Languages use different word orders (the order of the Subject, Object and Verb)
to organize and convey information. The distribution of word orders
documented in the world is uneven, favoring SOV and SVO (Dryer, 2013),
which may provide a window into biases shaping language. The visual-manual
modality has recently been taken up to study basic word order in the absence of
a system of language conventions. In silent gesture experiments, in which
hearing participants improvise ways of communicating motion events using
only their hands and no speech, SOV order is dominant for speakers of various
native languages (Goldin-Meadow et al., 2008). However, participants may
switch to SVO word order depending on the semantic properties of the event
(Schouwstra & de Swart, 2014; Gibson et al., 2013), or in the presence of a
vocabulary of conventional signs (Marno et al., 2015). As these findings appear
to be robust across different languages (native SOV and SVO speakers in the
aforementioned studies), silent gesture is increasingly used as a way to uncover
the cognitive biases playing a role in situations of emerging language.
Thus far, these experiments have been carried out using the manual modality
alone, leaving open the question whether the findings generalize to other
modalities.
1 Are the word order patterns observed in silent gesture essentially a
consequence of the fact that participants improvise in the absence of linguistic
rules? Or are they (partly) the result of modality specific production constraints
(such as suggested in e.g. Hall et al. 2013)? An answer to this question will have implications for the generality of the evolutionary claims to be made on the
basis of silent gesture results.
Here we use a vocal analog of the silent gesture paradigm to test whether biases
previously observed in silent gesture generalize to the vocal modality. We do
this by testing the word order preferences for extensional events (in which both
participants in the action are concrete, e.g., ‘the pirate throws the guitar’) vs
intensional events (in which the Patient is unspecific or dependent on the action,
e.g., ‘the pirate thinks of the guitar’). Schouwstra & de Swart (2014) showed a
preference for SOV for extensional events and SVO for intensional events in
silent gesture. Here, we tested participants in the vocal modality, to see if this
meaning-dependent word order pattern replicates outside the manual modality.
In experiment 1, we asked native SVO participants (N=20) to use non-linguistic
vocalizations to describe pictures while sitting on their hands. The stimuli were
line drawings that consisted of 32 extensional events (e.g. ‘a robot drops a drill’)
and 32 intensional events (e.g. ‘a diver loves a car’) that were selected to be
relatively easy to describe using non-linguistic vocalizations. In each
description, the elements were coded as subject, object or verb. For instance,
‘evil-laugh’ (denoting a witch) was coded as Subject; ‘boom’ (denoting
dropping of an object) was coded as Verb. Sequences with repeated consecutive
constituents were recoded as if having only one iteration of each constituent
(SVVO was recoded as SVO). Sequences with non-consecutive repetitions
remained as they were (VSVO remained VSVO). Word orders other than SVO
or SOV (14% of the total data set) were excluded from statistical analysis.
The findings confirm that constituent order in improvised vocalizations is
influenced by the semantic properties of the event: a greater proportion of SOV
descriptions are given for extensional events than intensional events, and a
greater proportion of SVO descriptions are given for intensional events than
extensional events (see Figure 1). Event type predicted the proportion of SVO
produced (b = 1.965, SE = .737, p < .01), mirroring what has been found in
silent gesture.
Strikingly, the results from the vocal modality follow what has been found in the
manual modality, namely that semantic properties of the event significantly
affect constituent order. However, the results of experiment 1 differ from
previous findings, where the preference for SOV for extensional events is much
greater, and SVO only becomes dominant for the intensional events (Schouwstra
& de Swart, 2014). A possible explanation for the difference in results is that in
the present study the vocal channel is used for both language and the improvised
vocalizations, and thus may be more susceptible to interference from the native
language of participants (SVO in the present study). Despite this possible
influence, the vocalizations of participants in the present study were conditioned
by meaning, and we thus replicated the basic finding of Schouwstra & de Swart
(2014).
In experiment 2, we build on work by Marno et al. (2015) who show an increase
in SVO order in silent gesture if participants (native SOV and SVO) are trained
on individual lexical items (e.g. ‘a girl’) prior to complex meanings, akin to the
extensional events in the present study (e.g. ‘a girl throwing a fish’; Marno et
al., 2015). We asked native SVO participants (N=20) to produce vocalizations
for individual items first, before going on to complete scenes. Using the same
coding scheme as for experiment 1, word orders other than SVO or SOV (9% of
the total data set) were excluded from statistical analysis.
Again, our results in the vocal modality reflect those in the manual modality: a
higher proportion of SVO is produced when participants first describe individual
items before continuing to full scenes. In experiment 2, SVO accounts for a
majority of extensional events (87%) and intensional events (99%) while SOV
accounts for a minority of extensional events (13%) and intensional events
(1%). Overall, more SVO is produced for both event types in experiment 2 than
in experiment 1; in a model combining experiment 1 and 2, we find that
experiment (b = 3.998, SE = 1.465, p < .01) is a stronger predictor of producing SVO than verb type (b = 1.693, SE = .714, p < .05); thus, the high incidence of
SVO in experiment 2 is more strongly predicted by the access to the lexicon
than the meaning of the event. Following the results of Marno et al. (2015) in
the manual modality, the lexicon triggers more SVO, showing similar results in
the vocal and manual modalities.
The silent gesture paradigm has been used to make claims about features of
language emergence, in the manual and vocal modalities, but could these
findings be specific to the manual modality? In the present study, we focused on
2 results from the silent gesture paradigm, and replicated both of the main
findings in the vocal modality. In the vocal modality, more SVO is produced
than in the manual modality, which may be attributed to interference from the
participant’s native language, but the overall finding is the same: word order is
conditioned by event type in improvisation. These findings provide the first
evidence that gestural and vocal improvisation yield similar results for basic
word order, demonstrating that the effects observed are modality independent.
| Period | 16 Apr 2018 |
|---|
| Event title | Modality Matters |
|---|
| Event type | Workshop |
|---|
| Location | Toruń, Poland |
|---|
| Degree of Recognition | International |
|---|