SeanTrott HonorsThesis .pdf
Original filename: SeanTrott_HonorsThesis.pdf
This PDF 1.5 document has been generated by TeX / pdfTeX-1.40.14, and has been sent on pdf-archive.com on 13/02/2016 at 00:38, from IP address 192.150.x.x.
The current document download page has been viewed 358 times.
File size: 6.4 MB (11 pages).
Privacy: public file
Download original PDF file
Within-Language Variation in Motion Event Categorization
Sean Trott (email@example.com)
Cognitive Science Program, University of California, Berkeley, CA 94704 USA
There is a key grammatical difference in the way that different languages express motion events. Satellite-framed languages, such as English, encode the Manner of motion in the
main verb, and the Path of motion in a particle or prepositional phrase, such as: ”He ran into the room.” Verb-framed
languages, such as Spanish, encode the Path of motion in the
main verb, and the Manner of motion in an optional auxiliary
phrase, such as: ”Entro la sala (corriendo)”. One prediction
is that this typological difference leads to differences in manner processing across languages and speakers, such as varying
degrees of expertise in encoding manner. Past research has
uncovered considerable variation in non-linguistic cognitive
tasks concerning motion event conceptualization, both within
and between languages; some of these results were consistent with the predictions of the typology, and others were not.
We investigated whether the within-language variation could
be explained by an individual speaker’s expertise in manner.
We explored whether an English speaker’s manner expertise
predicted their performance on a motion triads task by measuring linguistic and non-linguistic expertise. Ultimately, we
found no significant effect of linguistic expertise on motion
event categorization, but we did find a significant effect of nonlinguistic expertise, or how easily a participant identifies manner and path, on motion event categorization. Conflation of
manner with the fact of motion significantly predicted a preference for manner on a motion categorization task.
Keywords: motion event typology, manner expertise, manner
The Sapir-Whorf hypothesis claims that the structure of a language plays a role in the perception and concept formation
in the minds of its speakers (Whorf, 1941). Different languages represent reality in different ways; linguistic relativity
predicts that these typological differences between languages
lead to differences in non-linguistic cognition. Linguistic relativity has primarily been investigated by examining whether
typological differences in language lead to corresponding differences in the cognition of speakers of these languages. Typological divisions in domains such as color cognition (Kay
& Kempton, 1984; Regier et al, 2007; Kay & Regier, 2006)
and spatial conceptualization (Majid et al, 2004; Hespos &
Spelke, 2004) have been argued to correspond with differences in the non-linguistic behaviors of speakers. One lesserstudied domain of inquiry has investigated variation in the
expression of motion events.
Motion events involve the motion of a particular object
through space. There are four crucial components of a motion event: Figure, Ground, Path, and Motion (Talmy, 1985).
The Figure is the moving object. The Ground is the space
that the Figure moves across or along. The Path is the physical path that the Figure takes along the Ground. The Motion component refers to the fact or presence of motion in the
event itself. Additionally, the Manner of a motion event specifies the means or method by which the Figure moves (Talmy,
Languages differ in how they encode translative motion
events, motions in which the Figure crosses a salient boundary; different languages use different grammatical constructions for expressing manner and path (Talmy, 1985; cf.
Beavers et al, 2010). Two broad classes of languages can
be identified according to the constructions that they use for
identifying translative motion events. In verb-framed languages, such as Spanish, the path of motion is expressed in
the main verb, and the manner is expressed in an optional,
auxiliary phrase (2). In satellite-framed languages, such as
English, the manner of motion is expressed in the main verb,
and the path is encoded in a satellite to the verb.
(1) Patrick ran up the hill.
(2) Patrick subio por la colina (corriendo).
Patrick ascended the hill (by running).
In (1), ”ran” expresses the manner of motion, and ”up the
hill” expresses the path. In (2), ”subio” expresses the path of
motion, ”ascend”, and ”corriendo” expresses the manner, by
running’. Although it is possible to express path in the main
verb in English, using Latinate verbs like ”exit” or ”ascend”,
this construction is much less typical.
Cross Linguistic Variation in Manner Codability
The primary difference between verb-framed and satelliteframed languages lies in the codability of manner in each
language. While each language must encode path information to convey translative motion events, the languages differ
in the ease with which they encode manner information. This
has been referred to as a difference in the manner codability, or ease with which manner is expressed, across languages
The typological distinction in the expression of manner and
path has been argued to lead to differences in speakers’ fluency in manner, or ability to easily express and encode information about manner (Slobin, 2005). Manner is a more
salient component of motion events to satellite-framed speakers because satellite-framed languages, such as English, have
a higher codability of manner. This means that manner is
more easily encoded in descriptions of motion events (Slobin,
Satellite-framed languages contain grammatical constructions that make it very simple to include information about
manner; the manner of a motion event is easily expressed
as the main verb of the sentence, as in (1) above. In verbframed languages, speakers must append an unnecessary auxiliary phrase to express manner, such as in (2). In other
words, there is a lower processing cost of including manner
in satellite-framed languages than in verb-framed languages.
Conversely, there is a heavier processing cost of adding manner information in verb-framed languages, because the speakers must add a degree of syntactic complexity to the sentence.
For speakers, it is easy to encode manner linguistically, because adding this information comes at no additional processing cost; manner can simply be expressed in the main verb.
This encourages speakers to attend to manner when they are
processing a motion event, and also leads to the creation of
new manner verbs, which can convey more specific information to the listener (Slobin, 2006). The regularity of manner
expression leads listeners to expect a certain degree of information about manner, which then encourages speakers to include more.
Ultimately, this cycle leads to the creation of more manner
verbs, which is why English is such a manner-rich language
(Slobin, 2006). Consequently, the cycle perpetuates itself:
the high codability of manner leads to more attention paid to
manner, which encourages manner expertise and fluency in
manner processing; this results in more informational content and semantic space devoted to manner encoding, which
leads back to higher manner codability in satellite-framed
It has been argued that these differences in codability of
manner affect conceptualizations of motion events (Slobin,
2005). This theory predicts that speakers of languages with
higher manner expertise attend to more information about
manner, and will differ from speakers of languages with
lower manner expertise in their categorization of motion
events. Specifically, speakers of languages that encourage
higher manner expertise will demonstrate a greater attention
to manner than speakers of verb-framed languages, which
have lower manner codability.
Additionally, Slobin has argued that an increased codability of manner leads speakers to conflate information about
manner with the fact of motion, because the verb expresses
both the motion and the manner of motion (Slobin, 2005).
One prediction which can be drawn from this is that these
speakers view manner as inseparable from the fact of motion,
which might make it more difficult for them to perceive manner as a separate component of a motion event; thus, they
would find it more difficult to ignore irrelevant information
about manner when analyzing other components of a motion
event, such as path.
In this paper, I will attempt to address the apparent inconsistencies in the existing literature, as well as investigate
Slobins claims about expertise and conflation. I will focus
on the predictions that these claims entail: specifically, that
speakers with higher linguistic expertise in manner attend to
more information about manner, and that speakers of languages that conflate manner with the fact of motion will find
it difficult to construe manner as separate from the fact of motion.
Differences in Linguistic Expression of Motion
Cross-linguistic differences in codability of manner are predicted to produce significant differences between languages
in motion event conceptualization. Specifically, speakers of
languages that encourage greater manner expertise should attend to more information about manner, than speakers of languages that do not.
Much empirical research has tested this prediction in
purely linguistic studies. When asked to report mental imagery in response to linguistic stimuli, speakers of satelliteframed languages report more mental imagery for manner
(Slobin, 2005) and tend towards ”greater specification of
manner” than speakers of verb-framed languages (Berman
& Slobin, 1994). When asked to describe scenes of motion
events, satellite-framed speakers are more likely to include
manner information in descriptions of motion events (Slobin,
1991); additionally, satellite-framed speakers can generate
more manner verbs on demand than verb-framed speakers
(demonstrating greater linguistic expertise), and they tend to
add manner information when translating motion events from
a verb-framed language (Slobin, 2005).
Non-Linguistic Differences in Motion Cognition
One problem with purely linguistic studies is that they merely
indicate a difference that has already been observed in the typology. If the goal of the research is uncovering the underlying mental representations of motion events across different
languages, linguistic studies may not reveal any new information, other than that linguistic behavior follows the pattern of
the existing typology. Thus, many researchers have opted to
study this phenomenon using non-linguistic tasks.
With few exceptions, most past research investigating
cross-linguistic differences in non-linguistic cognition has
used judgments of similarity as a metric for non-linguistic differences in motion event conceptualization. The most common methodology used has been the forced choice paradigm,
also referred to as a ”triadic similarity” task, or simply a motion triads task (Papafragou et al, 2002; Gennari et al, 2002;
Bohnemeyer et al, 2006). In the case of motion events, this
triad task involves the presentation of a target motion event,
followed by two variants. The variants differed from the target event along exactly one dimension - manner or path. Participants are asked to select the variant they find most similar
to the target.
(a) Target: a man running into a room.
(b) Variant1: a man running out of the room (same manner).
(c) Variant2: a man walking into a room (same path).
The reasoning behind the triads task is as follows: if there
is a non-linguistic difference in motion event conceptualization between speakers of different languages, and that difference is modulated by differences in linguistic structure,
speakers of these different languages should differ in the criteria they use in judging similarities of motion events. One
Figure 1: Triads: Run In (Target), Run Out (Same-manner),
Walk In (Same-path
common prediction is that satellite-framed speakers should
prefer the same-manner variant more often than verb-framed
speakers (c.f. Kersten et al, 2010).
The use of the triad task to investigate non-linguistic differences in motion event conceptualization has found mixed
results. Papafragou (2002) found no effect of language on
categorization in either verbal encoding or nonverbal conditions; the only difference was in the verbal descriptions of the
pictures participants viewed, but this did not correlate with
Gennari et al (2002) found similar results. English and
Spanish speakers were separated into three conditions: verbal
encoding (labeling), free encoding, and shadowing (a verbal
working memory interference task). In the free encoding and
shadowing conditions, there was no effect of language; in the
verbal encoding condition, Spanish speakers showed a slight
preference for path, and English speakers behaved at random.
Additionally, in a memory task performed later, there was
no effect of language in any of the conditions, leading researchers to conclude that the effect of language was primarily as a ”cognitive strategy” (Gennari et al, 2002). In other
words, when language was used as a tool to solve the triad
task, the categorization of triads differed across languages;
otherwise, in absence of a prompt to use language, the categorization of motion events did not differ across speakers of
typologically distinct languages. Furthermore, this appears
to only be an ”online” effect: there was no effect of language
on memory, even with verbal encoding, suggesting that the
stored representations of the motion events were fundamentally the same.
The triad task was also used in an experiment involving
speakers of seventeen genetically and typologically diverse
languages (Bohnemeyer et al, 2006). While there were differences across languages in the categorization of events, the
speakers of these languages did not pattern in a way consistent with their typology; there was no simple categorical verbframed and satellite-framed distinction, in terms of ”samemanner” choices. Some verb-framed speakers chose significantly more same-manner choices than the satellite-framed
speakers, and some chose far less. Additionally, there was
a high degree of intra-typological variation in the similarity
judgments (Bohnemeyer et al, 2006).
Papafragou et al (2008) also investigated cross-linguistic
differences using an eye-tracking device, rather than the triad
task. Participants viewed animated motion events, and were
separated into verbal and nonverbal conditions. Cognitive attention to particular components of the motion events was
measured by the eye movements of the participants. In the
verbal task, English participants allocated more attention to
the manner components of the motion events. In the nonverbal task, participants allocated attention similarly regardless of language during the motion event, and focused on the
event not typically encoded in their language after the motion
ended. Papafragou et al (2008) concluded that these results
indicated that allocation of attention to motion events is only
affected by language when participants recruit language to
perform the task.
One study did find language-congruent effects on the triad
tasks (Finkbeiner et al, 2002); in a study involving English (satellite-framed), Spanish (verb-framed), and Japanese
(verb-framed) speakers, English speakers showed a significantly greater preference for same-manner variants than the
Spanish and Japanese speakers, following the predictions of
the typology. However, in an altered condition, in which
the target and variants were presented together, there was no
longer a significant effect of language.
Significant effects of language were also found in a study
involving motion triads and eye tracking, which measured the
eye movements of participants (Soroli, 2011). Researchers
added ”filler stimuli” to distract participants from the aim of
the task. In both labeling and non-labeling conditions, English speakers selected same-manner variants more often than
French speakers. Despite this difference, the eye tracking
data revealed a cross-linguistic difference only in the labeling
condition. Soroli (2011) concluded that language has a measurable effect on non-linguistic cognition, such as categorical
judgments (triad choices) and attention allocation (tracking
of eye movements).
A study measuring looking time, or gaze, during a triad
task found that satellite-framed speakers looked at the samemanner variant for longer than verb-framed speakers, while
verb-framed speakers looked at the path variant for longer
(Hohenstein, 2005). This suggests that satellite-framed languages encourage extra attention to manner components of
motion events, while path-framed languages encourage extra
attention to path.
Significant effects of language on motion event categorization were also found in a study involving different methodology (Kersten et al, 2010). English and Spanish speakers were
given a ”supervised classification task”, in which they had to
classify moving objects and events into four novel linguistic categories. In one condition, these categories were based
on manner of movements; in the other, they were based on
path. A strong effect of language was found: English speakers learned the manner-based categories significantly faster
than Spanish speakers (Kersten et al, 2010). There was no
effect of language in the path-based condition. Even when
the experiment was performed outside of the word-learning
context, English speakers learned the manner categories with
greater speed and accuracy than Spanish speakers, suggesting
that the conceptual difference exists beyond the purely linguistic level (Kersten et al, 2010). These results are substantially different from the majority of non-linguistic research
on motion event conceptualization. Most notably, Kersten
et al’s results are aligned with Slobin’s predictions; due to
enhanced linguistic expertise, English speakers demonstrated
greater cognitive expertise with manner categorization than
Spanish speakers, but not on path (Kersten et al, 2010).
Despite the wide array of studies, significant questions
about the effect of language on motion event conceptualization remain. Previous research has uncovered high degrees
of intra-typological variation (Bohnemeyer et al, 2010); in
a study comparing similarity judgments between English and
Spanish speakers, some English speakers demonstrated a very
strong preference for manner, while others demonstrated a
strong preference for path (Goss, 2014). This variation is
not aligned with the original predictions of Talmy’s typology,
but there may be an alternative explanation for differences
in motion event conceptualization, both between and within
Slobin’s theory of codability and manner expertise predicts
significant between-language differences. Slobin suggests
that languages be viewed on a cline of manner salience
(Slobin, 2004). The salience of manner in a particular language influences the facility with which a speaker processes
manner; satellite-framed languages contain constructions that
make the encoding of manner easy and relatively typical,
which encourages greater expertise in manner. Slobin argues
that this leads to cross-linguistic differences in motion event
If cross-linguistic differences in manner expertise predict
differences in motion event conceptualization, we propose
that within-language differences in manner expertise also predict differences in motion event categorization across speakers of the language. This would provide an explanation for the
high degree of within-language variation (Goss, 2014) and
intra-typological variation (Bohnemeyer et al, 2006) found in
Previously, no researchers have investigated whether
within-language differences in manner expertise accurately
predict performance on non-linguistic tasks. However, the
same logic should apply. Past research has demonstrated that
background knowledge or expertise in a particular subject domain affects object classification and categorization (Lin &
Murphy, 1997). If Slobin is correct that expertise in manner affects how much manner information a speaker encodes,
within-language differences in expertise should also have a
measurable effect on the amount of manner information encoded by speakers of the same language. Based on Slobin’s
reasoning, an English speaker who is very linguistically fluent
with manner will attend to more information about the manner of motion events. Thus, their conceptualization of a motion event will be markedly different from an English speaker
with a lower manner expertise; specifically, they will highlight more aspects about the manner of motion. Furthermore,
if higher expertise leads to an increased salience of manner,
it might also be the case that it leads to a higher tendency to
conflate manner with the fact of motion.
If Slobin’s theory of manner codability and expertise is
correct, there will be an observable difference in how English speakers with greater manner expertise categorize motion events, compared with the behavior of English speakers
with lower manner expertise. In particular, higher expertise
will predict a greater cognitive emphasis to manner as a component of motion; specifically, English speakers with higher
expertise will choose more same-manner variants as more
similar to the target motion event. Within-language variation in same-manner judgments should be predicted by differences in manner expertise.
Additionally, if speakers with higher expertise tend to conflate manner with the fact of motion more, as Slobin argues
is encouraged by satellite-framed languages (Slobin, 2005),
they will have a more difficult time conceptualizing a motion event with manner as a separate component from the
fact of motion. Consequently, they will find it harder to ignore irrelevant manner information, such as when they process information about path. Thus, these speakers will experience more difficulty in making category judgments about
path. Ultimately, this means that these speakers should prefer
the same-manner variant in a triad task, because they process
information about manner similarities faster than path similarities.
If Slobin’s theory is incorrect, manner expertise will not
be correlated with motion event categorization, and speakers
with higher expertise will not choose more same-manner variants as more similar to the target. Additionally, there will be
no relationship between expertise and the conflation of manner with the fact of motion.
Slobin’s theory predicts that greater expertise in classifying
manner should result in a greater cognitive salience of manner overall. We investigated whether this prediction should
hold within languages as well as between them. We examined whether differences in the manner expertise of speakers
within a single language predicted their preference for samemanner variants in a triad task. Despite the mixed results
from past research utilizing triad tasks, some researchers have
found positive effects of language (Finkbeiner et al, 2002;
Soroli, 2011), and we hoped to explore the source of variation
among other studies (Bohnemeyer et al, 2006; Papafragou et
Expertise was operationalized in two ways: a linguistic
task, in which participants demonstrated linguistic fluency
with manner (mirroring Slobin, 2005), and a non-linguistic
task, in which we established how easily a participant processed similarities and differences in manners and paths. If
Slobin’s theory is correct, linguistic expertise, or a speaker’s
fluency with expressing manner, should be significantly correlated with non-linguistic expertise, which is a speaker’s
ability to process manner cognitively.
We measured non-linguistic motion event conceptualization using the triadic similarity task, in which participants
were asked to select one of two variants as more similar to
a target motion event.
Linguistic expertise in manner was measured by asking
participants to generate as many manner verbs as they could
in 30 seconds; participants were also asked to generate as
many path verbs as they could in 30 seconds. We predicted that participants who could generate more manner
verbs would choose more same-manner variants in the triad
task. Additionally, participants with a great difference between the amount of manner verbs and path verbs (participants with higher linguistic manner expertise in relation to
their path expertise) should choose more same-manner variants in the triad task; furthermore, linguistic manner expertise
should be correlated with non-linguistic manner expertise.
Non-linguistic manner expertise was measured by testing
participants’ reaction time in two forced choice discrimination tasks. In the manner task, participants were shown pictures of two motions and asked to decide whether they had the
same or different manner. In the path task, participants were
shown pictures of two motions and asked to decide whether
they had the same or different path. We assumed that faster
reaction times in the manner condition indicated a higher degree of manner expertise. Furthermore, the degree to which a
participant was faster in manner than path categorization was
also an indicator of expertise in manner; although a speaker
with higher manner expertise should attend faster to manner
components, higher expertise may lead to a more significant
conflation of manner and the fact of motion, making it more
difficult to attend solely to path and ignore manner. This
means they would respond significantly slower when asked
to categorize paths than when asked to categorize manners.
We predicted that participants with higher manner expertise would prefer the same-manner variant in the triad task.
Specifically, faster reaction time on the manner categorization
task should predict a higher number of same-manner judgments. Similarly, participants who generated more manner
verbs, or who produced more manner verbs in relation to path
verbs, should also choose more same-manner variants.
10 native English speakers participated in this study. For
compensation, participants received either course credit, or
$10 in cash. All participants were undergraduates at UC
Berkeley, with normal or corrected vision. Participants were
told the study was an investigation into how different sociocultural backgrounds affected similarity judgments, so they
were not initially aware of the study’s purpose.
The experiment was designed on PsychoPy2, an open-source
software application that allows the presentation of stimuli
and collection of data. We used the software to present pairs
of photographic stimuli and record key presses by the participants.
Triad Task For the triad task, we presented eight different
motion event triads. Each triad contained three pictures, so
there were 24 pictures total. The stimuli were constructed
so that the two variants differed from the target in either the
Manner or the Path. In all cases, each set of target and variants
shared the same Figure, as well as the same Ground. This is
a replication of the procedure used by Goss (2014), and a
near replication of Gennari et al (2002). For example, one set
(a) A man running into a room.
(b) A man running out of a room.
(c) A man walking into a room.
Figure 3: Triads: Step On (Target), Jump On (Same Path),
Step Off (Same Manner)
Figure 2: Same-Manner judgments predicted by Linguistic
and Non-Linguistic Expertise
Sentences (a) and (b) share the same Manner (running),
while sentences (a) and (c) share the same Path (into). We
recorded whether a participant chose the same-Manner variant, or the same-Path variant.
Decision Task There were 4 within-task conditions on the
decision task, generated from each triad in the motion triads
task. These pairings were constructed from combinations in
the triad task, to allow for the presentation of all possible
relationships between the photos: target and same-manner
variant (Manner-Match / Path-Mismatch), target and samepath variant (Manner-Mismatch / Path-Match), target and
target (Identical), manner and path variants (Manner No
Match, Path No Match). These pairings were designed to
assess how similarities and differences in manner and path
were analyzed with respect to each triad in the triads task.
Further, this pairing ensured that there were an equal number
of similar and different stimuli, to dissuade participants from
forming a bias in response for either the ”f” or ”j” keys.
1. Match: the two pictures share the same Manner (in section
M) or the same Path (in section P), and have different Paths
(in section M) or different Manners (in section P).
2. Mismatch: the two pictures share the same Path (in section
M) or the same Manner (in section P), and have different
Manners (in section M) or different Paths (in section P).
3. Identical: the two pictures share the same Manner and
Path (in both sections M and P).
4. No Match: the two pictures do no share Manner or Path
(in both sections M and P).
Figure 4: From top: Manner Match, Manner Mismatch, Manner No Match, Manner Identical
In each picture, the relative direction of motion (to the right
or left of the screen) for each comparison was controlled for.
For each of the 32 base stimuli, the photos were flipped, cre-
ating 4 versions: both photos flipped, neither photos flipped,
or only one of the photos flipped. Thus, because there were
32 base stimuli, there were 128 possible pairings total in both
the manner decision task and path decision task.
Figure 5: Two identical pictures with the same Path, reversed.
Additionally, 128 filler stimuli were created. The procedure for developing these stimuli was identical to the core
items above, except that the motions conveyed in each contained differences in actor and location. These were intended
to dissuade participants from forming strategies for responding that were based on features of the background or actors,
rather than features of the motion itself. Thus, although the
motion event was the same, the stimuli featured different locations, landmarks, and actors. Again, there were 8 base triads, paired in 4 ways (match, mismatch, no match, identical),
and presented in 4 different directions.
The experiment consisted of three tasks: a motion triads task
to measure how participants categorized motion events, a decision task to measure non-verbal expertise, and a word generation task to measure verbal expertise. Each of these tasks
was preceded by training to familiarize participants with response procedures. The decision task was further divided
into decisions about manner and decisions about path. The
order of these differences were counterbalanced across subjects. The full order of tasks for each subject was:
1a. Practice triad task.
1b. Motion triad task.
2a. Training on manner-path distinction.
2b. Pretest for decision task.
2c. Manner decision task.
2d. Path decision task.
3a. Training for verbal expertise task (colors, mammals).
3b. Verbal expertise task (manners, paths).
Triad Task Before the triads task, participants were given
a ”practice triad task”, in which non-motion stimuli were
presented. After a crosshair displayed for 1500ms, participants were presented with a ”target” image for 3000ms. After 500ms, they were presented with two variants, and were
instructed to choose the variant they found most similar to
the target. They were told to press ”f” for the variant on the
left, and ”j” for the variant on the right. The target displayed
a water bottle, one variant displayed a can of soda, and the
second variant displayed a roll of paper towels. The variants
displayed for a maximum of 5s and disappeared when the
participants made their selection.
The same procedure was used in the motion triads task,
except that the stimuli were motion events. The target depicted a particular motion event, and the two variants differed
from the target in exactly one dimension either manner or
path. Before the target was displayed, a centered crosshair
was presented for 1500ms. Participants then viewed a target
image depicting a particular motion event for 3000ms. After
500ms, they were presented with both variants, which displayed for a maximum of 5s. Again, they were instructed to
press ”f” to select the variant on the left, and ”j” to select the
variant on the right.
Non-Verbal Expertise Task After completing the triad
task, participants given a training task to familiarize them
with the goals of the decision task. This consisted of two
sections: a familiarization task and a practice task.
In preparation for the discrimination task, the first section
of the practice familiarized participants with the different between manner and path. Participants were given the instructions: ”There are two parts to a motion - manner and path.
”Manner” is how someone gets from one point to another, as
in what type of action they are using. ”Path” is where they
are going, as in what direction they are headed. Now you will
view a series of pictures, with information about their relationships. Press ”Space” to proceed to each new picture.”
Participants then viewed a series of 9 pairs of pictures of
equal size ([.9, 1.125]), presented side by side on the screen.
Above the pictures, text described the relationship between
the pictures, such as: ”These pictures have the same manner”,
or ”These pictures have the same path”. Each pair of pictures
displayed indefinitely until the participant understood the relationship, and pressed ”Space”.
Figure 6: ”These pictures have the same manner.”
After the familiarization task, participants were given a
practice task. The goal of the practice task was to ensure (a)
that they had accurately learned the distinction between manner and path, and (b) to familiarize them with the format of
the upcoming discrimination task. In the familiarization task,
participants were again presented with pairs of pictures, but
these had no text above them. They were instructed to press
”j” if the pictures shared the same manner of motion, and ”f”
if they did not share the same manner. In order to proceed
past the training task, the participant had to successfully answer ten in a row. If an answer was correct, text flashed above
the pictures saying ”correct”, and the participant proceeded to
the next picture. If an answer was incorrect, the counter re-
verted to zero, and text flashed saying ”incorrect”; this task
continued in a loop until ten questions were answered in a
After completing the training, participants proceeded to the
decision task. This decision task mirrored the pretest; participants were asked to view two pictures, presented side by side,
and judge whether they shared the same manner. All participants performed both conditions. If they performed the manner task first, they received these instructions: ”There are two
parts to a motion - manner and path. ”Manner” is how someone gets from one point to another, as in what type of action
they are using. ”Path” is where they are going, as in what
direction they are headed. Now you will view a series of pictures. If they share the same manner, press ”j”. If not, press
”f”. Press ”Space” to proceed.” In a second condition (Path
categorization), participants were asked to judge whether two
pictures shared the same path. The procedure in this condition mirrored the manner categorization task, except that participants were asked to identify whether two pictures had the
same path or not.
Before the presentation of each pair of stimuli, a crosshair
was shown for 1500ms. After 500ms, the two pictures were
presented, and remained on the screen until the participant
pressed either ”j” or ”f”.
During both tasks, we measured the participant’s reaction
time, as well as whether or not their response was correct.
The aim of the decision task was to test how quickly the participant identified particular aspects of motion events, which
allowed us to investigate whether particular speakers attended
faster to manner or path. We used reaction time as a measure
of non-linguistic expertise in categorizing manners and paths.
Verbal Expertise Task After the path and manner decision
tasks were completed, the participant performed a linguistic
expertise task. The goal of this task was to assess the fluency
with which the participant encoded and identified manner.
Each participant was asked to write down as many manner
verbs as they could in a 30 second time period. For comparison, they were also asked to write down as many path verbs
as they could, during the same time period. We analyzed the
effects of both their individual expertise in path and manner,
as well as the difference between the two. In order to ensure
that differences in expertise were not due to typing speeds or
other unrelated factors, we also asked participants to demonstrate their expertise in other categories: types of mammals
Treatment of Data
We investigated whether within-language differences in manner expertise were correlated with similarity judgments on a
triad task. Additionally, we predicted that participants who
chose more same-manner variants on the triad task would also
tend to conflate manner with the fact of motion, which we
measured by the degree of difficulty a participant experienced
on the path decision task. Manner expertise was evaluated in
terms of both linguistic and non-linguistic expertise.
Overall, there was a high degree of variation in the percentage
of how many same-manner choices participants made on the
triad task. There was a mean of 63.75% same-manner choices
(Median = 62.5%), with a range (25%, 87.5%), and a standard
deviation of 21.6%.
Nonverbal Response Task
We measured non-verbal expertise in manner and path by
comparing the reaction times on the manner and path categorization tasks. Reaction time was interpreted as a measurement of task difficulty. As a measurement of expertise, we
used the difference between their reaction time on the manner
decision task and their reaction time on the path decision task.
This was intended to capture the extent to which a particular
participant was better at categorizing manners than categorizing paths.
In general, participants were overall faster at categorizing manners (M = 1678ms, SD = 324ms) than paths (M =
2039ms, SD = 590ms).
We applied linear regressions of Judgment, the proportion
of same-manner variants chosen in the triad task, as a function
of participants reaction times. In general, participants who
took longer overall on the categorization tasks chose more
same-manner variants in the triad task (R2 = .41, F(1, 8) =
5.668, p < .05). The mean reaction time of a participant over
both the manner and path decision tasks significantly predicted how many same-manner variants they would choose
(β = .59, p < .05).
There was a statistically significant effect of the difference
between reaction time on categorizing paths and categorizing
manners (R2 = .4388, F(1, 8) = 6.256, p = .04). As the degree of difference between the reaction times on the two tasks
grew, participants tended to choose more same-manner variants on the triad task (β = 1.66, p = .04). 1 . In other words,
participants who had higher manner expertise in comparison
to path expertise demonstrated a preference for manner on the
Figure 7: Difference in Non-Linguistic Expertise Predicts
Proportion of same-manner choices
Non-Linguistic Expertise Predicts Same-Manner Choices
To measure linguistic expertise in manner, we recorded the
number of manner a participant generated in 30 seconds, as
well as the number of path verbs. We used both the individual
variables (number of path verbs and number of manner verbs)
as predictors, as well as the difference between the number of
manner verbs and path verbs generated for each participant.
To measure non-linguistic manner expertise, we used the
difference in log-reaction time for each participant between
the manner and path categorization tasks. This tells us how
much better each participant is at categorizing manners than
paths. There were 2560 total responses to non-filler stimuli, across 10 participants. Incorrect responses accounted for
15% (377) of the data and were discarded for analysis. A
further 3% (65) of the responses were removed because they
fell 2 standard deviations beyond the mean RT. RT data was
log-transformed before the analysis.
Finally, we measured the proportion of same-manner judgments each participant made in the motion triads task, out of
eight total triads.
LogPathRT - LogMannerRT
The figure above shows the proportion of same-manner
judgments, plotted against the difference in reaction time between the path and manner categorization tasks. Although all
participants found the manner decision task easier than the
path decision task, the degree to which a participant found
the path task more difficult, which we measured with reaction time, predicted performance on the triad task.
Verbal Response Task
To measure linguistic manner expertise, we asked participants
to generate as many manner verbs as they could in 30 seconds; we also asked participants to generate as many path
verbs as they could in 30 seconds. Manner expertise was
judged both by the number of manner verbs generated, as
well as the difference between the number of manner verbs
and path verbs each participant produced. We applied linear
regressions of judgment with the number of manner verbs and
the number of path verbs, as well as the difference between
There was no significant effect of the number of manner
verbs generated on the number of same-manner judgments
participants made (R2 = .08, F(1, 8) = .7324, p = .24). Linguistic manner expertise did not significantly predict samemanner judgments (β = −.02, p = .42). The linear regression
of judgment predicted by the number of path verbs showed no
significant effect (R2 = .1364, β = .02, F(1, 8) = .2197, p =
1 A comparable effect was found in a 3X3 mixed effects model
with Judgment, Condition, and Presentation (Identical, Match, Mismatch, NoMatch) as predictors of reaction time; the model indicated
that participants who chose more same-manner variants took considerably longer on the path categorization task than on the manner
.65). Additionally, the difference in linguistic manner and
path expertise, measured by the difference in number of
verbs generated, did not significantly predict triad judgments
(R2 = −.03, β = −.03, F(1, 8) = 1.242, p = .29). This suggests there was no relationship between a participants linguistic expertise and preference for a same-manner or same-path
variant on the triad task.
Although the difference between the number of manner
verbs and path verbs generated did significantly predict reaction times on the path categorization task (R2 = .56, β =
−.03, F(1, 8) = 10.41, p = .01) and manner categorization
task (R2 = .65, β = −.05, F(1, 8) = 14.76, p < .01), there was
not a statistically significant effect of this measure of linguistic expertise on the predictor of non-linguistic expertise used
previously the difference in RT between the manner and path
categorization tasks (tasks (R2 = .28, β = −.02, F(1, 8) =
3.097, p = .12). In other words, linguistic expertise significantly predicted individual reaction times on the manner and
path tasks, but not the difference in reaction times between
the tasks. There was a slight, but not statistically significant,
Figure 8: Linguistic Manner Expertise Predicts Performance
on Non-Linguistic Categorization Task.
Non-Linguistic Expertise as a Function of Linguistic Expertise
The figure above illustrates the slight effect of linguistic
expertise on non-linguistic expertise. There seems to be a
general correlation between the two factors, but not a statistically significant one.
of Condition (manner categorization and path categorization).
Speakers tended to be better at identifying whether two pictures shared the same or different manners, than identifying
whether two pictures shared the same or different paths. One
interpretation of this result is that the rich semantic space of
manner in English encourages greater expertise and fluency
with identifying the manner component of motion events. Alternatively, it could just be the case that all speakers are better
at categorizing manners than paths, regardless of language.
Speakers who were faster at categorizing manners should
also show a preference for the same-manner variants on the
triad task. When we compared individual speakers nonlinguistic expertise in manner and path, we found that speakers who were considerably better at categorizing manners
than paths tended to choose more same-manner variants. As
the difference between reaction times on the manner decision
task and path decision task grew, speakers demonstrated more
of a preference for manner in the triad task. This interpretation of expertise takes into account not only the ability to
identify manners, but also the ability to identify paths; participants who struggled more with identifying paths chose less
One explanation for this behavior stems from Slobin’s expertise theory. He argues that satellite-framed languages
encourage the conflation of manner with the motion event
(Slobin, 2004). Because manner is conflated with the fact
of motion, these speakers do not view manner as a separate
component or as a co-occurring event. Rather, the expression of manner in the main verb encourages a compression
of the manner component with the motion itself. Furthermore, speakers with higher expertise in manner tend to conflate manner more with the fact of motion. These speakers
find it more difficult to view manner as a separate, distinct
component of motion events, and thus have a harder time ignoring irrelevant manner information when making category
judgments about path. Speakers with the highest expertise
in manner should actually have the most difficult time on the
path decision task, because it involves ignoring information
about manner, and focusing only on path; for speakers with
high manner expertise, manner is viewed as an integral component of motion events, and is less separable from the fact
This provides an explanation for behavior on the triad task.
Participants who chose same-manner variants appeared to
have a difficult time viewing path apart from manner, because
manner was an integral part of their motion event schema.
They viewed the triad task as a choice between a samemanner and different-manner variant, rather than a samemanner and same-path variant, due to the relative speed with
which they attended to and processed manner.
If Slobin’s theory is correct, English speakers should be better at categorizing manners overall than paths, because English encourages both linguistic and cognitive attention to
manner. This effect was confirmed in the significant effect
We predicted that participants with higher linguistic expertise
in manner would have faster reaction times on the manner categorization task. Participants who were able to generate significantly more manner verbs than path verbs did have better