22
JCPSLP
Volume 15, Number 1 2013
Journal of Clinical Practice in Speech-Language Pathology
greater activation might be provided to words with high
ND versus low ND. While such activation could potentially
inhibit naming of a lexical representation (i.e., via lexical
competition), it appears this relationship actually
facilitates
production during children’s naming. These results also
support an influence of ND at the lexical level in models of
speech production (e.g., Goldrick & Rapp, 2007).
In addition to finding facilitative effects of ND at
the semantic level, a similar pattern emerged at the
phonological level. Words with high ND were articulated
more accurately than those with low ND. Such productive
advantages have been demonstrated previously. Recall that
Munson and Solomon (2004) found that adults articulated
words with high ND to a greater degree than words with
low ND, related perhaps to a need for greater intelligibility
when words have multiple similarly sounding forms. Based
on the current results, it appears that a word’s ND can also
influence young children’s articulation of a word.
Note that participants in the present study
did
exhibit
developmental production errors on this task; such errors
were more likely to occur on words with low ND than
on words with high ND. This is a novel finding. ND may
then not only play a role in lexical processing, but also in
post-lexical processing at the articulatory level. Otherwise,
articulatory differences might not have been observed
between words with low and high ND. Revisiting Goldrick
and Rapp’s (2007) proposed “cascading” effect during
phonological processing, nontarget words sharing a target’s
phonological structure may also become activated during
productive acquisition of a word. As mentioned above, this
relationship seems to facilitate production. Words with high
ND presumably have more forms activated relative to words
with low ND, thereby resulting in more accurate articulation
for words with more phonologically similar forms relative to
words with few neighbours.
Clinical implications
Apart from theoretical contributions, there are also clinical
implications of the results. Since words with high versus low
ND were more accurately named at the semantic level,
these words may be ideal targets for children with word-
finding deficits. One might argue that having children
succeed early (with high ND words) in treatment may
increase motivation before presenting more challenging
activities. Speech-language pathologists could match word
naming strategies to target words based on their ND, with
phonemic cues provided when necessary to improve
naming of these forms. Future research is warranted to
determine how incorporating ND into treatment may assist
children with word-finding impairment.
Limitations
Although there was an attempt to control for many
confounding stimuli factors such as imageability, information
was not available for all items. It is possible that results may
have been impacted by variables for which data were
unavailable. Additionally, given that imitated and
spontaneous productions were analysed together for the
articulatory analyses, it cannot be ruled out that response
accuracy may have been inflated for some children after
hearing a model. Future studies of this nature can control
for this possibility.
Future directions
Work in the future should include older participants to rule
out the possibility that ND may be a developmental factor
affecting articulation only at an earlier age. It is uncertain
used for statistical tests to allow for multiple comparisons.
When applicable, spontaneous productions and imitations
were analysed together given that the majority of children’s
responses (> 75%) were spontaneous. Furthermore, past
studies have found no significant difference in articulatory
accuracy between imitations and spontaneous productions
of words (Andrews & Fey, 1986; Goldstein & Fabiano,
2004). Average accuracy rates for each dependent variable
are presented in Table 1 by condition.
Effect sizes were calculated for all analyses. Using
Cohen’s
d
(1988), effect sizes were considered to be small
(0.2–0.3), medium (0.5), or large (0.8).
Table 1. Mean percentage accuracy rates and
standard deviations for children’s naming by
analysis and condition
Semantic
Binary
Segmental
Low neighbourhood
density
75.50 (10.31)
70.81 (21.16) 95.78 (4.37)
High neighbourhood
density
81.80 (10.38)
77.66 (19.19) 96.88 (3.11)
Semantic accuracy
The first analysis was used to determine how ND might
influence children’s naming according to semantic accuracy.
A main effect of ND was found,
t
(36) = 4.55,
p
< .01,
d
=
0.61, noting a moderate-large effect size. Consistent with
predictions, words with high ND were named more
accurately than words with low ND.
Binary articulatory accuracy
The next analysis considered effects of ND on production at
the phonological level using a binary measure of phonetic
accuracy. There was a significant effect of ND,
t
(36) = 3.82,
p
< .01,
d
= 0.34. As predicted, children more accurately
articulated words with high ND versus low ND.
Segmental articulatory accuracy
The third analysis investigated how effects of ND might
impact production accuracy at a segmental level. Again, a
main effect of ND was discovered,
t
(36) = 3.58,
p
< .01,
d
= 0.29. Segmental accuracy on words with high ND was
greater than those with low ND.
Summary of results
In summary, children more accurately named (retrieved) and
articulated words with high ND versus low ND. This was
true at the phonological level using both a binary and
segmental measure of phonetic accuracy.
Discussion
The current experiment was designed to shed light on how
items in the developing lexicon may interact with one
another as a function of their phonological similarity:
competitively, facilitatively, or neither. Results will first be
discussed for the semantic analysis, followed by
interpretations of the articulatory findings.
Regarding an influence of ND on semantic accuracy,
children most successfully retrieved words with high ND.
This finding revealed a facilitative nature of the lexicon
during naming, such that words appeared to aid one
another. Consistent with original predictions and previous
findings (e.g., Storkel et al., 2006), words with many
phonologically similar forms were easier to retrieve than
those with few forms. Considering the possibility that lexical
forms must be activated in some manner prior to retrieval,