The Influence Of The Word Connection Type On The Facilitation Effect In The Lexical Decision Task

Authors: Milena Jakic1, Aleksandar Kostic1 and Dušica Filipovic-Ðurdevic1,2
Affiliation: 1Laboratory of Experimental Psychology, Faculty of Philosophy, University
of Belgrade, 2Department of Psychology, Faculty of Philosophy, University of Novi Sad
Email addresses: xmile@eunet.yu, akostic@f.bg.ac.yu, dmfilipo@f.bg.ac.yu

The Influence Of The Word Connection Type On The Facilitation
Effect In The Lexical Decision Task

Introduction
The results of numerous studies indicate that word recognition is faster when a target
word is preceded by the associatively or semantically connected prime word (cf. Meyer
& Schvaneveldt, 1971; Koriat, 1981; Neely, Keefe, & Ross, 1989; Schelton & Martin,
1992; Thompson-Schill, Kurtz & Gabrieli, 1998). The aim of this study is to answer two
questions: a) is there a facilitation effect that derives from the type of word relation over
and above the effect of associative connection, and b) is the facilitation effect between
associatively connected words symmetrical, or put differently, will the effect change if
we change the positions of the prime and target stimulus. In order to answer these
questions we performed two experiments in which we examined the facilitation effect
among two groups of word pairs: pairs in which the connection is purely associative and
pairs with an associative and semantic type of connection (synonymy, antonymy,
hyponymy). In order to compare different types of relations among the stimuli we have
chosen the theory of lexical semantics: componential analysis. Experiments will also test
the predictions of this theory.

Method
Twenty-eight participants (Experiment 1) and twenty-seven participants (Experiment 2)
were tested in Lexical decision task (lexical priming paradigm). All the participants were
first-year psychology students from the University of Belgrade. The pairs of stimuli were
presented on the computer screen (SOA period was 750 ms), and the participants had to
decide if the second stimulus was a word of Serbian language or not (choosing a
YES/NO button). The dependent variable was the reaction time, which was measured
from the beginning of the presentation of the second stimulus until the response.

In each experiment 100 pairs of stimuli were presented, half of which were
pseudoword-targets. Both word- and pseudoword-targets were preceded either by neutral
context (*****)1 or by the words (counterbalanced in Latin square design). Word pairs
were taken from the list of associative norms, i.e. The Associative Dictionary of Serbian
Language (Piper, Dragicevic & Stefanovic, 2005), while the word-primes for
pseudoword-targets were chosen to be unassociated to any of word-targets. In the first
experiment the word targets were stimuli from the word-association test, while the primes
were their most frequent2 associates. In the second experiment primes and targets

1 Since the strong facilitation effect is obtained even if the target is preceded by the unrelated word, we
decided to measure the facilitation effect of the related word comparing to the neutral context, which
should represent the reaction time for a discrete target word.
2 Frequency of associates was taken from the test of free associations (Piper et al, 2005), in which 800
students took part. Most frequent associates were the words that most participants gave as the response to
the stimulus given in a test, and the frequency is the number of participants that had given it.

---

reversed their positions. There were five groups of associatively connected stimuli pairs.
Three of them were also semantically connected: synonymy (kuca – dom meaning house
– home), antonymy (noc – dan meaning night – day) and hyponymy (jabuka –vocka
meaning apple – fruit ). Two groups were not semantically connected: a stronger
contextual (majmun – banana meaning monkey – banana) and weaker contextual
connection (svada – tašta meaning quarrel – mother-in-law). The difference between the
stronger and weaker contextual connection was based upon the associative frequency,
whose average values were significantly different. The criteria for the selection of the
stimuli concerning the relation of synonymy and antonymy was based upon the primary
lexical definitions from the dictionary of Serbian language (Recnik MS). Since the
criteria for the hyponymy was more clear than the criteria for the synonymy and
antonymy relation, it wasn’t necessary to base the choice of the stimuli on the primary
lexical definitions.

Results and discussion
The analysis of variance established the 57 ms effect of primed context comparing to the
neutral: F(1, 49)=47,414, p<0.0001. The rest of the analyses were performed on the
facilitation effect which was calculated by subtracting reaction times for the target
preceded by neutral context and the target preceded by related prime. In table 1 the
average values for the five types of relation between primes and targets are given. It is
obvious that the five groups of stimuli are very different with regard to distributional
statistics, but this couldn’t be avoided because of the criteria of stimuli selection. 3
However the differences between the groups were partialed out by means of the analysis
of covariance. The analys is of covariance performed on items showed that the facilitation
effect was significantly stronger when in addition to associative connection word pairs
were also semantically connected (the average difference was 35 ms; figure 1): (F(1,
43)=7.03, p<0.01).

Table 1. Average values for the stimuli in first experiment
rating of
Rt on target
Rt on target
Facilitation
Facilitation
relation
associative
target
prime
associative
(ms) in neutral
in related
effect (ms)
effect (ms)
type
connection
freq
freq
prime freq context (*****)
context (ms)
ANOVA
ANCOVA4
Antonymy
4,669
1317
1209
200
584
532
52
61
Hyponymy
5,741
47
57
110
627
562
65
55
Strong
context
5,994
397
622
145
605
568
37
40
Weaker
context
4,370
367
150
79
638
615
23
25
Synonymy
5,380
122
452
139
732
624
108
91


3 Besides the fact that there is a limited number of synonyms, hyponyms and antonyms in language, it was
also necessary to find the confirmation of strong associative connection between them (in Associative
dictionary).
4 Notice that facilitation effect ANCOVA is not simply obtained by subtracting reaction time on target
preceding neutral and connected prime, but also by partialing out (in analysis of covariance) factors known
to affect facilitation (frequency of prime and target, associative frequency of prime, rated associative
connection, length (number of graphemes) of the target).

---

The results of the first experiment also indicated that the relation type between words
(synonymy, antonymy, hyponymy, stronger contextual and weaker contextual

Figure 1. Average facilitation effect for different Figure 2. Average facilitation effect for different types
type of connections



of relation
0100
140
090
120
080
100
070
s
)
s
)
80
060
60
050
40
040
20
030
f
a
c
i
l
i
t
a
t
i
o
n

e
f
f
e
c
t

(
m
f
a
c
i
l
i
t
a
t
i
o
n

e
f
f
e
c
t

(
m
0
020
-20
010
-40
00
Antonym
Strong context
Synonym
associative and semantic
pure associative
Hyponym
Weaker context
type of connection

type of relation


connection) has a significant influence on the facilitation effect: F(4,40)=2.62, p<0.05
even if we partial out factors known to affect facilitation (lexical frequency of the prime
and target (taken from the Kostic, 1999), associative frequency of prime (taken from
Piper et al. 2005), target length, rated associative relation between prime and target)
figure 2). The strongest facilitation effect was in the group of synonymy. Note that the
average lexical frequency for the targets of the synonymy group is higher than average
lexical frequency for the targets of the hyponymy group, while the effect is much higher
for the synonymy group. On the other hand, the effects of hyponymy and antonymy
groups are almost the same, but the average frequency of targets is dramatically different
(see Table 1). These outcomes indicate that the observed effects were not due to target
frequency.
The differences in the facilitation effect among the five experimental situations,
observed ordinally, can be accounted for in terms of componential analys is (Lyons,
1977), a linguistic theory that describes the number of common semantic components
between words in different lexical relations. For example: synonymy pairs will have the
biggest percentage of common semantic components, then antonymy, hyponymy,
stronger contextual connection and the last will be the weaker contextual connection
group. From the perspective of componential analysis, using the Spearman’s correlation
coefficient, we explained the significant percentage of the facilitation effect variance: (r=
-0.38, t(3)=2.81, p<0.01). By the prediction of componential analysis, the facilitation
effect will be the same no matter of the direction of the association (forward or
backward). The second experiment tested this prediction.

The outcome of the second experiment indicates that the facilitation effects are
not symmetrical (table 2). This could be expected on the basis of the associative norms,
but it is not in accordance with previous studies. Some authors claimed that, unlike
associative relations, semantic relations are symmetrical (cf. Thompson-Schill et al,
1998). However, in the present study, although the overall facilitation effect is roughly
the same in both experiments (53 ms versus 54 ms) the correlation of the facilitation
effects in two experiments (including semantic relations) was not significant (r = 0.54, p
> 0.05).

---

Table 2. Average values for the stimuli in second experiment
Rt on target (ms) in
Rt on target in
neutral context
related context
Facilitation effect
Facilitation effect
relation type
(*****)
(ms)
(ms) ANOVA
(ms) ANCOVA
Antonymy
583
525
58
82
Hyponymy
625
583
42
32
Strong context
605
565
40
38
Weaker context
632
584
48
45
Synonymy
644
569
75
66

The results of the present study indicate that the strength of the facilitation effect depends
not just upon the connection type between words (pure associative or associative and
semantic connection) but also upon the relation type between them (synonymy,
antonymy, hyponymy, stronger or weaker contextual connection). The fact that the type
of word relation accounts for facilitation effect over and above the effect obtained by the
factors known to affect priming (frequency of target and prime, rated associative
connection between prime and target, associative frequency of the prime, length of the
target) suggests that the semantic (lexical) relations are cognitively relevant. Furthermore,
the theory of lexical semantics (i.e. componential analysis) provides good predictions of
the facilitation variation among different types of relations. However, the facilitation
effect is not symmetrical and varies in the direction of both the associative and semantic
relation. This, on the other hand, is not in accordance with previous studies and the
predictions of componential analysis.

References
Koriat, A. 1981. Semantic facilitation in lexical decision as a function of prime-target
association. Memory & Cognition, 9(6): 587-598.
Kostic, Ð. 1999. Frekvencijski recnik savremenog srpskog jezika, Beograd, Institut za
eksperimentalnu fonetiku i patologiju govora i Laboratorija za eksperimentalnu
psihologiju Filozo fskog fakulteta u Beogradu.
Lyons, J. 1977. Semantics, 1-2, Cambridge: Camb ridge University Press: 327-335.
Meyer, D.E. & Schvaneveldt, R.W. 1971. Facilitation in recognizing pairs of words:
Evidence of a dependence between retrieval operation. Journal of Experimental
Psychology, 90: 227-234.
Neely, J. H., Keefe, D. E., & Ross, K. L. 1989. Semantic Priming in the Lexical Decision
Task: Roles of Prospective Prime-Generated Expectancies and Retrospective
Semantic Matching. Journal of Experimental Psychology: Learning Memory and
Cognition, 15(6): 1003-1019.
Piper, P., Dragicevic, R., & Stefanovic, M. 2005. Asocijativni recnik srpskoga jezika,
Beograd: Beogradska knjiga, Službeni list SCG, Filološki fakultet u Beogradu.
Recnik MS, 1967-1976. Recnik srpskohrvatskoga književnog jezika t. I-VI, Novi Sad:
Matica srpska.
Schelton, J.R. & Martin, R.C. 1992. How semantic is automatic semantic priming?
Journal of Experimental Psychology: Learning, Memory & Cognition, 18: 1191-
1210.
Thompson-Schill, S. L., Kurtz, K. J., & Gabrieli, J. D. E. 1998. Effects of Semantic and
Associative Relatedness on Automatic Priming. Journal of Memory and
Language, 38(4): 440-458.

---