Review of Experiments at the Interfaces
|EDITOR: Jeffrey T. Runner, editor
TITLE: Experiments at the Interfaces
SERIES TITLE: Syntax and Semantics, Volume 37
PUBLISHER: Emerald Group Publishing Limited
Darcy Sperlich, Department of Applied Linguistics and Language Studies,
University of Auckland / School of English, Manukau Institute of Technology
Runner begins the book by giving a brief overview of the articles contained in
1 Interface Effects: Serbian Clitics -- Molly Diesing, Draga Zec
The first study investigates the Serbian second-position clitic ‘je’, which can
be placed either after the first word (1W) in the sentence, or after the first
phrase (1P), positions which have been considered simply interchangeable in
previous descriptions. The authors note the limitations of previous studies, and
explore argument-initial and predicate-initial sentences, together with
branching and nonbranching phrases where ‘je’ may be placed. They argue that the
position of ‘je’ is determined by discourse, structural and prosodic features.
Using corpus data (synchronic and diachronic), together with experimental
evidence, and conclude that differences were found between the argument and
predicate placement of ‘je’.
They then look at a previous study on the placement of ‘je’ (which confirmed
what the authors suspected), flowing to the main experiment. This experiment
presented biased and neutral contexts (combined with argument and
predicate-initial positioning) for ‘je’, and found that the basic pattern held.
This leads to the further investigation of context, finding that discourse,
structural and prosodic features all are important in determining the placement
of ‘je’ (1W/1P), namely in an argument versus predicate-initial of branching and
nonbranching phrases. They conclude that placement of ‘je’ is not simply
optional, but dictated by a combination of different interfaces.
2 I'm Leaking Oil and Looking for a Garage: Testing Conditions on Meaning
Transfer -- Sam Featherston, Klaus von Heusinger, Hanna Weiland
This study examines shifts in meaning with the German nominalizing suffix
‘–ung’, in three different experiments. The article starts by reviewing shifts
in meaning, such as ‘I am parked out back’, in which ‘I’ might mean ‘my car’ or
‘(I am) the driver of the car which is…’, which in turn represents a subject or
a predicate shift. The authors note that context can trigger either a Result or
Event reading, discuss issues of preferred versus forced interpretation,
controlling context versus lexical items, and the problem of how to generalize
The background on meaning shift leads to discussion of ‘-ung’ and ultimately the
aims of the study. Moving onto the first experiment (asking participants to
judge test sentences), they use the local optimum method by controlling the
context to suit the differing lexical items used -- testing the acceptability of
sentences, divided by Result/Event within either a NP/VP, and their relatedness.
Experiment two used the identity method by matching a small array of lexical
items to a particular context, and using the ‘Thermometer Judgments’ method to
test acceptability (an evolution from Magnitude Estimation). Results show that
firstly the identity method is superior to local optimum, which captures the
intuitions on the ambiguous readings of ‘-ung’. The final study focuses on the
factor of relatedness, finding its effect remains constant. Finally, the
observations of the experiments are bought to the fore, and the methodologies
3 Tracking the Preference for Bound-Variable Dependencies in Ambiguous Ellipses
and Only-Structures --Arnout W. Koornneef, Sergey Avrutin, Frank Wijnen, Eric
The authors begin by covering two different hypotheses about pronoun resolution,
a single route (e.g. via discourse) and a dual route (e.g. via logical syntax
and discourse). It is the latter hypothesis that current the study is conducted
against, using Reuland’s Primitives of Binding (POB) framework. The aim is to
investigate whether there is a processing difference in strict/sloppy readings
of pronouns in VP-ellipsis and ‘only’ structures, which is predicted by the POB
account on grounds of economy. The literature reviewed finds favourable evidence
for it, although not compelling.
The first experiment (off-line) tested four different conditions, bias to either
the strict or sloppy antecedent, within an ellipsis or ‘only’ structure. The
experiment presents a story (in Dutch) with the target sentence, and
participants are then given two sentences and asked to rate how the sentence
relates to the story on a scale of 1-5, indicating which sentence they
preferred. Also, they indicated the difficulty and plausibility of each story on
a 1-5 scale. Results show that bias worked well, while obtaining a sloppy
interpretation was slightly easier than a strict one (predicted by the POB to be
more economical), and the ellipsis and ‘only’ structure were treated much the same.
Experiment two (eye tracking) was based on the previous test materials, using
the most successful stories in inducing bias to either strict or sloppy
readings. Overall, sloppy interpretations are found to be more economical than
strict interpretations, which are clearly reflected in the measures. Among other
things, they also find reason to investigate a problem which can actually be
attributed to individual differences, as some decide to establish a
coreferential relation in the first conjunct, which helps them in the second
conjunct of the ellipsis/ ‘only’ structure region. They conclude by bringing in
converging evidence from neurolinguistic experiments, which largely support the
4 Most Meanings are Superlative --Hadas Kotek, Yasutada Sudo, Edwin Howard,
The chapter begins by discussing the superlative and proportional aspects of
‘most’, especially in subject position. They review the literature and argue for
a structural approach to the readings of ‘most’. They cover their
representations in semantic theory, which motivates their experiment. In each
experiment that follows, they used either 2 and/or 3 groups of colored dots in a
picture, with varying number of dots (e.g. 7 blue dots, 4 red and 4 yellow dots)
asking participants to match sentences such as ‘most/ more than half of/ the
dots are blue’ to the picture.
The first experiment uses the ‘covered box’ methodology, where two pictures are
shown and participants match sentences to one of them. If they believe that the
pictures do not match the sentence, they can opt for a third unseen picture
covered by a box (participants are told one of these pictures will match the
sentence) -- this helps in detecting a less preferred interpretation. The second
experiment includes three different colors compared to the previous two, and
instead of a covered box a rating task was opted for, as well as a different
method in presenting the dots. The last experiment includes reaction timing and
self-paced counting of dots. The authors provide a detailed discussion of
reaction times and confidence levels, and their results point to two different
readings for ‘most’ in the subject position, a proportional readings (dominant)
and a superlative one (latent, and previously unnoticed in the literature).
Finally, the authors overview all the experiments and compare them to another
experiment with different results, before discussing the semantics ‘most’ and
how this might interact with other components to provide the superlative reading.
5 Grammatical Illusions and Selective Fallibility in Real-Time Language
Comprehension -- Colin Phillips, Matthew W. Wagers, Ellen F. Lau
This study comments upon ‘linguistic illusions’, which are ungrammatical
sentences that appear to be grammatical at first glance, e.g. ‘The key to the
cabinets are on the table’. They begin by comparing two ways the processer could
retrieve information, searching in a syntactic tree to find subject-verb
agreement, for example. The second type is ‘content-addressable access’, where
the parser assesses the incoming information based upon cue strength, checking
for the closest match, which may lead to inaccuracies.
The authors then review structures made ungrammatical that are easy to detect
(no illusion), which is useful in measuring the performance of search
mechanisms. The first structure reviewed is island constraints, e.g.
wh-extraction *‘who do you wonder whether the press secretary spoke with?’,
which is referred to as an ‘active’ dependency formation. This means the
dependency is constructed in advance before information pertaining to the
position of the gap disambiguates it. The second is backwards anaphora, as in
‘As he wrote the review, John wondered about the word count’, which if accounted
for under Principle C of the Binding theory, helps to understand c-command
violations. The last is forwards anaphora, as in ‘The banker didn’t want the
judge to convict himself’, which has syntactic constraints (e.g. Principle A)
and pragmatic ones (e.g. neo-Gricean pragmatics).
The next topic is grammatical illusions, thought to be a result of constraints
not impacting on processing as much as those just discussed. Subject-verb
agreement, or the lack of it, tends to happen with marked pronouns, e.g. plural.
Incorrect case licensing in German may be detected but to different degrees,
e.g. dative problems are more detectable than nominative. Forward anaphora of
the pronoun type has trouble being accounted by Principle B (having pragmatic
constraints), where they look at child and adult comprehension studies that show
processing problems when number and gender are altered. The same can be said of
negative polarity items (e.g. any). The final item is comparative illusions,
‘More people have been to Russia than I have’, which have been shown to be very
hard to detect.
Finally, the authors discuss the possible reasons for these illusions;
directionality, locality, higher order representations, and structural priority,
the major proposal. It is structural information which leads to dependency
formation. When the parser comes across something incorrect it looks for other
cues, resulting in illusions. The authors sum up by applying this to the
6 Seeing what you Mean, Mostly -- Paul Pietroski, Jeff Lidz, Tim Hunter, Darko
Odic, Justin Halberda
This article continues the previous theme of investigating ‘most’, as in ‘Most
of the dots are yellow’, but instead focusing upon its semantic representation
and how this is constructed in thought. The authors first discuss how words
might be associated with meaning and how this is achieved. For instance, logical
vocabulary has an analytical component to its meaning, and in the case of ‘many’
its meaning may relate to the Approximate Number System (ANS), but it is not
entirely clear on which strategy a speaker might analyse ‘most’.
The authors present possible ways in which ‘most’ might be computed, focusing on
the theory that it is analysed in terms of ‘cardinalities’, and discuss how this
could be implemented among various theoretical constructs. Reviewing previous
studies, they reject the OnetoOnePlus model over the ANS, and discuss the
problems of ‘most’ meaning ‘significantly more’. They then introduce the ITT
(acronym only provided) from a previous study by Lidz et al., which is a model
of how verification procedures are computed, giving an in depth discussion. The
final section looks at other uses of ‘most’, e.g. with count nouns.
7 On the Representational Nature of Representational Noun Phrases -- Jeffrey T.
Runner, Micah B. Goldwater
The final article considers the English anaphor ‘himself’, comparing its use as
a direct object reflexive (DOR) as in ‘John hit himself’, and as a
Representational Noun Phrase’ (RNP) as in ‘John pointed at a picture of
himself’. They first review work on syntactic/semantic binding of ‘himself’ in
these two uses, and cover their interpretation in ellipsis. This leads to the
first experiment, testing DOR and RNP using a scene verification task
(identifying a sentence with a picture) in a variety of structures (with two
referents). Results show that RNP are interpreted more coreferentially than DOR
in ellipsis contexts, which they point out is unusual (for the DOR) where
Binding theory is concerned.
The authors cover proxy readings of reflexives, i.e. the ‘statue’ readings as in
‘Ringo Starr fell on top of himself’, when discussing the person himself and his
statue. This leads to a second experiment to tease out the differences between
proxy and NP in ellipsis contexts. They find in a similar vein to the first
experiment, that proxy interpretations attract more coreferential
interpretations than DOR.
The final experiment (eye tracking) tested various conditions of the reflexive.
Overall, proxy readings are interpreted to be more coreferential, and the eye
tracking reveals that there is a difference between how proxy and nonproxy are
processed, even though the final result appears to be similar (in choosing the
The authors conclude with an unified account of proxy and RNP reflexives,
considering a structural and a semantic approach.
We first discuss each chapter and offer our viewpoints, before discussing the
book as a whole.
Diesing and Zec’s study of clitic placements unravels quite succinctly the
factors that dictate the use of ‘je’. A point of criticism may be levelled at
the experimental process, in that when presenting the stimuli, together with two
sentences side by side with differing clitic placement, and asking which one
appears more natural, one may risk introducing response bias. The forced choice
could mask finer details between the two; perhaps they could have presented the
two sentences in separate sections of the survey, and ask to judge its
naturalness on its own, which would require a Likert scale (or magnitude
estimation). The use of prosody to show the differences between the ‘je’s
provide another angle but it is not noted whether the pitch chart shown applies
to various speakers, or just one.
Featherston et al.’s paper on the varied interpretations of ‘-ung’ is
interesting, especially the great extent they have mapped out all methodological
considerations for each experiment. Considering how the experiments evolve and
which principles guide them is an inspiration to anyone testing meaning; it is
exactly this kind of careful consideration of all the possible factors that an
experimenter would find valuable.
It was pleasing to see that Koornneef et al.’s experiments had their theoretical
roots in Reuland’s theory, rather than the older Binding theory. It is good
that the authors pilot the study first in an off-line task, rather than just
doing an on-line task first. Also of note is their examination of individual
differences, which were borne out to be true; what interests me is why we find
two groups of the same language speakers acting differently, one making use of
discourse strategies early on, while the other does not? (See Sequeiros (2004)
for a Second Language Acquisition perspective.)
Kotek et. al. provide a series of experiments which set out the
proportional/superlative meaning of ‘most’. The methods are most interesting,
from the ‘covered box’ methodology to alternating colours. These approaches are
invaluable for those who wish to tease out ambiguous meanings in words which
would not be apparent by looking at introspection alone.
Phillips et al. provide a select overview of grammatical illusions, and offer
their insight into what might be occurring. While there are many interesting
things to comment on, one which caught my eye was the tentative suggestion that
reflexives in argument positions should not be able to long distance bind (in
English at least), because only structural features are used in search for
antecedents. This means that the phi-features of reflexives are irrelevant, and
such a suggestion would have implications for theories of anaphora that rely on
morphological differences between reflexive types (e.g. Reuland 2011).
Pietroski et al. offer a deep theoretical analysis of ‘most’ while using
experimental studies to back up their claims. While not being particularly
conversant in the nuances of semantic representations, I found their article
quite enlightening as to how a speaker might represent ‘most’ mentally.
The final article by Runner and Goldwater on anaphora is interesting as again
eye tracking shows differences in how a speaker processes these reflexives in
different contexts, which would not be so obvious if we just looked at their
final product. Their experiment seems to fit in well with what Kennedy and Lidz
(2001) find, namely that in ellipsis constructions (in a comparative structure)
the elided anaphor has a long distance antecedent, hence the question is whether
or not this could be incorporated into their own analysis.
Considering the volume as a whole, the articles fit well with each other, are
sufficient in depth and explain the issues well. Furthermore, as noted in the
introduction, the hope is that these articles will inspire further research --
they do because of the wide range of methodologies used and questions asked are
worthy of further application. The only criticism I could offer is the narrow
range of topics covered, three on anaphora, and two on ‘most’. Runner does note
that only recently has the experimental component been used in conjunction with
formal linguistics, which is a good thing, and perhaps explains the relatively
narrow focus of the topics and the state of the field. To conclude, I would
recommend this volume to those interested in the syntax/semantic interface and
more specifically in experimental methodology.
Kennedy, C. and J. Lidz (2001). A (Covert) Long Distance Anaphor in English.
Proceedings of the 20th West Coast Conference on Formal Linguistics. K.
Megerdoomian and L.A. Bar-el. Somerville, MA, Cascadilla Press: 318-331.
Reuland, E. (2011). Anaphora and Language Design. Cambridge, Massachusetts, The
Sequeiros, X.R. (2004). ''Interpretation of reflexive anaphora in second language
VP-ellipsis: Relevance Theory and paradigms of explanation.'' Second Language
Research 20(3): 256-280.
ABOUT THE REVIEWER
| ABOUT THE REVIEWER:
Darcy Sperlich is currently a senior lecturer of ESOL in the School of
English at the Manukau Institute of Technology, in Auckland, New Zealand.
He is also a PhD candidate in his final year at the Department of Applied
Language Studies and Linguistics at the University of Auckland,
investigating anaphoric interpretation in Chinese Mandarin by speakers of
other languages, and whether or not this suggests an anaphoric
pragmatic/syntactic division of labour in the languages concerned,
involving a combination of psycho- and theoretical linguistics, as well as
SLA. His other research interests include Chinese comparative dialectology,
especially as related to syntax.