II. The problem: ancient and modern interpretations

The first thing to observe is that the three child-language productions reproduced by Strepsiades have all been given an accusative case-ending -ν. When stripped of it, we end up with βρῦ, μαμμᾶ and κακκᾶ, which Kenneth Dover calls ‘plausible baby-words’.Footnote ¹ Kakkā is easy to interpret (it has cognates in adult language), and it also fits a pattern of early speech production recognized by child-language researchers, consisting of a reduplicated consonant-vowel sequence.Footnote ² Mammā, too, follows this form, but its alleged meaning should give us pause. In fact, there is very little reason to think that Strepsiades’ interpretation of the utterance as (a need for) food, and its implementation as ‘bread’ is correct, although commentators have not been too concerned. Mark Golden lists μᾶ, μαῖα, μάμμα, μάμμη, μαμμία, μαμμίδιον as words referring to ‘mother, grandmother, old woman, nurse, midwife’, but accepts a separate entry μαμμᾶ, ‘food’, on the strength of our passage.Footnote ³ However, it is only reasonable to take the possibility into account that Strepsiades’ confident interpretation may itself be a source of humour, while his point (that he was doing what he could for his little son) remains intact.

Two of the three utterances, kakkā and mammā, appear to be straightforward and not entirely straightforward child productions, respectively. However, contrary to Dover’s claim, βρῦ is not a ‘plausible baby-word’ at all, and reflecting on this issue may be a meaningful point of entry into a different interpretation of the passage. When children start to articulate (proto-)words (and our passage clearly suggests that we are at a very early stage of language production), they tend to avoid consonant clusters, and in fact usually simplify the pronunciation of target words containing consonant clusters in order to get rid of them, as in ‘tain’ for the target word ‘train’.Footnote ⁴ The consonant cluster in [brü:] is thus highly unusual in early language use. So is the rounded front vowel /y/, which usually appears late in language development (in contrast to the /a/ in the other two articulations mentioned by Strepsiades).Footnote ⁵ In summary, the phonetic shape of the word is not easily compatible with actual early child-language productions.

Besides issues of phonology, there is also a semantic problem: the meaning of βρῦ is far from obvious, despite assurances to the contrary in ancient and modern commentaries, which may all have followed Strepsiades rather too trustingly in his confident (over-)interpretation of his baby son’s vocalizations.

Ancient interpretations that blandly equate βρῦ with ὕδωρ, ‘water’, or πιϵῖν, ‘to drink’, are all clearly directly derivative of our passage, and cannot therefore be considered conclusive,Footnote ⁶ although they may at least be less likely to shock a modern reader than the one identifying βρῦ with wine or beer.Footnote ⁷ Although, presumably and not unreasonably, drink and food may have been welcome to any baby, the ‘word’ βρῦ cannot simply mean ‘drink’: it takes the efforts and attentiveness of the father to come to this interpretation. Moreover, as we will argue, that is the very point of Strepsiades’ argument: he was willing and able to interpret his son’s wishes long before the boy had even heard of the fancy and corrupt rhetoric on which he now prides himself, at a time when he was virtually inarticulate. And although his guesses at what his son needs are not unreasonable, they may not reflect what the baby ‘says’.

Yet some ancient commentators do not just emphasize the semantic content of the ‘word’, but actually imply a form of intentionality. The scholia to Aristophanes mostly simply identify the word as a baby word,Footnote ⁸ but one comment also refers to the intentions of the baby (Σ Ar. Nub. 1382c):

Something similar may be going on when Phrynichus states s.v. βρῦ (Praep. soph. 55.13):

The ‘word’ βρῦ would then be onomatopoeic, not just in the sense of imitating the sound of drinking, but in that the baby makes drinking noises as a conscious indication that he wants to drink. However, this comment is probably better taken as referring to the caregivers’ register, which would also provide a more natural interpretation for ὑποκόρισμα. It would then mean that βρῦ is uttered in the sweet tone of voice used to talk to babies (‘a baby word’, a natural extension of the use of ὑποκόρισμα for ‘term of endearment’), and that it is said not ‘by’ but ‘to’ babies to refer to drinking.Footnote ¹⁰

Modern lexica and commentaries also allow themselves to be led by Strepsiades in interpreting βρῦ: LSJ lists it s.v. βρῦν and gives the whole construction βρῦν ϵἰπϵῖν, defined as ‘cry for drink, of children’.Footnote ¹¹ This suggests both intentionality and referential meaning of a sound production by the baby, not the caregiver. Jan van Leeuwen ad loc. explicitly derives his interpretation of βρῦν as a baby word from the context, and Alan Sommerstein ad loc. concurs that brū is ‘evidently a nursery word meaning “a drink”’.Footnote ¹² In his survey of child language and baby talk, Golden also simply gives ‘drink’ as the meaning of βρῦ.Footnote ¹³ Oliver Thomas takes it as ‘holophrastic’, representing a ‘condensed request’.Footnote ¹⁴

III. Child language in antiquity

Aristophanes’ rendition of the infant Pheidippides’ sounds is among the very first attempts in Western history to register baby or toddler (proto-)speech.Footnote ¹⁵ Herodotus’ famous report of Psammetichus’ experiment is another, but very different one: there, two babies taken from random parents are deprived of normal human interaction. Their only company is each other, some goats whose milk they drink, and the goatherd who comes in daily to tend to their needs, but who has been ordered to refrain from any communication. The experiment was set up in an attempt to establish what the children would utter as their first articulate utterance (ϕωνή) and thus to discover experimentally which language, and by extension what people, was the oldest on earth (Hdt. 2.2).Footnote ¹⁶ It is important that Psammetichus expressly excluded any ‘meaningless whimperings’ (ἀπαλλαχθέντων τῶν ἀσήμων κνυζημάτων).Footnote ¹⁷ He was looking for something articulate, and that only came after two years (διέτης χρόνος). It turned out to be the word βέκος, which both children said to the goatherd one day when he came in, falling at his feet and extending their hands to him. When this happened more frequently and Psammetichus had heard it for himself, he had inquiries made, and learned that βέκος was Phrygian for ‘bread’, and that settled the matter.Footnote ¹⁸

The differences with little Pheidippides are clear: he was interacting with a parent, a much more ecologically valid situation; there was no obvious deprivation of stimuli in the environment.Footnote ¹⁹ Strepsiades’ point and claim to fame is precisely that he could understand utterings that were jumbled or lisped (τραυλίζοντος, 1381); in fact, this increases the likelihood that these utterances were not articulate words. Note also that in contradistinction to the Herodotean passage, in which the children were two years old when they became articulate, the Aristophanic passage does not indicate an age.

The most relevant other ancient passage focusing on the interaction between a baby or toddler and a caregiver is Aeschylus’ Choephoroe 749–62. Believing that Orestes is dead, the nurse Cilissa remembers her interaction with him in his infancy. She recalls how much work caring for a baby is: they just eat, there is no sense yet and they cannot speak (755–60):

This tragic passage does not reproduce any actual utterings of the baby, but it does agree with the Herodotean and Aristophanic passages in insisting on the lack of clear articulation, and hence the fundamental incomprehensibility, of baby talk. Alexander Garvie ad loc. quotes the scholiast on Septem contra Thebas 348: τὰ νϵογνὰ οὐδέπω τὴν ϕωνὴν ἔναρθρον ἔχοντα …, ὥστϵ αὐτὰ ἄναρθρον ϕωνὴν καὶ ὥσπϵρ προβατώδηFootnote ²¹ προΐϵσθαι (‘babies cannot yet produce articulate sound … so that they utter an inarticulate and cattle-like sound’). It is reasonable to assume hunger, thirst or the need to urinate, but this does not correspond one-to-one with any particular utterance. This means that it takes almost mantic qualities to predict (in particular) when the child needs to urinate. And even the immediate caregivers, frequently the only ones to ‘read’ and understand a baby, may fail at this task. The nurse in this case failed regularly, and as a result had to do a lot of laundry. The passage is a wonderful counterpoint to the Aristophanic one: the three basic baby needs are identical and it is precisely the masterful feat of being fully focused on Pheidippides’ intentions (ὅ τι νοοίης) and (presumably) correctly and instantly guessing his son’s meaning (γνοὺς … ἂν οὐκ ἔϕθης ϕράσας, κἀγὼ …) that proves Strepsiades’ utter devotion as a parent. But if babies could talk so that their words were a precise indication of their meaning, there would be no special merit in understanding them.

IV. What is βρῦ?

Aristophanes is very much a child of his time in his interest in issues of language, ‘grammar’, rhetoric and argument. The Clouds in particular, with its parody of Socrates as a representative of all forms of new-fangled learning, including a sophistic focus on language and rhetoric,Footnote ²² offers a wonderful occasion to mock ideas of linguistic correctness, etymology and perversely complicated argumentation. However, like all other exponents of ancient linguistic thought, Aristophanes’ focus is primarily semantic and pragmatic: he thinks about language in terms of meaning and effect. This is a tendency that would even affect ‘technical’ grammar, once that had separated itself as a discipline from a more general interest in language.Footnote ²³ Ancient observations of language rarely touch on the domain of phonology for its own sake. This may be the reason why the true nature of what Aristophanes is saying here has escaped detection.

If we abandon the idea that we must be dealing with an actual baby word, uttered intentionally and with some form of referential meaning, βρῦ may prompt an entirely different interpretation. Along the lines of some of the ancient interpretations discussed above, we may regard it as the imitation of a sequence of sounds (onomatopoeia) which together constitute a baby sound that Aristophanes has of necessity tried to render with the alphabetic means available to him.Footnote ²⁴ This would put βρῦ in the category of conventionalized renditions of interjections, of which ancient literature (and grammar) is full.Footnote ²⁵ But if we ask what actual sound was at the basis of this conventionalization, it so happens that we find a perfect fit in what we know of first-language acquisition in infants.

So far, we have seen that whereas mammā and kakkā correspond to formations we may expect from a normally developing infant, brū appears abnormal: the front-round vowel is unusual in early child language and, if anything, a baby would omit the complex onset, certainly not add it to a target word originally lacking it (for example, hypothetically, πιϵῖν, ‘to drink’). Brū would only fit either a later stage, when the child has mastered consonant clusters, or a much earlier one. Now, of course, there is no reason why Aristophanes should not have simply combined utterances from different developmental phases as identified by modern observations. However, the solution that brū should belong to a later stage seems excluded by the fact that there is no plausible target word, since brū is not recognizably related to ‘normal’ Greek words for eating or drinking. Considering an earlier stage, on the other hand, provides a highly realistic and plausible solution for the whole conundrum: between four and six months, babies start experimenting with the production of different sounds. This stage is called the ‘expansion stage’.Footnote ²⁶ One of the results is what is referred to as an oral ‘raspberry’,Footnote ²⁷ technically a linguolabial trill [r̫] or a bilabial trill [ʙ]. In our own ‘regular’ orthography in the Latin alphabet ‘bwrrr’ is the closest approximation. It is produced by putting the tongue against or between the lips and blowing, which may produce a sound accompanied with many little spit bubbles. As a bilabial trill, it resembles the way in which we imitate the sound of a snorting horse. Examples of babies blowing raspberries can be found on the Internet.Footnote ²⁸ The question then becomes how Aristophanes would have rendered this sound in his own Greek alphabet,Footnote ²⁹ and it seems to us that βρῦ (the same would be true for βροῦ) is a fair attempt: it renders a bilabial trill with rounded lips. Given this succession of phonemes, and the fact that it is unlikely that βρῦ with its complex onset and front rounded vowel belongs to early child language, we propose that βρῦ is actually a conventionalized rendition in writing of the sound of a raspberry. This would make Aristophanes the oldest attestation of this baby sound production, well over 2,000 years earlier than the oldest references currently recognized. These belong to German diaries with parental observations of their children’s development, and their solution to the transcription problems match the Aristophanic one very well: they come up with renderings like ‘Brrr or Urrr’, or ‘brrr-há’.Footnote ³⁰

V. Performing Clouds 1382

If βρῦ(ν) represents a raspberry, new possibilities in performance present themselves: rather than articulating [brü:], the actor playing Strepsiades may have graphically mimicked the intended baby behaviour by blowing a prolonged raspberry himself, with his tongue between his lips, resulting in a lot of audible spit bubble production, until it comes to an end in the astonishingly grammatical accusative ending -ν. The unexpected grammaticality would add to the performative humour.Footnote ³¹ As would the claim that Strepsiades knew what this extraordinary sound actually ‘means’. Embedding a mimetic, performative element in a ‘normal’ sentence, such as a full-blown ‘raspberry’ here, is a recognized linguistic phenomenon, styled ‘depiction’.Footnote ³²

Clearly, one would need extraordinary mantic qualities to derive any kind of meaning from a raspberry, as the rest of the ancient tradition about baby language, scanty as it is, suggests. But long before Pheidippides even came close to becoming a rhetorician, his devoted father had guessed impossibly expressed (or unexpressed) desires and come running with anything a baby might need. Since this is a more remarkable feat by far than providing a drink when simply asked for a drink in so many words, this should inspire the feelings of obligation and debt in Pheidippides that would persuade him to desist from his perverse bad behaviour.

VI. A brief and speculative return to μαμμᾶν and κακκᾶν

Now that this new interpretation of βρῦ is available, what about μαμμᾶν and κακκᾶν? Although more plausible as linguistic child productions at an only slightly later stage of language development than βρῦ,Footnote ³³ there is still the issue that Strepsiades offers a suspect interpretation of μαμμᾶν, and that κακκᾶν is not objection-free itself: while it is certainly among the baby-talk words used by parents, the sound [k] is usually not favoured in the set of early produced consonants, although this also depends on the frequency of the sound in the surrounding language, and on individual preference.Footnote ³⁴

Given our interpretation of βρῦ and Strepsiades’ comically effective interpretation of what is at heart perfectly meaningless, we should at least acknowledge the unprovable possibility that all three ‘words’ should be taken to reflect roughly the same stage of infant linguistic experimentation, at the threshold between the expansion and canonical stages, although, obviously, there is no inherent reason why Aristophanes should be held to any standards of consistency here.Footnote ³⁵ In this period, not only do we hear raspberries (friction noises), but other sounds also appear: nasal murmurs and glottal stops ([ɂ]).Footnote ³⁶ Marginal babbling occurs at this stage as well, in which a closure of the vocal tract is opposed to a vowel sound. The typical reduplicated consonant-vowel-consonant-vowel structures appear next, at around seven months.Footnote ³⁷ The words typically associated with early child language, like ‘mamma’, ‘pappa’, ‘pipi’, result from parents assigning meanings to their infant’s babbling sequences!Footnote ³⁸

This opens up the possibility that Strepsiades imitates a baby in all three cases, with the ‘mamamama’ sequence suggesting munching on food (which in turn leads him to the interpretation ‘bread’), and the form ‘kakkā’ being the closest one could get in writing to a more onomatopoeic form of the action [ɂaɂa], involving the glottal stop, but possibly also suggesting the effort needed to relieve oneself. Especially in the last case, the conventional parental interpretation seems to have won out and led to the spelling with κ as well.

When performed in this way, all three sounds would mimic a baby of four to seven months of age, and the text would indicate a rather careful observation on the part of Aristophanes of the ‘expansion stage’ of language acquisition. However, given the relationships between mammā and kakkā to conventionalized lexical entitiesFootnote ³⁹ and the fact that Aristophanes’ priorities clearly lie with raising a laugh in any way possible rather than with phonology, there is no reason to insist. Our interpretation of βρῦ is sufficient to make of Aristophanes the first observer to put this infant sound production on our Western-European record. And in passing he is putting the widespread tendency of adult caregivers to over-interpret such infant vocalizations to good comic effect.