It’s all about gender: Hebrew speakers’ judgment of adjective plural agreement

Dorit Ravid1 · Rachel Schiff2
Received: 20 March 2013 / Accepted: 19 February 2015 / Published online: 4 March 2015
© Springer Science+Business Media Dordrecht 2015
Abstract The current study investigated the role of morpho-phonological and syntactic
factors in judging the marking of Hebrew adjectives in agreement with plural
nouns of differing categories of regularity. Participants were 36 literate adult speakers
of Hebrew, who were administered a judgment test of 144 sentences (half grammatical
and half ungrammatical), each containing a plural adjective in agreement with
a nominal head. The task sentences were divided into three different syntactic configurations
of the adjective in relation to the noun—predicative adjectives, attributive
adjectives in sentence-initial NPs and attributive adjectives in sentence-final NPs. The
agreement-assigning nouns were classified according to three morphological criteria:
Noun gender (masculine and feminine), Noun suffix (regular, irregular, and ambiguous),
and noun stem (changing and non-changing stems). Results showed that gender
was the most important factor affecting agreement judgment: adjectives agreeing with
ambiguous and opaque feminine nouns taking irregular suffixes yielded the lowest
accuracy scores and the longest reaction times. Syntactic position affected judgment
scores as well, with predicative and attributive-initial adjectives being more difficult
to process than attributive-final adjectives. These results support the view that categories
of number and gender emerge from the morpho-phonological properties of
words learned in their syntactic and semantic contexts.
Keywords Inflectional morphology · Hebrew · Nouns · Adjectives · Number/gender
agreement · Stem change · Syntactic position · Morpho-phonology
Supported by an Israel Science Foundation grant No. 79/08 to Dorit Ravid and Rachel Schiff.
B D. Ravid
doritr@post.tau.ac.il
1 School of Education, Tel Aviv University, Tel Aviv 69978, Israel
2 Bar Ilan University, Tel Aviv, Israel
328 D. Ravid, R. Schiff
Table 1 Hebrew adjective agreement with head nouns in the attributive (NP) position and in the predicative
position
Attributive position Predicative position
Masculine
Singular
sir gadol ‘pot big’
big pot
ha-sir gadol ‘the-pot big’
the pot is big
Masculine
Plural
sirim gdolim ‘pots big, Pl’
big pots
ha-sirim gdolim ‘the-pots big, Pl’
the pots are big
Feminine
Singular
sira gdola ‘boat, Fm big, Fm’
big boat
ha-sira gdola ‘the-boat, Fm big, Fm’
the boat is big
Feminine
Plural
sirot gdolot ‘boats, Fm, Pl big, Fm, Pl’
big boats
ha-sirot gdolot ‘the-boats, Fm big, Fm, Pl’
the boats are big
1 Introduction
Agreement—formal or semantic concord between two syntactic elements—is a
prevalent feature in the world’s languages (Comrie 1984; Corbett 2003, 2006; Steele
1978; Wechsler 2009). Linguistic analyses reveal that agreement phenomena share
the same framework across languages: the source of agreement, the controller, is
coupled with the target marked for agreement in a particular domain (Baker 2008;
Moravcsik 1978; Tiberius et al. 2002). Corbett (1998) designates three principles
for characterizing canonical agreement. First, that canonical agreement is redundant
rather than informative; second, that it is syntactically simple; and third, that its expression
is close to canonical inflectional morphology. Adjective agreement with
noun number and gender in Hebrew fulfills these three criteria. It is redundant, in
the sense that both the noun and adjective in agreement with it are marked with
a suffix that expresses the same morpho-syntactic properties—noun number (singular
or plural) and gender (masculine or feminine). It is syntactically simple, in
the sense that both attributive and predicative adjectives are marked for agreement
(see Table 1 for a detailed presentation). And moreover, gender is the most prevalent
morphological phenomenon in Hebrew nominals as the morpho-phonological
organizer of the noun lexicon (Glinert 2005), marking adjectives and participial
verbs in agreement with noun gender and number (Berman 1978; Ravid 1995;
Schwarzwald 2002). Cross-linguistic research indicates that such gender systems,
with the same two gender categories marking both grammatical and biological gender,
strongly affect their speakers’ cognition from early on (Sera et al. 2002). In such
gender systems, agreement is a major cue for acquisition.
Where number-gender agreement is so central in the language, it constitutes evidence
for a noun’s inherent gender (Dixon and Aikhenvald 2004), serving a major
role in typical and atypical morphological development (Lew-Williams and Fernald
2007; Rakhlin et al. 2014). Testing agreement marking on adjectives would thus
indicates whether a speaker, especially a child, knows the gender of a given noun
(Brehmer and Rothweiler 2012). In previous studies we explored how grammatical
gender interacts with suffix regularity and stem change in Hebrew plural noun
formation and adjective agreement within the noun phrase (Ravid and Schiff 2012;
It’s all about gender: Hebrew speakers’ judgment of adjective plural agreement 329
Schiff and Ravid 2012, 2013). In the current study, we investigate judgment of adjective
number-gender agreement with plural nouns at different syntactic positions in
Hebrew-speaking adults, as part of a large-scale psycholinguistic experimental study
of the development of adjective agreement in typically- and atypically-developing
children and adolescents.
The study was conducted within the usage-based approach to linguistics and psycholinguistics,
which holds that past experience with relevant linguistic phenomena
shapes subsequent linguistic performance in a learning system (Bybee 2007;
Chang 2009; Haskell et al. 2010). Studying literate adults’ processing of numbergender
agreement in Hebrew is meant to provide a robust baseline for the examination
of children’s developmental attainments in this central area of Hebrew morphology.
Adult input is the main source available to the child regarding the relative
distributions of words and morphemes in the ambient language (Behrens 2006;
Hoff-Ginsberg 1985; Malsen et al. 2004). While the current study’s findings do not
constitute direct adult input to children, the patterning of adults’ judgments of adjectives
in agreement with plural nouns with different morphological properties and in
different syntactic positions can serve to generate predictions regarding the role of
these factors in morphological acquisition (Da˛browska 2008; Krajewski et al. 2012).
Under the usage-based approach, morpho-lexical properties such as number and
gender emerge as higher-order regularities from the process of learning semantic and
phonological properties of words and word adjacencies (Gillespie and Pearlmutter
2011; Mirkovic et al. 2008), as well as morphological properties (in languages with
morphological gender assignment). It thus follows that the different linguistic factors
that affect the agreement process in a language need to be taken into consideration
in tasks testing agreement. In the current context, our investigation focused on the
interface of gender with noun morpho-phonology and different syntactic positions in
the sentence as situations that cause opacity, looking at correct judgment and reaction
time.
A major prediction of the study is that opaque agreement situations would cause
problems of judgments as to grammaticality among participants (Badecker and Kuminiak
2007; Hartsuiker et al. 2003; Vigliocco and Franck 2001). In the current
study, there were two such factors: (i) When the morphological form of the noun
does not provide unambiguous evidence as to its gender—a well-known source of
processing errors in Hebrew acquisition and usage (Berman 1985; Ravid and Schiff
2012); and (ii) when the syntactic positioning of the adjective in relation to the head
noun makes it harder to process agreement (Barlow 1991; Corbett 1979). Specifically,
we predicted (i) that nouns with irregular or ambiguous number-gender morphology,
and in particular feminine nouns, especially in conjunction with stem changes, would
impede judgment and slow down reaction time; and (ii) that distance between the
adjective and the head noun, or interference between them, would also result in lower
accuracy scores and longer RTs.
1.1 Gender-number in Hebrew nominals
The nominal Hebrew lexicon is classified into masculine or feminine gender for
both animate and inanimate nouns, typically heralded by singular noun phonology
330 D. Ravid, R. Schiff
Table 2 Regular Hebrew gender-number morpho-phonology in nouns
Masculine nouns Feminine nouns
Consonant-final -e final -a final -t final
Singular sar ‘minister’
sir ‘pot’
kone ‘buyer’
mixse ‘lid’
sara ‘minister, Fm’
mishpaxa ‘family’
ganévet ‘thief, Fm’a
mapit ‘napkin’
xanut ‘store’
Plural sarim ‘ministers’
sirim ‘pots’
konim ‘buyers’
mixsim ‘lids’
sarot ‘ministers, Fm’
mishpaxot ‘families’
ganavot ‘thieves, Fm’
mapiyot ‘napkins’
xanuyot ‘stores’
aAs most Hebrew nominals carry final stress, stress is marked only when penultimate.
Table 3 Irregular Hebrew gender-number morpho-phonology in nouns (plural -im is irregular on feminine
nouns, plural -ot is irregular on masculine nouns)
Masculine nouns Feminine nouns
Regular singular
phonology
Ambiguous
singular phonology
-a/t final
Regular singular
phonology
Ambiguous
singular phonology
Consonant-final
Singular rexov ‘street’
sulam ‘ladder’
taklit ‘record’
masa ‘journey’
mila ‘word’
tolá’at ‘worm’
kos ‘glass’
ir ‘city’
Plural rexovot ‘streets’
sulamot ‘ladders’
taklitim ‘records’
masa’ot ‘journeys’
milim ‘words’
tola’im ‘worms’
kosot ‘glasses’
arim ‘cities’
(Schwarzwald 1982). Gender distinctions affect both singular and plural formation
of Hebrew nouns, as summarized in Tables 2 and 3 and delineated below.
Singular morpho-phonology Native masculine nouns either end with a consonant
or else with a final stressed -e. Feminine nouns end with a stressed -a or else
with -t.1 Gender is an inflectional property of animate nouns, where the masculine
singular stem is almost always the basic noun form which can be inflected for feminine
gender (e.g., xatul/xatula ‘cat/Fm’,2 safran/safranit ‘librarian/Fm’). In inanimate
nouns, however, inherent gender is an property of noun morpho-phonology
(Glinert 2005; Schwarzwald 2002) expressed in the phonological forms of the two
major derivational devices in Hebrew nouns—non-linear Semitic patterns and linear
suffixes (Bolozky 2007; Ravid 2012; Schwarzwald 2005). For example, nouns in the
instrument pattern maCCeC (e.g., masrek ‘comb’) are masculine as the pattern ends
with a root consonant, whereas nouns in the corresponding pattern maCCeCa (e.g.,
mavxena ‘test tube’) are feminine, as the pattern ends with a stressed a.
1Spelled by š rather than by homophonous ˆ, which is not a function letter (Ravid 2012).
2Fm = feminine; Pl = plural.
It’s all about gender: Hebrew speakers’ judgment of adjective plural agreement 331
Virtually all animate nouns and most inanimate nouns are regular in the sense
that their basic stem phonology matches masculine gender,3 (top part of Table 2).
However singular inanimate noun phonology may be ambiguous, that is, mimic the
opposite gender phonology (a and t ending on masculine nouns, consonant ending on
feminine nouns), as shown at the top part of Table 3. For example, masculine taklit
‘record’ (root q-l-t
"
, spelled with ˆ) which mimics the feminine suffix -t (cf. genuine
feminine matlit ‘rag’, spelled with š),4 (Ravid and Schiff 2012). Spelling conventions,
which are inherently morphological, can help resolve singular noun ambiguity
in masculine nouns, as these lack the obligatory final „ and š of real feminines
(Ravid 2012), but they are not relevant to ambiguity in singular feminine nouns. This
difference in gender marking of singular nouns already generates the prediction that
masculine nouns should be easier to pluralize than feminine nouns with shorter RTs.
Plural morpho-phonology Choice of plural suffix generally follows noun gender:
Masculine nouns take the plural suffix -im, feminine nouns take the plural suffix -ot,
as shown in Table 2, bottom part. While animate nouns are overwhelmingly regular,
non-animate nouns sometimes exhibit irregular pluralization, that is, nouns taking the
opposite-gender suffix, as shown at the bottom of Table 3: masculine rexov/rexovot
‘street/s’ (instead of rexovim), feminine mila/milim ‘word/s’ (instead of milot). Irregular
plurals are not well balanced across the two genders, being skewed towards
masculine nouns which are more numerous than feminine nouns in the core lexicon
of Hebrew (Ravid et al. 2008; Tubul 2003) and with more irregular nouns (Levy
1980). In addition, irregular masculines tend to be cued by final voiced segments,
including sonorants, vowels and voiced obstruents (see examples in Table 3, bottom
part), while irregular feminines lack such cue segments (Ravid and Schiff 2009). It is
thus reasonable to predict that the less numerous, cue-less irregular feminines should
be more difficult to judge and take longer to do so than irregular masculines.
A second type of plural suffix irregularity derives from ambiguous singular marking
(see top of Table 3). Ambiguous singular nouns mimic the opposite gender
phonology, as in feminine gader ‘fence’ ending with a consonant or masculine mabat
‘glance’ ending with t (Ravid and Schiff 2012). Ambiguous nouns can take regular
plural suffixation, as in masculine taklitim ‘records’ and feminine kosot ‘glasses’
(bottom part of Table 3)—that is, ignoring the misleading phonological cues. But
they can also take irregular plural suffixation, as in masa’ot ‘journeys’ or arim ‘cities’
(Table 3). Therefore, we predict that ambiguous nouns should also hinder and slow
down correct judgment, especially feminine nouns.
3With very few exceptions of animate feminines with masculine phonology such as em ‘mother’ and a
number of Capra animals such as ez ‘goat’ or ya’el ‘ibex’.
4Irregular phonology invades even morphological patterns (Avineri 1976). For example, the CaCeC pattern
denotes masculine nouns, both inanimate (e.g., batsek ‘dough’ and yated ‘stake’) and animate (e.g.,
xaver ‘friend’, shaxen ‘neighbor’), but it also contains feminine nouns with the same masculine morphophonology,
such as inanimate gader ‘fence’ and xatser ‘yard’). And the ubiquitous penultimately marked
CéCeC, which is inherently masculine (bérez ‘tap’, mélex ‘king’), contains several feminine nouns such as
néfesh ‘soul’, régel ‘leg’, or ná’al ‘shoe’.
332 D. Ravid, R. Schiff
Table 4 Hebrew adjective agreement with regular and irregular nouns in the attributive (NP) position and
in the predicative position
Attributive position Predicative position
Masculine
Regular
sirim gdolim ‘pots big, Pl’
big pots
ha-sirim gdolim ‘the-pots big, Pl’
the pots are big
Masculine
Irregular
kirot gdolim ‘walls big, Pl’
big walls
ha-kirot gdolim ‘the-walls big, Pl’
the walls are big
Feminine
Regular
sirot gdolot ‘boats big, Fm, Pl’
big boats
ha-sirot gdolot ‘the-boats big, Fm, Pl’
the boats are big
Feminine
Irregular
milim gdolot ‘words big, Fm, Pl’
big words
ha-milim gdolot ‘the-words big, Fm, Pl’
the words are big
1.2 Stem types
Irregular suffixation interacts with yet another factor in pluralization—stem change.
While many stems remain unchanged under linear suffixation (e.g., tanur/tanurim
‘oven/s’), others undergo morpho-phonological changes including vowel reduction,
deletion, or change, stop/spirant alternation, and t omission (Ravid 1995, 2006a).
For example, singular masculine dli ‘bucket’ changes to dlay-im in the plural, while
feminine rakévet ‘train’ changes to plural rakav-ot. Stem changes have been shown
to hinder and complicate morphological acquisition in Hebrew-speaking children
(Levin et al. 2001; Ravid and Schiff 2009, 2012). This is because in cases of morphophonological
distance, it is difficult to relate the singular form of a noun such as
ish ‘man’ to its bound form in the plural, as in anash-im ‘people’ (Schiff et al. 2011).
When stem change is coupled with irregular or ambiguous plural marking, we predict
lower accuracy scores and longer RTs.
1.3 Number-gender agreement in adjectives
Hebrew adjectives follow the noun within the noun phrase, as in sir gadol ‘pot big =
big pot’ and sira gdola ‘boat big = big boat’ (Table 1). Plural adjectives, the focus
of the current analysis, agree with the number and gender of the head noun in both
the attributive and the predicative positions. Consequently, adjectives in agreement
with regular nouns carry the same number/gender marking as their head nouns, as
shown in Table 4. Even when the head noun has irregular plural marking, adjectives
are marked according to the inherent gender of the noun, creating a noun phrase or a
sentence with a predicate adjective with two opposing gender suffixes (see Table 4).
Given the larger number of masculine nouns in general and irregular masculines in
particular, the transitional probabilities for sequences such as N-ot A-im (irregular
masculine noun and adjective) are much larger than N-im A-ot (irregular feminine
noun and adjective), as elaborated in Ravid and Schiff (2012). This yields the prediction
that agreement with irregular masculines should be judged more correctly and
with shorter RTs.
Noun stem changes impact on the agreement process in the judgment of the N-A
plural phrase. The need to check singular noun gender in order to select the correct
adjective suffix arises only when the noun suffix is irregular (Table 3). In such cases,
It’s all about gender: Hebrew speakers’ judgment of adjective plural agreement 333
the singular noun stem needs to be retrieved. In case of non-stem changing nouns
(e.g., sulam/sulam-ot ‘ladder/s’), gender determination is easy. But where the stem
changes (e.g., katse/ktsav-ot ‘edge/s’), the morpho-phonological distance between
the free and bound stems interferes with the retrieval process and hence with suffix
determination. This should be especially daunting in singular stems which mimic
opposite-gender phonology (ambiguous marking), such as feminine ir/ar-im ‘city/s’.
1.4 Syntactic position of the adjective
In addition to the morphological variables characterizing the head noun (gender, suffix
regularity and stem change) the current study examines plural adjective agreement
in conjunction with yet another variable—the factor of syntactic adjective position.
Adjectives occupy two major positions in syntax—predicates indicating the nature
of the subject referent, as in the thoughts of the child were happy (predicative adjectives),
or as modifiers in the noun phrase, restricting the meaning of the head noun
(attributive adjectives), as in the child had happy thoughts. In judging the correctness
of plural adjective agreement, attention needs to be paid both to the morphological
properties of the head noun and to their expression in the plural adjective, and in
addition, to the syntactic position of the plural adjective.
Both predicative and attributive adjectives agree with the head noun, so that number/
gender agreement takes place both within the clause and within the noun phrase.
But there are many cases in Hebrew where predicates do not agree with subjects,
especially when in the predicate-first position (Berman 1980; Ravid 1995).5 Given
that agreement relies on a limited scope of planning (Gillespie and Pearlmutter 2011),
it makes sense to expect predicative adjectives to be more difficult and slower to judge
than attributive adjectives, due to distance between subject and predicate and to the
less robust nature of subject-predicate agreement.
Syntactic position is relevant to attributive adjectives as well. As shown in Tables 1
and 4, adjectives follow the head noun, so that the noun plural suffix always precedes
that of the adjective. However, subject noun phrases with attributive adjectives may
appear pre-verbally, at sentence initial position (e.g., ha-maxshavot ha-smexot hayu
be-roshi ‘the happy thoughts were in my mind’; but also in post-verbal position as in
la-ish hayu maxshavot smexot ‘to-the-man were happy thoughts = the man had happy
thoughts’. This has implications for their processing: it makes sense to expect plural
adjective agreement to be more accurate and faster in the post-verbal noun phrase
which appears in sentence final position than in the pre-verbal, sentence initial position.
This is because the sentence-initial position may be followed by more possible
gender-marked nouns, adjectives or verbs which can interfere with the agreement
process.
2 Method
The study examined the impact of noun gender, suffix (ir)regularity, and stem change,
as well as adjective syntactic position, on judgment of adjective agreement with plural
nouns.
5I am grateful to Zohar Livnat for pointing this out in the current context.
334 D. Ravid, R. Schiff
Participants were 36 Israeli monolingual native speakers of Hebrew aged 25–30,
all undergraduate students attending Bar Ilan and Tel Aviv Universities. They were all
from mid-high SES background, with no overt emotional problems and no recorded
developmental disorders or neurological problems, with normal hearing and sight (or
sight corrected by spectacles). In order to assure that they had no reading disability,
they were administered the Schiff and Kahta (2006) Single Word Reading Task, all
scoring above the mean level of 108.37 (4.19) on Accuracy and 87.54 (17.41) on
Speed.
Instruments consisted of the Judgment Task testing knowledge about the correct
assignment of plural marking on adjectives in agreement with nouns. Piloting of components,
structure and instructions was undertaken among 10 pilot participants to ensure
uniformity and exclude confounding variables, with subsequent corrections and
revisions of the materials. Sentences were recorded with the help of a professional
technician and played in random order by the SuperLab computer software. Participants’
responses were recorded by SuperLab and, separately, by student researchers
who had undergone training by the two authors.
Item selection The irregular and ambiguous nouns selected for the task consisted
of the exhaustive list on https://www.safa-ivrit.org/irregulars/index.php—so that, in
fact, we used all irregularly pluralizing nouns listed there. The rest of the nouns on
the task, which were regular, all ranked between 2–3 (out of 10) on the Noun Scale
(Ravid 2006b), and in addition were ranked no more than 2 on a 3-point familiarity
scale by 30 native-speaking students (not the same participants who were eventually
administered the Judgment Task). Nouns with the dual suffix, which is unmarked
for gender (Ravid 2012), were excluded from the task. Adjectives selected for the
task belonged to the core adjectival lexicon of Hebrew as designated in Ravid and
Levie (2010). Since Hebrew adjectives may also undergo stem change, we made sure
that the target adjectives either had nonchanging stems (tov/tovim ‘good/Pl’) or else
shared a-deletion, as demonstrated in the plural forms of gadol/gdolot ‘big/Pl.Fm’,
lavan/levanim ‘white/Pl’, and shaket/shketot ‘quiet/Pl, Fm’. This type of stem change
is among the earliest acquisitions in Hebrew-speaking children (Ravid 1995; Ravid
and Shlesinger 2001). All sentences contained 10–14 syllables so as to control for
sentence length which might affect working memory. Finally, we did our best to
minimize the occurrence of gender markers in the sentence. For example, we used the
gender-neutral past tense copula hayu ‘were’ rather than the gender-specific present
tense hem/hen ‘are, Masc/are, Fm’ copula.
The nouns and adjectives selected were used in the Judgment Task, which required
participants to judge the correctness of 144 sentences, each containing a plural adjective
in agreement with a nominal head. Instructions were as follows: you will hear
a series of sentences. Some of the sentences are correct and some are incorrect. You
have to determine for each sentence whether it is correct or incorrect and press the
appropriate button. To create Noun-Adjective sets for each sentence, 48 nouns were
used altogether, each occurring three times (see below regarding syntactic position),
coupled with 16 adjectives, each occurring 9 times. Half of the 144 task sentences
were grammatical, i.e., the adjective was in correct agreement with the head noun
(e.g., ha-maxshavot ha-mehirot hayu shelanu ‘the-thoughts, Fm the-swift, Pl, Fm
It’s all about gender: Hebrew speakers’ judgment of adjective plural agreement 335
were ours’) and half ungrammatical, so that the adjective was incorrectly marked
for a different gender than the noun (e.g., ha-maxshavot ha-mehirim hayu shelanu
‘the-thoughts, Fm the-swift, Pl, Masc were ours’).
The 144 sentences were divided into three different syntactic configurations of the
adjective in relation to the noun: (1) Sentence-initial NPs containing a plural noun and
an attributive adjective in agreement with it (e.g., grammatical ha-yalkutim ha-kxulim
hayu ba-pina ‘the sachels, Masc the-blue, Pl,Masc were in the corner’); (2) sentencefinal
NPs containing a plural noun and an attributive adjective in agreement with it
(e.g., ungrammatical le-Netan’el hayu kosot shkufim ‘to-Nathaniel were [=Nathaniel
had] cups, Fm transparent, Pl, Masc’); and (3) a plural noun as grammatical subject at
sentence initial position, and a predicative adjective in agreement with it at sentence
final position (e.g., ungrammatical ha-maxshavot shel ha-yeled hayu mehirim ‘thethoughts,
Fm of the-boy were swift, Pl, Masc’). The same noun-adjective set occurred
in all three positions.
The 48 agreement-assigning nouns were classified according to three morphological
criteria: Noun gender (masculine and feminine), Noun suffix (regular, irregular,
and ambiguous), and noun stem (changing and non-changing stems). Thus, the four
task variables tested were syntactic position, noun gender, noun suffix and noun stem.
For example, ha-tslalim ha-shxorot hayu me-axoray ‘the black shadows were behind
me’ is an ungrammatical sentence with an initial NP consisting of a masculine plural
noun shadows carrying a regular suffix and a changing stem (cf. singular tsel
‘shadow’), and an attributive feminine plural adjective ‘black’.
Procedure The task was preceded by four training items of grammatical and ungrammatical
sentences with adjectives requiring correct/incorrect decisions, e.g.,
yesh ba-kita harbe neyarot levanot ‘there (are) in the classroom many papers, Masc
white, Pl, Fm’ (decision: incorrect). After training, the actual experiment started.
Randomized stimuli were presented orally by the SuperLab software program, which
also recorded the response selected and marked the onset of the participant’s response.
To ensure students’ full attention, the Judgment task was administered in
three parts, each containing a randomized list of items, on three consecutive days.
Scoring Judgment measures were Accuracy (correct response on sentence grammaticality)
and Reaction Time to Correct Response. All scores were converted to
percentages. Where necessary, we conducted Bonferroni post-hoc pairwise comparisons
to determine significant differences and sources of interactions.
3 Results
3.1 Accuracy
Our hypotheses were confirmed. Judgment of adjective agreement was higher with
masculine nouns, nouns with regular suffixes, and at the final attributive position. The
statistical details, including the interactions, are presented below.
The following simple effects were found: Gender (F(1, 35) = 29.75, p < .001,
η = .46): Adjectives in agreement with masculine nouns scored higher (M =
336 D. Ravid, R. Schiff
Fig. 1 Interaction of gender × suffix × stem on accuracy
97.38 %) than with feminine nouns (M = 91.05 %). Suffix (F(2, 70) = 18.63,
p <.001, η = .35): Adjectives agreeing with nouns carrying regular suffixes scored
higher (M = 98.55 %) than with nouns bearing irregular suffixes (M = 92.07 %) and
ambiguous marking (M = 92.01 %). Position (F(2, 70) = 4.4, p < .02, η = .11):
Adjectives at final attributive position had a mean score of 95.31 %, significantly
higher than adjectives in predicative position (M = 93.23 %), with initial attributive
position non-differing from these two (M = 94.1 %). Stem did not have an effect.
Suffix interacted with stem (F(2, 70) = 14.95, p < .001, η = .3) and also with
gender (F(2, 70) = 21.78, p <.001, η = .38), both contained in the three-way interaction
of gender, suffix and stem (F(2, 70) = 7.19, p <.002, η = .17) presented in
Fig. 1.
The sources of the interaction presented in Fig. 1 essentially lie in the irregular
and ambiguous feminine items. Regular suffixes show no difference with either stem
type or gender. In general, irregular suffixes score lower than regulars, but in masculine
nouns this difference is not significant. Feminine irregulars with no changing
stems score higher than feminine irregulars with changing stems, one clear source of
this interaction. Ambiguous marking nouns with masculine gender score as high as
regulars, however the two ambiguous feminine categories are the second source of
the interaction with significantly lower scores than all other categories.
3.2 Reaction time to correct response
Again, our hypotheses were mostly confirmed. Correct judgment of adjective agreement
was faster with masculine nouns and nouns with regular suffixes. The statistical
details, including the interactions, are presented below.
The following simple effects were found: Gender (F(1, 35) = 14.29, p < .002,
η = .29): Adjectives in agreement with masculine nouns yielded a shorter reaction
time (M = 1329.57 ms) than those agreeing with feminine nouns (M = 1510.49).
Suffix (F(2, 70) = 219.89, p < .001, η = .36): Adjectives agreeing with nouns
It’s all about gender: Hebrew speakers’ judgment of adjective plural agreement 337
Fig. 2 Interaction of gender, suffix and stem on reaction time
carrying regular suffixes had shorter RTs (M = 1187.83) than with both irregular
(M = 1616.24) and ambiguous nouns (M = 1456.02). Stem and position did not
have simple effects.
Suffix interacted with stem (F(2, 70) = 15.04, p < .001, η = .3) and also with
gender (F(2, 70) = 37.29, p < .001, η = .52). Two three-way interactions were
significant—gender, suffix and stem (F(2, 70) = 14.4, p <.001, η = .29), and gender,
suffix and position (F(4, 140) = 5.03, p <.002, η = .13), presented in Figs. 2
and 3 respectively.
The interaction depicted in Fig. 2 derives from several sources. First, the regular
suffixes have the shortest reaction times. Second, irregular nouns take the most time,
except for feminine non-changing stems. And third, feminine ambiguous nouns take
the most time, and most particularly the category of feminine non-changing stems.
The interaction depicted in Fig. 3 derives from several sources. First, regular suffixes
have the shortest RTs across both genders, with no difference across the three
positions. Irregular nouns take longer, and the predicative position takes longest.
Finally, ambiguous marking nouns all take longer, with no difference among the three
positions.
4 Discussion
Previous investigations have focused on the developing ability of various Hebrewspeaking
populations to produce plural nouns (Ravid and Schiff 2009; Schiff et
al. 2011) or pluralize noun-attributive adjective pairs (Ravid and Schiff 2012;
Schiff and Ravid 2012). The current study differs from its predecessors in several
ways, which made it possible to draw more precise and systematic conclusions about
the impact of grammatical factors on marking Hebrew adjectives in agreement with
plural nouns beyond the simplex notion of noun irregularity. First, it targeted young,
literate adults as the population most likely to exhibit good command of inflectional
338 D. Ravid, R. Schiff
(a)
(b)
Fig. 3 (a) Interaction of gender, suffix and position on reaction time: Adjectives in agreement with masculine
nouns. (b) Interaction of gender, suffix and position on reaction time: Adjectives in agreement with
feminine nouns
processing in their native tongue and thus pinpoint those factors which lower their
performance. Second, it was a judgment rather than a production task, which required
participants to think about the grammaticality of sentences with various structures,
including those which could not have been elicited in a production task (such
as attributive-initial adjectives). Third, it focused directly on adjectives alone, with
properties of noun structure (gender, suffix, stem) taken into account as explaining
rather than dependent variables. Fourth, this study extended and fine-tuned the notion
of plural suffix irregularity to the two categories of irregular plural suffixation
and gender ambiguity. Finally, it broadened the scope of grammatical investigation
from the effect of plural noun morphology (gender, suffix regularity and stem change)
to the syntactic function of the adjective (attributive vs. predicative) and the site in
which it appeared (sentence initial vs. sentence final).
Syntactic position Recall that each of the adjectives appeared in one of three syntactic
conditions, illustrated here by adjective levanot ‘white, Pl, Fm’ in erroneous
It’s all about gender: Hebrew speakers’ judgment of adjective plural agreement 339
agreement with neyarot ‘papers’, a masculine noun taking an irregular feminine suffix
(cf. singular niyar), translated literally: (1) participating in a subject NP located at
the beginning of the sentence (attributive-initial), e.g. ha-neyarot ha-levanot hayu bakita
‘The-papers the-white were in-the-classroom’; (2) participating in a subject NP
located post-verbally at the end of the sentence (attributive-final), e.g. yesh ba-kita
harbe neyarot levanot ‘be in-the-classroom many papers white’; and as a predicative
adjective following a subject NP at sentence final position, e.g., ha-neyarot she-bakita
hayu levanot ‘the papers which-in-the-classroom were white’. In all of the above
cases, participants had to detect that the plural adjective should have been levanim
rather than levanot, in agreement with the inherent masculine gender of niyar ‘paper’.
As predicted, syntactic position was found to affect adjective plural agreement.
Despite the fact that all responses ranged at top percentage (between 93–95 %), adjectives
in attributive-final position scored significantly higher than either of the other
two positions. When interacting with noun gender and suffix irregularity (Fig. 3), the
predicative position emerged as taking the most time to judge. Taken together, these
findings indicate to what extent processing impacts inflectional judgment even in this
literate, adult sample: Implementing correct agreement requires paying close attention
to the properties of the noun while keeping the form of the adjective in mind.
This is more difficult in the predicative position, where the noun at the beginning of
the sentence and the adjective at its end are separated by all the intervening material
of the sentence, in a looser agreement relationship with the noun, as detailed in the
introduction. Also, it is not easy in the attributive-initial position, where the adjective
is directly adjacent to the noun but other nouns in the rest of the sentence vie for
the attention of the judge with their various properties. The attributive-final position
facilitates correct judgment in keeping the adjective adjacent to the head noun as the
most recent information to retain.
Gender and morphology Noun-Adjective pairs in the three syntactic positions
were further classified according to three criteria shown to be important in processing
noun and adjective plurals (Ravid and Schiff 2009, 2012; Schiff and Ravid 2012,
2013): by noun gender (masculine versus feminine), by noun plural suffix (regular,
and two irregular categories—irregulars and ambiguous marking), and by stem type
(changing and non-changing). The most important property affecting Hebrew plural
adjective agreement, which emerged in this study, was found to be noun gender.
In this literate population, average scores were over 80 %, and in most categories
ranged between 85 % to close to 100 %. In general, agreement with nouns having
regular suffixes yielded higher scores than with either of the irregular categories, and
adjectives in agreement with masculine nouns yielded higher scores than with feminine
nouns. But specifically, irregularity hardly affected masculine nouns, with lower
scores restricted to irregular feminine nouns. The vulnerability of agreement with
feminine nouns was highlighted in the relationship revealed between noun gender,
suffix type and stem type (Fig. 1): Hebrew-speaking adults had no trouble at all judging
the correctness of adjectives in agreement with nouns bearing regular suffixes of
both genders and stem types; and they had the same top scores judging agreement
with ambiguous masculine nouns bearing false-feminine singular phonology such as
yalkutim ‘sachels’ or meshotim ‘oars’. But they achieved lower scores on judging
340 D. Ravid, R. Schiff
adjectives in agreement with irregular feminine nouns of both types—irregular feminines
and ambiguous feminine nouns bearing false-masculine singular phonology
such as kosot ‘glasses’ or arim ‘cities’.
Accuracy results were supported by virtually mirror-image reaction times to correct
decisions. Judging adjectives in agreement with feminine nouns took longer than
with masculine nouns, and judging agreement with nouns bearing regular suffixes
was shorter than with the two irregular categories. The combined effect of the three
factors of gender, suffix and stem type yielded the same impact on latency: Agreement
with regularly suffixed nouns, with ambiguous masculines, and with feminines
with non-changing stems took the shortest time, with other irregulars taking much
longer. Feminines with changing stems and ambiguous feminines took the most time
to judge. Moreover, the impact of syntactic position was expressed in this study
through its interaction with noun gender and noun suffix type: While regular suffixes
yielded the same RTs across all positions and in the two genders, irregular feminines
of both types had very long RTs, whereas in the masculine nouns it was only the
predicative and attributive-initial positions that took longer.
Gender and suffix regularity The main source of vulnerability in processing Hebrew
plural agreement of adjectives with nouns, apparent even in highly literate adults
with no learning or language impairment, is thus feminine gender in combination
with suffix irregularity. As presented in the introduction, feminine nouns are less
clearly cued for gender marking and less frequent in their occurrence, and therefore
previous studies on younger or impaired populations showed that they present
a higher obstacle than masculine nouns to learning about agreement. While gender
did not impact agreement with nouns with regular suffixes in this literate adult population,
its effect was detrimental at a specific junction of agreement with feminine
nouns bearing irregular and ambiguous suffixes together with changing stems such
as tolá’at/tola’im ‘worm/s’ or éven/avanim ‘stone/s’. It is important to note that the
problem here was not the actual plural inflection of such nouns, but rather judging
plural adjective agreement with them. It required taking into account their inherent
feminine gender which should yield N-A combinations such as avanim xamot
‘stones, Fm hot, Pl, Fm’. All factors contributed to the opacity of this environment
and hence to slower and less efficient judgment, with the lowest accuracy scores and
the highest RTs on this task reflecting more processing effort: the -im masculine plural
suffix of the noun, in direct opposition to the -ot feminine plural suffix of the
adjective’; stem structure, which impacted plural agreement indirectly through the
morpho-phonological distance between singular and plural noun form; and the relative
rarity of N-im A-ot structures.
A second vulnerable site in processing plural adjective agreement again involves
feminine gender, specifically ambiguous feminine nouns of both stem types—
changing and non-changing. Recall that a feminine noun such as ir ‘city’ has singular
masculine (i.e., unmarked) phonology, whereas a masculine noun such as yalkut
‘satchel’ has feminine (i.e., final -t) phonology. In principle, misleading singular
phonology could cause problems at the stage requiring determination of noun gender.
However our results indicate that ambiguous masculine nouns did not pose any problems
in either accuracy or RT, whereas both changing and non-changing feminines
It’s all about gender: Hebrew speakers’ judgment of adjective plural agreement 341
had low accuracy scores and long reaction times. This is because unlike pseudomasculine
feminines, pseudo-feminine masculines mimic better masculine morphophonological
behavior: While real feminines omit the suffixal a and t before the plural
suffix (e.g., mana/man-ot ‘portion/s’, axot/axay-ot ‘sister/s’), pseudo-feminine a
and t, not being morphological components, are retained before the plural suffix—as
in mada/mada’-im ‘science/s’, mashot/meshot-im ‘oar/s’. That is, ambiguous masculine
nouns only superficially mimic feminine phonology while having masculine
root morphology, as evidenced by their written forms (Ravid 2012). Thus, with the
advent of literacy, ambiguous nouns will be even more robustly identified by the
written cues of their gender, while no such cues help disambiguate false-masculine
feminines such as kos ‘glass’, explaining why even well-educated adult speakers of
Hebrew had lower accuracy scores and longer RTs on feminine ambiguous nouns. In
the current study, the difficulty of retrieving feminine gender from singular feminine
nouns with ambiguous suffixes applied across the board to such nouns with nonchanging
stems (e.g., kos/kos-ot ‘glass/s’) as well as to those with changing stems
(e.g., gader/gder-ot ‘fence/s’). And it was present not only in ambiguous feminines
with irregular suffixes (e.g., ir/arim ‘city/s) but also in those taking regular suffixes,
such as kos/kos-ot ‘glass/s’.
5 Conclusion
The current study examined the judgment of number/gender agreement of adjectives
with plural nouns in a population of literate, educated Hebrew-speaking adults. Regarding
the notion of (ir)regularity, no evidence was found for rules sweeping across
the board. Participants correctly handled judgment of agreement with both regular
and irregular nouns with the same degree of accuracy and at the same speed. What
we found was degrees of irregularity and opacity coalescing at one permanent site—
the less frequent category of feminine noun gender, especially in conjunction with
suffix irregularity—ambiguous singular marking and plural irregular marking. Retrieving
feminine gender from nouns with this particular morpho-phonology in order
to determine the grammaticality of plural adjectives proved problematic even
for this population. At the same time, processing factors associated with the position
and syntactic function of the adjective in the sentence in relation to the head
noun proved to enhance this difficulty. These results support the view that categories
of number and gender emerge from the morpho-phonological properties of words
learned in their syntactic and semantic contexts (Gillespie and Pearlmutter 2011;
Haskell et al. 2010; Mirkovi´c et al. 2004, 2008).
References
Avineri, Y. (1976). Heyxal ha-mishkalim. Tel Aviv: Izre’el (in Hebrew).
Badecker, W., & Kuminiak, F. (2007). Morphology, agreement and working memory retrieval in sentence
production: Evidence from gender and case in Slovak. Journal of Memory and Language, 56, 65–85.
Baker, M. C. (2008). The syntax of agreement and concord. Cambridge: Cambridge University Press.
342 D. Ravid, R. Schiff
Barlow, M. (1991). The agreement hierarchy and grammatical theory. In Proceedings of the Seventeenth
Annual Meeting of the Berkeley Linguistics Society: General session and parasession on the grammar
of event structure (Vol. 17, pp. 30–40).
Behrens, H. (2006). The input-output relationship in first language acquisition. Language and Cognitive
Processes, 21, 2–24.
Berman, R. A. (1978). Modern Hebrew structure. Tel Aviv: Tel Aviv University Publishing Projects.
Berman, R. A. (1980). The case of an (S)VO language: Subjectless constructions in Modern Hebrew.
Language, 56, 759–776.
Berman, R. A. (1985). Acquisition of Hebrew. Hillsdale: Erlbaum.
Brehmer, B., & Rothweiler, M. (2012). The acquisition of gender agreement marking in Polish: A study
of bilingual Polish-German-speaking children. In K. Braunmüller & C. Gabriel (Eds.), Multilingual
individuals and multilingual societies (pp. 81–100). Amsterdam: John Benjamins.
Bolozky, S. (2007). Israeli Hebrew morphology. In A. S. Kaye (Ed.), Morphologies of Asia and Africa
(including the Caucasus) (pp. 283–308). Winona Lake: Eisenbrauns.
Bybee, J. (2007). Frequency of use and the organization of language. Oxford: Oxford University Press.
Chang, F. (2009). Learning to order words: A connectionist model of heavy NP shift and accessibility
effects in Japanese and English. Journal of Memory and Language, 61, 374–397.
Comrie, B. (1984). Reflections on verb agreement in Hindi and related languages. Linguistics, 22, 857–
864.
Corbett, G. (1979). The agreement hierarchy. Journal of Linguistics, 15, 203–224.
Corbett, G. (2003). Agreement: the range of the phenomenon and the principles of the Surrey Database of
Agreement. Transactions of the Philological Society, 101, 155–202.
Corbett, G. (1998). Morphology and agreement. In A. Spencer & A. M. Zwicky (Eds.), The handbook of
morphology (pp. 191–205). London: Blackwell.
Corbett, G. (2006). Agreement. Cambridge: Cambridge University Press.
D ˛ abrowska, E. (2008). Questions with long-distance dependencies: A usage-based perspective. Cognitive
Linguistics, 19, 391–425.
Dixon, R. M. W., & Aikhenvald, A. (Eds.) (2004). Adjective classes: a cross-linguistic typology. Oxford:
Oxford University Press.
Gillespie, M., & Pearlmutter, N. J. (2011). Hierarchy and scope of planning in subject–verb agreement
production. Cognition, 118, 377–397.
Glinert, L. (2005). Modern Hebrew: An essential grammar (3rd ed.) New York: Routledge.
Hartsuiker, R. J., Schriefers, H. J., Bock, K., & Gerdien, M. K. (2003). Morphophonological influences on
the construction of subject–verb agreement. Memory & Cognition, 31, 1316–1326.
Haskell, T. R., Thornton, R., & MacDonald, M. C. (2010). Experience and grammatical agreement: Statistical
learning shapes number agreement production. Cognition, 114, 151–164.
Hoff-Ginsberg, E. (1985). Some contribution of mothers’ speech to their children’s syntactic growth. Journal
of Child Language, 12, 367–385.
Krajewski, G., Lieven, E. V. M., & Theakston, A. L. (2012). Productivity of a Polish child’s inflectional
noun morphology: a naturalistic study. Morphology, 22, 9–34.
Lew-Williams, C., & Fernald, A. (2007). Young children learning Spanish make rapid use of grammatical
gender in spoken word recognition. Psychological Science, 18, 193–198.
Levin, I., Ravid, D., & Rapaport, S. (2001). Morphology and spelling among Hebrew-speaking children:
From kindergarten to first grade. Journal of Child Language, 28, 741–769.
Levy, Y. (1980). The acquisition of gender. Unpublished doctoral dissertation, Hebrew University,
Jerusalem, Israel (in Hebrew).
Malsen, R., Theakston, A., Lieven, E., & Tomasello, M. (2004). A dense corpus study of past tense and
plural over-regularization in English. Journal of Speech, Language and Hearing Research, 47, 1319–
1333.
Mirkovi´c, J., MacDonald, M. C., & Seidenberg, M. S. (2004). Where does gender come from? Evidence
from a complex inflectional system. Language and Cognitive Processes, 20, 139–168.
Mirkovic, J., Seidenberg, M. S., & MacDonald, M. C. (2008). Acquisition and representation of grammatical
categories: Grammatical gender in a connectionist network. In B. C. Love, K. McRae, & V. M.
Sloutsky (Eds.), Proceedings of the 30th Annual Conference of the Cognitive Science Society (pp.
1954–1959). Austin: Cognitive Science Society.
Moravcsik, E. A. (1978). Agreement. In J. Greenberg (Ed.), Universals of human language (Vol. 4,
pp. 331–374). Stanford: Stanford University Press.
Rakhlin, N., Kornilov, S. A., & Grigorenko, E. L. (2014). Gender and agreement processing in children
with developmental language disorder. Journal of Child Language, 41, 241–274.
It’s all about gender: Hebrew speakers’ judgment of adjective plural agreement 343
Ravid, D. (1995). Language change in child and adult Hebrew: A psycholinguistic perspective. NewYork:
Oxford University Press.
Ravid, D. (2006a).Word-level morphology: A psycholinguistic perspective on linear formation in Hebrew
nominals. Morphology, 16, 127–148.
Ravid, D. (2006b). Semantic development in textual contexts during the school years: Noun Scale analyses.
Journal of Child Language, 33, 791–821.
Ravid, D. (2012). Spelling morphology: the psycholinguistics of Hebrew spelling. New York: Springer.
Ravid, D., & Levie, R. (2010). Adjectives in the development of text production: Lexical, morphological
and syntactic analyses. First Language, 30, 27–55.
Ravid, D., & Schiff, R. (2009). Morpho-phonological categories of noun plurals in Hebrew: A developmental
study. Linguistics, 47, 45–63.
Ravid, D., & Schiff, R. (2012). From dichotomy to divergence: Number/gender marking on Hebrew nouns
and adjectives across schoolage. Language Learning, 62, 133–169.
Ravid, D., & Shlesinger, Y. (2001). Vowel reduction in Modern Hebrew: Traces of the past and current
variation. Folia Linguistica, 35, 371–397.
Ravid, D., Dressler, W. U., Nir-Sagiv, B., Korecky-Kröll, K., Souman, A., Rehfeldt, K., Laaha, S., Bertl,
J., Basbøll, H., & Gillis, S. (2008). Core morphology in child directed speech: Crosslinguistic corpus
analyses of noun plurals. In H. Behrens (Ed.), Finding structure in data (TILAR 6) (pp. 25–60).
Amsterdam: John Benjamins.
Schiff, R., & Kahta, S. (2006). Single-word reading test: Vowelized and unvowelized word reading. Ramat-
Gan: Haddad Center, Bar-Ilan University (in Hebrew).
Schiff, R., & Ravid, D. (2012). Linguistic processing in Hebrew-speaking children from low and high SES
backgrounds. Reading & Writing, 25, 1427–1448.
Schiff, R., & Ravid, D. (2013). Morphological processing in Hebrew-speaking reading-disabled students.
Journal of Learning Disabilities, 46, 220–229.
Schiff, R., Ravid, D., & Levy-Shimon, S. (2011). Children’s command of plural and possessive marking
on Hebrew nouns: A comparison of obligatory vs. optional inflections. Journal of Child Language,
38, 433–454.
Schwarzwald, O. R. (1982). Feminine formation in Modern Hebrew. Hebrew Annual Review, 6, 153–178.
Schwarzwald, O. R. (2002). Modern Hebrew morphology. Tel Aviv: The Open University (in Hebrew).
Schwarzwald, O. R. (2005). Inflection and derivation, stems and suffixes in Hebrew nouns. HaIvrit Veaxyoteha
[Hebrew and its sisters], 5, 229–246 (in Hebrew).
Sera,M. D., Elieff, C., Forbes, J., Clark Burch,M., Rodrıguez,W., & Dubois, D. P. (2002).When language
affects cognition and when it does not: An analysis of grammatical gender and classification. Journal
of Experimental Psychology: General, 131, 377–397.
Steele, S. (1978). Word order variation: a typological study. In J. H. Greenberg, C. A. Ferguson, & E.
A. Moravcsik (Eds.), Universals of human language: IV: Syntax (pp. 585–623). Stanford: Stanford
University Press.
Tiberius, C., Brown, D., & Corbett, G. (2002). A typological database of agreement. In Proceedings of
LREC2002, the Third International Conference on Language Resources and Evaluation, Las Palmas,
Spain (pp. 1843–1846).
Tubul, M. (2003). The distribution of plural morphemes in Biblical and Mishnaic Hebrew. Unpublished
doctoral dissertation, Bar Ilan University, Israel (in Hebrew).
Vigliocco, G., & Franck, J. (2001). When sex affects syntax: Context influences in sentence production.
Journal of Memory and Language, 45, 368–390.
Wechsler, S. (2009). Agreement features. Language and Linguistics Compass, 3, 384–405.