A bird's eye view of human language evolution

“…sequences for this bird’s song repertoire, e.g., ab efgffge cd ab h feb. Note that even though there are only a finite number of states in this network, because of loops between some states, there can be a countably infinite number of valid possible paths from the starting arrow to the end state.”

“Long-standing results (Chomsky,1956; Rabin and Scott, 1959) demonstrate that such patterns cannot be generated by any finite-state transition network, because, for example, in order ensure that each warble i on the left is matched with its corresponding rattle i on the right one must in effect be…”

“Long-standing results (Chomsky,1956; Rabin and Scott, 1959) demonstrate that such patterns cannot be generated by any finite-state transition network, because, for example, in order ensure that each warble i on the left is matched with its corresponding rattle i on the right one must in effect be able to match up warb…”

“Such descriptions are conventionally called finite-state transition networks (Kleene, 1956); see Figure 3C.”

“Such alternatives were originally advanced by Solomonoff (1964), Horning (1969), and later pursued by Berwick (1982, 1985), Stolcke (1994), De Marcken (1995, 1996), and, more recently, Chater and Christiansen (2010) and Hsu et al. (2011),among several others).”

“As for non-human antecedents or possible pre-adaptations for pushdown stack storage, it has been proposed that the requirement for this kind of auxiliary storage may have been driven by the requirements for animal navigation such as map following or foraging (Bartlett and Kazakov, 2005; Okanoya, 2007; Shettleworth, 20…”

“(This is true of even such otherwise divergent linguistic theories as Construction Grammar, Goldberg, 2006; Head-driven Phrase Structure Grammar, Sag et al., 2003; and modern generative grammar, Radford, 1997).”

“Turning to the interplay between common descent and convergent evolution, over the past decade many studies have confirmed that songbirds and humans possess homologous brain regions for auditory–vocal and motor-driven learning (Jarvis et al., 2005).”

“In this sense, FoxP2 serves as an example of “deep homology” – a shared trait involved as part of both human speech and songbird vocal learning (Bolker and Raff, 1996; Shubin et al., 1997; Fitch, 2011; Scharff and Petri, 2011).”

“This is an extremely favorable result from the standpoint of both perceptual processing and learning, since in general, learning finite-state transition networks is not possible even given a large number of positive examples, possibly exponential with respect to the number of states in the final, to-be-acquired networ…”

“…and more recently brain-imaging studies, has accumulated showing that this necessarily “linear” format is mapped to an internal representation that respects only hierarchical structure (see, e.g., Moro, 2008, 2011; for recent fMRI confirmation along these lines, see Pallier et al., 2011).”

“The AFP is essential for sensorimotor learning and adult song plasticity (Brainard and Doupe, 2002; Mooney, 2009).”

“FOXP2 is developmentally and seasonally regulated in songbirds and intact FOXP2 levels are required for normal song learning (Fisher and Scharff, 2009).”

“More precisely, it appears that one can characterize the formal complexity of birdsong sound systems as a so-called k-reversible finite-state transition network (Angluin, 1982; Berwick and Pilato, 1987; Berwick et al., 2011a).”

“More generally, the set of all finite-state transition networks generate the (syllable) stringsets called the regular languages, equivalently, stringsets defined by regular expressions (McNaughton and Yamada, 1960).”

“Depressed vocal learning in songbirds that has been attributed to FoxP2’s role in regulating other genes involved guiding neuronal development (Haesler et al., 2004; Vernes et al., 2011).”

“Earlier hypotheses that certain components of the vocal tract have been uniquely adapted for human speech, such as a descended larynx, now seem questionable (Fitch, 2005).”

“As we describe in Section “Human Language and Birdsong: The Key Differences” below, this distinction is too weak, because it appears that all birdsong can be described by a far narrower class of regular languages, that turn out to be easily learned from examples, an important point if birdsong is to be learned from ad…”

“(B) Diagram of a songbird brain giving approximate positions of nuclei and brain regions involved in vocal production and sensorimotor learning.”

“…Key Differences” below, this distinction is too weak, because it appears that all birdsong can be described by a far narrower class of regular languages, that turn out to be easily learned from examples, an important point if birdsong is to be learned from adult male tutors (Berwick et al., 2011a).”

“Regions in the songbird caudomedial pallium (including the NCM) contain the neural representation of tutor song memory that is formed in juvenile males (Bolhuis and Gahr, 2006), whereas nuclei in the song system are required for sensorimotor learning and song production (Brainard and Doupe, 2000).”

“The most recent common ancestor of birds and mammals, originating from the clade Amniotes, lived about 300 million years ago (Laurin and Reisz, 1995).”

“Similarly, Crain and Nakayama (1987) found that children acquired question formation rules that abided by hierarchical constraints, but never rules based on linear order.”

“Regions in the songbird caudomedial pallium (including the NCM) contain the neural representation of tutor song memory that is formed in juvenile males (Bolhuis and Gahr, 2006), whereas nuclei in the song system are required for sensorimotor learning and song production (Brainard and Doupe, 2000).”

“…activity in the superior temporal lobe,but not in the inferior frontal cortex, in response to human speech (Imada et al., 2006), while 3- to 12-month-old infants showed activation in both Wernicke’s and Broca’s areas in response to hearing speech (Dehaene-Lambertz et al., 2006; Imada et al., 2006).”

“In humans, syllable chunks have been found to be an integral part of perception, even in very young infants as early as 4 days old (Bijeljac-Babic et al., 1993).”

“In this sense, it is perhaps the “simplest” example of a language that cannot be recognized by an unadorned finitestate transition network, as noted by Rogers and Pullum (2011).”

“, 2011); or rely on the hand-construction of an initial grammar that is in any case covers but a small fraction of the human language system (Perfors et al., 2010).”

“Depressed vocal learning in songbirds that has been attributed to FoxP2’s role in regulating other genes involved guiding neuronal development (Haesler et al., 2004; Vernes et al., 2011).”

“, 2011) it has been proposed that the capacity for vocal learning in passerine (oscine) birds such as the zebra finch and the non-passerine vocal learning birds such as parrots is more likely to have evolved in a common Psittacopasseran ancestor as a unique evolutionary event, leading to shared genetic/neural componen…”

“…of view, FoxP2 comprises one of probably many necessary ingredients in a complex recipe for vocal learning and production, rather than a single “master gene” that sits at the top of a regulatory cascade as in the case of the well-known regulatory Pax-6 eyeless homeobox gene (Halder et al., 1995).”

“Following the lead of experiments carried out with non-human primates (Fitch and Hauser, 2004), these approaches have most often attempted to probe whether songbirds can learn to discriminate the strings of languages that are not describable by any finite-state transition network. In particular, researchers have focus…”

“For example, one currently fruitful line of research in child language acquisition has probed the nature of infants’ abilities to acquire particular sound patterns and word boundaries in part via statistical regularities (e.g., Saffran et al., 1996; Shukla et al., 2011, among much other work).”

“…as above, and a corresponding NP to its right, associated with, say, the starlings, the Lambek combinatory operator takes as input these two items, (NP§)/NP and NP, and output a new structure, NP§, corresponding to a traditional Verb Phrase, with the “NP§” notation indicating that a Subject NP is…”

“The song system is subdivided into two major pathways, the song motor pathway (SMP; Mooney, 2009) and the anterior forebrain pathway (AFP; Brainard and Doupe, 2000; Doupe et al.”

“…been known that people restructure sentences so as to avoid producing complex center-embedded structures, as well spontaneously using alternative strategies for solving tasks that would otherwise provide evidence for the processing of such structures (Langendoen, 1975, Perruchet and Rey, 2005).”

“…case, from psycholinguistic studies indicating that complete Verb Phrases, i.e., Verb–Object combinations, are “recycled” in human sentence processing, while there is no comparable evidence for this with respect to Subject–Verb combinations; see, e.g., Arregui et al. (2006), Mauner et al. (1995).”

“In fact, the contrary seems to be true, as noted by Berwick and Chomsky (2011), and by Petitto et al. (2004); Petitto (2005): the sensory–motor sequencing involved in the human sound system can be carried over in large measure to the domain of manually signed languages.”

“Indeed, as emphasized by Yip (2006), there are many basic questions regarding the connection between human and animal sound systems that remain unanswered, such as the precise role of statistical and prosodic features in birdsong, and their possible connection to the human language case.”

“For example, the combinatorial operator could still work on other elements, for example, syllables, in this way yielding the distinctive metrical patterning of sound melodies, rhythmic patterns, as suggested in the domain of human language by Halle and Idsardi (1995).”

“Indeed, it remains unclear whether birdsong even contains unbounded dependencies of the sort described in the previous paragraph, if we follow the results of Gentner and Hulse (1998) and others that first-order Markov processes, a more restricted network system, suffices to describe birdsong.”

“For example, Jackendoff (2010) argues that both components (1) and (3) have additional, separate interfaces to the mental repository of information about words, the lexicon, bypassing the syntactic component (2).”

“Even more recently it has been shown that bats possess “superfast” laryngeal muscles for echolocation that can work at frequencies greater than 100 Hz; such muscles are also found in songbirds (Elemans et al., 2011).”

“As noted by Scharff and Petri (2011), this system may be part of a “molecular toolkit that is essential for sensory-guided motor learning” in the relevant brain regions of both songbirds and humans.”

“Pushdown stack storage is sometimes assumed to be implemented by, for example, a set of decaying neural networks, with each offset in a net’s decay time corresponding to a different stack location (for more details see, e.g., Pulvermüller, 1993; Pulvermüller and Knoblauch, 2009).”

“…there is an accumulating body of more recent results confirming the advantages of asymmetry in sentence processing; see, e.g., Fong and Di Sciullo (2005); and for a recent perspective from the perspective of neurolinguistics, confirming the basic asymmetry of language, see Friederici et al. (2011).”

“…other findings suggest that Frontiers in Evolutionary Neuroscience www.frontiersin.org April 2012 | Volume 4 | Article 5 | 8 in songbirds there is a neural dissociation between song recognition and song production that is already apparent in juveniles (Gobes and Bolhuis, 2007; Gobes et al., 2010).”

“Importantly, the detailed study of human sound systems has established that only linear precedence relations are required for the description of such systems; see Heinz and Idsardi (2011) and Wohlgemuth et al. (2010) for further discussion.”

“In fact, the contrary seems to be true, as noted by Berwick and Chomsky (2011), and by Petitto et al. (2004); Petitto (2005): the sensory–motor sequencing involved in the human sound system can be carried over in large measure to the domain of manually signed languages.”

“Indeed, it would be surprising if this were not true, since this is in line with what is also known about the acquisition and use of human sound systems as well (Heinz and Idsardi, 2011).”

“Since both vocal learning and non-vocal learning birds possess identical FoxP2 genes (Webb and Zhang, 2005), and the birds’ FoxP2 genes are distinct from those of humans, differences in this gene alone cannot be what accounts for the vocal learning/non-learning distinction in birdsong.”

“Modified and reproduced, with permission, from Bolhuis et al. (2010) and Moorman et al. (2011), copyright 2010 Nature Publishing Group.”

“In particular, current alternative approaches either advance a method that has no corresponding constructive algorithm, let alone an efficient one (Solomonoff, 1964; Hsu et al., 2011); or rely on the hand-construction of an initial grammar that is in any case covers but a small fraction of the human language system (P…”

“As for non-human antecedents or possible pre-adaptations for pushdown stack storage, it has been proposed that the requirement for this kind of auxiliary storage may have been driven by the requirements for animal navigation such as map following or foraging (Bartlett and Kazakov, 2005; Okanoya, 2007; Shettleworth, 20…”

“…case, from psycholinguistic studies indicating that complete Verb Phrases, i.e., Verb–Object combinations, are “recycled” in human sentence processing, while there is no comparable evidence for this with respect to Subject–Verb combinations; see, e.g., Arregui et al. (2006), Mauner et al. (1995).”

“In further contrast with the Lambek system, note that this output set is by definition unordered, in this way reflecting the apparent lack of any syntactic predicates based on linear precedence.”

“For example, Suge and Okanoya (2010) demonstrated that Bengalese finches perceive songs in terms of syllable “chunks” that can be detected by placing a brief noise either at the boundary of chunks or in the middle of chunks, while training birds under operant conditions to react to the noise as…”

“…a bird’s behavioral repertoire, typically these transitions are defined probabilistically, so that a transition between states occurs only with some positive probability corresponding to the likelihood of observing such a transition in the behaviorally observed data (Kakishita et al., 2009).”

“…as one imputes a full two-pushdown stack system to a computational device, then this computational Frontiers in Evolutionary Neuroscience www.frontiersin.org April 2012 | Volume 4 | Article 5 | 20 machinery is as powerful as any general-purpose computer, i.e., it is as powerful as a Turing machine.”

“…Evolutionary Neuroscience www.frontiersin.org April 2012 | Volume 4 | Article 5 | 21 and Abe experiment as distinct in terms of memorization alone (Ikebuchi and Okanoya, 2000), by this argument we do not need to posit any novel syntactic ability for finches, a more parsimonious explanation for…”

“Although algorithms for contentaddressable memory are less well-known, even here, hierarchical representations can be readily developed, as described by, e.g., Oldfield et al. (1987).”

“Sasahara et al. (2006) have shown that one can in fact apply the same computer algorithms described by Berwick and Pilato to the problem of automatically inducing k-reversible transition networks from birdsongs.”

“…and more recently brain-imaging studies, has accumulated showing that this necessarily “linear” format is mapped to an internal representation that respects only hierarchical structure (see, e.g., Moro, 2008, 2011; for recent fMRI confirmation along these lines, see Pallier et al., 2011).”

“…a finite-state transition network with recursive subroutines, alternatively a context-free grammar, was not in fact of the right type to unambiguously demonstrate the conclusion that was sought, as noted by Corballis (2007), Friederici and Brauer (2009), and Friederici et al. (2011), among others.”

“Humans might have a strong inductive bias constraining language learning to some set of compositional systems, which may have emerged as a side effect of multiple evolutionary events in our lineage [134] or which may have been selected for directly [135].”

“In terms of evolutionary history, language is a recent arrival that shows clear signs of being a bio-cultural hybrid: a complex adaptive system in continuous cultural evolution within a landscape of cognitive, cultural, and communicative factors [56–59].”

“…evidence to date that birdsong matches the characteristic syntactic complexity of human language, arising from the composition of smaller forms like words and phrases into larger ones” (Berwick et al., 2012, p. 1), the type of syntax that linguists claim give rise to semantic compositionality.”

“Human language syntax has a hierarchical structure, which Berwick et al described as “composition of smaller forms like words and phrases into larger ones” (Berwick et al., 2012; Miyagawa et al., 2013).”

“Nor does birdsong contain the rich hierarchical structure characteristic of human language (Berwick et al., 2012).”

“However, the limitations of the language-birdsong comparisons are obvious, since birdsong lacks the semantics of language with its mapping of combinable syntactical elements on accordingly combined meaning (for reviews, see Berwick et al., 2011, 2012).”

“Several candidates of phoneme-like structures in nonhuman animals have been proposed, but defining features are either lacking or have yet to be demonstrated [8,11,17].”

“As Berwick et al. (2012) correctly observe, the hierarchy is both ‘‘too weak and too strong,’’ failing as it does to cut natural language at its joints.”

“…(and, in the case of Abe and Watanabe, 2011, their additional mistakes in experimental materials renders their findings extremely suspect (see Berwick et al., 2012)), demonstrating a general failure to distinguish between the ability to implement, learn, and have a preference for a Type 2…”

“And indeed Marler’s view leads to lexicocentrism, even if Berwick et al. (2012) appear to claim the contrary.”

“The expression layer is similar to birdsongs; birdsongs have specific patterns, but they do not contain words, so that birdsongs have syntax without meaning (Berwick et al., 2012), thus it is of the E type.”

“For some adherents of the gradualist approach (especially Jackendoff, 1999, 2002, 2009; Progovac, 2006, 2008, 2009, 2010, 2012; Progovac and Locke, 2009), human language has undergone a sequence of stages, initially a one-word stage, then a two-words stage—a proto-syntax—characterized by the…”

“Instead of focusing on the complexity of the overtly reinforced patterns, we ask how the perceptual groupings of pattern elements influence the generalization pattern knowledge.”

“The generic architectural properties of the language system according to that view are illustrated in Figure 1 (Berwick et al., 2012, 2013; Bolhuis et al., 2014).”

“(Berwick et al. (2012)) I too claim that a property of lexical units was the key, but this property is not valency, but rather, the absence of it (lexical items as ‘edge feature carriers,’ see Boeckx (2011b, To appear)).”

“A couple of contributions take the evolutionary perspective even further by including song birds into their comparative approach (Berwick et al., 2012; Kiggins et al., 2012; Petkov and Jarvis, 2012).”

“(Hauser, Chomsky and Fitch 2002; Jenkins 2000, 2004, Di Sciullo et al 2010; Di Sciullo & Boeckx 2011; Berwick et al 2012, Boeckx and Grohmann 2013, a.o.).”

“The cerebellum was generally smaller, and there was a large space where the striatum and thalamus should develop, above which there was a space in the skull.(39) There is also an indentation behind the eyes”

“Berwick et al. (2012) thought that the recursion of birdsong had nothing to do with the recursion of human syntax, that it was rather analogous to aspects of human speech, that is, the sensorymotor component of human language.”

“Human and Aves species diverged about 600 MYA [32], yet chicken embryonic-specific insulin uORF controls low-level and glucose-independent expression of proinsulin not only in the neuroepithelial cells of the ectoderm, but also in the mesoderm and endoderm layers [33].”

“As a result, the clear examples of controllable vocal languages can be found in birdsongs and cetaceans’ songs, instead of nonhuman primates’ vocal languages [22].”

“Combinatorial syntax allows a multitude of possible meanings to be generated from a relatively small number of base sounds, allowing a wide range of concepts to be expressed, as in human language (Berwick et al., 2012).”

“(2009) and Berwick et al. (2011) for a critical view].”

“We wish to reproduce aspects of the development of human natural language (e.g. Pinker & Bloom (1990), Berwick et al. (2012)).”

“We aligned INSU peptide and nucleotide sequences of human and chimp ( Pan troglodytes , NC_03689 0.1, diverged 6 MYA) (30) with other primate species and found a 16 bp deletion in the human-chimp ORF region, generating an INSU protein coding region that is disrupted by a frame shift in other primate species (Fig.”

“…to communicative capacities (Hauser et al., 2002); for example, Berwick ( 2011 ) conjectured that Merge might be the basic operation underlying the structure of birdsong, although he attributed its presence in two so dis - tant lineages to convergent evolution (see also Berwick et al., 2012 ).”