Subordinate questions in Swedish by 12- and 15-year-old Finnish immersion students

Abstract

The present study explores how 12- and 15-year-old immersion students
(n=75 and n=73) produce subordinate questions in Swedish on a written test.
Previous studies are sparse, but they report difficulties with both subject-verb
word order and use of the subjunctor om and the subject marker som
occurring in these clauses; informants with varying ages and competence
levels struggle with similar problems. However, the acquisition order between
these two types of constructions, a central theme in this study, has gained
less attention. Analyses of the actual data show significant differences with
varying effect sizes in accuracy between the different subcategories of
subordinate questions and both informant groups. Insertion of grammatical
words was mastered by significantly fewer informants than word order. Also,
effect sizes were large in these contexts. Older informants do better than the
younger ones, but the differences are not always statistically significant, as
certain constructions are already mastered at a high level by the younger
informants, whereas other constructions are still difficult for the older ones.

Full text

Subordinate questions
in Swedish by 12- and 15-
year-old Finnish immersion
students
Eeva-Liisa Nyqvist (University of Helsinki)
Abstract
The present study explores how 12- and 15-year-old immersion students
(n=75 and n=73) produce subordinate questions in Swedish on a written test.
Previous studies are sparse, but they report difficulties with both subject-verb
word order and use of the subjunctor om and the subject marker som
occurring in these clauses; informants with varying ages and competence
levels struggle with similar problems. However, the acquisition order between
these two types of constructions, a central theme in this study, has gained
less attention. Analyses of the actual data show significant differences with
varying effect sizes in accuracy between the different subcategories of
subordinate questions and both informant groups. Insertion of grammatical
words was mastered by significantly fewer informants than word order. Also,
effect sizes were large in these contexts. Older informants do better than the
younger ones, but the differences are not always statistically significant, as
certain constructions are already mastered at a high level by the younger
informants, whereas other constructions are still difficult for the older ones.
Keywords: immersion, Swedish as an L2, indirect questions,
complexity, usage-based grammar
© 2021 Eeva-Liisa Nyqvist. This is an open access article distributed under the Creative Commons
BY 4.0 license (https://creativecommons.org/licenses/by/4.0/deed.en).
Folia Scandinavica
VOL.30, 2021 (p. 15–25)
DOI: 10.2478/fsp-2021-0002

Folia Scandinavica Posnaniensia 30 (2021)
Nyqvist: Subordinate questions in Swedish by 12- and 15-year-old Finnish immersion students
16
1. INTRODUCTION
Immersion is a second language (L2) teaching programme with excellent results (Bergroth
2015, Lyster 2007), but it requires further development (Lyster 2007). Canadian and Irish
studies (Harley 1993, 1998; Ó Duibhir 2009) have revealed problems with grammatical accuracy.
Finnish immersion research has been manifold (Bergroth/Björklund 2013), but the development
of grammatical competence has gained minor attention.
This article explores how Finnish-speaking pupils in early total immersion (henceforth
immersion) master subordinate questions, i.e. a subcategory of subordinate clauses, in a written
test at the end of primary school (at age 12) and at the end of secondary school and immersion1
(at age 15) and what kinds of difficulties they have with these constructions. The data consist
of a written test. Subordinate questions play a central part in polite language (Lahtinen/Toro-
painen 2015); hence, it is important to be able to use them.
Factors behind difficulties in second language acquisition (SLA) can be feature-related
(properties of a linguistic construction), context-related (different learning conditions), and learner-
related (individual characteristics; Housen/Simoens 2016). This study focuses on learner-related
factors in the form of a comparison between younger and older informants. It also offers
information about which aspects of subordinate questions are the most difficult (i.e. feature-
related difficulty) and, hence, which must be the precise focus of more explicit instruction. It is
well founded for grammar instruction to focus on moderately difficult and difficult constructions
because the learners benefit the most from doing so (DeKeyser 2003). It is likewise important
to explore which constructions are most difficult by analysing language produced by L2
speakers instead of analysing only the grammatical descriptions of a language: what seems to
be difficult for a linguist describing a language does not need to be difficult for an L2 speaker
(Hammarberg 2008) and vice versa. It is also important to study immersion pupils separately
from other L2 speakers of Swedish as this especially intensive, input-rich and long-lasting learning
programme combines both communication and a focus on form that distinguishes it from many
of the other methods.
The direct questions usually follow the V2-rule in Swedish, i.e. they have subject-verb
inversion (henceforth inversion). Swedish subordinate questions, on the other hand, typically
have a canonical subject-verb word order (henceforth word order)2. Thus, transforming a direct
question into a subordinate one typically implies that the inverted word order of the direct
question must be cancelled, i.e. straightened (Teleman et al. 1999a). Both direct and subordinate
questions can be divided into two categories in Swedish: those beginning with an interrogative
word (henceforth vx-questions when the interrogative word is not the subject, e.g. Jag vet vad
du menar, and vs-questions when the interrogative word is the subject or part of it, e.g. Jag vet
vad som har hänt) and those lacking that interrogative word (henceforth yes/no questions)
(Teleman et al. 1999b.) In addition to the cancelled inversion, the subordinate yes/no questions
begin with the subordinator om, whereas vs-questions include the subject marker som in the
subordinate clause (Holmes/Hinchcliffe 1994, Hammarberg/Viberg 1979).
2. USAGE-BASED GRAMMAR AND ASPECTS OF DIFFICULTY
Children acquire their first language (L1) by memorising concrete utterances (constructs) that
combine both form and meaning. In time, they discover regularities in these constructs and start
varying them, ultimately discovering the abstract constructions behind them. The constructions
1 There are no immersion upper secondary schools in Finland (Bergroth 2015).
2 In spoken Swedish, inversion can also occur, especially in subordinate questions (Teleman et al. 1999b). This variation
is, nevertheless, stylistic, not evidence of language change (Källström 2000).

Folia Scandinavica Posnaniensia 30 (2021)
Nyqvist: Subordinate questions in Swedish by 12- and 15-year-old Finnish immersion students
17
become increasingly complex and schematised over time as the learners abstract on how the
parts link together and build relationships between them, which permits the production of brand
new utterances (Nistov et al. 2018). Thus, grammar is an implicit, cognitive organisation of a learner’s
actual language experience, and it develops when new constructions are added to the inventory
(Bybee 2008). In this article, I define both word order for subordinate questions and for the
grammatical words om and som as constructions.
SLA in immersion begins early on and takes place mostly as a spontaneous process when
the learner focuses on meaning. Therefore, it is appropriate to draw parallels between L1 and L2
acquisition, although the SLA of immersion pupils is already impacted by their L1, directing
whether they notice constructs in the input (N. Ellis/Wullf 2015). One’s L1 can hamper SLA if
the L1 lacks, e.g. grammatical morphemes occurring in the L2 (Collins et al. 2009), especially
at the earlier stages of the acquisition (Bybee 2008). Finnish subordinate questions have, e.g.
the same word order as direct questions3 (Hakulinen et al. 2004), and they lack equivalents for
om/som. Hence, Finnish-speaking L2 learners of Swedish may experience difficulties noticing
these words in the input as they tend to ignore them as redundant. Learners are also likely to
transfer elements from other L2s they have mastered when the distance between the L1 and the
target language is greater than between the target language and another L2 (De Angelis 2007).
Since the informants in this study also learn English at school, their acquisition of Swedish is
likely affected. Subordinate vx- and yes/no questions in Swedish and English are actually very
similar to one another: The word order is canonical in both languages, and the Swedish
subjunctor om equals the English subjunctor if. The subject marker som, on the other hand, is a mor-
pheme without an English equivalent.
Many constructions of an L2, though, cannot be learned without explicit instruction
(N. Ellis/Wulff 2015). It is easier to learn frequent traits, and repetition makes memory repre-
sentations more accessible. Extremely frequent sequences can be acquired as wholes as if they
were independent of a general pattern, and consequently, they can help the learner analyse similar
but less frequent forms. However, high-frequency elements tend to become phonologically
reduced, i.e. difficult to notice in the input, i.e. their salience is low (Bybee 2008). This explains
why several highly frequent grammatical morphemes in an L2 are difficult to acquire implicitly:
One cannot acquire what one has not noticed (Goldschneider/DeKeyser 2001).
Subordinate questions are rare in both spoken (Jörgensen 1978) and written Swedish
(Hultman/Westman 1977). Especially rare are vs-questions, as only a few interrogative words
can act as a subject in a sentence, e.g. vem (‘who’), vad (‘what’) and vilken (‘which’). Moreover,
these interrogative words can also act as other sentence constituents when the subject marker
som should not be used (e.g. Jag frågade vem du såg, ‘I asked whom you saw.’).
According to DeKeyser (2005), a grammatical construction can be difficult to learn for
three reasons. Firstly, meaning is difficult if it is abstract or novel for languages previously
acquired by the learner. Both om and som have abstract meanings; they only signal the fact that
a clause is subordinate. The subordinator om, however, has an equivalent in English, so it might
be easier than the subject marker som, which does not. Omission of om also leads to an obscure
utterance (*Jag vet inte han kommer), much like if the subordinator if were omitted in English
(*I don’t know he is coming).
Secondly, the source of the difficulty might lie in the form, especially if it is complex, i.e.
the learner is obliged to make many decisions concerning, e.g. different allomorphs (DeKeyser
2005). Choices of allomorphs are not applicable to subordinate questions, but the clauses in which
3 E.g. Tuleeko hän huomenna? ‘Is he coming tomorrow?’; En tiedä, tuleeko hän huomenna, ‘I don’t know if he comes
tomorrow’; Mitä hän tekee? ‘What is he doing?’; En tiedä, mitä hän tekee, ‘I don’t know what he’s doing.’

Folia Scandinavica Posnaniensia 30 (2021)
Nyqvist: Subordinate questions in Swedish by 12- and 15-year-old Finnish immersion students
18
the insertion of om/som is obligatory can be considered more complex than vx-questions: they
include two constructions inside each other, and hence, one must be able to use both the right
word order and the right grammatical word.
Thirdly, difficulty may stem from the relationship between form and meaning. These can
both be easy per se, but the link between them might still be problematic to grasp (DeKeyser
2005). In subordinate questions, the problem is opacity, where different constructions stand for
the same meaning. Direct questions usually have an inverted word order, whereas subordinate
questions follow a canonical word order, but both types are questions. This might be especially
perplexing for Finnish learners, as direct and subordinate questions have the same word order
in their L1. To use inverted and canonical word order in accurate contexts, they should also
manage the distinction between main and subordinate clauses, but this is often difficult (Rahkonen/
Håkansson 2008).
3. PREVIOUS RESEARCH
Philipsson’s (2007) doctoral dissertation is the most comprehensive study on subordinate
questions in L2 Swedish so far. The acquisition order in written production by the most advanced
informants was yes/no<vx<vs. The difficulty of vs-questions, according to Philipsson, was
caused by the omission of som, not by an inaccurate word order.
Nyqvist (2020) has previously analysed the actual data at the individual level (see section
4.2 below) with 12- and 15-year old Finnish-speaking immersion students and a control group
of 16-year-old formal learners of L2 Swedish as informants. As this analysis dealt with the
word order and the use of om/som separately, it led to an acquisition order of vs<vx<yes/no,
which differs from the one found by Philipsson (2007) but still supports his perception: the
word order is not the main problem in the vs-questions. Nyqvist’s (2020) results also showed
that the formal learners reached a higher level of accuracy in the word order than the immersion
groups. Inversion in subordinate questions was also documented by Hyltenstam/Lindberg
(1983) and Viberg (1990).
Only 53% of Rahkonen’s/Håkansson’s (2008) informants, who were Finnish-speaking
formal learners (17–18-year-old high school students that had received explicit instruction) of
L2 Swedish, mastered canonical word order in subordinate questions (the type of subordinate
question not specified), whereas 95% mastered inversion in direct questions. This was in line
with the order presented in Processability Theory, but Rahkonen/Håkansson also emphasised
the low frequency of the subordinate questions and the difficulty in distinguishing subordinate
and direct questions (cf. Philipsson 2007) from one another as factors explaining the order. The
fact that the reporting clauses preceding the subordinate questions were often direct questions
(e.g. Vet du när han kommer? ‘Do you know when he comes?’) might have also played a part.
Problems with L2 learners’ use of som by were noted by Philipsson (2007), Viberg (1990)
and Nyqvist (2020). Philipsson (2007), however, did not explore the equivalent phenomenon
with om in yes/no questions, and even Viberg (1990) only stated that om was sometimes
omitted, whereas Nyqvist (2020) showed that the use of om is problematic for the younger
immersion students, whereas the older immersion students and the control group reached a high
accuracy at the individual level.
4. DESIGN AND METHODOLOGY
4.1. INFORMANT GROUPS
The informants in this study make up two immersion groups (the same individuals as in
Nyqvist (2020)): the first consists of 12-year-old Finnish-speaking 6th graders (n=75, henceforth
IM6) and the second group of 15-year-old Finnish-speaking 9th graders (n=73, henceforth IM9)4.
4 Statistical power 0.8 and 0.9, respectively (Larson-Hall 2016).

Folia Scandinavica Posnaniensia 30 (2021)
Nyqvist: Subordinate questions in Swedish by 12- and 15-year-old Finnish immersion students
19
The informants started learning Swedish at daycare at 4–5 years of age (for the average starting
ages for immersion in Finland, see Bergroth 2007), so they have been learning Swedish for 8–
9 years (IM6) and 11–12 years (IM9), respectively.
The proportion of instruction given in Swedish varies in different grades. Immersion pupils
receive 50% of their instruction in Swedish in the comprehensive school. The actual proportion
is 50% in the 6th grade and 45% in the 9th grade (Bergroth/Björklund 2013). The standards set
for competence in the immersion language are essentially higher than in the non-immersion
context: students have to reach B-level on the CEFR scale in order to reach the level of ‘good’
at the end of immersion (i.e. in the 9th grade; Bergroth 2015). The informants have also been
learning English since the age of nine, and they live in continuous contact with it; hence, their
L2 Swedish might bear certain traits of their L3 English (FNBE 2014).
4.2. DATA AND ANALYSIS
The data consist of a written test during which the informants received help with vocabulary
but not with grammar. The test was part of an extensive package of grammatical tests the
informants wrote during one of their regular Swedish lessons. This package could not be too
long, so the part focusing on the subordinate questions is relatively short. There are 12 direct
questions that the informants transformed into subordinate ones, e.g. Hur kallt är det? → Jag vet
inte hur kallt det är (‘How cold is it outside?’ → ‘I don’t know how cold it is outside.’). Vx-, vs-
and yes/no questions have four obligatory occasions on the test. There are 900 subordinate
clauses in IM6 and 876 in IM9. One disadvantage of using a grammar test as data is that it does
not deal with practical communicative competence, which is essential in immersion. Due to the
low frequency of subordinate questions, however, the only way to obtain enough obligatory
occasions for all studied constructions is to use elicited data. The actual test also involves written
production, i.e. it is not a multiple-choice test where an informant would have a 50% chance to
answer correctly.
The data are analysed at the group level. Accuracy scores for the different types are calcu-
lated as a whole and separately from the perspective of word order and of the insertion of om/som.
The subordinate question (Jag undrar) om *var maten god (‘I wonder if *did the food taste good’),
for example, is inaccurate as far as word order is concerned, whereas the use of the subjunctor
om is accurate. The subordinate question (Jag undrar) *maten var god (‘I wonder *the food
tasted good’), in contrast, has the accurate word order, but the subjunctor om has been omitted.
However, all these clauses are classified as inaccurate when the overall accuracy has been cal-
culated. Accuracy scores are calculated by dividing the number of accurate obligatory occasions
by the total number of obligatory occasions: e.g. in IM9, 260 of the 292 vs-questions have an
accurate word order. Hence, the accuracy score is 89%. Acquisition sequences were established
in line with the principle wherein an acquisition sequence delivers an acquisition order; a high
accuracy implies early acquisition and, consequently, an easy construction (Collins et al. 2009).
Pearson’s χ² (value of p <0.05) will be used as a statistic test as it does not require Gaussian
distribution. As statistical significance sometimes occurs solely because of a high number of
informants (Sullivan/Feinn 2012), the effect sizes of results have also been calculated (Cramér’s
V; V <0.5 large, V=.3–0.5 medium, V <0.3 small when df=15; Cohen 1988:79–80)3. The central
research questions are:
1. Which type of subordinate question is easiest/most difficult?
2. Are om/som markers more difficult than word order?
5 df=1 in all cases in this study.

Folia Scandinavica Posnaniensia 30 (2021)
Nyqvist: Subordinate questions in Swedish by 12- and 15-year-old Finnish immersion students
20
5. RESULTS
In this section, I present quantitative data. Table 1 presents the accuracy scores, among which
there are no differences between the inaccuracies in word order and in the use of om/som.
Group vx yes/no vs
accurate/all % accurate/all % accurate/all %
IM6
148/300
49%
42/300
14%
0/300
0%
IM9 210/292 72% 194/292 66% 36/292 12%
Tab. 1 Overall accuracy scores for the different types of subordinate questions in the two informant groups
Overall accuracy scores in both groups are highest in the vx-questions and lowest in the vs-
questions (as in Philipsson 2007). The accuracy score of vs-questions in IM6 (0%) is signifi-
cantly lower than that of the two other subcategories in both groups (p <.001 in both cases).
The effect sizes are large in both groups when vs-questions are compared to vx-questions
(V=.572 in IM6; V=.603 in IM9). In IM9, the effect size is likewise large when compared to
yes/no questions (V=.554), in which IM9 reaches an accuracy score that is 54 percentage points
higher than in IM6.
Vx-questions are also mastered at a significantly higher level than yes/no questions in IM6
with a medium effect size (p <.001, V=.405), which also underlines the group’s low overall
accuracy score for yes/no questions. The accuracy scores of IM6 are lower than those of IM9
(p <.001 in all cases), but the effect size is large solely in yes/no questions (V=.535; V=.231 in
vx-questions and V=.258 in vs-questions). Table 2 presents the accuracy scores for word order.
Group
vs
v
x
yes/no
accurate/all
%
accurate/all
%
accurate/all
%
IM6 280/300 93% 148/300 49% 67/300 22%
IM9
260/292
89%
210/292
74%
194/292
66%
Tab. 2 Accurate word order in subordinate clauses in the two informant groups
When the results are explored from the word order’s point of view, both informant groups
reach the highest accuracy score in vs-questions and the lowest in yes/no questions; i.e. the
acquisition order differs from that in Philipsson (2007) but is similar to that in the analysis of
the same data at the individual level (Nyqvist 2020). It is not surprising that vs-questions have
the highest accuracy scores, as the word order is also canonical in the corresponding direct
question. The accuracy score, however, is not 100% in either group: a minority of informants
in both groups has inverted the word order in vain.
Accuracy is significantly higher for vs-questions than for both vx-questions and yes/no
questions in IM6 (p <.001 in all cases); effect size is large when vs- and yes/no questions are
compared (V=.719) and medium when vs- and vx-questions are compared (V=.487). Also in
IM9, the word order is significantly higher in vs-questions than in both vx-questions and yes/no
questions (p <.001 in both cases), but with small effect sizes (V=.216 and V=.272).
Consequently, differences between the different types of subordinate questions are less substantial
than in IM6.
IM9 reaches significantly higher accuracy scores than IM6 in vx- and yes/no questions
(p <.001 in both). A medium effect size is noted only in yes/no questions (V=.444), in which
the accuracy score in IM9 is as much as 44 percentage points higher than in IM6, whereas vx-
questions manifest only a small effect size (V=.231). In vs-questions, the difference between
IM6 and IM9 is almost significant (p=.065), but with a very small effect size (V=.076).

Folia Scandinavica Posnaniensia 30 (2021)
Nyqvist: Subordinate questions in Swedish by 12- and 15-year-old Finnish immersion students
21
Five of the reporting clauses are direct questions, and seven are declarative clauses. The
type of reporting clause has a nonsignificant impact with very small effect sizes on the accuracy
of the word order in both groups (p=.0812 and V=.008 in IM6; p=.373 and V=.03 in IM9).
Thus, there is no evidence of any kind for interrogative reporting clauses leading to increased
failure to use the accurate word order (cf. Rahkonen/Håkansson 2008). Table 3 summarises the
accuracy scores from the point of view of the use of om/som in the different informant groups.
Group
o
m
som
accurate/all
%
accurate/all
%
IM6 42/300 14% 0/300 0%
IM9
195/292
67%
42/292
14%
Tab. 3 Accurate use of om/som in the two informant groups
The omission of som in both groups is significantly more common than that of om (p <.001
in both). The effect size is large in IM9 (V=.534) and close to medium (V=.274) in IM6, where
the omission of om is also common. Accuracy scores for both grammatical words are signifi-
cantly higher in IM9 (p <.001 in both cases), but the effect size is large merely in the case of
om (V=.592); in som, only a small effect size is observed (V=.28) as accurate contexts for it are
still rare in IM9.
When accuracy scores for the insertion of om/som and word order are compared, one can see
that using som in both informant groups is significantly more difficult with large effect sizes than
the word order in the questions in which it occurs (i.e. vs-questions, p <.001 in both groups; V=.935
in IM6; V=.747 in IM9). This confirms the fact that the use of som is problematic for both informant
groups. The insertion of om is significantly more difficult than word order only in IM6 (p=.008),
and the effect size is small (V=.108), as the use of om is less problematic than som for them.
An essential difference between om and som is the fact that the omission of om co-occurs
with inaccuracies in word order in both groups: no fewer than 90% of the subordinate yes/no
questions with an inaccurate word order also lack om in IM6. In IM9, the equivalent proportion
is 99%. In subordinate vs-questions, only 7% of questions in IM6 and 10% in IM9 have both
an inaccurate word order and omission of som, i.e. the percentages are significantly lower than
those for the co-occurrence for om + inaccurate word order with very large effect sizes (p <.001
in both groups; V=.821 in IM6; V=.844 in IM9). Essentially, omission of the subordinator om
and an inaccurate word order in yes/no questions are closely intertwined.
Besides the omission of om/som, there are instances of overusing another particle, the
subordinating conjunction att (the primary meaning of which is ‘that’, as in ‘I think that he is
right.’). This occurs in both IM6 and IM9 (43 and 50 instances, respectively) and in all types
of subordinate questions. The phenomenon is explainable by the L1 of the informants: nowadays,
että (the Finnish equivalent for att) is commonly used as a linking word between a reporting
clause and the reported speech, including subordinate questions especially in spoken Finnish
but also in informal written language (Korhonen 2009). According to Teleman et al. (1999b),
the conjunction att in reported speech also occurs in informal Swedish, but their examples do
not include subordinate questions, which implies that the construction [att + subordinate question]
is unlikely and might confuse L1 speakers of Swedish.
6. DISCUSSION AND CONCLUSION
This study aimed to explore the extent to which subordinate questions are mastered at the group
level by Finnish-speaking 12- and 15-year-old immersion students who had completed a test
where they were expected to transform direct questions into indirect ones. Subordinate questions

Folia Scandinavica Posnaniensia 30 (2021)
Nyqvist: Subordinate questions in Swedish by 12- and 15-year-old Finnish immersion students
22
are low-frequency constructions (Jörgensen 1978, Hultman/Westman 1977), and they include
so many formal characteristics that they are typologically difficult to acquire (Källström 2000).
However, they play a part in communication as a politeness strategy (Lahtinen/Toropainen 2015),
making them important to master.
Results from the analyses in both informant groups are parallel to those obtained in the
analysis of the same data at the individual level (Nyqvist 2020). They reveal that subordinate
clauses basically involve two different learning tasks: use of the canonical word order typical
of Swedish subordinate clauses and the use of om/som occurring in yes/no and vs-questions,
respectively. The previous research (e.g. Philipsson 2007, Rahkonen/Håkansson 2008), however,
has mainly focused on word order. The word order of subordinate clauses is indeed a notorious
source of difficulty for L2 learners (e.g. Hyltenstam 1992), but it is likely to be especially diffi-
cult in subordinate questions due to opacity (cf. DeKeyser 2005), i.e. direct and subordinate
questions have basically the same meaning, but they typically have different word orders. It can
be especially difficult for Finnish-speaking learners, whose L1 has the same word order in both
types of questions (Hakulinen et al. 2004). Immersion students are often seen as privileged L2
learners because they receive rich input with plenty of occasions for meaningful communication,
but they still appear to need instruction in word order, although they have reached a higher level
than, e.g. Rahkonen/Håkanssons’s (2008) informants.
When accuracy scores at the group level are examined without distinguishing between
inaccuracies in word order and the use of om/som, vx-questions have a higher accuracy score
than both yes/no and vs-questions in both informant groups; i.e. the acquisition order is identical
to the one presented by Philipsson (2007), who also stated that the low accuracy of vs-questions
was not caused by the word order but by the subject marker som. This also explains the
acquisition sequence in the actual study. Vx-questions lack contexts for om/som, i.e. they are less
complex (cf. DeKeyser 2005) than other types of subordinate questions. Moreover, they are
likely to be a rather frequent type of subordinate question in the input received by the informants.
When the focus shifts to word order, however, both groups master vs-questions at a higher
level than vx- and yes/no questions (as in Nyqvist 2020). This is because the informants do not
have to cancel inversion in these, as the equivalent direct question also manifests the canonical
word order. However, it is important to note that the word order in vs-questions is not mastered
at 100%. An analysis of reporting clauses also revealed that their forms, i.e. whether they are
declarative or interrogative, do not impact the accuracy of the word order, although Rahkonen/
Håkansson (2008) speculated about their role.
Yes/no questions are the most difficult type of subordinate question in both groups when
the analysis focuses on word order, and similar results have been documented in an analysis at
the individual level (Nyqvist 2020); my results differ essentially from Philipsson’s (2007). This
might be because Philipsson’s informants living in Sweden have received even more input than
my informants. The use of inverted word order is widespread especially in IM6, and the omission
of om and inverted word order co-occur in both informant groups. Regardless of how the
accuracy is calculated, one can conclude that the different types of subordinate questions have
different levels of difficulty, although they have the word order. The fact that yes/no questions
have the lowest accuracy when focusing on word order might be because the subjunctor om
adds to the complexity of the clause type (cf. DeKeyser 2005) in a way that also impacts their
mastery of the word order.
In short, answering whether accurate word order is easier for the informants than the accurate
use of om/som is ‘yes’ for both groups. However, there are differences between these words.
The insertion of the subject marker som is especially problematic; its accuracy score is 0% in
IM6 and 14% in IM9, whereas 11% of informants master it (cf. Nyqvist 2020). In other words,
it is a very difficult construction: som is a short grammatical word with an abstract meaning and
low salience. It is also low frequency, it lacks equivalents in other languages commonly known

Folia Scandinavica Posnaniensia 30 (2021)
Nyqvist: Subordinate questions in Swedish by 12- and 15-year-old Finnish immersion students
23
by the informants and co-occurs with only a few interrogative words (those that act as a subject
in the subordinate clause), and it adds to the clauses’ complexity. Hence, the word is difficult
where both meaning and form are concerned (cf. DeKeyser 2005; see also N. Ellis/Wulff 2015).
Besides the inaccurate use of om/som, the ungrammatical use of att also occurs in subordinate
questions.
IM6 also masters the subjunctor om essentially at a lower level than IM9. The aspects of
difficulty concerning som are also applicable to om, but om has a direct equivalent in English
(if), and its omission makes the utterance difficult to understand, which might explain why the
older informants master it at a higher level than som. It is possible that the similarity between
Swedish and English and the omission’s dramatic effect on its comprehensibility increase the
salience of om. Similar results were found at the individual level in both immersion groups and
in the group receiving formal instruction (Nyqvist 2020). The insertion of som is, likewise,
more difficult than the use of canonical word order in both informant groups. A similar pheno-
menon can be detected in yes/no questions in IM6, whereas accuracy scores for the word order
in yes/no questions and for the insertion of om are at approximately the same level in IM9, i.e.
the older informants reach better learning results, contrary to som.
Hence, one can assume that feature-related factors such as frequency, salience and complexity
(Housen/Simoens 2016) play a crucial role in the acquisition of subordinate questions. A compari-
son between IM6 and IM9 showed that IM9 mastered the studied constructions to a greater
extent than IM6 in all cases except for word order in vs-questions, which are mastered at a non-
significantly higher level in IM6. Effect sizes for the differences are medium to large only for
word order in yes/no questions and in the use of om. Thus, learner-rated difficulty (Housen/
Simoens 2016) appears most visible in these constructions; in others, the informants already reach
a rather high level in IM6, whereas others are very difficult even for IM9.
In some cases, inaccuracies in subordinate questions indeed put comprehensibility in danger,
but they also label the speaker as an L2 speaker; therefore, there is a reason to focus on them in
second language instruction. In short, my informants have not yet completely mastered word
order in subordinate questions, but other problematic constructions are the low-frequency,
insalient grammatical words om and som.
It is therefore important to increase salience and noticing (see Bybee 2008) of subordinate
questions during Swedish lessons. Their low frequency can be resolved by providing the
learners with skewed input in which subordinate questions have several occurrences, as high
frequency strengthens memory representations (Goldberg/Casenhiser 2008, DeKeyser 2005).
DeKeyser (2005) also recommends conscious analyses of input as part of instruction, as
grammatical constructions might otherwise be at risk of being disregarded by L2 learners. Even
Swedish researchers (Håkansson et al. 2019, Prentice et al. 2016) have proposed an increased
focus on pattern recognition as an effective method of L2 instruction. A deliberate analysis of
subordinate clause examples can create aides-memoire, which can be utilised by the learners
when analysing other examples of the same construction (Bybee 2008, N. Ellis/Wulff 2015).
Accurate use of the formally complex subordinate questions should also be practised
intensively. Interleaving, i.e. practising multiple skills in an irregular order in the same exercise,
leads to better results than blocking in the long run; focusing on one construction at a time is
beneficial, as interleaving leads to deeper processing and also resembles actual language use
(Nakata/Suzuki 2019). As Swedish has three different types of subordinate questions with the
same word order but different accuracy scores, and as two of the types also include difficult
grammatical words, it is possible to create interleaving exercises with subordinate questions.
Enhancing grammar acquisition, however, ought to be a collective task for all teachers of
immersion students in Swedish, not only for their language teachers. According to the counter-
balanced approach (Lyster 2007), teachers in subjects other than languages should also be able
to shift the focus from their own subject to linguistic questions when needed (e.g. when a student

Folia Scandinavica Posnaniensia 30 (2021)
Nyqvist: Subordinate questions in Swedish by 12- and 15-year-old Finnish immersion students
24
faces a problem related to the language) or when it otherwise feels natural. The students’ attention
can be drawn to the language, e.g. with help from different fonts, by the teacher stressing certain
grammatical morphemes, or by explicit comments that, in these cases, need not include
advanced grammatical terminology (‘Did you notice the word som in this subordinate question?’).
Thus, utterances occurring in, e.g. history compendia can also generate formulaic sequences
acting as aides-memoires outside the lessons. When the rich input and meaningful communication
typical of immersion are combined with effective instruction, there is a good chance that learners
will reach a high level of competence.
REFERENCES
Bergroth, M. (2007). Kielikylpyperheet valokeilassa. Taustat ja odotukset. Vaasa: Publications of the University
of Vaasa.
Bergroth, M. (2015). Kotimaisten kielten kielikylpy. Publications of the University of Vaasa. Retrieved from
http://www.uva.fi/materiaali/pdf/isbn_978-952-476-617-3.pdf. (10.06.2020)
Bergroth, M., Björklund, S. (2013). Kielikylpyohjelman tutkimustuloksia Suomessa. In: L. Tainio & H. Harju-
Luukkainen (eds.), Kaksikielinen koulu – tulevaisuuden monikielinen Suomi/Tvåspråkig Skola – ett flerspråkigt
Finland i Framtiden (pp. 91–114). Jyväskylä: Finnish Educational Research Association.
Bybee, J. (2008). Usage-based grammar and second language acquisition. In: P. Robinson & N. C. Ellis (eds.),
Handbook of cognitive linguistics and second language acquisition (pp. 216–236). New York: Routledge.
Cohen J. (1988). Statistical power and analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence
Erlbaum Associates.
Collins, L. et al. (2009). Some input on the easy/difficult grammar question: An empirical study. Modern Language
Journal 93, 336–353. DOI: https://doi.org/10.1111/j.1540-4781.2009.00894.x
De Angelis, G. (2007). Third or additional language acquisition. Clevedon: Multilingual Matters.
DeKeyser, R. (2003). Implicit and explicit learning. In: C.J. Doughty & M. Long (eds.), The handbook of second
language acquisition (pp. 313–348). Oxford: Blackwell Publishing.
DeKeyser, R. (2005). What makes learning second language grammar difficult? A review of issues. Language
Learning 55/S1, 1–25. DOI: https://doi.org/10.1111/j.0023-8333.2005.00294.x
Ellis, N. C., Wulff, S. (2015). Usage-based approaches to SLA. In: B. VanPatten & J. Williams (eds.), Theories in
second language acquisition: An introduction (pp.75–93). New York: Routledge.
FNBE 2014=Finnish National Board of Education. (2014). Koulutuksen tilastollinen vuosikirja 2014. Helsinki:
Finnish National Board of Education. Retrieved from http://www.oph.fi/download/163331_koulutuksen_
tilastollinen_vuosikirja_2014.pdf. (10.06.2020)
Goldberg, A., Casenhiser, D. (2008). Construction learning and second language acquisition. In: P. Robinson &
N. C. Ellis (eds.), Handbook of cognitive linguistics and second language acquisition (pp. 197–215). New
York: Routledge.
Goldschneider, J., DeKeyser, R. (2001). Explaining the “natural” order of L2 morpheme acquisition in English: A meta-
analysis of multiple determinants. Language Learning 51, 1–50. DOI: https://doi.org/10.1111/1467-9922.00147
Hakulinen, A. et al. (2004). Iso suomen kelioppi. Helsinki: The Finnish Literature Society.
Hammarberg, B., Viberg, Å. (1979). Platshållartvånget, ett syntaktiskt problem i svenskan för invandrare.
Stockholm University, Institution of Linguistics.
Hammarberg, B. (2008). Konstruktioner som produkt och process – en studie av hur L1- och L2-talare utnyttjar
“det är”. Nordisk tidsskrift for andraspråksforskning 3, 79–107.
Harley, B. (1993). Instructional strategies and SLA in early French immersion. Studies in Second Language Acquisi-
tion 15, 245–260. DOI: https://doi.org/10.1017/s0272263100011980
Harley, B. (1998). The role of focus-on-form tasks in promoting child L2 acquisition. In: C. Doughty & J. Williams
(eds.), Focus on form in classroom second language acquisition (pp. 156–174). Cambridge: Cambridge
University Press.
Holmes, P., Hinchcliffe, I. (1994). Swedish: A comprehensive grammar (2nd ed.). New York: Routledge.
Housen, A., Simoens, H. (2016). Introduction: Cognitive perspectives on difficulty and complexity in L2
acquisition. Studies in Second Language Acquisition, 38, 163–175. DOI: https://doi.org/10.1017/s0272263116000176
Hultman, T., Westman, M. (1977). Gymnasistsvenska. Lund: Liber Läromedel.
Hyltenstam K. (1992). Non-native features of near-native speakers. On the ultimate attainment of childhood L2
learners. In: R. Harris (ed.), Cognitive processing in bilinguals (pp. 351–368). Elsevier Science Publishers.

Folia Scandinavica Posnaniensia 30 (2021)
Nyqvist: Subordinate questions in Swedish by 12- and 15-year-old Finnish immersion students
25
Hyltenstam, K., Lindberg, I. (1983). Invandrares svenska. En kritisk genomgång av materialet i projektet svenska
för invandrare (Josefson 1979), särskilt med avseende på dess vidare användningsmöjligheter. In: B. Hammar-
berg, (ed.), Studium av ett invandrarsvenskt språkmaterial (pp. 5–51). SSM Report 9, Stockholm University.
Håkansson, C. et al. (2019). Typfall och mönsterigenkänning – konstruktionsbaserad andraspråksundervisning
i praktiken. In: M. Bianchi et al. (eds.), Svenskans beskrivning 36 (pp. 107–117). Uppsala: Uppsala University.
Jörgensen, N. (1978). Underordnade satser och fraser i talad svenska: funktion och byggnad. Lund: Walter Ekstrands.
Korhonen, R. (2009). Että ja letkeät lainat. Kielikello 3/2009. Retrieved from https://www.kielikello.fi/-/etta-ja-
letkeat-lainat (10.06.2020).
Källström, R. (2000). Man vet inte riktit va e detta. In: A. G. Grönberg et al. (eds.), Sett och hört – en vänskrift tillägnad
Kerstin Nordenstam på 65-årsdagen (pp. 201–211). Gothemburg University.
Lahtinen, S., Toropainen, O. (2015). Användningen av frågor i en L2-korpus bestående av finska och svenska
inlärartexter. NordAnd 10, 81–108.
Larson-Hall, J. (2016). A Guide to doing statistics in second language research using SPSS. New York: Routledge.
Lyster, R. (2007). Learning and teaching languages through content: A counterbalanced approach. Amsterdam:
John Benjamins.
Ó Duibhir, P. (2009). The spoken Irish of sixth-class pupils in Irish immersion schools. Dublin: University of Dublin,
Trinity College, Centre for Language and Communication Studies.
Nakata, T., Suzuki, Y. (2019). Mixing grammar exercises facilitates long-term retention: Effects of blocking, interleaving,
and increasing practice. The Modern Language Journal, 103, 629–647. DOI: https://doi.org/10.1111/modl.12581
Nistov, I. et al. (2018). Bruksbaserte tillnærminger till andrespråkslærning. In: A-K. H. Gujord & G. Tveit Randen
(eds.), Norsk som andrespråk – perspektiver på læring og utvikling. Oslo: Cappelen Damm Akademisk.
Nyqvist, E-L. (2020). Interrogativa bisatser hos 12- och 15-åriga språkbadselever. En jämförelse med traditionell
undervisning. In: C. Wide et al. (eds.), Svenskan i Finland 18 (pp. 164–180). Turku: University of Turku.
Philipsson, A. (2007). Interrogative clauses and verb morphology in L2 Swedish. Theoretical interpretations of
grammatical development and effects of different elicitation techniques. Stockholm: Stockholm University.
Prentice, J. et al. (2016). Bortom ordklasser och satsdelar: Konstruktionsgrammatik i klassrummet. In: A. W. Gustafsson
et al. (eds.), Svenskans beskrivning 34 (pp. 385–397). Lund: Lund University.
Rahkonen, M., Håkansson, G. (2008). Production of written L2-Swedish – processability or input frequencies? In:
J-U. Kessler (ed.), Processability approaches to second language development and second language learning
(pp. 135–161). Cambridge Scholars Publishing.
Sullivan, G. M., Feinn R. (2012). Using effect size – or why the P value is not enough. Journal of Graduate Medical
Education 4, 279–282. DOI: https://doi.org/10.4300/JGME-D-12-00156.1
Teleman, U. et al. (1999a). Svenska Akademiens grammatik Vol. 1. Stockholm: Nordstedts Ordbok.
Teleman, U. et al. (1999b). Svenska Akademiens grammatik Vol. 4. Stockholm: Nordstedts Ordbok.
Viberg Å. (1990). Bisatser i inlärarsperspektiv. In: G. Tingbjörn (ed.), Andra symposiet om svenska som andra-
språk i Göteborg 1989 (pp. 388–362). Stockholm: Scriptor.
Eeva-Liisa Nyqvist
University of Helsinki
Finnish, Finno-Ugrian and Scandinavian studies
PB 24
00014 Helsingin yliopisto
Finland
eeva-liisa.nyqvist@helsinki.fi