Acquisition of polarity items in Czech children: An experimental study


39

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

LINDA DOLEŽÍ1 DOI: 10.15290/CR.2020.30.3.03
Masaryk University
ORCID: 0000-0003-4617-5490
PETRA KURFÜRSTOVÁ2
Masaryk University
ORCID: 0000-0002-2846-956X
IVETA ŠAFRATOVÁ3
Masaryk University
Georg-August-Universität Göttingen
ORCID: 0000-0001-8437-1738

Acquisition of polarity 
items in Czech children: 
An experimental study
Abstract. The Czech polarity items i and ani are traditionally treated as English even. This paper deals 
with the acquisition of these polarity items in Czech children. These focus/scalar particles are specific 
for their sensitivity to probability. We aim to find out whether Czech children at primary school ( jun-
ior school age) have already acquired i/ani and whether they are able to connect them correctly with 
alternatives on the scale of probability. The research was conducted with children from the second 
and the fourth grade at primary school. The paper represents an initial insight into this area since no 
similar research has been done in the Czech language so far. 

Keywords: language acquisition; focus particles; scalar particles; experimental linguistics; language 
teaching.

1. Introduction
The particle even has been studied for its polarity and unlikelihood properties for 
a long time in several languages. Although many issues and questions concerning 
the behaviour of even still remain, it has been consistently shown that even is sensi-
tive to the polarity of a sentence. Namely, there is even that can appear only in negative 

1 Address for correspondence: Faculty of Arts, Masaryk University, Arne Nováka 1, 602 00 Brno, Czech 
Republic. E-mail: dolezi@med.muni.cz

2 Address for correspondence: Faculty of Arts, Masaryk University, Arne Nováka 1, 602 00 Brno, Czech 
Republic. E-mail: kurfurst.petra@gmail.com

3 Address for correspondence: Faculty of Arts, Masaryk University Arne Nováka 1, 602 00 Brno, Czech 
Republic. E-mail: safratova@mail.muni.cz



40

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

sentences, and therefore it is called Negative polarity item (NPI). In contrast, another 
even so-called Positive polarity item (PPI) may occur only in positive sentences (Krifka 
1995). It is tough to detect this distinction in English since there is only one lexical item 
for even, see (1). However, there are two (or more) lexical items corresponding to even 
in many languages, e.g. German and Slovenian, among others. The Czech language 
also belongs to the group of languages where positive even and negative even are distin-
guished lexically.

(1) a. Even Charles came to the party.
 b. Even Charles didn’t come to the party.

Moreover, even is a focus sensitive particle that is associated with an F-marked 
expression. Polarity items, as well as focus sensitive particles, introduce alternatives 
(Rooth 1985). Basically, there is a set of alternatives, and the particle even picks out one 
alternative.4 The set of possible alternatives for the example (2-a) (repeated from (1-a) 
is in (2-b)). Note that the focused expression is highlighted in capital letters. The parti-
cle even chooses the one alternative in a certain way. The principles under which even 
selects an alternative are introduced in section 1.1.

(2) a. Even CHARLES came to the party.
  b. the set of alternatives:
 {Charles}
 {Jane}
 {Richard}

We have dealt with polarity and focus properties of even. In the next section we 
focus on even in Czech, we describe two main “evens” and show how they differ from 
each other. In section 3 we introduce two experiments on child acquisition of English 
even and in section 4 we present the experiment on child acquisition of Czech even 
and discuss the results.

2. Theoretical background
English even corresponds to (at least) two lexical expressions in Czech, namely i and ani.5 
These two items are nicely divided according to their polarity properties. The first Czech 

4 Note that there is a requirement for the alternatives to be of the same semantic type.

5 In addition, the particles i and ani may function as conjunctions and also marginally interjections. 
We leave these two other usages of the expressions aside and deal with i and ani as particles only.



41

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

even is i, and it is considered to be PPI since i may occur only in affirmative sentences, 
and it is ungrammatical in negative sentences, see (3-a) and (3-b), respectively.

(3) a. Na večírek přišel   i Richard.
   to party come.3SG.PST  even Richard
   ‘Even Richard came to the party.’
 b. *Na   večírek  nepřišel  i Richard.
   to party  NEG.come.3SG.PST even Richard.
   ‘Even Richard didn’t come to the party.’

The second Czech even is ani. Historically, i is claimed to be the basic, whereas ani 
is the modified version of i in such a way that ani contains the basic i and the negative 
element -n. (Lamprecht et al. 1986). This pattern nicely explains that ani is a negative 
counterpart of positive i. Since ani is NPI, it behaves in exactly the opposite way than i; 
it can appear only in negative sentences and it is ungrammatical in positive sentences, 
see (4-a) and (4-b), respectively.

(4) a. Na večírek  nepřišel  ani Richard.
   to party  NEG.come.3SG.PST even Richard
   ‘Even Richard didn’t come to the party.’
 b. *Na večírek  přišel   ani Richard.
         to party  come.3SG.PST  even Richard.
       ‘Even Richard came to the party.’

Both i and ani exhibit the same focus properties as English even. Unlike in English, 
Czech i/ani have to occur immediately before the F-marked expression in a sentence. Since 
the occurrence of i/ani is sensitive to the polarity of sentences, they belong to the group 
called polarity items. Therefore, they introduce alternatives just like English even.

We do not go into details of the complicated theory of alternatives (for more details 
see Rooth 1985) but some background is needed for understanding our experiment. 
Concerning alternatives introduced by even the likelihood and entailment play a role. 
For simplification, only a closed set of alternatives are taken into account. Since even 
is generally considered to work with the likelihood, the alternatives are ordered 
on the probability scale, i.e. it is given by the context that one alternative is more likely 
or less likely than the other. Consequently, one alternative entails the other.6 Consider 
the following example:

6 Note that there is no entailment relationship between alternatives in some alternative sets, but still, 
the alternatives are ordered by probability because there is a likelihood relationship between them. 



42

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

(5)    Richard        přečetl           i        pět knih.
         Richard        read.3SG.PST   even five books
         ‘Richard has read even five books.’

We can easily imagine that Richard is a student and he has to read books for an exam. 
The alternative set is restricted by the context where the maximum number of books he 
is supposed to read is five. Of course, many students do not read all the required books. 
The use of i in (5) indicates that Richard has read the maximum number of the required 
books, and this is considered to be a significant achievement. The alternatives and their 
entailment and likelihood relationship are schematically captured in (6).

(6) a. the alternative set: {1 book, 2 books, 3 books, 4 books, 5 books}
 b.  the entailment relationship: read 5 books → read 4 books → read 3 books → 
   read 2 books → read 1 book 
 c.  the likelihood relationship: read 5 books < c read 4 books < c read 3 books < 
   c read 2 books <c read 1 book 

There is the closed set of alternatives in (6-a), and the alternatives are in entailment 
relationship in such a way that the proposition Richard has read 5 books entails the prop-
osition Richard has read 4 books and so on but not vice versa. Consequently, the alter-
native 5 books is the strongest alternative from the given set of alternatives since read 
5 books entails all the other alternatives. At the same time, the alternative 5 books is 
the least likely alternative because reading 5 books is the most unlikely situation (again 
from the given alternative set).

We conclude that i always associates with the least likely alternative and, simulta-
neously, the strongest alternative from the alternative set given by the context. Let’s 
focus on ani and check whether it associates with an alternative of the same type. Now, 
consider the example with ani.

(7)    Richard        nepřečetl               ani    jednu   knihu.
         Richard        NEG.read.3SG.PST even one   book
         ‘Richard hasn’t read even one book.’

We imagine the same context, but Richard is a lazy student now. The alternative set 
remains the same. However, we observe that ani associates with a different alternative 

Since we use only alternative sets where there were both entailment and likelihood relationships be-
tween alternatives, we leave the alternatives without entailment relationship aside.



43

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

than i. Let’s first analyze the entailment and likelihood relationships between the alter-
natives before we make a conclusion.

(8) a. the alternative set: {1 book, 2 books, 3 books, 4 books, 5 books}
 b. the entailment relationship: not read 1 book → not read 2 books → 
   not read 3 books → not read 4 books → not read 5 books
 c. the likelihood relationship: not read 1 book < c not read 2 books <
   c not read 3 books <c not read 4 books <c not read 5 books

The proposition Richard hasn’t read 1 book entails the proposition Richard hasn’t read 
2 books and so on, but not vice versa.7 And, at the same time, the alternative not read 
1 book is the least likely alternative because read 1 book is the minimum you can do. 
Imagine that a student wants to pass an exam. Then, it is more likely (and highly recom-
mended) to read 2 books than only 1 book. Consequently, the alternative not read 1 book 
is the strongest alternative since it entails all other alternatives and it is also the least 
likely alternative from the contextually given alternative set.

The fact that ani associates with the strongest and the least likely alternative as well 
as i is caused by the presence of negation in sentences containing ani. Negation inter-
venes between the F-marked expression and even and works as a scale reversing opera-
tor. In other words, negation reverses the scale of entailment and likelihood (for more 
technical details and formalizations see, e.g., Karttunen & Karttunen 1977 and Crnič 
2011). Therefore, both i and ani associate with alternatives of the same type concerning 
entailment and likelihood, but the alternatives differ lexically.

The unified type of alternatives concerning entailment and likelihood for both i and 
ani is well theoretically justified. Moreover, Šafratová (2018) focused on adults and their 
perception of structures with Czech even and showed that Czech adults treat i/ani precise-
ly as would be expected. Since even works with probability, which is part of pragmatics, 
the question that comes to mind is how and when adults learn how to use particles such 
as even. To solve this question we turned our attention to children and experimentally 
tested whether Czech children at a certain age have already acquired the particles i/ani. 
Note that the acquisition of i/ani is not affected by Czech language education because 
according to two main education programs for primary education in Czech (Rámcový 
vzdělávací program pro základní vzdělávání, Školní vzdělávací program), no explic-
it attention is paid to the particles in primary education. However, there is no study 

7 We are aware of the fact that this is not true since the proposition Richard hasn’t read 1 book may, 
of course, mean that Richard has read 2 books. However, this is not relevant for explaining how the entail-
ment and likelihood work for the particle even. For more details, we refer readers to the theory of scalar 
alternatives (Sauerland 2004; Fox & Hackl 2006; Spector 2007, among others).



44

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

about the acquisition of polarity items in Czech children we could build our research 
on. That is why we follow two relevant experimental research studies on child acquisi-
tion of English even that are introduced in the next section. Our experiment is presented 
in section 4.

3. Child acquisition of English even 
As far as we know there are only two experimental research studies on child acquisition 
of English even, and interestingly, the two research studies came to different results. 
Both experiments tested English even in both affirmative and negative sentences. To 
avoid confusion we call even in positive sentences simply even, and even in negative 
sentences not even.

3.1. Kim’s experiment
The first experiment was run by Kim (2011). She experimentally tested English even 
in two syntactic positions, namely the pre-subject position (9-a) and the pre-object posi-
tion (9-b). We focus only on the results of even in the pre-object position since we tested 
Czech even in this syntactic position. 

(9) a. Even Petr ate chocolate.
 b. Petr ate even chocolate.

Kim tested 30 English speaking children aged 4-5 using a “guess who game”. There 
were three pictures with even, and three pictures with not even, i.e. six tested pictures 
in total. Besides, there were six filler sentences and six control sentences to distract 
the children. The children were asked to listen to stories with pictures in such a way that 
the last sentence of a story contained even/not even. There were always three characters 
differing in size (the smallest one, the biggest one and a character of the middle size), 
and the task was to show the appropriate character in the picture based on the story. 
For instance, there was a picture of three different-size bears trying to reach a biscuit. 
Based on the pragmatic feature (height), children should have been able to recog-
nize that the biggest bear has the greatest chance to reach the biscuit since the tested 
sentence contained not even.

Kim concludes from the experimental results that English children aged 4-5 have not 
acquired English even/not even yet. Moreover, the middle characters (options) did not 
appear as an answer at all. The number of children who answered correctly to each 
item and the total percentage of correct answers is in Table 1 (taken from Kim 2011: 96).



45

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

Table 1. The correct answers

 Total

Affirmative even
41/90
45.6 %

Negative even (not even)
42/90
46.7 %

The question that comes to mind is why do English children have a problem 
with the usage of even? There are (at least) three possible answers or their combina-
tion: (i) The problem is lexical. Children have to learn the semantic meaning of even. 
(ii) The problem is syntactic. English even can stay quite freely in a sentence, unlike 
in Czech. (iii) The problem is pragmatic. Even has an unlikelihood presupposition 
and children have to count with the likelihood. Before we try to answer the question 
let’s have a look at the second experiment on English even and compare the results from 
both experiments.

3.2. MIT experiment
Following Kim (2011), Newman et al. (2018) asked the following research questions: (i) 
How do children think when considering sentences with even? (ii) Why did none of the 
children choose middle options? (iii) What happens when we change scalar types when 
choosing different options?

They tested English speaking children aged 3−6. There were 88 children participat-
ing in the experiment in total. The experiment included four different scales: reach-
ing stories, lifting/wight stories, fitting stories, and filling/capacity stories. The scales 
are presented in more detail in section 4. The authors also used a “guess who game” 
by telling stories and pictures. There were four pictures with affirmative even and four 
pictures with negative even (not even), therefore eight pictures in total. Moreover, 
the experiment contained four sample filler stories.

The experiment shows the polarity effect between even and not even. Children were 
more successful in tested sentences with not even, see Figure 1 taken from Newman 
et al. (2018). 



46

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

Figure 1. A plot of the rate of adult-like responses by age group, separated by polarity

Surprisingly, middle responses appeared in the results of the experiment, unlike 
the results of Kim’s experiment. The results of MIT’s experiment show that children 
answered not only the left or right option but also the middle option. Middle responses 
also show the polarity effect, but the opposite, i.e. middle responses, appeared more 
in tested sentences with even, see Figure 2, taken from Newman et al. (2018).

Figure 2. Middle responses decrease with age, and they decrease 
faster for negative even than positive even



47

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

Overall experimental results show that:
1. children were more successful in tested sentences with not even;
2. middle responses appeared more in tested sentences with even;
3. children consider the likelihood, i.e. the unlikelihood presupposition,  

for even/not even;
4. children’s usage of not even is close to the adults’ usage of it;
5. children acquire not even at around 4 years of age, whereas they acquire  

even at around 6 years of age.
It is evident that Newman et al. (2018) gained different results than Kim (2011). 

However, we follow both experiments in our research, especially concerning the design 
of the experiment. We were aware of the fact that the experimental results of Czech i/
ani could differ from the results of English experiments because i and ani are lexically 
distinguished as opposed to only one lexical item even in English. To the best of our 
knowledge there is no research on children acquisition of Czech scalar particles i/ani, 
and therefore the experiment was designed as a “mapping territory” project.

4. Experiment
The main focus of our experiment was to find out whether children aged 7−10 have 
already acquired the i/ani particles in their scalar function. 20 children from grade 2, 
and 20 children from grade 4 took part. As stated above, when designing the experi-
ment, we followed similarly oriented research, in particular the study by Kim (2011) 
and with the kind permission of the authors also the experiment developed by MIT 
linguists Newman et al. (2018). 

Four hypotheses were stated:
• children in grade 2 will not have adult-like or similar understanding of i/ani 

expressions;
• children in grade 4 will have acquired i/ani expressions fully;
• there will be middle variants of the responses, especially in grade 2 children;
• there will be a difference in the acquisition of i and ani expressions. 

4.1. Procedure and participants
We ran the experiment in two phases. In phase 1 we tested 20 adults to see whether our 
tasks are well designed and comprehensible and also to check whether the language 
phenomena we studied were fully acquired by adult speakers. In phase 2 we carried 
the experiment out with children (20 children age 7−10 from grade 2 and 20 children 
age 9-10 from grade 4). Children were chosen randomly by their teachers, parents 
had to provide their consent. The testing ran according to the same scenario as with 
the adults. Before the beginning they were informed about the form of the experiment – 
the pictures of animals in three different sizes and the task – to choose to which animal 



48

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

the text relates to. The participants were not aware of the fact that they are taking part 
in research on scalar particles i/ani, neither were shown any training pictures so that 
they were not biased. The accompanying text was read aloud, however, children could 
also read it for themselves. Their answers were noted down into answer sheets without 
stating explicitly to the participants whether their answers were correct or not so that 
their answers were not influenced by this information. One experiment with one person 
took approximately 10 minutes.

4.2. Design and material
We adapted thematic story areas created by Newman et al. (2018) and used the same 
number of items. Furthermore, we profited from their experience with testing and thus 
avoided possible problems. We created our own unique illustrations (Bukovjan 2019) 
and texts including the tested expressions i/ani. The experiment consisted of two parts 

– the first part were stories testing i/ani, the second part were filler stories that were 
meant to distract the participants’ attention. 

The individual pictures depicted animals of three different sizes in the given contex-
tual situation (in the given thematic story) in the first experiment part. The context 
provided a pragmatics feature ensuring that only one animal should be chosen assum-
ing that children interpret the sentences with i/ani pragmatically correctly.

The thematic story areas were as follows: 
• Reaching stories: the motive is the effort to reach for something/somewhere;
• Lifting/weight stories: the motive is to try to pick up something;
• Fitting stories: the motive is to fit somewhere;
• Filling/capacity stories: the motive is an effort to fill something (e.g. a basket).
Every story area contained one picture testing the i-expression and one picture testing 

the  ani-expression, there were 8 tested pictures together. An example of the illustra-
tions and texts can be found below in Picture 1 and Picture 2.

Picture 1: V prodejně rybiček zkoušeli, jak velká rybka se vejde do skleněného stolního 
akvária. Bylo tak prostorné, že se do něj vešla i Dona. Poznáš, která rybka je Dona?



49

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

‘In a fish shop they tried how big a fish could fit into a glass table aquarium. It was so 
spacious that even Dona could fit into it. Do you know which fish is Dona?’

Picture 2: Opice se vsadily o trs banánů, že ze země zvednou velikou těžkou kládu. Kláda 
byla ale tak moc těžká, že ji neuzvedl ani opičák Rocky. Poznáš, která z opic je Rocky?

‘Monkeys bet a bunch of bananas that they will be able to lift a large, heavy log from 
the ground. But the log was so heavy that even Rocky didn’t lift the log. Do you know 
which monkey is Rocky?’

In both examples of the test pictures, the accompanying text referred to the biggest 
animal (i.e. the biggest fish, the biggest monkey) because it is clear from the context 
that it is the least probable alternative (and with these the scalar particles bind). 
Generally speaking, the correct answers, however, were not only the biggest animals 
but also the smallest or the middle ones in the case of fillers. This is how we prevented 
the participants from guessing or seeing through the experiment. 

In addition to tested pictures, we used four sample filler stories in the second part 
of the experiment. To prevent the children from observing a pattern in which the exper-
iment works and find their own algorithm according to which they would answer, we 
ordered the pictures randomly so that two pictures of the same story category did not 
appear next to each other. All together there were 12 pictures accompanied by texts, i.e. 
8 tested pictures and 4 filler pictures.

4.3. Results 
The whole experiment was evaluated by several means ranging from the most ordi-
nary descriptive statistics to proper statistical analyses using programming language R 
in RStudio. The responses were modeled in lmerTest, which is not a part of RStudio but it 
is a standard model and it was added as a package into the RStudio. By using such a wide 
range of statistical tools we tried to get as many pieces of information as possible. As 
to the results of the research carried out on adults – they were successful in completing 



50

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

the task in almost 100%. Only in one case did an adult make a mistake probably due 
to lack of concentration rather than linguistic incompetence.

In children the results were as follows:
1. in the case of 100% success in all items testing i/ani as well as the fillers, the results 

were:
• grade 2: only 25% of children were 100% successful (5/20)
• grade 4: 55% of children were 100% successful (11/20)

2. the results of 100% success in tested i/ani items (all together):
• grade 2: 30% of children were 100% successful
• grade 4: 60% of children were 100% successful.

From this perspective it seems that children in grade 4 were two times more success-
ful than children in grade 2. We also had a look at the number of mistakes made by chil-
dren. Some of them made only one mistake which might have been caused by lack 
of attention or hesitation. In case we recalculated the results and included also the “1 
mistake children” the success rate would increase a lot (especially in grade 2): 

• grade 2: 60% of children with almost 100% success
• grade 4: 75% of children with almost 100% success

3. Success rate was further analysed for every even-variant separately − i-variant 
and ani-variant. The results showed a higher success rate in items testing under-
standing i-variant in children from both grades. Both variants − i/ani − were easier 
for children from grade 4, which might mean that they understand these expres-
sions better. Nevertheless, the i-expression success rate was 65% in grade 2, which 
already proves a very good understanding. On the other side, understanding ani-
expression was in grade 2 only 40%, which could mean that this expression is 
more difficult to acquire and is thus acquired later.

a) Results of items with i:
100% successful: a child correctly understood all items with i

• grade 2: 65% of children were 100% successful (13/20)
• grade 4: 80% of children were 100% successful (16/20)

b) Results of items with ani:
100% successful: a child correctly understood all items with ani

• grade 2: 40% of children were 100% successful (8/20)
• grade 4: 65% of children were 100% successful (13/20).

4) We also compared the results according to their success rate in individual story areas. 
Every picture was evaluated as to the success rate separately for i- and ani-expressions. This 
comparison showed a high success rate for both expressions and children in both grades 
seem to understand these two expressions well. Even though the partial results showed 
a difference between the two classes as well as the two expressions, the children responded 
correctly in about 80%.



51

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

Table 2. The success of children in each story in sentences with i. Story 1 = reaching stories, 
story 2 = lifting/weight stories, story 3 = fitting stories and story 4 = filling/capacity stories.

story 1 story 2 story 3 story 4 in total %

grade 2 17/20 17/20 16/20 18/20 68/80 85 %

grade 4 19/20 17/20 18/20 20/20 74/80 93 %

Table 3. The success of children in each story in sentences with ani.  
Story 1 = reaching stories, story 2 = lifting/weight stories, story 3 = fitting stories  

and story 4 = filling/capacity stories8

story 1 story 2 story 3 story 4 in total %

grade 2 12/20 15/20 20/20 16/20 63/80 79 %

grade 4 15/20 18/20 20/20 16/20 69/80 86 %

The preliminary summary is the following: 
1) Children make fewer mistakes in sentences with i.
2) It is easier for children to understand sentences with i. Several possible explanations 

may affect the understanding of i. First, since ani interacts with negation, it might make ani 
more complex to understand. Second, the interaction with negation causes the reverse scale 
of likelihood and therefore both i and ani associate with the least likely alternative. Third, 
the correct understanding of i and ani depends on pragmatic abilities of children. 

3) Children in both grade 2 and grade 4 have acquired both expressions i and ani.
For getting a more detailed analysis we processed the obtained data in programming 

language RStudio. Answers were modified as follows:
• 1 – the expected responses (the least likely alternative)
• 0.5 – the middle option
• 0 – the opposite end of the scale (the most likely alternative).

To model the data we constructed a mixed linear model that tested whether 
the subjects’ answers can be predicted from a condition (fixed effect) and whether 
the conditions differed in their statistical significance. There were following conditions:

1) 2-i: sentences with i in grade 2
2) 2-ani: sentences with ani in grade 2

8 The attentive reader may notice that there is a relatively large difference between the success in story 1 
and story 3. Looking at success in the stories, we see a big drop between story 1 and other stories. There 
is not much difference in success between other stories. Since only story 1 seems to be problematic for 
children, we blame this story 1 for failure. Even though story 1 was based on the same principle as the 
other stories, children may have seen a difference between story 1 and others or children just did not like 
the story. Thanks to an anonymous reviewer for raising this point. 



52

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

3) 4-i: sentences with i in grade 4
4) 4-ani: sentences with ani in grade 4.
The success rate of responses depending on the conditions is between 0−1. The statis-

tical differences between sentences with i and sentences with ani were modeled 
in lmerTest. The model had one predictor, i.e., the reference level condition: 2-ani. 
The success of the other conditions was measured against the reference level condition. 
The output of the model is reported in Table 4. The t-value states how big the difference 
between two conditions is, the p-value says how likely it is that the difference between 
the two conditions is random. As we can see from the results, no fixed effects were 
statistically significant. 

Table 4. The statistical output

Fixed effects:

 Estimate Std. Error df t value p value

(Intercept) 0.80000 0.05601 10.35250 14.283 3.81e-08 ***

Condition 2-i 0.06875 0.07754 9.60427 0.887 0.397

Condition 
4-ani

0.06875 0.05028 291.00007 1.367 0.173

Condition 4-i 0.13125 0.07754 9.60427 1.693 0.123

---
Signif. Codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Box plots of the individual conditions are shown in Figure 3. Means and medians 
of each condition are in Table 5.



53

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

Figure 3. Detailed graph: means of conditions

Table 5. Means and medians of conditions

Condition Means Medians

2-ani 0.80000 1

2-i 0.86875 1

4-ani 0.86875 1

4-i 0.93125 1

The statistical analysis shows (i) that the mean success rate of tested sentences 
with i and ani was very high, (i) that there is no statistical difference between grade 2 
and grade 4 in sentences with i, (iii) that there is no statistical difference between grade 
2 and grade 4 in sentences with ani, and last but not least (iv) that children aged 7-10 
have already acquired both expressions i and ani and they understand the scale of like-
lihood (the pragmatic aspect).

5. Summary and conclusion
The experimental results show a tendency of children aged 7-10 to have already acquired 
i/ani, which was proved by statistical analysis. Our initial four hypotheses were not 
confirmed. The experimental results show that children in grade 2 understand scalar 
expressions. Both i and ani are acquired approximately at the same time, although 



54

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

children made fewer mistakes in sentences with i. The middle options were marginal, 
which corresponds to Kim’s (2011) finding. Even though there is only one lexical item 
even in English as opposed to Czech i/ani, children have to learn the pragmatic aspect, 
i.e., the likelihood, in both languages. Moreover, children have to learn how the likeli-
hood changes with respect to positive/negative sentences. 

Since the particles are generally marginally included in the methodological plan for 
language education at primary schools, we assume that the acquisition of expressions 
i/ani depends on the individual development of the child’s cognitive and pragmatic 
competences independently of language education. Getting to know how language is 
acquired is crucial for many areas connected to child development. In the current Czech 
linguistic environment and renewed interest in child language the main attention has 
been paid to pre-school age language acquisition, e.g. Smolík & Bláhová (2017); Smolík 
& Seidlová Málková (2014); Saicová Římalová (2013); Doleží (2014); Mertins et al. (2014). It 
might seem that once children enter school there are not many things going on linguis-
tically speaking. Opposite is the case, though. With our study we tried to contribute 
to the understanding of how later phases of acquisition work and support it by concrete 
results and analyses and not impressions. The results might have interesting implica-
tions for mother tongue as well as second or foreign language teaching in understand-
ing whether, when and how to teach or confront children with particular linguistic 
phenomena so that they can master them fully. To see when and how exactly the i- 
and ani-expressions are acquired by Czech children we have to pay attention to younger 
children and carry out further research with first grade and pre-school age children. 

References
Crnič, L. 2011. Getting Even [Unpublished PhD dissertation]. Massachusetts Institute 

of Technology. 
Doleží, L. 2014. Reduction and elaboration in Czech coordinate structures: An elicit-

ed imitation task. In: L. Veselovská & M. Janebová (eds.), Complex Visibles Out There. 
Proceedings of the Olomouc Linguistics Colloquium 2014: Language Use and Linguistic 
Structure, 387-400. Olomouc: Univerzita Palackého.

Fox, D. & Hackl, M. 2006. The universal density of measurement. Linguistics 
and Philosophy 29(5): 537-586.

Karttunen, L. & Karttunen, F. 1977. Even questions. In: NELS 7: Proceedings of the Seventh 
Annual Meeting of the North Eastern Linguistic Society, 115-34. Cambridge: MIT.

Kim, S. 2011. Focus Particles at Syntactic, Semantic and Pragmatic Interfaces: 
The Acquisition of only and even in English [Unpublished PhD dissertation]. University 
of Hawaii at Manoa, Manoa. 

Krifka, M. 1995. The semantics and pragmatics of polarity items. Linguistic Analysis 
25 (3-4): 209-57.



55

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

Lamprecht, A., Šlosar, D. & Bauer, J. 1986. Historická mluvnice češtiny. Praha: Státní peda-
gogické nakladatelství. 

Mertins, B. et al. 2014. Identifikace konstituentů kompozit u předškolních děti.́ Naše řeč 
97(3): 146-165.

Newman, E., Gowda, Y., Rosenstein, L. & Hackl, M. 2018. Polarity sensitivity of even 
in early child grammar. GALANA 8, Indiana University, Bloomington. Poster.

Rámcový vzdělávací program pro základní vzdělávání [online]. Praha: MŠMT, 3-2017. 
Available from: http://www.nuv.cz/t/rvp-pro-zakladni-vzdelavani.

Rooth, M. 1985. Association with Focus [Unpublished PhD dissertation]. University 
of Massachusetts. 

Saicová Římalová, L. 2013. Když začínáme mluvit... Lingvistický pohled na rané projevy 
česky hovořícího dítěte. Praha: Práce Filozofické fakulty Univerzity Karlovy, Varia.

Sauerland, U. 2004. Scalar implicatures in complex sentences. Linguistics and Philosophy 
27(3): 367-391.

Smolík, F. & Bláhová, V. 2017. Comprehension of verb number morphemes in Czech 
children: Singular and plural show different relations to age and vocabulary. First 
Language 37(1): 42-57.

Smolík, F. & Seidlová Málková, G. 2014. Vývoj jazykových schopností v předškolním věku. 
Praha: Grada.

Spector, B. 2007. Scalar implicatures: Exhaustivity and Gricean reasoning. Questions 
in Dynamic Semantics 17: 225-249.

Šafratová, I. 2018. Negation, NPI and Scales of Likelihood: An Experimental Study. 
[Unpublished Master’s thesis]. Masaryk University.

Školní vzdělávací program pro základní vzdělání ZŠ a MŠ Rovečné. 2015. Rovečné.

* * *
Linda Doleží is a language teacher and teacher trainer. She works as an Assistant 
Professor at Masaryk University Language Centre and at the Department of Czech 
Language of Faculty of Arts at Masaryk University. She focuses on teaching method-
ology, language acquisition and psycho- and neurolinguistics in multilingual contexts. 
Her current interests include language attrition phenomena and the role of emotions 
and trauma in language acquisition and attrition.

Petra Kurfürstová is a graduate of upper secondary school teacher training in Czech 
language and literature at the Faculty of Arts at Masaryk University. In her M. A. thesis 
she dealt with the acquisition of scalar particles i and ani in Czech children. During 
her studies she was interested not only in child language acquisition and in teaching 
methodology but also in teaching Czech as a second language. 



56

.........................................................................................CROSSROADS. A Journal of English Studies 30 (2020) (CC BY-NC-SA 4.0) 

Iveta Šafratová is a second-year doctoral student in a joint supervision of doctoral 
studies at the Department of Linguistics at Masaryk University and at the Seminar for 
English Philology at Georg-August-Universität Göttingen. She focuses on the syntax-
semantic-pragmatic interface and she is also interested in experimental linguistics. 
She is working on negation, especially scalar particles and polarity items in Czech.