Sensitivity Comes from Age and Experience:

1
Sensitivity Comes from Age and Experience Age
and Span Effects on the Ability to Utilize
Contextual Information Thomas A. Farmer1, Karen
A. Kemtes2, and Morten H. Christiansen1 1Cornell
University, 2University of Nevada-Las Vegas
And more broadly, 2) Span tasks measure
language experience, not capacity. A
capacity-based account of individual differences
cannot accommodate these data. No a priori reason
exists to believe that high, but not low
capacity individuals possess a bias against
RC-resolution.
Introduction
Fifty-three college undergraduates were presented
counterbalanced materials in a self- paced
reading task. Participants also completed a
reading span task (Waters Caplan, 1996).
Results Before considering span, the 2 (SC vs.
RC Context) x 3 (SC vs. RC vs. UnAmb) x 3
(ambiguity vs. disambiguation vs. carry-over)
ANOVA yielded a 3-way interaction, F(4, 208)6.0,
plt.001.
Traditional accounts of individual differences in
sentence processing have focused on individual
differences in verbal working memory as a key
explanatory variable (Just Carpenter, 1992, for
example). Alternatively, MacDonald and
Christiansen (2002) proposed that individual
differences in language comprehension are, in
part, a product of differences in language
experience reading span tasks, traditionally
thought to measure verbal working memory
capacity, actually measure differences in
language experience. Although these two
perspectives overlap considerably in terms of the
predictions they make, the experience-based
approach often predicts more subtle interactions
between particular structural elements and
specific lexical items. Here, we capitalize on
such fine-grained predictions in order to explore
the manner in which individual differences in
reading experience influence on-line processing
performance.
Study 2 Older Adults
Figure 1a illustrates that indeed, when context
supports RC-resolution, the garden-path effect is
reduced, although not eliminated. Figure 1b
illustrates that when context supports SC
resolution, and the ambiguity is resolved in
accordance with the SC interpretation,
participants read significantly faster than
baseline at disambiguation. In Figure 1b, it is
also evident that when context supports SC
resolution, but the ambiguity is RC-resolved,
participants are extremely sensitive to the
incongruency. The results of Altmann et al.
replicate here. Although the garden-path effect
on RC sentences is still present, it is
substantially reduced in the RC-supporting context
We administered the same materials to older
adults (65 years or older). Under the assumption
that older adults have more experience with the
statistical regularities of naturally-occurring
language, we predict that they, too, will exhibit
a bias against RC-resolution, even when it is
supported by context. Currently, our sample of
older adults is relatively small. As such, we
cannot yet make any claims regarding statistical
reliability. However, the predicted trends are
strongly present.
These RTs are length-adjusted (Ferreira
Clifton, 1986). This procedure should strongly
reduce the effect of longer RTs associated with
aging, given that it adjusts RTs on the basis of
an individual. In Figure 3a, we see again that
older adults are strongly biased against
RC-resolution, even when it is supported by
context. In Figure 3b, it is evident that older
adults are able to use context, given that they
read SC-resolved sentences faster than at
baseline when supported by context (not the case
when context supports RC-resolution, 3a). This
trend in the data supports the experience-based
approach to individual differences in sentence
processing, and would not be predicted by
capacity-based approaches.
Study 1 Young Adults
Altmann, Garnham, and Dennis (1992) demonstrated
that the referential properties of the
discourse-context can influence the processing of
Sentential Complement / Relative Clause (SC/RC)
ambiguous sentences created by sentence stems
such as He told the woman that hed
misunderstood SC-Resolution the nature of her
question. RC-Resolution to repeat her last
question. UnAmb Control He asked the woman
that hed misunderstood to repeat her last
question. Without the presence of any contextual
information, individuals typically prefer the SC
over the RC interpretation, and exhibit a
garden-path effect when the ambiguity is
RC-Resolved. However, certain types of context do
seem to help readers avoid this garden-path
effect. 2 Related Entities (RC Supporting
Context) A bank manager was giving financial
advice to two women. They were asking about the
benefits of a high-interest savings account. The
bank manager had misunderstood one of the womens
questions about the account but understood the
other perfectly. A context containing 2 very
similar (related) entities sets up an expectation
for an RC-resolution. Participants avoided the
garden-path effect on RC-Resolved sentences in
this context. 2 Similar Entities (SC Supporting
Context) A bank manager was giving financial
advice to a man and a women. They were asking
about the benefits of a high-interest savings
account. The bank manager had misunderstood the
womens question about the account but understood
the man perfectly. A context containing 2
distinct entities helps to confirm a bias toward
SC resolution. HOWEVER An interesting aspect
of the original materials was that in all but one
case, the verb told was used to create the
ambiguity. Spivey and Tanenhaus (1994), through
a series of corpus analyses, found that every
time an SC/RC ambiguity was created by the verb
told, it was ALWAYS resolved in accordance with
the SC interpretation. Prediction Given that
high-experience individuals have had more
exposure to the statistical regularities of
naturally-occurring language, they should exhibit
a bias against RC-resolution, even when context
supports it. Low experience individuals should
not exhibit as much of a bias as they have
reduced knowledge of naturally-occurring
language patterns. Notably, this prediction
cannot be made by capacity-based accounts of
individual differences in sentence processing.
There is no a priori reason to assume that
individuals labeled high capacity should
exhibit an overwhelming bias in favor of
SC-resolution (against RC-resolution) on these
materials. That is, only an experience-based
approach can make such a prediction.

Figure 1. Overall results, before taking into
account individual differences. The results of
Altmann et al. do replicate somewhat.
Experience / Span Effects


Span/experience effects were assessed by
predicting RT difference scores (target sentence
RTs - UnAmb counterpart RTs) from span scores.
These analyses allowed us to avoid dicohtomizing
the continuous span score variable. We elicited
two span score effects. Span significantly
predicted 1) the RT difference between
RC-resolved ambiguous sentences and the
unambiguous counterpart in the SC-supporting
context, t(52)2.04, p.047. 2) the difference
between RC-resolved sentences and the unambiguous
counterpart at carry-over in the RC-supporting
context, t(52)2.54, p.014. Notably, both of
these effects are on the RC-resolved sentences of
interest.
Figure 3. Older adults also seem to be strongly
biased against RC-resolution, even when it is
supported by context.
Conclusion
Here, we exploited the fact that in the Altmann
et al. (1992) materials, a strong statistical
bias in favor of SC-resolution (and against
RC-resolution) existed. The presence of this bias
allowed us to test between capacity and
experience-based theories of individual
differences in sentence comprehension. Through
the evidence presented here, from both young and
older adults, it is evident that individual
differences in language experience, as measured
by a reading span task, or by age, is the true
predictor of individual differences in
processing. Once again, capacity-based accounts
cannot accommodate these trends. We are currently
collecting more older adult data, and do predict
that older adults labeled high span will indeed
be the most sensitive to RC-resolution.
Figure 2. Differences scores (target point -
unambiguous counterpart) for high and low span
individuals.
References
Figure 2 illustrates that at these two points,
high span individuals are more sensitive to
RC-resolution than are low span individuals.
That is, even when context supports
RC-resolution, high span individuals have a more
difficult time with RC-resolution. These
results demonstrate two points in support of the
experience-based framework 1) High experience
individuals, more attuned to the strong
statistical bias in these materials, are always
biased against RC-resolution when the ambiguity
is created with told. This statistical bias, in
high experience individuals, overrides the
context effect.
Altmann, G. T. M., Garnham, A., Dennis, Y.
(1992). Avoiding the garden path Eye movements
in context. Journal of Memory and Language, 31,
685-712. Ferreira, F., Clifton, C. (1986). The
independence of syntactic processing. Journal of
Memory and Language, 25, 348-368. Just, M. A.,
Carpenter, P. A. (1992). A capacity-based theory
of comprehension New frontiers of evidence and
arguments. Psychological Review, 103,
773-780. MacDonald, M. C., Christiansen, M. H.
(2002). Reassessing working memory Comment on
Just and Carpenter (1992) and Waters and Caplan
(1996). Psychological Review, 109, 35-54. Spivey,
M. J., Tanenhaus, M. (1994). Referential
context and syntactic ambiguity resolution. In C.
Clifton, L. Frazier, K. Rayner (Eds.),
Perspectives on Sentence Processing. Hillsdale,
NJ Erlbaum. Waters, G. S., Caplan, D. (1996).
The measurement of verbal working memory capacity
and its relation to reading comprehension.
Quarterly Journal of Experimental Psychology,
49, 51-79.