“Liz can buy a croissant or donut. That’s both together, right?”
Distinguishing target Free Choice from non-target Modal Conjunction in Child French

Antoine Cochard, Angeliek van Hout & Hamida Demirdache∗

Abstract. Several acquisition studies have reported that children draw free choice
inferences at adult-like rates from modal disjunctive statements. This study explores
an alternative explanation for children’s seemingly adult-like behavior: modal con-
junction, which shares verifying and falsifying conditions with free choice. However,
existing experimental setups were not able to distinguish the two. With our novel de-
sign, we were able to set apart modal conjunctive interpreters from genuine free choice
interpreters, using a new type of condition: a mutually exclusive context. The results
revealed that free choice inferences are not so early acquired as previously thought. In
contrast to the earlier studies, only half of the children between 4 and 6 were genuine
adult-like free choice interpreters. The other children either show the basic inclusive
interpretation of disjunction, or, as hypothesized, a modal conjunctive interpretation.
Keywords. Acquisition; Disjunction; Free Choice; Modal conjunction; Mutually
exclusive context

1. Introduction. Plain disjunctive statements as in (1) typically yield an exclusivity inference (1a)
generated by strengthening of the basic logical inclusive meaning of disjunction in (1b). Thus, (1)
is commonly taken to mean that exactly one of the two disjuncts is true, that is Liz either bought
the croissant or the donut. It has been reported in several acquisition studies that children perform
non-adult-like with such statements in that they do not generate the exclusivity inference (1a) and
allow the logical inclusive meaning of disjunction (1b) (Chierchia et al. 2001, 2004; a.o.), or/and
derive an incorrect conjunctive inference (1c) (Braine & Rumain 1981, Cochard et al. 2023, Paris
1973, Singh et al. 2016, Tieu et al. 2017).

(1) Liz bought a croissant or a donut. ‘C or D’
a. Exclusivity inference:

Liz bought a croissant or a donut, but not both. (𝐶 ∨ 𝐷) ∧ ¬(𝐶 ∧ 𝐷)
b. Logical inclusive meaning:

Liz bought a croissant or a donut, possibly both. (𝐶 ∨ 𝐷)
c. Conjunctive inference:

Liz bought a croissant and a donut. (𝐶 ∧ 𝐷)
When disjunction is embedded under a possibility modal, henceforth modal disjunctive state-

ments, as in (2), it is inferred that both disjuncts must each be individually permitted. This type
of inference has been referred to as a free choice inference (Kamp 1973). So, (2) is taken to mean
that there are two pastries that Liz can buy, namely a croissant and a donut (2a). On top of the
free choice inference, an additional, although optional (Simons 2005), exclusivity inference can be

∗For helpful discussion and feedback, we would like to thank reviewers & audiences at ELM3, the LLING
and the CLCG Acquisition Lab. Authors: Antoine Cochard, University of Groningen & Nantes University (an-
toinecochard@outlook.com); Angeliek van Hout, University of Groningen (a.m.h.van.hout@rug.nl); Hamida Demir-
dache, Nantes University (hamida.demirdache@univ-nantes.fr).

Proceedings of ELM 3: 104-116, 2025

c©2025 Antoine Cochard, Angeliek van Hout, and Hamida Demirdache
Published by the LSA with permission of the author(s) under a CC BY license. 104

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


derived according to which it is not possible for Liz to buy the croissant and the donut, (2b).

(2) Liz can buy a croissant or a donut. ‘can C or D’
a. Free choice inferences:

i. { Liz can buy a croissant. ^𝐶 } ^𝐶 ∧ ^𝐷ii. { Liz can buy a donut. ^𝐷
b. Exclusivity inference:

{ Liz can buy a croissant or a donut, but not both together. ^(𝐶 ∨ 𝐷) ∧ ¬^(𝐶 ∧ 𝐷)
In contrast to children’s non-adult-like interpretation of plain disjunctive statements, recent

studies focusing on modal disjunctive statements suggest that children correctly derive the free
choice inference at adult-like rates (Huang & Crain 2020, Tieu et al. 2016, Zhou et al. 2013).
2. Disentangling Free Choice from Modal Conjunction. Experiments that investigated the ac-
quisition of free choice inferences did so by falsifying one of the two disjuncts. In this section, we
show that this is not enough to confidently assert that children know how to derive such inferences.
To be more precise, we investigated the idea that children’s allegedly adult-like responses might
also arise in a non-adult-like grammar where they posit modal conjunction, as paraphrased in (3).

(3) Liz can buy a croissant and a donut. ^(𝐶 ∧ 𝐷)
As a starting point, take (4) adapting the experiment setup from Huang & Crain (2020). In

the control condition (4a), the character is permitted to do both actions: the target response is
to accept the test sentence (4) since the free choice inference is satisfied. To the contrary, in the
critical condition (4b), the character is only permitted to perform one of the two actions: the target
response is to reject the test sentence, since the free choice inferences are no longer both satisfied.

(4) Liz can buy a croissant or a donut.
a. Control condition:

Liz has the possibility to buy a croissant. ^𝐶 } ^𝐶 ∧ ^𝐷She also has the possibility to buy a donut. ^𝐷
b. Critical condition:

Liz has the possibility to buy a croissant. ^𝐶 } ^𝐶 ∧ ¬^𝐷But she does not have the possibility to buy a donut. ¬^𝐷
However, the conclusion that accepting the test sentence in (4a) and rejecting it in (4b)

necessarily reflects a free choice inference is too fast. Indeed, this answer pattern may also arise
for another reason than correctly deriving a free choice inference: if the test sentence is incorrectly
interpreted as a modal conjunctive statement, on a par with (3), then the same response pattern is
expected, as illustrated in (5) for the control condition and (6) for the critical condition.

(5) Expected responses for the control condition (4a):
True if understood as ^𝐶 ∧ ^𝐷 Free Choice
True if understood as ^(𝐶 ∧ 𝐷) Modal Conjunction

(6) Expected responses for the critical condition (4b):
False if understood as ^𝐶 ∧ ^𝐷 Free Choice
False if understood as ^(𝐶 ∧ 𝐷) Modal Conjunction

Proceedings of ELM 3: 104-116, 2025

Antoine Cochard, Angeliek van Hout, and Hamida Demirdache:
Distinguishing target Free Choice from non-target Modal Conjunction in Child French. 105

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


Nevertheless, a modal conjunctive interpretation is not equivalent to a free choice interpretation.
This is because modal conjunction entails free choice, but not conversely. Under the experimental
setup presented in (4), they do indeed both have the same satisfying and falsifying conditions, but
this is not always the case. A different experiment setup – namely, mutually exclusive scenarios
such as (7) – allows us to tease apart their different truth conditions.

(7) Mutually Exclusive scenario:
a. i. Liz has the possibility to buy a croissant. ^𝐶 } ^𝐶 ∧ ^𝐷ii. She also has the possibility to buy a donut. ^𝐷
b. Liz does not have the possibility to buy both pastries. ¬^(𝐶 ∧ 𝐷)

In this situation, the character has the possibility of buying either the croissant or the donut:
he is free to buy whichever he chooses, as stated by (7a). But, critically, the character does not have
the possibility to buy the croissant and the donut, as stated in (7b). In this scenario, free choice is
satisfied (8a), but modal conjunction is not (8b).

(8) Liz can buy a croissant or a donut. Mutually exclusive scenario
a. True if understood as ^𝐶 ∧ ^𝐷 Free Choice
b. False if understood as ^(𝐶 ∧ 𝐷) Modal Conjunction

Given that some children tend to assign a conjunctive interpretation for simple disjunctive
statements, it is plausible that they may also assign a modal conjunctive interpretation to modal
disjunctive statements. Importantly, previous studies investigating the acquisition of free choice
inferences did not include any sort of mutually exclusive condition. Thus, it is premature to
conclude that 4 to 6-year-olds actually derive genuine free choice inferences at adult-like rates.
Instead, some of the allegedly adult-like children might have given an adult-like response that
arises for a non-adult-like reason. Addressing this issue, we developed a new experimental design
distinguishing free choice inferences from modal conjunctive ones.
3. A novel setup to investigate free choice inferences. This section presents a novel design
taking inspiration from Liu (2017) to investigate the derivation of free choice inferences. This
design introduces two new conditions not used before in acquisition studies to our knowledge: a
Mutually Exclusive condition adapted that allows to detect modal conjunctive interpreters and a
Package condition that investigates whether the exclusivity inference is computed.
3.1. Participants. 68 native French children participated in the experiment. To be included in
the study, participants had to be native French speakers and not have known cognitive, visual or
hearing impairments. In addition, 40 native French adults were recruited on Prolific1.
3.2. Procedure. A truth-value judgment task (Crain & Thornton 1998) was used for this study.
Participants had to judge a sentence describing the picture of a shop-selling situation as in Figure
1. Each participant was told that the shop employee (the character at the bottom of Figure 1)
tries to help customers (e.g. the character at the top) by telling them what they can buy with the
number of coins in their purse. Participants had to decide whether or not the shop employee gave

1The present experiment received approval from the CEDIS ethics committee of Nantes Université (IORG0011023)
and was assigned the identification number 06112023-1.

Proceedings of ELM 3: 104-116, 2025

Antoine Cochard, Angeliek van Hout, and Hamida Demirdache:
Distinguishing target Free Choice from non-target Modal Conjunction in Child French. 106

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


Figure 1: Sample item for Liz can buy a croissant or a donut.

a successful description of the picture by pressing a thumbs-up button (correct) or a thumbs-down
button (incorrect). The children were tested individually in a quiet space at their school using a
tablet. The adults participated in a web-based version of the experiment.
3.3. Stimuli. Each picture contains several key elements. The bottom character, the koala, is a
shop employee and appears from one item to the next; he utters the test sentence. Participants were
told that the shop employee is a newcomer and, as such, he can make mistakes since it is his first
day at work. The top character is a customer, who varies across pictures. Each customer comes to
the shop with a certain number of coins in their purse. For instance, Figure 1 depicts a condition
where the bear arrives in the shop with one coin in her purse. Finally, the shop stand shows two
spaces, each of them containing a food item and under it the price. Here, the croissant costs one
coin and the donut costs two coins. After hearing the test sentence, the thumb-up and down buttons
appear on the right side of the screen. The position of those buttons was counterbalanced between
participants.
3.4. Design. The experiment included the three critical conditions in Figure 2 and the two control
conditions in Figure 3. The conditions varied the price of the food and the number of coins in the
purse. It was a within-subject design and participants saw six items of each condition (18 critical
items and 12 control items, 30 items in total). This was needed to determine the consistency in
individual answer patterns comparing a given participant’s interpretation across the three critical
conditions.

(a) Free Choice False (b) Mutually Exclusive (c) Package

Figure 2: Critical conditions

Proceedings of ELM 3: 104-116, 2025

Antoine Cochard, Angeliek van Hout, and Hamida Demirdache:
Distinguishing target Free Choice from non-target Modal Conjunction in Child French. 107

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


(a) Control True (b) Control False

Figure 3: Control conditions

For each condition, participants heard a modal disjunctive statement preceded by a lead-in
fragment introducing the number of coins in the purse as in (9).

(9) Avec
with

une
une

pièce,
coin

Liz
Liz

l’ours
the.bear

peut
can.3sg.prs

acheter
buy.inf

un
a

croissant
croissant

ou
or

un
a

donut.
donut

‘With une coin, Liz the bear can buy a croissant or a donut.’

The Free Choice False condition was designed in the same spirit as previous studies: one of
the disjuncts is accessible while the other is not. In figure 2a, with one coin, it is true that the
customer can buy a donut, but it is false that she can buy a croissant, as the latter costs two coins.
Thus, the free choice inferences do not hold here. Moreover, the modal conjunctive interpretation
does not hold either, for the same reason, and also because the customer cannot buy the croissant
and the donut together. So, rejection of this condition reflects either a free choice interpretation or
a modal conjunctive interpretation. Furthermore, a yes-answer in this condition is a clear signal
of a non-adult-like interpretation. But at this stage it is not yet possible to distinguish an inclusive
interpretation, where at least one disjunct is permitted, from an exclusive interpretation, where
exactly one disjunct is permitted.

The Mutually Exclusive condition is the critical condition which, alongside the Free choice
false condition, was needed to disentangle a free choice interpretation from a modal conjunctive
one. In this condition, the possibility of buying both food items is excluded, falsifying modal
conjunction. Crucially, it allows a free choice interpretation, as it is possible to buy one or the
other pastry, whichever she chooses. In Figure 2b, with one coin, it is true that the customer can
buy a croissant and it is also true that she can buy a donut. But, it is false that she can buy both the
donut and the croissant as she does not have enough coins. Only on the basis of an answer pattern
where the participant gives a consistent yes-answer to the Mutually Exclusive condition alongside
a consistent no-answer to the Free Choice false condition, can it be confidently concluded that the
free choice inference is derived.

The display of the pastries and price in the food stand in the Package condition differs from
the other two conditions. Here, both food items must be bought together in a package. In Figure
2c, the package containing both the donut and the croissant costs one coin. It is not possible to get
the croissant without getting the donut, and conversely. This condition serves two purposes: on the
one hand, it checks whether the optional exclusivity inference shown in (2b) is derived. If so, then
rejection is expected. Combined with the two other conditions, the Package condition is needed
to distinguish two potential adult-like grammars that both yield genuine free choice inferences but
differ in the derivation of the exclusivity inference. On the other hand, although not critical in
distinguishing a genuine free choice interpretation from a modal conjunctive one, it allows the latter
and gives a chance to a modal conjunctive interpreter to give a positive answer in one of the critical

Proceedings of ELM 3: 104-116, 2025

Antoine Cochard, Angeliek van Hout, and Hamida Demirdache:
Distinguishing target Free Choice from non-target Modal Conjunction in Child French. 108

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


conditions.
Table 1 below summarises the expected responses patterns for each of five possible grammars.

Free Choice false Mutually exclusive Package

Ta
rg

et

+ Free Choice + Exclusivity
(^𝑃 ∧ ^𝑄) ∧ ¬^(𝑃 ∧𝑄) NO YES NO

+ Free Choice − Exclusivity
(^𝑃 ∧ ^𝑄) ∧ ^(𝑃 ∧𝑄) NO YES YES

N
on

-T
ar

ge
t

Modal Conjunction
^(𝑃 ∧𝑄) NO NO YES

− Free Choice − Exclusivity
^(𝑃 ∨𝑄) YES YES YES

− Free Choice + Exclusivity
^(𝑃 ∨𝑄) ∧ ¬^(𝑃 ∧𝑄) YES YES NO

Table 1: Response patterns and their corresponding possible grammar

If the conclusion from previous studies that children can derive a genuine free choice inference
is correct, then it is expected that most children will follow one of the two possible target patterns
depending on whether they draw the exclusivity inference: the [+ Free Choice + Exclusive] or the
[+ Free Choice – Exclusive] response pattern. However, if our hypothesis is on the right track
and children interpret modal disjunctive statements as modal conjunction instead, then a subset of
children are expected to fit the modal conjunctive response pattern.

Besides the three critical conditions, twelve filler items served the purpose of balancing the
number of yes and no-responses. Two types of fillers were created for different purposes. Six asked
about elements of the display to make sure that participants were paying attention to the picture
and to the recorded sentence, e.g. by stating the price of a given disjunct as in (10). The other six
checked if participants understood the coin system, for example by stating what could be purchased
with a certain amount of coins as in (11).

(10) Un
a

croissant
croissant

coûte
cost.3sg.prs

une
one

pièce.
coin

‘A croissant costs one coin.’

(11) Avec
with

une
one

pièce,
coin

Liz
Liz

l’ours
the.bear

peut
can.3sg.prs

acheter
buy.inf

un
a

croissant.
croissant

‘With one coin, Liz the bear can buy a croissant.’

4. Results. Exclusion of participants from analysis was determined based on their performance
on the controls and fillers. Participants were not allowed to make more than one error out of six
true controls, no more than one error out of six false controls and no more than three errors out of

Proceedings of ELM 3: 104-116, 2025

Antoine Cochard, Angeliek van Hout, and Hamida Demirdache:
Distinguishing target Free Choice from non-target Modal Conjunction in Child French. 109

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


twelve obvious yes/no filler items. After exclusion, 57 children (range = 3;11–6;9, M = 5;5) and 37
adults (range = 22-67, M = 34) remained for analysis. Group results and mixed model analyses are
reported in section 4.1 while individual analyses are reported in section 4.2.
4.1. Group analyses. Figure 4 shows that children performed differently from adults in the three
critical conditions. To determine whether children’s responses significantly differed from adults,
a generalized linear mixed-effects model was fitted in R (R Core Team 2022) using the glmer
function from the lme4 package (Bates et al. 2015). The model included the interaction2of group
(adults vs. children) and condition (Free Choice False, Mutually Exclusive and Package) as fixed
effects. The controls conditions were not introduced in the model since individual variation is
expected across conditions. Participants were included as random effects with a random slope on
conditions. Taking adults as the reference point, this model revealed a significant main effect of
the group (𝑝 < 0.0001), suggesting that children’s responses differed from the adults at the group
level (Table 2).

Figure 4: Bars show the mean percentage of acceptance for each condition and each group.
Whiskers represent a confidence interval of 95%

To further investigate all contrasts, Tukey-corrected pairwise comparisons were computed
(Table 3). The comparisons confirmed that children’s responses to the Free Choice false condition
differed from those of adults (𝑝 < 0.0001). As for the Mutually Exclusive condition, although
no difference was found (𝑝 = 0.76), the ceiling acceptance in the adult results is problematic as
it prevents us from confidently reaching the conclusion that children performed at adult-like rates
for this condition.3Furthermore, the comparisons also confirmed that children and adults differed
significantly in their acceptance rate of the Package condition, with children being more likely to
accept this condition than adults (𝑝 < 0.0001).

2An ANOVA for linear models showed that an interaction model was more successful in explaining the variance
than an additive model (𝜒2 (4) = 56.5, 𝑝 < 0.0001). Items were not included as random effects since each item features
a unique combination of character and food. A model including items as random effects returns little to no variance at
all for this factor; which means that no particular items stand out and bias the results.

3Indeed, as broadly discussed in Bates et al. (2015), and more specifically for linguistic data in Clark et al. (2023),

Proceedings of ELM 3: 104-116, 2025

Antoine Cochard, Angeliek van Hout, and Hamida Demirdache:
Distinguishing target Free Choice from non-target Modal Conjunction in Child French. 110

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


## Formula: answer ˜ condition * group + (1 + condition | part_number)

## Fixed effects:

## Estimate Std. Error z value Pr(>|z|)

## FC_False:adults (Intercept) -7.317 1.227 -5.963 0.00000000247 ***

## Mutually_Exclusive 28.027 61.202 0.458 0.64700

## Package 4.727 1.571 3.010 0.00262 **

## Children 5.063 1.232 4.109 0.00003977355 ***

## Mutually_Exclusive:Children -23.217 61.204 -0.379 0.70443

## Package:Children 2.236 2.134 1.048 0.29472

Table 2: Mixed model output with adults and children

## Formula: pairwise ˜ group * condition, adjust = "tukey"

## $‘simple contrasts for group‘

## adults - children

## estimate SE df z.ratio p.value

## condition = FC_False: -5.06 1.23 Inf -4.109 <.0001 ***

## condition = Mutually_Exclusive: 18.15 61.21 Inf 0.297 0.7668

## condition = Package: -7.30 1.82 Inf -4.001 0.0001 ***

Table 3: Pairwise comparisons output

Finally, in order to check whether age was a driving factor in explaining children’s responses,
a generalized linear mixed-effects model was computed including only the children. This model
included the critical conditions and the age in months as fixed effects. Similarly to the previous
model, participants were included as random effects with a random slope on conditions. This model
revealed that age played no role (𝑝 = 0.495), Table 4.
4.2. Individual results. To answer our research questions, it is necessary to run an analysis of
the individual results to determine which interpretive patterns were at stake. To do so, we defined
the following criterion: if a participant consistently accepted a condition at least 5 times out of 6,
she is considered as an accepter. If the condition was accepted 0 or 1 time, she was considered as a
rejecter. Any other number was considered to be chance behavior. Table 5 shows the distribution
of participants after applying this criterion.

For the adults, the individual analysis matched the overall results: all except one rejected the
Free Choice False condition and all accepted the Mutually Exclusive condition. However, while
they agreed on the first two conditions, the group split into two uneven groups for the Package
condition, the majority of them rejecting it and about a quarter accepting it. In contrast, for the
children, the individual analysis showed a more diverse distribution. Indeed, although the majority
in the Free Choice False and the Mutually Exclusive conditions resembled the adult cohort, there
were some children who consistently accepted the former and/or rejected the latter. In the Package
condition, the children also split into two groups but in a different manner than adults: while most

model convergence and overfitting issues can be aggravated by specific data scenarios such as ceiling data as it becomes
nearly impossible for the model to estimate the effects of predictors accurately. To mitigate the issue of ceiling data,
an option would be to compute a separate penalized regression using Bayesian priors for instance. We propose such
models in Cochard et al. (2024).

Proceedings of ELM 3: 104-116, 2025

Antoine Cochard, Angeliek van Hout, and Hamida Demirdache:
Distinguishing target Free Choice from non-target Modal Conjunction in Child French. 111

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


## Formula: answer ˜ condition + Age_in_months + (1 + condition | part_number)

## Fixed effects:

## Estimate Std. Error z value Pr(>|z|)

## FC_False (Intercept) -1.4576 0.2857 -5.101 0.000000337 ***

## Mutually_Exclusive 3.2327 0.2570 12.580 <.000000001 ***

## Package 3.7126 0.2727 13.616 <.000000001 ***

## Age_in_months -0.1711 0.2509 -0.682 0.495

Table 4: Mixed model output within children

Free Choice false Mutually Exclusive Package

Adults
Accepter 0 37 9
Rejecter 36 0 25

At chance 1 0 3

Children
Accepter 11 41 42
Rejecter 38 11 5

At chance 8 5 10

Table 5: Distribution of individual patterns across conditions

adults consistently rejected this condition, the majority of the children accepted it and only a few
of them consistently rejected it.

The next step was to determine combined patterns by checking the conditions together, focusing
on the Free Choice False and the Mutually Exclusive conditions. Tables 6 and 7 show the distribution
of response patterns for adults and children,4respectively. Unsurprisingly, all adults fell into the
free choice interpretation category (bottom left cell of Table 6). To the contrary, children were
divided across three distinct but consistent patterns: 22 children fell into the same category as
adults. 11 children accepted both the Free Choice False and the Mutually Exclusive conditions,
corresponding to a [– Free Choice – Exclusive] interpretation, and 11 other children rejected both
conditions, revealing a modal conjunctive interpretation.
5. General discussion. The goal of this study was to investigate how children interpret modal
disjunctive statements in French, putting to test the earlier claims that children derive the associated
free choice inference at adult-like rates. Specifically, we tested the hypothesis that alleged adult-like
children might instead be modal conjunctive interpreters, and, furthermore, investigated whether
children draw the additional exclusivity inference. To carry out these goals, we developed an
innovative design that included two new conditions: the first condition is the Mutually Exclusive
condition which allows us to tease apart an (adult-like) free choice interpretation from a (non-adult-

4To preserve space, at-chance participants in at least one of the two conditions do not appear on Tables 6 and 7.

Proceedings of ELM 3: 104-116, 2025

Antoine Cochard, Angeliek van Hout, and Hamida Demirdache:
Distinguishing target Free Choice from non-target Modal Conjunction in Child French. 112

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


‘Liz can buy a
croissant or a donut.’

^(𝐶 ∨ 𝐷) Mutually Exclusive
Accept Reject

Free Choice False

Accept 0 Inclusive
^(𝐶 ∨ 𝐷)

0
Logically impossible

Reject 36 Free Choice
^𝐶 ∧ ^𝐷

0 Modal Conjunction
^(𝐶 ∧ 𝐷)

Table 6: Distribution of the number of adult individual response patterns for the Free choice False
and the Mutually Exclusive conditions (N = 36/37)5.

‘Liz can buy a
croissant or a donut.’

^(𝐶 ∨ 𝐷) Mutually Exclusive
Accept Reject

Free Choice False

Accept 11 Inclusive
^(𝐶 ∨ 𝐷)

0
Logically impossible

Reject 22 Free Choice
^𝐶 ∧ ^𝐷

11 Modal
Conjunction
^(𝐶 ∧ 𝐷)

Table 7: Distribution of the number of child individual response patterns for the Free choice False
and the Mutually Exclusive conditions (N = 44/57).

like) modal conjunction interpretation, which previous experimental designs failed to differentiate.
The results for this condition validate our hypothesis by showing that some children, who appear
adult-like in rejecting the Free Choice False condition, are in fact modal conjunctive interpreters.
Using mutually exclusive scenarios thus proved essential in teasing apart two subgroups who agreed
on the Free Choice False condition but differed on the Mutually Exclusive condition, successfully
identifying a subset of children who gave a modal conjunctive interpretation of modal disjunctive
statements. The second condition, the Package condition, allows us to further test for the exclusivity
inference, be it on the target (free choice), or the non-target (modal conjunctive) interpretations of
modal disjunction.

Going back to Table 1, four out of the five the expected response patterns were attested,
summarized in Table 8 below.6

Proceedings of ELM 3: 104-116, 2025

Antoine Cochard, Angeliek van Hout, and Hamida Demirdache:
Distinguishing target Free Choice from non-target Modal Conjunction in Child French. 113

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


Children Adults

Target [+ Free Choice]
[+ Exclusivity] 4/44 (9%) 25/36 (69%)
[– Exclusivity] 16/44 (36%) 9/36 (25%)

Non-Target
Modal Conjunction 11/44 (25%) 0/36 (0%)

[– Free Choice – Exclusivity] 11/44 (25%) 0/36 (0%)

Table 8: Summary of the four attested patterns

Starting with the adult-like patterns, our results show that half of consistent 4 to 6-year-olds
(22/44; 50%) computed a free choice inference in an adult-like way, successfully rejecting the
Free Choice False condition and accepting the Mutually Exclusive condition. However, while this
subset corresponds to the largest group of children in our study, it does not reach the adult-like
rates observed in previous studies where the majority of children appeared to derive a free choice
inference (e.g. 98% of rejection of a condition that falsifies free choice inferences in Huang
& Crain 2020). Furthermore, adding the Package condition to the discussion highlights a stark
contrast between adults and children: while most of the adults rejected the Package condition and
derived an exclusivity inference (25/34; 73%), only 4 [+ Free Choice] children (4/44; 9%) did
so. To the contrary, 16 children (16/44; 36%) accepted the Package condition, not deriving the
exclusivity inference.

Turning now to the non-adult-like patterns, our study found that children can face difficulties
with modal disjunctive statements: first, 11 children (11/44; 25%) allowed a [–Free Choice –
Exclusivity] interpretation, accepting both the Free Choice False and the Mutually Exclusive
conditions. These children show the basic inclusive interpretation of disjunction, drawing neither
a free choice, nor an exclusivity inference in modal contexts. Second, the last 11 children (11/44;
25%) allowed a modal conjunctive interpretation, rejecting both the Free Choice False and the
Mutually Exclusive conditions. These results were confirmed in child Romanian by Bleotu et al.
(2024) with an experimental setup replicating our conditions.

Conjunctive interpretations of disjunction have long been reported in the child language liter-
ature, going back to Paris (1973), for matrix disjunction in plain, affirmative contexts (see §1),7 but
also for disjunction embedded under every (Singh et al. 2016). The novel finding reported here is
that these non-target interpretations extend to modal free choice contexts.8

References
Bates, Douglas, Martin Mächler, Ben Bolker & Steve Walker. 2015. Fitting linear mixed-effects

models using lme4. Journal of Statistical Software 67(1). https://doi.org/10.18637/j
ss.v067.i01.

62 out of the 22 [+ Free Choice] children and 2 out of the 36 [+Free Choice] adults were at chance on the Package
condition. For a complete report of the Package condition, see Cochard et al. (2024).

7For recent eye tracking evidence with toddlers, see Lobina et al. (2023).
8This finding is indeed expected, once we assume Cochard et al. (2024) uniform analysis of children’s conjunctive

interpretations of disjunction, across plain/affirmative, universally quantified and modal contexts, straightforwardly
extending Singh et al. (2016).

Proceedings of ELM 3: 104-116, 2025

Antoine Cochard, Angeliek van Hout, and Hamida Demirdache:
Distinguishing target Free Choice from non-target Modal Conjunction in Child French. 114

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


Bleotu, Adina Camelia, Lyn Tieu, Andreea Nicolae, Anton Benz, Gabriela Bı̂lbı̂ie & Mara
Panaitescu. 2024. Free choice and exclusivity in child and adult romanian. Presented at
the ”Free Choice Inferences: Theoretical and experimental approaches” workshop.

Braine, Martin & Barbara Rumain. 1981. Development of comprehension of “or”: Evidence
for a sequence of competencies. Journal of Experimental Child Psychology 31(1). 46–70.
https://doi.org/10.1016/0022-0965(81)90003-5.

Chierchia, Gennaro, Stephen Crain, Maria Teresa Guasti & Rosalind Thornton. 2001. ”some” and
”or”: a study on the emergence of logical form. In S. Catherine Howell, Sarah A. Fish &
Thea Keith-Lucas (eds.), Proceedings of the 24th annual Boston University Conference on
Language Development, 22–44. Cascadilla Press. Sommerville, MA.

Chierchia, Gennaro, Maria Teresa Guasti, Andrea Gualmini, Luisa Meroni, Stephen Crain &
Francesca Foppolo. 2004. Semantic and Pragmatic Competence in Children’s and Adults’
Comprehension of Or. In Ira A. Noveck & Dan Sperber (eds.), Experimental Pragmatics, 283–
300. London: Palgrave Macmillan UK. https://doi.org/10.1057/9780230524125_13.

Clark, Robert, Wade Blanchard, Francis Hui, Ran Tian & Haruka Woods. 2023. Dealing with
complete separation and quasi-complete separation in logistic regression for linguistic data.
Research Methods in Applied Linguistics 2(1). 10–44. https://doi.org/10.1016/j.rm
al.2023.100044.

Cochard, Antoine, Hamida Demirdache & Angeliek van Hout. 2023. Interpreting Disjunction
across Positive and Negative Contexts: Evidence from Child French. In Paris Gappmayr
& Jackson Kellog (eds.), Proceedings of the 47th annual Boston University Conference on
Language Development, 159–172. Somerville, MA: Cascadilla Press.

Cochard, Antoine, Angeliek van Hout & Hamida Demirdache. 2024. Acquiring Modal Disjunctive
Statements: Free Choice vs. Modal Conjunction. [Unpublished manuscript].

Crain, Stephen & Rosalind Thornton. 1998. Investigations in Universal Grammar: A Guide to
Experiments on the Acquisition of Syntax and Semantics. Cambridge, MA: MIT Press.

Huang, Haiquan & Stephen Crain. 2020. When OR is assigned a conjunctive inference in child
language. Language Acquisition 27(1). 74–97. https://doi.org/10.1080/10489223.2
019.1659273.

Kamp, Hans. 1973. Free Choice Permission. Proceedings of the Aristotelian Society 74. 57–74.
Publisher: Aristotelian Society, Wiley.

Liu, Ying. 2017. Interpreting Disjunction under Deontic Modals: An Experimental Investigation:
Utrecht University MA thesis.

Lobina, David J., Hermann Bulf & Maria Teresa Guasti. 2023. The Comprehension of Conjunction
and Disjunction by Toddlers. https://doi.org/10.17605/OSF.IO/S7K26. [OSF
Preregistration].

Paris, Scott G. 1973. Comprehension of language connectives and propositional logical relation-
ships. Journal of Experimental Child Psychology 16(2). 278–291. https://doi.org/10.1
016/0022-0965(73)90167-7.

R Core Team. 2022. R: A Language and Environment for Statistical Computing. R Foundation for
Statistical Computing. https://www.R-project.org/.

Simons, Mandy. 2005. Dividing things up: The semantics of or and the modal/or interaction.

Proceedings of ELM 3: 104-116, 2025

Antoine Cochard, Angeliek van Hout, and Hamida Demirdache:
Distinguishing target Free Choice from non-target Modal Conjunction in Child French. 115

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/


Natural Language Semantics 13(3). 271–316. https://doi.org/10.1007/s11050-004
-2900-7.

Singh, Raj, Ken Wexler, Andrea Astle-Rahim, Deepthi Kamawar & Danny Fox. 2016. Children
interpret disjunction as conjunction: Consequences for theories of implicature and child
development. Natural Language Semantics 24(4). 305–352. https://doi.org/10.1007/
s11050-016-9126-3.

Tieu, Lyn, Jacopo Romoli, Peng Zhou & Stephen Crain. 2016. Children’s Knowledge of Free
Choice Inferences and Scalar Implicatures. Journal of Semantics 33(2). 269–298. https:
//doi.org/10.1093/jos/ffv001.

Tieu, Lyn, Kazuko Yatsushiro, Alexandre Cremers, Jacopo Romoli, Uli Sauerland & Emmanuel
Chemla. 2017. On the role of alternatives in the acquisition of simple and complex disjunctions
in french and japanese. Journal of Semantics 34(1). 127–152. https://doi.org/10.109
3/jos/ffw010.

Zhou, Peng, Jacopo Romoli & Stephen Crain. 2013. Children’s knowledge of free choice inferences.
In Todd Snider (ed.), SALT 23, 632–651. Cornell University.

Proceedings of ELM 3: 104-116, 2025

Antoine Cochard, Angeliek van Hout, and Hamida Demirdache:
Distinguishing target Free Choice from non-target Modal Conjunction in Child French. 116

https://journals.linguisticsociety.org/proceedings/index.php/ELM/issue/archive
https://www.elm-conference.net/