Abstract
The timing of participants' contributions in conversational turn-taking is well organized, with usually very short gaps between turns [1, 2]. How such timing is possible was recently described as a psycholinguistic puzzle, since comprehension of the incoming turn and planning of the next turn must run in parallel in next
speakers' minds to achieve short gaps between turns
[3]. Psycholinguistic studies have shown that incoming linguistic material interferes with
production planning in monologic tasks [4], giving
rise to the question of what the magnitude of
this interference is in conversational situations, given the observed timing of turn-taking.
Planning of the next turn can usually only start after the previous turn's message was
understood, which is often already possible before the end of the previous turn. If more
uninformative (i.e. reaction-irrelevant) material follo
ws after this point, one hypothesis is that it
does not interfere as severely with planning as was observed in monologic tasks. Hence,
production planning could start without interfer
ence as early as the
incoming message is
understood. Alternatively, the interference observed in monologic tasks could also hold in
conversational situations, with uninformative turn-final material giving next speakers more time
to plan their response than it costs them. In this case, next speakers should still gain some time
for planning while further, uninformative material is coming in. Contrary to both these
hypotheses, production planning could be
delayed by interfering incoming speech.
This paper tests these hypotheses, using a no
vel task-oriented dialogue paradigm with a
confederate and pre-recorded critical sentences in German, combining ecological validity with
experimental control. The looking behavior and vo
cal response latencies of 38 participants were
analyzed. Participants saw 96 critical items with 3 to 5 natural objects on a screen, some of
which were named by the confederate. They then had to identify and name the objects that had
not been named. The following syntactic structures of German main clauses were exploited to
manipulate the informativity (task-redundant sentence final verb form or not) and predictability
(ambiguous main verb/auxiliary or unambiguous
main verb/modal verb in second position) of
the critical sentences' turn ends to investigate their effect on the timing of participants' language
comprehension and planning:
Ich habe
(I have)
+ object list.;
Ich habe
(I have)
+ object list +
besorgt
(gotten).;
Ich sehe
(I see)
+ object list.;
Ich kann
(I can)
+ object list +
besorgen
(get).
Gaze direction data was analyzed using growth-curve analysis. In all conditions,
proportions of looks towards the objects that were named first by the participants increased as
soon as the last object of the confederate's tu
rn was recognizable. However, these proportions
increased faster in conditions without a task-redun
dant verb form at the turn end. Yet, in these
conditions, participants looked significantly longer at the objects they would name before they
came up with their labels than in conditions without redundan
t, sentence-final verbs. These
findings were not modulated by the predictability of the uninformative sentence-final verb form.
The study's results suggest (1) that planning of one's turn is launched as early as
possible, i.e. as soon as the previous turn's
message was understood, i
rrespective of further
uninformative material, and (2) th
at planning during comprehension of further incoming speech
is less efficient than during silence, even with task-redundant material, but still shortens the gap
between turns. This study thus shows that even though comprehension also interferes with
production in a conversational setting, uninform
ative material before the turn end buys next
speakers time to be able to start to articulate thei
r next turn in a timely fashion. This effect could
be strategically exploited by interlocutors to manage smooth turn-timing in conversation.