• No results found

Chapter 6 General Discussion

6.5. Potential limitations

In this section, I will address three potential limitations of the research presented in this the- sis. These include: the suggestion that explicit memory tasks may not be the most efficient and ap- propriate for revealing specificity effects; the relatively constrained implications arising from the glimpse account; the failure to replicate the original sound specificity effect reported by Pufahl and Samuel (2014); and finally, the limitations in explaining what I refer to as the “when, how, and where” conundrum, that deals with the main questions revolving around sound specificity effects.

6.5.1. The memory task

When it comes to the methodology used in measuring indexical and sound specificity effects, the memory test has been a matter of controversy. As discussed in Chapter 2, the debate has fo- cused on two main issues: 1) the reliability of the explicit vs. implicit tasks: implicit tasks have been found more sensitive; and 2) the suitability (validity) of these tasks to measure specificity ef- fects that have implications for lexical representations in memory: explicit tasks are considered to tap into episodic memory, rather than the lexicon (Goh, 2005; Pufahl & Samuel, 2014).

With respect to the first matter, the results in this thesis indicates the opposite pattern. Name- ly, the recognition memory task (explicit) was successful in revealing the voice specificity effect (Experiment 1: Chapter 2), as well as sound specificity effects in two cases (Chapter 3: Experi- ment 3, and Chapter 4: Experiment 4). On the other hand, the identification memory task used in one study failed to reveal the anticipated sound specificity effect (Chapter 5: Experiment 7).

The issue of suitability is more complicated. As I argued in Chapter 2, the validity of the dependent measure depends on theoretical assumptions about the structural organization of episod- ic and lexical memory. Judging from the ongoing debate in the literature, it is still unclear what constitutes strictly lexical vs episodic memory, how these two types of memory interact with each other, and whether there is a need for drawing boundaries at all. Perhaps the best arguments to be made in favour of using a recognition task are: 1) ultimately, the task is about the word per se, and recognising the word entails lexically accessing it; and 2) both explicit and implicit tasks have been used extensively to measure specificity effects and inform models of spoken word recognition.

6.5.2. The fragility of the glimpse account

As noted earlier in the discussion, the case of highly contrasted glimpses provides evidence in support of the possibility that the degraded versions of the words are retained in memory, with- out the actual sounds being encoded. In doing so, it offers an alternative explanation to the claims

Approach can be adapted to accommodate speech-extrinsic auditory information present in its auditory context. This adaptation leads to a triple-route approach, in which linguistic, indexical and speech-extrin- sic information can be processed simultaneously, affecting recognition/comprehension of the spoken word. I have not dissociated between recognition and comprehension in this instance, since for the sake of simplicity, I am assuming that recognising the word entails comprehending it. However, it should be pos- sible to modify that level, if required. The parts highlighted in blue represents the third route that I posit adding to Sumner et al. (2014)’ approach (the components in grey), in order to accommodate the present sound specificity effects. The dotted lines in the case of the third route represent the evidence that speech- extrinsic sound/noise information may not always make it to affect processing, given the seemingly tran- sient and context-selective nature of the respective specificity effects.

that mere co-occurrence is sufficient for the emergence of a sound specificity effect. However, this account is fragile in two respects:

1) The calculated glimpse proportions from the two sounds were quantitatively different in both the experiment that revealed an effect (Chapter 3: Experiment 3) and the one that did not (Chapter 3: Experiment 2 (A and B)). I argued that the crucial factor in explaining the appearance of an effect was the quality of glimpses, rather than their quantitative measure. Specifically, in the case of an effect, the glimpses of the same word resulting from the masking of the two sounds were both quantitatively and qualitatively different, due to the masking contrast created as a result of a joint change in both the sound pitch and its temporal overlap with the word.

2) The absence of a study where only the degraded versions (glimpses) of the words are played as stimuli, instead of the word-sound pairs, also contributes to the limitation of the account. Similarly, another study in which the same sound would have led to different glimpses of the same word (i.e., the same car horn sound, in different temporal overlaps with the word) could have strengthened the case for the encoding of the glimpses in memory.

Therefore, despite making a plausible case for the role of energetic masking in sound specificity effects, this account needs more elaboration and additional empirical work.

6.5.3. The context uniqueness puzzle

The failure to replicate the original sound specificity effect found by Pufahl and Samuel (2014) despite using the same environmental sounds, the same encoding and memory tasks, the same filtering technique, and a similar (though, not the same) experimental design, is a puzzling result and could be considered as a potential weakness of the present work (Chapter 5: Experiment 7). However, it is important to consider the following two observations that arise from both these studies. First, the lack of an effect does not constitute a problem only for the present work, but also for the Pufahl and Samuel’s original study. Given the novelty of the effect, any failure to replicate it weakens the strength of claims made with regard to its implications. Second, taken together, the pattern of results in these studies suggests that these types of effects are small and fragile in gener- al.

6.5.4. The “when, how, and where” conundrum

This issue is complex and almost impossible to address within a single study. As already mentioned, the results in this thesis provide more evidence for the when question regarding the emergence of sound specificity effects. Specifically, these results identified conditions that con- strain when the speech and sound stay together.

As to where in memory these effects may reside, this work provides little direct evidence. Although the implications of specificity effects in general have been frequently interpreted in terms of lexical representations, it is far from clear whether the sounds, or even voices, co-exist with the words in long-term lexical memory I suggested several ways in which these effects could be ac- commodated in the memory system by providing visualisations, but these are speculative at this stage.

With respect to how speech-extrinsic auditory information is processed alongside linguistic and indexical information, again, the present work provides little-to-no direct evidence. Perhaps the only insight is the integrality effect, as reminiscent of the common fate principle of integrating dif- ferent sources of information that follow the same pattern of change over time.

I addressed this question indirectly in section 6.4.3.3, where I discussed how the sound specificity effect could be accommodated in a recently proposed dual-route theoretical approach of speech perception (Sumner et al., 2014). This account attempts to explain how different sources of infor- mation in the context of spoken words, namely linguistic and indexical, can be integrated in paral- lel during processing.

As it currently stands in the literature, a better understanding and interpretation of sound specificity effects in the face of this multi-faceted conundrum will need to be supported by consid- erably more evidence in future studies.