3 RESULTADOS Y DISCUSIÓN
3.2. Implementación fase medir
3.2.2. Medición de la capacidad del proceso antes de la
3.2.2.6. Resumen índices de capacidad L1 utilizando el
Below is a summary of the results as they pertain to the predictions made by the
individual theories.
1. Direct Access View
This view (Gibbs, 1986; Gibbs, 2002) did not receive empirical support from the
experiment presented. Sarcastic Prosody interacted with Frequency in the Balanced
homophone reaction time data, indicating that it was being processed with a different
time frame than a Sincere interpretation. Due to the lack of support for this framework
within the analysis, the remaining experiment focused on testing Graded Salience and
2. Graded Salience Hypothesis
Using this hypothesis, (Giora, 1997), several testable predictions were generated
at the onset of the experiment. These predictions followed the notion that salience would
be added to the interpretations in the case of Sarcasm interacting with previous factors,
when discourse interpretations continually compete until participants are forced to make
a choice. The previous cues that Sarcasm was considered to interact with included
Frequency, Dominance, and Social Contextual information.
Biased Homophones: Graded Salience Hypothesis
When examining processing of Biased homophones using the eye-movement
record, no effect of Sarcasm was seen within the data, either in the reported analyses or in
additional behavioral analyses that were carried out including a direct test of Sarcasm as
interacting with Frequency and Dominance. Instead, the effects of Frequency and
Dominance alone persisted in directing attention and choice throughout the spoken
discourse and decision of applicable context. This suggests that Sarcasm does not
contribute salience given these conditions, and in order to be interpreted within the
Graded Salience Framework, one would argue that Sarcasm is not viewed as a salient
cue.
Balanced Homophones: Graded Salience Hypothesis
Within the eye-movement record, when tested Sarcasm did not interact with
Frequency throughout as originally predicted (when tested using alternative models).
Instead, during the processing of Sentence 2, it interacted with variables such as Change,
generating a shorter fixation duration when there will be a shift in viewing contextual
Contextual Ratings scores, increasing fixation time as the score increased. Together, this
suggests participants are considering the Sarcasm they are presented with in the context
of the overall conversation and its social implications, rather than as a highlight for
specific lexical information (such as Frequency classification) as processing unfolds.
However, this interaction does support the Graded Salience (Giora, 1997) prediction of
being considered a salient cue given Balanced homophone conditions. However, given
Sarcasm did not interact with Frequency, the interpretation is not without issue.
Behavioral: Graded Salience Hypothesis
Within the behavioral data, there was again no effect of Sarcasm in the Biased
homophones (only Frequency and Dominance), thus it appears that the Bias or
Dominance itself (Dominant or Subordinate meaning) interacts with Frequency and is
deemed sufficient information to base choice of interpretation on. Or, another way to
view these results in support of a Graded Salience approach (Giora, 1997), is that
Dominance information contributes sufficient salience to choose an interpretation, and
renders Sarcasm unnecessary. This could be an effect of the fact that Sarcasm does not
directly interact with the mention of the homophone in the design.
In the case of Balanced homophones, a poverty of lexical information associated
with the homophone, appeared to increase reliance on the Sarcasm as an informational
cue soon after it was offered in Sentence 2. This was demonstrated through to final
processing, as measured by reaction times. In reaction time measures, Sarcasm did
interact with Frequency, (as was predicted to also occur within the eye-movement
measures, according to Graded Salience). Additionally, the interaction does appear to
Frequency alternative, and increases reaction time when selecting a Low Frequency
alternative (and removes the significant difference between the conditions) as compared
to Sincere prosody, thus reversing the pattern. This suggests that Sarcastic Prosody did
interact with the Frequency information on some level, even though it was not clear in the
eye-movement record.
In summary, this data suggests that Sarcasm is considered useful in particular
situations in the contexts created within the current experiment. As Sarcasm did not
affect final choice for either Biased or Balanced homophones, it appears that as a salience
cue, while in some cases it does appear to highlight an alternative, the alternative only
receives additional consideration. Together, this results in partial support of the Graded
Salience Framework as an explanation for Sarcasm resolution.
3. Relevance Theory
This framework (Sperber & Wilson, 1986; Sperber & Wilson, 1995) also
predicted an interaction between factors Dominance, Frequency, and Sarcasm when
choosing an interpretation of the homophone.
Biased Homophones: Relevance Theory
In the case of Biased homophones, when considering the lexical information
associated with the homophone (Frequency, Dominance status), it appears that these cues
create a discourse interpretation that is difficult to overcome. Indeed, within the data, it
appears that listeners do not require sarcasm to resolve the ambiguous reference, even
when the cue is available. However, Relevance Theory did predict that this was a likely
outcome, since the information provided by Dominance and Frequency was strong.
altered in the case of Biased homophones (as in Experiment 3), sarcasm can serve as a
stronger contextual effect.
Balanced Homophones: Relevance Theory
Relevance Theory framework (Sperber & Wilson, 1986; Sperber & Wilson, 1995)
predicts that only in some cases will sarcasm be beneficial. Here, it is only in the case of
Balanced homophones where we see a strong effect of sarcasm as an informational cue
being considered by listeners. The effect is observable from when Sarcasm is introduced
in Sentence 2 (such that it interacts with Social Contextual Ratings, increasing processing
of highly rated meanings), through to reaction time data (where it reverses the choice
time pattern by Frequency seen in the Sincere prosody), as reported in the previous
section. This singular effect of aiding in one context (Balanced homophones) was
predicted from the Relevance Theory Framework.
Behavioral: Relevance Theory
The null effect of sarcasm in reaction times and final choice of Biased
homophone interpretations is predicted by Relevance Theory by the prediction that the
cue is not relevant enough to listeners. In the case of Balanced homophones, we see an
interaction of Sarcasm and Frequency in reaction time data (again, such that time to
choose between High and Low frequency alternatives is lower in the High Frequency
condition and longer in the Low Frequency condition than the Sincere control). Here, the
information is considered relevant, considering the lack of other available information.
While Relevance Theory would predict differences between the final choice proportions
Summary
Thus, as described above, it appears that both Graded Salience (Giora, 1997) and
Relevance Theory (Sperber & Wilson, 1986; Sperber & Wilson, 1995) receive some
support from the current experimental findings. In the case of Biased homophones, there
was no observed effect of Sarcasm. While predicted by Relevance Theory, it is unclear
whether sarcasm has no effect on highlighting alternatives for interpretation (given the
fairly consistent Frequency by Bias interaction), or the design simply did not render
sarcasm important enough to be considered, given listeners have access to other
information and conversations were short. Thus, it is possible the design encouraged
ambiguity to be resolved in Biased homophones based solely on Frequency and
Dominance information because the homophone was not introduced with Sarcasm. Both
Graded Salience and Relevance Theory predicted increased processing of one alternative
in the Balanced condition, and it was here we observed the increase of processing
meanings which were considered more Socially Contextually relevant.
When a homophone is directly mentioned using Sarcasm, the likelihood a speaker
intends a different interpretation to be considered is elevated. Thus, the design of
Experiment 3 introduced this manipulation, and also reduced the time from the
introduction of the homophone to the decision on a context, to encourage listeners to
utilize the prosody information further as a guiding heuristic in judgment of context.
Additionally, this further differentiates Graded Salience and Relevance Theory