University of Eastern Philippines

DALL 904:Psycho-sociolinguistics and Language Learning

1st Semester SY 2021-2022

Syntactic Sentence
Processing

APRIL JOY S. CASCANO Dr. Villa C. Carpio

DALL 2 Subject Professor
 1

Introduction

2
Introduction This topic will consider how readers and listeners
assign syntactic structure to string of words.
▹ Determining syntactic structure is an important
aspect of sentence comprehension.
▹ Markers of syntactic structures can help readers
and listeners in this process;
▹ There are claims that comprehenders have certain
preferred strategies that they can use in assigning
syntactic structures to sentences.

3
Outline ▹ The focus is on how we build a sentence structure
using the sequence of words we have recognized.
1. Complexity and sentence processing
2. The clausal hypothesis
3. Explicit syntactic markers
4. Strategies for syntactic processing
5. Garden paths and the sausage machine
6. Syntactic category ambiguity
7. Cross-linguistic evidence for processing
strategies

4
 1
Complexity and
Sentence Processing

5
 Complexity
and Sentence
Processing  It is claimed that certain sentences are
more difficult to process than others.
 The more difficult structures were
those that had undergone some
change(s) in the history of their
derivation.

6
▹ Derivational Theory of Complexity
- Notion that difficulty in the production
and/or comprehension of a sentence
is related to how different the surface
structure is from the deep structure,
according to the rules of transformational
(derivational) grammar.

7
 Complexity of passive sentences
-It was argued that a passive was derived from an
‘underlying form’ that expressed the basic idea of the
sentence and which was closer in structure to the
active version
1. The dog chased the boy.
2. The boy was chased by the dog.
-It was predicted that passive sentences would be
more difficult to process and the comprehension
would involve greater processing demands.
8
 Application of a negative transformation
3. That ball is blue.
4. That ball is not blue.
 Sentence 4 was predicted to be more
difficult than sentence 3 because a negative
transformation has applied.

9
 Relative clauses in which the relative
pronoun is the object of the embedded verb
were predicted to be more difficult than
those in which the relative pronoun is the
subject.
5. The boy who chased the dog ran home.
6. The boy who the dog chased ran home.

10
Evidence of ▹ The greater complexity of passive sentences
complexity: came from a task which measures how long
participants take to match sentences with
pictures showing the events described in
those sentences.
▹ The result showed that the task took longer
for the passive versions of the sentences than
for the active versions.

11
Evidence of
complexity: ▹ The same findings with the other sentence
pairings with different syntactic contrasts in
sentence 3 and 4.
▹ However, further research found that if the
perspective shown in the picture was
changed the result was that the responses to
passives were no longer slower than those to
actives.
▹ It shows that grammatical structure is not the
only factor that influences comprehension.
12
Evidence of
▹ Another experimental tests of derivational complexity by
complexity:
Miller and McKean. Participants were given examples of
transformationally related sentences. The labels show the
nature of the transformational relationship.
7. he caught her – he didn’t catch her (AA NA)
8. he caught her – she was caught by him (AA AP)
9. he caught her – she wasn’t caught by him (AA NP)
AA –Affirmative Active NA- Negative Active
AP – Affirmative Passive NP – Negative Passive

13
Evidence of
complexity: ▹ For each trial, Miller and McKean presented
one example pair such as sentence 7-9 as a
model.
▹ The participants are given an Affirmative
Active version of a different sentence and
asked to transform it in their heads, pressing
a response button to indicate when they had
performed that transformation.

14
Evidence of
complexity: ▹ The study found that the time taken to press
the response button increased in line with the
number and complexity of the transformation
involved.
▹ In Memory task it was demonstrated that
sentences were more confusable with one
another the more closely related they were in
terms of transformations.

15
Evidence of
complexity: ▹ More recent research has focused on
the possibility that there is a bias
towards statements, since statements
are closer to the ‘underlying’
affirmative active forms of
sentences, which provide the
propositional content required for
comprehension.

16
Evidence of
complexity: ▹ Further memory task, participants
were required to memorize both a
sentence and a list of unrelated words.
It was found that the more complex the
sentence was, in transformational
terms, the fewer the words the
participants could reliably remember.

17
Criticisms:
▹ It only provide indirect evidence for
greater complexity in language processing.
▹ Carrying out a transformation from one
sentence to another requires some mental
operations, but critics question whether
these are the mental operations ordinarily
involved in sentence comprehension.

18
Criticisms:
▹ Memory tasks are not measures of the
immediate processing of sentences.
▹ From an experiential point-of-view,
readers and listeners have far greater
exposure over their lifetimes to active
than to passive sentences.

19
 1

The Clausal Hypothesis

20
 Clausal
Hypothesis ▹ -claims that the clause is the basic unit of
analysis in language comprehension.
▹ The clause is a group of words in a
sentence that includes verb.
▹ There are finite clauses where a verb can
carry markings for tense and for number
agreement and non-finite clauses with
verbs that cannot be marked in this way.

21
 Clausal
Hypothesis
11. John walked.
FINITE CLAUSES
12. They ride on donkeys.
13. John wanted to leave work early.
NON-
14. The man who John saw was very tall. FINITE
Finite clauses can include main verb clauses
as in sentence 10 and 11, but also subordinate
clauses such as the relative clause underlined
in sentence 13.
22
 Clausal
Hypothesis
▹ The clausal hypothesis is a special instance,
for language, of a ‘chunking’ process. In an
experiment, subjects are better able to
remember number sequence if they are
encouraged to chunk them, as in
remembering a phone number.
14. 463 5631
15. 4635631

23
 Clausal
Hypothesis
▹ It was hypothesized that language chunking
might proceed on the basis of sentence
structure, i.e. that sentences are chunked into
units like clauses.
▹ A click location experiments were done to
test this. Participants were given a printed
version of a sentence, but without the
additional markings (/ and .)

24
 Clausal
Hypothesis ▹ The participants then heard a recording of the
sentence, onto which a click (or a beep) had
been added, for instance at the position
indicated in the examples by |. Their task was
to mark on the printed version where the
click had occurred.
▹ What the experiments were interested in was
whether the nature of syntactic boundary
close to the click would affect the likelihood
that the click would ‘migrate’.
25
 Clausal
Hypothesis 16. The teacher who taught the biggest cla/ss/
was given extra pay by his principal.
17. The very pretty and talented blonde gi/rl/
was the runner up in the Miss America contest.
▹ They found that participants were more likely
to erroneously report clicks at the boundary
between the subject and predicate of the main
clause, marked by (/) in the examples in
sentence 16 than in sentence 17.

26
 Clausal
Hypothesis 16. The teacher who taught the biggest cla/ss/
was given extra pay by his principal.
17. The very pretty and talented blonde gi/rl/
was the runner up in the Miss America contest.
▹ This is because in 16 the subject predicate
boundary is also the end of the relative clause
who taught the biggest class.
▹ The experimenters claimed that this result
provides evidence for the clausal hypothesis.

27
 Clausal
Hypothesis ▹ Clausal Structuring and Clausal Processing
▹ Clausal structuring is a claim that language
Criticism:
is segmented in to clauses at some stage
during comprehension but that processing
can carry on during a clause
▹ Clausal processing is a stronger claim about
clause structure i.e. that processing is
concentrated at clause boundaries.

28
 Clausal
Hypothesis ▹ Evidence for a claim that processing does
not take place solely at these points but is
carried out within clauses comes from online
studies of comprehension, which track word-
by-word analysis the sentence.
▹ One instance of this is the word monitoring
task employed by Marsie-Wilson and Tyler,
(1980).

29
 Clausal
Hypothesis ▹ In this task participants, are told to listen for a
word which may or may not occur in a
sentence which they then hear that word, if it
is present.
18. Some thieves stole most of the lead off the roof.
19. No buns puzzle some in the lead off the text.
20. Some of the no puzzle buns in lead text the off.

30
 Clausal
Hypothesis 18. Some thieves stole most of the lead off the roof.
19. No buns puzzle some in the lead off the text.
20. Some of the no puzzle buns in lead text the off.
▹ The same target word is used (shown in bold)
and it is in the same serial position in each
example. Across the experiment, the position
of the target words was deliberately varied so
the effect of accumulating information across
the utterance could be measured.

31
 Clausal
Hypothesis ▹ The results show an overall advantage for
normal prose over anomalous prose and for
anomalous prose over scrambled prose.
▹ In terms of position effects, the results show
first that there is no advantage for one
sentence type over any others at the first
word position.

32
Reasons:
1. When the target appears at this position it
does not yet have any contextual support.
2. In normal prose sentence 18 there is a
marked decrease in word monitoring times,
although the change across the sentence is
less dramatic than for non normal prose.
3. Scrambled sentences show practically no
serial position effect.

33
Key Results:
1. When the sentence has structure, a word
within that sentence can be responded to
more rapidly that this response facilitation
increases as more of the sentence is heard;
2. The availability of both syntactic and
semantic (meaning) structure results in a
faster decrease in response times than the
availability of syntactic structure alone.

34
▹ These results supports the claim that sentence
processing does not need to wait until major
structural boundaries, but can take place in a
cumulative way as the sentence is heard.

35
 1
Explicit Syntactic
Markers

36
 Explicit ▹ Sentence processing can be helped by words
Syntactic or affixes which explicitly mark the
Markers syntactic structure, but these are often left out
in English, without making the utterance
ungrammatical.
21. I told Mark the woman I met had red hair.
22. I told Mark that the woman who I met had red
hair.

37
 Explicit ▹ But leaving out syntactic markers can affect
Syntactic processing.
Markers
▹ Response times in a phoneme monitoring task
were faster when explicit markers of syntactic
structure were present.
▹ Phoneme monitoring is similar to word
monitoring is similar to word monitoring, but
requires participants to listen to particular
speech sound rather than for a word.

38
 Explicit When participants listened to a version of
Syntactic sentence 23 with the structure make explicit by
Markers the words shown in square brackets in the
example, they took less time to detect the target
phoneme (/p/) than when those words were left
out.
23. The children |that were| playing in the
hayloft startled the farmer’s wife.

39
 Explicit ▹ The claim is that listeners find it easier to
Syntactic complete the syntactic analysis when syntactic
Markers structure is more explicitly marked and are
therefore better be able to pay attention to the
phonemes in the input.
▹ Studies have shown the usefulness of the
explicit marking of syntax.

40
 Explicit
24. John knew the answer was wrong.
Syntactic 25. John knew that the answer was wrong.
Markers
26. John knew the answer by heart.
▹ Evidence for the effect on processing of the
presence of the complementizer (that) is that
reading times for the word was were much longer
in 24 than n 25.
▹ This is because it is only at that point in 24 that
the sentence structure becomes clear and different
from that in 26

41
 Explicit ▹ The more complex a sentence structure is, the
Syntactic more helpful explicit markers of syntax tend to
Markers be.
These are also easier to recognize and provide
useful anchor points during processing since
explicit markers are words and grammatical
endings that occur very frequently in the language
inflectional endings, determiners conjunctions,
etc. as well as generative.

42
Prosody and ▹ Spoken sentences also carry further cues to
Punctuation syntactic structures through the intonation and
phrasing used in speech.
▹ Prosodic phrase structure cues perform much
the same function as syntactic markers, i.e.
they make the syntactic structure of an
utterance more explicit.
▹ Punctuation provides a similar marking of
syntactic structure.

43
Prosody and 27.What is this thing called love?
Punctuation 28. What is this thing called, love?
29. What, is this this thing called love?
▹ Note that as you read these out or hear them in
your head you will notice how their prosody
and phrasing reflects the punctuation
differences.

44
Prosody and ▹ Line breaks can also have an effect on
Punctuation processing.
30. When I’ve finished running
the class can we all go home?
31. When I’ve finished running the class can we all go home?
▹ Formatting issues such as these have practical
implications, such as in the design of signs.

45
 1
Strategies for Syntactic
Processing

46
Strategies for
Syntactic
Processing How, then do we carry out the processing of
sentences?
How do we build sentences using the input
words?

47
Strategies for 1. Discontinuous constituents are difficult to
Syntactic
Processing process.
32. Lou ran her friend up.
33 Lou rang her friend in the Outer Hebrides up.
34. Lou rang her friend in the Outer Hebrides
that she hadn’t seen since their trip together to
Japan the summer before last up.
▹ The more material that intervenes between
rung and up, the more difficult it is to
process. the sentence.
48
Strategies for 2. Native speakers seem to have clear
Syntactic
Processing preferences in the structures they assign to
sentences.
▹ Non-native speaker’s use approaches to the
processing of sentences is unambiguous, it is
problematic.
▹ Most readers do a double-take, and have to
go back and read the sentence again to get
the right analysis.

49
Strategies for ▹ By contrast sentence 36, is ambiguous, but it is
Syntactic often hard to detect the ambiguity.
Processing
35. Sandy said that Terry will take the cleaning
out yesterday.
36. Pat brought the book that I’ve been trying to find
for Chris.
It is argued that these effects come about because
human sentence processing automatically builds the
words of a sentence into a particular preferred
structures.

50
 1
Garden paths and
sausage machines

51
 ▹ Garden Path Sentence is one which leads the
Garden Paths
and Sausage reader/listener ‘up the garden path’ by initially
Machines inducing an interpretation which turns out to
be incorrect.
▹ Garden path sentences involve a misleading
syntactic analysis.
37. The horse raced past the barn fell.
38. The horse which was raced past the
barn fell.

52
 ▹ 37. The horse raced past the barn fell.
Garden Paths
and Sausage 38. The horse which was raced past the
Machines barn fell.
▹ A successful revision will result in raced past the
barn being interpreted as a relative clause, with
raced as a participle rather than a past tense verb.
▹ This relative clause reading is made clear by
which was in 38. Because these words are missing
from 37, this sentence is said to contain a reduced
relative.

53
 ▹ Other interpretation suggested interpretation
Garden Paths
and Sausage of example 3
Machines ▹ fell as meaning a type of hill, as in fell-
running and of barn fell as therefore a type of
fell or the name of a particular fell)
▹ Or require additional punctuation, e.g.
semicolon after past (the horse raced past; the
barn fell.)

54
 ▹ Example from newspaper headline:
Garden Paths
and Sausage 39. Man held over betting shop killings freed
Machines Eye drops off shelf
Two sisters reunited after 18 years in
checkout queue
Hospitals are sued by 7 foot doctors
▹ The removal of grammatical information such as
function words in newspaper add to the level of
ambiguity by removing explicit markers of syntax
make an important contribution to comprehension.

55
Garden Paths
Sausage Machine - derived from the way the
and Sausage human sentence processor packages the input
Machines words into strings of phrases much as a machine
manufacturing sausages packages the contents
into strings of sausages.
▹ The sausage machine is a parser, i.e. it
analyses sentences according to their syntactic
structure.

56
 Key principle:
Garden Paths
and Sausage 1. The goal of its operation is to build a syntactic
Machines tree, also known as a phrase marker
2. The parser is deterministic. It can merely build one
tree at a time. And can be contrasted with parallel
processing models in which multiple
interpretations can be entertained at the same time.
3. Tries to keep the syntax as simple as possible. It
also tries not to leave too much material unattached,
i.e. not to built into the syntactic tree under
construction.
57
 ▹ LATE CLOSURE: When possible, attach
Late Closure
and Minimal material into the clause or phrase currently
Attachment being processed.
▹ MINIMAL ATTACHMENT: Attach incoming
material into the phrase marker being
constructed, using the fewest possible nodes
consistent with the well-formedness rules of the
language consideration.

58
 42. Before the police stopped the driver he was getting
Late Closure nervous.
and Minimal
Attachment

59
 42. Before the police stopped the driver was getting
Late Closure
and Minimal nervous
Attachment

60
 ▹ Syntactic parsing approach-The syntactic
SYNTACTIC
CATEGORY processor cumulatively links words into a
AMBIGUITY syntactic tree structure as they are encountered
▹ The processor needs to know what kind of
word it is dealing with at each stage. However,
a large proportion of words in all languages are
ambiguous.
▹ Many involve category ambiguity, which is
when the same word-form may represent more
than one syntactic category
61
 ▹ Syntactic parsing approach-The syntactic
SYNTACTIC
CATEGORY processor cumulatively links words into a
AMBIGUITY syntactic tree structure as they are encountered
▹ The processor needs to know what kind of
word it is dealing with at each stage. However,
a large proportion of words in all languages are
ambiguous.
▹ Many involve category ambiguity, which is
when the same word-form may represent more
than one syntactic category
62
 ▹ Karen knew the schedule by heart.
SYNTACTIC
CATEGORY
AMBIGUITY

63
 ▹ Karen knew the schedule was wrong
SYNTACTIC
CATEGORY
AMBIGUITY

64
 ▹ The evidence for these processing strategies
SYNTACTIC
CATEGORY comes from garden path experiences.
AMBIGUITY ▹ These experiences can be measured in eye-
tracking experiments that record participants
eye movements as they read.
▹ If a participant is reading a sentence like 43,
then when they reach the word was their gaze
lingers on that word, and they show regressions
(backtracking or backwards eye movements) to
earlier parts of the sentence.
▹ Consider
65
▹ This does not happen when they reach the
word in the equivalent position (he) in 42.
▹ The stopped and this analysis is pursued in
both sentences. Such an analysis becomes
problematic in 43 once the word was is
encountered ,because this signals that the
driver should be interpreted as the subject of
the second clause (and stopped is being used
intransitively, i,e, the police stopped
themselves not someone else). Similar
findings apply to Minimal Attachment
66
▹ 46. Sam figured that Max wanted to take the
book out.
▹ This is reflected in a preference for interpreting 46
with the second of the meanings given above, i.e.
that Max wanted to take the book out.
▹ That interpretation has out belonging to take, with
two intervening words (or one intervening NP
constituent), rather than to figured, with seven
intervening words (and a string of intervening
constituents). These attachments are illustrated in
47.
67
68
▹ Right Association - states that new nodes are
preferentially attached to the lowest node ( if
the strategies of Late Closure and Minimal
Attachment do not indicate otherwise). This is
shown by the lower of the two dashed
connecting liens in 47.

69
 ▹ Syntactic parsing approach-The syntactic
SYNTACTIC
CATEGORY processor cumulatively links words into a
AMBIGUITY syntactic tree structure as they are encountered
▹ The processor needs to know what kind of
word it is dealing with at each stage. However,
a large proportion of words in all languages are
ambiguous.
▹ Many involve category ambiguity, which is
when the same word-form may represent more
than one syntactic category
70
 ▹ The partial tree structure would allow words
SYNTACTIC
CATEGORY from more than one category.For example, if
AMBIGUITY the word trains is encountered in the fragment
in 48 it could be either a verb
▹ I know that the desert trains…
▹ I know that the desert trains young people to be
especially tough.
▹ I know that the desert trains are especially
tough on young people.

71
 ▹ Because of parsing operations like Late Closure,
SYNTACTIC
CATEGORY Minimal Attachment or Right Association, the parser
AMBIGUITY may prefer one structure- and therefore one syntactic
category for the ambiguous word- over another.
▹ However, these parsings strategies do not lead to a
structural preference for one analysis. For such
cases, one solution that has been proposed is that the
processor puts off attaching any additional words
into the current syntactic tree until further words in
the input clarify the category ambiguity

72
 1
Cross Linguistic
Evidence for Processing
Strategies

73
 ▹ It has been suggested that there are two competing parsing
SYNTACTIC strategies, and that the cost associated with one of these differs
CATEGORY across languages.
AMBIGUITY ▹ Recency Principle-states that there is a preference to attach
material to a recent point in the preceding structure (for
alternatively that there is an increasing processing cost with
more distant attachments).

▹ The second principle, Predicate Proximity, results in a

preference to attach the relative clause to the noun phrase that is
closest to the verb in the main clause, or that is otherwise
highest in the structure (i.e. the daughter)/hija as the object of
the verb in (56 and 57), and to painting/pintura) in 58 and 59),
and this has to be learned through exposure during acquisition.
In some languages, -such as Spanish- the cost of violating 74
 ▹ native speakers of English relied primarily on
SYNTACTIC
CATEGORY word order, native speakers of German on
AMBIGUITY animacy and to a lesser extent on noun-verb
agreement, and native speakers of Italian on
noun-verb agreement, and to a lesser extent on
animacy.
▹ Clearly a key factor on learning a strategy for
processing a language is the reliability of the
cues employed by that strategy

75
 1. Recency Principle-states that there is a preference to attach
SYNTACTIC material to a recent point in the preceding structure (for
CATEGORY alternatively that there is an increasing processing cost with
AMBIGUITY more distant attachments). The second principle,
2. Predicate Proximity, results in a preference to attach the relative
clause to the noun phrase that is closest to the verb in the main
clause, or that is otherwise highest in the structure

76
 1

Summary

77
 ▹ We have seen in this topic a number of basic
SUMMARY
properties of sentence comprehension
▹ It is widely assumed that in order to understand
a sentence we need to build a syntactic structure
for that sentence, and that we do this
incrementally, word by word;

78
 ▹ Explicit markers of syntactic structures help this
SYNTACTIC
CATEGORY process, reducing ambiguity and increasing the
AMBIGUITY efficiency with which readers and listeners can
construct syntactic trees;
▹ Readers process sentences following a set of
parsing preferences, which determine how input
words are built into the syntactic tree being
constructed;

79
 ▹ These parsing strategies include Late Closure,
SYNTACTIC
CATEGORY Minimal Attachment, and Right Association, all of
AMBIGUITY which are based on structural preferences;
▹ Evidence for these preferences comes from garden
path experiences for which measurable data are
provided by studies of eye movements during
reading;
▹ Parsing strategies are assumed to apply in all
languages, but the details on how they apply reflect
differences, between the grammatical languages.

80
 Thank You For Your Indulgence!

Any questions?
You can find me at @april04cascano@gmail.com

81