The prehistory of speech and language is revealed in brain damage

Abstract

The aim of this paper is to develop further the idea that symptoms that emerge in speech and language processing following brain damage can make a contribution to discussions of the early evolution of language. These diverse impairments are called aphasia, and this paper proposes that the recovery of a non-fluent aphasia syndrome following stroke could provide insights into the course of the pre-history of human language evolution. The observable symptoms emerge during recovery, crucially enabled by (dis)inhibition in parallel with a range of impairments in action processing (apraxias), including apraxia of speech. They are underpinned by changes in cortical and subcortical status following brain damage. It is proposed that the observed recovery mimics ontogenic and phylogenic processes in human speech and language. The arguments put forward provide insights tending to support the motor-gestural model of speech and language evolution.

This article is part of the theme issue ‘Reconstructing prehistoric languages’.

1. Introduction

Research in the evolution of speech and language has taken many approaches in recent decades. However, there has been little investigation of the contribution that pathologies of speech and language can add to the general debate. A limited number of studies have recently explored the role of a range of disorders and conditions caused by some kind of brain damage [1–5]. In the current volume, four related cognitive disorders/conditions (schizophrenia, autism, synaesthesia, Tourette's syndrome) are all shown to display altered patterns of inhibition/disinhibition, where reactive aggression (verbal or physical) and altered cross-modality cluster together [4]. These, sometimes co-occurring, conditions involve the same cortical and subcortical brain circuits that evolved relatively recently. The authors conclude that language evolution was enabled by a fine-tuning of (dis)inhibition in the cortico-subcortical circuits, particularly cortico-striatal networks [4]. A different approach based on neurological data proposed three broad stages in language evolution: primitive communication systems (animal communication); systems using sound combinations (lexicon) but without grammar; and advanced systems including word-combinations (grammar) [5]. Many evolutionary linguists believe that a form of proto-speech and pre-syntactic proto-language emerging in pre-history paved the way for the emergence of recursive syntax and modern human language [6]. A proto-language is made up of utterances comprising two or three words without syntactic structure beyond word order. It has been observed that some of the common symptoms of aphasia resemble proto-language, purely in terms of their linguistic structure [1,2].

This paper aims to further the argument that some specific and related impairments in language (aphasia), action (apraxia) and speech action (apraxia of speech) following brain damage may provide insights that can contribute to the pre-history language debate. More specifically, the paper explores the recovery pattern in selected common symptomology during recovery from stroke. Cognitive factors, particularly (dis)inhibition, attention and working memory impairments, have a significant impact. Underlying the behavioural patterns of change are impaired neuro-cortical circuits running from relatively recently evolved neocortex to the ancient subcortical structures of the basal ganglia [7,8] and limbic system [9,10]. The neural circuits that link the inferior frontal cortex with the basal ganglia and other subcortical structures are engaged in the regulation of motor control, including speech motor control and syntax [8].

The core argument proposed is that this evolution of symptoms with recovery uncovers prehistoric fossils that might contribute to models of speech and language evolution.

This argument is informed by a range of relevant models of brain, speech, language and action evolution. First, it is grounded in the claim that language has a motor origin and that communicative gesture and tool use preceded speech and language in evolution. This theory is currently among the most discussed in evolutionary studies of speech and language [8,11–15]. Evidence is presented that a range of aphasic and apraxic symptoms can be explained by damage to the interactional mechanisms of the action-motor system and the speech and language system. The engagement of the separate parts of the left inferior frontal cortex (traditionally Broca's area, BA44/45) and related frontal structures in this system is explored.

An over-arching neurological framework for this exploration is Hughlings Jackson's model of the central nervous system [16], which is a fundamental base for modern human neurology and the evolution of the brain [17]. In his levels of representation model, brain functions are organized and represented hierarchically at different levels within the central nervous system; an arrangement established by evolution and regulated by inhibition. With cortical damage, evolutionarily older subcortical systems can be released from the inhibitory control of later evolving higher cortical levels. Jackson saw cognitive deficits as evidence of the inverse of evolution, dissolution. (Dis)inhibition is the crucial controlling process in the model. For him, symptoms, for instance, aphasic speech automatisms (discussed in detail below), are expressions of lower levels released from the inhibition of higher cortical levels, which could represent fossils of evolution [2]. Informed by his work with aphasia, Jackson originated the idea that language can be propositional and referential or it can be non-propositional and automatic. Propositional and referential speech makes reference to persons, objects and ideas, etc. [18], whereas speech that is non-propositional may be automatic or formulaic [19]. This distinction between automatic and voluntary uses of language is now well established, with wide evidence that two forms of language can be differentially impaired in brain damage [18,20]. The distinction may reflect the extent of cognitive control/resources required to build representations due to simpler or more frequently used computations. Jackson himself was the first to suggest that the right cerebral hemisphere was responsible for the production of automatic and non-propositional language and the left hemisphere for the propositional and referential [16].

The crucial role of (dis)inhibition in the evolution of speech and language has been examined through exploration of a number of conditions/disorders [1–4]. It was noted above that schizophrenia, autism spectrum disorder, synaesthesia and Tourette's syndrome display altered patterns of inhibition/disinhibition with the proposal that speech and language evolution was enabled by a fine-tuning of (dis)inhibition in cortico-subcortical circuits [4]. The functions of (dis)inhibition and attention in the patterns of change observed in speech and language impairments is pivotal to the general proposal presented here.

2. Non-fluent aphasia and its recovery

Patterns of symptoms change with recovery and the emergence of aphasic symptoms is determined to varying degrees by the cause, site and extent of brain lesioning [21], and by the severity of the aphasia and the operation of neuroplasticity and compensatory strategies. These symptoms have been traditionally used to classify aphasia into different types, and there are a number of classifications that have been proposed, including the classical Wernicke–Lichtheim model [22]. Classifications come with theoretical baggage, and many prefer to avoid them, opting to limit classification to a non-fluent form, occurring mostly from damage to inferior frontal areas of the left brain, and a fluent form arising mainly from damage to left posterior-superior temporo-parietal areas. Fluency is a major divider between the classic Broca's (non-fluent) and Wernicke's (fluent) aphasias. The parameters of speech production that distinguish non-fluent from fluent aphasia are the rate of utterance, phrase length, articulatory agility, speech effort, stereotypes or speech automatisms and apraxia of speech (hence, AoS), the latter being the largest predictor of non-fluency [23]. We are concerned with the recovery of non-fluent aphasia, which is sometimes referred to as (Big) Broca's aphasia, motor or expressive aphasia, because, as the parameters described above suggest, in a large proportion of cases it represents impairments to the processes underlying both speech action and pre-articulatory language production.

What follows is an examination of the closely related major classic symptoms of a non-fluent aphasia syndrome: speech automatisms, agrammatism and AoS, and a mapping of their emergence and progress over time during recovery. With recovery, these trajectories mimic ontogenetic processes of change in development, and unsurprisingly a number of theorists have made comparisons between acquired aphasia and child speech and language to examine the claims of the regression hypothesis, that a dissolution of language in aphasia in the reverse order to which it is acquired can be observed [24–27]. This pathological variant of ontogenesis could be seen as a form of pathogenesis.

(a). The core symptoms of the non-fluent aphasia syndrome

(i). Aphasic speech automatisms

Speech automatisms are classic core symptoms of non-fluent aphasia [28]. Indeed, the first case to be described as having ‘Broca's aphasia', Louis Victor Leborgne [29] had what Broca described as ‘aphemia’, what today most call AoS; a disorder Broca emphasized was not of language, but of speech. Leborgne had little speech except famously the repeated non-lexical speech automatism, /tan, tan/. Not generally known is that Leborgne had a second automatism, that was the lexical speech automatism, Sacre Nom de Dieu (Sacred name of God), made up of real words [30]. He also had impairments in communicative limb gesture.

A potential role for speech automatisms in the evolution of speech and language was explored in earlier work [1,2]. Research has revealed two types, the lexical and the non-lexical speech automatism, introduced above [20,31]. A large corpus of lexical and non-lexical automatisms was analysed, which included emotionally charged pronoun + modal stems like, ‘I can…', ‘I can't…' and ‘I want…', as well as many expletives and oaths like, ‘bugger, bugger' and ‘Sacré Nom de Dieu' [20]. Non-lexical examples consisted predominantly of always repeated consonant-vowel (CV) syllables like /ba, ba/, /da da/ and so on. Automatisms are mostly made up of high-frequency words and phonemes, and they do not break the rules of syntax or the phonotactic rules of the language, despite their limited contents. The pronoun + modal/auxiliary subtype was the largest, and linguistically most complex.

Individuals with automatisms were mainly classified as having Broca's or global aphasia, what we can now call Big Broca's aphasia (see below), and were severely impaired with no or little other speech. It is proposed [20] that both subtypes of automatism are produced holistically and ready-made, without input from componential language processes. The utterance is usually fluently and invariantly produced, emerging the same each time speech is attempted. The speaker has little or no other speech and there is no sign of AoS within the utterance itself, which is probably masked by the holistic nature of the automatism. The invariability in production suggests that the utterance is stuck in a loop, in a buffer [27]. A total of 14 people from the sample of 75 with an automatism had more than one automatism and for half of those the second utterance was phonologically, syntactically or semantically related to the first utterances [20]. The appearance of a second and third variation on the first automatism is a sign that recovery is taking place. Many of the lexical subtype would be considered as examples of formulaic language [19]; if, that is, they had been produced voluntarily in normal discourse. It seems likely that utterances are released from the inhibition of higher centres, and research, discussed in some detail below, suggests that automatisms emerge from subcortical neural levels.

As noted above, non-lexical automatisms consist predominantly of repeated consonant-vowel (CV) syllables (e.g. /tan tan/). It has been proposed that syllabification is hard-wired in the brain, particularly in the left brain [32], suggesting that each consonant and vowel position is associated with a specific cell assembly network. The CV syllable is particularly robust and can survive the severest brain damage, including the complete surgical removal of the left hemisphere [33]. Lesion evidence suggests that, if syllabification is hard-wired, it is likely to be hard-wired either subcortically or is diffusely represented throughout the brain [33,34].

TThe meeting of phonation and articulatory activity would seem to be a particularly crucial early event in the pre-history of speech and language evolution. The Frame/Content theory was proposed for the early evolution of speech [25], from the close–open cycle of the mandible that originally evolved for mammalian chewing, sucking and licking. This basic mandibular cycle underpins the speech frame which was exapted and paired eventually with vocalization to become the basis for CV combinations into syllables. A comparison of infant babbling and the aphasic non-lexical automatisms described above, revealed significant parallels with infant babbling, providing support for the general claim that high-frequency CV syllable utterances may have been among the earliest human utterances in the pre-history of language [26].

The idea that damaged frontal attentional–inhibitory processes underlie the failure to suppress production of automatisms was proposed [28], and the re-emergence of attentional awareness, shown by attempts to check the utterance, seems critical for recovery. For those who make some recovery, the next stage is initially laboured, agrammatic speech with AoS. Attentional disinhibition in automatism production was recently reported in progressive aphasia/AoS-limb apraxia [35].

(ii). Agrammatism

Agrammatism is a core symptom of non-fluent aphasia. It describes impairments in the use of syntax and sentence processing. Opinion is divided as to whether it is a primary disorder of syntax per se or emerges from the impact of compensatory processes, neuroplasticity, impaired or inaccessible cognitive resources. It manifests itself differently across individuals and languages ([36,37], for contemporary reviews). Agrammatism in English is characterised mostly by omitted grammatical words low in information (grammatical ‘function' words), like ‘and’, ‘on’, ‘the’, and impaired verbal inflections. The resulting pattern is called telegraphic speech, because it resembles a now obsolete telegraph or telegram where each word was charged individually. To reduce expense, a user would limit themselves to the content words, leaving out the function words. The metaphor has significant theoretical implications. It has been argued that agrammatic speakers employ economy of effort, where the speech production system is so impaired that the speaker's options are limited to producing only the important information-carrying elements—the content words [38]. The principle of economy was introduced by George Zipf [39], who termed it the principle of least effort (maximum benefit with minimum cost) and regarded it as governing all human behaviours. Adaptation theory proposes that agrammatic surface structure does not directly reflect the underlying deficit, but is the result of unconscious systemic adaptation within the language system. Agrammatic output was compared with that of the developing elliptic speech of 2- to 3-year-old boys and significant similarities were found in the two forms, suggesting the operation of an unconscious choice by agrammatic speakers for normal elliptical speech that functions to prevent computational overload [40].

A motor basis to agrammatism has been proposed by a number of aphasiologists [38,41–43]. A ‘stress-saliency' hypothesis was developed for ‘motor agrammatism', based on experimental findings that the words omitted are the unstressed function words occurring in sentence-initial positions (e.g. The, A) [42]; such speakers have particular problems with initial mobilization of the motor-speech production system.

It has been contended that the underlying pattern of deficit in non-fluent aphasia is an interaction of agrammatism and AoS [43]. There is a convergence of this relationship in the KE family, some members of which share inherited AoS, buccofacial apraxia and syntactic deficits due to impaired expression of the FOXP2 gene for family members with the condition, but not those without it. The buccofacial apraxia and AoS constitute core elements of the condition [44–46]. FOXP2 may have contributed to the evolution of human speech and language by adapting and tuning cortico-bulbar circuits during evolution [46].

The KE family findings suggest that a core inter-relationship between facial action, speech-action and syntax evolved neural programming stages within Broca's complex, providing evidence in support of theories claiming a motor basis to syntax [e.g. 7–9,45].

Thus, we observe a staged recovery process from (i) single CV utterance or a lexical automatism stuck unchanging in a buffer, partly due to disinhibition; (ii) a second and third utterance emerging, demonstrating some control over the emergence of the first utterance and the strengthening of attention and cortical inhibition; and (iii) early proto-stages of agrammatism. With continued recovery in cognitive control processes, the agrammatism reduces further, articulatory control improves and a mild naming impairment may persist.

(iii). The apraxias and apraxia of speech – impairments in motor-speech planning and programming

Apraxia refers to impairment in the ability to perform purposeful, voluntary movements, although there is no neuromuscular disorder. It represents damage to the processes that transform mental phenomena in mind (language) into physical reality in body (e.g. speech). It is understood as an impairment in (i) the planning, (ii) the initiation and (iii) the coordination of an action or action sequences in the absence of neuromuscular impairment [47,48]. A range of disorders of action processing can occur, including limb, buccofacial-oral, gait, as well as AoS. Limb apraxia can impair the production and comprehension of imitation, pantomime, symbolic gesture and tool use. Liepmann [49] was the first to describe apraxias in detail, including AoS, and to develop a classification. He emphasized a marked feature of all apraxias is a disassociation of voluntary and involuntary actions. Cognitive models of action have developed employing modular information processing stages. These modules process action perception and expression. Modules include an action semantics, input and output action lexicons, various buffers, etc., ending in physical action implementation [47,48]. AoS can impair initiation of speech, with observable articulatory groping and searching behaviour, as well as prosodic disturbances. Severity can range between mild to very severe. Involuntary and automatic actions are relatively preserved. Thus, someone with an apraxia that affects buccofacial-oral mechanisms may be unable to lick his or her lips to command or to imitation, but will automatically lick their lips when drinking. Speech can feature omissions, substitutions, additions and distortions of speech phones and problems of sequencing the order of phone production, together with disturbed prosody. Classic features are that errors increase as the complexity of the motor task increases, as seen in most apraxias. Some of these factors contribute to the idea that common CV–CV syllable structure that was established previously is a potential prehistoric feature.

Thus, in AoS, vowels are easier than consonants to produce and single consonants are easier than clusters of consonants; errors occur more on low- than high-frequency consonants; errors increase with word length and occur less in automatic than in more propositional speech [50]. These errors in selection and combination of phones to form syllables, and the coordination of them for execution, are all influenced by the complexities of the motor task. It is not enough to simply have the ability to utter syllables to produce speech. As discussed in more detail below, selection and combination processes are essential to turn individual articulatory gestures into coordinated speech and seem probable early features of a speech production system in the process of evolving.

Liepmann [51] saw AoS as a variety of limb apraxia and suggested that ‘the word limb here, refers to the tongue, palate, and oral mechanism' (p. 56). Speech, he claimed, ‘is a parasite making use of a preformed sensory-motor mechanism which is also involved in other (Liepmann's italics) functions, particularly that of eating; it likewise shares in any impairment sustained by this same mechanism' (p. 56). This view was further developed by Kimura [52–54] proposing that the close relationship between speech and praxic impairment is explained with reference to the finding that both language and praxis are processed by the left hemisphere, and speech is highly dependent upon praxic skills. She argued that it is complex motor programming that the left hemisphere is specialized for, not language. The evolution of the capacity to speak is built on an earlier pre-existing capacity for action and gesture. Serial control of behaviour plays a central role in Kimura's hypothesis: both the sounds of language and the sequences of action are organized temporally and ‘the left hemisphere is particularly well adapted, not for symbolic function per se, but for the execution of some categories of motor activity which happen to lend themselves readily to communication' [52, p. 154]. Lieberman [7,8] emphasized the close evolutionary relationship between motor control, syntax and cognition, with the parallel evolution of the cortico-basal neural circuits. The basal ganglia, he suggests, constitute a sequencing engine that accesses stored motor pattern generators in other parts of the brain.

Greenfield [11] claimed that in the first two years of life Broca's area is a common site for the development of speech production and also the ability to manipulate objects and tools. Speech production and object manipulation develop more autonomously with increasingly modular organization after 2 years of age. More recently [55], it has been suggested that the mechanisms within the brain for both praxis and language share an evolutionary affiliation with the mirror neuron system, represented in Broca's area. An expanded mirror neuron system hypothesis based on imitation was proposed, the later stages of which involve the evolution of manual proto-sign communication to a proto-language, which combined proto-sign and proto-speech, before the emergence of full language competence.

Pure apraxias are rare, with apraxias most often appearing in various combinations, such as aphasia with AoS, or oral with limb apraxia. Buccofacial-oral apraxia (affecting the face, tongue, larynx, lips, pharynx, etc., in non-speech actions) is found in the majority of those with AoS also [56].

The main characteristics of the non-fluent aphasia syndrome have been outlined, but not all those effected will experience all the symptoms. The symptoms, including the cognitive and behavioural, do not all necessarily emerge in all individuals. For example, lexical and non-lexical speech automatisms seldom mix and those with the non-lexical subtype are significantly less likely to make good recoveries and will only evolve later symptoms if recovering.

3. The neural architecture of non-fluent aphasia and apraxia of speech

Convergence between data from lesion studies, structural and functional imaging and electrophysiological studies is the goal of much research in neuropsychology and neurolinguistics. We now have a broader understanding of the responsibilities of Broca's area than in 1861, and so it is with speech, voice and language functions in general, which are known to have wide representation throughout the left, right, anterior, posterior cortical and subcortical brain [57–60]. Our brief sketch concentrates on the neural substrate relevant to the recovery of non-fluent aphasia, AoS and limb and buccofacial apraxias: namely, the left inferior frontal cortex and related frontal and subcortical neural structures. Intensive research into the responsibilities of this cortico-basal network demonstrates its essential functional control of speech, language and action. The evolution from older primitive subcortical action processing to more finely tuned neocortical control, allowed the emergence of modern speech and language from limited and elemental action elements, including actions of what would evolve into the phono-articulatory system.

A large seminal lesion study using computerized tomography concluded that Broca's aphasia did not result from a lesion limited to the classical Broca's area, but resulted from a larger lesion involving the area of supply of the upper division of the left middle-cerebral artery. This produced a global aphasia–AoS affecting all or most modalities [61]. Damage included the operculum, the third frontal gyrus, anterior parietal region, insula and both sides of the central Rolandic fissure, extending deep into the underlying white matter and subcortical structures. This extensive damage produced what the authors called ‘Big Broca's aphasia’ or the operculum syndrome, with a severe and persisting AoS with either mutism or speech automatisms, with the later emergence of agrammatism and severe reading and writing problems, if there is any recovery.

Imaging and electrophysiological studies with neuro-typical volunteers have identified an enlarged left frontal region engaged in various aspects of speech, language and action, including, but not limited to, Broca's area in inferior frontal cortex (BA 44, pars opercularis and 45 pars triangularis). It includes the supplementary motor area in ventral premotor cortex (BA 6). In this enlarged Broca's area, called ‘Broca's complex’ [62], activation is seen while participants engage in semantic, syntactic and phonological processing and speech planning. A cognitive selection and binding or unification role for Broca's complex is identified, with a particular role in maintaining information online while the binding operations take place, achieved with working memory. This combination process is invoked by Goldenberg [48] who argued that underlying all apraxias is a fundamental segmentation and combination process impairment. Segmentation and combination are, he suggests, the basic components of cognition. In addition, the inferior frontal cortex is considered to be the area where the articulatory loop component of working memory engages in storage and manipulation [63,64], and the region has been identified as where the learnt procedural knowledge for the production of the speech sound patterns is assembled and stored in development. It evolved into the neural basis of speech motor planning [65].

As noted earlier, it is possible that an adaptation involving FOXP2 contributed to the tuning of the cortico-bulbar circuit [39], and it is claimed that the gene governs the embryonic development of the basal ganglia and other subcortical structures in the network [8]. Subcortical and limbic structures have been implicated in automatism production, and the basal ganglia have received particular attention, which has strong interconnections with the limbic system [66]. Descriptions of the limbic system vary in the structures included, but the hippocampus, amygdala and ventral striatum are primary [10]. Damage here has been implicated in the production of speech automatisms, swearing (a common aphasic automatism), coprolalia in Tourette's and pallilalia [67]. Tourette's may result from cortical and limbic system imbalance, where the coprolalia and other behaviours emerge uninhibited by the cortical system [3,4].

Deficits in propositional speech and language arise most commonly from left hemisphere damage. However, dissociation between automatic and propositional speech have been reported with basal ganglia damage [67–69], causing impairments to automatic non-propositional speech.

Basal ganglia damage appears to be crucial for the production of speech automatisms [70]. Twelve of 26 aphasic participants had either a lexical or non-lexical speech automatisms and scans showed all 26 had basal ganglia involvement. Neither type occurred in the participants without basal ganglia damage. Additionally, automatisms did not occur in those with only subcortical (including basal ganglia) damage. Thus, a large left hemisphere lesion involving inferior frontal cortex and basal ganglia appears to be required for an automatism to emerge [70]. There is a confusing paradox here: how can lesions of the basal ganglia have some special role in both causing automatisms and causing impairments in automatic non-propositional tasks, like counting, prayers, etc.? Jackson famously pointed out that impairments emerging from damage to specific parts of the brain do not indicate what the normal function of that brain part might be. A lesion cannot show what a neural component does, but what the entire brain does without it. In the current case, we find automatisms caused by basal ganglia and inferior frontal cortex lesion, and problems with automatic speech production in basal ganglia lesions without inferior frontal damage. Again, the actions of (dis)inhibition are central in (i) causing automatisms and (ii) causing impairments in automatic non-propositional tasks [67–70]. This is not to ignore the recognized limitations of the lesion model of brain function in its interpretative power.

An extensive left anterior cortical and subcortical neural network underlies the production of speech and language, involving Broca's complex—left inferior frontal cortex, supplementary motor area, insula, basal ganglia and thalamus. Damage to this network from stroke or progressive disease can cause the emergence of symptoms that can suggest ways in which speech and language may have evolved.

4. Discussion

It has been suggested that brain damage can reveal patterns in symptoms mimicking evolutionary processes in evolution, providing potential insights for theories of speech and language evolution. Evidence suggests that the common patterns observed in the emergence of symptoms during recovery of non-fluent aphasia and AoS are underpinned by changes in the cortico-subcortical network. These processes point to a course for evolution from more ‘primitive’ forms of communication to more complex.

The arguments presented are grounded in a prominent neurological model of the evolution of brain and language featuring powerful (dis)inhibitory controls and provides support for a motor-gestural history for language. In turn, a scenario is supported in the pre-history of language of an ancient, primitive, subcortical brain, which seems to be heavily involved in automatic, emotional, sexually charged vocalizations. The patterns described appear to be pointing to where speech and language began. It also suggests where it ends in the case of progressive non-fluency/AoS—in cortico-bulbar syndrome in the ancient subcortical areas of the brain.

Aphasia is notoriously heterogeneous, hampering research efforts. Symptoms and syndromes and their putative relationship to cognitive-linguistic domains take us only so far, especially given the major differences that exist between underlying neuro-pathologies. Identical symptoms can result from different causes, and indeed do. Aphasia classifications are unreliable and the non-fluent recovery syndrome sketched also presents variably: not everyone recovering from non-fluent aphasia will have an automatism, AoS or agrammatism, for instance. Agrammatism is not a unitary condition, and the evidence suggests that it does not represent a pure impairment of syntactic processing. It can present variably in different individuals, and from acute to chronic stages, changing the pattern of syntactic deficit from severe to milder forms, as well as manifesting differently in different languages.

While the arguments proposed appear to support motor-gesture characterizations of speech and language evolution, other models are available. It is for future attempts to test the suggestions made here in the light of other proposals.

5. Conclusion

It has been argued that the emerging symptoms of aphasia and AoS in recovery from brain damage provide insights into where speech and language began. We outlined the parallels pertaining between neural damage and the emergence of cognitive-linguistic symptoms, and the central role of (dis)inhibition. Recovery progresses from simple to complex; from proto-sign and proto-speech/language to recursive syntax; from automatic to propositional. We noted limitations to the approach taken. Speech and language impairments following brain damage are particularly heterogeneous, and it was noted that types, subtypes, syndromes and the symptoms that define them, have never remained static in the history of aphasia and are still the subject of active research. Examining symptoms can go only so far and a computational approach focusing on specific deficits could provide further insights. Lesioning a computational model to mimic features of impaired speech and language could provide a basis for questions about the evolution of speech and language.

Nonetheless, there appear to be fossils revealed by brain damage that can make a contribution to debates on the evolution of speech and language in pre-history. It has been argued that the patterns of recovery of speech and language in non-fluent aphasia/AoS can mimic the phylogenic processes of language, in a form of pathogenesis. As part of this process, fossils can emerge following damage to various neural regions and networks. Strong candidates appear to be repeated non-lexical speech automatisms, among the most primitive human utterances. These phono-articulatory gestures constitute rudimentary CV–CV structures. These emerging processes during evolution were, in Liepmann's words, ‘parasites', built on a much older pre-existing motor capability. This capability was exploited by early humans, beginning with an early proto-sign formed for gestural communication leading to rudimentary speech action.

Data accessibility

This article has no additional data.

Competing interests

We declare we have no competing interests.

Funding

We received no funding for this study.