“Double dissociations are convincing evidence for a modular organization of cognitive processing.”

Discuss this statement with reference to the English Past Tense debate and/or Language Development and Processing in Atypical Populations (e.g., Aphasia, Down Syndrome, Williams Syndrome, Developmental Language Disorder).

According to the APA Dictionary of Psychology (n.d.), a double dissociation demonstrates “two separable psychological or biological systems, such as differentiating between types of memory or the function of brain areas. One experimental variable is found to affect one of the systems, whereas a second variable affects the other”. Van Orden, Pennington, & Stone (2001) state that double dissociations can relate brain lesions to functional behaviours thus identifying modular systems of cognition. Fodor 1983 (as cited in Stojanovik, 2014) proposed the theory of modularity, which argues that aspects of cognition with semiautonomous functions are localised in different brain regions.

This essay discusses double dissociations and their implications for modularity in reference to the English past tense debate and atypical populations. Section 1, ‘English Past Tense’, will outline the theories of Parallel Distributed Processing and the Dual Processing ‘Words and Rules’ Theory, discussing them in relation to evidence from the stages of past tense acquisition, generalization of regular inflection to novel words, and brain mapping studies including LORETA analysis, PET and fMRI scans. Section 2 ‘Language Disorders’ will analyse double dissociations found in the language disorders aphasia, William’s syndrome, Down syndrome andDevelopmental Language Disorder. Finally, suggestions for further research will be given to investigate evidence for modularity theories.

English Past Tense

The English past tense has been identified as an important example of double dissociations, which can be used to explain cognitive processing (Pinker & Ullman, 2002). The English Past tense uses both irregular and regular verb inflections; the irregular exceptions also exhibit quasi-regularity meaning that some degree of regularity exists (McClelland & Patterson, 2002b; Pinker & Ulman, 2002).

These different strategies for inflecting tense are used to support opposing arguments of modularity. Some linguists argue for a dual processing model of words and Rules (WR) which is made up of a lexical module stored in memory and a computational module that allocates rules (e.g. adding -ed onto regular past verbs); the opposing theory is the connectionist model of Parallel Distributed Processing (PDP), which argues language is a result of a distributed single network operating in parallel (McClelland & Patterson, 2002a; Pinker & Ullman, 2002).

Acquisition of the English past tense has been identified as occurring in three stages forming a U-shaped curve of learning (Sampson, 1987). In stage 1, children can produce some common regular and irregular verbs. At stage 2, the ‘ed’ past tense rule is learned resulting in the over regularization of irregular verbs such as ‘go’ to ‘goed’ instead of ‘went’ as well as the correct use of past inflections -d, -id, and -t on unfamiliar verbs. Stage 3 results in continuing the regularization of unfamiliar verbs and relearning of unlearnt common irregulars (e.g., relearning ‘went’ as the past tense form) (Sampson, 1987).

This acquisition process supports the PDP model. The model works by gradually adjusting connection weights between a network or interconnected units where each input node is connected to each node of output (McClelland & Patterson, 2002a). The past tense is formed by first breaking down the word into its phonological features (or Wickel features), which enables the model to recognise inflection patterns from their similarities across words (McClelland & Patterson, 2002a) The Model is capable of learning by comparing outputs created from the original stem to the correct past tense form in a pattern associator network (McClelland & Patterson, 2002a). Outputs that are incorrect, result in the model modifying the weights at the responsible nodes whilst connections that are correct are strengthened (McClelland & Paterson, 2002 b). At the output, the Wickel features are decoded back into phonemes by which the past tense can be formed (McClelland & Paterson, 2002a). The model creates categories of words with similar features enabling it to predict a word’s past tense form and generalise to novel words; for example, categories such as ‘cling’ to ‘clung’ enables the model to generate ‘spling’ to ‘splung’ (Pinker & Ullman, 2002). This learning ability mimics the stages of developmental learning of the past tense, thus supporting the model’s application (McClelland & Paterson, 2002a). 

Sampson (1987) explained that researchers conducted an experiment inputting regular and irregular verb lists with varying frequency levels into the PDP model. The model produced similar results to the stages of past tense acquisition, with over-regularisation, relearning frequent irregulars, and using statistical learning based on phonological features to predict correct past tense forms (Sampson, 1987). For example, irregular verbs that do not change with the past tense often have a base form ending in either ‘t’ or ‘d’ such as ‘cut’. Research found verbs with this base formwere exempt from over regularization in child language acquisition and in the PDP model (Sampson, 1987). 

Whilst regularising novel words is common in individuals, the PDP model can fail to produce a past tense or produce unusual blends such as ‘mail to membled’ (Pinker & Ullman, 2002). Some recent models solve this by designing pattern associators based on the distribution of phonological space for regular and irregular verbs. If there are ‘categories’ of phonologically similar irregulars ( e.g., break, wake) and many phonologically dissimilar regulars (e.g. walk, dance), nodes can be implemented  to navigate dissimilar phonology which will then apply to novel words. Despite modifications, the PDP model does not work for languages where phonological fields are shared between regulars and irregulars (e.g., Hebrew) (Pinker & Ullman, 2002).

In opposition to the PDP model, Pinker and Ullman (2002) proposed the WR theory. This model suggests that irregular verbs are stored in a lexical module found in declarative memory whilst regular inflection is rule-based and stored in a separate syntactic module in procedural memory (Pinker & Ullman, 2002). Declarative memory stores information on facts and events, utilises the hippocampus and other medial temporal lobe structures to form new memories (Pinker & Ullman, 2002).  Irregulars are learned when a complex word’s affix and stem are not parsed and rather the word itself is memorised, storing it in declarative memory (Pinker & Ullman, 2002). Alternatively, motor control, cognitive skills and sequential information are stored in procedural memory, localised in the frontal cortex and basal ganglia as well as Broca’s area and anteria cortical regions (Pinker & Ullman, 2002). The syntactic module functions as a system which forms words and complex sentences from morphemes (Pinker & Ullman, 2002). This is different to the PDP model as past tense inflections are not learned through pattern association (Pinker & Ullman, 2002). Instead, the syntactic module possessing the past tense ‘rule’ is inhibited so that irregular verbs can be recalled from declarative memory (Pinker & Ullman, 2002).

Support for systems of modular organization can be found in brain mapping techniques. Lavric, Pizzagalli, Forstmeier, & Rippon’s (2001) study conducted a LORETA analysis, which determined left temporal cortex activity predominance for irregular verbs and right prefrontal cortex activity predominance for regular verbs. PET scans in Damasio, Gabrowski, Tranel, Hichwa, and Damasio (1996) and fMRI scans in Ni et al. (2000) have also linked lexical retrieval to activity in the left temporal region. This supports the dual-mechanism model which argues firstly that lexical retrieval and regular suffixation occur simultaneously, but then, one system is overridden (Lavric et al., 2001). Suffixation overrides lexical retrieval for regular items and vice versa for irregular forms (Lavric et al., 2001).

However, the predominance of right hemispheric activity for regular verbs is unusual given previous research has linked the left hemisphere to syntactic processing (Lavric et al., 2001). Instead, Lavric et al. (2001) discuss the possibility of a more complex process of regular verb formation occurring in the central executive of working memory. The right dorsolateral PFC has been found to hold a function in the manipulation of working memory and other executive functions (Lavric et al., 2001). Under the WR model this is plausible as the past tense suffix for regular verbs must be retrieved whilst the verb’s stem is held in working memory; alternatively, irregular verbs are retrieved then replaced by the past tense form and are not held in working memory (Lavric et al., 2001).

Further evidence for the WR theory comes from ERP studies which have found that brain activity like N400 (associated with lexical anomalies) occurs when irregular forms are incorrectly applied or not used (Pinker & Ullman, 2002). Similarly, a response alike to Left Anterior Negativity (occurring with syntactic violations) occurs with the omission of a regular suffix or an irregular substitute (Pinker & Ullman, 2002). This demonstrates the use of two systems, a lexical module for irregulars and a regular syntactic module (Pinker & Ullman, 2002).

Language Disorders

Aphasia

The subdivision of the language faculty into a dual-processing model gains support from Aphasic patients. Van Orden et al. (2001) demonstrates double associations found in types of fluent and non-fluent aphasia as well as how these symptoms map onto different areas of the brain.

Aphasia is an acquired language disorder which affects individuals’ language and speech, often a result of a head injury or stroke (Aphasia, 2021). Broca’s aphasia (non-fluent) displays symptoms of lost syntactic knowledge but preserves conceptual knowledge. Alternatively, Wernicke’s aphasia (fluent) preserves syntactic knowledge whilst conceptual knowledge is lost (Van Orden et al., 2001). Wernicke’s patients utter grammatically correct yet meaningless sentences whilst Broca’s patients understand speech but lack syntactic knowledge, omitting function words whilst keeping content words (Van Orden et al., 2001; What is Aphasia?, 2017).

Aphasia caused by brain damage in Wernicke’s area (temporal-occipital region) and damage to Broca’s area (left frontal cortex) suggest that syntax and lexical concepts are localised in separate regions of the brain (Van Orden et al, 2001). Marslen-Wilson & Tyler’s (1997) study demonstrated that left frontal lobe damage (in individuals with Broca’s aphasia) caused no impairment of irregular morphology but deficits in regular forms. Therefore, brain damage in these areas causes a dissociation between irregular and regular past tense neural works, thus lending support to a dual mechanism model of cognition (Marslen-Wilson & Tyler, 1997).

Alternatively, Bird et al. (2003) showed that controlling for phonological complexity eliminated the disadvantage for regularly inflected forms. A connectionist model of past tense employs knowledge of phonology and semantics; it argues that the disadvantage of regular inflection in individuals with non-fluent aphasia results from an impaired phonological system struggling with the greater complexity of the regular past tense (Bird et al., 2003) Bird et al. (2003) state that a central phonological deficit was responsible and not a result of a damage to a dual mechanism.

Developmental Disorders

 Double dissociations are also found in William’s (WS) and Down Syndromes (DS) (Stojanovik, 2014). William’s syndrome is a non-hereditary congenital disorder caused by chromosome 7’s partial deletion, language related symptoms include low cognitive ability but high verbal skills and excessive talking (William’s Syndrome Explained, 2019). Down syndrome is a genetic disorder caused by a copy of chromosome 21, which results in memory difficulties, a delay in language development, and cognitive impairment (World Down Syndrome Day, 2018).

These syndromes are often compared as they present different linguistic symptoms whilst expressing similar non-verbal IQs (Stojanovik, 2014). Symptoms of WS include an impairment of pragmatic language features (e.g., conforming to social rules) but unimpaired prosodic skills (rhythm and intonation) and syntactic skills (Belugi, Lichtenberger, Jones, Lai, & St. George, 2000; Laws and Bishop 2004; Stojanovik, 2006). Alternatively, DS is characterized by impaired syntactic and prosodic skills whilst pragmatic skills remain intact (Bellugi et al. 2000; Laws and Bishop, 2004)

The neuroconstructivist approach argues that both disorders result from changes to individuals’ epigenetic landscapes, which results in basic-level language impairment (D’Souza & Kermiloff-Smith, 2017). Waddington’s epigenetic landscape states that interactions between the genome and the developmental environment result in cascading affects influencing an individual’s phenotype (Tronick & Hunter, 2016). The neuroconstructivist theory states that neural structure becomes increasingly complex during development through external input (Thomas & Baughman, 2014). The theory argues that the genetic component of these disorders has a cascading effect on multiple language related domains, as opposed to a specific language module (D’Souza & Kermiloff-Smith, 2017). This approach opposes a language module (proposed by nativist theories) and adopts a more complex view of language where language related structures develop and change over time (D’Souza & Kermiloff-Smith, 2017; Stojanovik, 2014).

Another language disorder in which double dissociations are found is Developmental Language Disorder (DLD; also, Specific Language Disorder or SLI). DLD is a developmental communication disorder that results in symptoms including difficulty learning new words, an impaired grammar, and a delay forming sentences (Developmental Language Disorder, 2022).DLD is often compared to WS due to their double dissociations. Pinker 1999 (as cited in Karmiloff-Smith, Scerif and Ansari, 2016) famously quoted “The genes of one group of children [SLI] impair their grammar while sparing their intelligence; the genes of another group of children [WS] impair their intelligence while sparing their grammar”. This suggests that language is independent of intelligence and localised in different modules (Musolino & Landau, 2012).

Clahsen & Almazan’s (1998) study found dissociations between regular and irregular morphology in WS individuals, which included impaired irregular inflection and unimpaired regular inflection and syntax. Double dissociations were found when comparing results with SLI individuals, who displayed difficulty with regular inflection and syntax whilst sparing irregular morphology. WS individuals were able to access syntax whilst aspects of lexical information were impaired. These double dissociations support a dual mechanism theory of language, with an associative memory system and a rule-based, computational system (Clahsen & Almazan, 1998).

Despite this, WS and DLD have been found to not demonstrate clear cut double associations (Stojanovik, 2014). Thomas et al. (2001), found no difference in performance in WS individuals for regular and irregular morphology when controlling for mental age along a developmental trajectory. This provides evidence against a dual mechanism theory of modularity, as there is no evidence that WS individuals dissociate between irregular and regular inflection as claimed to be found in other studies (Thomas et al., 2001; Stojanovik, 2014). Moreover, Karmiloff-Smith et al. (2016) state that studies have found that intelligence in SLI individuals develops atypically with subtle impairments and WS individuals to have aspects of impaired language. This is the opposite of what Pinker stated as WS having “intact” language and DLD having “spared” intelligence, which undermines the claimed double dissociation between the disorders (Karmiloff-Smith et al., 2016).

Conclusion

In conclusion, double dissociations found in the English past tense and between language disorders demonstrate evidence for and against modularity theories. Stojanovik, (2014) suggests conducting developmental longitudinal studies on children with WS to investigate differences to typical populations as evidence has suggested that the language’s neural organisation may be different in WS. Pinker & Ullman (2002) emphasis that researchers should be careful not to limit their analysis of data into supporting one model for the truth could be combinatorial or more complex.

References

Apa Dictionary of Psychology. (n.d.). Retrieved March 31, 2023, from https://dictionary.apa.org/double-dissociation

Aphasia. (2021, November 18). Retrieved April 15, 2023, from https://www.nhs.uk/conditions/aphasia/

Bellugi, U., Lichtenberger, L., Jones, W., Lai, Z., & St. George, M. (2000). I. the neurocognitive profile of Williams Syndrome: A complex pattern of strengths and weaknesses. Journal of Cognitive Neuroscience, 12(Supplement 1), 7-29. doi:10.1162/089892900561959

Bird, H., Ralph, M. A., Seidenberg, M. S., Patterson, K., & McClelland, J. L. (2003). Deficits in phonology and past-tense morphology: What's the connection? Journal of Memory and Language, 48(3), 502-526. doi:10.1016/s0749-596x(02)00538-7

Clahsen, H., & Almazan, M. (1998). Syntax and morphology in Williams syndrome. Cognition, 68(3), 167-198. doi:10.1016/s0010-0277(98)00049-3

Damasio, H., Grabowski, T. J., Tranel, D., Hichwa, R. D., & Damasio, A. R. (1996). A neural basis for lexical retrieval. Nature, 380(6574), 499-505. doi:10.1038/380499a0

Developmental language disorder. (2022, December 27). Retrieved April 8, 2023, from https://www.nidcd.nih.gov/health/developmental-language-disorder

Down syndrome. (2018, March 08). Retrieved April 15, 2023, from https://www.mayoclinic.org/diseases-conditions/down-syndrome/symptoms-causes/syc-20355977#:~:text=The%20genetic%20basis%20of%20Down%20syndrome,-There%20are%2023&text=In%20Down%20syndrome%2C%20there%20is,partial%20copy%20of%20chromosome%2021.

D'Souza, H., & Karmiloff‐Smith, A. (2016). Neurodevelopmental disorders. Wiley Interdisciplinary Reviews: Cognitive Science, 8(1-2). doi:10.1002/wcs.1398

Karmiloff-Smith, A., Scerif, G., & Ansari, D. (2003). Double dissociations in developmental disorders? theoretically misconceived, empirically dubious. Cortex, 39(1), 161-163. doi:10.1016/s0010-9452(08)70091-1.

Lavric, A., Pizzagalli, D., Forstmeier, S., & Rippon, G. (2001). A double-dissociation of English past-tense production revealed by event-related potentials and low-resolution electromagnetic tomography (Loreta). Clinical Neurophysiology, 112(10), 1833-1849. doi:10.1016/s1388-2457(01)00615-0

Laws, G., & Bishop, D. V. (2004). Pragmatic language impairment and social deficits in Williams Syndrome: A comparison with Down's syndrome and specific language impairment [Abstract]. International Journal of Language & Communication Disorders, 39(1), 45-64. doi:10.1080/13682820310001615797

Marslen-Wilson, W. D., & Tyler, L. K. (1997). Dissociating types of mental computation. Nature, 387(6633), 592-594. doi:10.1038/42456

McClelland, J. L., & Patterson, K. (2002a). Rules or connections in past-tense inflections: What does the evidence rule out? Trends in Cognitive Sciences, 6(11), 465-472. doi:10.1016/s1364-6613(02)01993-9

McClelland, J. L., & Patterson, K. (2002b). ‘words or rules’ cannot exploit the regularity in exceptions. Trends in Cognitive Sciences, 6(11), 464-465. doi:10.1016/s1364-6613(02)02012-0

Musolino, J., & Landau, B. (2012). Genes, language, and the nature of scientific explanations: The case of Williams syndrome. Cognitive Neuropsychology, 29(1-2), 123-148. doi:10.1080/02643294.2012.702103

Ni, W., Constable, R. T., Mencl, W. E., Pugh, K. R., Fulbright, R. K., Shaywitz, S. E., . . . Shankweiler, D. (2000). An event-related neuroimaging study distinguishing form and content in sentence processing. Journal of Cognitive Neuroscience, 12(1), 120-133. doi:10.1162/08989290051137648

Pinker, S., & Ullman, M. T. (2002). The Past-Tense Debate: The past and future of the past tense. TRENDS in Cognitive Science, 6(11), 456-463.

Sampson, G. (1987). [Review of Parallel Distributed Processing: Explorations in the Microstructures of Cognition, by D. E. Rumelhart, J. L. McClelland, & The PDP Research Group]. Language, 63(4), 871–886. https://doi.org/10.2307/415721