A commentary on Congruency between word position and meaning is caused by task-induced
spatial attention by Pecher, D., Van Dantzig, S., Boot, I., Zanolie, K., and Huber,
D. E. (2010). Front. Psychol. 1:30. doi: 10.3389/fpsyg.2010.00030
Pecher et al. (2010) presented targets (e.g., helicopter, submarine) up and down on
a computer screen. Participants were either asked to indicate whether these objects
were typically found in the ocean, or typically found in the sky. The authors examined
whether congruency effects between the vertical position of words and their meaning
were best accounted for by mental simulations or by polarity benefits (default asymmetries
in the way people process dimensions). I believe their conclusion that polarity benefits
cannot account for the interaction in reaction times between the meaning and the position
of words is at best premature. Moreover, instead of explaining language understanding
in terms of either simulation processes or linguistic input, a more fruitful approach
might be to examine when meaning emerges from simulation processes, and when meaning
is extracted from linguistic information (see Louwerse and Jeuniaux, 2010).
Pecher and colleagues state that “Certain dimensions are always coded in the same
way.” (p. 2). Although it is true that people are by default faster to process above
relationships compared to below relationships (Clark, 1969), studies have shown that
this default processing benefit can easily be reversed, depending on the nature of
the task and characteristics of the stimuli. For example, Banks et al. (1975) asked
participants to indicate which of two dots was located above (vs. below) the other.
These dots were referred to as balloons, or as yo-yos. Participants were faster to
indicate which of the two dots was the higher balloon, whereas they were faster to
judge which of the two dots was the lower yo-yo. The crossover pattern observed in
this study supports the idea that the default polarity codes of up as +polar and down
as −polar can be reversed by framing the task in a different way. Proctor and Cho
(2006, p. 428) similarly conclude: “In two-choice tasks with orthogonal stimulus and
response sets, the responses are coded relative to multiple frames of reference.”
A similar reframing occurs in the study performed by Pecher and colleagues. When thinking
of skies, as in thinking of balloons, high skies are the default. Therefore, up becomes
the default endpoint of the vertical dimension, and words presented up on the screen
are responded to more quickly. When thinking of oceans, as when thinking of yo-yos,
the deep ocean is the default. Therefore, down becomes the default endpoint of the
vertical dimension, and words presented down on the screen are responded to more quickly.
This means that the interaction between task and position reported by Pecher and colleagues
is an a priori prediction of (and not evidence against) a polarity explanation.
The second polarity effect Pecher and colleagues investigate is based on the assumption
that right-hand key-presses receive a reaction time benefit because right is +polar.
This prediction does not immediately follow from previous theoretical or empirical
work. One study has revealed that responses are faster for arrows pointing toward
the right than to the left, but importantly, these effects are eliminated when arrows
are substituted by words (Olson and Laxar, 1973), and seem to be limited to explicit
spatial categorizations where right key-presses are mapped onto stimuli presented
up on the screen (Weeks and Proctor, 1990; Cho and Proctor, 2002). Furthermore, several
researchers explicitly note that left-to-right space is symmetric, and should by default
not reveal polarity differences (Clark, 1973; Seymour, 1974; Làdavas, 1988). Reaction
time benefits for right key-presses are not ubiquitous in the literature, nor is such
a main effect present in the study performed by Pecher and colleagues. Therefore,
it is doubtful whether right key-presses should be coded as +polar.
Overall, the results of Pecher and colleagues provide little evidence against a polarity
explanation for congruency effects between word meaning and vertical position. Given
that previous studies that have investigated the vertical representation of concepts
often show that +polar words (e.g., powerful, moral) are categorized faster than −polar
words (i.e., powerless, immoral), and words presented up are categorized faster than
words presented down, it seems premature to exclude polarity effects from influencing
categorization times(e.g., Schubert, 2005; Meier et al., 2007). Indeed, as predicted
by the polarity correspondence principle (Proctor and Cho, 2006), these experiments
typically do not reveal differences in the categorization times for –polar words presented
up or down. At the same time, studies have revealed that words can direct attention
upward or downward (Richardson et al., 2003; Bergen et al., 2007; Van Dantzig, 2009).
Instead of attributing effects to either simulation processes or polarity effects,
several researchers have recently proposed models of conceptual processing that rely
on both semantic and embodied information (Barsalou et al., 2008; Andrews et al.,
2009; Louwerse, in press).
Such an interplay between linguistic and embodied sources of meaning might be able
to provide a more integrative model of conceptual thought, especially for more abstract
concepts. As an example, considerLakoff and Johnson's (1980) Conceptual Metaphor Theory
which describes how abstract concepts (e.g., power, valence, morality) are mapped
onto concrete dimensions (e.g., verticality, size, brightness). These mappings are
argued to result from experiential co-occurrence of the abstract and concrete domains,
such as drooped postures when we feel down. This reasoning cannot explain the fact
that practically all conceptual metaphors map abstract + polar concepts onto concrete + polar
endpoints of dimensions (e.g., happy = up, warmth = affection, powerful = large) and
vice versa (see Krzeszowski, 1997). Perhaps linguistic structures such as polarity
differences constrain the specific mappings that emerge between abstract and concrete
dimensions, thereby shaping how concepts are grounded.
Activated word associations, statistical regularities in language, and linguistic
structures can guide simulation processes (for examples, see Barsalou et al., 2008;
Louwerse and Jeuniaux, 2010), and should be taken into account when aiming to explain
language comprehension. Linguistic and concrete information both contribute to conceptual
thought; the challenge for future research lies in specifying how meaning emerges
from their dynamic interaction.