

isIumlcIuml e

tdo

cIumlotilde

20m

arke

dD

erL

offe

lsuc

htde

nSo

hn

Loz

Iumlka

isIumlcIumle

tsy

na

21O

VS

unm

ark

unm

ark

Die

Tor

tesu

cht

die

Mut

ter

Tor

tisIuml

cIuml et

mat

otilde 22

mar

ked

Die

Tor

tesu

cht

der

Vat

er

Tor

tisIuml

cIumlet

otec

23

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eM

utte

rT

arel

kuisIuml

cIuml et

mat

otilde 24

mar

ked

Den

Tel

ler

such

tde

rV

ater

T

arel

kuisIuml

cIumlet

otec

168

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1

APP

END

IX2

(Con

td)

Des

ign

and

exam

ple

sent

ence

sus

edin

the

pict

ure

choi

ceta

sk

Cel

lno

N1-

Ani

mac

yN

2-A

nim

acy

Con

gur

atio

nN

1-M

arki

ngN

2-M

arki

ngG

erm

anR

ussi

an

25in

anim

SVO

unm

ark

unm

ark

Die

Blu

me

such

tdi

eT

orte

C

veto

kisIuml

cIuml et

tort

26

mar

ked

Die

Blu

me

such

tde

nT

elle

rC

veto

kisIuml

cIumlet

tare

lku

27m

arke

dun

mar

kD

erL

offe

lsuc

htdi

eT

orte

L

ozIumlk

aisIuml

cIumlet

tort

28

mar

ked

Der

Lof

fels

ucht

den

Tel

ler

Loz

Iumlka

isIumlcIumle

tta

relk

u29

OV

Sun

mar

kun

mar

kD

ieT

orte

such

tdi

eB

lum

eT

ort

isIumlcIumle

tcv

etok

30

mar

ked

Die

Tor

tesu

cht

der

Lof

fel

Tor

tisIuml

cIuml et

lozIuml k

a31

mar

ked

unm

ark

Den

Tel

ler

such

tdi

eB

lum

eT

arel

kuisIuml

cIumlet

cvet

ok

32m

arke

dD

enT

elle

rsu

cht

der

Lof

fel

Tar

elku

isIumlcIumle

tlo

z Iumlka

169








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1








171

APP

END

IX4

F-va

lues

and

effe

ctsi

zes

inth

ere

actio

ntim

esfo

rth

ehu

man

data

and

the

num

ber

ofcy

cles

until

crite

rion

inth

eco

nnec

tioni

stne

twor

k

Rea

ctio

ntim

esad

just

edto

the

end

of1s

tno

unR

eact

ion

times

adju

sted

toth

een

dof

verb

Rea

ctio

ntim

esad

just

edto

the

end

of2n

dno

un

cycl

es

var

netw

ork

Eff

ect

F 1(1

26)

F 2(1

193

)F 1

(12

6)F 2

(11

93)

F 1(1

26)

F 2(1

193

)

N1-

Ani

mac

y24

8

21

5

4

3(6

0)

(60

)(1

10)

48

N2-

Ani

mac

y4

12

34

72

08

7

113

(1

6)

(25

)N

1-M

arki

ng53

3

18

81

50

1

16

73

49

9

16

75

(4

56)

(45

2)(4

21)

(46

5)(3

73)

(36

8)81

4N

2-M

arki

ng21

8

249

205

221

83

8

0

(66

)(6

0)

(61

)(6

1)

(22

)(1

8)

Lan

guag

eN

1-A

nim

acy

102

31

123

2

913

5

72

(2

3)

(24

)(2

3)

(16

)2

3L

angu

age

N1-

Mar

king

38

105

3

39

3

32

93

(2

5)

(26

)(2

0)

30

Con

gur

atio

nN

1-M

arki

ng25

1

51

3

25

1

45

7

24

6

46

1

(1

37)

(12

3)(1

30)

(12

7)(1

12)

(10

1)C

ong

urat

ion

N2-

Mar

king

130

89

10

4

79

126

9

1

(18

)(2

1)

(15

)(2

2)

(16

)(2

0)

N1-

Mar

king

N2-

Mar

king

184

138

181

123

19

5

12

1

(42

)(3

3)

(39

)(3

4)

(35

)(2

7)

Lan

guag

eC

ong

urat

ion

30

82

3

57

6

35

39

N1-

Ani

mac

y(2

9)

(21

)(1

0)

Lan

guag

eC

ong

urat

ion

44

9

2

45

82

4

48

3

N1-

Mar

king

(24

)(2

2)

(23

)(2

0)

(18

)C

ong

urat

ion

N1-

Ani

mac

y4

13

64

43

05

14

7N

2-M

arki

ng(1

0)

Con

gur

atio

nN

1-M

arki

ng13

5

55

126

4

914

2

49

N2-

Mar

king

(17

)(1

3)

(15

)(1

3)

(15

)(1

3)

Eff

ect

size

s(i

npe

rcen

tage

ofex

peri

men

talv

aria

nce)

are

give

non

lyif

the

effe

ctor

inte

ract

ion

acco

unte

dfo

rm

ore

than

1of

vari

ance

Plt

000

1

Plt

001

P

lt0

05n

oas

teri

sk0

05

ltP

lt0

1

processing of morphological and semantic cues in russian ... · 134 kempeandmacwhinney...

Documents