Deverbal Nominals: Do we process all we know about them

1
Deverbal Nominals Do we process all we know
about them?

• Christina Manouilidou
• Department of Psychology, Concordia University
• Douglas Mental Health University Institute

2
What is a Deverbal Nominal?

• Nominal nouns, adjectives
• Deverbal nominal a nominal derived from a verb
• teach gt teacher
• destruct gt destruction
• do gt doable
• Denominal nominal a nominal derived from a noun
• rain gt rainy
• child gt childish
• mother gt motherhood

3
What do we know about Deverbal Nominals?

• Deverbal Nominals (DNs) are fairly rich in terms
  of linguistic features.
• Complex morphological structure.
• They are specified for Thematic Features (TFs).
• TFs are inherited features and not inherent.
• TFs are related to both semantics and syntax.
• TFs are typically associated with the grammatical
  class of verbs and not of nominals.

4
Aim

• Investigate whether Thematic Features (TFs) of
  Deverbal Nominals (DNs) are processed during
  lexical access.
• Investigate the effect of TFs in the lexical
  access of Pseudo-Words (Pseudo-Ws) violating
  thematic specifications.
• Address the implications of the experimental
  investigation with respect to the
  psycholinguistic theory of lexical access and the
  morphological theory of word formation.

5
Argument Structures ofVerbs and DNs

• (1) The enemy destroyed the city
• agent theme
• The enemys destruction of the city
• agent theme
• (2) The king imprisoned the foreigners
• agent theme
• The imprisonment of the foreigners by the king
• theme agent

6
Argument Structures ofVerbs and DNs

• (3) The choir rehearsed the chant
• agent theme
• The choirs rehearsal of the chant
• agent theme
• (4) All patients can access this room
• agent theme
• This room is accessible to all patients
• theme agent

7
Theoretical Approaches to Deverbal Word Formation

• Lexical
• Verbal TFs are passed on to the nominals derived
  from them during the derivation process.
• Inheritance or Percolation of the verb Argument
  Structure (AS) to the nominal.
• (Lieber, 1980, 1990 Selkirk, 1982 Williams,
  1981 Hoekstra, 1986 DiSciullo Williams, 1987
  Baker Bobaljik 2002 Spencer, 2005).

8
Theoretical Approaches to Deverbal Word Formation

• Radically Syntactic
• The presence of AS is only syntactically
  represented.
• Deverbal nominals contain a VP node which is
  syntactically active. (Borer, 1993)
• The presence of AS is regulated by various
  functional projections, such as vP and AspectP,
  of the extended projection of VP
• (Grimshaw, 1990 Van Hout and Roeper, 1998
  Alexiadou, 2001 Malouf, 2000 Roeper, 2005)

9
Implications - Predictions

• Lexical Approach
• Semantic Participants
• TFs are part of the meaning structure of lexical
  items.
• TFs part of the lexical entry of DNs, thus
  available in, out-of-context, lexical access.
• Syntactic Approach
• TFs are mapped onto syntactic positions as
  syntactically realized arguments
• Argument structure properties available only in
  syntactic context.
• TFs do not available in out of context lexical
  access.

10
Derivation of Deverbal Nominals

• katakto ltAgiltThgtgt -tis Noun ltRi ltEvgtgt
• conquer -er
• kataktitis ltRiltThgtgt
• conqueror

11
TFs in the Lexical Access of Verbs

• The mental representation of verbs includes
  information about their TFs, specifying the
  number and type of their arguments
• (Jackendoff, 1990 Pinker, 1989).
• Accessing this information has a salient effect
  on the lexical access of verbs
• (Shapiro et al., 1993 Friederici Frisch,
  2000 Joanette Brownnell, 1990 Saffran et al,
  1980 Mauner et al., 2002).

12
Processing Dissociations between Nouns and Verbs

• Sentence comprehension (Mauner and Koenig, 1999)
• ERPs (Friederici Frisch, 2000)
• Verb-Noun dissociation in aphasia
• Production (Berndt et al., 1997)
• Naming (Luzzati et al., 2002)
• Explanation for V-N differences
• internal complexity of Vs
• argument structure features or TFs
• TFs accessed at early stages of word recognition
  (Joanette Brownnell, 1990 Saffran et al.,
  1980).

13
When it comes to DNs

• Assumption 1
• The mental representation of DNs includes
  thematic information similar to that of verbs.
• Question Does the processor make use of TFs
  when retrieving a DN?
• Assumption 2
• Thematic Information should also impose
  constraints in Deverbal Word Formation.
• Question Can we detect the processing of TFs in
  on-line processing of deverbal novel word
  formations.

14
Experimental Approach

• Exp 1. Processing of TFs in Deverbal Adjectives
  
• Exp 2. Processing of TFs in Deverbal Nouns
• Exp 3. Processing of TFs in Pseudo-Words which
  violate constraints on derivation
• Task Simple lexical decision
• Language Modern Greek (MG)
• Participants Native speakers of MG, 24-30 years

15
Exp 1 Accessing TFs of Deverbal Adjectives

• Objectives
• Detect the processing of TFs through longer
  Reaction Times (RTs) during the lexical access of
  Deverbal Adjectives (DVAdjs).
• Differentiate RTs due to processing of TFs from
  the effects of the derivational process itself,
  via a comparison to Denominal Adjectives
  (DNAdjs).

16
Methodology

• Experiment
• On-line simple lexical decision task.
• Dependent Variables
• a) Reaction times measured in milliseconds
• b) Error Rate
• Participants
• 24 native speakers of Modern Greek who attend a
  Greek University.

17
Stimuli - DVAdjs

• DVAdj-simos
• anatrepsimos reversible
• DVAdj-menos
• kalymmenos covered
• DVAdj-tikos
• perioristikos restrictive
• DVAdj-tos
• ypofertos bearable

18
Stimuli

• Denominal Adjectives (DNAdjs)
• DNAdj-eros
• vroxeros rainy
• DNAdj-ikos
• emporikos commercial
• Non-Derived Adjectives (NDAdjs)
• thetikos positive

19
Exp.1 Hypotheses

• 1. Derived forms will yield longer RTs than
  non-derived forms due to derivation.
• DNAdjs/DVAdjs gt NDAdjs
• comprehensible/agricultural gt intelligent
• 2. DVAdjs will yield longer RTs than DNAdjs due
  to the processing of TFs.
• DVAdjs gt DNAdjs
• comprehensible gt agricultural

20
Procedure

• Software PsyScope 1.2.5
• Pause 150 ms
• Mask 200ms
• Pause 150ms
• Target efarmo-simos
• Response Press Yes/N o unlimited time
• The main experiment was preceded by a set of
  training trials and pretest items.

21
Analysis

• PRIOR TO DATA ANALYSIS
• Erroneous responses were removed (3).
• Responses exceeding 1500 ms were removed.
• Outliers, (RTs two Standard Deviations above or
  below the overall mean) were removed (2-7).

22
Exp.1 Cumulative Results

• DVAdj vs. DNAdj
• 725ms 668ms
• plt0.0001
• DVAdj vs. NDAdj
• 725ms 652ms
• plt0.0001
• DNAdj vs. NDAdj
• 668ms 652ms
• NS

23
Exp. 1 Further Analysis

• WHY to see how the issue of derivation may have
  interacted with the lexical access route used.
• HOW Each group of adjectives were compared to
  its non-derived controls

24
Further Analysis
The derived groups that did not differ in RTs
from their controls were considered nondecomposed
with respect to their access. The derived groups
that differed in RT from their controls were
considered decomposed with respect to their
access. Separate analyses were conducted for
Hypothesis 2

• decomposed DVAdjs versus
• decomposed DNAdjs

nondecomposed DVAdjs versus nondecomposed DNAdjs
25
Exp. 1 Results Hypothesis 2-Partially supported

• RTs for decomposed groups
• Significant Difference
• F 6.236, p 0.01

Decomposed SIG

• RTs for nondecomposed groups
• No significant difference
• F 2.637, p 0.12

Nondecomposed NS
26
Exp. 1 Preliminary Conclusion and Ensuing
Question

• TFs processed only in DVAdjs accessed through
  decomposition.
• Processing of TFs in DVAdjs can only be observed
  after morphological parsing has taken place.
• Is it the viewing of the verbal root that
  activates the processing of TFs?

27
Exp 2 Low Frequency Deverbal Nouns (DVNs)

• AIM
• To determine whether decomposition and the
  consequent viewing of the verbal root
  facilitate accessing the TFs of DVNs.
• STIMULI
• Low frequency DVNs used to assure decomposition
  of derived forms
• Nonderived Nouns (NDNs) as controls for effects
  of decomposition
• A group of Denominal Nouns matched as controls to
  each group of DVNs to detect processing of TFs

28
Exp. 2 Stimuli Deverbal Nouns

• -tis subject nominalizations (n16)
• katakto to conquer ? katakti-tis conqueror
• -si process/result nominalizations (n16)
• diorthono to correct ? diortho-si
  correction.
• -ma/-simo mainly activity and, more rarely,
  result (n16)
• ravo to sew ? rap-simo sewing

29
Exp. 2 Stimuli Denominal Nouns
-as occupation or profession (n16) sfouggari
sponge ?sfouggar-as sponge- diver -ia
quality associated with the base N
(n16) zitianos beggar ?zitian-ia
beggary -adiko place associated with the base
N (n16) psari fish ? psar-adiko fishery
NDNs (Nonderived Nouns) exanthima rash
(n48)
30
Methodology/Procedure

• Simple Lexical Decision Task.
• Participants 24 native speakers of Modern Greek
• Software PsyScope 1.2.5
• Pause 150 ms
• Mask 200ms
• Pause 150ms
• Target katakti-tis
• Response Press Yes/N unlimited time
• The main experiment was preceded by a set of
  training trials and pretest items.

31
Exp 2 Hypotheses

• Due to structural complexity, derived forms (DVNs
  and DNNs) will yield longer RTs than NDNs.
• DVNs, DNNs gt NDNs
• katakti-tis, sfouggar-as gt adartis
• conqueror, sponge-diver gt partisan
• 2. If TFs are processed, then RTs will be
• longer for DVNs than for DNNs.
• DVNs gt DNNs
• katakti-tis gt sfouggar-as
• conqueror gt sponge-diver

32
Exp.2 Cumulative results

• Derived forms yielded longer RTs
• DNNs vs. NDNs gt p0.0316
• DVNs vs. NDNs gt p0.0131
• Decomposition and potential activation of the
  verbal root.
• However, this did not trigger the processing of
  TFs.
• DVNs vs. DNNs gt p0.9064

33
Within category variation in DVNs
34
Within category variation for DNNs as well
35
Results Hypothesis 1 Partially Supported

• Not all groups of DVNs and DNNs showed longer RTs
  than NDN controls.
• The derived groups that did not differ in RT from
  their controls were considered nondecomposed with
  respect to their access.
• The derived groups that differed in RT from their
  controls were considered decomposed with respect
  to their access.

36
Therefore separate analyses for
Hypothesis 2 nondecomposed DVNs (-si)
versus nondecomposed DNNs (-ia)

• decomposed DVNs (-ma)
• versus
• decomposed DNNdjs (-adiko)

• decomposed DVNs (-tis)
• versus
• decomposed DNNdjs (-as)

37
Exp. 2 Results
Nondecomposed

• Nondecomposed
• DVNs -si vs DNNs-ia
• No RT difference (p0.74)
• Decomposed
• DVNs-ma vs DNN Controls -adiko
• Significant RT difference (p0.03)
• Decomposed
• DVNs -tis vs DNN Controls -as
• No RT difference (p0.49)

NS
Decomposed
SIG
Decomposed
NS
38
Towards INTERPRETATION

• DVNs-ma/simo decomposition and processing of
  TFs.
• DVNs-tis decomposition did not trigger
  processing of TFs.
• ?Decomposition might be a necessary but not a
  sufficient condition for the processing of TFs.

39
Hypothesis 2 Partially Supported

• Could the different sentential and AS properties
  of the stimulus groups explain why the processing
  of TFs is not evident in both decomposed groups
  of DVNs?

40
Properties of Individual Items-tis (-er)

• Decomposed-tis e.g. xeiris-tis user
• -subject nominalizations
• -do not permit either adverbial modification or
  by-phrases
• O katharistis tou ktiriou prosektika.
• the cleaner of the building carefully
• O katharistis tou ktiriou epi ena mina.
• the cleaner of the building for a month
• (Alexiadou, 2001 129)

41
Properties of Individual Items-si (-ion)
Nondecomposed -si e.g. anatinak-si
explosion - process/result nominalizations -
do not tolerate aspectual modification
denoting repetition i syxni anatinaksi tis
gefyras apo to strato the frequent blowing up
of the bridge by the army
42
Properties of Individual Items-ma/simo (-ing)

• Decomposed-ma/-simo e.g. ply-simo washing
• - mainly activity
• - permit by-phrases
• - permit adjectival aspectual modification
  denoting repetition
• to syxno plysimo ton piaton apo to Jianni
• the frequent washing of the dishes by
  Jiannis

43
Exp. 2 Interpretation

• -ma/-simo (e.g. rapsimo sewing)
• decomposed access
• increased verb-like properties compared to the
  other DVN groups
• only DVN group to show evidence of the processing
  of TFs.
• -tis (e.g. kataktitis conqueror)
• decomposed access
• does not have verb-like properties
• does not show evidence of the processing of TFs

44
THEREFORE
- specific properties of a particular item
(verbal/eventive character) may be responsible
for the processing of its TFs. - decomposition
may be a necessary, but not a sufficient
condition for the processing of TFs.
45
How about the adjectives?

• Significant Difference
• DVAdjs-menos/-simos
• vs. DNAdjs
• gt Processing of TFs
• No significant difference
• DVAdjs-tos vs.DNAdjs
• gt Lack of Processing of TFs

46
Verbal Properties of Deverbal Adjectives

• -menos
• 1. To parathyro itan anoig-meno me losto
• The window was openedMENOS with crowbar
• 2. To fagito itan proxeira mageire-meno
• The food was casually cookedMENOS
• -tos
• 1. To parathyro itan anoig-to me losto
• The window was openedTOS with crawbar
• 2. To fagito itan proxeira mageiref-to
• The food was casually cookedTOS

47
Verbal Properties of Deverbal Adjectives

• -simos
• To vamvaki einai epekserga-simo apo eidikes
  mixanes
• the cotton is processable by special
  machines
• To vamvaki einai efkola epeksergasimo
• the cotton is easily processable

48
Conclusion

• Exps. 1 2 TFs affect processing only for those
  DNs with an increased verbal character, with the
  decomposition access route possibly playing a
  facilitatory role.

49
Compounds with menosKehayia, Manouilidou, Ralli
(2004)

• Noun-menos (N-menos)
• pondiko-fagomenos mice-eaten
• Adverb-menos (Adv-menos)
• katharo-grammenos clearly-written
• Noun-Noun-Stress Change (NN-SC)
• tsimendo-lithos cement-block

50
Does Theta Role Saturation increase processing
load?

• N-menos compounds yielded longer RTs than both
  Adv-menos and N-N compounds.

51
Compounds with tisManouilidou, Fyndanis,
Kehayia, Ralli (2007)

• Investigate whether the presence of an explicit
  argument in the position of the first constituent
  will trigger the processing of the TFs of the
  second constituent.
• vivlio-kritis book reviewer
• kapno-kalliergitis tobacco cultivator
• thirio-damastis beast-tamer

52
Central Questions

• Does the processing of the TFs of the second
  constituent (-kritis) depend on whether there is
  an explicit argument in the form of first
  constituent (vivlio-)?
• Can we detect any processing effects of Theta
  Role Saturation (TRS) in NN-tis compounds?

53
Method

• EXPERIMENT
• One on-line lexical-decision task with priming
• the compound acted as prime for the deverbal
  second constituent.
• PARTICIPANTS
• 28 native speakers of MG.
• 17 females, 11 males.
• Average age 31.5 years.
• Average education 12 years.

54
Stimuli

• Noun-Noun (NN-tis) with TRS (n9)
• vivlio-syllektis gt syllektis / agorastis
• book-collector gt collector / buyer
• Adjective-Noun (AN-tis) without TRS (n9)
• kalo-douleftis gt douleftis / glyptis
• strenuous-worker gt worker/ sculptor
• NN with non-derived second constituent, (n10)
• xion-anthropos gt anthropos / adelfos
• snow-man gt man / brother

55
Hypotheses

• Priming effects
• Not necessarily strong, since we assume that the
  processing of thematic information will act as
  inhibitory factor for an easy decomposition of
  the second constituent.
• Longer raw RTs
• due to the processing of thematic information.
• vivliosyllektis gt kalodouleftis gt xionanthropos
• book-collector gt strenuous worker gt
  snowman

56
Results Mean differences

• Significant priming effect for
• NN compounds
• NN-tis compounds
• Tendency (p0.077) for AN compounds.

57
Results Raw RTs

• Longer RTs for the compounds with deverbal second
  constituent (NN-tis and AdjN-tis) compared to
  those with a non-derived second constituent.
• vivliokritis, kalodouleftis gt xionanthropos
• No difference between NN-tis compounds and
  AdjN-tis.
• vivliokritis kalodouleftis
• book reviewer strenuous worker

58
Interpretation

• Even in a favorable environment (when a potential
  argument is present), the TFs of the deverbal
  nominals with the suffix -tis are not activated.
• Specific properties of the suffix -tis and its
  derivatives (either in derivation or compounding)
  which may block the surfacing of its thematic
  properties.
• Reference to an entity and not to an event.
• Thus, it is not the case that what is described
  as Theta Role Saturation has a processing
  correlate with all the suffixes that create
  deverbal nominals or compounds.

59

Summary

• Assumption 1
• The mental representation of DNs includes
  thematic information similar to that of verbs.
• Question Does the processor make use of TFs
  when retrieving a DN?
• Answer TFs appear to influence the lexical
  access of only those DNs with increased eventive
  properties or a more verb-like character.

60
What is the status of TFs in Deverbal Nominals?

• Difference between DVNs and DNNs.
• Processing Dissociation within the class of DVNs.

61
Difference between DVNs and DNNs
1) The mental representation of DVNs appears to
involve thematic information. 2) This information
increases complexity, and, when accessed, results
in longer RTs. cf. processing differences between
Nouns and Verbs
62
Dissociation within DVNs

• Why didnt we detect processing of TFs for EVERY
  nominal we tested?
• 1) The representation of these DNs does not
  include thematic information.
• 2) The representation of these DNs includes
  thematic information, but this was not accessed
  during the experiment.
• WHY?
• Cf. Verbs vs. Nouns?

63
Non-accessed TFs

• Lexical access consists of the incremental
  identification of components until the
  participant is ready to make the lexical decision
  and press the button to identify the word as an
  existing word of a given language.
• In the case of DVNs-si, DVNs-tis, and DVAdjs-tos,
  their identification takes place before the
  participant accesses the component with the
  thematic information.
• Accessing the TFs of these particular nominals is
  not a necessary prerequisite to identifying them.
  
• Thematic information is not as prominent as the
  other types of information constituting their
  mental representation.

64
Accessed TFs

• Eventive/verb-like character.
• The more eventive the DN is, the more prominent
  the status of the thematic component in its
  mental representation and the more likely that
  component is to be accessed.
• Accessing TFs is necessary for the word
  recognition of nominals with increased eventive
  character.

65
Representation vs. Access of DNs

• Although there is uniformity regarding the number
  of components which form the mental
  representation of these nominals stemming from
  the percolation mechanism, there is no uniformity
  when it comes to their prominence and their
  significance for a particular nominal. As a
  result, accessing a DN with an eventive character
  requires going through its thematic component,
  while accessing a DN with a diminished verbal
  character does not.

66
Implications for lexical access and the lexicon

• The access of a words meaning is not an
  all-or-none event that happens at a specific
  moment in the time course of word recognition.
• Information stored in mental representations is
  organized into components whose access during
  word recognition is relative to their prominence.
• Various features are activated at different
  time-course and according to their prominence.
• Information related to the grammatical class of a
  specific nominal does not appear to preclude
  accessing of other features in lexical access.

67
THANK YOU!!